Temperature Field Simulation And Prediction On Whole Post Weld Heat Treatment Of Large Pressure Vessel By Internal Combustion Method

Welding residual stress increases significantly during the assembly welding of the large pressure vessels.In this sense,a way to reduce residual stress is to perform post weld heat treatment.However,it is difficult to retrieve physical information such as temperature from inside the pressure vessels.The engineering experience may also not guarantee the quality of the heat treatment.In this paper,a numerical model of internal combustion method which can reflect temperature rising in the heat treatment process was established.Also the temperature field during the heat treatment was simulated and predicted.An extended domain was utilized in the calculations in such a way that,the whole heat treatment regions including the pressure vessel,the flue gas and the heating preservation cotton,were treated as one closed system for purposes of modeling.Besides,the coupling effects between the pressure vessel,the flue gas and the heating preservation cotton had been considered.A three-dimensional transient heat transfer model,including standard k-? turbulence equations and DTRM radiation equations,was set up.Then,it was applied to simulate the temperature field of the vertical as well as the horizontal pressure vessel.The simulation results were obtained and compared with the measured ones.The relative error of the horizontal pressure vessel was less than 1.2% whereas this value for the vertical pressure vessel was within 6.3%.These figures were small enough to verify the accuracy of the model.In order to perform the calculations,a back-propagation(BP)neural network containing 15 hidden layers was established.The flame temperature of oxygen enriched combustion and the positions of the different layers were the input variables.The temperature distribution of the whole pressure vessel was the output variable.This way,different outlet temperatures of the burners represented different oxygen enrichment conditions.Heat transfer model was then used in order to calculate the temperature distribution of the pressure vessel at different outlet temperatures.Four groups of the simulation results were used to train the BP neural network,one group was used to test it.The relative error between the prediction results obtained with the modeling and the simulation results was 0.34%.The good accordance between predicted and simulated values obtained from this combination of neural network and numerical simulation avoided a mass of FLUENT calculation,saving time and costs greatly.Based on the accuracy and reliability of the three-dimensional transient heat transfer model,the design of the baffle for the horizontal pressure vessel was optimized.The maximum temperature difference during the heating process should meet the actual heat treatment requirements.The results showed that the maximum temperature difference of the pressure vessel with baffle was less than 140? during the heating process and was less than 80 ? after heating for 9 hours,which provided a good foundation for the heat preservation process.For larger pressure vessels,the heat treatment generally demanded the use of multiple burners simultaneously.In this case,the pattern search method was introduced and used to optimize the inlet velocity of the burners.The optimized inlet velocity curve was obtained considering an error threshold of 2%.The maximum temperature difference was less than100? during the heating process and was less than 60 ? after heating for12 hours.Then,the optimized inlet velocity was converted into fuel consumption in order to retrieve the fuel consumption curve.The testresults showed that the measured temperatures fitted well with the simulated ones and it verified the accuracy of the pattern search method used to optimize the parameters of the numerical simulation.