Application Cases

Experiment Name: Quantitative Characterization of Aluminum Plate Damage Based on Anomaly Index

Research Direction: Damage Quantification

Objective:
Structural damage detection and quantitative assessment are crucial for ensuring the safety of infrastructure in fields such as aerospace, shipping, petrochemicals, and defense industries, garnering widespread attention. Among existing structural monitoring technologies, ultrasonic guided wave testing stands out for its broad detection range, high speed, sensitivity to internal structural damage, safety, convenience, and low cost, making it highly promising for non-destructive testing and health monitoring of thin-plate components.

Figure 1: Time-Domain Damage Signal Extraction from Sensor Signals

Ultrasonic guided wave signals exhibit short-term non-stationarity and are susceptible to external interference. To enhance the detection of weak defect signals, researchers have employed methods such as time-reversal focusing, phased array focusing, and virtual reverse focusing to improve the signal-to-noise ratio of Lamb waves and amplify defect signals. However, these techniques are not only operationally complex but also prone to generating additional clutter during the focusing process, which can affect damage detection accuracy. In general, raw damage features exhibit low sensitivity to initial damage and subtle degradation trends. Individual damage features can only reflect partial or specific aspects of a structural system, making comprehensive and accurate damage assessment challenging. High-dimensional multi-domain features often contain significant redundant information, leading to the "curse of dimensionality" and impacting the accuracy of damage state evaluation. Therefore, it is necessary to establish a comprehensive damage assessment index through feature fusion based on sensitive damage feature screening, enabling unified and accurate evaluation of structural damage states. This study introduces a sensitive damage feature fusion method based on Self-Organizing Feature Mapping (SOM) to construct a unified index for comprehensive assessment of aluminum plate damage.

Testing Equipment:
ATA-8202 RF power amplifier, signal generator, laser vibrometer, aluminum plate under test.

Experimental Procedure:
To better validate the superiority of AI in assessing aluminum plate damage, this section demonstrates its effectiveness through a bending fatigue cycle experiment on an aluminum plate, as illustrated below. Lamb waves generated by the excitation device propagate through the aluminum plate and are received by the sensor. The experimental material is an aluminum plate measuring 400 mm × 40 mm × 15 mm, placed unrestrained on sponge blocks. A function generator produces a 10-cycle Lamb wave excitation signal with a center frequency of 500 kHz, which is transmitted into the aluminum plate via the ATA-8202 RF power transmitter and received by a full-field scanning laser vibrometer. At bending fatigue cycles of 0, 1, 2, 3, 4, 5, 6, 7, 8, and 9, guided wave signal excitation-reception experimental data are recorded, and 18 sets of feature parameters are extracted.

Figure 2: Schematic Diagram of the Aluminum Plate Bending Fatigue Cycle Experiment

Experimental Results:
To achieve consistent characterization of aluminum plate bending damage, sensitive feature vectors from 30 sets of non-damaged signals obtained under identical test conditions were used as training samples to establish an SOM model. The anomaly index (AI) for each bending-damaged plate was then calculated. The results are shown in Figure (a) below. It is evident that the proposed AI exhibits a stronger linear correlation with the number of bending cycles. To validate the superiority of the SOM-based AI for damage state assessment, the feature vector composed of the average values of sensitive features from 30 measurements under non-damaged conditions was used as the baseline. The Euclidean distance between the feature vector F under damaged conditions and the baseline was used as the damage evaluation metric. The results are shown in Figure (b) below. As seen in Figure (a), the AI values increase linearly with the number of bending fatigue cycles and are distributed around the fitted curve, showing a high linear correlation with fatigue bending cycles. In Figure (b), although the damage values based on Euclidean distance generally increase with the number of bending cycles, they are mostly distributed below the fitted curve, exhibiting poor linear correlation and low sensitivity to damage evolution. The strong linear correlation between the AI index and bending cycles in Figure (a) demonstrates the AI index's capability to quantitatively characterize the degree of aluminum plate bending damage, confirming its effectiveness and sensitivity in assessing aluminum plate fatigue damage.

Figure 3: Quantitative Assessment Results of Aluminum Plate Damage

Aigtek ATA-8000 Series RF Power Amplifier:

Figure: Specifications of the ATA-8000 Series RF Power Amplifier