基于PCA和随机森林的故障趋势预测方法研究

Study on the fault trend prediction method based on PCA and random forest

摘要:

故障预测和健康管理技术(PHM)在现代工程系统中能够在系统具备较高复杂度的情况下,有效保障其可靠性和安全性。在机械故障诊断中对于采集到的原始数据的高维特征量的处理较为复杂,并且在实际应用中趋势预测的精度要求较高,针对该问题提出一种基于主成分分析(PCA)与随机森林算法的轴承故障趋势预测方法。该方法利用PCA对提取的原始轴承数据特征量进行线性降维,并选取其中主成分特征量,输出非线性时间序列数据。原始数据经过PCA处理得到非线性时间序列,将该序列作为随机森林算法的输入进行故障趋势预测,并把预测结果与BP神经网络模型预测的结果进行对比,结果表明随机森林在故障趋势预测上在精度相较于BP神经网络有显著提高,是一种有效的故障趋势预测方法。

Abstract:

Fault prediction and health management system(PHM) can effectively guarantee the reliability and safety of modern engineering system under the condition of high complexity. In mechanical fault diagnosis with high dimensional characteristic quantity of raw data collected for the more complex, and in the practical application trend prediction precision, this paper presents a method based on principal component analysis (PCA) method to predict bearing fault trend and random forests algorithm. The method uses PCA to reduce the characteristic data of the original bearing data, and selects the principal component characteristic quantity to output the nonlinear time series data. The original data are processed by PCA nonlinear time series, the sequence as a random forest algorithm input fault trend prediction, and the prediction results were compared with the BP neural network model prediction results, results show that the random forest in the fault trend prediction in precision compared with the BP neural network is improved, is a method of forecasting an effective fault trend.