矿用电动轮自卸卡车制动器异常检测

Abnormal detection for brake of mining electric wheel dump-trucks

针对矿用电动轮自卸卡车制动器异常工况检测需求,设计了制动器状态采集系统,并基于长短时记忆(LSTM)网络实现了监测数据的特征提取及异常检测.制动器状态采集系统主要由车载监控端、云服务端、用户端构成,实现制动器的数据采集、数据传输、异常预警等.基于监测数据,采用长短时记忆网络构建了异常工况检测模型,对矿用自卸卡车实况作业产生的多元时间序列数据进行特征提取.模型将早期输入序列信息传播到较后的记忆单元中,有效解决了时序数据的长期依赖性问题.实验结果表明,所提方法对异常工况的识别准确率高于93%,明显优于基于阈值的检测方法.

In response to the demand for abnormal working condition detection of the brake system of mining electric wheel dump trucks,a brake status acquisition system was designed,and feature extraction and abnormal detection of monitoring data were achieved based on a long short-term memory network.The brake status acquisition system is mainly composed of an onboard monitoring terminal,a cloud service terminal,and a user terminal,which realizes data acquisition,transmission,and abnormal warning of the brake system.Based on the monitoring data,a long short-term memory network was used to construct an abnormal working condition detection model,which extracted features from the multi-dimensional time series data generated during the actual operation of mining trucks.The model propagates early input sequence information to later memory units,effectively solving the long-term dependency problem of time series data.Experimental results show that the proposed method has an accuracy rate of abnormal condition recognition of over 93%,which is significantly better than threshold-based detection methods.