Validating numerical models - Quantifying error

The validation of numerical models often progresses incrementally from previous models or other numerical solutions or is undertaken by comparison with experimentally obtained reference measurements. Notwithstanding the accuracy of the reference results, quantification of the error between the two is important information in deciding the quality of the model. It is frequently the case that this estimate of error is done by eye. However, for purposes of traceability and objectivity, interest has started to focus on techniques to quantify this error in an algorithmic manner in a way that agrees with the general observations of experienced engineers. This paper reviews two of the most promising techniques, namely Feature Selective Validation (FSV) and Integrated Error against Logarithmic Frequency (IELF), putting them in the context of correlation and reliability functions.