Development and application of a precision feeding

Development and application of a precision feeding program using electronic
sow feeders and effect on gestating primiparous sow performance
by
Robert Quincy Buis
A Thesis
Presented to
The University of Guelph
In partial fulfilment of requirements
for the degree of
Master of Science
in
Animal Biosciences
Guelph, Ontario, Canada
© Robert Quincy Buis, July 2016
ABSTRACT
DEVELOPMENT AND APPLICATION OF A PRECISION FEEDING PROGRAM USING
ELECTRONIC SOW FEEDERS AND THE EFFECT ON GESTATING PRIMIPAROUS SOW
PRODUCTION
Robert Quincy Buis
University of Guelph, 2016
Advisor:
Professor C.F.M. de Lange
Current commercial gestation sow feeding strategies do not consider the sow as an individual; they are
generally based on using a single gestation diet for all sows regardless of parity or stage of gestation.
Computer controlled electronic sow feeders (ESF) allow precision feeding (PF) of individual, gestating
sows housed in groups. This thesis evaluated the effects of PF primiparous sows using the NRC (2012)
nutrient requirement model, on sow body weight (BW) and back fat (BF) changes during gestation, animal
well-being, lactation performance and environmental impact.
A performance trial demonstrated ESF technology using blended feeding of two basal diets to meet
the unique energy and lysine requirements of individual sows in a group housing system is possible. In this
study, PF of sows did not affect overall gestation BW and BF gain, sow feeding behaviour, lactation
performance or nutrient balances. However, patterns of BW gain for PF sows more closely reflected the
gains of fetus and reproductive tissues which could have effects on long term sow performance.
ii
ACKNOWLEDGMENTS
Many people were instrumental to the completion of this thesis and I would like to express my
thanks towards the contributions of every one of them. First, to Dr. C.F.M. de Lange, I would like to
show my sincerest gratitude; without his support, guidance and encouragement this thesis would never
have come to fruition. I would also like to thank my advisory and defence committee members, Dr.
Robert Friendship, Dr. Stephanie Torrey, Dr. Ira Mandell, and Dr. Vern Osborne, as their input has
greatly improved the quality of this thesis. I am also grateful for the assistance and patience of the staff at
the Arkell Swine Research station.
Acknowledgements are due to those who provided financial support: OMAFRA, Canarm
Agsystems, PigCHAMP (a farms.com company), OMAFRA/University of Guelph HQP Scholarship, and
Wallenstein Feed and Supply Graduate Scholarship. A special thanks to Mike Fisher, Curtiss Littlejohn,
Tymon Hardman, Scott Keys, Angie Bowman and Martin Widdowson for hours of technical support in
designing and troubleshooting the Electronic Sow Feeder system.
I would like to extend my thanks to fellow graduate students and lab mates, Emily Miller, LeeAnne Huber, Wilfredo Mansilla, Marko Rudar, Melissa Wiseman, Heather Reinhardt, Peter Park, Adam
Totafurno, and Youngji Rho. As well as technicians and summer students, Doug Wey, Julia Zhu, Hannah
Golightly, and Kayla Silva. Not only for their technical assistance during experimentation, statistical
analysis, and writing, but for their friendship and support throughout many challenges.
I would especially like to thank my friends who made my academic career enjoyable and who put
up with my continuously distracting them with adventures. Lastly and most importantly, I would like to
thank my family for their patience, sense of humour and support throughout this crazy journey.
iii
TABLE OF CONTENTS
1.0. LITERATURE REVIEW ..................................................................................................... 1
1.1. Introduction .......................................................................................................................... 1
1.2. Determinants of energy and nutrient requirements of gestating sows ................................. 2
1.3. Different feeding strategies throughout gestation aimed at improving sow performance ... 4
1.3.1. Increasing feed intake or bump feeding during gestation ............................................. 5
1.3.2. Altering energy levels throughout gestation ................................................................. 8
1.3.3. Effect of changing protein or amino acid intake levels during gestation ..................... 9
1.4. ESF in group housing systems as an alternative to gestation stalls ................................... 12
1.4.1. Determining optimal time of mixing post breeding for group housed sows .............. 14
1.5. Conclusion ......................................................................................................................... 16
2.0. RESEARCH HYPHOTHESIS AND OBJECTIVES ....................................................... 18
3.0. RESEARCH ELECTRONIC SOW FEEDER DEVELOPMENT.................................. 19
3.1. Introduction ........................................................................................................................ 19
3.2. Structural modifications and feeding logic development .................................................. 20
3.2.1. Entrance gate ............................................................................................................... 21
3.2.2. Mid-Way gate ............................................................................................................. 22
3.2.3. Exit gate ...................................................................................................................... 23
3.2.4. Feed bowl .................................................................................................................... 24
3.2.5. RFID Antenna ............................................................................................................. 25
3.2.6. Feeding logic ............................................................................................................... 27
3.2.7. Feeding settings .......................................................................................................... 28
3.2.8. Safety modifications ................................................................................................... 29
3.3. Developing and implementing Stepper Motors for accurate feed delivery ....................... 29
3.3.1. Calibration................................................................................................................... 30
3.4. Developing and evaluating scale feature ........................................................................... 31
3.4.1. Body weight scale testing ........................................................................................... 32
3.4.2. Summary ..................................................................................................................... 33
3.5. Conclusions and implications ............................................................................................ 34
4.0 FEEDING PROGRAM DEVELOPMENT ........................................................................ 48
4.1. Introduction ........................................................................................................................ 48
4.2. Modifying the NRC model to calculate required daily ME intake .................................... 49
4.3. Determining feeding level and recipe of diet for each sow ............................................... 50
4.4. Adjusting feeding curves for individual animals deviating from predicted growth curve 52
4.5. Summary ............................................................................................................................ 54
5.0 EFFECTS OF PRECISION FEEDING PRIMIPAROUS SOWS ON BODY WEIGHT
CHANGES, LACATION PERFORMANCE, BEHAVIOUR AND LAMENESS AND
iv
NUTRIENT BALANCES........................................................................................................... 59
5.1. Introduction ........................................................................................................................ 59
5.2. Materials and Methods ....................................................................................................... 60
5.2.1. Animals, housing, diets and experimental design ....................................................... 60
5.2.2. Experimental observations .......................................................................................... 63
5.2.3. Calculations and statistical analysis ............................................................................ 63
5.3. Results ................................................................................................................................ 64
5.4. Discussion .......................................................................................................................... 67
5.5. Conclusions and implications ............................................................................................ 70
6.0. GENERAL DISCUSSION, IMPLICATIONS AND FUTURE CONSIDERATIONS . 79
LITERATURE CITED .............................................................................................................. 88
v
LIST OF TABLES
Table 3.1. Canarm Research ESF Feeding settings used for described research ......................... 36
Table 3.2. Results of testing different motor calibration techniques for Canarm ESF stepper
motors, using Unit 2 as example. .................................................................................................. 36
Table 3.3. Scale 1 BW measurements during feed delivery and feed bowl retraction for random
subsample of 4 sows ..................................................................................................................... 38
Table 3.4. Scale 1 BW measurements during the time after the sow had entered the ESF but
before feed delivery had begun for random subsample of 4 sows. ............................................... 39
Table 4.1. Default values for target daily Ld (g/d) from average maternal Ld at typical energy intake levels
across parties and as outlined in NRC (2012). ..................................................................................... 55
Table 4.2. Example adjustment for a parity 1 gestating sow (gilt; 140kg initial BW, anticipated
12.5 pigs born, 1.40 kg birthweight) of feed intake for mid and late gestation based on higher
than anticipated BW gain between d 1 and 28 of gestation. ......................................................... 56
Table 5.1. Ingredient composition (%) of High and Low Lys gestation diets, as well as the
lactation diet. ................................................................................................................................. 71
Table 5.2. Calculated and analyzed nutrient contents (%, as is basis) in High and Low protein
experimental gestation diets (HP and LP respectively), as well as common lactation diet. ......... 72
Table 5.3. Amount of high protein (HP) and low protein (LP) feed consumed during entire
gestation period by primiparous sows on PF and CON treatments. ............................................. 73
Table 5.4. Change in mean BW, ADG and BF for primiparous sows feed on PF and CON
treatments throughout gestation. ................................................................................................... 74
Table 5.5. Litter and Lactation performance for primiparous sows feed PF and CON treatments
throughout gestation...................................................................................................................... 75
Table 5.6. Mass Balance calculation and feed cost for primiparous sows on PF and CON throughout
gestation.......................................................................................................................................... 76
vi
LIST OF FIGURES
Figure 1.1. NRC (2012) predicted total protein gain (Pd) in separate pools for a gestating gilt. 17
Figure 3.1. Primary flow designs and examples of commercially available ESF in 2014 .......... 40
Figure 3.2. Pen layout of loose housing pens containing ESF at the Arkell Swine Research
Station (University of Guelph, Guelph, ON, Canada). ................................................................. 41
Figure 3.3. Flow diagram of Canarm, SowChoice Systems ESF unit operations. ...................... 42
Figure 3.4. Labelled view of Canarm Research ESF. .................................................................. 43
Figure 3.5. Original entrance Gate and described issues. ............................................................ 44
Figure 3.6. Mid-way gate designs for Canarm research ESF. ..................................................... 45
Figure 3.7. Half open exit gate (springs removed) on Canarm research ESF............................. 46
Figure 3.8. Feeding bowl and RFID Antenna on Canarm research ESF. .................................... 46
Figure 3.9. ESF sow scale design 1. ............................................................................................ 47
Figure 4.1. Estimated energy intake (ME) and SID lysine (Lys) requirements according to the
original (Original) and modified (Mod) NRC (2012) model. ....................................................... 57
Figure 4.2. Example of .csv excel file that is generated using the modified NRC (2012) gestating
so model and imported into PigCHAMP sow management software. ......................................... 58
Figure 5.1. Average sow BW of Precision Feed primiparious sows compared to average
predicted BW using modified NRC 2012. .................................................................................... 77
Figure 5.2. Average sow BW of Control Feed primiparious sows compared to average predicted
BW using NRC 2012. ................................................................................................................... 77
Figure 5.3. Average sow ADG of Precision Feed primiparious sows compared to average
predicted ADG using modified NRC 2012. .................................................................................. 78
Figure 5.4. Average sow ADG of Control Feed primiparious sows compared to average
predicted ADG using NRC 2012. ................................................................................................. 78
vii
LIST OF APPENDICES
Appendix 1. Colour codes of light indicator for ESF status ........................................................ 84
Appendix 2. New calibration tool computer interface developed for ease of accurate individual
feed deliver motor calibration ....................................................................................................... 85
Appendix 3. Using Matrix multiplication, to develop equations that will blend 3 separate isocaloric feeds to achieve calculated Energy, Protein(Lys) and Phosphorus requirements............. 86
viii
LIST OF ABBREVIATIONS
ADFI
average daily feed intake
ADG
average daily gain
BF
back fat
BW
body weight
°C
degrees Celsius
cm
centimeter
CON
control
CP
crude protein
.csv
comma separate values file
d
day
DE
digestible energy
ESF
electronic sow feeder
FCR
feed conversion rate
g
gram
h
hour
HP
high protein
kcal
kilocalories
kg
kilogram
Ld
lipid deposition
LP
low protein
Lys
lysine
Mcal
Megacalories
ix
ME
metabolizable energy
min
minute
Mj
Megajoule
Mod
modified
N
nitrogen
NE
net energy
P
phosphorus
Pd
protein deposition
PF
precision fed
RFID
radio frequency identification
s
second
SID
standardized ileal digestibility
SQL
structured query language
WEI
weaning-to-estrus interval
x
1.0. LITERATURE REVIEW
1.1. Introduction
Changes in maintenance requirements and growth in both the maternal body and products
of conceptus (i.e. fetus, placenta, etc.) contribute to the dynamic changes in nutrient requirements
of sows throughout gestation (NRC, 2012). Current commercial gestation sow feeding strategies
do not consider these dynamic changes and do not reflect unique needs of individual sows; they
are generally based on using a single gestation for all sows regardless of parity or stage of
gestation. As a result, most sows are fed above their nutrient requirements for the majority of the
gestation period, which in many cases can be costly for the producer and results in excessive
nutrient losses into the environment. In addition, dietary nutrient levels may be below
requirements during late gestation and in young sows when dietary nutrient requirements are
greatest. The latter may compromise long-term sow reproductive performance. Further research
on adjusting sow nutrition during gestation to meet the requirements of the individual sow and
products of conceptus is required, especially when considering long-term sow productivity.
Phase and parity-specific feeding of gestating sows provides a means to more closely meet the
dynamic changes in nutrient requirements during gestation and across parities (Moehn et al.,
2011), resulting in reduced feed costs, improved nutrient utilization efficiencies and potentially
better sow productivity, including piglet quality at birth. Based on a review of the literature, there
is a lack of studies evaluating the effect of gestation feeding regime on long-term sow
reproductive performance.
Gestation sow management is of increasing importance as pork producers are encouraged
to carefully consider the welfare of gestating sows and the possible shift away from certain
conventional practices i.e. gestation crates. According to the Canadian Codes of Practice for the
1
Care and Handling of Pigs (NFACC, 2014), the conventional system of housing sows in
gestation stalls will be banned by 2024 and new facilities built after 2014 must incorporate group
housing for sows. The switch to group housing systems for gestating sows has resulted in the
development of new feeding systems such as electronic sow feeders (ESF), which provide an
opportunity to meet the dynamically changing energy and nutrient requirements of individual
sows. An ESF is a computerized feeding station where group housed sows are individually
identified when they enter and receive and consume their feed in a protected enclosure.
Technology is currently under development to create multiple feed options within an ESF that
are based on blending high and low nutrient diets to closely meet the dynamic changes in
nutritional demands of each individual gestating sow. When combined with optimal group
housing management , unique ESF feeding strategies can lead to improved sow performance and
longevity.
This review will examine the determinants of nutrient requirements of gestating sows as
well as altering gestating feeding regimes to meet these dynamic requirements. The opportunities
and challenges of managing group housed sows with ESF will also be addressed.
1.2. Determinants of energy and nutrient requirements of gestating sows
Over-feeding nutrients can be wasteful and costly, while under-feeding during gestation
results in the development of metabolic stress to the sow and may compromise sow productivity.
Accurately determining energy and nutrient requirements of sows is critical in order to
implement an optimal feeding program to avoid both over- and under-feeding. Based on the
model by Dourmad et al. (2008), the NRC (2012) developed a mathematical model which
calculates daily response to energy intake and nutrient requirements based on specified
2
conditions for gestating sows. The NRC (2012) model demonstrates that maintenance
requirements along with protein and lipid deposition (Pd and Ld, respectively), the main
determinants of energy and nutrient requirements in the maternal body and the products of
conceptus, change throughout gestation.
Gestating sows will partition feed energy first to satisfy maternal maintenance needs,
followed by growth of the conceptus and lastly maternal Pd and Ld (NRC, 2012). The maternal
maintenance energy requirements are calculated as a function of total body weight (BW), but are
also influenced by factors such as activity levels and environmental temperature. It has been
previously documented that a measurable increase in maintenance energy requirements occurs
when sows spend more than four hours per day standing or walking (Dourmad et al., 2008).
Additionally, environmental conditions, especially cold stress, increase energy requirements.
Sows require extra feed energy for maintaining a constant body temperature when the
environmental temperature is below the lower critical temperature, which is estimated at 20 and
16oC for individual and group housed sows, respectively, and when housed on solid concrete
flooring and in draft-free environments (NRC, 2012). Energy retention in the conceptus
increases exponentially with time, and varies with litter size and average piglet birth weight
(Farmer 2015). According to NRC (2012), energy intake over and above requirements for
maintenance and growth of products of conceptus will be used for both maternal Ld and Pd. The
ratio between maternal Ld and Pd increases with parity, but even in first parity sows, an increase
in energy intake will increase body fatness. Over-feeding of energy during gestation will
increase body fatness, which may contribute to excess body condition at the time of farrowing
and may compromise subsequent reproduction. If under-fed, the sow will mobilize her own body
lipid stores to supply energy to meet energy requirements for maintenance and products of
3
conceptus, which may lead to compromised reproductive performance as well.
