Prediction of litter performance in lactating sows using machine learning, for precision livestock farming
In: Computers and electronics in agriculture: COMPAG online ; an international journal, Band 196, S. 106876
36 Ergebnisse
Sortierung:
In: Computers and electronics in agriculture: COMPAG online ; an international journal, Band 196, S. 106876
In: Productions Animales 2 (8), 135-144. (1995)
Dans les régions de productions animales intensives, l'élimination des déjections constitue un problème crucial, en raison principalement des risques de pollution des eaux par les nitrates et de l'air par l'ammoniac. On a envisagé récemment, comme alternative ou en complément du traitement des effluents, des solutions préventives visant à réduire les rejets azotés à la source. Il s'agit principalement de mieux adapter l'apport protéique de l'aliment et d'améliorer les performances des animaux. Dans le contexte de la nouvelle politique agricole commune (PAC), nous avons tenté d'évaluer le coût de cette approche préventive. Les résultats montrent que la réforme de la PAC induit des modifications importantes dans la formulation des régimes. On constate ainsi une réduction de 1 à 1,5 point de la teneur en protéines des aliments qui s'accompagne d'une diminution de l'excrétion azotée de 0,2 à 0,4 kg/porc. Cependant, cette évolution est très sensible aux rapports de prix entre sources de protéines et d'énergie. La modification de la conduite de l'alimentation permet de réduire le rejet azoté jusqu'à 500 g/porc, tout en réduisant le coût " matières premières " de l'aliment de 8 à 13 F/porc (3 à 5%). Mais cette approche induit des investissements supplémentaires au niveau de l'élevage en terme de stockage et de distribution d'aliment. L'amélioration de l'équilibre du régime en acides aminés s'accompagne d'une augmentation du coût alimentaire d'autant plus importante que l'on s'éloigne de la solution optimale. Une réduction du rejet azoté de 500 g/porc entraîne ainsi un coût supplémentaire de 2 à 4 F /porc. Si l'on combine l'amélioration de la stratégie d'alimentation et l'amélioration de l'équilibre protéique du régime, on peut réduire le rejet azoté d'environ 20 à 25 % sans augmenter significativement le coût matières premières de l'aliment. ; In regions where intensive animal production is found, the management and disposal of manure directly affects the extent and the risk of water pollution by nitrates and air pollution by ammonia. As an alternative or complement to the treatment of effluents, solutions have been sought to reduce nitrogen pollution at its source of production. One approach that has been pursued has been to better adapt protein available in feeds to animal needs. In the context of the reformed CAP, this paper presents an evaluation of the cost of such a preventive approach to nitrogen pollution control. The results show that the CAP reform has introduced important modifications in the formulation of feed rations. A 1 to 1.5 point reduction in the protein content of feeds is estimated to result in a diminution of nitrogen excretion by pigs in the range of between .2 to .4 kg/pig. Further, it is shown that these results are very sensitive to the relationship between the prices of protein and energy. Modifications in the feeding strategy (number of diets) induce a decrease in nitrogen excretion in the range of 500 g/pig, as well as a reduction in the cost of the feedstuffs by 8 to 13 FF/pig (3-5%). However, this approach requires supplementary investments at the farm level for storage and distribution of the feeds. The improvement of the amino acid balance of the diet is generally associated with an increase in total feed costs. It is estimated that a reduction in nitrogen output by 500 g/pig would involve additional costs of 2 to 4 FF/pig. Jointly considering the benefits of improved feeding program and the improved protein balance, it is clear that nitrogen production can be reduced about 20 to 25 %, without substantilly increasing the cost of primary feed ingredients.
