Options
2015
Conference Paper
Titel
A concept for sensitivity analysis and parameter calibration of coupled nonlinear PDEs and its application to an industrial glass forming model
Abstract
Many industrial and environmental processes are characterized as complex spatiotemporal systems. Such systems, which often modeled with nonlinear coupled PDEs, are often highly complex and their relationships between model inputs, parameters and output may be poorly understood. Moreover the solutions of physics-based models commonly differ from the real measurements. Hence, aim of this work is the development of a concept which provides support in understanding of the system behavior and the parameter calibration. Recently the simulation considering uncertainties in models known as uncertainty quantification framework is an active research area. The uncertainty quantification framework can perform the sensitivity analysis of the uncertain influenced parameter to the quantities of interest. The probabilistic description offers also the Bayesian inference to handle the discrepancy between the model prediction and the real measurement. The efficient numerical solution is achieved by means of the generalized Polynomial Chaos expansion (gPC). With the gPC we perform the global sensitivity analysis and the model parameter calibration. We introduce the calculation of the local sensitivity analysis, which normally computed by analytical differential, by using the gPC as surrogate model. The considered industrial process in this paper is a complex rheological forming process producing glass tubes and accordingly rods which are pre-products for optical fibers. The material parameters of the process are temperature dependent that lead to nonlinear PDEs. These dependencies are mostly empirical, thus the parameters are considered as the uncertainties of the model. The sensitivity analysis results improve the understanding of the process, i.e. the coupling of the parameters to the outputs. By the means of the concept parameters were optimal calibrated to fulfill user defined performance criteria.