Due to its continuous motion, control valve performance tends to deteriorate over time due to the presence of static-friction or also known as stiction. This, in turn, leads to high variability in product quality and an increased frequency of valve maintenance. Model Predictive Control (MPC) based stiction compensation methods can remove oscillations caused by stiction but it assumes that stiction is known a �priori to exist in the related loops. To overcome this limitation, an integrated framework is proposed to automate the detection of stiction using only process variable and controller output while compensating for stiction with MPC. The detection algorithm, which was validated using industrial data, uses an adaptive neuro-fuzzy inference system (ANFIS). Out of the 78 benchmark industrial loops tested, the proposed Intuitive ANFIS-based Method (IAM) has a detection accuracy of 65, placing it on par with the best of the currently available methods reported in the literature for loops with stiction. Within the proposed framework, the detection component only activates the MPC-based compensation when needed. In a simulation of a multivariable process, it is demonstrated that the dual-mode MPC manages to eliminate oscillation caused by stiction with no chattering. This results in better overall performance when controlling a loop throughout the service life of the valve. © 2019 Elsevier Ltd