Grad Colloquium'24 Artificial Intelligence, İstanbul, Türkiye, 04 Haziran 2024, ss.36
In industries such as machine tools that demand substantial power and motion control, permanent magnet synchronous motors are extensively employed on account of their dynamicperformance and high power density. To ensure efficient control of this machinery, it is necessary to take into account equivalent circuit parameters. The online methods presented for determining these parameters primarily depend on laborious measurement processes, whereasthe analytical methods necessitate numerous computational processes. By eliminating the need for intricate computations and experiments, this research intends to develop a cascade-forwardneural network-based parameter estimation method. By relying solely on data obtained from manufacturer catalogs, the study was able to generate usable parameter estimation results in less than three minutes, with correlation values exceeding 0.9 at power levels in the range of 0.75 to 22 kW. Thus, the approach is a potential, accurate alternative. Future investigations may benefit from modifying the approach to generate estimates for a wider variety of powers. The prediction performance can be enhanced by combining neural networks with other machine learning or traditional methods to develop novel hybrid models