BASE
In regions where intensive animal production is found, the management and disposal of manure directly affects the extent and the risk of water pollution by nitrates and air pollution by ammonia. As an alternative or complement to the treatment of effluents, solutions have been sought to reduce nitrogen pollution at its source of production. One approach that has been pursued has been to better adapt protein available in feeds to animal needs. In the context of the reformed CAP, this paper presents an evaluation of the cost of such a preventive approach to nitrogen pollution control. The results show that the CAP reform has introduced important modifications in the formulation of feed rations. A 1 to 1.5 point reduction in the protein content of feeds is estimated to result in a diminution of nitrogen excretion by pigs in the range of between .2 to .4 kg/pig. Further, it is shown that these results are very sensitive to the relationship between the prices of protein and energy. Modifications in the feeding strategy (number of diets) induce a decrease in nitrogen excretion in the range of 500 g/pig, as well as a reduction in the cost of the feedstuffs by 8 to 13 FF/pig (3-5%). However, this approach requires supplementary investments at the farm level for storage and distribution of the feeds. The improvement of the amino acid balance of the diet is generally associated with an increase in total feed costs. It is estimated that a reduction in nitrogen output by 500 g/pig would involve additional costs of 2 to 4 FF/pig. Jointly considering the benefits of improved feeding program and the improved protein balance, it is clear that nitrogen production can be reduced about 20 to 25 %, without substantilly increasing the cost of primary feed ingredients. ; Dans les régions de productions animales intensives, l'élimination des déjections constitue un problème crucial, en raison principalement des risques de pollution des eaux par les nitrates et de l'air par l'ammoniac. On a envisagé récemment, comme alternative ou en ...
BASE
In: Computers and Electronics in Agriculture, Band 188, S. 106329
The progress of technologies (sensors, automates) in precision livestock farming enables the development of innovative feeding techniques such as precision feeding of individual animals. In addition to the design of adapted feeders, precision feeding requires decision-support tools to manage data and apply nutritional models that calculate the optimal feed composition and allowance. These calculations require to forecast body weight (BW) and feed intake (FI) of individual pigs according to past performance. To select the mostaccurate forecasting method, three statistical methods were tested on a dataset of measurements of BW and FI for 117 pigs: the double exponential smoothing (DES) method, multivariate adaptive regression splines (MARS), and the knearest neighbours (kNN) method. These methods were tested in relation to data sampling frequency (i.e., daily or weekly measurements) and data availability. The capacity to forecast BW or FI was evaluated through the mean error of prediction. The kNN method appeared suitable if few historical data are available as it requires not more than 3 historical data. The MARS method was better than the DES method to forecast daily BW, but the DES method was better in forecasting the daily cumulated FI. The DES method also seemed more appropriate for weekly BW data, requiring only 3 historical data to make a forecasting. These methods can be used for performance forecasting in a decision-support tool for precision feeding. This study was performed in theFeed-a-Gene Project funded by the European Union's H2020 Programme (grant agreement no 633531).
BASE
In: Precision Livestock Farming '17. 2017; 8. European Conference on Precision Livestock Farming (ECPLF), Nantes, FRA, 2017-09-12-2017-09-14, 574-583
The progress of technologies (sensors, automates) in precision livestock farming enables the development of innovative feeding techniques such as precision feeding of individual animals. In addition to the design of adapted feeders, precision feeding requires decision-support tools to manage data and apply nutritional models that calculate the optimal feed composition and allowance. These calculations require to forecast body weight (BW) and feed intake (FI) of individual pigs according to past performance. To select the most accurate forecasting method, three statistical methods were tested on a dataset of measurements of BW and FI for 117 pigs: the double exponential smoothing (DES) method, multivariate adaptive regression splines (MARS), and the knearest neighbours (kNN) method. These methods were tested in relation to data sampling frequency (i.e., daily or weekly measurements) and data availability. The capacity to forecast BW or FI was evaluated through the mean error of prediction. The kNN method appeared suitable if few historical data are available as it requires not more than 3 historical data. The MARS method was better than the DES method to forecast daily BW, but the DES method was better in forecasting the daily cumulated FI. The DES method also seemed more appropriate for weekly BW data, requiring only 3 historical data to make a forecasting. These methods can be used for performance forecasting in a decision-support tool for precision feeding. This study was performed in the Feed-a-Gene Project funded by the European Union's H2020 Programme (grant agreement no 633531).
