Artificial Neural Network-Assisted Optimization for Four-Pole Waveguide Filter Design

The optimization of microwave devices is very important in microwave and millimeter-wave frequency bands. Traditional microwave device optimization generally uses electromagnetic (EM) simulation, which requires repeated simulations to adjust the values of geometrical parameters. This makes the simulation-based optimization inefficient. In order to speed up the optimization of microwave devices, this paper proposes state-of-art artificial neural network (ANN) optimization for microwave devices. In the proposed method, an ANN model is developed to accurately represent the relationships between the inputs and outputs of microwave devices. After model development, the trained model can be used to perform the design optimization. The trust region algorithm is used to guarantee the convergence of the proposed technique. By using this method, the optimization can converge much faster than the existing simulation-based optimization. A four-pole waveguide filter example is illustrated to verify the effectiveness of the proposed technique.