基于机器学习的压力传感器故障诊断及剩余寿命预测模型研究  

作　　者：白同元 黄瀚宇 林琳淳 钟晓莹 赵志红[1] 

机构地区：[1]北京理工大学珠海学院,广东珠海519088

出　　处：《科技创新与应用》2021年第25期32-34,37,共4页Technology Innovation and Application

基　　金：2020年国家级大学生创新创业训练计划项目(编号:202013675010);广东大学生科技创新培育专项资金资助项目(编号:pdjh2020a0748);校级重点科研机构(编号:XJ-2018-05)。

摘　　要：压力传感器在各大工厂的生产线上大量应用,由于工作环境高温,存在粉尘、水汽等情况,会导致压力传感器寿命缩短,设备出现故障。针对此问题文章以压力传感器运行数据为研究对象,通过对数据处理后,分别建立时间序列预测模型和LSTM模型,设定数据分类判定准则,对比模型的预测准确率,选择最优的模型对未来设备运行状态进行预判,并进行剩余寿命预测。试验结果表明:为了对模型精度进行计算,将AR(7,0,0)作为拟合模型,该模型能够准确地预测传感器未来的运行趋势且模型的准确率达到96.14%,说明模型具有较高的可靠性,预测结果较准确,计算得到压力传感器运行周期为412h,并在265h附近开始退化。Pressure sensors are widely used in the production lines of major factories.Due to the high temperature of the working environment,the existence of dust,water vapor and other conditions,the life of the pressure sensor will be shortened,resulting in equipment failure.In order to solve this problem,taking the running data of the pressure sensor as the research object,after processing the data,the time series prediction model and the LSTM model are established respectively,the data classification criteria are set,the prediction accuracy of the model is compared,the optimal model is selected to predict the future running state of the equipment,and the remaining life is predicted.The test results show that:through the accuracy of the model,AR(7,0,0)is used as the fitting model,the model can accurately predict the future operation trend of the sensor,and the accuracy of the model is 96.14%,indicating that the model has high reliability and the prediction result is more accurate.It is calculated that the operating period of the pressure sensor is 412h and begins to degenerate around 265h.

关 键 词：故障诊断 时间序列 分类准则 寿命预测 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]