BASE
International audience ; Nutritional requirements of lactating sows mainly depend on milk yield and greatly vary across individuals. Moreover, because the same diet is generally fed to all sows, and feed intake is low and highly variable, nutrient supplies are often insufficient to meet the requirements, especially those of primiparous sows. Conversely, sows with high appetite may be fed nutrient in excess. Acquiring data on sows and their environment at high-throughput allows the development of new precision feeding systems with the perspective of improving technical performance and reducing feeding cost and environmental impact. The objective of this study was thus to design a decision support tool that could be incorporated in automated feeding equipment. The decision support tool was developed on the basis of InraPorc® model. The optimal supply for a given sow is determined each day according to a factorial approach considering all the information available on the sow (i.e. parity, litter size, milk production, body condition…) or predicted from real-time data (i.e. expected feed intake). The approach was tested using data from 817 lactations. Precision feeding (PF) with the mixing of two diets with different nutritional values was then simulated in comparison with conventional feeding (CF) with a single diet. In sows fed in excess PF reduced average digestible lysine excess from 10.9 to 2.7 g/d, whereas in deficient sows the deficiency was reduced from -5.7 to -2.1 g/d. Overall, PF reduced average lysine intake by 6.8%. At the same time, with PF, lysine requirement was met for a higher proportion of sows, especially in younger sows, and a lower proportion of sows, especially older sows, received excessive supplies. PF also reduced average phosphorus intake while limiting the occurrence of excess and deficiency. This study confirms the potential of precision feeding in order to better achieve nutritional requirements of lactating sows and reduce their nutrient intake and excretion. This project has received funding from the European Union's Horizon 2020 research and innovation program, grant agreement no. 633531.
BASE
In: Book of Abstracts of the 69th Annual Meeting of the European Federation of Animal Science. (24)2018; 69. Annual Meeting of the European Federation of Animal Science (EAAP), Dubrovnik, HRV, 2018-08-27-2018-08-31, 334
Nutritional requirements of lactating sows mainly depend on milk yield and greatly vary across individuals. Moreover, because the same diet is generally fed to all sows, and feed intake is low and highly variable, nutrient supplies are often insufficient to meet the requirements, especially those of primiparous sows. Conversely, sows with high appetite may be fed nutrient in excess. Acquiring data on sows and their environment at high-throughput allows the development of new precision feeding systems with the perspective of improving technical performance and reducing feeding cost and environmental impact. The objective of this study was thus to design a decision support tool that could be incorporated in automated feeding equipment. The decision support tool was developed on the basis of InraPorc® model. The optimal supply for a given sow is determined each day according to a factorial approach considering all the information available on the sow (i.e. parity, litter size, milk production, body condition…) or predicted from real-time data (i.e. expected feed intake). The approach was tested using data from 817 lactations. Precision feeding (PF) with the mixing of two diets with different nutritional values was then simulated in comparison with conventional feeding (CF) with a single diet. In sows fed in excess PF reduced average digestible lysine excess from 10.9 to 2.7 g/d, whereas in deficient sows the deficiency was reduced from -5.7 to -2.1 g/d. Overall, PF reduced average lysine intake by 6.8%. At the same time, with PF, lysine requirement was met for a higher proportion of sows, especially in younger sows, and a lower proportion of sows, especially older sows, received excessive supplies. PF also reduced average phosphorus intake while limiting the occurrence of excess and deficiency. This study confirms the potential of precision feeding in order to better achieve nutritional requirements of lactating sows and reduce their nutrient intake and excretion. This project has received funding from the European Union's Horizon 2020 research and innovation program, grant agreement no. 633531.
