【摘要】 在轨道交通车辆门系统远程监控技术背景下,研究基于KMEANS特征匹配的门系统亚健康预测方法。提取监测数据时域统计特征值,与正常数据的标准特征值同时输入KMEANS分类器当中聚类,根据聚类空间距离排序结果筛选出差异较大的特征集合,以特征值序号和正负号代表特征值及相对变化趋势,并构造特征向量;将特征向量输入规则库当中匹配,根据亚健康类型的频度和置信度大小综合给出预测结果。台架测试验证结果表明:对于训练后的亚健康类型预测准确度很好。KMEANS特征匹配的门系统亚健康系统在广州地铁正线试验运行。试验运行结果表明,该系统不仅能够满足现场状态修要求,而且还能不断学习新亚健康类型,并扩充数据库,使预测结果更准确稳定。更多还原

【Abstract】 According to the remote monitoring technology of urban rail transit door system, The sub-health prediction method for the door system based on KMEANS(k-means clustering algorithm) feature algorithm is studied. Firstly, the time domain statistical features are extracted from the monitoring data, and imported into the KMEANS classifier with the standard statistical features simultaneously, thus a set of distinct features is selected according to the result of KMEANS clustering. The feature vector is formulated based on its serial numbers and relative changing trend. Secondly, the feature vector is matched with the rule base, and the prediction result is outputted on the basis of frequency and confidence level of sub-health category. The verification results of door system show that the sub-health prediction method has higher accuracy, which could meet the requirements of maintenance of Guangzhou Metro door system on the main line test. The prediction will be more accurate and more stable with a period of running and training, together with continuous learning and data base extension.更多还原