Plume phenomenology code suite validation via code-to-code sensitivity propagation techniques

A chain of plume codes (e.g., rocket motor, flowfield, radiation atmospheric propagation) is required to properly model plume phenomenology. These codes have hundreds of user-selected input parameters and include close to a million database parameters such as molecular cross section, gases thermodynamic properties (JANNAF), chemical kinetic rate constants, etc. In practice, there are large ranges of uncertainties associated with these parameters, and running the models requires expertise in selecting the right values for the input variables. Verification and validation of the chain of codes can be accomplished by attempting to match the model predictions to actual sensor measurements data. However, such a process is largely one of trial and error that can take months, if not years. OKSI has demonstrated an automated technique for accomplishing this arduous process via enhancement of the plume code by automatic differentiation (AD) whereby the codes are used to compute the full Jacobian of the output sensitivities, with respect to the various inputs. With the enhanced code, it is possible to solve a multidimensional optimization problem, improving the accuracy of the computer model predictions as compared to the actual sensor measurements by automatically "tweaking" the model inputs. Without enhancing the codes, this optimization solution would take hopelessly long to converge on a solution. However, the technique to automatically propagate sensitivities (and uncertainties) among the codes in a chain is still lacking. Under Phase-I, this capability will be demonstrated with two plume phenomenology codes. The approach will be expanded under Phase-II to include a complete suite of codes starting from engine models and including flowfield, radiation, atmospheric propagation, and other such models.