基于PCA-IPSO-INN的离心风机噪声预测  

Noise Prediction of Centrifugal Fans Based on PCA-IPSO-INN

作　　者：张长伟 蒋淑霞[1] 隆波 刘文 刘梦安 ZHANG Changwei;JIANG Shuxia;LONG Bo;LIU Wen;LIU Mengan(School of Mechanical and Electrical Engineering,Central South University of Forestry and Technology,Changsha 410004,China;Hunan Liancheng Rail Equipment Co.,Ltd.,Zhuzhou 412000,Hunan,China)

机构地区：[1]中南林业科技大学机电工程学院,长沙410004 [2]湖南联诚轨道装备有限公司,湖南株洲412000

出　　处：《噪声与振动控制》2022年第2期73-78,84,共7页Noise and Vibration Control

基　　金：湖南省自然科学基金资助项目(2019JJ60076);湖南省科技创新计划重点研发经费资助项目(2018NK2065)。

摘　　要：因离心风机噪声超限造成返厂,会影响用户体验,同时增加制造成本,故提出一种新的分析方法,旨在生产前对离心风机噪声进行智能预测。首先基于相关性分析和主成分分析法(Principal Component Analysis,PCA)对风机15个性能和几何参数进行降维处理,然后提出一种具有非线性惯性权重的粒子群算法,并用于优化反向传播(Back Propagation,BP)神经网络的初始权值和阈值,最后寻找最佳隐层,确定BP神经网络4×8×1的结构,从而简化网络结构,提高收敛速度和预测精度。经实验验证,该方法较其他方法预测精度更高,平均误差仅为0.76%,且在生产企业得到很好应用,在工程实践中具有实际意义。Some centrifugal fans have been returned to the company due to their excessive noise. It will affect user’s experience and increase the manufacturing cost. Therefore, a new analysis method is proposed to intelligently predict the noise of centrifugal fans before production. Firstly, the dimensions of the fan’s fifteen performances and geometric parameters are reduced by using correlation analysis and principal component analysis(PCA). Then a particle swarm optimization algorithm with nonlinear inertia weight is proposed to optimize the initial weights and thresholds of BP neural network. Finally, the optimal number of hidden layers is found. The structure of 4×8×1 BP neural network is determined, the network structure is simplified and the convergence speed and prediction accuracy are improved. The experimental results show that this method has higher prediction accuracy than the other methods, and the average error is only 0.76 %. This method has been well applied in production enterprises and has practical significance in engineering application.

关 键 词：声学 离心风机 主成分分析 改进粒子群算法 噪声预测 改进神经网络 

分 类 号：TH432[机械工程—机械制造及自动化]