According to NRC (2012), the main determinants of dietary bio-available amino acid
requirements are protein deposition, basal endogenous gut losses (related to feed intake and
dietary fibre level), integument losses (a function of body weight), and the post-absorptive
efficiency of amino acid utilization. Gestating sow Pd is broken down into six pools which each
have their own unique amino acid profiles: fetus, placenta and fluids, uterus, mammary tissue,
time dependent maternal Pd, and energy intake dependent maternal Pd. The rate of protein
deposition and amino acid profile in the first four pools, i.e. those directly associated with
reproduction, are reasonably well characterized (Figure 1.1) and can be predicted from
anticipated litter size, piglet birth weight and stage of gestation. The rate of body protein
deposition in the maternal body is influenced by parity, energy intake, stage of gestation and
likely pig genotype; the effects of the latter especially remain to be better characterized (Miller et
al., 2015). Especially in first parity sows, maternal body protein deposition is a major
determinant of dietary amino acid requirements (Figure 1.1). Similar to energy, if the sow is
underfed amino acids she will mobilize her own stores to accommodate the needs of conceptus.
1.3. Different feeding strategies throughout gestation aimed at improving sow performance
Research focusing on relating patterns of intake of total feed, energy or protein (i.e.,
amino acids) during gestation to sow reproductive performance has yielded varying results. In an
extensive review of the scientific literature, Campos et al. (2012) reported that providing extra
feed or energy during late gestation only marginally improved piglet birth weight, and effects
were not consistent between different studies. Several studies demonstrated no effect, while
others indicated improvements in various aspects of production, such as litter size, gestation sow
4
BW gain, lactation sow BW loss and feed intake during lactation. Differences in results amongst
these studies could be attributed to different levels of energy and nutrients supplied, as well as
different durations of time and periods of supplementation. Another important factor to consider
is the use of primiparous sows compared to multiparous sows, which are known to have
differences in nutrient partitioning. There are numerous factors that can contribute to sow
performance and responses to feeding regimes including: initial sow BW and body condition,
litter size, piglet birth weight, litter growth rate, sow milk production, milk components, feed
intake during gestation and lactation, return to estrus interval, and BW changes during lactation.
In summarizing and interpreting data from the literature, studies evaluating the effects of
changing the feeding regime during gestation will be categorized into three different approaches:
altering intake of feed, energy, or amino acids.
1.3.1. Increasing feed intake or bump feeding during gestation
The simplest method of meeting the increasing energy and nutrient requirements of sows
during late gestation is increasing the level of feed supplied during late gestation. It is proposed
that simply increasing feed intake better meets the increasing nutrient demands of the sow. NRC
(2012) outlined a 400 g/day (about 20%) feed intake increase after day 90 of gestation based on
energy requirements, while this increase should be about 40% based on lysine requirements. A
cooperative research study by Cromwell et al. (1989) concluded that additional feed supplied
during late gestation improved reproductive performance. The study involved 1,080 litters where
multiparous sows in the treatment group were fed 1.82 kg/d of a corn and soybean meal based
diet (3.2 Mcal ME, 14% CP) in addition to the levels received by the control group (summer
1.82, winter 2.27 kg/d) from day 90 of gestation until farrowing. Sows fed extra feed in late
5
gestation farrowed an average of 0.35 more pigs/litter, as well as slightly heavier pigs at birth
(1.48 kg vs 1.44 kg) and at weaning (18 days) (Cromwell et al., 1989). A more recent study by
Shelton et al. (2003) yielded slightly conflicting results when 0.9 kg/d of extra feed
(corn/soybean meal based diet containing 3.26 Mcal ME, 0.57% SID lysine) was given after day
90 of gestation (2.09 vs 2.95 kg/d,) . These authors found that increasing feed intake during late
gestation led to a decrease in piglet birth weight in multiparous sows, but an increase in piglet
birth weight in gilts. Additionally, gilts offered extra feed had an increase in subsequent
conception rate compared to the control, whereas sows fed extra feed had reduced conception
rate in subsequent parities. Only in second parity sows did an increased feeding level during late
gestation slightly increase litter weight at weaning. The added cost of extra feed was predicted to
be $3.50-5.00 (US$) per sow per parity, with little to no improvement in sow performance. In
both Shelton et al. (2003) and Cromwell et al. (1989), the sows’ body weight increased with
additional feed intake, as expected. Arguments have been put forward that improved sow
condition at farrowing can increase piglet performance through increased milk output of the sow;
this is supported by evidence from Cromwell et al. (1989), but not Shelton et al. (2003). It
should be noted that in these two studies the increase in daily energy intake during late gestation
was greater than what is recommended by NRC (2012). It is, therefore possible that the extra
feed supplied during late gestation resulted in over-conditioning of sows at farrowing, which can
compromise sow reproductive performance (Young et al., 1991; NRC, 2012).
Increasing sow feeding levels in early to mid-gestation may have an effect on muscle
fiber number, affecting growth rate and feed efficiency of progeny. Secondary muscle fibers
known to improve growth rates in piglets from between day 54 and 90 of gestation and can be
influenced by environmental factors such as nutrition (Dwyer et al., 1994). To study these
6
effects, Lawlor et al. (2007) subjected 238 multiparous sows to one of five different dietary
treatments accomplished by altering feed allowances of a common gestation diet: 1) 30 MJ
DE/day throughout gestation (Control), 2) same as Control but with 60 MJ DE/day from day 25
to 50, 3) same as Control but with 60 MJ DE/day from day 50 to 80, 4) same as Control but with
60 MJ DE/ day from day 25-80 and 5) same as control but with 45 MJ DE/ day from day 80-112.
Across treatments, sows were fed the same diet (3017 kcal/kg DE; 0.62% total lysine). They
found that treatment had no effect on the number of piglets born alive, mean birth weight,
variation in birth weight, weaning weight, or ADG from birth to weaning, although the number
of piglets born dead was significantly greater for treatment three than any other dietary treatment
(Lawlor et al., 2007). Similarly, Cerisuelo et al. (2008) used 90 gilts to evaluate the effect of
increasing feed intake by 50% (2.5-3.0 kg/d based on individual sow BCS, plus 50% for
treatment; diet containing 2899 kcal/kg ME, 0.62% SID lysine) between days 45 and 80 of
gestation on lean gilt body reserves, and performance. Their results indicated that litter weights
at farrowing were slightly greater (17 kg vs. 16.2 kg) at the higher feeding level, but not enough
to justify the additional feed costs. Additionally, increasing feed intake increased the fat reserves
as compared to the controls. Weaning to estrus interval (WEI) and subsequent parity farrowing
rates were not different between treatment groups.
The increased benefits observed by Cromwell et al. (1989) may not be as prominent in
similar studies conducted many years later because of advances in genetics and nutritional
management practices over time. It has been established that sows will catabolize body stores to
meet demands for milk production during lactation; however, debate exists as to whether a
period of body protein or body lipid mobilization during late gestation reduces subsequent sow
performance (Goodband et al., 2013). Alternatively, it has been suggested that sows build up
7
adequate body reserves in early- to mid-gestation and are thus able to minimize the negative
impact of feeding below nutrient requirements during late gestation. Overall, increasing feed
intake during gestation increases sow BW and body fatness. As mentioned earlier, supplying
too much energy to sows may lead to excessive body fatness and compromise subsequent
reproduction. Although altering feed levels is the simplest method to accommodate dynamic
changes in nutrient requirements, it may also result in wasted nutrients if the demands of
individual nutrients change at different rates.
1.3.2. Altering energy levels throughout gestation
A limited number of studies with gestating sows have been conducted focusing on energy
intake only, as intake of energy and other nutrients are often confounded. Most of the studies
mentioned in the previous section may in fact be considered energy intake studies; in these
studies the supply of essential nutrients was generally above estimated requirements, with the
exception of amino acid supply during late gestation.
In a study by Laws et al. (2007), 48 multiparous sows were assigned to one of six
different diets on day 60 of gestation, with the control diet (3088 kcal/kg ME, 13.1% CP) fed at
3 kg/day. The treatment diets were the control diet plus 10% extra ME derived from either: more
of the control diet, palm oil, olive oil, sunflower oil or fish oil. This study found that total litter
size at farrowing, number of piglets born alive and total litter weight were the same across all
treatment groups. Growth rates were also similar between birth and slaughter for pigs from all
dietary treatments. These results demonstrated that increasing maternal energy intake during the
second half of gestation over and above 9.40 Mcal/d ME did not result in improved piglet
performance (Laws et al., 2007). These results are similar to those of Pond et al. (1981). In that
8
study, daily energy intake at day 100 of gestation was doubled (from 6000 kcal DE to 12000 kcal
DE) through the addition of corn starch to a standard corn/soybean meal diet, and no effects on
piglet birth weight, 28-day weaning weight or piglet survival rate were observed.
From these studies it is not apparent if sow productivity can be improved by increasing
the dietary energy content or daily energy intake in gestation sows beyond a constant level (e.g.,
NRC 2012) to meet requirements for gestation and provide sufficient energy to support
maintenance energy requirements and typical gains of the maternal body and products of
conceptus. It can, however, be argued that any metabolic stress on gestating sows may be
avoided by increasing dietary energy content or daily energy intake. To do this, a modest
increase in energy intake during late gestation is recommended to eliminate the mobilization of
body energy reserves for satisfying the increasing energy requirements of the rapidly growing
fetus (NRC, 2012). Altering energy level throughout gestation does not appear to have a positive
correlation with reproductive performance. A common thought is that feeding too much energy
during gestation will greatly increase fat stores and decrease feed intake during lactation,
therefore increasing lactation sow BW change which has negative effects on subsequent parities.
Lawlor et al. (2007) observed a decrease in feed intake, but it was not correlated to increased
weight loss in this case.
1.3.3. Effect of changing protein or amino acid intake levels during gestation
According to McPherson et al. (2004), the amino acid requirements of a sow may not be
reached by providing a constant amount of dietary protein during gestation, as the requirements
for amino acids of both the sow and its fetuses increase throughout gestation. In terms of amino
acid requirement studies for gestating sows, the main focus has been on lysine (NRC, 2012),
9
which is typically the first limiting amino acid in cereal grain and soybean meal based diets. The
NRC (2012) recommendation for dietary SID lysine requirements are 0.52% before day 90 of
gestation and 0.69% thereafter for a ‘typical’ first parity gilt, but these recommendations change
with parity and stage of gestation.
Zhang et al. (2011) conducted a thorough study examining the effects of lysine intake
during middle to late gestation on multiparous sows. On day 30 of gestation, 200 sows were
assigned to one of four dietary lysine levels of 0.46, 0.56, 0.65 and 0.74%, while feed intake was
kept constant at 2.2 kg/d from d 30-80 and 3.0 kg/d from d 80-110 using diets based on corn,
wheat, and soybean meal. Gestation BW gain increased with increasing levels of lysine fed; sows
fed the 0.56, 0.65 and 0.74% lysine diets gained 21.6, 26.6, and 31.2% more body weight,
respectively, compared to sows fed the 0.46% lysine diet. Increasing lysine intake did not affect
the total number of pigs born, but did increase average piglet birth weights (1.28 kg, 1.35 kg,
1.46 kg and 1.47 kg, respectively). Another study by Yang et al. (2008) changed the diets of 36
multiparous sows on day 80 of gestation to have a high or low lysine level (8.0 or 6.0 g/kg),
while maintaining three levels of energy intake (3274, 3322, 3394 kcal/kg ME) of a corn and
soybean meal based diet fed at 3.0 kg/d. In that study, the total number of piglets born was not
affected; however, piglet body weights were 1.45 to 2.4 kg heavier from sows fed the diet with
higher lysine inclusion across all three energy levels. Both studies support that increasing lysine
intake during mid- and late gestation increases piglet body weights, while it has no effect on the
number of piglets born.
Clowes et al. (2002) phase-fed protein to gestating gilts over three successive parities
with control diets containing 0.44% calculated SID lysine. Gilts on the phase-feeding treatment
receiving 0.34% calculated SID lysine from day 0-38, 0.40% calculated SID lysine between day
10
39-74 and 0.56% calculated SID lysine from day 75-115. In this study, phase-feeding protein
during gestation had no significant effect on total gestation maternal growth, litter size, piglet
birthweight, weaned litter size, WEI or subsequent litter size. More recent recommendations for
required dietary lysine levels (NRC, 2012) indicated that the lysine level used in Clowes et al.
(2002) was too low for first and second parity gilts. Based on nitrogen balance observations, it
was concluded that feeding sows to meet their gestational amino acid requirements reduced
urinary and total nitrogen excretion (Clowes et al., 2002).
Amino acids are not only building blocks of protein synthesis but are also used as
precursors for nitrogenous substances essential for whole-body homeostasis (Wu et al., 2010).
There is strong evidence that the members of the arginine family of amino acids have an
important role in placental vascularization and development, especially during the middle of
pregnancy (Wu et al., 2007). This theory is supported by Mateo et al. (2007) who supplemented
1% L-arginine to a corn and soybean meal based diet after day 30 of gestation to gilts and found
that supplementation increased the number of pigs born alive by 22%. However, a study by Li et
al. (2010) observed supplementation of 0.8% L-arginine to a corn and soybean meal diet
between day 0 and 25 of gestation reduced the total number of fetuses at day 25. The timing of
feeding arginine and interactions with other amino acids, needs to be better understood before
clear recommendations can be made about optimum dietary arginine levels for gestating sows
(e.g., NRC 2012).
Based on the above observations, it is clear that increasing dietary amino acid levels is
more beneficial than increasing feed intake, especially during late gestation, as it does not
contribute to excess maternal body lipid deposition which may reduce subsequent sow
reproductive performance. While studies have clearly demonstrated that the amino acid demands
11
of gestating sows change throughout gestation, more research is needed to clarify if more closely
meeting these changing amino acid requirements will improve (long-term) sow reproductive
performance and ultimately, profitability.
1.4. ESF in group housing systems as an alternative to gestation stalls
Conventional practices restrict sow feed intake during gestation to help maximize
economic performance and prevent sows from over-eating and becoming obese, which can
compromise subsequent reproductive performance. Restricted-fed sows often show hunger
motivated activities, such as aggression or stereotypic behaviour, which can lead to reduced
welfare (Terlouw et al., 1991). Traditionally, gestating sows were housed individually in stalls
to allow for individual feeding, which eliminated aggression among feed-restricted sows.
However, due to welfare concerns with regard to space allowance and social behaviour, gestation
stalls are being phased out in Canada (NFACC, 2014). Gestating sow welfare in group housing
has been extensively reviewed (Rhodes et al., 2005). Many large North American corporations,
such as Smithfield Foods and Maple Leaf, have already voluntarily required their pork products
to be sourced from gestation stall free production systems by 2017. As the industry moves
towards reducing the use of stalls during gestation, there is a need for increased research into
feasible alternative housing systems. Group housed sows can be fed collectively (by troughs or
on the floor), or individually using alternative feeding systems (i.e., ESF, free stalls, trickle
feeders or walk-in lock-in stalls). An extensive review on these alternative systems was recently
published by Bench et al. (2013b).
ESF systems are gaining popularity as a tool for feeding group housed sows as they allow
for individualized sow management. Sows entering an ESF station are identified by a
12
transponder and the unit communicates with a central computer, which then instructs the ESF to
deliver a predetermined amount of feed or to refrain from dropping feed if the sow has already
eaten. Current commercial ESF technology has the capacity to provide for the needs of 60 sows
per feeder; this provides ample time for all sows to be fed in a 24 hour period after which the
feed allowances will reset for another day. Barns can use ESF systems with static or dynamic
groups of sows and have an advantage of providing protection to a sow during eating. However,
use of ESF requires the sows to eat sequentially rather than simultaneously as a group (Bench et
al., 2013a), which in certain barn layouts can create competition at the entrance of the feeder. For
example, the design of the entrance gate of an ESF has been shown to influence the incidence of
vulva biting in sows (Bench et al., 2013a). Therefore, both barn and ESF designs are very
important and are continuously improving to increase sow welfare (Brooks, 2003).
Adding fibrous feedstuffs to gestation diets has been implemented to increase satiety of
sows and reduce hunger related behaviour (Farmer, 2015). However, adding fibrous feedstuffs in
ESF diets increases feeding time, which puts more pressure on the ESF in large groups (Bench et
al., 2013a), and could lead to more sow confrontation due to crowding at the entrance to the
ESF.
