PREDICTING TRIBOLOGICAL PERFORMANCE OF GRAPHITE-PLUGGED BRONZE BUSHINGS USING AN ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM
Ali K. A. Aljboury, Amir Alsammarraie, Maki H. Zaidan
Abstract: Graphite-plugged bronze bushings are widely used in industrial applications where conventional lubrication is not feasible, yet predicting their tribological behavior remains difficult due to the nonlinear interactions among material composition, plug geometry, and operating conditions. This paper presents two independent ANFIS models—one for coefficient of friction (COF) and one for material loss—developed and validated against dry sliding experimental data from CuAl10Fe5Ni5 bushings running against an AISI 410 stainless steel shaft. The training dataset came from 81 full-factorial tests across four input variables: graphite surface coverage (10, 20, 30%), plug diameter (8, 10, 12 mm), rotational speed (250, 500, 750 rpm), and applied load (50, 100, 150 kg). Each model uses a zero-order Takagi-Sugeno-Kang fuzzy structure with triangular membership functions on a 3×3×3×3 grid, producing 81 IF-THEN rules, trained over 200 epochs with a hybrid least-squares/backpropagation algorithm. MATLAB ANFIS training converged to final RMSE values of 0.00097889 for COF and 0.00636561 g for material loss. Validation against three previously unseen operating conditions produced maximum prediction errors of 1.20% and 3.927%, respectively—indicating satisfactory interpolation accuracy within the investigated parameter range. The results suggest that ANFIS can accurately predict the nonlinear tribological behavior of graphite-plugged bronze bushings and can reduce the extent of physical testing required during preliminary design evaluation across the interpolated operating range.
Keywords: ANFIS; Graphite-plugged bronze bushings; Coefficient of friction; Material loss; Tribological prediction; Self-lubricating bearings
DOI: 10.24874/PES08.03.020
Recieved: 03.06.2026 Revised: 10.07.2026 Accepted: 18.08.2026
UDC:
Reads: 2 
