An online hybrid (combined) approach for control loop performance assessment mainly based on normalized Harris-CLPA index, on linear principal component analysis (PCA) and on neural network predictor, is the main contribution proposed in the paper. The proposed approach permits the detection and isolation of intermittent faults, transients in control loops and PI(D) controller detuning, for single-input single-output systems. Experimental results obtained with the nonlinear three-tank system (DTS-200 benchmark) using a PI(D) controller are presented, in order to evaluate the overall system performance.