Fault Diagnosis of Fire Hydraulic System based on 1DCNN-Bayesian Optimization

Abstract

To solve the problem of low efficiency and scarce fault data caused by the traditional fault diagnosis of fire hydraulic system relying on manual experience, this paper proposes an intelligent diagnosis method integrating simulation modeling and 1DCNN-Bayesian optimization. Methods: Firstly, a fire hydraulic system model was built based on Simscape, and the effectiveness of the model was verified by comparing the performance curve of the fire pump. Secondly, a 1DCNN model was constructed, and the Bayesian optimization algorithm was introduced to automatically search for the optimal network structure. Results: The root mean square error of the head-flow performance curve of the fire pump was 0.13m, and the root mean square error of the effective shaft power-flow performance curve was 13.05w, indicating the data effectiveness of the simulation model; the accuracy of the 1DCNN model after Bayesian optimization reached 98.51%, which was 1.50% higher than the original model. The optimal configuration was: three-layer convolution (kernel size [20,5,3], number of filters [32,64,128]), Dropout parameter was 0.4643, optimizer type was Rmsprop, learning rate was 0.0023, and batch size was 32. Conclusion: This method generates fault data through simulation to make up for the lack of samples, combines Bayesian optimization to improve the model's adaptability, enhances the accuracy of fault diagnosis, and provides technical support for the intelligent operation and maintenance of fire hydraulic systems.