ESF systems are not without their challenges, including the overhead cost of the system,
which drives producers to attempt to maximize efficiencies. However, increasing the sow
capacity and reducing the feeding time allowed, increases aggressive sow behaviour and vulva
biting (Olsson et al., 2010). As an operating system, ESF require maintenance. In one study
(Olsson et al., 2010), up to 50% of sows visits are non-feeding visits as the sows had already
consumed their daily ration; these extra visits increase wear on the ESF units. Sows also require
training, and although the rate varies greatly between farms, several farmers report in discussion
13
that between 0 and 15% of sows cannot be trained to use ESF and must be culled. This has led to
new builds including equipment for gilt training, to begin training at a younger age, at an added
cost. Lastly a switch to ESF will require education of the barn staff operating the system and
modifications to sow management style.
In addition to managing feed delivery, ESF may also be used for other aspects of sow
management. Work has been conducted with limited success to use feeding order and ESF data
for detection of estrus and health disorders, which could assist in reducing staff workload and
improving sow management (Bressers et al., 1993; Cornou et al., 2008). New technology that
can also be added to ESF offers the possibility to separate selecting and separating due to farrow
animals which can be a powerful management tool (Olsson et al., 2010).
Despite some of the challenges associated with this system, ESF feeding systems allow
the needs of a single sow to be met and for sow management to become computerized and
individualized. Next generation ESF systems will be able to blend multiple diets in order to more
closely match changing, unique nutrient requirements of individual sows, providing benefits to
both the sow and producer.
1.4.1. Determining optimal time of mixing post breeding for group housed sows
With the switch to group housing, new management challenges have arisen. A critical
consideration is the timing of mixing of sows relative to breeding. It has been well established
that stress during the immediate post-breeding period can reduce conception rate and litter size
(Knox et al., 2014).
In certain countries, such as the Netherlands, regulations require sows to be group housed
by day four post-insemination (Spoodler et al., 2009), while mixing by five days post14
insemination is required to be defined as loose-housed in Australia (Australia Pork, 2013).
Furthermore, it appears that reducing stress between the second and fourth weeks post-breeding
is of the utmost importance for successful reproduction. According to Australia Pork (2013), it is
suggested to mix sows on the fifth day after breeding to avoid the recognized two critical
periods: the first critical period being immediately after mating when fertilization occurs and the
second critical period being during week 2 to 4 after breeding when embryo implantation occurs.
A non peer-reviewed study conducted in the Netherlands involved a telephone survey of 900
farms and an additional 70 on commercial farm visits to establish support with current within
country regulations and examined the effect of mixing sows post-insemination. This study
indicated that good sow reproductive performance was achievable when sows were mixed within
4 days post-insemination in any group housing system (van der Peet-Schwering et al., 2009).
Two separate studies found no difference in farrowing rate or litter size when sows were mixed
on days 2, 7, 14, 21, or 28 after breeding (Kirkwood and Zorella, 2005 involving 309 sows;
Cassar et al., 2007, involving 617 sows). In contrast, Knox et al. (2014) found that conception
rates and farrowing rates were lower when sows were mixed between day 3-7 post-service, but
not when mixed between day 13-17, when compared to mixing at day 35. They also reported no
differences in litter size from successful matings.
Another study reported an 8% improvement in farrowing rate from mixing sows in the
post-implantation period (37-46 days post-breeding) compared to pre-implantation (2-12 days
post-breeding) was reported (Strawford, 2006). The same study reported that sows mixed preimplantation engaged in more aggressive encounters than sows mixed post-implantation.
Reducing sow aggression at mixing is thought to be achievable through genetic selection
(Lovendahl, 2005), which may reduce stress at mixing and increase reproductive success.
15
It is clear that multiple factors contribute to reproductive success aside from time of
mixing, including feeding method, environmental conditions, time of year, group familiarity,
static vs. dynamic groups, sow conditioning and sow handling. This could explain the varying
observations among different sow grouping practices. Independent of housing system, with good
management, high levels of production can be achieved for early or late mixing for gestating
sows.
1.5. Conclusion
Gestating sow nutrient requirements are dynamic throughout gestation, as well as across
parity, sow body size and condition, and anticipated level of reproduction. Different feeding
strategies have attempted, though with limited success, to increase sow performance by
modifying diet regime to more closely meet dietary needs. Feeding extra amino acids during late
gestation appears to increase piglet birthweights; this strategy appears to be more beneficial than
providing extra energy (or feed). It is difficult to find supportive research that indicates that
increasing feed intake during gestation is economically beneficial for the sow’s nutritional
management. However meeting sows’ requirements more closely reduces excess nutrients,
which in turn will decrease the environmental impact of sow production. The push towards
group housing currently occurring in the Canadian swine industry has been accompanied by
various challenges, but provides the opportunity to individually feed and manage sows through
improving technology such as ESF.
16
Figure 1.1. NRC (2012) predicted total protein gain (Pd) in separate pools for a gestating gilt. 1
140
120
100
Pd mammary
g/d
80
Pd @ E intake
Pd Uterus
60
Pd @ time
Pd Placenta & fluids
40
Pd Fetus
20
0
1
21
41
61
Day
81
101
Gestating gilt, 140 kg at breeding, anticipated litter size 12.5, anticipated birth weight 1.40 kg, feed intake 2.21 kg/d
of a diet containing 2518 Mcal/kg NE and 5% fermentable fiber content.
1
17
2.0. RESEARCH HYPHOTHESIS AND OBJECTIVES
Improving ESF technology can accommodate the unique and dynamic changes in energy
and nutrient requirements of individual sows throughout gestation. This thesis is based on the
hypothesis that productivity, sow welfare, and nutrient utilization efficiencies can be improved
for group-housed sows by using ESF technology to control the blend of two extreme diets to
meet the changing nutrient requirements of individual sows throughout gestation.
The two main objectives of the work included in this thesis are:
1) Further develop a prototype ESF to take advantage of promising new precision-feeding
technology and improved design in order to maximize sow welfare and minimize
technical failures.
2) Explore the effects of more closely meeting the predicted protein (lysine) and energy
demands of gestating, primiparous sows by blending two feeds according to NRC (2012)
model in comparison to a one phase gestation diet program, on sow reproductive
performance, welfare and nutrient utilization efficiencies.
Sow welfare and reproductive performance were monitored based on sow behaviour (i.e.,
frequency of visits to feeder and incidence of lameness), changes in body weight and back fat
thickness of individual sows, total number of piglets born and born alive per litter, individual
piglet weight at birth and weaning, sow feed intake and body weight change during lactation.
Nutrient utilization efficiencies were calculated as differences in intake and retention.
18
3.0. RESEARCH ELECTRONIC SOW FEEDER DEVELOPMENT 1
Contributiors to this chapter and ESF development include: Curtiss Littlejohn and Mike Fisher (Canarm
Agsystems, Arthur, Ontario), Tymon Hardman (Roberts Onsite, Kitchener, Ontario), Scott Keys (B&R Automation,
Concord, Ontario).
1
3.1. Introduction
Electronic sow feeders (ESF) have existed for more than 25 years as a tool for
individualized feeding of gestating sows in group-housed conditions. With pressure within the
Canadian swine industry to move towards group-housed systems and with significant
advancements in computer technologies, ESF are once again gaining interest as a tool for sow
management. Based on an examination of commercially available units in 2014, key features of
ESF are: (1) overall lay out (Figure 3.1.), (2) design of entry, mid-way and exit gates, (3) feeding
bowl design, (4) sow identification, (5) feeding logic and (6) compatible sow management and
operating software. These features affect aspects of functionality of the system, such as
robustness, capacity (i.e. the maximum number of sows per unit), potential aggression among
sows entering and exiting the system (including injuries), sow comfort while eating, and ease of
use by sows and operators (including training).
In partnership with Canarm Agsystems (Arthur, Ontario) and PigCHAMP (Guelph,
Ontario, Canadian representative; a farms.com company), work was conducted to further
develop an ESF system at the Arkell swine research station (University of Guelph, Guelph, ON,
Canada). Working with local companies provided the advantage of close proximity and the
ability to provide service in a timely manner which is crucial during experimentation. The layout
of this ESF is depicted in Figure 3.1(A) and is advertised to have greater sow capacity than the
other layouts depicted in Figure 3.1. After evaluation and modification of key ESF features
previously mentioned, these ESF would be used for precision feeding research described in
19
further chapters as well as future projects. These feeders are unique in their capability to blend
multiple diets for creating different diets for individual sows. Attempts have also been made to
incorporate a sow body weight scale into the ESF, which would allow for real time adjustments
to individual sow feeding regimens. In total, four ESF were installed in existing partially slatted
group housing pens at the Arkell swine unit, with one station per pen as per the floor plan
outlined in Figure 3.2. These pens allowed for feeding of static groups of 20 to 30 sows each,
well below the advertised capacity of 60 to 80 sows per feeder to ensure access to feeding station
would not be a limiting factor to compromise sow performance.
In this chapter, a detailed discussion of the progression of feeder alterations and
development follows, focusing on (1) structural modifications and feeding logic development,
(2) the development of precision feeding technology, and (3) the preliminary work on scale
design.
3.2. Structural modifications and feeding logic development
For effective use of ESF, the system must be robust to withstand wear and tear induced
by large hungry sows, along with being reliable to minimize downtime. Feeding logic refers to
the sequence of events an ESF completes during a feeding cycle which includes recognition of
individual sows, and delivery of feed and water. This is important to maximize usage (capacity)
of the system, while ensuring all sows are able to consume their daily feed allotment with
minimal interference from other sows. The ESF feeding logic is described in a flow diagram
(Figure 3.3); this figure along with Figure 3.4, outlining the parts of an ESF, are to allow for
better understanding as a detailed description of changes to the ESF structure follows.
20
3.2.1. Entrance gate
Sows must enter the ESF through the entrance gate and when a sow is eating, the
entrance gate serves as a barrier protecting the sow from others trying to enter. The entrance gate
is center opening with two half, saloon style gates powered by one electric actuator, a motor that
converts energy into motion and is responsible for moving a mechanism. Several rebuilds or
modifications to the entrance gate were required to withstand the force of hungry sows
attempting to entering the ESF. The original design and key issues addressed in this section can
be viewed in Figure 3.5. The original doors were opened by flat bar gate arms; these were
quickly bent by the force of sows prying the gates open with their snouts. When the arm strength
of the gate was improved by increasing the thickness of steel and adding reinforcement, it put
more force on the brackets attaching the arms to the gates, which then twisted out of shape. To
remove some of the torsion on the gates when sows were trying to push them open, door stops
were placed on the floor. The first door stop became an obstruction to sows walking over it, so
the final version has no sharp edges and a height of less than 3 cm, decreasing the chance of
sows to stub their forelimb or hooves as they move into the feeder.
The entrance gate electric actuators were upgraded to a higher strength (Linear Actuator
LA36, LINAK, Nordberg, Denmark), as the initial models (High Speed Linear Actuator PA-15,
Progressive Automations, Richmond, British Columbia, Canada) were not strong enough to
prevent sows from prying open the gate. The higher power electric actuators now also contain a
clutch function; this prevents any sow from being squeezed when the gate is closing, which may
lead to injuries. A slide mechanism is present that moves with the actuator to allow the gates to
open and close; this mechanism required strengthening and heavier bolting to the ESF frame.
21
Another change to the entrance gate was the location and type of door closing sensor
(different from the sow sensor that initiates door closing). Originally, a metal detecting sensor
was positioned to sense the position of the door arm. Due to the high incidence of damage,
several alternative locations and directions were tried before abandoning use of a metal detecting
sensor. Now magnetic sensors are attached directly onto the new actuators to monitor the gate
position (Magnetic Field Switches, LINAK, Nordberg, Denmark), which has proven more
reliable. A sow sensor detects when a sow has entered the station and signals for the entrance
gate to begin closing; this sensor was also upgraded to improve reliability and add adjustability
for sow height.
After the completion of this research project, the ESF entrance gates have subsequently
been modified further by the manufacturer, changing the leverage on the bar gate arms, and
replacing the slide mechanism.
3.2.2. Mid-Way gate
The Mid-Way gate consists of a door that is placed just beyond the feeding area. This
gate along with the length of the ESF creates a post-feeding area, where a sow is protected after
feeding and before re-entering the pen. The door styles described in the following section are
presented in Figure 3.6 for clarification.
Originally, the door for the Mid-Way gate extended ¾ of the way across the feeding
station. The problem noted with this door was that sows present in the post-feeding area were
able to open the mid-way gate and enter the feeding area if they got in going the wrong direction
past the exit gate. Thus, previously fed sows were given the opportunity to steal feed and cause
sow congestion inside the ESF. Initially this ¾ door was replaced by a full door to prevent sows
22
from re-entering the feeding area. Installation of a full door created a challenge when training
sows to open it. Sows could often be seen trying to push their snout through the gap on the hinge
side of this door, as that was visually the largest opening, which did not open the door and this
could lead to sow injury and frustration. As well, due to the added width of a full door, it failed
to close behind sows that had moved into the post-feeding area; this allowed sows to back up and
steal feed from subsequent sows that had just accessed the feeding area.
Subsequently a saloon style Mid-Way gate was designed to allow sows to easily open in
order to leave the feeding area. Saloon style doors are consistent with the other gates on the
unit, which open through the center. These doors also leave the largest possible gap in the center
directing sows where to go. This gap is increased at the bottom of the gate, which allows sows
with natural rooting instincts, to more easily open the gate. While gate stops were placed on the
ESF frame at the top of the gates, gate stops for the saloon style doors were changed to be
upward facing to reduce the risk of sow injuries. The reduced length of the individual saloon
doors enables the gate to close more quickly behind sows leaving the feeding area, reducing the
incidence of sows re-entering the feeding area from the post-feeding area.
3.2.3 Exit gate
The exit gate is located at the end of the ESF; the main purpose of the exit gate is to
prevent sows from entering the unit from the wrong end of the feeder. Figure 3.7 provides a
labelled diagram of a half open exit gate (springs removed for clarity). This gate is similar in
design to the entrance gate, but is opened by sows leaving the post-feeding area; rather than
using an actuator, the exit gate is closed by two springs attached to the doors. These springs are
placed between the ESF frame and bolts in the top bar of the exit doors. Extra holes were added
23
to the top bar of the door on the exit gates for two reasons; (1) they allow for adjustable spring
tension and, (2) with the addition of a longer bolt in an unoccupied hole, a pinch free handle is
created for use when training sows to use the system. Opaque panels were added to the exit gate
to direct sows to push in the middle of the gate where it opens, instead of trying to push their
heads through the largest space visible which was between bars within the doors. A doorstop was
added to the exit gate, similar to the one at the entrance gate; the doorstop provided more
strength to the gate and greatly reduced the number of sows entering the station through the exit
gate. The door stop also required alteration to a height less than 3 cm as it was noticed that the
original door stop obstructed the sows’ rooting motions that are necessary to open the door. Just
before the exit gate, an anti-lie down bar was added in the post-feeding area to prevent sows
from lying down in the ESF, and to decrease congestion.
3.2.4. Feed bowl
A key feature of the straight flow ESF, is the moveable feed bowl, which allows for sows
to move through them in a straight line for increased speed. The feed bowl is kept retracted until
an eligible sow (one that has not yet consumed its daily feed allowance) is present in the feeding
area; at that time the bowl slides into the feeding area and is filled with feed and water. After two
actuators failed, the feed bowl actuator was upgraded to handle presenting and retracting the feed
bowl; the electric actuator has magnetic sensors similar to the actuator used at the entrance gate
(Linear Actuator LA36, LINAK, Nordberg, Denmark). The location for feed delivery into the
bowl was changed which improved speed of feed consumption by sows (Figure 3.8). Originally
the feed was delivered into the back corner of the bowl where it was more difficult for sows to
reach. By changing the location of feed delivery to the front edge of the bowl, the feed was more
24
accessible to the sow and allowed her to consume feed in a more natural position. This change
also increased the mixing of feed and water within the feed bowl, increasing moisture content of
feed and decreasing consumption time. A lip was added onto the entire top edge of the feed bowl
to reduce feed wastage that resulted from sows rooting feed onto the floor.
