基于机器学习的设备故障预测分析方法

数据准备Data preparation——数据处理Mergingdata sources——特征工程Featureengineering:lagfeature,static feature——建模Modeling:Bin-class, regression,multi-class——训练、仿真Training,Simulation——决策DecisionBinaryclassificationfor predictivemaintenance:to predict theprobabilitythatanequipmentwill failwithin a futuretime period.Regressionforpredictivemaintenance:tocompute theremaininguseful life(RUL) ofanassetMulti-classclassificationforpredictivemaintenance:to assignanasset to one of themultiplepossibleperiodsof timetogivearangeoftime to failureforeach asset,and toidentifythe likelihoodoffailureina futureperiod due tooneofthemultiplerootcauses.