Automated Neural Network-Based Multiphysics Parametric Modeling of Microwave Components

Abstract

This paper proposes a novel technique for automated neural network based multiphysics parametric modeling of microwave components. For the first time, we propose to utilize automated model generation (AMG) algorithm in the field of electromagnetic (EM) centric multiphysics parametric model development to improve the neural-based multiphysics modeling efficiency. All the subtasks in developing a neural network based multiphysics parametric model, including EM centric multiphysics data generation, neural network structure adaptation, training and testing, are integrated into one unified and automated framework, thus converting the conventional human-based manual modeling into an automated computational process. In the proposed algorithm, automated EM centric multiphysics data generation is realized by automatic driving of multiphysics simulation tools. Parallel computation technique is incorporated to further speedup the data generation process by driving multiple EM centric multiphysics simulations on parallel computers simultaneously. In addition, automated neural model structure adaptation algorithm for multiphysics parametric modeling is also proposed. In this way, the proposed technique automates the neuralbased multiphysics model development process and significantly reduces the intensive human effort and modeling time demanded by the conventional manual multiphysics modeling methods. The achieved neural model can be used to provide accurate and fast prediction of the EM centric multiphysics responses of microwave components in high-level multiphysics design. Examples of multiphysics parametric modeling of two microwave filters are presented to show the advantage of this work.INDEX TERMS Design automation, multiphysics modeling, neural networks, parallel computation, parametric modeling.

Mechanistic and Machine Learning Modeling of Microwave Heating Process in Domestic Ovens: A Review

“…The CPU time for predicting a test sample by using a trained machine learning model, compared to the multiphysics model, was reduced by 98%. Given the characteristic of automatic hyperparameter selection and updating, the hybrid model could be used without users’ understanding of the model structure or predefined parameters [ 90 ].…”

Reliable Structural Failure Detection in Eye Bolts using Reflectometry Signals

“…The techniques and applications of artificial neural networks have been constantly improved and demonstrate accuracy comparable or better than humans in several recognition tasks [9]. ANNs has already been used in microwave applications such as modelling [10]- [12], design of devices [13], calibration [14], and fault detection [15].…”

AI versus Classic Methods in Modelling Isotopic Separation Processes: Efficiency Comparison

“…Due to the approached separation process complexity and due to the necessity of the future implemented control systems (for controlling the 18 O isotope concentration) to gentableerate high control performances (both from technological and from economic reasons), the intelligent control figure usage becomes feasible and necessary. In practice, many intelligent control strategies are based on using the neural networks [67][68][69][70]. Consequently, the necessity to train neural controllers occurs.…”