基于相似样本特征提取的装备性能退化研究

为了实时监控装备的性能变化规律, 对可能出现的突发情况进行预测, 本文首先考虑工程中装备在线监控预警系统中历史样本数据来源广泛、数据形式不规范问题, 基于B样条插值算法对历史数据进行了规整。其次, 针对性能退化研究中存在的随机性误差、性能指标变化不规律问题, 基于自组织映射和堆栈自编码器对相似样本集的退化特征进行提取与重构, 提出了基于最小特征圆的指标构造方法, 再采用双指数模型对装备的性能退化规律和寿命进行预测。最后, 使用仿真数据对模型的正确性进行验证。结果表明, 本模型和方法可以有效预测装备的性能退化规律, 并能体现装备遭受冲击时性能的退化与恢复规律。

Research on equipment performance degradation based on feature extraction of similar samples

In order to monitor the performance change rule of equipment in real time and predict the possible emergencies, this paper firstly considers the problems of extensive data sources and non-standard data forms in the online monitoring and warning system of engineering equipment, and then organizes the historical data based on the B-spline interpolation algorithm. Secondly, in view of the random error and irregular change of performance indicators in performance degradation, this paper extracts and reconstructs degradation characteristics of similar sample sets based on self-organizing mapping and stacked autoencoder. An index construction method based on minimum characteristic circle is proposed. The double exponential model is adopted to predict the performance degradation law and life of the equipment. Finally, the simulation data are used to validate the correctness of the model. The results show that the model and method can effectively predict the performance degradation law of equipment, and can reflect the performance degradation and recovery law of equipment under impact.