This paper is about a regularization method applied to a chemical process considered as a nonlinear system. The Volterra model is used to represent such a nonlinear system. Volterra kernels are expanded on generalized orthonormal function bases in order to provide a quite flexible and accurate model. Tikhonov's regularization method, optimized on a sliding window, is then applied to the Volterra model in order to restore an unknown quantity from measurements of another quantity. Such an inverse method is then applied to a simulated pollutant oxidation process