Parallel Numerical Method Research And Application Of Automatic Balanced Partition Based On Structure Collapse In Building Fire

Building fire accidents occur frequently,and the fire spreads rapidly and causes great harm.In building fires,there are often structure collapses,which will have a great impact on the development of fire.Field model is usually used in building fire simulation,and the number of grids is often up to millions or even tens of millions,which requires high computing power,so serial computing is difficult to meet the computing demand.Aiming at the above problems,parallel numerical method of automatic balanced partition based on structure collapse in building fire is studied and applied in this thesis.The main contents include:Firstly,the mathematical model of fire dynamics based on structure collapse and numerical modeling method based on grid transformation are propose,and the tool to realize the method is developed.According to the control equations of combustion and collapse,the coupling mathematical model of collapse solid and fire based on contact transfer is given,and then the mathematical model of building fire based on structure collapse is established.Aiming at the mathematical model of building fire dynamics based on structure collapse,the method of modeling building structure in the field model based on the grid transformation mechanism and its application in discrete simulation are described.According to the mathematical model of fire dynamics based on structure collapse and the method of modeling collapse solid based on grid transformation,the corresponding function extension of the open-source fire numerical simulation software is carried out to support the modeling and numerical simulation of building structure collapse in fire.Secondly,automatic balanced partition method based on load quantification in fire simulation is proposed,and the tool to realize the method is developed.In view of the large number of field model grids and the need of message communication between different processes in parallel computing,a quantitative estimation method for computing load of sub region in parallel computing is proposed.Four computational load influence factors are proposed,and an estimation function of computational load is given by taking them as variables.The load of sub region is estimated by the value of the estimation function.For the case of indoor fire,a set of parameters which make the estimation function value having a good linear relationship with the actual simulation time are determined by orthogonal test method,and the specific form of the computational load estimation function in the indoor building fire is determined.Based on the method of load estimation,this thesis proposes an algorithm of automatic balanced partition in building fire numerical simulation,which can realize the automatic balanced partition for the building fire model.Finally,the numerical modeling method and automatic partitioned method proposed in the thesis are applied to the actual building fire simulation.Based on an express logistics center,a full-scale field model is established.Parallel simulate the model using automatic balanced partition method based on load quantification,then analyze the influence on parallel efficiency of automatic balanced partition method based on load quantification.The analysis results show that the automatic balanced partition method proposed in this thesis effectively improves the parallel efficiency and shortens the computing time.By using this method,the influence of structure collapse on the fire is further studied,and the influences on temperature,carbon dioxide concentration,oxygen concentration and smoke concentration are analyzed under different collapse positions,different collapse starting time and different measuring points.