Advanced Parametric Modeling of the Four-Pole Waveguide Filter Using Decomposed Sub-ANN-models

This paper presents an innovative technique for developing wide-range parametric models for microwave components by utilizing the decomposition method. In this method, a systematic algorithm is proposed to decompose the overall ranges of geometrical parameters into multiple small sub-ranges. We can build very accurate sub-model by good training for each sub-range. However, discontinuity issues may arise at the boundaries of adjacent sub-models. To address this, a new algorithm has been proposed that resolves these discontinuities and integrates all sub-models into a single, comprehensive model. By employing parallel computation, the model development process is accelerated. This proposed method not only improves accuracy but also reduces modeling time compared to methods that do not use decomposition. This technique is validated through the example of a four-pole waveguide filter.