A post-processing method for feature sensitivity coefficients

An economical postprocessing method for estimating feature sensitivity coefficients is presented. This method uses precomputed elementary sensitivity coefficients calculated by standard methods to obtain a perturbed solution using a linear approximation to the Taylor series expansion. The perturbed solution is then used for simple feature sensitivity calculations. This approach avoids costly repetitive integrations for each model parameter and can be readily implemented with available computer codes which solve for species profiles and elementary sensitivities. The method was tested using a practical example of a CH flame profile and shown to be in good agreement with a brute-force sensitivity analysis performed for the same conditions. It is also suggested that the method can be readily extended to simultaneous sensitivity analysis of multiple responses or to sensitivity analysis with respect to a functional combination of model parameters. © 1995 John Wiley & Sons, Inc.