Nové prediktivní diagnostické metody automobilových řídicích systémů

Abstract

During the 4th Industrial Revolution with a large number of technologies and systems in operation, predictive maintenance is becoming a significant industry for solutions optimizing maintenance costs. In the automotive industry and the car electronics sector, by implementing predictive diagnostic methods, it is possible to achieve an increased operational reliability before the occurrence of a defect, and a fundamental increase in the probability of identifying a defect once it occurred. This approach provides a major advantage, as modern cars are very complicated and difficult to diagnose. Continuous analysis of sensor and actuator data together with machine learning techniques for fault prediction are beneficial for the current development of on-board diagnostic systems.

The goals of this dissertation thesis are the development of an automotive control system platform, implemented using a fully programmable unit and a research of fault prediction methods based on algorithms of supervised machine learning. To achieve the mentioned goals, an extensive SW/HW analysis of the selected internal combustion engine management was conducted, along with a development of an on-board diagnostics algorithm including a predictive diagnostic system with classification and degradation algorithms based on speed, accuracy and scalability.