Effect of deep learning prediction accuracy on sensor fault detection

摘要：

　　A number of sensor fault detection techniques have been implemented for detecting defects in civil infrastructures instead of onsite human inspections in structural health monitoring.However,the methods based on deep learning are rarely used in sensor fault detection and structural health monitoring.In order to consider the effect of deep learning prediction efficiency on sensor fault detection,this paper presents a deep learning-based method with a new architecture of Stateful Long Short Term Memory Neural Networks(S-LSTM NN)for detecting sensor fault without going into details of the fault features.The prediction efficiency of the proposed S-LSTM NN is compared with the traditional network(LSTMs),in addition study the effect of deep learning prediction efficiency on sensor fault detection.The detection of four types of sensor faults are studied in this paper.Non-stationary acceleration responses of a three-span continuous bridge are researched.The response residuals between the true value and the predicted value of the deep S-LSTM network was statistically analyzed to determine the fault threshold of sensor.Numerical study indicated that the proposed S-LSTM NN shows better detection performance than the LSTMs.

关键词： structural health monitoring sensor fault Long Short-Term Memory Networks prediction accuracy deep learning

作者: Lili Li Gang Liu Liangliang Zhang Qing Li

作者单位: School of Civil Engineering,Chongqing University,Chongqing 400045,China;The Key Laboratory of New Te College of Computer Science,Chongqing University,Chongqing 400045,China

会议类型: 国际会议

会议名称: The 7th World Conference on Structural Control and Monitoring(7WCSCM)(第七届结构控制与监测世界大会)

会议地点: 青岛

会议语种:英文

页码: 1940-1950

在线出版日期: 2018-07-22（万方平台首次上网日期，不代表论文的发表时间）