Simulation and Optimization of Pollutants in Enclosed Kitchen Based on BP Neural Network and Genetic Algorithm
The lampblack pollutants produced by Chinese cooking will cause harm to human health.The existing research on kitchen pollutants mainly uses numerical simulation and onsite measurement.These two methods consume a large amount of calculation and a long research time.Therefore,this paper used the surrogate model to replace the numerical simulation software for the research and evaluation of kitchen pollutants,and then uses a genetic algorithm to find the optimal parameters.First,according to the relationship between the door and window stoves,566 practical cases of single-row kitchens were classified,and the most representative single-row kitchen prototype model was established.Then,the method of establishing any sample model through the four control variables of kitchen bay,depth,the distance between the window and the wall,and the distance between the range and the wall was proposed.After collecting 130 actual sample model parameters,the prediction model for kitchen pollutant concentration between kitchen size parameters and pollutant concentration was established using the BP neural network and Fluent.Finally,a genetic algorithm was used to seek the minimum value of the model,and the independent variable value at this time was the optimal parameter combination of the control variable,and the accuracy of the optimal parameter combination was verified.Based on the building modulus coordination standard,the range of values for the optimal parameter combination was given:kitchen width[1.7 m,1.8 m],kitchen length[3.9 m,4.0 m],distance from window to wall[0.17 m,0.18 m],distance from stove to wall[0.34 m,0.35 m].Of the four control variables,the distance from stove to wall played a more important role in reducing the concentration of pollutants.In future kitchen graphic design,the space requirements of building module coordination and cooking operations should be considered,and the appropriate layout scheme can be selected based on the results of this study.
NETL
NSTL
万方数据
