Local wavenumber based damage imaging for C/SiC composites using adaptive selection for frequency and mode of guided wave

Carbon-fiber-reinforced silicon carbide (C/SiC) composites are widely used in aerospace thermal structures exposed to extreme environments, imposing higher requirements on non-destructive testing. Laser ultrasonic testing is well suited to C/SiC inspection because of its high sensitivity to damage, while local wavenumber estimation (LWE) shows strong imaging potential by characterising local wavenumber variations. However, empirical selection of the narrowband center frequency and wavenumber range reduces imaging quality. This study reformulates narrowband center frequency selection as an uncertainty driven optimization problem and proposes a strongly constrained Bayesian optimization framework for LWE imaging (SCBO-LWE). The center frequency is treated as the optimization parameter. The selected mode is automatically identified and adaptively filtered using the radially averaged energy amplitude of the 2D wavenumber spectrum. Mean difference contrast is used to evaluate damage imaging quality. A probabilistic posterior distribution is constructed to guide center frequency optimization. It is embedded into an adaptively varied expected improvement acquisition function with strong constraints to reduce sampling in unsuitable frequency regions and improve convergence. SCBO-LWE is validated on two thin-walled C/SiC composite plate specimens. The results show that SCBO-LWE finds the center frequency that produces the best damage imaging performance, whilst verifying the effectiveness of the proposed method.