Application of support higher-order tensor machine in fault diagnosis of electric vehicle range-extender

It is difficult to establish accurate mathematical models to describe the range extender electric vehicles due to the non-stationary, non-linear and interconnection of the monitoring signal sources resulted from the massive moving parts and complex architecture in range-extender. And the support vector machine (SVM) and other algorithms would lead to the destruction of the natural structure and the correlation in the original data. In order to solve the above problems, Support Higher-order Tensor Machine (SHTM) is proposed. In order to verify the feasibility of SHTM in the fault diagnosis of electric vehicle extender, firstly, the GT-Crank lattice dynamics model was used to establish the samples of normal, single cylinder fire failure and misalignment failure, and then SHTM model was used to classify the samples. The results show that SHTM makes full use of the structural information and correlation in the engine state parameters, the test accuracy is high, the learning time is short, and the convergence speed is faster.