Elsevier

Biosystems Engineering

Volume 173, September 2018, Pages 143-156
Biosystems Engineering

Special Issue: Engineering Advances in Precision Livestock Farming
Research Paper
Online warning systems for individual fattening pigs based on their feeding pattern

https://doi.org/10.1016/j.biosystemseng.2017.08.006Get rights and content

Highlights

  • Four warnings systems based on individual pig feeding patterns were validated.

  • Time-varying individual limits seem most promising.

  • Sensitivity was 58.0%, specificity 98.7%, accuracy 96.7% and precision 71.1%.

  • Severe problems were detected on average already within 1.3 days.

  • Further improvements are warranted to make the system more practical.

For sustainable pork production and maximum pig welfare, all health, welfare and productivity problems in the barn should be detected as early as possible. In this paper, an automated monitoring and warning system is proposed. Based on measurements of the feeding pattern, it is able to generate daily alerts for individual fattening pigs. Using historical data, the following types of warning systems were developed: (1) fixed limits that treat all pigs and all days equally; and (2) time-varying individual limits using the concept of Synergistic Control. These types of limits were constructed either for the number of registrations per pig or the average interval between feeding visits of a pig, leading to four warning systems in total. These warning systems were used to generate alerts during an online validation period. During an entire fattening period, all pigs were individually monitored to establish true alerts, false alerts and missed problems. The best performance was achieved for the Synergistic Control method on the number of registrations, with a sensitivity of 58.0%, specificity of 98.7%, accuracy of 96.7% and precision of 71.1%. Severe problems were detected on average within 1.3 days from the start of the problem. These are promising results that provide a solid basis for the development of a system for individual pigs but further improvements are warranted to make the system more practical.

Introduction

In pig farming, disease control, animal welfare and production efficiency are important factors to help ensure sustainable pork production and maintain an economically viable farm. Therefore, it is important that health, welfare and productivity problems in the barn are detected and treated early. As the sector intensifies and farms and groups of pigs become larger, visual monitoring of the pigs as a sole tool for problem detection could be suboptimal. Visual monitoring gives only a snapshot-view on the animals appearance (Heitkämper, Schick, & Fritzsche, 2011; van den Heuvel, Hoofs, Binnendijk, Bosma, & Spoolder, 2004) and is often more focused on the group level than the individual level in pig farming. Automated monitoring or Precision Livestock Farming (PLF) (see for example Wathes, Kristensen, Aerts, and Berckmans (2008) and Banhazi et al. (2012)) allows to monitor the livestock online and continuously (Matthews, Miller, Clapp, Plötz, & Kyriazakis, 2016). The automatically gathered measurement data can be transformed into information for the farmer and support the farmer's decision making-process (Cornou & Kristensen, 2013). Using the right techniques, automated monitoring can also be done at the individual pig level, allowing for individual, custom-made care.

Disease, welfare and productivity problems can have an impact on the feeding pattern of a pig (Brown-Brandl et al., 2013, Hart, 1988, Hessel and Van den Weghe, 2011), such as a reduced feeding time or longer intervals between visits. Therefore, a system to measure individual pigs' feeding patterns has recently been developed and validated (Maselyne et al., 2014a, Maselyne et al., 2014b). Using high frequency (HF) Radio Frequency Identification (RFID), each pig's attendance at the feeder is registered (Maselyne & Saeys, 2014). From these raw data, feeding pattern variables such as the number and duration of feeding visits and pauses between feeding visits of a single pig throughout the day can be calculated (feed intake was not measured) (Maselyne et al., 2016). The present study investigated whether abnormal changes in the feeding pattern of a pig can be detected automatically and used as an (early) indicator for health, welfare and productivity problems.

To detect abnormal changes in the feeding pattern of a pig, fixed limits (the same limit for all pigs and days) can be constructed. However, it has been shown that using a Synergistic Control (SGC) procedure can be a better, alternative option for monitoring livestock production systems (Mertens, Decuypere, De Baerdemaeker, & De Ketelaere, 2011). SGC combines the power of Engineering Process Control (EPC) and Statistical Process Control (SPC) (Montgomery, 2009). In SPC, control limits allow to differentiate abnormal variation from normal variation (due to age, seasonal effects, etc.). The EPC step pre-treats the raw livestock production data to meet the assumptions of the statistical control chart in the SPC step. Thanks to this combination, the online SGC procedure allows to use pig-specific control-limits, which can be updated with every new measurement. Any abnormal variation detected can then be signalled to the farmer as an alert for a specific pig. Promising results have already been obtained with this SGC approach for monitoring process parameters of flocks of laying hens (Mertens et al., 2008, Mertens et al., 2009) and milk yield of individual dairy cows for mastitis detection (Huybrechts, Mertens, De Baerdemaeker, De Ketelaere, & Saeys, 2014).

Therefore, the aims of the present study were (1) to develop several warning systems with fixed limits or variable, individual limits on promising variables of the feeding pattern, based on historical data; (2) to validate and compare these warning systems online by comparing the alerts with detailed observations.

Section snippets

Animals and housing

The pigs were housed in an automatically ventilated barn at the experimental farm of ILVO (Melle, Belgium). They were housed in four identical pens. Each pen measured 4.3 m by 9 m with approximately 40% slatted concrete floor and 60% solid concrete lying area. In addition to natural light, artificial lighting was provided from 7:00 to 21:00. Water was supplied ad libitum via nipple drinkers. Dry pelleted feed was automatically supplied using Swing MIDI feeders (Big Dutchman Pig Equipment GmbH,

Overview of the historical dataset

In the historical dataset, days with technical problems related to the RFID measurements or the feed and water supply were not considered in the further data analysis. These were four days for the entire barn and an extra 2, 2, 4 and 19 days for the separate pens. Also pigs which had lost an ear tag were removed (four pigs in total).

Data of pigs that were found dead, euthanised or removed were used until the day of removal. Seven pigs of the first batch died and one pig was euthanised. The

Discussion

The overall best performing warning system was the Synergistic Control method on the number of registrations (SGC # reg). This system had a sensitivity of 58.0%, specificity of 98.7%, accuracy of 96.7% and precision of 71.1%. The average time until a first false alert was 101.0 days (in-control average run length ARL0) and severe problems were detected within 1.3 days on average (out-of-control average run length ARL1).

These results are promising and are in line with the results for other

Conclusion

To detect problems in individual fattening pigs, warning systems were developed to detect changes in the feeding patterns of individual fattening pigs pointing towards health, welfare and productivity problems. The individual feeding patterns were measured using an RFID system at the feeder trough. Both fixed limits (one threshold for all pigs and days) and individual, time-varying limits constructed using Synergistic Control were developed. The best performance was achieved for the Synergistic

Acknowledgments

Jarissa Maselyne has been funded by a PhD grant from the Agency for Innovation by Science of Technology (IWT Flanders – project SB 111447). The authors would like to thank Liesbet Pluym, Annelies Michiels and Liesbeth De Wilde for their help during this study. Special thanks also go to the technical staff of ILVO for the work and technical support provided and to the animal caretakers for their daily care of the pigs.

References (29)

Cited by (0)

View full text