Based On Neural Network Level Organic Light Tube Working Medium Flow Boiling Heat Transfer Research

Nowadays, the global energy crisis and environmental issues are becoming more and more serious. Hence, people pay more attention to how to save energy, reduce the greenhouse gas emissions, protect the ozone layer, and green low carbon life at present. Flow boiling heat transfer is one of the highly efficient heat exchanger manners in daily life and industrial production. To correctly predict the organic alkyl flow boiling heat transfer performance is very important for the optimization design of evaporators, because the detailed heat transfer characteristics in boiling can avoid the drastic results due to under-design or over-design evaporators. And it is extremely significant for improving energy efficiency and safe operation to the heat transfer equipments. The flow boiling heat transfer of organic alkyl inside tubes became more and more popular in engineering applications, which is a very important branch of the flow boiling heat transfer. However, the mechanism of the flow boiling heat transfer is very complex to represent. For example, the flow boiling influenced by many factors, strong non-linearity, and not yet formed a uniform cognition. The existing flow boiling heat transfer correlations for the refrigerants are all most empirical or semi-empirical, which came from the in-tubes flow boiling heat transfer experiments by the researchers themsleves or others’. However, the existing correlations can not be able to correctly predict in many cases, especially for some new organic alkyls.Artificial neural network is a kind of information processing system based on biological neural network structure features. It can avoid to analysis the complex internal mechanism in the process of heat transfer, and study with not completely, not accurate, with strong noise data, owning the formidable self-organizing, self-learning, nonlinear mapping capability features. The artificial neural network has already been applied to directly correlate experimental data as a function of the working conditions.In order to promote the accuracy of flow boiling heat transfer for refrigerants, a prediction model of the flow boiling heat transfer for pure refrigerant R245fa and refrigerant mixture R407C inside horizontal smooth tubes was proposed based on aritifical neural network in this papar. This study do some investigations as follows:the research status of flow boiling heat transfer inside tubes, mainly included the experimental research and correlations, described the refrigerant flow boiling heat transfer process inside horizontal tubes and the main factors affecting the flow boiling heat transfer. After analysising the existing flow boiling heat transfer correlations, we made them calculate the corresponding conditions flow boiling heat transfer coefficient of R245fa and R407C, compared with the experimental results.This paper employs the saturation boiling heat transfer coefficient of R245fa, R407C and the corresponding work condition from the published literatures. The database covers a wide range of the operating conditions.The prediction models of the flow boiling heat transfer for R245fa inside horizontal smooth tubes based on RBF and GRNN network were established respectively. And the prediction results of the networks showed significantly improved, compared with the calculation results of Chen, Guangor-Winteron, Liu-Winteron and Shah correlations respectively. Although the results of Liu-Winteron correlation was better than the other three correlations for calculating R245fa, the prediction results of the RBF network has been further improved than Liu-Winteron correlation, and it was much better than other thee correlations. The prediction accuracy of GRNN network is also better than the traditional correlations, even though its performance is not so good as the RBF network. If the source of current experiment data is insufficient or it could be more, the advantages of GRNN network will be better to display. It proved that the RBF and GRNN network can predict the flow boiling heat transfer performance for R245fa inside horizontal smooth tubes.The prediction models of the flow boiling heat transfer for R245fa inside horizontal smooth tubes was established based on RBF network.The K-means clustering algorithm was applied to design RBF network. In addition, the prediction result was significantly improved than the four frequently used conventional correlations. For the network model of heat transfer, the average deviation, absolute average and root-mean-square deviations are-0.9%,5.5%and10.9%, respectively. Hence, the simulation results revealed that the modeling method based on RBF neural network was feasible to calculate the flow boiling heat transfer coefficient, and it may provide some worthy guidelines for the optimization design of tube evaporators for R407C. The research works show that the simulation model for flow boiling heat transfer of organic alkyl inside horizontal smooth tubes based on neural network is feasible. It can accurately predict the flow boiling heat transfer coefficient under different conditions, avoid to analysis the flowing boiling heat transfer mechanism of refrigerant, and reduce the workload. For the tube-evaporators optimization design and safe operation to the low temperature waste heat power organic ranking circulation system and R407C refrigeration system, the method has a massive significance in theory and engineering application.