An Investigation Of Control For A Virtual Low Pressure Aluminum Wheel Die Casting Process

The low pressure die casting process is the main method for production of aluminum wheels. The predominant defect found in die-cast wheels is porosity. These defects have a relation with the die temperature. Stabilizing the die temperature can reduce the defect. Process modeling is becoming a key technique to improve and optimize industrial processes. And model predictive control (MPC) is an advanced control methodology which is making a significant impact on industrial control engineering. The die casting process is a batch process, and MPC can be used to control the process.By using process modeling, this thesis developed a control strategy for low pressure aluminum wheel die casting process. And an advanced control strategy, MPC, was used as the control solution is. Based on ProCAST and MATLAB, the development is entirely offline, the expenses associated with developing such a system in an operating plant are not incurred.Based on the 3-D geometry model of a wheel, a 3-D finite element model was developed. The filling and solidification process of the wheel were simulated by ProCAST, and the simulation could predict the temperature history of the die and the forming of choking defect. A system was developed to transform the ProCAST simulation into a virtual process with communications. The communication function allowed the reading/writing of process data and control data to and from the virtual process. By using a Perl script, the system enabled the single cycle ProCAST simulation to be run repeatedly to simulate a continuous cyclic process.The die casting process was a batch process, and its process inputs and the process outputs could be well represented by a state-space model. N4SID was a kind of system identification method, and was used to generate the state-space model derived from the input-output data of the casting process. To perform the system identification, the process variables, the control variables and the feedforward variables of the casting process were determined. And then a state-space model was generated from the input-output data. By validation, the identified model could accurately predict the input-output behavior of the process. Based on this model, by using the feedforward variables the MPC controller could predict the effects of the disturbances before the disturbances propagated through the process. To optimize the closed loop performance of the low pressure die casting process, the MPC controller was tuned, and the weighting matrix in the cost function, the prediction horizon, and the control horizon were defined. And the control variable constraints were also specified. The implementation of the MPC controller was implemented in the MATLAB.To confirm the effect of implementation of MPC for low pressure die casting process, the MPC controller was evaluated on the virtual process. To evaluate the control solution, three disturbance scenarios were applied to both the controlled and uncontrolled virtual process. The first disturbances was a ramping metal temperature input disturbance, which simulated the temperature of the molten metal varies in the industrial process. The second scenario was variations in the length of time the die remained open after the wheel was ejected from the dies. The third scenario is a combination of the metal temperature and die open time disturbances. These disturbances caused the die temperatures to deviate from their optimal values, which may result in defective wheels if the deviations were large enough. The evaluation results shows that, the control solution developed improved the process performance, could well reject the simulated disturbances and maintain the die temperatures near their optimal values. And the choke volume generated was also reduced, resulting the reduction of defective wheels.