Research On Fault Diagnosis Method Of Marine Auxiliary Boiler Based On SOM-BP Neural Network

As one of the important parts of the vessel,the marine auxiliary boiler,its safe and stable operation is of great significance to the ship’s economy and safety.Due to the numerous components of the auxiliary boiler system,the complex structure,the harsh operating environment,the difficulty of obtaining samples,and the many fault characteristic parameters,it is difficult to implement comprehensive and effective fault detection and diagnosis for the boiler nowadays.Therefore,exploring a stable and efficient method suitable for auxiliary boiler fault diagnosis is of great significance to ensure the safety of life and property at sea.In recent years,diagnosis methods based on neural networks have sprung up in the field of fault diagnosis,providing a new idea for auxiliary boiler fault diagnosis.On the basis of analyzing the composition and function of auxiliary boiler system,working process,summarizing common failure phenomena and analyzing failure causes,comprehensively considering the actual requirements and operating characteristics of auxiliary boiler fault diagnosis,Self-Organizing feature Map（SOM）neural network and Back Propagation（BP）neural network constitute a hybrid neural network auxiliary boiler fault diagnosis method is proposed.This article focuses on the Alfa Laval MISSION^TM marine D-type water tube boiler,and the DMS-VLCC large tanker simulator is used as the experimental platform.The platform is used to simulate the working process of the 10 operating conditions required for the experiment and extract the experimental data.Use IBM’s SPSS Statistics software to preprocess the extracted data,and the data will be used in subsequent diagnostic experiments.This paper establishes a fault diagnosis model based on a hybrid neural network,uses processed sample data to perform simulation experiments on the model,and analyzes the results of fault diagnosis.The results show that the hybrid neural network diagnosis model has achieved certain effects,but it still has the shortcomings of low network convergence efficiency and low diagnostic accuracy.The performance of the hybrid neural network model is easily affected by the initial connection weight and threshold of the network.The diagnosis method based on Particle Swarm Optimization（PSO）to optimize the hybrid neural network and the diagnosis method based on Differential Evolution（DE）algorithm to optimize the hybrid neural network are proposed.With the help of the optimization ability of the optimization algorithm,the weight threshold is configured for the model.It aims to improve the accuracy of fault diagnosis and network convergence efficiency of this model.The corresponding diagnosis model was constructed and the simulation experiment was carried out.The experimental results show that it is feasible to use the hybrid neural network fault diagnosis method for the nine kinds of faults studied in this paper for the marine auxiliary boiler system,and the effectiveness of the two optimization models is verified at the same time.Both optimization models can realize the fault diagnosis of the marine auxiliary boiler,and the accuracy and convergence efficiency of the diagnosis are improved compared with a single hybrid neural network.By comparing the three diagnosis methods,the results show that the model diagnosis optimized by the differential evolution algorithm has a higher accuracy and efficiency,and is suitable for the monitoring and diagnosis of marine auxiliary boiler faults.To a certain extent,it guarantees the safe operation of marine auxiliary boilers,and has certain theoretical research significance and engineering application value.