Immune intelligent online modeling for the stretching process of fiber

Abstract

An intelligent online modeling approach using the immune mechanisms for the stretching process of fiber is proposed in this paper. Linear parameter varying model is utilized as the process model, and the parameters are estimated under the framework of the expectation maximization algorithm. The proposed approach is composed of offline modeling and online modeling. For the offline modeling, the parameters of the local models as well as the weighting functions are estimated simultaneously, and this is the process of producing the first antibodies. Since the process is dynamic and continuous, when new operating point is different from the historical data, the online modeling approach is applied to estimate the parameters of new local models and weighting functions, and this is the process of producing the forthcoming antibodies. The proposed intelligent online modeling approach has the capability of learning and evolution, and can update the parameters adaptively. Apply the proposed approach to the stretching process and compare with other modeling methods, and then the Friedman and the Nemenyi post-hoc tests for assessing the statistical significance of differences in performance are analyzed. The feasibility and efficiency are demonstrated.