摘要
利用斯皮尔曼秩相关系数(SRCC)、贝叶斯(Bayesian)、k最近邻(KNN)算法提出了一种新的航空发动机剩余使用寿命预测方法。为解决关键特征提取不足问题,首先,利用SRCC方法对发动机的历史多元监测特征进行筛选,提取出衰退性能趋势明显的监测特征作为预测模型的输入;其次,构建了基于欧式距离的k最近邻回归预测模型,利用贝叶斯更新公式对KNN中的超参数模型进行训练,求解目标函数并返回训练模型最优超参数值与最小均方根误差;最后,推导航空发动机剩余使用寿命(RUL)概率密度函数解析式,得到发动机RUL预测结果。采用CMAPSS数据集验证所提方法的有效性,结果表明,与其他预测方法相比该方法具有更优的预测性能,有效提升了发动机RUL预测的精确度。
This paper proposes a new method for predicting the remaining useful life of aircraft engines using the Spearman’s rank correlation coefficient(SRCC),Bayesian,and k-nearest neighbors(KNN)algorithms.Firstly,to address the issue of inadequate feature extraction,the SRCC method is used to select historical multidimensional monitoring features of the engine,extracting monitoring features with significant degradation performance trends as inputs for the prediction model.Secondly,a KNN regression model based on Euclidean distance is constructed.The Bayesian optimization algorithm is utilized to train the KNN model with multiple parameters,solving the objective function and returning the optimal parameter values and minimum root mean square error of the trained model.Lastly,the probability density function of the aircraft engine’s remaining useful life(RUL)is derived to obtain the RUL prediction results.Using the CMAPSS dataset to validate the effectiveness of the proposed method.Experimental results demonstrate that compared to other prediction methods,the proposed method in this paper exhibits superior predictive performance and effectively improves the accuracy of aircraft engine RUL prediction.
作者
李东君
王海瑞
李亚
朱贵富
LI Dongjun;WANG Hairui;LI Ya;ZHU Guifu(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650504,China;Information Construction Management Center,Kunming University of Science and Technology,Kunming 650504,China)
出处
《陕西理工大学学报（自然科学版）》
2024年第6期27-35,共9页
Journal of Shaanxi University of Technology:Natural Science Edition
基金
国家自然科学基金项目(61863016)。
