基于混沌粒子群优化LS-SVM的发动机磨损状态监测研究

Research on Monitoring of Engine Friction and Wear based on Chaos PSO Algorithm Optimization Parameters of LS-SVM

基于混沌PSO算法优化最小二乘支持向量机实现航空发动机磨损状态监测;通过小波包分解消除润滑油光谱数据的噪声,获取LS-SVM的训练与测试样本;针对最小二乘支持向量机解决大规模数据样本回归问题时所出现的训练时间长、收敛速度慢等缺点,提出了混沌PSO算法优化LS-SVM的模型参数;该方法不仅克服了传统PSO算法早熟、容易陷入局部最小值等缺点,同时显著提高了最小二乘支持向量机的预测能力;最后,将一般LS-SVM和GM(1,1)模型的预测结果与文中预测结果进行对比,该方法构建的模型对测试样本产生的预测误差仅为0.0441,验证了该方法在预测精度上具有明显优势。

Based on chaos PSO algorithm optimization parameters of LS- SVM, monitoring of Aviation--engine friction and wear was realized. Noises in lubricating oil spectrum measurement data were eliminate by Wavelet decomposition, learning and testing samples for LS - SVM were also acquired. Directed towards LS - SVM solving large scale data regression led to long learning time and slow convergence speed, chaos PSO algorithm optimization parameters of LS - SVM was proposed. The disadvantages of earliness and tending to get into local solution in traditional PSO algorithm were overcome by this method. It also remarkable improved forecasting ability of LS-SVM. At last, we compare the results of general LS-SVM and GM （1, 1） forecasting model with the results of this article, the error of the forecasting module based on this method is only 0. 0441, and it proved this method has a transparent superior in forecasting precision.