Research On Lubricant Formulation Optimization By The Artificial Neural Networks Combine With Genetic Algorithm

This article involves lubricant formulation research field and mainly studies the application of the artificial neural networks (ANN) and genetic algorithm (GA) in lubricant formulation optimization research. First it studies the fundamental principle of ANN and GA from internal and external pertinent materials, emphasis is put on the analysis of the structure, construction methods, operation procedure and realizing on MATLAB platform of BP ANN and GA, put forward the basic idea of combination of ANN and GA. Then build a BP ANN simulation model of viscosity-temperature performance of the SAE CI-4OW-40grade diesel engine oil on MATLAB platform base on the sample data, which was collected from a testing program designed by uniform design method, and evaluate the generalization ability of the model. The evaluation results show that the BP ANN model has higher fitting precision and prediction precision.Next, with the viscosity-temperature performance as constraint condition, with formulation cost as optimization objective, with the addition of composite additive as constant, with the additions of composite base oil and viscosity index improver as variable, combine with the BP ANN simulation model, the GA which can optimize SAE CI-4OW-40grade diesel engine oil formulation is built on the MATLAB platform. The results show the optimum formulation is53.16%PAO,21.14%TMP,10.7%viscosity-temperature improver,15%CI-4grade composite additives and the formulation cost is about6times mineral oil cost. At last, evaluate viscosity-temperature performance of the optimum formulation and results show that three viscosity-temperature performance values of the optimum formulation:kinematic viscosity(100℃), pour point, viscosity index, are consistent with measured values basically and meet the requirements of related quality standards of OW-40grade diesel engine oil from GB11122-2006. All above explain the formulation optimizing method is feasible and reach the optimization objective.