Enhancing the Goman–Khrabrov dynamic stall model through flow delay analysis

The complete dynamic stall process encompasses a series of complex developmental stages, such as flow separation, leading edge vortex shedding, and reattachment. Unlike static stall, dynamic stall exhibits hysteresis, rendering phenomenological models as complex nonlinear state-space systems, often accompanied by numerous empirical parameters, which complicates practical applications. To address this issue, the Goman–Khrabrov (G-K) dynamic stall model simplifies the state space and retains only two empirical parameters related to time delays. Our study finds that different developmental stages of dynamic stall exhibit various time delay scales. The G-K dynamic stall model, which utilizes a first-order time-invariant inertia system, forcibly unifies the time scales across different stages. Consequently, this leads to intractable nonphysical modeling errors. This paper introduces the latest revised G-K model that employs a time-varying state space system. This model not only maintains a concise form but also eliminates the nonphysical modeling errors previously mentioned. In response to the challenge of identifying empirical parameters, this paper presents a parameter identification method for both the original and revised G-K models utilizing a Physics-Informed Neural Network. The revised model was validated through dynamic stall load prediction cases for mild, moderate and deep dynamic stall on various airfoils, achieving a maximum accuracy improvement of up to 74.5%. The revised G-K model is capable of addressing a broader range and more complex practical applications.