In the process of rare earth molten salt electrolysis,the working relationship between the automatic auxiliary components of electrolysis feeding is large,the faults are complex and diverse，and the effect of using a single fault diagnosis method is not ideal.To solve this problem,a fault diagnosis method for powder blanking equipment based on fault tree and LSTM-SVM was proposed by analyzing the working relationship between the components of the feeding automatic auxiliary machine.A multi-layer fault tree was built,the fault modes were analyzed,then the fault modes with high importance were extracted according to the fault tree data,a long-term and short-term memory neural network fault diagnosis model was established,the diagnosis results were output according to the weight size defined by the fault tree analysis results after fault location,and SVM was used to grade the non-fault abnormal working state.The test results show that the model has a high accuracy of fault identification.