基于LSTM的硬盘剩余寿命预测  

Hard Disk Remaining Life Predict Based on LSTM

作　　者：李顺 李君[2] 吴鑫 郎一辉 梅碧舟 LI Shun;LI Jun;WU Xin;LANG Yi-hui;MEI Bi-zhou(Zhejiang Wanli University,Ningbo Zhejiang 315100;Zhejiang Easy Forging Precision Machinery Co.,Ltd.,Xiangshan Zhejiang 315700)

机构地区：[1]浙江万里学院信息与智能工程学院,浙江宁波315100 [2]浙江万里学院,浙江宁波315100 [3]浙江易锻精密机械有限公司,浙江象山315700

出　　处：《浙江万里学院学报》2020年第4期69-77,共9页Journal of Zhejiang Wanli University

基　　金：浙江省公益技术研究项目(2017C31040);宁波市科技厅惠民项目(2017C50028);国家级大学生创新创业训练计划项目(201910876035)。

摘　　要：随着云计算和大数据的迅速发展,硬盘在存储系统中的可靠性和稳定性越来越重要。硬盘故障导致的数据损坏会给企业和用户带来一定的损失。因此,必须提前检测出故障硬盘。文章提出了一种基于深度学习长短期记忆神经网络(Long Short Term Memory,LSTM)的硬盘剩余寿命预测方法。该方法相对于传统的机器学习方法能够捕获硬盘特征的序列信息。实验结果表明:文章建立的LSTM模型可以在训练样本和测试样本上分别达到0.27和1.85的平均绝对误差(Mean Absolute Error,MAE),并且比传统的机器学习方法提前更长的时间检测出故障硬盘。With the rapid development of cloud computing and big data,the reliability and stability of hard disks in storage systems is becoming increasingly important.Data corruption caused by hard disk failure will bring certain losses to enterprises and users.Therefore,a failed hard disk must be detected in advance.This paper proposes a hard disk remaining life prediction method based on deep learning LSTM(Long Short Term Memory)neural networks.Compared with traditional machine learning methods,this method can capture the sequence information of hard disk features.The experimental results show that the LSTM model established in this paper can reach MAE(Mean Absolute Error)of 0.27 and 1.85 on the training samples and test samples,respectively,and detect the failed hard disk earlier than the traditional machine learning method.

关 键 词：硬盘故障 深度学习 长短期记忆 剩余寿命预测 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]