Layer thickness estimation in layered media using step-frequency continuous wave radar

Layer thickness estimation (LTE) of subsurface layers or multi-layered structures is one of the most important applications of ground penetrating radar (GPR). LTE of the roadway's pavement layers can evaluate their mechanical response to suggest the required repair work. Early and effective non-destructive LTE of transportation infrastructures is vital for economic and social development by saving time and reducing maintenance costs. The need for LTE becomes more urgent and prominent to inspect the structure of the old dams and bridges. GPR-based LTE is a fast and safe way to protect workers' lives and make work easier in places like mines, where sudden collapses frequently occur, and many casualties occur as a result. Several signal processing methods using different radar technologies have been developed for this purpose. Most of these methods require an experienced radar operator with good experience to analyze the output data or images of the GPR system and extract correct layer thickness estimations especially in the case of strong random noise and interference signals. In this study, a VNA based step frequency continuous wave (SFCW) radar has been designed to perform measurement and signal processing in the 2-4 GHz and 1.5-5.5 GHz frequency bands by examining a multi-layered medium consisting of polyethylene layer, air layer and reflective metal plate. A new LTE method has been proposed which applies principal component analysis (PCA) and range-FFT calculation to the baseband output signals of SFCW radar. By applying PCA as a blind source separation (BSS) technique to the SFCW baseband signals, the peak signals reflected from the layer boundaries are clearly obtained. Singular value decomposition (SVD) technique has been used in PCA calculations. Due to the separate processing of the principal components, the thresholding process can be actively applied in the range-FFT matrices of these principal components. Thus, the separation of radar signal peaks related to the layer boundaries has been simplified and the LTE error has been reduced. Despite the presence of random noise and interference signals, the proposed method has yielded good results with average maximum LTE error of 1.37% for the 2-4 GHz frequency range and 2.14% for the 1.5-5.5 GHz frequency range. Comparing with surface reflection method (SRM) that has been applied for LTE using SFCW signals in the two frequency ranges mentioned, the proposed PCA-based LTE method has given LTE errors lower than those of the SRM by 5.63% and 2.94% in the frequency ranges 2-4 and 1.5-5.5 GHz, respectively.