Abstract: This paper proposes a damage detection method based on an improved DAS imaging algorithm by introducing time difference due to anisotropy of composite material. First, the finite element characteristic frequency method is used to obtain the dispersion curve of the composite plate, and the validity of the dispersion curve is verified. Next, the average phase velocity of the Lamb wave at mixed modes in the composite plate is obtained by two-dimensional Fourier transform (2-D FFT). The mixed modal group velocity is calculated according to the corresponding phase velocity, the mean change rate of the phase velocity and the dispersion curve obtained by simulation. The time difference due to anisotropy of composite material is investigated, and the damage location is estimated by the delay-and-sum (DAS) imaging algorithm. Experiments on carbon fiber multilayer composite plates verify the effectiveness of the proposed method.
Introduction
Composite materials, such as carbon fiber reinforced polymers (CFRP), are prized in aerospace and automotive industries for being lightweight yet incredibly strong. However, detecting internal damage like delamination or debonding is notoriously difficult. Engineers often use Lamb waves—ultrasonic waves that propagate through thin plates—for non-destructive testing.
In simple materials like aluminum, these waves travel at the same speed in all directions (isotropy). In composites, however, the wave speed changes depending on the direction of the fibers (anisotropy). Furthermore, Lamb waves are dispersive (speed changes with frequency) and exist in multiple modes simultaneously. Most existing methods struggle to separate these modes in multilayer composites. This paper tackles the challenge of imaging damage using raw, "mixed-mode" signals by adjusting the imaging algorithm to handle the material's directional quirks.
Methodology
The authors propose a multi-step workflow to improve the classic Delay-and-Sum (DAS) imaging algorithm:
- Dispersion Curve Modeling: Finite Element Method (FEM) software was used to calculate theoretical dispersion curves—graphs showing how wave velocity changes with frequency—for specific fiber orientations (0°, 45°, and 90°).
- Velocity Extraction via 2D-FFT: To bridge theory and reality, a Two-Dimensional Fourier Transform (2D-FFT) was applied to experimental signals. This allowed the researchers to measure the actual average phase velocity of the mixed-mode waves.
- Group Velocity Calculation: Using the phase velocity and the slopes of the simulated dispersion curves, they calculated the group velocity (the speed at which the wave energy actually travels).
- Anisotropy Correction: By comparing signals across different paths on a healthy baseline plate, they calculated a "time difference" ($t_n$) caused by anisotropy. This essentially calibrates the algorithm for the specific layup of the composite.
- DAS Imaging: Finally, these corrected time delays were used to map the "scattered" signals onto a pixel grid, where the highest intensity indicates the damage location.
Experimental Setup
The study relied on experimental data from multilayer T700 carbon fiber composite plates (300 mm x 300 mm x 6 mm) featuring a 0°/90° cross-ply arrangement. Internal damage was simulated by debonding a PZT piezoelectric sheet at a specific coordinate (15 cm, 15 cm) within the epoxy layer. An Agilent signal generator and a Tektronix oscilloscope were used to excite and receive 200 kHz Lamb wave signals.
Key Results and Insights
The proposed method proved highly accurate in locating the simulated debonding:
- Velocity Accuracy: Experimental group velocities matched theoretical values with very low error rates, ranging from 0.45% to 1.62%.
- Localization Precision: The researchers tested two different sensor layouts. Layout 1 estimated the damage with an error of only 0.986 cm, while Layout 2 achieved an error of just 0.787 cm.
- Embracing Mixed Modes: Instead of struggling to isolate a single wave mode (like S0 or A0), the study showed that using an average group velocity of mixed modes is sufficient for accurate imaging when combined with a robust dispersion model and directional calibration.
Conclusion
By incorporating directional time differences into the DAS algorithm, this research provides a more effective way to monitor the health of anisotropic composite structures. While the current method requires a healthy baseline signal, it bypasses the traditionally difficult task of modal separation. Future work will focus on reducing baseline dependency and improving noise robustness for real-world applications.
Summary updated