Ayman PredictiveMaintenance

February 8, 2016

467 views
186 downloads

Algorithms

Report Abuse
Simple and Generic training model and experiment for Predictive Maintenance (Preventative Maintenance)
Simple and Generic training model and experiment for Predictive Maintenance (Preventative Maintenance). The training model is supervised model as features are engineered to drive the learning using a generic Stress Factor number (0:100) that represents the operational driver for failure or healthy state. A Planned maintenance state is denoted with **StressFactor**=0. Predictions can be obtained for a State attribute created in the model using R Module as a result of the relationship between StressFactor+StatusType+Cause field values and a Positive **DownTime** Value The Model uses Multiclass Logistic Regression algorithm to find out three probable value predictions for the State Where -1=Planned maintenance, 1=Healthy State and -2=Failure State prediction **Sample Data Format**:- AssetName,StressFactor,StatusType,Cause,Year,Month,DayOfYear,Hour,DownTime Cisco Platinum AC Power Supply,0,88,0,2011,1,1,10,381