PATTERN has expertise from four key areas, i.e., electromagnetic compatibility (EMC), medical engineering, risk management, and sustainability management, along with specific expertise from MedTech sectors and a complete value chain. PATTERN will be the first European initiative where the people with all this expertise work together to make AI a step changer in innovative and sustainable solutions to manage EMI and set new design guidelines for electromagnetic reliability and the safety of the European products of tomorrow

 A total of 16 individuals doctoral candidate (DC) projects have been defined. Based on their projects and (common) secondments, each DC will have a mix of requirements, design, and system-level verifications. While each DC project has been defined as a stand-alone research line with its own unique contribution, several collaboration points have been planned to bring complementary results together and give extra value-added to the project. The inter-relation between each DC project and their corresponding core WPs are described below.

WP1: AI driven EMC design

 The goal of this work package is to unravel unknown PATTERNS in how the EM noise interacts with medical products by exploring AI approaches never used so far in EMC to create deeply structured knowledge (“class 3” AI systems): these causality models are a prime example on how AI can support EMC expertise to progress into gaining structural knowledge and an understanding of the complex mechanism and pattern of EM propagation. To ensure adequacy with the reality of design requirements, PATTERN will focus on utilizing Neural Networks and optimal control methods to enhance electromagnetic compatibility in complex electronic systems. It consists of accurately learning about variations within electronic systems that can potentially lead to EMI. Feature selection of the data combined with suitable machine-learning methods involving hybridization with support vector machines (i.e., kriging) will be used. By leveraging state-of-the-art, data-efficient, Machine Learning approaches and developing novel Bayesian active learning (BAL) techniques, the goal is to reduce the time of well-known, lengthy EMC tests.

WP2: AI for EMC Risk assessment during Design

This WP will focus on the phase of integrated EMC design in medical products. It starts at the heart of the EMI generation with the integrated circuit (IC) with its Printed Circuit Board (PCB) towards the whole hierarchical scale of components propagating these EMIs. Five DCs combine the SSbD approach in AI-based tools to accurately predict the EM propagation and performance of these diverse electronic parts. The developed methodologies will build on the fusion of existing first-principal knowledge and the modelling capabilities of deep-learning methods to provide physically driven, interpretable prediction tools with reliable performance guarantees.

WP3: AI for EMC Risk assessment during Test and Post-Deployment

It is focused development of diagnostic tools for EMC to test the postdeployment phase of safety-critical medical products from immediate release to aged devices. Diverse metadata of the product will be combined and analyzed by AI tools to comprehend the propagation of EMI in a diverse deployed installation, used in different EM environments. Aging and mechanical fatigue are among other known factors degrading the initially predicted performance of EMC design solutions. 

The developed methods will exploit already-existing physical knowledge in existing “EMC expert systems” and use them to rapidly learn diagnostic capabilities for specific products with unprecedented interpretability in EM engineering. DCs will research how AI can help the design engineer to identify the origin and the best solution for an EMC problem with self-learning diagnostic tools. The algorithm would build up its experience based on sets of EMC measurements for a given set of implemented design techniques in the EMC expert system of medical products. 

Data collection of EM-field surrounding devices is performed on uniform grids to ensure that the maximum emission is found. A specific focus is given to the optimization of the near-field sampling of a deployed product as it contains all the radiant energy threatening the EM integrity of the complete deployed system. The goal is to reduce the time of well-known, lengthy EMC tests and open a new window of opportunity for in-situ testing: the unwanted emission or unwanted susceptibility of complex systems are also caused by human factors (during installations), deterioration in lifetime, corrosion, shock and vibration or high-level system interactions between various sub-systems. The environment in which the product is used also brings unpredicted aspects. Therefore, we will use the in-situ measurement results to update guidelines and good practices towards a pre-design synthesis, based on collected knowledge that was never available before the PATTERN DN.

WP4: KPIs of the SSbD approach for Safety, Functionality and Circularity

All DCs will exchange and use their findings in 3 tracks: the “AI driven design”, the “Design Track” and the “Test and Post-Deployment Track” to ensure the continuity of solutions over the complete lifecycle of EMC solutions. These data will be actively shared and combined by the DCs through their own research project and common secondments. These will enable a final assessment of the safety, functionality, and circularity of the optimized EMI solutions and Design guidelines that have been developed in PATTERN for the MedTech markets. .