Fast GO/PO RCS calculation: A GO/PO parallel algorithm implemented on GPU and accelerated using a BVH data structure and the Type 3 Non-Uniform FFT

The purpose of this PhD research was to develop and optimize a fast numeric algorithm able to compute monostatic and bistatic RCS predictions obtaining an accuracy comparable to what commercially available from well-known electromagnetic CADs, but requiring unprecedented computational times. This was realized employing asymptotic approximated methods to solve the scattering problem, namely the Geometrical Optics (GO) and the Physical Optics (PO) theories, and exploiting advanced algorithmical concepts and cutting-edge computing technology to drastically speed-up the computation. The First Chapter focuses on an historical and operational overview of the concept of Radar Cross Section (RCS), with specific reference to aeronautical and maritime platforms. How geometries and materials influence RCS is also described. The Second Chapter is dedicated to the first phase of the algorithm: the electromagnetic field transport phase, where the GO theory is applied to implement the †œray tracing†�. In this Chapter the first advanced algorithmical concept which was adopted is described: the Bounding Volume Hierarchy (BVH) data structure. Two different BVH approaches and their combination are described and compared. The Third Chapter is dedicated to the second phase of the calculation: the radiation integral, based on the PO theory, and its numerical optimization. Firstly the Type-3 Non-Uniform Fast Fourier Transform (NUFFT) is presented as the second advanced algorithmical tool that was used and it was indeed the foundation of the calculation of the radiation integral. Then, to improve the performance but also to make the application of the approach feasible in case of electrically large objects, the NUFFT was further optimized using a †œpruning†� technique, which is a stratagem used to save memory and computational time by avoiding calculating points of the transformed domain that are not of interest. To validate the algorithm, a preliminary measurement campaign was held at the headquarter of the Ingegneria Dei Sistemi (IDS) Company, located in Pisa. The measurements, performed on canonical scatterers using a Synthetic Aperture Radar (SAR) imaging equipment set up on a planar scanner inside a semi-anechoic chamber, are discussed.