C2022-006

About

The disclosed invention integrates acoustic emission (AE) sensing with deep multimodal fusion models to enable fast, accurate, robust, and non-intrusive detection of faults during operations of cooling systems. By taking advantage of the raw AE signals and the converted spectrograms using different channels of the deep learning model, it will yield unprecedentedly high accuracy and robustness of fault detection. The optimized deep learning algorithm also enables real-time monitoring and fast response. The invention will not only make an impact on fundamental thermal transport research, but also industrial production of electric vehicles, hybrid vehicles, power electronics, and power generation, among others. Specifically, this technology will provide reliable means for thermal management of compact, lightweight, high-frequency, and high-power-density transformers, of inverters, and converters, and of integrated motor, battery, and power electronics, which are critical to the electrification and decarbonization of vehicles, ships, and airliners. This technology will also be useful for other electronics using wide band gap semiconductors, e.g. the mitigation of interpulse instability in GaN power amplifiers in radar systems. The invention will be beneficial to the safe operations of cooling systems of nuclear power plants. With real-time detection of departure from nucleate boiling, it will effectively avoid overheating issues of the fuel claddings and allows reduced departure from nucleate boiling ratio to improve the thermal efficiency of power generation.