Parametric studies in CIVA

CIVA software enables to model and simulate a configuration of non-destructive evaluation with very satisfying performances in terms of computation time. These performances make the intensive use of CIVA affordable in the frame of parametric studies, for example with the aim to optimize and/or justify the performances of an inspection method. These tools strengthen the role of the simulation in which a model parameter can easily be modified, whereas it can become very costly and time-consuming in the frame of experimental studies.

The design of a parametric study (available for all the NDE techniques modeled in CIVA) starts with the definition and simulation of a nominal configuration of interest. Once this file is defined and computed, the user can go to the “Parametric study” tile from CIVA Desk, and initiate the parametric study by loading the nominal case as reference. A new configuration of *.var extension is then automatically created by CIVA, and the user has the possibility to select, using a right-click, any numerical parameter available in CIVA GUI to include it in the design of experiments of the parametric study.

 

 

In the “Variation” tab, the user can then define the design of variations of the studied parameters depending on different sampling methods: deterministic distributions (for example with linear or logarithmic spacings, following a given statistical law, or a manual draw), random distributions based on various statistical laws, or a parameter distribution ruled by a formula. Regarding the analysis, a data post-processing (calibration, signal processing, etc.) can also be applied to the simulated results of each combination of parameters. Moreover, various extraction features are available, in order to evaluate the impact of the parameters set on a given analysis criterion, for example the maximal amplitude of a defect signal, or even custom criteria which can be specified by a Python script (available with the “CIVA Script” feature).

 

 

Once the parametric study is computed, the user can access to each individual simulation results for each studied parameters set (if asked before computation), and to curves and images allowing to visualize the influence of the studied parameters on the defined analysis criteria.

 

 

Before or after computations, the user can also benefit from advanced features for parameters sampling and results interpolation (metamodel), in order to predict continuous variations of the analyzed output outside the computed parameters values (between min and max bounds initially defined by the user). These advanced features allow to predict the influence of the studied parameters on the analysis criteria very easily and efficiently, without having to simulate every combination of parameters values (which can become very time-consuming depending on the number of parameters studied). Such features also give access to sensitivity analysis tools, which can be very useful to highlight the most influential parameters on the analyzed output, and they can also be very valuable for reliability studies (POD curves, search for the “worst case” scenario, etc.) and/or optimization work of the simulated END method.

 

 

Do not hesitate to play with these very effective features to carry out your parametric studies in CIVA! We remain at your disposal in case of any question through CIVA support, and we also provide practical exercises about these specific features during our CIVA training sessions and “Reliability in NDE” training course.