أرسل هذا في رسالة قصيرة: Physics-Informed Neural Networks with Runge–Kutta Discretization: Stability, Accuracy, and Applications