BASE
International audience ; Nutritional requirements of lactating sows mainly depend on milk yield and greatly vary across individuals. Moreover, because the same diet is generally fed to all sows, and feed intake is low and highly variable, nutrient supplies are often insufficient to meet the requirements, especially those of primiparous sows. Conversely, sows with high appetite may be fed nutrient in excess. Acquiring data on sows and their environment at high-throughput allows the development of new precision feeding systems with the perspective of improving technical performance and reducing feeding cost and environmental impact. The objective of this study was thus to design a decision support tool that could be incorporated in automated feeding equipment. The decision support tool was developed on the basis of InraPorc® model. The optimal supply for a given sow is determined each day according to a factorial approach considering all the information available on the sow (i.e. parity, litter size, milk production, body condition…) or predicted from real-time data (i.e. expected feed intake). The approach was tested using data from 817 lactations. Precision feeding (PF) with the mixing of two diets with different nutritional values was then simulated in comparison with conventional feeding (CF) with a single diet. In sows fed in excess PF reduced average digestible lysine excess from 10.9 to 2.7 g/d, whereas in deficient sows the deficiency was reduced from -5.7 to -2.1 g/d. Overall, PF reduced average lysine intake by 6.8%. At the same time, with PF, lysine requirement was met for a higher proportion of sows, especially in younger sows, and a lower proportion of sows, especially older sows, received excessive supplies. PF also reduced average phosphorus intake while limiting the occurrence of excess and deficiency. This study confirms the potential of precision feeding in order to better achieve nutritional requirements of lactating sows and reduce their nutrient intake and excretion. This project has received funding from the European Union's Horizon 2020 research and innovation program, grant agreement no. 633531.
BASE
Les travaux de recherche sur l'alimentation des porcs ont été conduits par l'INRAgrâce au soutien financier de partenaires privés via des collaborations bilatérales avec des firmes services (ou des producteurs et vendeurs d'additifs) ou des collaborations plus larges regroupant une grande partie des acteurs de l'alimentation animale. Les principaux outputs de ces travaux de recherche ont été : (1) des concepts et méthodes de référence pour caractériser la valeur des ressources et les besoins des animaux, (2) des outils d'aide à la décision (OAD) pour faciliter la diffusion et l'appropriation des connaissances auprès des acteurs de la filière au niveau national et international. Les résultats de ces travaux ont été utilisés et diffusés hors de la sphère académique grâce à des activités d'intermédiation (expertises, réunions techniques) réalisées d'abord par les chercheurs eux-mêmes. Les firmes services françaises (notamment celles associées au GERNA ) ont rapidement intégré et adapté les nouveaux concepts d'évaluation des aliments pour que leurs clients bénéficient des progrès réalisés, tant pour l'énergie (énergie digestible vers énergie nette (EN)), que pour l'azote, sur leur marge nette de production. Certaines entreprises privées (Ajinomoto Eurolysine) ont financé la diffusion des résultats (aide à la production d'outils d'aide à la décision) et ont également utilisé leur réseau commercial en France et dans le monde pour aider à leur mise en pratique. Enfin, la diffusion de la connaissance sur l'adéquation entre les besoins et les apports nutritionnels pour limiter les rejets a été en grande partie effectuée grâce aux recommandations successives du CORPEN dans lequel l'INRA a eu un rôle déterminant. Au niveau de la filière porcine française et européenne, les systèmes d'alimentation chez le porc que l'INRA a conçus ont eu des impacts sur les trois piliers de la durabilité. Au niveau économique et social, ils ont permis d'améliorer la rentabilité de différents maillons de la filière (firmes services, ...
