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（Decision）

++Binaryclassificationfor predictivemaintenance:to predict theprobabilitythatanequipmentwill failwithin a futuretime period.

++Regressionforpredictivemaintenance:tocompute theremaining

useful life(RUL) ofanasset

++Multi-classclassificationforpredictivemaintenance:to assignanasset to one of themultiplepossibleperiodsof timetogivearangeoftime to failureforeach asset,and toidentifythe likelihoodoffailureina futureperiod due tooneofthemultiplerootcauses.