Deep digital twin of electromechanical actuator heat transmission by using physics-informed neural network

作者：Dong Liu, Shaoping Wang, Jian Shi, Di Liu, Yaoxing Shang · 发表于：IET conference proceedings. · 年份：2025 · DOI：10.1049/icp.2024.2840 · 被引用次数：1 · 研究领域：Iterative Learning Control Systems、Advanced machining processes and optimization、Laser and Thermal Forming Techniques

With the rise of electric aircraft, the application of electromechanical actuators in the aerospace field has received widespread attention and has broad development prospects. By utilizing digital twin technology, it can be used for the state monitoring and maintenance of the entire lifecycle of electromechanical actuators. Deep digital twin is based on deep learning method to create the virtual entity directly from the operation data. The digital twin based on physical information neural network is a feasible method for deep digital twin, which has attracted the attention of researchers. Especially in the field of digital twins, research hotspots include the reconstruction of mechanical fields, temperature fields, and electromagnetic fields. This paper proposed a deep digital twin method for the electromechanical actuator heat transmission by using the physics-informed neural network. At first, the governing equations of the heat transfer problem were introduced which are included the heat generation functions of the inner components. Then, the physics-informed neural network was established to solve the heat transfer problem which is driven by both the prior knowledge and the operation data. In the end, the method was verified through practical examples.