Design of Adaptive Steady-State Fading Memory Filter for Servo Drive Systems Under Maneuvering Conditions
To address the state observation requirements of permanent magnet synchronous motor servo systems under complex maneuvering conditions, this article proposes an artificial intelligence–enhanced state observation approach. First, by integrating the fading memory filter (FMF) model with covariance steady-state conditions, the state observation equation for servo systems based on the steady-state FMF is established for the first time. Second, comprehensive transfer function analysis reveals the influence of filter gain on the observer's responsiveness and smoothness, providing theoretical guidance for parameter optimization. To enhance the algorithm's adaptive capability, the innovation sequence is employed as a maneuverability assessment metric, while an optimized long short-term memory neural network architecture is implemented to preserve estimation accuracy under computational efficiency constraints. This framework enables dynamic filter gain adaptation, robust state estimation performance, and high computational efficiency. Comprehensive simulation and experimental results demonstrate the superior performance characteristics of the proposed adaptive steady-state FMF method.