BASE
Nutritional studies indicate that nutrient requirements for pregnancy largely differ among sows and according to the stage of pregnancy, whereas in practice the same diet is generally fed to all sows from a given herd. In this context, the availability of new technologies for high throughput phenotyping of sows and their environment, and of innovative feeders that allow the distribution of different diets, offers opportunities for a renewed and practical implementation of prediction models of nutrient requirements, in the perspective of improving feed efficiency and reducing feeding costs and environmental impacts. The objective of this study was thus to design a decision support tool that could be incorporated in automated feeding equipment. The decision support tool was developed on the basis of InraPorc® model. The optimal supply for a given sowis determined each day according to a factorial approach considering all the information available on the sow: genotype, parity, expected prolificacy, gestation stage, body condition (i.e. weight and backfat thickness), activity and housing (i.e. type of floor and ambient temperature). The approach was tested using data from 2500 pregnancies on 540 sows. Energy supply was calculated for each sow to achieve, at farrowing, a target body weight established based on parity, age at mating and backfat thickness (18 mm). Precision feeding (PF) with the mixing of two diets was then simulated in comparison with conventional (CF) feeding with a single diet. Compared to CF, PF reduced protein and aminoacid intake, N excretion and feeding costs. At the same time, with PF, amino acid requirement was met for a higher proportion of sows, especially in youngersows, and a lower proportion of sows, especially older sows, received excessive supplies. This project has received funding from the European Union's Horizon 2020 research and innovation programme, grant agreement No 633531. The data used for the simulations were issued from a project conducted within theAgriInnovation Program ...
BASE
In: Book of Abstract of the 68th Annual Meeting of the European Federation of Animal. 2017; 68. Annual Meeting of the European Federation of Animal Science (EAAP), Tallinn, EST, 2017-08-28-2017-09-01, 319
Nutritional studies indicate that nutrient requirements for pregnancy differ largely among sows and according to the stage of pregnancy, whereas in practice the same diet is generally fed to all sows in a given herd. In this context, the availability of new technologies for high throughput phenotyping of sows and their environment, and of innovative feeders that allow the distribution of different diets, offers opportunities for a renewed and practical implementation of prediction models of nutrient requirements, in the perspective of improving feed efficiency and reducing feeding costs and environmental impacts. The objective of this study was thus to design a decision support tool that could be incorporated in automated feeding equipment. The decision support tool was developed on the basis of InraPorc® model. The optimal supply for a given sow is determined each day according to a factorial approach considering all the information available on the sow: genotype, parity, expected prolificacy, gestation stage, body condition (i.e. weight and backfat thickness), activity, and housing (i.e. type of floor and ambient temperature). The approach was tested using data from 2,500 pregnancies on 540 sows. Energy supply was calculated for each sow to achieve, at farrowing, a target body weight established based on parity, age at mating and backfat thickness (18 mm). Precision feeding (PF) with the mixing of two diets was then simulated in comparison with conventional (CF) feeding with a single diet. Compared to CF, PF reduced protein and amino acid intake, N excretion and feeding costs. At the same time, with PF, amino acid requirements were met for a higher proportion of sows, especially in younger sows, and a lower proportion of sows, especially older sows, received excessive supplies. This study is part of the Feed-a-Gene project and received funding from the European Union's H2020 program under grant agreement no. 633531. The data used for the simulations were issued from a project conducted within the AgriInnovation Program from Agriculture and Agri-food Canada.[br/]
BASE
In: Precision Livestock Farming '17. 2017; 8. European Conference on Precision Livestock Farming (ECPLF), Nantes, FRA, 2017-09-12-2017-09-14, 584-592
Nutritional studies indicate that nutrient requirements for pregnancy largely differ among sows and according to the stage of pregnancy, whereas in practice the same diet is generally fed to all sows from a given herd. In this context, the availability of new technologies for high throughput phenotyping of sows and their environment, and of innovative feeders that allow the distribution of different diets, offers opportunities for a renewed and practical implementation of prediction models of nutrient requirements, in the perspective of improving feed efficiency and reducing feeding costs and environmental impacts. The objective of this study was thus to design a decision support tool that could be incorporated in automated feeding equipment. The decision support tool was developed on the basis of InraPorc® model. The optimal supply for a given sow is determined each day according to a factorial approach considering all the information available on the sow: genotype, parity, expected prolificacy, gestation stage, body condition (i.e. weight and backfat thickness), activity and housing (i.e. type of floor and ambient temperature). The approach was tested using data from 2500 pregnancies on 540 sows. Energy supply was calculated for each sow to achieve, at farrowing, a target body weight established based on parity, age at mating and backfat thickness (18 mm). Precision feeding (PF) with the mixing of two diets was then simulated in comparison with conventional (CF) feeding with a single diet. Compared to CF, PF reduced protein and aminoacid intake, N excretion and feeding costs. At the same time, with PF, amino acid requirement was met for a higher proportion of sows, especially in younger sows, and a lower proportion of sows, especially older sows, received excessive supplies. This project has received funding from the European Union's Horizon 2020 research and innovation programme, grant agreement No 633531. The data used for the simulations were issued from a project conducted within the AgriInnovation Program from Agriculture and Agri-food Canada.
