https://www.celus.io/ POWER ELECTRONICS DESIGN 21 www.power-mag.com Issue 3 2026 Power Electronics Europe semiconductor can affect not only electrical performance but also heat generation, cooling requirements, and physical layout. These interconnections mean engineers need to consider the design as a system rather than a collection of individual components. AI-based tools are able to assist by identifying these relationships earlier. Instead of simply answering whether a component meets a particular specification, an intelligent system can help engineers investigate the wider implications of its selection. Managing Thermal and EMI Constraints Thermal management is one example of where this approach can provide practical value. Power losses ultimately become heat, closely linking thermal performance to component selection and switching behavior, as well as overall efficiency. Engineers may therefore need to consider pertinent factors that include thermal resistance, heat dissipation, cooling arrangements, and the physical placement of components. AI can help bring these considerations into the design process earlier by connecting component information with the requirements established for the system. A candidate component that looks attractive in terms of its electrical performance might warrant further investigation because of its thermal characteristics. EMI presents a similar challenge. Fast switching is able to improve power density and efficiency, but also introduces highfrequency noise and makes PCB layout, filtering, and parasitic effects increasingly challenging. For the engineer, it is not simply about selecting the right semiconductor, but understanding how that semiconductor behaves within the complete power stage and physical implementation. AI does the heavy lifting, bringing relevant information together and highlighting relationships that might otherwise require extensive manual investigation. From AI Assistant to Engineering Agent While an AI assistant can answer a question, an engineering agent has the potential to do considerably more. It can take a defined objective, break it into stages, use appropriate tools or information sources, and return the results within the context of the original requirements. For power electronics, this workflow could mean starting with a set of system requirements and identifying relevant functional blocks. It could also entail investigating suitable components and supporting the creation of an initial design. In addition, the agent could maintain awareness of the constraints established at the start rather than treating every subsequent task as an isolated request. This factor is relevant because power electronics design is inherently iterative. A change made to improve one characteristic can create a new problem somewhere else. An agentic approach could therefore support a feedback loop. If a proposed component or design change affects another requirement, the agent could identify and present the potential consequences to the engineer for consideration. Verification Remains Essential It is not possible to design power electronics safely by assuming that a plausible-looking result is a correct one. Electrical behavior, thermal performance, switching characteristics, EMI, and physical implementation all require careful consideration. AI-generated recommendations must therefore remain connected to the requirements that produced them. Component choices require traceability to relevant specifications, while proposed design changes should be capable of being checked against the constraints established by the engineer. This is another area in which an agentic workflow could prove useful. Rather than generating an output and leaving the engineer to determine its validity, the system can potentially support successive
RkJQdWJsaXNoZXIy MjQ0NzM=