Study on Temperature Prediction Method of Electron Collector Based on Reduced-Order Model

The electron collector in high-power microwave devices require effective thermal management under pulsed heat sources. However, the complexity of transient heat sources leads to high computational costs in simulations, hindering efficient thermal analysis. In this paper, the thermal characteristics at the end of the pulse heating period and intermittent period in electron collectors are first carried out. Subsequently, a reduced-order model (ROM) for predicting the electron collector's temperature is proposed, utilizing singular value decomposition and multiple interpolation methods. The construction process of this model is initiated by generating training datasets through numerical simulations under various conditions. Singular value decomposition is employed to identify dominant thermal features, reducing the dimensionality of the dataset while preserving critical thermal characteristics. Multiple interpolation methods, including polynomial regression, least squares, and Kriging interpolation, are systematically implemented to improve the prediction accuracy of the reduced-order model. Furthermore, the influence of thermal power and water flowrate on the maximum temperature and thermal uniformity of the electron collector by this model is examined. Results indicate that the prediction maximum error of this model remains below 1%. The reduced-order model demonstrates remarkable computational acceleration, achieving a 3400-fold performance enhancement by reducing simulation duration from 120min per computational fluid dynamics (CFD) case to 2.1s. The proposed approach is expected to significantly reduce computational costs while maintaining high accuracy, and thus, it is an effective calculation for engineering applications in high-power microwave devices.