A hybrid numerical/machine learning model development to improve the bimetal performance in the electric circuit breakers

Abstract

Bimetals are widely used as a thermal tripping mechanism inside the miniature circuit breakers (MCBs) products when an overload current passes through the circuit for a certain period. Experimental, numerical, and, recently artificial intelligence methods are widely used in designing electric components. However, developing the bimetal for MCB products somewhat differs from developing other conductor items since they are strongly related to the electrical, mechanical, and thermal performance of the MCB. The conventional experimental and numerical approaches are time-consuming processes that cannot be easily utilized in optimizing the product's performance within the development lead time. In this study, a simple, fast, robust, and accurate novel methodology has been introduced to predict the temperature rise of the bimetal and other related performance characteristics. The numerical model has been built on the time-based finite difference method to frame the theoretical thermal model of the bimetal. Then, the numerical model has been consolidated by the machine learning (ML) model to take advantage of the experiments to provide an accurate, fast and reliable model finally. The novel model agrees well with the experimental tests, where the maximum error does not exceed 8%. The model has been used to redesign the bimetal of a 32 A MCB product and significantly reduce the maximum temperature by 24 °C. The novel model is promising since it considerably reduces the required design time, provides accurate predictions, and helps to optimize the performance of the circuit breaker products.