Control performance assessment (CPA) is an important strategy to establish the quality of industrial control loops. However, most CPA methods are mainly restricted to linear systems, ignoring common process nonlinearities such as valve stiction. To tackle this problem, a control performance assessment method based on the improved radial basis function (RBF) nonlinear timeseries model in the presence of valve stiction nonlinearities was proposed in this paper. By applying the Hinich test method into the RBF learning, which guarantees the removing of the nonlinear part of the output, the performance benchmark was estimated using standard time series identification techniques. The simulation of an integral process verifies the validity of the proposed method.