A challenge with moving parts is that they create pinch points for both humans and sows.
To avoid sow snouts or feet from getting caught when the feeder is retracted, a programming
code was inserted to return the feed bowl to the feeding area when too much force is required to
retract the bowl, such as when a foot or snout is caught in the bowl. The programming code
enables the system to then try retracting the bowl again, giving the sow a chance to move their
foot or snout.
A stainless steel plate is located beneath the extended feed bowl to allow for smooth bowl
movement when the bowl is made available for providing feed and water, or to be retracted when
the individual sow has consumed her daily feed allotment or has left the feeding area. This
stainless steel plate created a slippery surface for sows to cross post-feeding. Anti-slip traction
bars were added to decrease sows from slipping and to increase sow safety.
3.2.5. RFID Antenna
The antenna on ESF is responsible for identifying the sow in the feeding area; individual
sow identification is then communicated to the computer to initiate feed delivery or to open the
entrance gate if the sow has already consumed her daily allotment of feed. The antenna is
positioned on the sidewall of the feeding area and immediately above the feed bowl (Figure 3.8).
Signal interference from steel structures or other electronic devices created a challenge for
getting accurate and consistent tag signal readings. At the Arkell research station, the change
25
from Full Duplex to Half Duplex sow RFID tags (Allflex technologies, Texas, USA) improved
tag signal recognition. Originally, the ESF included a beeper on the antenna that signalled a
successful tag signal reading. This beeper was removed but eventually replaced on research
units by a light with variable colours to reflect the feeder status, providing quick feedback to
staff at the unit (Appendix 1).
Considerable effort was made to improve the consistency of tag signal readings.
Changing internal pin configurations of the antenna reader changed the amperage of the antenna
and different configurations worked better in different locations. The details and rationale behind
this remains unclear but is thought to be linked to amount of potential grounding, and electrical
and radio interference present in different locations and on different units. Trial and error testing
was used to establish the final settings for each individual ESF. Multiple alterations of size,
shape and location of the antenna were also made in an attempt to increase accurate reading
frequency.
Each ESF in a feeding system for a barn will have its own RFID reader and antenna to
identify sows. While the Arkell system had 4 ESF with 4 RFID readers, it was necessary to add
a fifth RFID reader without an antenna. This fifth reader resided in the central power box and
did not function; however, it served as the ‘master’ antenna which allowed the computer
program to operate all the other readers simultaneously and independently.
Even though improvements in tag readings were observed, continued effort will be
required in this area. This applies in particular to the Arkell unit where antennas are placed on
either side (left or right) of the feeding area. Effective tag signal reading requires the tag to be in
the ear that is closest to the antenna.
26
3.2.6. Feeding logic
The feeding logic is the sequence of events an ESF completes during a feeding cycle,
beginning with a sow entering the ESF and ending with the entrance gate opening for the next
sow (Figure 3.3). Several modifications were made to improve ESF functionality, and influence
sow behaviour to improve animal flow.
Originally, the ESF operational software program was coded to present the feed bowl
immediately following the closing of the entrance gate after a sow had entered the feeding area.
This was adjusted to only present the bowl after the sow was identified and when the sow present
in the feeder had not yet consumed fully its daily feed allowance. This reduced the incentive for
sows to cycle back into the ESF to clean up any residual feed that was left in the feed bowl by
other sows, reducing the amount of time sows spend in the ESF. This is important as upwards of
50% of sow visits to ESF can be non-feeding events (Olsson et al., 2011). Next, a delay was
added from the time the entrance gate closed behind a sow until the RFID reader attempts to
capture the tag signal from the sow in the feeder. This delay allows the sow that just entered the
feeder more time to push out the previous sow or to get into a comfortable stance with her head
near the feed bowl which brings her ear tag transponder closer to the antenna. Initially, water and
feed were delivered with their own separate timing cycles; this was modified, pairing water
delivery directly to feed delivery such that both would be dispensed at the same time creating a
more uniform blend. This was done to ensure water was delivered when needed and to increase
eating rate. Another logic alteration was to ensure feed and water delivery were stopped if a sow
left the ESF before she fully consumed her daily feed allowance for feed and water. Based on
using the system for several months, other minor glitches were fixed as well including the ESF
becoming stuck in various states or failing to start at the beginning of a new day.
27
3.2.7. Feeding settings
The ESF feeding settings are adjustable and include time settings for the following
parameters: the timing and frequency of feed and water delivery, the amount of time allowed to
clean the bowl, and the amount of time for the sow to clean the floor (whiling time). All of these
settings contribute to the overall feeding time per sow. Olsson et al. (2011) conducted an
objective study of 3 commercial operations using ESF and found the balance between shortening
feeding time and increasing feeder capacity had a large effect on sow welfare as indicated by
lesion and fight scoring. After exploring various feed cycle settings and in consultation with an
expert (T. Parson; University of Pennsylvania, personal communications, 2015), the values were
derived and used with minor adjustments in the experiment described in the following chapters
(Table 3.1).
These settings resulted in a feed delivery rate that encouraged sows to eat quickly as feed
was delivered at a quick pace. However, the increased amount of time between the last delivery
of feed (and water) provided enough opportunity for sows to clean out the bowl, and even leave
the feeding area before the bowl was retracted. In this manner, the likelihood that the next sow is
allowed entry into the feeding area before the previous sow leaves is reduced, lowering the
chance of sow congestion in the feeding area.
The timing of antenna searching for tag readings was adjusted to get the most accurate
readings and eliminate lost readings and feedings if a sow walks out of the ESF for any reason.
Lastly, flow regulators were added to the water line such that water flow into the feed bowl is
consistent regardless of pressure on the line due to other water uses in the barn, which allows the
settings to be more reliable
28
Currently, the Canarm system only has one setting for feed cycle times for all the units on
a particular swine unit. Providing an opportunity to adjust these settings per ESF unit may be
beneficial based on sow-to-sow variation in feeding behaviour (i.e. timid, primiparous sows
versus multiparous sows). These settings may also be altered when a new group of sows is first
introduced and trained on the unit.
3.2.8. Safety modifications
Sow and barn staff safety is of the upmost importance in operating ESF. Any sharp
edges were removed after laser cutting of stainless steel plates while standard hex head bolts
were switched to carriage head bolts wherever possible, to reduce risk of injuries. An important
feature for future installations is the capability to have overriding manual controls at the unit to
assist with training but also for shutting down the ESF at the unit itself, in emergency situations.
3.3. Developing and implementing Stepper Motors for accurate feed delivery
The major difference between precision feeding ESF and current commercial units is the
capability of the ESF to blend precise amounts of feed for individual sows from up to four
different basal diets that are delivered to each ESF. For increased precision, the regular electric
motors for feed delivery were replaced with stepper motors. A stepper motor is a brushless
electric motor where the rotation is divided into a number of equal steps (Personal
communications T. Hardmon, 2014). The motor moves in discrete steps, due to multiple electric
coils which are organized into phases; the motor then moves by energizing each of these phases
individually, providing increased control of the number of rotations, speed and torque at low
speeds (Personal communications T. Hardmon, 2014). Rather than simply running a motor for a
29
set amount of time, the motor can be controlled to execute a predefined number of rotations. For
precision feed delivery, a different stepper motor was used for each of the basal diets (Figure
3.4), and was connected to an auger. Initially the spacing between auger flightings was
minimized, but this caused pelleted feed to be stuck and resulted in inconsistent feed delivery
rates. Some effort was required to position and secure the auger and stepper motor in such a way
that it did not cause any friction with the feed bins and to reduce leakage of feed when the unit
was shaken but the motor was not operating; this was accomplished by adding a collar to the
auger, and thus eliminating gravitational flow.
For the current research project, the ESF was set to blend two different basal diets
(Chapter 4). For each feed delivery and each individual sow, the firmware identified the feed
allotted values for both basal diets, which were operator generated inputs (Chapter 4). The
system then calculated how many turns of each motor are required to deliver the allotted feed,
based on a previously calibrated dispense rate for each motor per 180 degree rotation. The speed
of the motor associated with less feed delivery was adjusted to ensure the relative blend for the
two basal diets was consistent.
3.3.1. Calibration
Accurate feed delivery relies on accurate calibration of each motor’s dispense rate. The
original method for calibration started and stopped each motor for a movement of 180 degrees.
This required multiple starts and stops of each motor with pauses between individual movements
to collect the amount of dispensed feed and determine the average amount of feed dispensed per
180 degrees rotation. The value was then used as a fixed value and was included directly into the
software code. However, it was discovered using a manually positioned identification tag that
30
this approach did not properly represent the amount of feed dispensed to sows, when feed
delivery was continuous rather than intermittent. Moreover, this calibration needs to be
conducted repeatedly to account for variation in feed density across basal diets and time. To
allow for routine calibration of each stepper motor, a routine was developed that also included an
interface in the software that requires simple data entry from ESF operators (Appendix 2).
Essentially, the interface is used to deliver a specified and typical amount of feed (e.g., 1000 g);
this feed is then collected, weighed and the amount entered into the interface. The discrepancy
between targeted and actual feed delivery is then calculated, and adjustments to account for this
discrepancy are made automatically. Results from this simple calibration method are presented
in Table 3.2; there was good repeatability of feed delivery based on the % error calculated as the
difference between prescribed amount of feed and actual amount delivered.
3.4. Developing and evaluating scale feature
A sow body weight scale feature was tested in the precision feeding ESF, to allow for
routine measurement of sow body weight and to adjust feeding regimes in real time based on
discrepancies between actual and targeted sow body weight changes. Body weight scales are
used widely in finishing pig auto sorters that are placed in growing-finishing barns (Whittington
and Schneider, 2004). The placement of a functional scale within ESF presents multiple
challenges, as sows are larger than growing pigs and also more active around the time of feeding.
The original scale (design number 1; Figure 3.9) used a hanging basket style with load cells
placed at either end, similar to what is used in growing pig auto sorters. The scale was placed in
the feeding area, just behind the area where the feed bowl was presented. The scale recalibrated
itself once a day during the system reset before any animals were fed.
31
3.4.1 Body weight scale testing
The original logic to eliminate erratic readings was to discard the highest and the lowest
body weight values if they differed greater than 25 kg from the average, and then provide one
weight measurement that is the mean of the remaining values. However, this method gave
unreliable BW measures. So to test scale design 1, 17 multiparous gestating sows (mean parity =
3) were used. They were previously trained on a ESF unit, and were placed with a unit that was
fitted with the weigh scale. The sows were weighed manually using a separate scale on day 1
before they had consumed their feed and the ESF scale was recalibrated using the bodyweights
of two humans. During a 7 day period, body weights of each sow were measured and recorded in
a SQL table. Body weight readings were obtained at 2 second intervals and began when the
entrance gate closed behind an animal until the end of a feeding cycle or the opening of the
entrance gate if the sow had already eaten. Sows were weighed manually again after they had all
consumed their daily allotted feed on day 8.
On day 8 during routine observations, it was noticed the scale had become unbolted from
the frame of the unit so any data collected on day 8 and later day 6 and 7 was disregarded.
Body weights recorded on days 1 to 5 yielded a larger number of extreme and erratic values. In
order to eliminate erratic values, only weight values that were recorded when sows were actively
feeding or between the time when feed delivery ended and prior to feed bowl retraction. The
daily mean, median, mode, maximum and minimum body weights for each feeding visit to the
ESF during this time period were calculated and compared to the manually obtained initial and
final body weights for individual sows. The mean, standard deviation, and range using this
approach (Table 3.3), did not yield accurate measures of BW for individual visits to the ESF. In a
random subsample of 4 animals across 5 days of BW measurements, the range of BW measures
32
failed to capture the actual measured BW 55% of the time. It appears the activity of eating
interferes with obtaining reasonably accurate estimates of BW.
The second approach to accurately measure sow BW was based on trying to avoid the
time during eating for BW measurements as perhaps part of the sow’s BW is resting on the lip of
the feed bowl. Body weights were used after the sow had entered the feeder but before feeding
began; which included the time when waiting for a sow to finish entering, looking for tag read,
and searching PigCHAMP information. Events where a previously fed sow recycled through the
feeder were discarded. Again, the daily mean, median, mode, maximum and minimum body
weights for each feeding visit to the ESF during this time period were calculated and compared
to the manually obtained initial and final BW for individual sows. The mean, standard deviation,
and range using this approach (Table 3.4) did not yield accurate measures of BW for individual
visits to the ESF. In a random subsample of 4 animals across 5 days of BW measurements, the
reported range of BW measures failed to capture the actual measured BW 65% of the time.
There appears too much motion for accurate BW measurements when the sow enters the station
and is waiting for feed delivery.
There was no observable pattern for errors using either method of weight measurements
between days or related to time of day.
3.4.2 Summary
Based on these preliminary observations, substantially more development is required to
obtain accurate BW measurements for sows using ESF. An important approach to consider is to
place the BW scale outside of the ESF in the case of an exit lane for multiple ESF. It will be
difficult to get accurate sow weights within the feeding station given the large amount of sow
33
movement while sows are eating, the still high incidence of sow congestion in the ESF during
testing, sows stepping off the scale to reach into the feed bowl and the occurrence of the sows’
chin/neck resting on the bowl while eating.
Suggested modifications to scale design 1 to improve reliability include: changing the
floor plate shape and design to allow for greater clearance underneath the scale floor, reducing
debris build up underneath the scale, placing more anti-slip bars on the scale floor, changing
design to allow for better drainage, increasing the number of load cells from 2 to 4, adding a pin
beneath the scale to reduce scale movement, moving the scale closer to the bowl to reduce sows
stepping off the front edge of the scale, and adding a moveable joint above the load cell
designed to maintain the scale floor in a horizontal position. These and other design changes
were implemented in scale design 2 but research was completed before further testing could be
conducted in a meaningful manner.
3.5. Conclusions and implications
Development of a precision feeding ESF unit began with structural modifications and
feeding logic development that included changes to the entrance, mid-way and exit gates, the
feed bowl, RFID Antenna, and the feeding logic in which all changes focussed on increasing
ESF robustness and reliability. These changes were accompanied by fine tuning of feeding
settings for improved sow flow though the ESF. Many of these alterations are also applicable and
have been implemented on commercial ESF.
The development and implementation of stepper motors for accurate feed delivery sets
these units apart from commercial units as it allows the blending of multiple feeds at very small
and precise amounts. The retooled interface allows for quick and accurate calibration to ensure
34
high levels of accuracy even with changing feed densities between batches. Lastly preliminary
development and evaluation of a scale feature was attempted to advance to real time adjustments
of feed intake based on changing sow BW in feeding programs.
In conclusion, the modified research ESF unit is sufficiently developed for evaluation of
the concept of precision feeding gestating sows (Chapters 4 and 5), but without the scale feature.
35
Table 3.1. Canarm Research ESF Feeding settings used for described research.
Parameter
Time from gate closed until searching for tag reading begins, s 1
Frequency of tag readings during cycle, s1
First feed delivery duration, s
Feed delivery interval, s
Other feed delivery duration, s
Interval before bowl retraction, s
Whiling time, s
Water delivery duration, s2
Approximate total feeding time, if free of tag read errors, min
25
7
10
40
10
210
35
6
8-12
These are not available as adjustable parameters at this time, as they are embedding in the operating software but
adjustments were made alongside of programmers.
2
Currently labelled as ‘First Dispense water duration’ but was modified for University of Guelph system to start
delivery at the same time as feed.
1
36
Table 3.2.Results of testing different motor calibration techniques for Canarm ESF stepper motors, using
ESF Unit 2 as an example.
Dispense rate1 , g/180o
Prescribed Feed2, kg
Test 13, kg
Test 2, kg
Test 3, kg
Test 4, kg
Test 5, kg
Average % error4
New Dispense Rate5, g/180o
Prescribed Feed2, kg
Test 1, kg
Test 2, kg
Test 3, kg
Average % error
1
18.80
2.00
1.670
1.726
1.734
1.746
1.758
-13.76
16.23
2.00
1.960
1.977
2.020
-0.72
Motor
2
19.0
2.20
1.554
1.630
1.644
1.640
1.682
-19.28
15.34
2.20
2.104
2.064
2.054
2.67
Dispense rate determined by original methodology of catching feed from 180 degree motor jog.
