Joint time–frequency sparse optimization of laser ultrasonic arrays

Laser-acoustic array technology offers distinct benefits for inspecting microdefects in advanced industries like aerospace and additive manufacturing, including its non-contact nature, ability to excite multiple ultrasonic modes, and high imaging resolution. Nevertheless, its practical, real-time application is hindered by the substantial data loads, extended acquisition periods, and heavy computational demands associated with complete array data capture and full-focus image reconstruction. This paper introduces an adaptive sparse imaging approach that integrates joint time–frequency variance analysis with constraints from a minimum redundancy linear array (MRLA). First, the time-domain and frequency-domain characteristics of different laser ultrasound modes (LL, SL, LS, SS) were analyzed, revealing the intrinsic complementary mechanism whereby time-domain variance is sensitive to defect location, while frequency-domain variance is sensitive to defect size and material properties. Based on this, a joint time–frequency domain variance feature function was developed to evaluate the imaging contribution of each transmit element. Combined with the MRLA criterion to optimize transmit element selection, an adaptive sparse array was constructed, significantly reducing the amount of data acquisition. Secondly, a nonlinear enhancement phase-shift migration method was introduced into frequency-domain imaging to suppress incoherent noise through sign consistency discrimination and geometric averaging of amplitude; this was further combined with phase-coherent weighting to enhance defect contrast to noise ratio (CNR) and signal-to-noise ratio (SNR). Experimental validation was conducted on 0.3 mm and 0.5 mm SDH defects in 1060 aluminum alloy specimens. The results show that, under the four modes (LL, SL, LS, and SS), the CNR and SNR of the combined time–frequency domain variance method outperform those of single-domain methods. To address the poor imaging quality in the SS and LS modes, dual-mode fusion imaging combining LL and SL was employed, and the −6 dB drop method was used to quantify defect sizes. The absolute errors in detecting 0.3 mm and 0.5 mm defects were + 0.04 mm and + 0.07 mm, respectively, with relative errors less than 14 %. The method described in this paper effectively reduces the burden of data acquisition and processing while ensuring imaging quality, providing a new approach for real-time, high-resolution inspection using laser ultrasonic arrays.