Research On Compressor Fault Detection Algorithm

Compressor fault detection is an important part of fault diagnosis technology,and has received extensive attention in the field of compressor technology research.Since compressor failure will cause economic damage and threaten people's lives and property,compressor fault detection has been developed Technology is very necessary.Compressor fault detection uses the signals generated by the compressor during operation,and uses the related signals to find the faulty compressor.The traditional method is to use the detection method of BP neural network to predict the type of compressor failure using the recognition features extracted from the compressor voice signal.This method has good feasibility,but the BP neural network has the problem of falling into the local optimal value,Which leads to the problems of poor stability of prediction results and low prediction accuracy.For the problems of BP neural network in compressor fault detection,this paper has done the following research:First,this paper proposes the compressor fault detection algorithm of IBFOA and BP neural network.The improved bacterial foraging algorithm is used to deeply optimize the weights and thresholds of the BP neural network,thereby improving the accuracy and prediction of the BP neural network.stability.Firstly,the initial bacterial flora is generated by the method of chaotic sequence,so as to ensure that the individuals are more evenly distributed in the search range,and secondly,the replication operation and the migration operation are combined,and the migration operation and the chaotic sequence method are adopted for half of the dead individuals in the replication operation.Adaptive step size,which effectively strengthens the ability of global search.In the later period,the optimal direction is selected.The step size is related to the amount of change in the adaptation value.Half of the eliminated individuals in the copy operation are regenerated near the better individual,thus The local development capability will be effectively strengthened,so that the precision of the optimization results will be higher,and finally the improved bacterial foraging algorithm and BP neural network model will be applied to the compressor fault detection.Secondly,this paper proposes the compressor fault detection algorithm of IWOA and BP neural network.First,by optimizing the probability that the searched prey in the whale optimization algorithm is executed and the process of gathering the prey to the optimal individual,thereby improving the global search ability and convergence speed of the whale optimization algorithm,and secondly,the improved whale optimization algorithm is applied to the BP neural network.The weights and thresholds are optimized in depth to decode and assign the optimal whale individual to the BP neural network,thereby improving the accuracy of the BP neural network prediction results.Finally,the model combining the improved optimization algorithm and BP neural network,through the compressor fault detection experiment,compared with the model in other literature,reuses the measurement indicators such as test error,training error and iteration number to prove the improved optimization algorithm And BP neural network has the advantages of high prediction accuracy and good stability.