Phased array failure detection and correction

• Main Author: Yeo, Beng Kiong.
• Other Authors: Lu Yilong
• Format: Theses and Dissertations
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Antennas, wave guides, microwaves, radar, radio; DRNTU::Engineering::Computer science and engineering::Computing methodologies::Simulation and modeling
• Online Access: http://hdl.handle.net/10356/52691
• Institution: Nanyang Technological University

This Ph.D. thesis is focused on the element failure detection of high frequency surface wave radars (HFSWR) using analog beamformer, and the failure correction techniques for adaptive digital beamforming arrays. By the end of our research, we have developed an SVM-based method for HFSWR failure detection, and an iterative convex optimitization (ICO) technique for the failure correction of uniform linear arrays (ULA) and planar arrays with digital beamforming. In mission-critical applications, timely failure detection and correction of phased arrays for surveillance is vital in an increasingly complex electronic warfare environment. As a consequence, there is considerable research interest and demand within the antenna community for the failure detection of high-power phased arrays using analog beamforming, and the failure correction of digitally beamformed phased array radars in the modern battlefield. The failure isolation and detection for analog arrays with analog beamforming architectures are particularly important in certain military applications, where there is significant preference for digitally-controlled high-power analog phased arrays, such as the HFSWR, with large one-dimensional receiving arrays that typically span over hundreds of meters to a few kilometers. Since time is of the essence, it is highly desirable to devise a better approach to detect and identify the presence of failed element(s) that pinpoint the failure(s) without the costly dismantling of components. We have demonstrated that a multi-class one-againstone (OVO) support vector machine (SVM) using a kernel of low-degree polynomial would surpass RBF-based kernels in both (i) the speed of training and (ii) the accuracy in subsequent predictions, especially under less than ideal signalto- noise ratio (SNR) situations where noisy test signals are received. For the failure classifications of both uniform linear arrays and planar arrays, weemonstrated that the feature space is better approximated by low-degree polynomial functions. The low-degree polynomial kernels provide much better fit for the same amount of training data, in spite of noise, and consistently outperformed the RBF-based kernels. Coupled with a binary-coded scenario system using dot-hex representation, the proposed multi-class SVM classifier is capable of rapid failure identification for small HFSWR deployments, in noisy environments. In situations where immediate repairs or replacements are not possible, the capability for continued operations of a lightly damaged surveillance phased array radar is very much preferred, thereby allowing ground forces to maintain their tactical advantage. Over the last decade, numerical algorithms employing hybrid and non-deterministic methods, ranging from Genetic Algorithms (GA), Particle Swarm Optimization (PSO), to Inverse Fast Fourier Transform (FFT), etc. have been proposed to restore a malfunctioned digital beamforming phased array system. Nonetheless, they may require considerable computational power with exponentially longer execution time as the number of failures increases, and/or the peak sidelobe level cannot be fully recovered. We have developed a fast technique which leverages on convex optimization for the failure correction of uniform linear arrays (ULA) and planar arrays. Using this iterative convex optimization (ICO) method, usable solutions for the average correction of small to medium-sized arrays are produced in a matter of minutes on a laptop with a powerful processor. Its solutions ensure the graceful degradation of far-field main beam pattern for average failure scenarios. More importantly, the main beam of the corrected array is steerable, while the original peak side-lobe level (SLL) is always maintained. Additionally, the ICO can further optimize the corrected beam pattern of the ULA, to steer wide-null(s) within the re-optimized pattern, allowing for continuing operations with anti-jamming capability.