Read more
As the technology advances and vehicles become more computerized, the complexity of automotive electronics grows exponentially, thus creating a new level of challenge for detecting faults in systems with such overwhelming number of active components. The proposal of this research was to analyze the data from the Controller Area Network from two vehicles, evaluate different techniques for data mining and feature extraction and devise possible methods for fault detection that could be employed in any system with this level of complexity. The data used in this research was collected from an ordinary vehicle and an experimental vehicle test bed, the LabCar. The techniques evaluated in this research were based on proven mathematical methods that had original application in areas of statistics and digital signal processing. A detailed analysis of each of the techniques studied and their suitability for analysing complex non-linear systems were documented with basis on the data collected. The final stages of this research consisted of using these techniques to devise a model of the Controller Area Network bus and use it for finding faults in the network.
About the author
Graduated with First Class Hons in Robotics and Automated Systems at The University of Plymouth, followed a distinct degree in MSc Electronics systems at the University of Warwick. Worked in many areas from automated testing systems to intelligent design for bipedal robot gait control.