Prescribed feed is the amount of feed requested to be delivered by PigCHAMP software.
3
The amount of feed actually delivered during individual tests.
4
Error was calculated by the % difference from prescribed and actual feed delivered.
5
New Dispense rate determined by using measure of error to adjust previous dispense rate.
1
2
37
Table 3.3. Scale 1 BW measurements during feed delivery and feed bowl retraction for a random
subsample of 4 sows.
Sow
Manual
measurement
of BW (kg)
1
336.4
2
302.8
3
300.4
4
330.0
BW measurements (kg) using ESF scale
Day
Mean
SD
Range
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
422.9
341.2
279.9
429.6
289.3
150.1
230.3
341.8
322.3
254.1
430.0
285.4
239.4
358.4
181.3
415.0
306.0
257.0
347.4
214.7
19.9
10.0
27.7
10.2
50.2
29.3
22.1
23.3
21.2
18.7
35.0
29.7
27.5
31.6
35.9
14.1
10.2
9.9
8.3
164.4
*Days where measured BW were not within the range of weights given from the scale.
38
311-455
298-363
183-318*
392-450*
138-353
79-209*
150-260*
280-377
266-353
194-278*
326-499*
213-323
163-279*
278-408
73-256*
370-444*
245-326*
179-274*
329-367
-326-384
Table 3.4. Scale 1 BW measurements during the time after the sow had entered the ESF but before feed
delivery had begun for a random subsample of 4 sows.
Sow
Manual
measurement
of BW (kg)
1
250.3
2
305.6
3
227.8
4
300.4
BW measurements (kg) using ESF scale
Day
Mean
SD
Range
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
166.9
12.3
-17.9
105.4
31.9
367.1
224.2
193
305.8
265.7
333.3
110.7
41.8
108.4
-46.6
444.7
243
229.3
339.6
165.2
99.2
58.5
79.6
89.2
109.8
49.7
48.1
35.2
26.7
60.8
11.2
53.9
95.1
86.3
54.4
36.7
46.8
17.0
42.0
10.5
*Days where measured BW were not within the range of weights given from the scale.
39
100-378
-46-98*
-89-154*
36-292
-42-233*
257-405
128-259*
158-253*
269-335
130-305
319-351*
42-191*
-68-131*
17-282
-89-66*
389-489*
154-284*
211-251*
275-395
154-179*
Figure 3.1. Primary flow designs and examples of commercially available ESF in 2014.
A. The sow enters the station and a moveable bowl becomes available, retracting when feeding
is finished allowing the sow to continue out of the feeder in a straight path.
Examples
-Canarm, Sow Choice System
-AP Systems/Schauer, Compident
B. The sow enters the station eating from a stationary trough on her left or right side, exiting
when feeding is finished on an angle opposite to the feeding trough.
Examples
-Nedap, Velos
-PigTek/Chore-Time, Mannebeck
-Osborne, TEAM feeder
-BigDutchmen, Callmatic
-WEDA
C. The sow enters the station eating from a stationary trough directly in front of her allowing a
straight posture, exiting to her left or right side when feeding is finished.
Examples
MPS, Cheiftian
SKIOLD, Datamix
Insentec, Compufeeder
40
Figure 3.2. Pen layout of loose housing pens containing ESF at the Arkell Swine Research
Station (University of Guelph, Guelph, ON, Canada).1
Room consists of four partially slatted pens each containing one ESF (indicated in blue) with an area of 66.5 m2 for
sow housing. Grey indicates slatted flooring, white is solid concrete floors, dark black lines 4 ft concrete walls, light
grey spindle penning, and four hanging water nipples where above the slatted area in each pen.
1
41
Figure 3.3. Flow diagram of Canarm, SowChoice Systems ESF unit operations.
42
Figure 3.4. Labelled view of Canarm Research ESF.
43
Figure 3.5. Original entrance gate and described issues.
44
Figure 3.6. Mid-way gate designs for Canarm research ESF.
45
Figure 3.7. Half open exit gate (springs removed) on Canarm research ESF.
Figure 3.8. Feeding bowl and RFID Antenna on Canarm research ESF.
46
Figure 3.9. ESF sow scale design 1.
47
4.0 FEEDING PROGRAM DEVELOPMENT
4.1. Introduction
The NRC (2012) provides a mathematical model which calculates the response to energy
intake and requirements of gestating sows for amino acids, calcium and phosphorus based on key
determinants of energy and nutrient requirements. The primary use of the gestating sow model,
programmed in Microsoft ExcelTM, is to estimate nutrient requirements of gestating sows under
varying dietary, environmental, and performance conditions. With increased pressure to make
swine production both more cost efficient and environmentally friendly, it has become important
to more closely meet nutrient requirements of sows. To achieve the latter, a revised version of
the NRC (2012) gestating sow nutrient requirement model may be integrated with computer
controlled, electronic sow feeder (ESF).
The objective of the following work was to modify the programming code of the NRC
(2012) gestating sow nutrient requirement model to generate daily energy and nutrient
requirements for individual gestating sows, and to link this model to use with ESF for precision
feeding management of individual sows that were housed in groups. The modified NRC (2012)
gestating sow nutrient requirements model provides the potential to adjust both the feeding level
and diet composition for individual sows according to sow body weight, desired body condition
score, parity, activity level, anticipated litter size, and anticipated mean piglet birth weight.
Ultimately, this approach would allow individual sow feed management decisions to be made in
real time using computer controlled ESF. This could potentially reduce feed costs and nutrient
losses into the environment, while improving sow reproductive performance, sow welfare, and
quality of newborn piglets.
48
4.2. Modifying the NRC model to calculate required daily ME intake
The NRC (2012) model determines nutrient requirements and predicts sow growth based
on the user-specified level of energy intake. Modification of the model was required to change
the use of the program and adapt it into a tool for feeding gestating sows.
First, the model needed to be changed from predicting a response based on energy intake
to predicting energy requirements, and by extension feed intake. In its current form, the model
calculates the rate of maternal body protein deposition (Pd) and body lipid deposition (Ld), based
on energy intake over and above energy requirements for maintenance and growth of products of
conceptus. To change the model to an energy intake prediction model, daily Ld was held
constant throughout gestation (to ensure that model predicted energy intake well above all sow
requirements) and treated as a model input, with default values suggested for each parity. Then a
rearrangement and substitution of equations for Pd (Eq. 4-1. NRC 2012), and associated Ld (Eq.
4-2. NRC 2012) yields calculated ME requirements for each day of gestation (Eq 4-3).
@
=
(
=
(
,
,
−
−
−
@
)
Eq 4-1 (NRC, 2012; Eq. 8-62)
∗
)
Eq 4-2 (NRC, 2012; Eq. 8-64)
,
−
=
∗
(1 −
∗
∗
)
+
∗
Eq 4-3
The abbreviations in these equations represent the following:
Pd @ ME intake: Maternal body protein deposition associated with ME intake (g/d)
Slope: slope of linear relationship between maternal body protein deposition versus ME intake (g/Kcal)
ME intake, required: ME required from the diet to fulfill needs of all Pd and Ld
49
MEm0: ME maintenance on day 0
MEreq1: ME intake to support ME requirements for maintenance, growth of products of conceptus and
maternal body protein deposition defined by day of gestation (Kcal/d)
MEp: Energy cost of Pd, kcal/g
MEf : Energy cost of Ld, kcal/g
When targeting similar mean daily Ld over the entire gestation period for parity 1 sows,
differences between default values (i.e., feed intake) according to the original NRC (2012) model
and the revised model are very small (Figure 4.1). However, the distribution of energy intake is
slightly different between the two models, which is the result of lower energy intake during early
gestation in the revised model, and therefore reduced body weight gains and lower maintenance
energy requirements. This is followed by greater energy intakes and energy supplied to meet
increasing demands of conceptus in late gestation. As a result, overall energy intakes, feed
intake- dependent Pd, and thus maternal Pd, are predicted to be slightly lower in the revised
model.
For the purposes of the research described in subsequent chapters in this thesis, the
targeted daily Ld was based on the average maternal Ld across the gestation period of the generic
sow outlined in NRC (2012), with unique values for parity 1, 2, 3 and 4+, respectively (Table
4.1)
4.3. Determining feeding level and recipe of diet for each sow
The model was further modified to generate a database, representing the target feeding
level and blend between two basal diets for each day of gestation for up to 200 individual sows.
To achieve a blended diet for each day of gestation and each sow, two isocaloric basal diets were
used: one high and one low in protein, with standardized ileal digestible (SID) lysine contents of
0.80 and 0.20%, respectively. These two SID lysine levels were more extreme than the highest
50
and lowest estimated SID lysine requirements of any sow and on any day of gestation. The
energy content was kept identical between the two diets, and it was ensured that other nutrients
were not limiting in either of these two basal diets. The required intakes of each of the two basal
diets were calculated using knowledge of SID lysine and ME contents for these diets, as well as
model generated estimates of SID lysine and ME requirements. The calculated intakes were
stored in the data base for each sow and each day of gestation (knowing Eq 4-4, Eq 4-5, Eq 4-6
and Eq 4-7 can be used to determine the amounts of Feed A and Feed B).
A*a1+ B*b1=TL
Eq 4-4
A+B= T
B=T∗
A= T-B
( −
( −
Eq 4-5
)
)
Eq 4-6
Eq 4-7
The abbreviations in these equations represent the following:
A: Amount of Feed A (kg/d)
a1: Lys level of Feed A (g/kg)
B: Amount of Feed B (kg/d)
b1: Lys level of Feed B (g/kg)
T: Total amount of feed (kg/d)
L: Lys requirements (g/kg)
In a similar manner, and created using more extensive mathematics, three basal diets may
be used to generate blended diets that satisfy requirements for three nutrients (i.e., energy, lysine
and phosphorus). The mathematics for this scenario are presented in Appendix 3.
For the purpose of this research project, calculations were conducted using the revised NRC
(2012) using two basal diets and run manually for each individual sow, identified by sequential
numbers between 1 and 200. A ‘recipe.csv’ file was then created and imported into the
PigCHAMP program that controls ESF feed delivery. The original PigCHAMP feeding curves
51
were overridden to allow each sow to be fed for each individual day of gestation and according
to the data in the imported .csv file (Figure 4.2).
4.4. Adjusting feeding curves for individual animals deviating from predicted growth curve
Due to biological variation between animals, there will be discrepancies between model
predicted and observed animal performance, which will be reflected in differences between
observed and predicted sow BW gain and back fat thickness. In gestating sows, this discrepancy
may be attributed largely to between animal variations in maintenance energy requirements,
which are a main contributor to total dietary energy requirements and is influenced by factors
such as animal activity (NRC, 2012). Furthermore, Ld that is associated with Pd, in the so-called
energy intake dependent maternal Pd, is known to vary with parity and possibly genotype (NRC,
2012). The original NRC (2012) model provides an opportunity to manually calibrate the model
to observed performance to better predict requirements for sows, by making proportional
adjustments to maintenance energy requirements or the slope of the linear relationship between
maternal Pd and energy intake. However, these adjustments are based on BW and back fat
thickness at breeding and farrowing only, which does not allow for periodic feeding program
assessment within a gestation period. With further advancement towards precision feeding, this
logic can be incorporated into the model to automatically adjust for discrepancies between
observed and predicted changes in BW gain and back fat thickness throughout gestation.
In the original model and when energy intake is kept constant, an increase in maintenance
energy requirements will result in reduced maternal Ld (and Pd). In contrast, in the revised
model and when target Ld is kept constant, an increase in maintenance energy requirements will
result in an increase in required energy and feed intake calculation. This key difference will
52
have consequences when implementing automated calculations to calibrate the model to
individual sows, i.e. to match model predicted and observed changes in BW gain and back fat
thickness. Such calibration should become part of real-time precision feeding whereby sow BW
(and possibly back fat thickness) is monitored continuously and changes to feeding levels and
diet compositions are calculated automatically.
For precision feeding, some logic is required to adjust feeding programs for individual
sows based on previously observed performance. For example, when actual BW gain during
early gestation is greater than targeted BW gain, then the maintenance energy requirements
should be adjusted downwards and energy intake during the remainder of the gestation period
should be decreased to ensure that BW gain over the entire gestation period matches target BW
gain. In general terms, first the model is calibrated to match observed BW gain, and the
discrepancy between targeted and actual Ld is calculated, and then the targeted Ld is adjusted for
the remainder of the gestation period. Provided that the discrepancy between actual and observed
BW gains can be attributed to Ld - which will be revealed through calibrating maintenance
energy requirements and the slope of the linear relationship between and maternal Pd and energy
intake - then the discrepancy between observed and predicted Ld can be calculated based on
monitored past performance (i.e., within the first part of the gestation period). This will then
allow for adjustment of the targeted Ld, and as a consequence targeted energy and feed intakes,
for the remainder of the gestation period (Eq. 4-8).
=
−
(
−
)∗
Eq 4-8
The abbreviations in these equations represent the following:
matLdn: New Ld input for remainder of gestation period
matLdi: Initial targeted Ld prior to model calibration
matLde: Estimated Ld for monitored period from model calibration
53
days: number of days performance monitored
days remaining: number days remaining in gestation period
These adjustments can be applied to adjust energy and feed intakes for the remainder of
gestation period. An example of such a calibration is provided in Table 4.2. Interestingly, in this
example the reduction in BW gain after adjustments is quite large, because lower energy intakes
also means less maternal Pd and associated water gain. This method needs verification and if
successful, further development would be needed to include it into the precision feeding logic or
software creating a real-time feeding management system.
4.5. Summary
In this chapter, modifications to the NRC (2012) nutrient requirement model for gestating
sows are described, which allow its integration with precision feeding systems for individual
sows that are housed in groups and managed using ESF. Inputting a constant maternal Ld value
into the modified NRC (2012), now allows for the prediction of energy requirements and based
on specified diet energy densities, targeted feed intakes. Routines were added to generate a
database of a specific blend of two basal diets for each sow on each individual day of gestation;
this database can then be inputted into the sow management software, PigCHAMP. The modified
NRC (2012) model provides the basis for the study reported in the next chapter. The modified
model may be further expanded by including routines to automatically calibrate key animal
parameters to match targeted with observed performance, and to make adjustments to the feeding
program in real-time to correct for discrepancies between observed and predicted performance.
54
Table 4.1. Default values for target daily Ld (g/d) from average maternal Ld at typical energy intake levels
across parities and as outlined in NRC (2012)1.
Trait
Average total Ld, g/d
Average maternal Ld, g/d
1 (140)
108
105
Parity (body weight at breeding, kg)
2 (165)
100
96
3 (185)
84
80
4 (205)
32
28
5 (205)
45
41
In addition to the NRC (2012) recommendations, the following environmental parameters were also applied when
calculating default values to represent group housed conditions: Dietary fiber, 9%; Sow standing time, 240 min/d;
Housing, Group; Floor type, other; Effective environmental temperature, 20°C.
1
55
Table 4.2. Example adjustment for a parity 1 gestating sow (gilt; 140 kg initial BW, anticipated 12.5 pigs
born, 1.40 kg birthweight) to feed intakes for mid- and late gestation based on higher than anticipated BW
gain between d 1 and 28 of gestation.
Calculated
Average daily gain, g/d
Average daily feed allowance,
kg/d
Inputs
Maternal Ld, g/day
Adjustment to MEm , %
D 1-28
D 29-114
Targeted
(default)
(d 1-28)
Observed
(d 1-28)
Estimated
calibration
using model1
345
1.919
380
1.919
383
1.921
577 (640)2
2.037(2.297)2
138
-10
94.4[eq4-8]
-10
105
0
Adjusted
inputs
Values obtained by calibrating the model through adjusting Maternal Ld and ME Maintenance (MEm) to match
observed during d 1-28 of gestation.
2
Value in brackets represents default values over the same days.
1
56
Figure 4.1. Estimated energy intake (ME, kcal) and SID lysine (Lys, g) requirements by day of
gestation according to the original (Original) and modified (Mod) NRC (2012) models.
7000
25
6000
20
5000
15
g
3000
10
2000
5
1000
0
1
5
9
13
17
21
25
29
33
37
41
45
49
53
57
61
65
69
73
77
81
85
89
93
97
101
105
109
113
kcal
4000
0
Days of gestation
Mod Energy Intake
Original Energy intake
57
Mod Lys
Original Lys
Figure 4.2. Example of .csv excel file that is generated using the modified NRC (2012) gestating
sow model and imported into PigCHAMP sow management software.
