Long-term guided wave structural health monitoring in an uncontrolled environment through long short-term principal component analysis

Abstract

Environmental effects are a significant challenge in guided wave structural health monitoring systems. These effects distort signals and increase the likelihood of false alarms. Many research papers have studied mitigation strategies for common variations in guided wave datasets reproducible in a lab, such as temperature and stress. There are fewer studies and strategies for detecting damage under more unpredictable outdoor conditions. This article proposes a long short-term principal component analysis reconstruction method to detect synthetic damage under highly variational environments, like precipitation, freeze, and other conditions. The method does not require any temperature or other compensation methods and is tested by approximately seven million guided wave measurements collected over 2 years. Results show that our method achieves an area under curve score of near 0.95 when detecting synthetic damage under highly variable environmental conditions.

Guidelines for effective unsupervised guided wave compression and denoising in long-term guided wave structural health monitoring

“…16,46 The experiment started in March 2018 and has collected over 11 million measurements by now. More experiment details can be found in Yang et al 16 and Kim et al 46 Collected guided waves from fair weather and precipitation are provided in Figure 2. We also illustrate examples of guided waves with synthetic damage and noisy conditions with an SNR of −10 dB.…”

“…where y ^ i is the i th guided wave of our synthetic damage data, and γ is the scatter factor. 16,46 The y i and y ′ i signals are taken from the same experiment but from different receivers. The y i measurements are taken from the shortest path (0.20 m) between sensors in the experiment and the y ′ i measurements are taken from the longest path (0.44 m) between sensors in the experiment.…”

“…A value of 0.95 is significant since weak damage is often shown to cause a drop of less than 0.05 in the guided wave correlation coefficient with optimal baseline comparison. 16 Hence, Figure 5 suggests that to maintain a correlation coefficient above 0.95, the measurement interval should be less than 430 s (approximately 7.2 min).…”

“…However, this often requires that we accumulate huge datasets. 8 For example, methods based on baseline signal selection, [9][10][11] singular value decomposition, [12][13][14] principal component analysis (PCA), [15][16][17] independent component analysis, 18,19 and deep learning [20][21][22][23][24] utilize collections of previous measurements to make decisions about new measurements. As datasets grow in size, many of these methods become more effective, but also more increasingly difficult to use computationally.…”

Continuous monitoring with a permanently installed high-resolution ultrasonic phased array

“…Previous authors have suggested that temperature variations strongly affect ultrasonic data. 10,39,40 For example, the ambient temperature difference between the two experimental measurements as shown in Figure 5 is 11.5 °C and this could be responsible for the changes in the ultrasonic datasets.…”