Design of software-oriented technician for vehicle’s fault system prediction using AdaBoost and random forest classifiers

Abstract

Detecting and isolating faults on heavy duty vehicles is very important because it helps maintain high vehicle performance, low emissions, fuel economy, high vehicle safety and ensures repair and service efficiency. These factors are important because they help reduce the overall life cycle cost of a vehicle. The aim of this paper is to deliver a Web application model which aids the professional technician or vehicle user with basic automobile knowledge to access the working condition of the vehicles and detect the fault subsystem in the vehicles. The scope of this system is to visualize the data acquired from vehicle, diagnosis the fault component using trained fault model obtained from improvised Machine Learning (ML) classifiers and generate a report. The visualization page is built with plotly python package and prepared with selected parameter from On-board Diagnosis (OBD) tool data. The Histogram data is pre-processed with techniques such as null value Imputation techniques, Standardization and Balancing methods in order to increase the quality of training and it is trained with Classifiers. Finally, Classifier is tested and the Performance Metrics such as Accuracy, Precision, Re-call and F1 measure which are calculated from the Confusion Matrix. The proposed methodology for fault model prediction uses supervised algorithms such as Random Forest (RF), Ensemble Algorithm like AdaBoost Algorithm which offer reasonable Accuracy and Recall. The Python package joblib is used to save the model weights and reduce the computational time. Google Colabs is used as the python environment as it offers versatile features and PyCharm is utilised for the development of Web application. Hence, the Web application, outcome of this proposed work can, not only serve as the perfect companion to minimize the cost of time and money involved in unnecessary checks done for fault system detection but also aids to quickly detect and isolate the faulty system to avoid the propagation of errors that can lead to more dangerous cases.