Bond Graph Based Fault Diagnosis And Prognosis For Electric Scooter

As the social economic development and the acceleration of aging population,electric scooter for elders has received increasing attention.Electric scooter is especially designed for the older and disabled people to keep an active lifestyle.It gives people facility on work and life but as the same time brings a security risk,or even leads to a serious consequence.Therefore,it is imperative to develop a fault diagnosis and prognosis approach for the electric scooter,which can ensure the safety and reliability of the system.This thesis proposes an uncertain bond graph(UBG)based compound fault(including progressive and abrupt faults)detection and isolation as well as an enhanced unscented Kalman filter(EUKF)based fault estimation and remaining useful life(RUL)prediction for an electric scooter in the presence of parameter uncertainties.Firstly,an adaptive threshold considering parameter uncertainties is built from UBG to achieve reliable fault detection.In addition,the concept of dependent analytical redundancy relations(ARRs)is introduced to improve the fault isolation capability under multi-fault conditions.Secondly,after a set of possible faults is isolated,fault estimation is enabled where the EUKF is proposed to ensure the prompt state tracking ability by enhancing the covariance matrix when a sudden change of fault value is detected by the second derivative of the sum of ARRs.Finally,for RUL prediction,a degradation model with time-varying decay rate is used to approximate the progressive fault,and the model coefficients are identified from the previously estimated fault value,the sudden change time and fault occurrence time.Once the degradation model coefficients are determined,the degradation trend of the fault can be used to predict the RUL statistically.The effectiveness of the proposed methodologies is validated by simulation and experiment results in the electric scooter platform.