Application Cases

Experiment Name: Application of the ATA-2031 High-Voltage Amplifier in Ultrasonic Non-Destructive Testing of Composite Plates

Experiment Direction:Composite Materials, Ultrasonic Guided Waves, Non-Destructive Testing, Signal Processing, Deep Learning

Experimental Equipment:ATA-2031 High-Voltage Amplifier, Arbitrary Function Waveform Generator, Oscilloscope, Ultrasonic Signal Preamplifier, CFRP Composite Plate, etc.

Experiment Objective:Carbon Fiber Reinforced Polymer (CFRP) composite materials are widely used in the aerospace field due to their excellent mechanical properties. However, the detection and localization of damage in composite materials remain highly challenging. This study introduces a novel method using a Multimodal Gated Recurrent Unit Neural Network (MGNN) model for damage detection and localization in CFRP composite materials.

Experiment Process:

The generation and reception of LW signals use a single actuator-receiver pair. The study selected a five-cycle sine toneburst signal modulated by a Hanning window. For signal generation and data acquisition, the center frequency of the input signal was 80 kHz. Ultrasonic guided wave signals were sampled at a rate of 10 MHz, and the acquisition duration was adjusted to ensure the capture of the activated wave modes and their boundary reflections, considering the group velocity of a single wave mode. The excitation voltage was set to 20 V, which was then boosted to a peak of 80 V through the high-voltage amplifier. The amplification rate of the preamplifier was set to 60 dB.

To improve the signal-to-noise ratio (SNR) of the received signals, the received signals were averaged 16 times to reduce random noise and further suppressed high-frequency noise using a four-level scale decomposition based on the "DB6" wavelet. To simplify the acquisition of the dataset, additional mass blocks were considered to simulate actual damage.

For damage assessment of the composite plate, the inspection area was set as a 400×400 mm square area and evenly divided into 100 small squares. Damage points were located starting from the lower left corner of the inspection area, with a 4 cm interval. Each damage point was centered in a small square area, and only one simulated damage was introduced at a time. It is worth noting that there were no damage points at the sensor locations. This setup generated 119 different damage signals.

Experiment Results:

The proposed MGNN outperformed other models and had the fastest detection speed. The proposed MGNN model had an MSE of 0.8804 and required the least amount of time (only 0.0562 seconds), demonstrating the best damage detection capability of the MGNN.

Specifications of the ATA-2031 High-Voltage Amplifier Used in the Experiment: