| The engine cooling fan of commercial vehicles plays a vital role in engine cooling.In mechanical engineering,it is necessary to pursue cost and quality in order to enhance the core competitive power of an enterprise.For the processing and production of fan parts,the traditional design mode is based on the experience of engineers,iterative design-analysis-verification,until a satisfactory design scheme is obtained,with high design cost and waste of resources,It does not conform to the current development trend of intelligent manufacturing.A new researchidea is proposed in this paper.On the basis of the existing automatic modeling and simulation tools,the automatic simulation template is designed by programming,and the parametric model of fan scheme design is combined.On this basis,the mapping data set between fan structure scheme parameters and performance indicators is obtained through a large number of simulations,and then the sensitivity analysis of relevant parameters is carried out based on the range method,The influence of the key parameters of the fan on several important performance indicators of the fan is analyzed.The approximate response surface of the fan performance indicator mapping model is fitted by BP neural network.Based on the response surface,the multi-objective optimization algorithm is used to quickly search the optimal solution.At the same time,the accuracy is ensured and the optimal scheme is found quickly,thus the design efficiency is improved.In this paper,the sensitivity analysis of parameters is analyzed by regression analysis.Change each parameter or several parameters jointly at a time,and analyze the individual or joint impact of each parameter on fan performance.Determine the parameters that greatly affect the fan performance,and analyze the parameter sensitivity through the range method.Then,the paper analyses the feasibility and advantages of the cooling fan,and introduces the neural model and its algorithm,select the appropriate input and output variables,use the simulation data to bring into the neural network for training,calculate the mean square error below 10-2,reflecting the good generalization ability of the neural network.At the same time,the relationship between the number of neurons in the hidden layer of neural network and the performance of neural network is discussed.Thirdly,the parameters of the BP neural network are optimized by genetic algorithm,the influence of the randomness of the initial weight and threshold of the neural network on the prediction results is solved,the structural performance of the original network is improved,and the prediction results of the standard,and improve the structural performance of the original network.In contrast to the prediction results of the standard and modified BP neural network,the average absolute error percentage calculated by the standard BP neural network is 1.15%,The average absolute error percentage of BP neural network optimized by genetic algorithm is 0.74%,the error is reduced,the prediction accuracy is further improved.Finally,based on the response surface,the optimal solution set is found through multi-objective optimization,reducing the search range of design variables,converting the design variables corresponding to the optimal solution of the cascade into the initial design variables of the impeller,further optimizing the impeller design scheme on the automatic simulation platform,obtaining several sets of optimal schemes,and verifying the optimal solution through detailed simulation and bench tests.Compared with the original fan,under the design point operating conditions,Power consumption decreased by 17.6%;The static pressure has increased by 12.7%,the noise has decreased by 3.6 d B,and the static pressure efficiency has increased by 48.6%.It is verified that the multi-objective optimization method proposed in this paper can be effectively applied to the performance optimization of commercial vehicle engine cooling fans. |
