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

Area of Science:

• Electrical Engineering
• Materials Science
• Computational Methods

Background:

• Bimetals are crucial for thermal tripping in miniature circuit breakers (MCBs) during overloads.
• Traditional design methods (experimental, numerical) are time-consuming, hindering optimization within development timelines.
• Optimizing bimetal performance is complex due to intertwined electrical, mechanical, and thermal requirements.

Purpose of the Study:

• To introduce a novel, efficient methodology for predicting bimetal temperature rise and performance characteristics in MCBs.
• To develop a hybrid numerical-machine learning model for accelerated and accurate bimetal design.
• To validate the model against experimental data and demonstrate its effectiveness in product redesign.

Main Methods:

• Developed a time-based finite difference numerical model for the bimetal's thermal behavior.
• Integrated a machine learning (ML) model with the numerical model, leveraging experimental data for enhanced accuracy.
• Applied the consolidated model to redesign a bimetal for a 32 A MCB.

Main Results:

• The novel hybrid model demonstrated high accuracy, with a maximum error of 8% compared to experimental tests.
• Redesigning the bimetal for a 32 A MCB resulted in a significant reduction of maximum temperature by 24 °C.
• The methodology proved to be simple, fast, robust, and reliable for predicting bimetal performance.

Conclusions:

• The developed numerical-machine learning model offers a significant improvement over conventional methods for bimetal design in MCBs.
• This approach considerably reduces design time and enhances the optimization of circuit breaker performance.
• The validated model provides accurate predictions, facilitating the development of more efficient and reliable MCB products.