Sensitivity analysis in nuclear power engineering

Within the reactor analysis and design calculation, sensitivity analysis offers to a nuclear engineer unique insight into investigated system. Estimation of the change of the system response, due to change in some input parameter, can identify important processes and evaluate the influence of variation in this parameter. Defining the critical eigenvalue or reactivity as our response, allows us to use the perturbation theory to determine the sensitivity coefficients of the system. Main constraint in this methodology is the assumption that a considered perturbation is small enough to cause a change in the neutron flux. Otherwise it is not necessary to perform new calculation for the perturbed system, which decreases the computational time requirements. The sensitivity coefficients can be further used for calculation of k_eff or other response uncertainty coming from cross section data; or sensitivity coefficients of reactivity response by using critical eigenvalue sensitivities for perturbed and unperturbed state. The responses which can be addressed with perturbation theory are; multiplication factor, peak to average power, control rod worth, breeding ratio, delay neutron fraction and moderator void coefficient. Sensitivity analysis methods offer to nuclear engineer variety of application. Easily can be identified important processes, reactions or isotopes through sensitivity coefficients, presented in Fig. 1. By the direct perturbation calculation method, evaluator can validate his results calculated by SPT, which was demonstrated in Tab.1. The exact perturbation is able to decomposed a reactivity effects to contributed processes. By using two sets of sensitivities, describing some reactivity change, sensitivity of this reactivity to cross section data can be calculated. Finally, all sensitivity profiles are suitable for evaluation of the corresponding response uncertainty coming from the cross section data. But final uncertainties dependent mainly on the covariance data because sensitivities in this process serve as a weighting function. (authors)