点击次数：

论文类型：会议论文

发表时间：2016-12-06

收录刊物：EI

页面范围：985-990

摘要：Independent Component Analysis (ICA)-based statistical monitoring has the potential for sensor fault detection prior to Structural Health Monitoring (SHM). There are, however, two major disadvantages of traditional ICA: the ignorance of dynamic property and the problem of dimensionality reduction. This paper proposes a Dynamic Independent Component Analysis (DICA) methodology to address these two disadvantages. Based on the proposed DICA, two fault detection statistics are defined for detecting the potential sensor faults occurred in SHM systems. A case study using benchmark structure is considered to verify the availability and effectiveness of the proposed method. The fault detection results demonstrate that the consideration of the dynamic property in DICA enhances the fault detection ability and performance compared with the traditional ICA. © 2017 Taylor & Francis Group, London.