Knowledge-Based Conditional Generative Adversarial Network for Conformal Antenna Array Diagnosis

In this letter, we propose a novel machine learning (ML)-based method for real-time diagnosis of impaired conformal antenna arrays. An improved conditional generative adversarial network (cGAN) is first applied to the array diagnosis. Specifically, a generator is used to generate the diagnosed excitations, and a discriminator is used to determine whether the diagnosed excitations are real. The impaired far-field pattern is sampled to be fed as conditions into the discriminator and generator. Additionally, as prior knowledge, the sparsity of the impaired pattern is used to improve the proposed network and to enhance diagnostic accuracy. Examples of diagnosis and comparisons with existing ML-based techniques demonstrate that the proposed approach has a higher diagnostic accuracy, even with a smaller number of measurements