基于大数据的液压支架电液控制系统故障诊断

Fault diagnosis for electro-hydraulic control system of hydraulic support based on big data

针对液压支架电液控制系统故障人工排查方式无法准确定位某些随机故障或个别系统故障的问题,对传统电液控制系统硬件设备进行智能化改造:增加了对系统核心关键部件电气参数的采集传输功能;从大数据采集、传输、处理等方面,阐述了基于Hadoop的大数据决策分析服务平台的构建;设计了大数据故障诊断引擎,以并行算法为核心对各类故障进行识别和诊断,基于MapReduce对C4.5决策树分类算法进行改进,并通过后剪枝技术解决算法容易过度拟合且不稳定的问题,通过多分类器融合技术提高算法准确性。测试结果表明,通过C4.5决策树分类预测引擎提取的电磁先导阀、控制器、压力传感器及行程传感器故障特征曲线存在较大差异性,通过动态比较匹配,依据故障特征曲线变化规律可识别出故障类型。

In view of problem that manual trouble shooting method of electro-hydraulic control system of hydraulic support cannot accurately locate certain random faults or individual system faults,hardware equipment of traditional electro-hydraulic control system was transformed by intelligent technique:Collection and transmission function of the electrical parameters of key components of the system was added;Construction of big data decision analysis service platform based on Hadoop was expounded from aspects of big data collection,transmission and processing;Big data fault diagnosis engine was designed which used parallel algorithm as the core to identify and diagnose various faults.Based on MapReduce,C4.5 decision tree classification algorithm was improved,the post-pruning technique was used to solve the problem of instability and being easy to overfit of the algorithm,and multi-classifier fusion technology was used to improve accuracy of the algorithm.The test results show that fault characteristic curves of electromagnetic pilot valve,controller,pressure sensor and stroke sensor extracted by C4.5 decision tree classification prediction engine have great differences,and through dynamic comparison and matching,fault type can be identified according to the change law of the fault characteristic curves.