Diese Website verwendet Cookies. Mit der Nutzung der Website stimmen Sie deren Verwendung zu. Weitere Informationen erhalten Sie in unserer Datenschutzerklärung.

Statistical Analysis of Process Chains: Novel PRO-CHAIN Components

The robustness of production processes and the quality of resulting products suffer from variations in important material and process parameters, geometry and external influences, which can have substantial and critical influences. Therefore these variations have to be analyzed and transferred over process steps in order to achieve considerably better forecasting quality. We developed the PRO-CHAIN strategy for statistical analysis of sensitivity and stability as well as multi-objective robust design-parameter optimization of whole process chains, even for simulation results on highly resolved grids. PRO-CHAIN constructs an ensemble of simulation results; this data base reflects local variations of functionals. Newly developed PRO-CHAIN components deal with transforming and ensemble compression of the data base via a fast principal component analysis with user-controlled accuracy. Essential features are the classification of design parameters into importance and nonlinearity classes in order to reduce the design space and to get an adequate accuracy for an efficient optimization. In this paper we address the importance of this classification and appropriate kinds of classification measures. Another main novel PRO-CHAIN component is the fast and accurate interpolation of new designs on the whole grid. This interpolation works also for nonlinear applications like crash if the design of experiments is adequate for a high-quality metamodel. The interpolation is based on a nonlinear metamodel with radial basis functions accelerated by a specialized principal component decomposition. Summarized, PRO-CHAIN is now able to fully locally analyze a chain consisting of several process steps with regard to sensitivity and robustness and to predict new designs with user-controlled accuracy. In each step, the influence of parameters onto criteria is classified and sensitivity is measured. PRO-CHAIN is able to propagate the essential scatter due to parameter uncertainty locally over the steps, keeping the necessary number of simulation runs small. Additionally, PRO-CHAIN allows for predicting new designs fully locally, allowing for immediate answers to what-if scenarios, without additional time-spending simulation runs. Thus PRO-CHAIN is a very efficient strategy for statistical analysis of process chains, involving parameter uncertainties, in order to get a robustly optimized solution. Recently, we integrated the efficient interpolation method described into DesParO along with LS- DYNA d3plot readers/writers: on one hand, as a so-called “mixing functionality” for constructing and dumping interpolated results, on the other hand into the novel DesParO Geometry Viewer. Now, DesParO allows for an interactive exploration of the design space, connected with direct interpolation and visualization of the new design and its functionals, like thickness, effective plastic strains and damages as well as statistical measures, locally on the whole grid. Results are presented for the forming-to-crash process chain for a ZStE340 metal blank of a B-pillar. In detail, results of importance and nonlinearity classifications in each process step are shown as well as the prediction of new designs by means of DesParO.