Simulation has become a critical part of the design process for engineers around the world. Coupled with advanced design optimization software, it can drive development to improve product performance significantly. However, simulation predictions are meaningless if the underlying model can’t first be matched to real-world baseline performance. Calibration of the simulation model is a critical first step. In this process, responses from real-world experiments are compared with simulation responses. Then simulation input parameters are adjusted until the simulation results match the experimental ones.
Using trial and error, it can be difficult and time-consuming to decide which calibration adjustments need to be made. With the automated simulation calibration tools within ANSYS optiSLang, you can conduct sensitivity analysis to determine which parameters have the most influence on the simulation results, and determine the relationships between the inputs and responses. Advanced tools in optiSLang generate charts showing these relationships and predicting the correct calibration.
During a calibration, additional input parameters that are unknown or associated with uncertainties should be identified. Automating the identification workflow can save substantial costs compared to manual trial-and-error procedures. The procedure should also filter for the most relevant variables and generate the best possible metamodel for each response with a minimum number of solver calls. Inverse strategies are generated by using simulation models that correspond to the existing experimental geometry, constraints and test procedure. The unknown parameters will be determined by an iterative calibration between experimental data and simulation results. After a suitable set of input parameters and result variables are identified, a successful model calibration can be conducted using global and local optimization methods, such as gradient-based or adaptive response surfaces, as well as evolutionary and genetic procedures.
The analysis of the significance and sensitivity of the input parameters in the calibration process also reveals opportunities for system and product optimization. These opportunities give design engineers the freedom to explore “what-if” variations and their tolerances to fulfill customer requirements. Design study results can be used as reduced order models (ROMs) that enable engineers to respond quickly to changing customer concerns or requirements, and instantly determine if available designs will satisfy those requirements or if a new design is needed. All methods are multidisciplinary and can be integrated into the virtual development processes of all industrial fields.
If you would like to learn more about efficient parameter identification, please attend the 15th Weimar Optimization and Stochastic Days 2018 conference, June 21–22. Experienced engineers will be discussing these techniques and sharing ways in which you can apply CAE-based sensitivity analysis, optimization and robustness evaluation using optiSLang.
More information about the conference can be found at https://www.dynardo.de/en/wost.html.
The post High-Caliber Calibration Yields More Representative ROMs appeared first on ANSYS.