A Comparative Study with J48 and Random Tree Classifier for Predicting the State of Hydraulic Braking System through Vibration Signals

Even though hydraulic brakes are valuable safety elements for riders, they are necessary for braking in a good condition. The state of the brake components can be determined by using vibrational signatures. In this proposed study, interactive condition monitoring with a piezo-electric transducer and the dynamic data acquisition mechanism is suggested as a potential solution to such problems using a machine learning approach. For good and poor braking conditions, the Ford EcoSport setup was used to get the vibration signal. The empirical statistical descriptive characteristics of the vibration signals were obtained and the J48 decision tree model was used in the selection. A systematic decision does not specify the sufficient number of features necessary to characterize a certain problem. Therefore, to find the right number of features a rigorous analysis is required. The failure study of the hydraulic braking system of Ford EcoSport has been determined using the decision tree classification J48 and the random wood classification. The findings were compared and showed that a random forest classifier with a computational time of 0.43s had a maximum classification accuracy of 98.5%.