Smart Fault Tracing: Learning with Diagnostic Data for Predicting Vehicle States
Abstract: Software applications, as the control centre of vehicle functionality, is becoming much important in recent years. The failure of downloading software applications can cause the loss of vehicle functionality. A tool that effectively estimates the software downloading statuses and locates root causes when failure happens, is highly required. The thesis investigates supervised learning methods, proposes a quantitative and data-driven approach to estimate software downloading statuses and measure the effect of software parts on software downloading process. The goal is to find out if classification models can be used to predict software downloading statuses, and help locate the part numbers that cause the failures in software downloading process. The experiment results indicate that the classification models can help predict the statuses with high prediction performance, and can help locate the causes. The trained models can be used to predict upcoming failures on other vehicles that have the same ECUs. The proposed method also gives a hint that classification tools can help evaluate the statuses of other components of the vehicle system, and help suggest vehicle maintenance planning. A tool that automatically analyses vehicle statuses will be helpful, which can be future work.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)