Abstract
With growing population and limited land for cultivation, advanced farming techniques and mechanization are gaining more traction worldwide. Technology also helps overcome the hurdles a typical farmer faces, which is also attracting the younger generation to cultivation. India, whose economy largely depends on agriculture; uses a tractor and implements mainly in land preparation, puddling, seeding, and harvesting. Since the average farm size is small, the widely used tractors are under 80HP. Owning and maintaining tractors for small lands are not viable which created a rental ecosystem of the tractor and corresponding equipment. A Power Take Off (PTO) shaft is provided in such tractors to use tractor engine power for multiple agriculture or non-agriculture applications such as driving implements like rotavators or driving irrigation pump sets. This PTO is engaged or disengaged by levers in the tractor. Real-time detection of PTO ON or OFF condition is useful for understanding tractor utilization and which also has a correlation with service and maintenance. For the owners who are renting, the charges will vary according to the number of PTO operation hours. From dealer and service center point of view, the PTO usage hours is useful to correlate the failures during services. Current practice in tractor models to track PTO usage is using a dedicated switch or sensor. Often with the rough field conditions; the switches get damaged; which makes it difficult to track the PTO usage. Telematics devices which collect tractor sensor information and aggregate it at a server for the end user enabling real-time tracking of PTO operation with a web portal or a mobile application. Dedicated PTO switch and telematics are typically available in higher HP tractor models or on higher price variants. With increasing demand for connected vehicles and reduced costs of data with mobile networks; telematics penetration on tractors is expected to grow up in coming years. In our study, we use the collected data from tractors with telematics and PTO switch to build a supervised machine learning model to predict PTO operation. This model can be then applied in predicting PTO usage for tractors where PTO switch is not available, or the switch is damaged or non-functional.