Frame Structure Fault Diagnosis Based on a High-Precision Convolution Neural Network

• Published in: Sensors (Basel, Switzerland) Vol. 22; no. 23; p. 9427
• Main Authors: Xue, Yingfang, Cai, Chaozhi, Chi, Yaolei
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 02.12.2022; MDPI
• Subjects: Accuracy; Algorithms; Analysis; anti-noise capability; Civil engineering; Classification; Computational linguistics; convolution neural network; Deep learning; Engineering; Fault diagnosis; Fault lines; Frame structures; Humans; Language processing; Mechanical engineering; Methods; Natural language interfaces; Neural networks; Neural Networks, Computer; Noise; Records; Sensors; Steel; Structural health monitoring; vibration signal; Wavelet transforms; fault diagnosis; structural health monitoring; vibration signal; anti-noise capability; convolution neural network
• ISSN: 1424-8220; 1424-8220
• DOI: 10.3390/s22239427

Structural health monitoring and fault diagnosis are important scientific issues in mechanical engineering, civil engineering, and other disciplines. The basic premise of structural health work is to be able to accurately diagnose the fault in the structure. Therefore, the accurate fault diagnosis of structure can not only ensure the safe operation of mechanical equipment and the safe use of civil construction, but also ensure the safety of people's lives and property. In order to improve the accuracy fault diagnosis of frame structure under noise conditions, the existing Convolutional Neural Network with Training Interference (TICNN) model is improved, and a new convolutional neural network model with strong noise resistance is proposed. In order to verify THE superiority of the proposed improved TICNN in anti-noise, comparative experiments are carried out by using TICNN, One Dimensional Convolution Neural Network (1DCNN) and First Layer Wide Convolution Kernel Deep Convolution Neural Network (WDCNN). The experimental results show that the improved TICNN has the best anti-noise ability. Based on the improved TICNN, the fault diagnosis experiment of a four-story steel structure model is carried out. The experimental results show that the improved TICNN can obtain high diagnostic accuracy under strong noise conditions, which verifies the advantages of the improved TICNN.