ADS-79D-HDBK Downloaded from https://www.everyspec.com In the modeling perspective (shown as the left branch), a mathematical model representing the physics nature of the target problem is developed, and detailed numerical algorithms are implemented into the computational model. Verification of the computational model is performed to check functionality and reliability of the code. Once the fidelity of the computational model is satisfied, numerical simulations are performed to generate numerical outcomes. Further verification of the computational model is followed to ensure desirable robustness and accuracy of the code.
Often, actual data and/or case studies with known solutions are used at this stage. Due to the inherent scatter associated with input parameters and model assumptions and simplification of the mathematical models resulting from incomplete knowledge of the physical nature of the problem, variability in the response of interest is often expected and rigorous uncertainty quantification needs to be carried out. The simulation outcomes obtained from verified and calibrated computational frameworks form a database for further comparison against the outcomes of physical modeling. The right branch in Figure K-3 highlights the process of physical testing.
The essential element in the process is to validate mathematical models through physical experiments. The purpose of validation experiments is to generate baseline ―ground truth‖ data needed to assess the accuracy of the mathematical model; therefore, all assumptions behind the mathematical model should be understood, well defined and controlled. The process starts with a basic physics-based model describing the underlying problem. Often, preliminary assessment using the computational model would be performed to identify the critical factors for further experimental design.
During this stage, preliminary sensitivity analysis and associated uncertainty quantification are performed to help determine the location and type of measurements needed for validation. Once a validation testing plan has been established, the experiments will be performed and measurement data from various instruments will be collected and further processed. The extracted features will form the baseline for direct comparison with the simulation outcomes. Repeat experiments are generally required to quantify uncertainty due to lack of repeatability and inherent variability.
The experimenter then performs uncertainty quantification to quantify the effects of various sources of uncertainty on the experimental data. Once experimental outcomes and simulation outcomes for the actual test conditions have been generated, the validation assessment will be conducted by comparing these two sets of outcomes. The metrics for comparing experimental outcomes and simulation outcomes, as well as the criteria for acceptable agreement, will have been specified in V&V plan stage. The degree to which the model accurately predicts the data from validation experiments is the essential component of the overall assessment of the model‘s predictive capability.
This leads to further accredit the benefit to be obtained from a CBM using the predictive model. Note, however, the ―acceptable agreement‖ provides a subjective decision point for initiating improvements in the conceptual, mathematical, and computational models and in the experimental designs. Sometimes, the outcomes of simulation and physical observation may not result in agreement. If the discrepancy between the mathematical model and physical reality exceeds a pre-defined acceptance limit, the outcomes from the mathematical model and the physical model need to be carefully examined.
The key assumptions for the conceptual model, its mathematical representation, and the computational framework must be reviewed. The scope and setup of 194