Segmentation-Oriented Compressed Sensing for Efficient Impact Damage Detection on CFRP Materials

Abstract

The emerging task-oriented compressed sensing (CS) provides a possibility to revolutionize traditional separate sensing-processing models by jointly considering task with data and system. This article presents an efficient mechatronic system for impact damage detection on carbon fiber reinforced polymer materials. By jointly designing the damage region segmentation task with the near-field microwave imaging system, the overall efficiency is improved by one order of magnitude without upgrading the hardware. The proposed segmentation-oriented CS greatly down-samples the specimens with a specially designed binary measurement matrix, followed by a corresponding segment-orthogonal matching pursuit process that obtains the defect region directly from under-sampled data. Compared to raster scan and the traditional CS designs, the proposed method eliminates the redundant process of obtaining the full data and defect region extraction algorithms. Segmentation-oriented CS achieves the best speed and accuracy that single method cannot, and it is much more robust to noise or measurement fault. This integrated sensing-processing method can inspire more efficient mechatronics applications.