| The control of material-flow using an adjustable blank holder force (BHF) enhances part quality and consistency in stamping. Using the punch force as an in-process variable, a process controller can adjust BHF based on tracking a reference punch force. The goal of this research is to investigate the design and implementation of a multi-input multi-output (MIMO) adaptive stamping process controller to improve tracking performance and to improve part quality in the presence of plant variations during the manufacture of stamped parts.;All procedures to develop a process model, to parameterize the model using system identification, and to validate experimentally the MIMO process controllers designed through simulation are presented. A MIMO fixed-gain PI process controller can be used to adjust BHF based on tracking of a reference punch force to improve part quality. However, experimental tests with a fixed-gain PI controller required manual controller fine-tuning, which can be time-consuming and expensive.;Auto-tuning and model reference adaptive control (MRAC) are investigated not only to eliminate the manual tuning but also to improve part quality, in the presence of process variations. The direct MRAC, whose controller gains are adjusted to accommodate changes in process variations, is presented, including simulation-based robustness analysis of the adaptation law with constrained estimation. In experimental validation with the MRAC process controller, good tracking performance and significant part quality improvement are achieved in the presence of disturbances.;However, the direct MRAC filter uses nominal process parameters, and so requires some knowledge of the process. Consequently, an indirect adaptive control (AC), which does not need the nominal process parameters, is also considered. However, due to the simple PI controller selected, the computation of the controller gains from the estimated parameters requires an optimization, which is not amenable to real-time implementation. Hence, the indirect AC is implemented using a look-up table, with controller gains that are pre-computed off-line via optimization. The indirect AC, with a sufficiently high level of discretization in the look-up table, performs well in simulations. However, due to extensive memory requirements, a smaller look-up table is used in the experiments, where it is outperformed by the direct AC. |
