Deep learning-based method for detecting anomalies of operating equipment dynamically in livestock farms

The scale of livestock farms has grown significantly and the amount of livestock being reared is also increasing recently. As a result, interest in automated livestock smart farms is huge. To realize a smart farm, the environment for livestock (limited to pigs in this paper) in the barn is properly maintained for growth conditions, thereby increasing its productivity and animal welfare. To maintain such a suitable environment, many and various equipment are built and operated inside and outside the pig house. However, due to poor environments, the failures of lots of environment equipment are high. Furthermore, there are difficulties in detecting its malfunctions during equipment operation. In this paper, we provide a mechanism to simultaneously detect anomalies in such various and lots of equipment and to quickly detect them, which is adaptable to the environment of each pig house. Data from lots of equipment (environment sensors and controllers, etc.) installed in a pig house are collected. Through the data to predict malfunctions of each equipment, the learning model is built using RNN. When something goes wrong with the sensor, there is a difference between the predicted value and the measured value, which shows that the models can work well at the same time. It is possible to increase the productivity of pigs in various types of livestock farms where various and lots of equipment is built.