Multi-feature integration and machine learning for guided wave structural health monitoring: Application to switch rail foot

Abstract

Switch rails are weak but essential components of high-speed railway systems that have urgent nondestructive testing requirements owing to aging and the associated potential for fatigue damage accumulation. This study presents a multi-feature integration and automatic classification algorithm for switch rail damage using guided wave monitoring signals. A combination of piezoelectric transducers and magnetostrictive patch transducers is adopted to improve the monitoring performance and meet actual monitoring requirements. Furthermore, multiple features extracted from various signal processing domains—such as the time domain, power spectrum domain, and time–frequency domain—are proposed and defined according to the structure and characteristics of the switch rail and guided wave to represent the complex nature of the damage. A damage index is defined to eliminate the influence of various environmental and operational conditions, signal power, and other factors. In addition, a feature selection method based on binary particle swarm optimization with a new fitness function is proposed to select the most damage-sensitive features and eliminate irrelevant and redundant features to improve the classification performance. Moreover, considering that the results are easily influenced by experts’ subjective judgment and experience, the least-squares support-vector machine is used to construct automatic classification models to reduce the probability of artificial incorrect diagnosis and improve the generalization ability to unknown environments. Finally, three types of experiments on the foot of a switch rail are presented to evaluate the proposed method. The results indicate that the proposed method is capable of identifying damage in challenging cases and is superior to conventional methods.

Application of compressed sensing in the guided wave structural health monitoring of switch rails

“…Real-world monitoring experiments were conducted to evaluate the proposed method using our previous SHM system (TKMRGW-20, Zheda Jingyi Tech, Ltd Hangzhou, China) [8] (figure 12(a)). To apply the proposed CS method and not increase the cost of hardware when adapting to the existing SHM system, the guided wave data were first collected using the system.…”

“…In addition, guided waves can detect large areas in a short time (at the speed of several kilometers per second), which meets the requirements of extremely fast scanning of the HSR. This study is based on our previous works, which proved the effectiveness of guided waves in the switch rail SHM, and the guided wave SHM system has been established [7,8]. It has been operating since 2020 at the national railway-testing center in Beijing, China.…”

“…This further aggravates the issue that more guided wave data must be collected, transmitted, and stored, often accompanied by higher power consumption. For example, in our previous guided wave SHM system of the HSR [8], a substantial number of transducers (multichannel) are required because of the large scale and complexity of the switch rails, and the sampling rate is 1 MHz. The sampling interval between each channel is 12 s, and each channel generates approximately 700 MB data per day.…”

Unsupervised data normalization for continuous dynamic monitoring by an innovative hybrid feature weighting-selection algorithm and natural nearest neighbor searching

“…For the problem of SHM, Buckley et al 39 evaluated different feature selection methods based on the criterion for obtaining the maximum classification performance from some supervised classification algorithms. Liu et al 40 proposed a feature selection method based on binary particle swarm optimization with a new fitness function to choose the useful features and disregard irrelevant and redundant features to improve the classification performance. Wang and Song 41 utilized a feature selection approach called joint mutual information maximization in the problem of bolt-looseness detection.…”

An efficient approach for guided wave structural monitoring of switch rails via deep convolutional neural network-based transfer learning

“…As guided waves can propagate over relatively long distances, cover large areas and are sensitive to defects, they enable the long-range evaluation of rails from a single test point and have demonstrated great potential in nondestructive evaluation [2][3][4]. Many studies using guide waves focused mainly on the stock rail [5][6][7][8][9][10][11][12], and only a few studies [13][14][15][16][17] have focused on the switch rail. This study builds on our previous works that demonstrated the effectiveness of guided waves in structural health monitoring (SHM) of switch rails [14,15].…”