Multi-objective HDD Failure Prediction Method Based on AE-LSTM
Hard disk drive(HDD)failure prediction is used to avoid data loss or service interruption,which sends a warning be-fore HDD failures occur,it improves the reliability and security of data center.However,most HDD failure prediction models convert HHD failures into binary classification tasks,ignoring the gradual deterioration of HDD and lacking of fault diagnosis function.Therefore,an HDD failure prediction method based on auto encoder and long short term memory(AE-LSTM)is proposed to achieve the HDD multi-objective tasks of health status multi-classification,remaining useful life(RUL)prediction,and fault diagnosis.The regression decision tree model is used to intelligently label the HDD health status.Then,the robust hidden variables are extracted through the AE-LSTM model,the RUL model and HDD health status classification model are built.The HDD fault diagnosis is im-plemented by computing the difference between the input and output of the AE module.By evaluating the random forest(RF),LSTM and AE-LSTM algorithms on the Backblaze public dataset,the experimental results show that the AE-LSTM algorithm has the effectiveness and advantages in multi-objective HDD failure prediction.
hard drive failure predictionhard drive fault diagnosisremaining useful lifeLSTMAE
万方数据
维普
