An Adaptive Artificial Neural Network Model for Predicting Friction and Wear in Polymer Matrix Composites: Integrating Kragelsky and Archard Laws

This study presents a hybrid modeling approach that integrates Kragelsky’s friction law and Archard’s wear law with an artificial neural network (ANN) to predict the coefficient of friction (COF) and specific wear rate (SWR) in epoxy-based self-lubricating composites reinforced with graphite and MoS₂. Given the complex, nonlinear interactions among tribological parameters such as contact pressure, sliding speed, hardness, and filler composition, traditional empirical models often fail to capture wear behavior accurately. The proposed ANN architecture comprises an input layer, three hidden layers employing sigmoid, ReLU, and power activation functions, and an output layer predicting COF and SWR. The network is trained using a feed-forward method with backpropagation to minimize prediction error. SEM analysis reveals that graphite imparts superior wear resistance compared to MoS₂. The ANN achieved significantly higher prediction accuracy for graphite-reinforced composites. For COF, graphite yielded an MSE of 0.00073 and R² of 0.9047, while MoS₂ showed an MSE of 0.00318 and R² of 0.5567. For SWR, graphite attained an MSE of 1.3351 and R² of 0.9809, compared to MoS₂ with an MSE of 1.6993 and R² of 0.8271. The reduced performance in MoS₂ predictions is attributed to its oxidative degradation forming MoO₃. The model also offers 3D surface simulations, aiding in composite design optimization and reducing experimental costs.