A closed-loop evaluation identification algorithm for rubber mixing process based on physical information and deep learning

Rubber mixing process is a crucial link in tyre manufacturing. Accurate mixing model will greatly affect the control performance, thus affecting the final yield and quality of tyres. However, due to the complexity of the mixing process, there is still a great challenge to establish an accurate mixing process model. In this paper, a closed-loop evaluation identification algorithm for mixing model is proposed. On the basis of the state space model established by heat transfer mechanism, the unknown parameters are identified based on CNN-LSTM deep network. A fuzzy evaluation method is utilised to comprehensively evaluate the identification effect in view of the lack of uniform evaluation indexes. The GRU network is used to correct the error of identification results. The deep neural network hyperparameters are iteratively updated based on the genetic optimisation algorithm to achieve the feedback optimisation. Finally, simulation verifies the effectiveness and merits of the proposed algorithm.