Estimation Of Heat Transfer Coefficient Of Binary Vapor Mixtures Condensation On Horizontal Tube By Using SVM

Heat transfer coefficient is one of the most important parameters in the thermalhydraulic design and safety analysis of various heated equipment such as evaporator,fuel cell cooling system, nuclear reactors, etc. When binary vapor mixtures of positivesystem condenses on a solid surface, the Marangoni condensation, which caused bythe surface tension difference associated with the local concentration and temperaturedifferences, would be formed. Marangoni condensation is significative to improve theheat transfer performance and efficiency in a number of industries, such as lowtemperature heat resource, geothermal energy, oil gas and aviation aerospace.Water-alcohol binary vapor mixtures condensation phenomenon on horizontal tube isa typical Marangoni condensation phenomenon. In order to analysis the heat transfercoefficient. In this article the effects of vapor pressure, vapor concentration on thecondensation heat transfer characteristics of ethanol-water mixture vapor onhorizontal tube were investigated experimentally. Visual results were showed in thisexperiment. And condensation forms of water-alcohol binary vapor mixture related tosuper-cooling degree and concentration of alcohol. While the prediction of heattransfer coefficient has high requirements, it’s need to solve the small sample learningand the non-linear relationships problem between heat transfer coefficient andsuper-cooling degree. By the characteristics of support vector machine (SVM), it isvery suitable for application in this problem. This paper introduces the construction ofthe experiment system, deduced the calculation formula of coefficient of heat transfer.At the same time, this paper introduces the principle of support vector machine (SVM)and the application of genetic algorithm of parameter optimize for v-SVR, applyingthe model to predict the heat transfer coefficient. Results show that the model and theexperimental measurement data, a better precision is higher, verify the correctnessand feasibility of the model. Results show that the model prediction and experimental measurement data better match, high precision, verify the correctness and feasibilityof the model. And analyzed prediction results under different pressure and differentalcohol concentration. Compared the result of using genetic algorithm and particleswarm optimization (pso) for parameter optimization according to the effect of themodel of contrast. Found the v-SVR model based on genetic algorithm has betterfitness and accuracy.