Set-membership identification recursive algorithm based on adaptive noise bounding

Set-membership identification recursive algorithm (SMIR) is proposed based on adaptive noise bounding in order to achieve complete dynamic modeling for a linear ordinary dynamic process without increment. Here we describe the output set as an ellipsoid under the framework of set-membership. First, we select an CARMA model with a single-input-single-output (SISO) system, and the measurement noise is unknown but bounded. The long pieces of historical data are selected based on the complexity of the dynamic characteristics of the actual object under study. Second, the designed algorithm can significantly determine the optimal noise bound-tuning coefficient on this model, making the noise bound more compatible with the actual noise environment. Next, the stability of the algorithm is proved using the Lyapunov method. Then, simulation experiments are conducted to model and predict both linear input processes and actual process data on a liquid-level control loop of the distillation tower section. Finally, the experimental results on two general examples verify the effectiveness of the proposed algorithm.