The development of renewable and eco-friendly biolubricants can address the environmental challenges posed by petroleum-based lubricants. At the same time, it is possible to improve the tribological properties of lubricants through alternative sources. To overcome these problems, castor oil is a potential basis for biolubricants due to its high viscosity, natural lubricity, and biodegradability. In the current work, castor oil was chemically modified by the epoxidation process. This process has improved the tribological properties of castor oil through the epoxidation method. In this method, the presence of hydrogen peroxide, acting as a catalyst, converts the unsaturated double bonds present in the oil into oxirane rings. At the same time, this modification enhanced the thermal stability and tribological applications in harsh operating conditions. The tribological behavior of the bio-lubricant was evaluated using a four-ball tribometer test. In the present test, pressure, temperature, nanoparticle concentration, and rotational speed were considered as process parameters, while the Coefficient of Friction (COF) and Wear Scar Diameter (WSD) were evaluated as the primary responses. The Response Surface Methodology (RSM) and Artificial Neural Network (ANN) models were adopted for planning, analyzing, and validating the factors and their interactions in the experiments. The RSM model developed to optimize the COF showed a high correlation of R² = 0.999, which was found to be highly consistent with the experimental results. This model provides more than 90% prediction accuracy and also highlights important interactions between various factors. The best operating conditions were determined to be 0.2 wt.% CuO nanoparticles, 1300 rpm, 25 mbar, and 120 °C, at which a minimum COF of 0.0409 and 0.6288 mm WSD was obtained. The Levenberg–Marquardt training algorithm of the ANN model was developed for predicting and validating data of RSM. The dataset was divided into 70% training, 15% validation, and 15% testing sets. A single hidden layer with 10 neurons was inserted in the network structure. Following the experiments, high prediction accuracy is achieved for both COF and WSD, with an error rate of less than 1%. The results show that CuO-reinforced castor oil significantly improves friction and wear performance and may prove to be a consistent alternative to conventional lubricants. The combined use of RSM and ANN provides a reliable path of process optimization for bio-lubricant development.
Keywords: Artificial Neural Network, Bio-lubricants, Coefficient of Friction, Response Surface Methodology, Wear Scar Diameter.