 
Summary: 1
Polynomial regression with derivative information in nuclear reactor
uncertainty quantification*
M. Anitescua
, O. Rodericka
, P. Fischera
, W.S. Yangb
a
Mathematics and Computer Science Division, Argonne National Laboratory, Argonne,
IL, USA
b
Nuclear Engineering Division, Argonne National Laboratory, Argonne, IL, USA
Abstract. We introduce a novel technique of uncertainty quantification using polynomial regression with
derivative information and apply it to analyze the performance of a model of a sodiumcooled fast reactor. We
construct a surrogate model as a goaloriented projection onto an incomplete space of polynomials, find
coordinates of projection by collocation, and use derivative information to reduce the number of sample points
required by the collocation procedure. This surrogate model can be used to estimate range, sensitivities and the
statistical distribution of the output. Numerical experiments show that the suggested approach is significantly
more computationally efficient than random sampling, or approaches that do not use derivative information, and
that it has greater precision than linear models.
