一种自适应的控制棒驱动机构滚轮丝杠副剩余寿命预测方法研究

针对反应堆控制棒驱动机构（CRDM）滚轮丝杠副剩余使用寿命（RUL）预测中如何选取有效的健康状态指标和合理构建预测模型的难题，提出了一种新的滚轮丝杠副RUL预测模型。采用基于生成拓扑映射算法（GTM）的负对数似然概率（NLLP）指标作为滚轮丝杠副的健康状态指标，利用K均值聚类算法对NLLP指标进行状态划分。利用历史数据和在线监测数据构建基于Markov模型和最小均方算法（LMS）的自适应预测模型，根据设定的阈值预测得出剩余寿命。通过实验验证，结果表明本文选取的健康状态指标能够有效地反映设备状态，所给出的自适应预测模型比一般的预测模型的预测精度高，为合理构建RUL预测模型提供了依据。 

Research on an Adaptive Method for Predicting the Remaining Life of the Roller Screw Pair of the Control Rod Drive Mechanism

Aiming at the problem of how to select effective health indicators and reasonably construct the prediction model in the remaining useful life (RUL) prediction of roller screw pair of reactor control rod drive mechanism (CRDM), a new RUL prediction model of roller screw pair is proposed. The negative logarithmic likelihood probability (NLLP) index based on the generated topology mapping algorithm (GTM) is used as the health index of the roller screw pair, and the K-means clustering algorithm is used to divide the NLLP index. Using historical data and online monitoring data, an adaptive prediction model based on Markov model and least mean square algorithm (LMS) is constructed, and the remaining life is predicted according to the set threshold. Through experimental verification, the results show that the health status indicators selected in this paper can effectively reflect the equipment status. The prediction accuracy of the proposed adaptive prediction model is higher than that of the general prediction model, which provides a basis for the reasonable construction of RUL prediction model.