Non-Invasive Real Time Error State Detection for Tractors Using Smart Phone Sensors & Machine Learning

Condition Monitoring is the process of identifying any significant change in operating parameters of a machine, which can be indicative of a failure in future. This paper discuss a non-invasive condition monitoring methodology for sensing and investigating the problems which could be identified by noise and vibrations. This could be an easy solution for predicting failures in tractors which are operational in the field. An example of engine tappet is used to demonstrate the methodology. A disturbed setting causes a distinguishable noise, referred to as “tappet rattle”. Android smartphones (with inbuilt sensors - accelerometer, gyroscope and microphone) are used to record noise and vibration from tractors in good condition as well as in disturbed condition. Time series data analysis is done to extract relevant features and then Fourier Transform is applied to the signals for extracting frequency domain signatures. Frequency domain-based features have shown significant improvements in the model prediction accuracy. This paper compares different classification algorithms and evaluate the results on performance and accuracy. The trained model is then used along with a smartphone application to do the real-time detection without any additional sensors.