多传感器和故障率隐半马尔可夫模型的剩余寿命预测方法

Method for Residual life Prediction Using Multiple Sensors/Failure Rates and Hidden Semi Markov Model

为解决复杂设备剩余寿命预测精度不高的问题,提出了一种基于多传感器和故障率隐半马尔可夫模型的剩余寿命预测方法。使用极大似然线性回归变换方法表示多传感器之间的差异,通过故障率与健康状态转换矩阵相结合的方法,建立多传感器和故障率隐半马尔可夫模型预测有效剩余寿命。实验结果表明,通过提出的故障率方程,利用某火炮炮管的多传感器历史监测数据,炮管的实际剩余寿命与预测剩余寿命的平均相对误差可以降低至6.6852%,提高炮管剩余寿命预测精度1.3%左右。

In order to improve the prediction accuracy of the remaining life of complicated equipment,the paper presents a method for predicting remaining life by using the Hidden Semi Markov Model as well as multiple sensors and failure rates.The maximum likelihood linear regression transform is used to represent the differences between multiple sensors.Through combining the failure rate with health state transition matrix,the Hidden Semi Markov Model with multiple sensors and failure rate is established to predict the effective remaining life.The experimental results show that the average relative error between the actual remaining life and the predicted remaining life of the gun barrel can be reduced to 6.6852%by using the proposed failure rate equation and historical monitoring data of the multiple sensors.The prediction accuracy is improved by about 1.3%.