A review in guided-ultrasonic-wave-based structural health monitoring: From fundamental theory to machine learning techniques

Zhengyan Yang,Hongjuan Yang,Tong Tian,Deshuang Deng,Mutian Hu,Jitong Ma,Dongyue Gao,Jiaqi Zhang,Shuyi Ma,Lei Yang,Hao Xu,Zhanjun Wu

DOI: https://doi.org/10.1016/j.ultras.2023.107014

IF: 4.2

2023-04-27

Ultrasonics

Abstract:The development of structural health monitoring (SHM) techniques is of great importance to improve the structural efficiency and safety. With advantages of long propagation distances, high damage sensitivity, and economic feasibility, guided-ultrasonic-wave-based SHM is recognized as one of the most promising technologies for large-scale engineering structures. However, the propagation characteristics of guided ultrasonic waves in in-service engineering structures are highly complex, which results in difficulties in developing precise and efficient signal feature mining methods. The damage identification efficiency and reliability of existing guided ultrasonic wave methods cannot meet engineering requirements. With the development of machine learning (ML), numerous researchers have proposed improved ML methods that can be incorporated into guided ultrasonic wave diagnostic techniques for SHM of actual engineering structures. To highlight their contributions, this paper provides a state-of-the-art overview of the guided-wave-based SHM techniques enabled by ML methods. Accordingly, multiple stages required for ML-based guided ultrasonic wave techniques are discussed, including guided ultrasonic wave propagation modeling, guided ultrasonic wave data acquisition, wave signal pre-processing, guided wave data-based ML modeling, and physics-based ML modeling. By placing ML methods in the context of the guided-wave-based SHM for actual engineering structures, this paper also provides insights into future prospects and research strategies.

acoustics,radiology, nuclear medicine & medical imaging

What problem does this paper attempt to address?

The main problems that this paper attempts to solve are the challenges faced by the existing guided - wave - based structural health monitoring (SHM) techniques when applied to actual engineering structures, especially the inefficiency and unreliability of signal feature mining methods. Specifically:

1. **Complex Propagation Characteristics**: The propagation characteristics of guided waves in engineering structures in actual service are very complex, making it difficult to develop accurate and efficient signal feature mining methods.
2. **Damage Identification Efficiency and Reliability**: The efficiency and reliability of existing guided - wave - based methods in damage identification cannot meet engineering requirements.
3. **Application of Data - Driven Methods**: With the development of machine learning (ML), researchers have proposed many improved ML methods that can be incorporated into guided - wave - based diagnostic techniques to enhance the practical application effect of SHM.

To solve these problems, the paper provides a review of the latest progress in machine - learning - based methods, covering multiple aspects from basic theory to specific applications, aiming to improve the effectiveness and reliability of guided - wave - based SHM techniques. Specifically, it includes the following aspects:

- **Guided - Wave Propagation Modeling**: Understanding the propagation characteristics of guided waves in complex waveguide structures.
- **Guided - Wave Data Acquisition**: Optimizing sensor layout to obtain high - quality guided - wave data.
- **Wave Signal Pre - processing**: Modifying and synthesizing data for input into ML models.
- **ML Modeling Based on Guided - Wave Data**: Using ML models for anomaly detection.
- **Physics - Based ML Modeling**: Combining the advantages of physical models and ML models to improve diagnostic accuracy.

Through the combination of these methods, the paper hopes to provide new ideas and development directions for future guided - wave - based SHM techniques.