Home > Research > Research Areas  
Active Fault Diagnosis System for In-Wheel Motor Electric Vehicle
 
The number of actuators and sensors increases in Electric Vehicle (EV) system. One of the remarkable advantages of the EV is applicability of In-Wheel motor (IWM) as well as Brake-by-wire and Steer-by-wire. Therefore EV has much larger number of actuators and sensors compared to Internal Combustion Engine Vehicle (ICEV). However fault probability also increases. Therefore fault diagnosis system is needed.

Figure. Electric Vehicle Components


EV is over actuated system (redundantly actuated system) which means that the number of control/actuation is greater than system mobility. Using redundant actuation is a common approach to satisfy the continuously increasing demand on system performance. Over-actuated feedback control system has redundant actuation so control input distribution is not unique to achieve desired control objectives. We propose active fault diagnosis system for over-actuated feedback control system to improve fault diagnosis performance.

Until 2013, we studied about degree of fault isolability and active fault diagnosis using weighting changes.

First, fault isolability is a structural property related with system dynamics and composition of actuators and sensors. The cases of existing researches of testing fault isolability involve checking whether the system is isolable or not, i.e., binary nature. However, even if it meets a mathematical fault isolation condition, it will be hard to classify it as a real fault due to disturbance, signal noise and model uncertainty. Therefore, it is necessary to introduce degree of fault isolability, which is continuous value rather than binary metric. We propose a degree of fault isolability for multi-input-multi-output linear feedback control system by analyzing residual vectors in parity space. Degree of fault isolability can be applied to evaluate how much isolable a given system is, in continuous value.

Second, procedure of active fault diagnosis system is as follows. In normal condition, controller decides control input allocation to optimize desired control strategy. In fault condition, fault diagnosis system adjusts control input distribution for the purpose of fault diagnosis. By using active fault diagnosis method, much more system information can be obtained so performance of fault diagnosis can be improved. Changing weighting factors in control input allocation is employed for adjustment strategy of control input distribution.

However, degree of isolability is only applicable to ideal system. The external disturbance or measurement noise degrades the performance of degree of isolabilty. Also, current active fault diagnosis system use just sequential strategy which is inefficient and physical meaning is weak. Therefore from 2014, we are studying about robust degree of isolability and optimal active fault diagnosis system.