Fault diagnosis is guarantee of modern industrial processes. However, most of the existing fault diagnosis methods are based on a single-mode process and are not adapted to multimode processes. To solve the problem that the fault is difficult to diagnose in the multimode process and improve the efficiency of the fault diagnosis, this paper proposes an ensemble fault diagnosis approach based on variational modal decomposition (VMD), improved fuzzy c-means clustering for just in time learning (IJITL) and recursive least squares support vector machine(RLSSVM). First VMD is introduced to denoise the collected data, and then IJTL method is used for modal distinction, and finally RLSSVM is designed to achieve fault diagnosis. Experimental simulation is carried out in the process of continuous stirring heater (CSTH). The simulation results show that the ensemble method improves the diagnostic efficiency and performance.