An imminent bearing fault detection can reduce downtime or avoid the failure of rotating machinery. Modern condition monitoring demands simple but effective bearing failure diagnosis by integrating dynamic models with intelligence techniques. In view of this, an integrated approach of an Adaptive Neuro-Fuzzy Inference System (ANFIS) and Dimensional Analysis (DA) is demonstrated to diagnose the size of the bearing faults. Several experiments were performed with artificially damaged bearings at different operating conditions on the developed rotor-bearing test rig. With a 0.1 mm error band to fix minor bugs, a two-performance indicator evaluated the model accuracy. Comparison of the performance of DA and ANFIS models with experimental and Artificial Neural Network (ANN) validated the potential of the present approach. The results showed that the simplicity of the DA and the superior performance of the ANFIS contributes to detecting bearing fault effectively and accurately in the diagnosis of industrial machines. © 2020 Elsevier Ltd