Comparison of Electromagnetic Data With Irregular or Discontinuous Surfaces Using the Feature Selective Validation Method

In the field of Computational Electromagnetics, validation is a formal process to ensure the expected behaviour of the model. The Feature Selective Validation (FSV) method was originally developed to aid the validation of computational electromagnetics, and particularly electromagnetic compatibility (EMC). Since then, it has been adopted by the IEEE Standard for Validation of Computational Electromagnetics Computer Modelling and Simulations 1597.1 and used in a variety of other applications. The FSV method quantifies the difference between two sets of original data using a reliability function based on the decomposition of the data into a number of component parts that are then combined using a weighted scheme, giving a number of presentations of the comparison data. From the literature review, it is identified that the FSV method has been applied for various structured data like the S-parameter, radiation pattern, efficiency, and gain of an antenna. Since the original development of the FSV in the field of computational electromagnetics, more complex data like surface current, electric, and magnetic field outputs from Ultra High Frequency (UHF) devices are represented in 2- (or higher) dimensional image formats with irregular shapes, which are non-rectangular structures including features such as spaces or gaps within the device structure. However, performing FSV on such image data is challenging, particularly to avoid non-contributing spaces or voids in the image structure dominating the comparison results. This thesis discusses in detail a methodology developed to perform 2-Dimensional FSV on 2-Dimensional regular (rectangular and square) shaped images by segmenting the image into multiple blocks. In each block, 2-Dimensional FSV is performed separately, and then the outputs from the segmented blocks are concatenated to form the original 2-Dimensional image. Finally, the FSV outputs from the segmented approach are compared with the FSV outputs obtained from the original (full) structure, and the results are analysed using a non-parametric statistical test, which shows that the approach leads to results that support the proposed solution of comparison by segmentation and recombination. The proposed method is demonstrated on regular and irregular images to allow a detailed analysis.