基于DAE-MD的电源车柴油机健康状态量化评估方法

DAE-MD-based quantitative evaluation method of health state of diesel engine of power supply vehicle

针对柴油机健康状态的综合评估问题,结合降噪自动编码机(DAE)深度网络与马氏距离(MD)算法,提出一种基于深度学习与距离融合的电源车柴油机健康状态量化评估方法.该方法首先通过DAE深度网络提取柴油机不同健康状态下的特征序列,建立柴油机健康状态识别模型;为进一步量化评价柴油机的健康水平,又将MD算法引入,度量DAE模型输出状态与正常状态特征序列之间的距离,并将其归一化为健康指数,给出柴油机健康状态的量化评估结果.最后,借助于120 kW电源车模型仿真系统,验证了所提方法的有效性.文中方法通过DAE的分类和MD的计算实现了电源车柴油机健康状态定性与定量评估的融合,为其开展视情预防维护提供了依据.

Aimed at the comprehensive evaluation problem of the diesel engine health status and combined with deep network of noise reduction auto-encoder(DAE)and Mahalanobis distance(MD)algorithm,a method for quantifying health status of diesel engine is proposed in this article based on deep learning and distance fusing.In this method,the feature sequence of the diesel engine with different health conditions is extracted by means of DAE deep network to establish diesel engine health status recognition model,firstly.Then,in order to further quantify the health level of the diesel engine,the MD algorithm is introduced to measure the distance from output states to normal state of DAE model in feature sequence and this distance is normalized as a health index,so that the result of the quantitative assessment of health status of diesel engine is given.Finally,with the help of the simulation system of an 120 kW power supply vehicle model,the effectiveness of the proposed method is verified.By using the method in this paper,the fusion of qualitative with quantitative assessment of health status of diesel engine of power supply vehicle is implemented by means of DAE classification and MD calculation,and a basis for developing situation-based preventive maintenance is provided.