Application of an AI-assisted Approach for Efficient EMI Source Characterization based on the Spherical Wave Expansion Technique

Accurate estimation of unwanted electromagnetic radiation is becoming nowadays a hot topic for a large set of cutting-edge electronic devices and applications, like for instance high-performance computing systems and server units. The possibility to reconstruct the radiation from a source is a fundamental feature for modern system designers, capable of paving the way also to the possibility of estimating the overall radiation coming from a set of radiation sources, like stacks of server units in a data center. However, most of the time, the radiation source can be characterized (measured) only in terms of the amplitude of the radiated field and this fact prevents the possibility of accurately predicting the overall radiation from several sources. This work aims to present an approach based on Machine Learning techniques and suitable to efficiently characterize a radiation source in terms of its Spherical Wave Coefficients only by using field amplitude data as input. Since the scope of the work is to present the mere proof-of-concept, well-established Neural Network-based Machine Learning algorithms will be used in the proposed method. The general framework and some simulation analyses to validate the method are proposed.