基于IPSO-BSVM的小样本数据不均衡下的设备健康预测研究

Equipment health prognosis based on IPSO-BSVM under small and imbalanced sample data

针对设备的健康预测缺乏大量样本且存在样本不均衡问题,提出基于改进粒子群优化算法优化均衡支持向量机(IPSO-BSVM)的健康预测模型。首先,提出动态非线性惯性权重对PSO进行优化;其次,提出了一种非线性多分类均衡支持向量机BSVM,以减小由于样本量不均衡引起的误差;然后利用改进后的PSO对BSVM参数进行优化;最后利用建立的IPSO-BSVM模型对设备进行状态识别及剩余寿命预测。仿真结果表明,提出方法能够有效解决小样本数据不均衡下的设备健康预测问题。

In view of the lack of abundant sample data and sample unbalanced problem in equipment health prognosis,this paper proposed a health prognosis model based on IPSO-BSVM.Firstly,it put forward the dynamic non-linear inertia weight to improve the PSO algorithm.Secondly,it introduced the balanced SVM to improve the performance of SVM.Thirdly,it applied the improved PSO algorithm to optimize the parameters of the balanced SVM.Finally,it used the IPSO-BSVM model for equipment state identification and residual useful life prognosis.The simulation results show that the proposed method can effectively solve the problem of equipment health prognosis under small and imbalanced sample data.