Abstract: The following paper proposes a novel Multiple Models Method for observer design to solve the problem of state and parameter estimation of uncertain nonlinear time-varying parameters systems with unknown but bounded disturbance. Classically speaking, an interval observer is a special class of observers that generates a bounded interval vector for the real state vector in a guaranteed way under the assumption that the uncertainties are unknown but bounded; it gives an upper and lower estimate for the system states at each time instant (determining a certain interval for the estimated state variations). Several approaches have been developed and adapted to different kinds of models (linear, nonlinear, fuzzy, etc.). However, in the proposed approach, the objective is not to design an interval observer, but rather a classical Luemberger observer, based on an interval multiple model of the nonlinear system model. The novelty introduced in the paper is about proposing a new interval Multiple Models representation of the uncertain nonlinear system. The observer’s gains are developed based on the Lyapunov stability theory proving that the state and parameter estimation errors are stable and converge to an origin-centred ball of a given radius to be minimized. The design conditions are formulated into linear matrix inequalities constraints, which can be efficiently solved. A numerical example is given to illustrate the design and validate the performance of the interval observers.Abstract: The following paper proposes a novel Multiple Models Method for observer design to solve the problem of state and parameter estimation of uncertain nonlinear time-varying parameters systems with unknown but bounded disturbance. Classically speaking, an interval observer is a special class of observers that generates a bounded interval vector for th...Show More
Abstract: The remanufacturing of rolling mill rolls offers significant economic, environmental, and societal benefits. However, the uncertainty surrounding the performance degradation of retired rolls and its associated timeline poses challenges to the efficiency and cost-effectiveness of roll remanufacturing operations. Therefore, the real-time monitoring of the degradation status of rolling mill rolls is of paramount importance. This study presents an approach that combines multi-sensor data fusion with a multilayer perceptron (MLP) model, which takes into account economic considerations to predict the degradation status of hot-rolling mill work rolls and make online decisions for active remanufacturing. The degradation process of rolling mill rolls is analyzed, and degradation performance indicators are established. Eddy current signals and torque signals from the rolling mill surface are collected during the roll degradation experiments. The friction coefficient and energy of the Hilbert spectrum of the eddy current signal are used as online input features for the MLP model, which is trained using the degradation experiment data. The superiority of the proposed MLP model is validated through rolling mill roll degradation experiments. Based on the predictions of the MLP model, the optimal timing for remanufacturing rolling mill rolls in the time domain is evaluated using Wiener and update-reward theories. This approach enables the online monitoring and quantitative characterization of the comprehensive degradation of high-speed steel work rolls and facilitates online decision-making regarding the optimal timing for active remanufacturing.
Abstract: The remanufacturing of rolling mill rolls offers significant economic, environmental, and societal benefits. However, the uncertainty surrounding the performance degradation of retired rolls and its associated timeline poses challenges to the efficiency and cost-effectiveness of roll remanufacturing operations. Therefore, the real-time monitoring o...Show More