A Multi-Layer Perceptron Neural Network for the Automated Design of Frequency Selective Surface (FSS) Based Absorbers

Frequency selective surface (FSS) based absorbers are used on stealth platforms in order to reduce their observability from rival radar systems. However, their traditional design approach is very time consuming and demands substantial computational resources. In this regard, this paper explores the application of deep learning in the design of FSS based absorbers, reducing reliance on computationally expensive electromagnetic simulations. A multi-layer perceptron (MLP) neural network (NN) is proposed to predict the FSS unit cell given target reflection characteristics. The network is trained on a synthetically generated dataset of FSS geometries and their corresponding reflection responses. Experimental results show that the model achieves a prediction accuracy of 51.21% on test data and closely approximates the desired frequency response, demonstrating its effectiveness. The predictive performance of the model is illustrated using suitable test cases. The findings highlight the potential of deep learning in the design of stealth components.