58
5.0 EFFECTS OF PRECISION FEEDING PRIMIPAROUS SOWS ON BODY WEIGHT
CHANGES, LACATION PERFORMANCE, BEHAVIOUR AND LAMENESS AND
NUTRIENT BALANCES
5.1. Introduction
Commercial feeding practises commonly involve only one diet that is fed to all gestating
sows within a unit, leading to over feeding of nutrients during early gestation and, in the case of
primiparous sows, under feeding of nutrients during late gestation. Phase and parity-specific
feeding of gestating sows provides a means to more closely meet the dynamic changes in
nutrient requirements during gestation and across parity (Moehn et al., 2011). The effects of
gestation feeding regimes on sow reproductive performance remains unclear in the literature.
Campos et al. (2012) reported that effects of feeding extra feed or energy during late gestation
are not consistent between different studies. A factor that may contribute to inconsistencies
across studies is sow parity; in general, young sows that undergo considerable maternal body
weight (BW) gain during gestation appear more sensitive to alterations in feeding regimes (NRC,
2012). However, even if gestation feeding programs do not alter sow performance, more closely
meeting nutrient requirements at the various stages of gestation, may reduce feed costs as well as
nutrient losses into the environment (Moehn et al., 2011; NRC, 2012).
According to NRC (2012), the daily requirements for protein (i.e., essential amino acids;
lysine typically being the first limiting essential amino acid) increase more quickly than
requirements for energy as gestation progresses. Therefore, simply increasing the daily feed
allowance towards the end of gestation is not sufficient to meet the different dynamic changes in
energy and lysine requirements. It is possible to meet the individual needs of sows through the
use of an electronic sow feeder (ESF), which could allow for precision feeding management. A
novel research ESF, capable of blending feeds at the station (Chapter 3), allows for tailoring
59
diets to individual sows. Precision feeding may be able to better accommodate the requirements
of the gestating primiparous sow which would reduce negative energy balance before lactation
by providing nutrients when the needs for the conceptus are the greatest.
It is hypothesised that more closely meeting the nutrient requirements of sows throughout
gestation, will improve sow reproductive performance, welfare, and nutrient utilization
efficiencies. The objective of this work was to determine the effects of precision feeding
primiparous sows using the computer controlled ESF. This ESF will be used to blend two diets
to more closely meet dynamic changes in energy and lysine requirements according to NRC
(2012) for examining changes in sow BW and back fat thickness, feeding behaviour (number
visits to ESF), lameness, subsequent lactation performance, and nutrient balances during
gestation in comparison to sows on a one diet gestation feeding program.
5.2. Materials and Methods
5.2.1. Animals, housing, diets and experimental design
All work was conducted at the Arkell Swine Research Station (University of Guelph,
Guelph, ON, Canada). The experimental protocol was approved by the University of Guelph
Animal Care Committee (AUP #3237). Over a 6 month period, six blocks of primiparous sows
(n = 140; 125 F1 (Landrace x Yorkshire), 15 Yorkshire) were selected between d 2 and 9 post
breeding, weighed, and tagged with reusable radio frequency identification (RFID) transponder
(half-duplex signalling technology; Allflex technologies, Texas, USA). The sows were then
moved into group-housed pens after feeding with 20 to 30 sows per pen, and remained there until
day 101- 107 of gestation. Each pen was fitted with one ESF (Canarm Agsystems; Arthur, ON;
60
Chapter 3). Periodic testing of ESF feed delivery motors was performed and motors were
recalibrated when they exceeded 5% error (Chapter 3).
Within each pen, sows were randomly assigned to one of two dietary treatments, while
balancing for littermates, genetics and initial BW across treatments. The main feed ingredients in
the basal diets were corn, soybean meal and soy hulls (Table 5.1); soy hulls inclusion levels were
adjusted to achieve similar dietary fiber levels. The experimental gestation diets were
manufactured at the Arkell Feed Mill (University of Guelph, Guelph, ON, Canada); the lactation
diet was a commercial diet (Floradale, ON, Canada; Table 5.1). Each batch of feed was analyzed
and the nutrient composition for the average of all feed used presented in Table 5.2. For sows on
precision feeding (PF), the feeding level and blend of two basal iso-caloric diets (2518 Kcal/kg
NE; 0.80 vs. 0.20% SID Lys for high and low Lys, respectively; diets HP and LP, Table 5.2)
were adjusted daily for each animal to accurately meet estimated energy and Lys requirements.
Thus the total quantity and blend of HP and LP feed varied throughout gestation. A detailed
description of the calculations used to estimate daily and individual sow requirements based on
the NRC (2012) model can be found in Chapter 3. The remaining sows, managed as a control
(CON), received a constant amount of feed (2.20 kg/d) throughout gestation: a blend of 1.32 and
0.88 kg/d of HP and LP diets, respectively (mean SID Lys 0.56%). For CON, feed intake and the
dietary Lys level were established based on the average estimated requirements of primiparous
sows on PF over the entire gestation period. In this manner the targeted total BW gain
throughout gestation and the cumulative Lys intake were identical for the two treatments
(Chapter 3).
After the sows entered into the pen, the entrance gate to the ESF was kept open and fully
operational and all sows were freely allowed to explore the ESF. During the next three days (i.e.,
61
the training period) the ESF were set to be fully operational, and all sows were manually assisted
through the ESF to ensure they had consumed feed. The daily feeding cycle started at 0730 and
ended at 0700 the next day. After the training period, any sows that had not visited the feeding
station were manually moved through the ESF in the afternoon of the second day if they still had
not visited the station. Sows could consume all or parts of their meal in a singular visit to the
ESF; if a sow left the feeder before all feed was dropped, she could return within the same daily
feeding cycle and receive the remaining amount of feed. There was no carry over of missed
feedings between daily feeding cycles.
All sows received Farrowsure GOLD (Pfizer, Kirkland, Quebec), and a LitterGuard LTC (Zoetis, Kirkland, Quebec) vaccination during gestation, as per standard operating procedures
at the research unit. Around day 28 of gestation, pregnancy was verified using ultrasound; open
sows were removed from the pens. After the final BW measure between d 101 and d 107 of
gestation, sows were weighed, washed and moved to the maternity wing, where they were placed
in farrowing crates. All sows were then fed once daily approximately 2.5 kg of the common
lactation diet (Table 5.1) until farrowing. After farrowing, sow feed intakes were gradually
increased until ad libitum feed intake was achieved from d 5 post farrowing until weaning,
between d 18 and d 23 of lactation. Litters and sows were processed according to standard
procedures: teeth were clipped, males were castrated, and all cross fostering completed within 24
hours of birth. Starting on d 7 after birth, piglets were provided ad libitum access to a
commercial creep feed (Floradale Feedmill Ltd, Floradale, Ontario).
62
5.2.2. Experimental observations
Dietary nutrient analyses were conducted at Agri-food laboratories (Guelph, ON; Table
5.2). The BW of individual sows was determined once weekly using a floor scale (Atlas Scale,
Kitchener, Ontario). Weighing was done early in the morning before the beginning of the daily
feeding cycle to minimize variation in gut fill. Backfat (BF) thickness and loin depth were
measured using ultrasound every 4 weeks throughout gestation. During gestation, feed
consumption was recorded daily using the ESF, while it was recorded manually during lactation.
Sow and litter weights were determined within 24 h after birth prior to cross-fostering, and at
weaning; all deaths or removal of piglets, and their final weights, were recorded.
Sows were scored for skin scratches, lameness and mobility every 2 weeks during
gestation and when sows were manually moved to the weigh scales. In terms of mobility, sows
were scored as either sound (score 0) or lame (score 1). Sows were described as lame if they
favoured, had swelling or open wounds on their limbs, or displayed an unnatural gait.
The number of ESF feeding visits and feed delivery occurrences was recorded daily for
each sow. Unique feeder visits, defined as any visits that were greater than 5 minutes apart, were
recorded and used to distinguish between attempted, unsuccessful feedings (i.e. when no feed
was supplied), and actual feedings. Failure to enter the ESF, failure to achieve or maintain RFID
tag reads, or system malfunction resulting in missed or incomplete feedings were also recorded.
5.2.3. Calculations and statistical analysis
Weekly body weights were used to calculate sow average daily gain (ADG) and to
compare the PF group to predicted body weight values from the adapted NRC model. Whole
body mass balance calculations for nitrogen (N) and phosphorus (P) were conducted for
63
individual sows according to NRC (2012). In short, equations from NRC (2012) were used to
calculate N and P content in the empty maternal body, fetus and placental stores for individual
sows based on BW, BF, and day of gestation, as well as actual litter size at farrowing and mean
piglet birthweight (Eq. 8-49. 8-50, 8-51, 8-54, 8-56, 8-57, 8-58, 8-61, 8-67, 8-68, 8-69 in NRC,
2012), and then summed. These calculations were conducted for d 2 to 9 (beginning of gestation)
and for d 102 to 107 (end of gestation). Cumulative dietary N and P intakes were calculated for
individual sows from ESF-recorded feed intakes and analyzed dietary N and P contents, between
the beginning and end of gestation. Cumulative N and P losses into the environment were then
calculated as the difference between cumulative intake and cumulative retention.
Data were analyzed using the PROC MIXED function of SAS (version 9.4, SAS Institute
Inc., Cary, NC). In the model, dietary treatment, breed, and time were used as fixed effects, and
block (pen) and sow were considered random effects. Back fat thickness, BW and ADG values
were analysed as repeated measures over time using the Kenward and Roger adjustment to
estimate degrees of freedom. Initial back fat and BW were tested as covariates but were not
found to be significant (P >0.05) and were excluded from the final model. Data are presented as
the least square means for the two treatments ± standard error, and significance was accepted
when P < 0.05; P-values between 0.05 and 0.10 were accepted as trends.
5.3. Results
Feed was determined to be accurately prepared from nutrient analyses (Table 5.2). The
first two blocks of primiparous sows (n = 24 and n = 23) were removed from all observations
after weaning due to technical difficulties and inconsistencies with operations of the ESF. In the
remaining blocks, 7 sows (PF = 3, CON = 4) were removed after being confirmed not pregnant
64
and 6 sows (PF = 3, CON = 3) were removed due to severe lameness. Sows that did not finish
gestation were excluded from statistical analysis related to gestation growth performance
(number of sows used, Table 5.4; PF = 43, CON = 42). During farrowing and lactation there
were 5 mortalities (PF = 2, CON = 3) and 7 sows were culled due to farrowing or lactation
complications (PF = 4, CON = 3). Only sows that completed lactation were used to compare
litter and lactation performance (number of sows used, Table 5.5; PF = 37, CON = 36). For feed
intake and mass balance calculations for N and P as well as feeding costs (Table 5.6), sows
without complete gestation feed intakes due to technical errors or lost ear tags were also not used
(final number of sows, Table 5.3, Table 5.6; PF = 30, CON = 25).
In addition to the sows removed due to severe lameness, an additional 34 sows (PF = 15,
CON = 19) were deemed to be lame during gestation, but remained on trial. The average number
of days without feeding or incomplete feeding during the 28 days post-training was not affected
by treatment (PF = 3.5 days vs. CON = 3.5 days, P = 0.92). The ESF data containing number of
daily visits was lost before a common date due to technical difficulties, leaving different amounts
of data for each block. However, in the final month of gestation the number of daily visits was
not different even though sows received different amounts of feed (PF = 2.0 vs CON = 2.2
visits/day, P = 0.65).
Total feed consumption per sow between the beginning and end of gestation was similar
for PF and CON (203.7 vs. 202.3 kg/sow, respectively, P = 0.74; Table 5.3).
A total of 85 primiparous sows (PF = 43, CON = 42) that finished gestation were used for
statistical analysis of gestation performance. BW change throughout the entire gestation period
was not different between PF and CON (63.9 vs. 60.2 kg/sow, P = 0.17). However, PF sows
tended to gain less weight during early (d 4 to 33) gestation (PF = 10.0 vs. 12.4 kg/sow, P =
65
0.09), but gained more weight during late (d 68 to 102) gestation (29.2 vs. 23.9 kg/sow, P <
0.01). Figures 5.1 and 5.2 demonstrate the BW change throughout gestation was similar to the
model (NRC, 2012) predicted values for both PF and CON. A similar pattern was found for
ADG across the early, mid and late gestation periods. During early gestation, sows on PF tended
to grow at a slower rate in comparison to the CON sows (357 vs. 441 g/d, P = 0.09) but grew at a
faster rate during late gestation (842 vs. 675 g/d, P < 0.01, Table 5.4). Figures 5.3 and 5.4
demonstrate that weekly ADG are much more erratic than model predicted curves for both PF
and CON. Changes in back fat depth during gestation were not different between the two
treatments (3.3 vs. 2.7 mm, P = 0.31).
A total of 73 primiparous sows (PF = 37, CON = 36) were used in statistical analysis
based on finishing lactation with piglets weaned (Table 5.5). Although sow ADG (PF = -789 vs.
CON = -929 g/d) and voluntary ADFI (PF = 5.1 vs. CON = 4.7 kg) during lactation were
numerically greater for PF sows, there were no significant differences in sow growth and feed
intake throughout lactation (Table 5.5). There were no significant differences in the number of
piglets born alive (PF = 12.2 vs. CON = 12.4), piglet birth weight (PF = 1.45 vs. CON = 1.50
kg/piglet), and litter growth rate (PF = 2.49 vs. CON = 2.59 kg/litter/d; Table 5.5).
Mass balance calculations revealed the PF sows retained 2.5% more N than CON,
although this was not significantly different (PF = 26.3 vs. CON = 23.7%, P = 0.16, Table 5.6).
This resulted in numerically less N losses into the environment (PF = 1337 vs. CON = 1389 g, P
= 0.16). A similar pattern was observed for P (Table 5.6). Given the costs of the main feed
ingredients, feeding costs per sow did not differ between treatments (P = 0.91, Table 5.6). The
feed conversion was not significantly different between treatments (P = 0.57, Table 5.6).
66
5.4. Discussion
In this study computer controlled ESF were used to PF sows according to dynamic
changes in energy and lysine requirements throughout gestation according to a modified NRC
(2012) model (Chapter 4). This was done to assess the effects of more closely meeting the sow’s
nutrient requirements on changes in sow BW, feeding behavior (visits to ESF), lameness,
subsequent lactation performance, and nutrient balances during gestation, in comparison to a one
diet gestation feeding program. Current commercial gestation sow feeding strategies are
generally based on using a single gestation diet for all sows regardless of parity or stage of
gestation, and often do not consider the needs of a sow as an individual. In the literature, the
benefits of changing feeding strategies as a means to improve gestating sow performance
remains unclear (Chapter 2). However, the NRC (2012) gestating sow nutrient requirement
model demonstrates large differences in nutrient requirements across parities, and dynamic
changes in nutrient requirements throughout gestation. To our knowledge, blended feeding of
different gestation diets to closely meet nutrient requirements of gestating sows according to the
NRC (2012) model has not been reported in the literature.
Overall, the results from this experiment demonstrate that more closely meeting the
predicted protein (Lys) and energy demands of gestating primiparous sows in comparison to a
one phase gestation diet program provided no benefit to primiparous sow production or wellbeing. However, the pattern of sow BW gain more closely reflected the growth of fetus and
tissues involved in reproduction (e.g. uterus and uterine fluids, udder) based on the NRC 2012
model and suggests less negative energy balance before farrowing, which could have long term
effects on the animal’s reproductive life.
As previously mentioned, this is a novel feeding approach and as such, required the use
67
of prototype technology, which is partly responsible for the loss of observations. As described in
Chapter 3, the ESF went through major changes to reach the current functionality. The first two
blocks of sows completed lactation but the data were discarded because the accuracy of feeding
during gestation could not be presented with confidence. Furthermore, some experimental feed
usage records were lost due to inaccurate records as a result of software modification
complications. For example, the software did not record the cumulative feed intake of sows that
lost ear tags and were retagged, even though they were maintained on the same feeding program
There were no differences in the measures of sow lameness or feeding behaviour (visits
to ESF). The ESF went through many modifications to reduced sow injury, influencing the
incidence of lameness between the beginning and end of the project (Chapter 3). As handling of
sows for weighing may also have influenced body lesion scoring, these data were discarded. The
number of daily feeding visits was examined only for the last month for two reasons. The first
reason was an error with data recording software; computer observations for the full gestation
period existed only for 1 block of sows while data for the last month of gestation was available
for 4 blocks of animals. The second reason is due to the hypothesis that in the last month of
gestation, sows receiving less feed in a period of high nutrient requirements may seek out feed
more often and therefore have more visits to ESF. This was not the case and there was no
treatment effect between CON and PF sows.
