DESIGN OF A MACHINE LEARNING BASED FRAMEWORK FOR REALISTIC WEAR AND TEAR ESTIMATION OF THE BRAKING SYSTEM OF A VEHICLE

Abstract

Commuting from place to place is among one of the basic needs of human beings. Vehicles are the general mode of transportation since their commercialization. The major safety hazard in vehicles is the braking system which cannot only endanger the lives of the people travelling in it but also the others who are on the roads or in other vehicles. A realistic wear and tear analysis of the braking system can help in avoiding severe accidents. We have devised a framework which will provide the required hardware and software tools to estimate the realistic wear and tear of the braking system of a vehicle. The machine learning algorithms are used to determine different modes of the transportation of a vehicle like moving, stationary, braking, turning left or right. A high shock survivability and high resolution accelerometer is used to get the raw data and then with the help of machine learning algorithms specific criteria is set to get a threshold. When the decelerations cross that particular threshold, an indication is set to let the user know about the time for the check up of the braking system which could be the replacement of the brake pads or other component. The framework is flexible enough to be used to get the wear and tear of any part of the vehicle which has a direct impact upon change in acceleration. The framework provides a learning mode which helps to define a specific criteria and a monitoring mode which actually indicates that the specific criteria has met.