Research On Model Optimization And Nanofluid Property Prediction Based On Data Minin

With the progress of science and technology,heat transfer issues are gradually emerging in various engineering fields such as microelectronics,military,and chemical industries.The excellent heat transfer performance of nanomaterials has also entered the vision of researchers.Adding a single or multiple nanoparticle to a liquid can effectively improve the heat transfer efficiency during convective heat transfer.However,the addition of nanoparticles increases the viscosity of the liquid,and the thermal conductivity and viscosity of nanofluids are influenced by various conditions,which require extensive experimental discussion and analysis,and a large number of experiments not only increase experimental costs but also generate a large amount of waste.Extracting useful knowledge from data through data mining,where machine learning plays a crucial role in data segmentation.Therefore,using machine learning methods to establish and improve the thermal conductivity and viscosity preheating physical properties models of small amounts of single and mixed nanofluids has important research significance.This article adopts a combination of experimental and simulation methods to prepare single and mixed nanofluids,and measures the thermal conductivity and viscosity of the two nanofluids.It introduces several machine learning algorithms in data mining and uses machine learning models to model,predict,and optimize the experimental data.The main research content is as follows:（1）A two-step method was used to prepare 0.08,0.25,0.6,1.1,and 2vol%Ti O₂single nanofluids and 0.02 vol%MWCNTs-Ti O₂/water nanofluids,and the factors affecting their thermal conductivity and viscosity were studied.The results show that the thermal conductivity of nanofluids increases with increasing temperature and volume fraction,while the viscosity decreases with increasing temperature and decreasing volume fraction;It was found that the particle ratio of MWCNTs-Ti O₂ has a nonlinear effect on the thermal conductivity and viscosity in the MWCNTs-Ti O₂mixed nanofluid.（2）Using machine learning algorithms to mathematically model the thermal conductivity of two types of nanofluids and using grid search cross validation method to optimize the model.Ti O₂ nanofluids use temperature and volume fraction as input conditions,while MWCNTs-Ti O₂nanofluids use temperature and mixing ratio as input conditions.Multiple linear regression（MLR）,BP artificial neural network,radial basis function optimized BP artificial neural network（RBF-BP）,support vector machine regression（SVR）Least Squares Support Vector Machine（LS-SVM）predicts the changes in thermal conductivity of two nanofluids,compares the models,and analyzes the impact of different input conditions on the accuracy of machine learning prediction data.Through research,it was found that the LS-SVM model has the highest accuracy in predicting the thermal conductivity of nanofluids among the five models.（3）Five machine learning models were used for viscosity prediction of two nanofluids,with predicted thermal conductivity input at the same temperature and volume fraction as well as temperature and mixing ratio.It was found that the LS-SVM model was found to have the highest accuracy in predicting the viscosity of nanofluids by comparing five models for four evaluation metrics.The LS-SVM model was found to exhibit extremely good results in predicting nanofluids,allowing the establishment of a database of prediction models with the LS-SVM model as the underlying layer.