Structural damage identification using strain mode differences by the iFEM based on the convolutional neural network (CNN). (15th February 2022)

• Record Type: Journal Article
• Authors: Li, Mengying; Jia, Dawei; Wu, Ziyan; Qiu, Shumao; He, Wei
• Is Part Of: Mechanical systems and signal processing. Volume 165(2022)
• Publication Date: 2022-02-15
• Subjects: Damage identification -- Inverse finite element method -- Strain mode differences -- Reconstruct deformation -- Convolutional neural network; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2021.108289
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: A new damage identification method based on the iFEM for SHM is proposed. Strain mode difference damage index by the iFEM shows a high damage identification accuracy. The iFEM based on iQS4 provides a high accuracy for structural deformation reconstruction. With the consideration of noises under different damaged conditions, the proposed damage identification method based on iFEM by CNN improves recognition efficiency. Abstract: Structural damage identification is the key problem of structural health monitoring (SHM), which has been widely used in many fields. Traditional damage detection methods usually require complete measured data and modal information before and after damage, resulting in difficulty in practical applications. Strain-based damage indicators have attracted widespread attention due to their high sensitivity to local damage. In this paper, using the inverse finite element method (iFEM), a damage identification method based on strain mode differences is proposed. The global strain data can be performed only by the limited strain measuring points, which will greatly improve the efficiency of direct use of strain data to achieve damage detection, and it can identify and locate the damage just using the post-damage strain modal data. It is noteworthy that this paper investigates the application of the convolutional neural network (CNN) training with input data contaminated by random noises in structural damage estimation. The results show that the CNN-based damage detection method using strain mode differences as the inputs has a high accuracy under different damage conditions, i.e., the proposed method not only has a significant damage localization ability, but also has a relatively high damage quantification prediction accuracy, which provides a new research approach for the structural health monitoring system.