Intelligent Optimization Control Method Research Of Titanium Alloy Bar Rolling Process

This thesis is based on the project of the titanium alloy rod rolling line. The characteristics of bar rolling technology, rolling method, the application in the field of intelligent optimization control technologies and the model of the rolling process are summarized and analyzed. Due to the increasingly high demand for domestic and foreign users of the product quality, accurate control of bar rolling is particularly important. Main research methods of this thesis include speed optimization control and speed compensation of Roughing mill based on NMPSO algorithm, tension fuzzy optimization control of intermediate mill and finishing mill, bar rolling force prediction algorithm based on ACPSO optimized SVR and fault diagnosis of rolling process. The main research methods are used to actual rolling and have achieved good results. The main research contents and conclusions are as follows:(1) Combined with the actual production condition, the composition of titanium alloy rod rolling system and the framework of overall structure are studied. The mathematical model of rolling process included dc motor model, tension model, rolling force model, deformation resistance model and the temperature model.(2) Aiming at the size fluctuation problem of the inlet of intermediate mill causing by rough mill’s speed mismatch, an optimization algorithm based on NMPSO optimized speed PID and compensation strategy are proposed. Firstly, the principle of PSO algorithm and non-uniform dynamic mutation particle swarm optimization algorithm are studied, and then the Rosenbrock function and Rastrigrin function are tested. Test result shows that NMPSO algorithm convergence speed is faster than the standard PSO algorithm. The simulink model of Roughing mill speed output and load disturbances is presented finally. The model combined with NMPSO algorithm is applied to roll speed compensation research. Simulation results and experimental showed that optimization and compensation strategy of speed can overcome the speed drop and which can make rough rolling process smooth.(3) Aiming at the heap materials and pull material phenomenon caused by tension’s mismatch, an optimization algorithm based on ISTPSO optimized fuzzy tension PID control algorithm is proposed. Due to the mutual coupling of speed and tension, it is not easy to establish accurate mathematical model. Fuzzy control algorithm does not rely on accurate mathematical model of controlled object, and which has good robustness and adaptability, so tension controller optimized by fuzzy algorithm in the intermediate mill and finishing mill were used. In order to overcome the disturbance of the actual control, an1STPSO algorithm was proposed which can optimize the quantization factor and scaling factor. Combined with tension soft measurement method, the ISTPSO algorithm is applied to rolling system, the simulation and experimental results verify the effectiveness of the proposed method, and which solved the heap materials and pull material phenomenon caused by tension fluctuation.(4) Aiming at the accurate prediction problem of titanium alloy bars rolling force, an optimal approach of support vector regression (SVR) parameters is proposed based on the accelerate convergence particle swarm optimization (ACPSO) algorithm. Firstly, this algorithm improve convergence and convergence speed through making particle’s speed deviate from a small angle in each speed iterative process and making particle’s position deviate from a small step in each position iterative process. Secondly, ACPSO algorithm is applied to optimize the three parameters synchronously, which make the ACPSO-SVR model good prediction accuracy and generalization capabilities. Through simulation experiment and particle date comparison, the validity of the method is validated. The results show that the ACPSO-SVR algorithm can effectively predict rolling force, and is superior to PSO-SVR at prediction speed and adaptability, and the ACPSO-SVR algorithm is better than BPNN, SVR and PSO-SVR in precision of prediction and the average error rate decreases from±9%achieved by the BP neural network to less than±4%by using ACPSO-SVR algorithm.(5) Under the background of application of titanium alloy rod continuous rolling process, combined with lifting wavelet data noise reduction method, LW-RLSSVM algorithm and LW-PNN algorithm are put forward, which are used for the fault diagnosis of rolling process. Simulation results verify the validity of the proposed method, and enhance the reliability of bar rolling optimal control system.The content of this research is focused on the control of titanium alloy rod rolling process, which has a strong practical. The research result in the thesis has been confirmed in the work field and the production is improved.