The Research On Noise Prediction Based On IPSO-BP Neural Network And Matching Of Centrifugal Fan

High noise is a major problem in the use of centrifugal fans,which seriously affects people’s comfort in life,and excessive noise will cause centrifugal fans to be returned to the factory for recasting,increasing manufacturing costs and production cycles.In order to realize the noise reduction design of the centrifugal fan,the fan factory must estimate the fan noise before producing the centrifugal fan.Precise prediction of fan noise before fan production is of great significance to protect the physical and mental health of users and improve the economic benefits of enterprises.The paper comprehensively analyzes the current research status at home and abroad,selects the factors affecting the noise of the centrifugal fan,and optimizes the BP neural network prediction model through dimensionality reduction and combined with the improved particle swarm algorithm to complete the accurate prediction and matching selection of the centrifugal fan noise.The main research contents and conclusions are as follows:(1)Comprehensively analyze the influence of centrifugal fan geometric parameters and performance parameters on noise,use stepwise regression analysis to correct collinearity to screen out 10 centrifugal fan noise influencing factors,and use principal component analysis to reduce dimensionality to construct a construction consisting of 4 principal components as The input noise prediction model of the centrifugal fan.Due to the small data sample of the centrifugal fan noise prediction problem,and the strong non-linear mapping ability and self-learning and self-adapting ability of the BP neural network model,the prediction error is only 1.5%,which is better than the linear regression prediction model and the long short-term memory network prediction model.Finally,the BP neural network model is selected to predict the noise of the centrifugal fan.(2)This paper proposes an improved non-linear weighted particle swarm algorithm by modifying the inertia weight of the particle swarm algorithm.After 5 test functions,it is concluded that the improved particle swarm algorithm has stronger global search ability and convergence speed.Faster.(3)This paper uses principal component analysis combined with improved particle swarm algorithm to optimize the BP neural network model to complete the prediction of the centrifugal fan noise.In order to verify the excellence of the model,the model is combined with the PCA-BP neural network and the PSO-BP neural network.Compare.The final result shows that the prediction error of the PCA-IPSO-BP noise prediction model used in this paper is only 1.12%,which meets the actual accuracy requirements and has strong practical value.(4)This paper proposes a fan matching method based on dimensionless parameters and cluster analysis to solve the problem of time-consuming,laborious and low efficiency in traditional fan matching selection.Calculate the flow coefficient,total pressure coefficient and specific speed through the flow rate,total pressure,speed and diameter provided by the customer,calculate the correlation coefficient between the dimensionless parameters and the dimensionless parameters of the existing fans in the fan library,and match the top four correlations As the fan to be selected,the most suitable fan model is finally selected.If some parameters need to be modified,some parameter values should be reasonably modified and put into the neural network prediction model for centrifugal fan noise prediction.In this paper,the research content is designed as a fan matching and noise prediction system based on neural network,and it has been well practiced in manufacturing enterprises,which has better improved order delivery and product quality.