参考教程案例flow3d软件

1、Optimizing Casting Process Parameters for Homogeneous Aggregate Distribution in Self-Compacting Concrete using Flow3D: A Preliminary StudyJon Spangenberg, Cem C. Tutum, and Jesper H. HattelProcess Modelling GroupFTP project Partners:10/06/20102DTU Mechanical Engineering, Technical University of Denm

2、arkOutlineSelf-Compacting Concrete (SCC) CastingsNumerical modelContinuous phaseParticlesOptimization Classical methods vs. evolutionary algorithms Applied evolutionary algorithm Coupling of the optimization algorithm to Flow3D ResultsSummary/Conclusion10/06/20103DTU Mechanical Engineering, Technica

3、l University of DenmarkSelf-Compacting Concrete CastingsThe difference between SCC and conventional concrete is the fluidityApplications: House Building Bridges Skyscrapers Modern Architecture10/06/20104DTU Mechanical Engineering, Technical University of DenmarkNumerical ModelSimulation of SCC casti

4、ng of a 3 m long and 0.5 m high beam SCC calculated with an one way momentum coupling between continuous phase and particlesSpangenberg, J. et al., IEEE Congress on Evolutionary Computation , 2011, New Orleans, USA10/06/20105DTU Mechanical Engineering, Technical University of DenmarkNumerical Model:

5、 Continuous PhaseContinuous phase Calculated as a single phase Paste and 75% of finest aggregate Yield stress 50 Pa, plastic viscosity 50 Pas Bingham material model (bi-viscosity)Thrane, L. N., PhD Thesis, DTU, 200710/06/20106DTU Mechanical Engineering, Technical University of DenmarkNumerical Model

6、: ParticlesParticles Representing 25% coarsest aggregate Modeled as spheres Average diameter of 0.014 m Particles represented as a volume fraction scalar The settling of particles calculated explicitly based on velocity fields from continuous phase calculation 10/06/20107DTU Mechanical Engineering,

7、Technical University of DenmarkNumerical Model: ParticlesFlowchartCalculation of velocity and pressure fieldAdvection of volume fraction scalarFlow3DSettling of volume fraction scalarSubroutineCalculation of global flowCalculation of volume fraction scalarInputOutputTime stepSettling velocityequatio

8、nVolume fractionequation10/06/20108DTU Mechanical Engineering, Technical University of DenmarkOptimizationAll engineering problems can be optimized Trial and error (manually) Algorithms (autonomous)Questions to ask one self before choosing How much should the solution be optimized How much time does

9、 one have How much engineering thinking is needed in order to make the input10/06/20109DTU Mechanical Engineering, Technical University of DenmarkOpt.: Clas. vs Evolutionary AlgorithsClassical MethodsEvolutionary AlgorithmsMethods/AlgorithmsNewton-Raphson,Conjugated-Gradient,.Non-Dominated Sorting G

10、enetic Algorithm (NSGA II),Multi Objective Genetic Algorithm (MOGA),.MethodologyGradients/Curvatures,Deterministic search Population,Stochastic operatorsPros Fast convergence AccuracyBlack-boxDiscrete valuesNoiseParallelization ConsRobustnessParallelizationSpeed of convergenceAccuracy10/06/201010DTU

11、 Mechanical Engineering, Technical University of DenmarkOpt.: Applied Evolutionary AlgorithmObjective: Get an uniform aggregate distributionObjective in mathematical terms:Minimize the standard deviation of the volume fraction scalarSingle Objective Genetic Algorithm (SOGA) Binary encoding Genetic o

12、perators: Binary tournament selection Single-point crossover Uniform mutationBest solution preserved to accelerate convergence (Elitism) 10/06/201011DTU Mechanical Engineering, Technical University of DenmarkOpt.: Couple opt. Algorithm to Flow3DCode to run the optimizationvia Matlab:Flowchart of opt

13、. problem VERY IMPORTANT: Investigate mesh andtime step sensitivitySpangenberg, J. et al., IEEE Congress on Evolutionary Computation , 2011, New Orleans, USA10/06/201012DTU Mechanical Engineering, Technical University of DenmarkResultsOptimization with three design variablesBest Solution:X-position

14、0.88mInlet size 7.5 cmMass Rate 1.66 kg/sSpangenberg, J. et al., IEEE Congress on Evolutionary Computation , 2011, New Orleans, USA10/06/201013DTU Mechanical Engineering, Technical University of DenmarkResultsBest solution vs. solution when casting from the middleSpangenberg, J. et al., IEEE Congres

15、s on Evolutionary Computation , 2011, New Orleans, USA10/06/201014DTU Mechanical Engineering, Technical University of DenmarkResultsOptimization with two design variablesSpangenberg, J. et al., IEEE Congress on Evolutionary Computation , 2011, New Orleans, USA10/06/201015DTU Mechanical Engineering, Technical University of DenmarkSummary/Conclusion Showed how to make a dynamic segregation subroutine for SCC Showed how to couple an evolutionary algorithm with Flow