Research And Implementation Of Fan Intelligent Diagnosis System Based On Improved Fireworks Algorithm

The progress of modern industry is closely related to the development of information technology,which is becoming more intelligent.Large industrial equipment often goes through continuous operation without shutdown for a long time.During operation,there may be collision between equipment,or friction between parts of equipment.Even the operation environment of equipment may corrode the parts of equipment,resulting in hidden trouble.In the long run,major faults will occur,resulting in huge economy loss and casualties of staff.If you know the running state of the parts at any time,you can get the wear degree of the parts.Before major failure,you can replace the worn parts to minimize the loss and save the cost.In recent years,with the rapid development of machine learning and artificial intelligence,it has been gradually applied in various fields.The combination of machine learning and traditional fault diagnosis can not only remove the hidden trouble,but also use machine instead of human troubleshooting,which can save labor costs,increase the safety and reduce the difficulty of equipment maintenance.It has gradually become a hot research direction.In view of the limitations of the existing fault diagnosis system,this paper designs an intelligent fan diagnosis system which improves the fireworks algorithm(ImFWA).Because of the shortcomings of the existing BP neural network,such as slow convergence speed,this paper proposes an improved fireworks algorithm to optimize the BP neural network model,using intelligent optimization algorithm to calculate the better weight and threshold,and improve the performance of BP neural network.Aiming at the problem that fireworks algorithm is easy to fall into the local optimum to some extent,this paper uses the improved Gauss density function to optimize the enhanced fireworks algorithm,avoiding the false image that original algorithm is easy to cause the intelligent accelerated convergence,enhance the randomness and accuracy of spark.At the same time,the system includes five modules,which are login registration module,fan fault diagnosis system managementmodule,data processing module,fan fault diagnosis model building module,fan fault diagnosis module,and provide a large number of fan bearing data as the data set of training fault diagnosis model.Data preprocessing and fault diagnosis model building are the core functions of the system.The algorithm used in the data preprocessing function is the maximum correlation kurtosis deconvolution method.Four features in the time domain are selected as the final feature selection results.The preprocessed eigenvector is used as the training sample set of the fault diagnosis model.In the fault diagnosis model building module,the improved BP neural network algorithm is used to train the fault diagnosis model.The improved BP neural network model greatly improves the accuracy of fault classification.Through the test,the system can accurately carry out fault diagnosis on the bearing data of the fan,and meet the user's requirements in function and performance.Compared with the traditional BP neural network and RBF neural network,the improved fireworks algorithm has higher accuracy.