Overall, BW and BF changes throughout gestation relative to feed intake patterns for the
two treatments were as anticipated; sows receiving greater nutrients at specific times grew at a
greater rate. Certain discrepancies in BW occurred, between the predicted NRC (2012) model
and observed animal performance, which is likely due to biological variation between animals.
In the future, this could be accounted for by real-time feeding strategy changes as described in
68
Chapter 4, where the feeding program changes for sows growing differently than model
prediction. Generally, the observed changes in BW, ADG and BF thickness validates and
provides reasonable confidence in the NRC predictions for these measures.
Compared to the most recently released PigCHAMP Benchmarking data (2014)
summaries for Canada, piglets born alive and weaned per litter were slightly below industry
average, but reasonable compared to Arkell Research Station herd averages. Litter size, growth
rate and weaning rates were not affected by feeding regimen. Further work should monitor
closely sow daily feed intake and body weight changes during lactation as a numerical increase
in feed intake and reduction in BW loss was noted in PF sows, although not significantly
different. This is the opposite to the response reported by Shelton et al. (2003), where lactation
feed intake in gilts decreased when feed allowance was increased after d 90 of gestation. The
usefulness and effect of bump feeding on sow performance and lactation feed intake are
discussed in detail in Chapter 1.
Although not a specific objective of the current study, an after experiment review of herd
breeding records revealed average weaning to breeding interval of 11.5 and 14.6 days (P = 0.50)
for PF and CON, respectively (number of sows rebred PF = 23, CON = 26). However, the Arkell
research herd is managed in a batch farrowing system where breeding only occurs once per
month, which skews non-productive day measures. In future studies, this is a production aspect
that deserves greater attention, perhaps using a percent bred by 7 days post weaning measure.
The N retention efficiency calculated in this study (23.7%) was similar to the 23%
reported by Dourmad et al. (1999), although separate gestation and lactation values were not
reported. As well, the observed increase to 26.3% N retention efficiency in PF sows fits well
with described patterns for multiple-phase feeding regimens (Monterio et al., 2010). Although no
69
significant differences were reported between PF and CON for N retention efficiency or feed
costs, no differences were expected as the CON feeding level was intentionally set as the average
of the PF treatment, resulting in similar total feed and N intakes. However, it is important to note
that the current study only examined feeding primiparous sows, balancing the diets only for
energy and protein intakes. Using the NRC (2012) model as a predictive tool, it is estimated that
larger cost and environmental savings would exist in higher parity sows, especially when
compared to commercial feeding strategies.
5.5. Conclusions and implications
The results of this study demonstrated that the PF technology was functional. Although
this study indicates no difference in primiparous sow productive performance, there were
differences in patterns of BW gain in PF sows during gestation that mirror more closely the
progression of pregnancy, which could have an effect on long term sow performance. Future
trials should be conducted with a larger number of animals and a multi-parity approach to
determine the effects of PF on long-term sow performance and cost benefits, especially in higher
parity sows.
70
Table 5.1. Ingredient composition (%) of high and low protein gestation diets (HP and LP respectively),
as well as the lactation diet1.
Ingredient
Dry Corn
Soybean Meal
Soy Hulls
Chopped Barley
Chopped Wheat
Wheat Shorts
A/V Fat
Mono-Calcium phosphate
Limestone
Salt
Arkell Micro Premix
Commercial Micro Premix
L-Lysine (50%)
Pelltech
Threonine
Methionine (88%)
Integral AL
Rononzyme GT hiPhos 2500
Gestation Diets
HP
LP
68.66
25.71
85.84
0.89
7.50
2.03
1.15
1.58
0.37
0.50
1.82
1.54
1.51
0.40
0.50
Lactation1
48.46
21.60
10.00
2.50
10.00
3.20
0.75
1.65
0.49
0.30
0.43
0.40
0.10
0.05
0.05
0.03
The experimental gestation high protein (HP) and low protein (LP) diets were mixed and pelleted at the Arkell Feed
Mill (University of Guelph, Guelph, ON, Canada). The lactation diet was formulated, mixed and pelleted at a
commercial facility, Floradale Feed Mill, Ltd. (Floradale, ON, Canada).
1
71
Table 5.2. Calculated and analyzed nutrient contents (as is basis) in high and low protein experimental
gestation diets (HP and LP respectively), as well as the common lactation diet. 1
Gestation Diets
Component
Net Energy (kcal/kg)
Dry Matter, %
Crude Protein, %
SID Lysine, %
Calcium, %
Phosphorus, %
Sodium, %
HP
Calculated1 Analyzed
2518.00
89.00
87.89
17.93
18.53
0.80
0.84
0.89
0.61
0.62
0.18
0.16
LP
Calculated2 Analyzed
2518.00
88.80
88.19
8.27
8.15
0.20
0.84
0.89
0.56
0.59
0.18
0.18
Lactation
Calculated3 Analyzed
2520.06
87.78
87.76
16.74
17.60
0.744
0.86
0.89
0.55
0.59
0.22
0.20
All components expressed on a % basis except for Net Energy.
Calculated values represent contributions from each feed ingredient according to NRC (2012).
3
Calculated values presented as provided from commercial feed mill supplier, Floradale Feed Mill, Ltd. (Floradale,
ON, Canada).
4
Not provided, calculated in same manner as 2.
1
2
72
Table 5.3. Amount of high protein (HP) and low protein (LP) feed consumed during entire gestation
period by primiparous sows on precision feeding (PF) and control (CON) treatments.
Number of sows1
HP, kg
LP, kg
Total, kg
PF
30
120.0 ±1.62
83.8 ±1.7
203.7 ±2.8
CON
25
122.6 ±1.6
79.7 ±1.9
202.3 ±3.1
P value
0.27
0.11
0.74
Number of sows (observations) per treatment (n). Actual n is lower than expected because data from one group
was unavailable due to an error with the recording software. In addition there are missing data from sows where ear
tags were lost and replaced and cumulative feed intake data were not recorded. Data only included sows that
farrowed.
2
Standard error.
1
73
Table 5.4. Changes in mean BW, ADG and BF for primiparous sows feed on precision feeding (PF) and
control (CON) treatments throughout gestation. 1
Parameter
Number of sows
Initial BW d 4 ±3, kg
Early BW d 33 ±3, kg
Mid BW d 68 ±3, kg
Late BW d 102 ±3, kg
Early ADG, g/day
Mid ADG, g/day
Late ADG, g/day
Total ADG, g/day
Initial BF, mm
Final BF, mm
Total BF change, mm
1
2
PF
43
143.2 ±2.12
152.4 ±2.2
177.7 ±2.6
207.1 ±2.8
357 ±34
700 ±34
842 ±28
654 ±18
14.0±0.3
17.1±0.4
3.3 ±0.5
CON
42
145.4 ±2.2
156.8 ±2.2
181.6 ±2.7
205.3 ±2.9
441 ±35
709 ±35
675 ±29
619 ±18
13.1±0.3
16.4±0.4
2.7 ±0.5
Only data from sows that finished entire gestation are included in analysis.
Standard error.
74
P value
0.40
0.15
0.30
0.66
0.09
0.86
<0.01
0.17
0.06
0.28
0.31
Table 5.5. Litter and lactation performance for primiparous sows managed on precision feeding (PF) and
control (CON) treatments throughout gestation. 1
Parameter
Number of sows
Sow BW at farrowing, kg
Sow BW change, kg
Sow ADG kg/d
Sow voluntary ADFI, kg/d
# Piglets born alive
Piglet birth weight, kg
# Piglets weaned
Piglet weaning weight, kg
Piglet ADG3, g/d
Litter growth rate3, kg/d
PF
37
193.7 ±3.12
-15.7 ±2.8
-0.789 ±0.139
5.05 ±0.24
12.2 ±0.5
1.5 ±0.03
10.4 ±0.5
6.3 ±0.1
242 ±7
2.49 ±0.14
CON
36
192.6 ±3.2
-18.0 ±2.9
-0.929 ±0.141
4.71 ±0.24
12.4 ±0.5
1.5 ±0.03
11.0 ±0.5
6.3 ±0.2
237 ±7
2.59 ±1.4
P value
0.79
0.34
0.25
0.14
0.86
0.34
0.15
0.83
0.55
0.35
Only data from sows that finished lactation are included in analysis.
Standard error.
3
Actual n for these measurements is 36 for PF group and 31 for CON due to recording errors during data collection.
1
2
75
Table 5.6. Mass Balance calculation and feed cost for primiparous sows managed on precision feeding (PF)
and control (CON) throughout gestation.1
Parameter
Number of sows
Nitrogen retained, % of
intake
Excess Nitrogen, g/sow
Phosphorus retained, % of
intake
Excess Phosphorus, g/sow
Feed costs, $/sow3
Feed conversion ratio
PF
30
26.3 ±1.22
CON
25
23.7 ±1.3
P value
1337 ±25
18.8 ±1.1
1389 ±27
16.6 ±1.2
0.16
0.18
1238 ±17
49.85 ±0.67
3.2 ±0.1
1231 ±19
49.73 ±0.74
3.3 ±0.1
0.77
0.91
0.57
0.16
n is lower than expected because data from one group became unavailable due to an error with the recording
software. In addition there are missing data from sows where ear tags were lost and replaced and cumulative feed
intake data were not recorded.
2
Standard error.
3
Ingredient costs ($/tonne): Soybean meal, 433; Corn, 211;, Soy Hulls, 105; Mono-Calcium, 90; Limestone, 3.7;,
Salt, 5; A/V fat, 650;, and Premix, 130.
PF, sows managed on precision fed treatment, and CON, sow managed on control treatment.
1
76
Figure 5.1. Average sow BW for precision fed (PF) primiparous sows compared to average
predicted BW using the modified NRC (2012) model (mod NRC 2012).
210
200
190
Kg
180
170
160
150
140
130
120
5
12
19
26
33
40
47
54
61
Day of Gestation
PF
68
75
82
89
96 103
mod NRC 2012
Figure 5.2. Average sow BW for control fed (CON) primiparous sows compared to average
predicted BW using the NRC (2012) model (NRC 2012).
210
200
190
Kg
180
170
160
150
140
130
120
5
12
19
26
33
40
47
54
61
Day of Gestation
CON
68
NRC 2012
77
75
82
89
96 103
Figure 5.3. Average sow ADG for precision fed (PF) primiparous sows compared to average
predicted ADG using the modified NRC (2012) model (mod NRC 2012).
1200
1000
800
g/day
600
400
200
0
-200
1
2
3
4
5
6
7
8
9
10
11
12
13
14
-400
-600
Week
PF
mod NRC 2012
Figure 5.4. Average sow ADG for control fed (CON) primiparous sows compared to average
predicted ADG using the NRC (2012) model (NRC 2012).
1200
1000
800
g/day
600
400
200
0
-200
1
2
3
4
5
6
7
8
9
-400
-600
Week
CON
NRC 2012
78
10
11
12
13
14
6.0. GENERAL DISCUSSION, IMPLICATIONS AND FUTURE CONSIDERATIONS
The Canadian pig industry is committed to the transition to group housing systems that
will offer more freedom of movement for gestating sows. This commitment was made clear in
the Canadian Code of Practice for the Care and Handling of Pigs (2014). Due to the transition to
group housing for gestating sows, producers will be forced to re-evaluate how they manage sows
as it is more difficult to ensure each individual sow within a group receives adequate care. The
concept of individual feeding of sows within a group is not new and was first proposed in a 1969
US technology patent (3,473,515. Atchinson, P.) to provide ‘means of feeding each of a
succession of sows’. Electronic Sow Feeders (ESF) are gaining in popularity and paired with
modern technology create the opportunity to further improve how individual sows are fed and
managed within groups. Current gestation feeding strategies are more of a “one size fits all”
approach with a single gestation diet used for all sows regardless of parity from breeding to
farrowing. The objectives of the study reported in this thesis were to (1) further develop a
prototype ESF and feeding program to take advantage of promising new precision-feeding (PF)
technology and improved design in order to maximize sow welfare and minimize technical
failures, and (2) explore the effects of more closely meeting the predicted protein (lysine) and
energy demands of gestating primiparous sows, by blending two feeds according to the NRC
(2012) model, in comparison to sows managed on a one-phase gestation diet program, on sow
reproductive performance, welfare and nutrient utilization efficiencies.
An automatic and intelligent precision feeder that blends feed and distributes two
premixes to growing pigs to optimize production has been described by Pomar et al. (2009).
However, applying blending technology for PF gestating sows through ESF is a novel concept
(Chapter 3). Due to their size and strength, sows offer a larger challenge than feeding growing
79
pigs when designing equipment. As outlined in this thesis, structural modifications and feeding
logic development were required to increase ESF robustness and decrease technical failures.
Implementing stepper motors for accurate feed delivery and blending within the ESF were
essential for delivering balanced diets that minimize excess nutrient supply as part of the PF
approach. The feed delivery system required fine tuning and development of accurate calibration
techniques. This technology is now capable of blending feed in very small amounts (min 14 g)
under research settings. Current attempts to design a scale for measuring BW within the ESF
were unsuccessful. If the ultimate goal remains to include a weigh scale within the ESF,
researchers should consider modifications of forelimb and hindlimb scales used for evaluating
grower pig body weights (e.g.. Ramaekers et al., 1995) along with modern technology. An
accurate, ground-based hindquarter scale could eliminate problems with moving parts, and sows
stepping off the front of the scale to reach the feed bowl. Furthermore, designing adjustability or
separate units for gilts is a necessity with ESF due to the large difference in size between gilts
and older sows. In general, technology for PF using ESF was developed and proved functional;
however further work is needed to make the unit commercially feasible.
For this work, a revised NRC (2012) gestating sow model was adapted and used as the
basis for a feeding program for the PF treatment (Chapter 4). Suggestions were outlined for
possible logic that could better fit the feeding program to individual animals based on changes in
BW growth when compared to predicted BW. In theory, this will also account for differences
between genotypes, farms and physiological differences between animals. However, the NRC
(2012) model is not perfect for predicting BW gain. For example, it predicts protein and lipid
gains, and does not account for changes in body water mass associated with growing maternal
organs and the patterns of maternal body protein gain need to be better defined (Miller et al.,
80
2015). On the other hand, possible discrepancies in actual compared to predicted BW, could
originate from more managerial-based roots. The initial BW used to calculate feed requirements
and predict BW could be affected by gut fill, which in this experiment was accounted for by
standardization of eating and weighing times. Secondly, initial BW are often taken when mixing
sows together to allow for synergy of barn tasks which takes place between d 2 and d 9 after
breeding; this may result in up to a 6 kg difference from the BW at breeding, which is the
parameter the current NRC (2012) accepts. This difference may become even larger if sows are
mixedafter confirming pregnancy, which causes the model to feed a ‘heavier gilt’ than what
actually exists. Before commercialization, further adjustment is required to the NRC (2012)
model to allow input of a sow BW on days other than day1 of gestation for more practical farm
management. Lastly, post-mixing BW and ADG decreased in the present study for both
treatments as compared to the predicted curve (NRC, 2012). This may be attributed to sows
learning to eat from an ESF, and is not yet accounted for in NRC (2012) model. It is hoped this
decline in BW and ADG will be reduced in higher parity animals or through better training.
Although nutritional models are at times limited, it remains unclear which growth pattern
differences are due to social/immune stress or the ability to meet nutritional requirements. In
summary, progress was made in this research adapting the NRC (2012) model to generate
practical feeding programs.
