A Simulation-Data-Driven Tolerance and Structure Integrated Design Method for Planar Arrays

Planar arrays are composed of large numbers of discrete elements. The position errors of these elements are inevitable and strongly influence the electrical performance. Based on the geometric requirements, many methods have been proposed to deal with the tolerance design of position errors using experimental data. However, the geometry-based tolerance design methods may not satisfy the performance requirements. What's worse, the experimental data are often not available due to the high measurement cost. Here, a simulation-data-driven tolerance and structure integrated design method is proposed to reduce the degradation of electrical performance based on the performance requirements. First, a simulation dataset including position errors and the corresponding performance parameters is generated based on the conical sampling method and the antenna theories. Then, based on the extreme gradient boosting (XGBoost) algorithm, prediction models are built and the weights of discrete elements for each electrical performance parameter are determined. Finally, according to the practical performance requirements, the whole array is divided into several subarrays based on the clustered weights and the tolerance in each subarray is optimized with the grid searching method. Taking an antenna with $20\times 20$ elements as an example, simulation experiments are conducted and compared with existing tolerance design methods to verify the effectiveness of the proposed method.