Experimental analysis of recycled engine oil tribofilms synthesized at mild temperature and comparison of optimal conditions using ANN and RSM

• Engine Oil Degradation: Examines SAE 15W-40 oil degradation over 13,500 km, showing increased friction and wear compared to fresh oil. • ZDDP Antiwear Additive Depletion: Analyzes ZDDP breakdown, revealing reduced phosphorus and zinc, and increased iron in tribofilms. • Tribofilm Composition Changes: Fresh oil forms tribofilms with higher phosphorus and zinc; used oil shows increased iron, reducing wear protection. • Friction and Wear Performance: Used oil samples show deteriorating friction and wear properties due to ZDDP depletion. • Optimization of Used Engine Oil: Uses RSM and ANN to optimize recycled oil yield, achieving 65.2% under optimal conditions. • Comparison of RSM and ANN: ANN demonstrates higher accuracy in predicting recycled oil yield compared to RSM. This study examines the wear and friction characteristics of used 15W-40 engine oil throughout its lifespan, focusing on tribological films in various oil states. The oil was drained from a turbo-boosted petrol engine, which has 420% more wear performance and 9% more friction than fresh oil. The degradation of the ZDDP (zinc dialkyl dithiophosphate) anti-wear additive was also analyzed, revealing a correlation between ZDDP depletion and increased iron levels along with reduced zinc and phosphorus levels in the tribofilms. Artificial neural networks (ANN) and surface response methodology (RSM) were used to optimize the output of recycled oil with the range of 15W-40.The ANN model used three input layers, ten hidden layers, and one output layer (3-10-1), whereas the RSM used a Box-Behnken design to create design matrix. Temperature, solvent-to-oil molar ratio, and reaction duration were process variables. With a molar ratio of 1:5.5 and a reaction period of 50 minutes, the optimum conditions yielded a maximum of 65.2% recycled oil. RSM analysis revealed a coefficient of determination (R 2 ) value of 0.9588 and an adjusted R 2 value of 0.928. The ANN with an R-value of 0.9993 and a mean squared error of 4.48e-18 were trained using the Levenberg-Marquardt technique. Fourier-transform infrared spectroscopy (FTIR) assessed the efficacy of the activated carbon catalyst, while gas chromatography-mass spectrometry (GC-MS) confirmed that presence of important components (natural base oil, additives, and) in the recycled oil. The recovered oil's physicochemical characteristics compared with ASTM standard.