The hypothesis that accommodating the unique and dynamic changes of energy and
nutrient requirements of individual sows throughout gestation would improve sow welfare and
productivity was not supported by the results of this study. There were no significant differences
between PF and CON treatments in sow feeding behaviour (i.e. feeder visit frequency) or
lameness, total changes in BW and back fat thickness of individual sows, number of total piglets
81
born and piglets born alive per litter, individual piglet weight at birth and weaning, sow feed
intake or body weight change during lactation. However, the results of this study indicate that PF
managed sows had different patterns of ADG throughout gestation, when compared to the
control feeding program (Chapter 5). Precision feeding managed sows exhibited greater ADG in
late gestation which could have potential beneficial effects on sow longevity, a topic for which
limited scientific studies exist. Interestingly the voluntarily ADFI during lactation was
numerically greater along with decreased BW loss during lactation for PF managed sows. These
are very important sow production parameters and should be monitored closely in future
precision feeding studies.
Increasing nitrogen utilization efficiency by delivering an adequate supply of protein
according to the sow’s physiological state, is an approach to decrease excretion and
environmental impact (Monterio et al., 2010). However, no differences were anticipated in cost
savings or environmental impact between PF and CON managed sows as the CON feeding levels
were determined as the anticipated average for PF. Thus, overall both management groups of
sows received similar total intakes of feed and protein. As predicted there were no significant
differences in nitrogen or phosphorus retention or feeding costs between treatments in this
experiment.
In conclusion, this thesis documents the development of a precision ESF capable of
blending feed and an accompanying feeding program incorporating a revised NRC (2012) sow
nutrient requirement model. Using these technologies with primiparous sows, it was
demonstrated that precision feeding gestating sows is possible and by doing so, patterns of
primiparous sow BW gain can be altered to better reflect increasing demands during late
pregnancy. A larger scale study across multiple parities with a larger number of animals is
82
recommended to evaluate predictions of greater economic and environmental effects in older
sows as well as determine the effects on long-term sow performance. Overall, this research was a
positive contribution to the possibilities for precision livestock farming and eventual fully
automated monitoring and management systems for livestock.
83
Appendix 1. Colour codes of light indicator for ESF status. 1
1
Created in partnership with T.Hardman of Roberts Onsite, personal communications (2014)
Green
White
Yellow
Light Blue
Light Blue Flashing
Dark Blue
Purple
Purple Flashing
Red Flashing
Feeding
Open- Waiting for Sow
Feeder Disabled
Looking for Tag read
Lost read searching for tag
Tag read successful
Delay before bowl retraction
Whiling time to clean up
Feeder Trouble/ Empty Hopper.
84
Appendix 2. New calibration tool computer interface developed for ease of accurate individual
feed deliver motor calibration.1
1
Created in partnership with S. Keys of B&R Automation, personal communications (2015)
85
Appendix 3. Using Matrix multiplication, to develop equations that will blend 3 separate isocaloric feeds to achieve calculated Energy, Protein(Lys) and Phosphorus requirements. 1
1
Developed in partnership with R.Rath BSc., personal communications (2015)
Aa1a2+Bb1b2+Cc1c2=TLP
Where:
A= Amount of feed A, a1=Lys level of feed A, a2=P level of Feed A
B= Amount of feed B, b1=Lys level of feed B, b2=P level of Feed B
C= Amount of feed C, c1=Lys level of feed C, c2=P level of Feed C
T=Total amount of feed, L= Lys requirements, P= P requirements
A+B+C=T
Aa1+Bb1+Cc1=L
Aa2+Bb2+Cc2=P
a1R1-R2
a2R1-R3
1
0
0
R2/(a1-b1)
0 1
0 0
(
)(
)
1 0 −
0 1
0 0
−
−(
−
(
−
−
−
−
−
−
1
)
)(
−
+1 −
−
−
1
−
−
−
−
−
+1
−
−
−
− )
−( −
−
−
−
−
1
1
−
−
1
1
0
1 0
1
1
−
−
1
0
-R2+R1
(a2-b2)R2-R3
R3/
1
R1
R2
R3
+
−
−
−
−
)(
When
+ 1 R3 + R1
86
−
(
=
)(
−
(
−
+
−
−
−
− )
−(
)
− )
− )(
− )
−(
( − )
( − )( − )
−
−
− )
+
−
R3 + R2
1
0
0
Therefore:
=
−
−
=−
−
−
0 0
1 0
0 1
−
−1 ( )−
−
−
(
( )+
−
−
−
−
− )(
− )
−(
− )
( − )
−1
( − )( − )
− +
−
( − )(
− )
−(
− )
−
( − )
( − )( − )
−
− +
−
( − )(
− )
−(
− )
( − )
=
( − )( − )
− +
−
87
+
−
+
−
−
−
−
+
LITERATURE CITED
Australia Pork, Pork CRC, and Australian Pork Limited. 2013. Feeding pregnant sows in group
housing systems- An update. www.porkcrc.com.au (accessed Sept 1, 2014)
Atchinson, P. C. Means for feeding each of a succession of sows. US Patent 3,473,515. Oct 21,
1969.
Bench, C. J., Rioja- Lang, F. C., Hayne, S. M., and Gonyou, H. W. 2013a. Group gestation
housing with individual feeding -I: How feeding regime, resource allocation, and genetic factors
affect sow welfare. Livest Sci. 152: 208-217.
Bench, C. J., Rioja- Lang, F. C., Hayne, S. M., and Gonyou, H. W. 2013b. Group gestation sow
housing with individual feeding – II: How space allowance, group size and composition, and
flooring affect sow welfare. Livest Sci. 152: 218-227.
Bressers, H. P. M., Te Brake, J. H. A., Engel, B., and Noordhuizen, J. P. T.M. 1992. Feeding
order of sows at an individual electronic feed station in a dynamic group-housing system. Appl
Anim Behav Sci. 36: 123-134.
Brooks, P. H. 2003. Group housing of sows- the European Experience. London Swine
Conference proceedings: 37-60.
Campos, P. H. R. F., Silva, B. A. N., Donzle, J. L., Oliveria, R. F. M., and Knol, E. F. 2012.
Effects of sow nutrition during gestation on within-litter birth weight variation: a review.
Animal, 6 : 797-806.
Canada 2014-year end summary. PigCHAMP- Benchmarking. www.pigchamp.com. (accessed
May 1, 2016)
Cerisuelo, A., Sala, R., Gasa, J., Chapinal, N., Carrion, D., Coma, J., and Baucells, M. D. 2008.
Effects of extra feeding during mid-pregnancy on gilts productive and reproductive performance.
Span J Agric Res. 6: 219-229.
Cassar, G., Kirkwood, R. N., Seguin, M. J., Widowski, T. M., Farzan, A., Zanella, A. J.,and
Friendship, R. M. 2007. Influence of stage of gestation at grouping and presence of boars on
farrowing rate and litter size of group-housed sows. J Swin Health Prod. 16: 81-85.
Clowes, E. J., Kirkwood, R., Cegielski, A., and Aherne, F. X. 2002. Phase-feeding protein to
gestating sows over three parities reduced nitrogen excretion without affecting sow performance.
Livest Prod Sci. 81: 235-246.
Cornou, C., Vinther, J.,and Kristensen, A. R. 2008. Automatic detection of oestrus and health
disorders using data from electronic sow feeders. Livest Sci. 118: 262-271.
88
Cromwell, G. L., Hall, D. D., Clawson, A. J., Combs, G. E., Knabe, D. A., Maxwell, C. V.,
Noland, P. R., Orr, D. E. Jr., and Prince, T. J. 1989. Effects of Additional Feed During Late
Gestation on Reproductive Performance of Sows: A Cooperative Study. J Anim Sci. 67: 3-14.
Dourmad, J. Y., Etienne, M., Valancogne, A., Dubois, S., van Milgen, J,. and Noblet, J.2008.
InraPorc: A model and decision support tool for the nutrition of sows. Anim Feed Sci Tech. 143:
372-386.
Dourmad, J. Y., Seve, B., Latimier, P., Boisen, S., Fernandez, J., van der Peet-Schwering, C.,
and Jongbloed, A. W. 1999. Nitrogen consumption, utilisation and losses in pig production in
France, The Netherlands and Denmark. Livest Prod Sci, 58 : 261-264.
Dwyer, C. M., Strickland, N. C., and Fletcher, J. M. 1994. The Influence of maternal nutrition on
muscle fiber number development in the porcine fetus and on subsequent postnatal growth. J
Anim Sci. 72: 911-917.
Farmer, C. 2015. The Gestating and Lactating Sow. Wageningen Academic
Publishers,Wageningen, The Netherlands.
Goodband, R. D., Tokach, M. D., Goncalves, M. A.D., Woodworth, J. C., Dritz, S. S., and
DeRouchey, J. M. 2013. Nutritional enhancements during pregnancy and its effects on
reproduction in swine. Animal Frontiers. 3 (4): 68-75.
Kirkwood, R., Zanella, A., 2005. Influence of gestation housing on sow welfare and
productivity. National Pork Board Final Report.
http://old.pork.org/filelibrary/researchdocuments/03-055-kirkwood-abs.pdf (accessed Jan 10,
2015).
Knox, R., Salak-Johnson, J., Hopgood, M., Greiner, L., and Connor, J. 2014. Effect of day of
mixing gestating sows on measures of reproductive performance and animal welfare. J Anim Sci.
92: 1698-1707.
Laws, J., Laws, A., Lean, I. J., Dodds, P. F., and Clarke, L. 2007. Growth and development of
offspring following supplementation of sow diets with oil during mid to late gestation. Animal. 1
: 1490-1496.
Lawlor, P. G., Lynch, P, B., O’Connell, M. K., Namara, L. M., Reid, P., and Strickland, N. C.
2007. The influence of over feeding sows during gestation on reproductive performance and pig
growth to slaughter. Arch. Tierz, Dummerstorf. 50: 82-91.
Li,X., Bazer, F. W., Johnson, G. A., Burghardt, R. C., Erikson, D. W., Frank, J. W., Spencer, T.
E., Shinzato, I and Wu, G. 2010. Dietary Supplementation with 0.8% L-Arginine between Days
0 and 25 of Gestation Reduces Litter Size in Gilts. J Nutr. 140: 1111-1116.
Lovendahl, P., Damgaard, L. H., Nielsen, B. L., Thodberg, K., Su, G., and Rydhmer, L. 2005.
Aggressive behaviour of sows at mixing and maternal behaviour are heritable and genetically
89
correlated traits. Livest Prod Sci. 93: 73-85.
Mateo, R, D., Wu, F., Bazer, F. W., Park, J, C., Shinzato, I., and Kim, S, W. 2007. Dietary LArginine supplementation enhances the reproductive performance of gilts. J Nutr. 137: 652-656.
McPherson, R. L., Ji, F., Wu, G., Blanton, R, Jr.,and Kim, S. W. 2004. Growth and
compositional changes of fetal tissues in pigs. J Anim Sci. 82: 2534-2540.
Miller, E.G., Wey, D., Levesque, C. L., and de Lange, C. F. 2015. Whole body and maternal
protein deposition in gestating gilts at two feeding levels. Mid-West American Society of Animal
Science Meeting. Des Moines, Iowa, March 2015. Journal of Animal Science 93 (supplement 2;
abstracts): 172-173. https://asas.confex.com/asas/mw15/webprogram/Paper11571.html
Moehn. S., Franco, D., Levesque, C, Samuel,R., and Ball. R. O. 2011. New energy and amino
acid requirements for gestating sows. Advances in Pork Production. 22: 1-10.
Monterio, D. O., Pinheiro, V. M. C., Mourao, J. L. M., and Rodrigues, M. A. M. 2010. Strategies
for mitigation of nitrogen environmental impact from swine production. R. Bras. Zootec, 39:
317-325.
National Farm Animal Care Council and Canadian Pork Council. 2014. Code of Practice for the
care and handling of pigs. https://www.nfacc.ca/pdfs/codes/pig_code_of_practice.pdf
NRC. 2012. Nutrient requirements of swine. 11 th edition. National Academic Press,
Washington, DC.
Olsson, A. C., Andersson, M., Botermans, J., Rantzer, D., and Svendsen, J. 2011. Animal
interatction and response to electronic sow feeding (ESF) in 3 different herds and effects of
function settings to increase capacity. Livest Sci. 137: 268-272.
Pomar, C., Hauschild L., Shang, G., Pomar, J., and Lovatto, P. A. 2009. Apply precision feeding
techniques in growing-finishing pig operations. R. Bras. Zootec. 38: 226-237.
Pond, W. G., Yen, J. T., Maurer, R. R., and Christenson, R. K. 1981. Effect of doubling daily
energy intake during the last two weeks of pregnancy on pig birth weight, survival and weaning
weight. J Anim Sci. 52: 535-541.
Ramaekers, P.J.L., Huiskes, J. H., Verstegen, M. W. A., den Hartog, L. A., Vesseur, P.C., and
Swinkels, J. W. G. M. 1995. Estimating individual body weights of group-housed growingfinishing pigs using a forelegs weighing system. Comput Electron Agric, 13: 1-12.
Rhodes, R. T., Appleby, M. C., Chinn, K., Douglas, L., Firkins, L. D., Houpt, K. A., Irwin, C.,
McGlone, J. J., Sundberg, P., Tokach, L., and Wills, R. W. 2005. A comprehensive review of
housing for pregnant sows. JAVMA, 227 (10): 1580-1590.
Shelton, N. W., DeRouchey, J. M., Neill, C. R., Tokach, M. D., Dritz, S. S., Goodband, R. D.,
90
and Nelssen, J. L. 2003. Effects of increasing feeding level during late gestation on sow and litter
performance. Sow Herd Nutrition and Management. 38-50.
Spoolder, H. A. M.,Geudeke, M. J., Van der Peet-Schwering, C. M. C. and Soede. N. M.
2009.Group housing of sows in early pregnancy: A review of success and risk factors. Livest Sci.
125:1-14.
Strawford, M. L., Li, Y. Z., Gonyou, H. W. 2008. The effect of management strategies and parity
on the behaviour and physiology of gestating sows housed in an electronic sow feeding system.
Can J Anim Sci. 88: 559-567.
Strawford, M. L. 2006. Social Factors That Affect the Behaviour and Productivity of Gestating
Sows in an Electronic Sow Feeding System. University of Saskatchewan. Saskatoon. MSc
Thesis.
Terlouw, E. M. C., Lawrence, A. B., and Illius, A W. 1991. Influences of feeding level and
physical restriction on development of sterotypies in sows. Anim Behav. 42: 981-991.
Van der Peet-Schwering, C., Hoofs, A., Soede, N., Spoolder, H., and Vereijken, P. 2009. Group
Housing of sows during early gestation. Wageningen UR Livestock Research. Research report.
Whittington, L., and Schneider, K. 2004. Considerations for large group housing of finishing
pigs. Advances in Pork Production, 15:109-118.
Wu, G, Bazer, F. W., Davis, T, A., Jaeger, L, A., Johnson, G, Kim, S. W., Knabe, D. A.,
Meininger, C.J. Spencer, T. E., and Yin, Y. 2007. Important roles for the arginine family of
amino acids in swine nutrition and production. Livest Sci. 112: 8-22.
Wu, G., Bazer, F. W., Burghardt, R. C., Johnson, F. A., Kim, S. W., Li, X. L., Satterfield, M. C.,
and Spencer, T. E. 2010. Impacts of amino acid nutrition on pregnancy outcome in pigs:
Mechanisms and implications for swine production. J Anim Science. 88: E195-E204.
Yang, Y., Heo, S., Jin, Z., Yun, J., Shinde, P, Choi, J, Yang, B., and Chae, B. 2008.Effects of
dietary energy and lysine intake during late gestation and lacation on blood metabolites,
hormones, milk compostion and reproductive performance in multiparous sows. Arch Anim
Nutr, 62 :10-21.
Young, L. G., King, G. J., Shaw, J., Quintin, M., Walton, J. S., and McMillian, I. 1991.
Interrelationships among age, body weight, backfat and lactation feed intake with reproductive
performance and longevity of sows. Can. J. Anim. Sci. 71: 567-575.
Zhang, R. F., Hu, Q., Li, P. F., Xue, F., Piao, X. S., and Li, D. F. 2011. Effects of Lysine Intake
during Middle to Late Gestation (Day 30-110) on Reproductive performance, colostrum
composition, blood metabolites and hormones of multiparous sows. Asian Austral J Anim. 24 :
1142-1147.
91