Optimal Method Research For Rubber Vulcanization Process System Of Water Lubricated Bearings

In the rubber vulcanization process of water lubricated bearings, the determinationof the optimal parameters is always the big issue. In order to solve this issue, this paperdetermines the optimal parameters using the technology of optimal modeling to ensurethe performance index of water lubricated bearings on the basis of no change ofvulcanizing equipment and technological process. The rubber vulcanization process ofwater lubricated bearings is a process system with strong nonlinearity, what makes theseproblems in the optimal modeling:Traditional modeling method can only establish a static approaching model whichis inaccuracy, because of the dynamic characteristics of the vulcanizing process. And theprocess of vulcanization may be affected by some uncertain interference factors. In thiscase, the traditional optimal method can only find out the optimal solution in the idealcondition, however, in the actual operation, disturbed by the uncertain interferencefactors, the control parameters are bound to deviate from the set value, and theperformance of bearings must be affected.In order to solve these issues, the artificial neural network is trained by theunscented kalman filter (UKF) which has dynamic performance to obtain the model forthe process system. And robust optimization is used to determine the optimal parameterswith the former model.In this paper, the main research achievements are as follows:First, considering the strong nonlinearity and dynamic characteristics in the processof vulcanization, based on the test data which is obtained in the former vulcanizing test,UKF is combined the theory of strong tracking filter to train the artificial neuralnetwork, bringing in the time-varying fading factor to adjust the filter gain matrix online,and the tracking ability can be further improved. Through the coMParison of the modelswhich are established by BPNN, UKFNN and improved UKFNN, the model which isestablished by the accuracy of static modeling method BPNN is far worse than themodel established by the dynamic modeling method. And the improved UKF is morebetter than the UKF because its better tracking ability.Secondly, considering the influence of uncertain interference factor which leads toexcursion of the performance index of the bearings, robust optimization which is basedon the former improved UKFNN accurate dynamic model is used and the non-dominated sorting genetic algorithm-II(NSGA-II) is used as the method of robustoptimization to determine the optimal parameters. And considering the global searchingability of NSGA-II is weak, the crossover operator is improved to guarantee thepopulation diversity.The dynamic model and robust optimization strategy are established for the rubbervulcanization process system of water lubricated bearings which lay a foundation forimproving its performance and production efficiency.