ROBUST PATTERN RECOGNITION BASED FAULT DETECTION AND
ISOLATION METHOD FOR ABS SPEED SENSOR |
Ayad Qays Abdulkareem , Abdulrahim Thiab Humod , Oday Ali Ahmed |
Department of Electrical Engineering, University of Technology-Iraq |
|
|
|
| |
ABSTRACT |
Anti-lock braking system (ABS) is considered an essential safety system in electric vehicles that works to
grant a reliable vehicle driving experience, and it is very important to ensure the security of such an onboard safety system.
This work presents a detailed analysis associated with a comparison that includes several techniques based on pattern
recognition for biasing fault detection in wheel and vehicle speed sensors. These techniques are K-nearest neighbor (KNN),
support vector machine (SVM) and decision tree (DT), which were selected among other pattern recognition techniques that
have been studied. The MATLAB Simulink model for the ABS system was implemented, and data was extracted from
healthy and unhealthy operating conditions in order to be used to train each technique individually. An offline test was
applied to these trained FDI models using the same implemented ABS Simulink model to express the performance of each
one. Specifically speaking, accuracy and sensitivity were used in the algorithm’s efficiency comparison, with 99.9 %
accuracy in the Fine KNN, 75 % accuracy in the Coarse Gaussian SVM, and 61.5 % accuracy in the Coarse Tree. From the
result, and considering the ABS issues mentioned above, it can be concluded that the KNN classifier is superior to both the
SVM and TREE classifiers. |
Key Words:
Anti-lock braking system, Sensor fault, Fault detection, Pattern recognition, KNN, SVM, Decision tree |
|