BASE
Precision feeding is a promising way to improve feed efficiency and thus economic and environmental sustainability of livestock production. A decision support system (DSS) was built to determine in real-time the nutritional requirements of animals and feed characteristics (i.e. composition, amount) for an application of precision feeding in pig and poultry commercial farms. This tool, dedicated to animals managed individually or in groups, is designed with a modular structure for adaptation to different feeder devices, species and production stages. The modules are built to perform specialized tasks in a cooperative way. It includes a data management module with a proper characterization of data by meta-data definition for precision feeding. It ensures standard encoding to allow data interoperability from any platform. Other modules are dedicated to data verificatrion and correction for inclusion in a database, prediction of most probable body weight (BW) gain and feed intake (ad libitum or restricting feeding), and the calculation of nutritional requirements. The BW and feed intake prediction is based on dynamic data analyses. For that, specific methods have been evaluated and selected depending on the number of available data, data type (BW or feed intake), and recording frequency. The calculation of nutritional requirements is performed using nutritional models specific for a species or production stage. These two modules are currently designed for healthy animals and will be refined to extend prediction to a larger range of field situations (e.g. health problems, climatic conditions) with nutritional models in development/refinement in other workpackages of the project. The general specifications of this DSS and dynamic data analyses will be illustrated for growing pigs. This study is part of the Feed-a-Gene project and received funding from the European Union's H2020 program under grant agreement no. 633531.
BASE
In: Book of Abstract of the 68th Annual Meeting of the European Federation of Animal. 2017; 68. Annual Meeting of the European Federation of Animal Science (EAAP), Tallinn, EST, 2017-08-28-2017-09-01, 319
Precision feeding is a promising way to improve feed efficiency and thus economic and environmental sustainability of livestock production. A decision support system (DSS) was built to determine in real-time the nutritional requirements of animals and feed characteristics (i.e. composition, amount) for an application of precision feeding in pig and poultry commercial farms. This tool, dedicated to animals managed individually or in groups, is designed with a modular structure for adaptation to different feeder devices, species and production stages. The modules are built to perform specialized tasks in a cooperative way. It includes a data management module with a proper characterization of data by meta-data definition for precision feeding. It ensures standard encoding to allow data interoperability from any platform. Other modules are dedicated to data verificatrion and correction for inclusion in a database, prediction of most probable body weight (BW) gain and feed intake (ad libitum or restricting feeding), and the calculation of nutritional requirements. The BW and feed intake prediction is based on dynamic data analyses. For that, specific methods have been evaluated and selected depending on the number of available data, data type (BW or feed intake), and recording frequency. The calculation of nutritional requirements is performed using nutritional models specific for a species or production stage. These two modules are currently designed for healthy animals and will be refined to extend prediction to a larger range of field situations (e.g. health problems, climatic conditions) with nutritional models in development/refinement in other workpackages of the project. The general specifications of this DSS and dynamic data analyses will be illustrated for growing pigs. This study is part of the Feed-a-Gene project and received funding from the European Union's H2020 program under grant agreement no. 633531.
BASE