Vibration Data Mining and Machine Learning for Anomaly Detection of Cycling Trails Using Instrumented Bike

The instrumented bike has been seen as a useful tool for condition surveys of cycling trails. This paper introduces the recent development of instrumented bike and its application in quality assessment of cycling trails through cycling data collection, data mining/processing, data analysis, anomaly detection and mapping. A machine learning-based computing algorithm using LSTM method is presented to demonstrate how vibration patterns are screened and anomalies are identified. A cycling test site was selected at Northern Arizona University, USA. Four cyclists were recruited to participate in the cycling data collection using their individual instrumented bike. The results show the LSTM method is capable of analyzing vibration patterns and identifying anomalies along the cycling trails. The computing algorithm is suitable in the development of instrumented bike.