In order to improve the fault diagnosis ability of hydraulic pump under complex conditions,the health evaluation of hydraulic pump was realized by combining sensor data fusion and cascade forest model.The feature level and decision level fusion technology were used to realize the rapid fusion of sensor information of the piston pump.The random forest model was used to evaluate the importance of the preliminary features and the initial feature parameters with high importance were selected.The cascade forest model was used to classify the health detection results of the hydraulic pump.The results show that the cascades forest model constructed by multi-sensor information fusion method can accurately diagnose the health status of the hydraulic pump.When only 5% training set is set,the health diagnosis result of the hydraulic pump can achieve 99.5% accuracy.When a single temperature feature cannot meet both precision and recall conditions,the prediction accuracy and recall of the combined mode are higher than those of other modes,the combined form of temperature fusion flow has greater advantages.