Application of Genetic Algorithms in Fluid Flow through porous colloidal suspensions to optimize the Fractal Dimension of Primary Particle Agglomerates. Project Guide: Dr. N. Chakraborty Course: Genetic Algorithms in Engineering Process Modeling

Presented by

P.V.Kiran Kumar 05SI2023

Shailvee Panjiyar 05BT1010

Sourav Dutta 05MA2024

Suman Bikash Mondal 05BT3012

Sushmita Dutta 05GG2002

Genetic Algorithms Term Project

Page 1

Index Contents

Page No.

1. Abstract………………………………………………………………………………….03 2. Introduction…………………………………………………………………………...03 3. Mathematical Relations…………………………………………………………...05 4. Working Principle……………………………………………………………………06 5. Elements of GA used and their advantages…………………………………07 6. Mathematica Code…………………………………………………………………..09 7. Results…………………………………………………………………………………..12 8. Practical Applications……………………………………………………………..16 9. References……………………………………………………………………………..17 10. Acknowledgements…………………………………………………………………17

Genetic Algorithms Term Project

Page 2

ABSTRACT In this paper we determine the optimal values of Fractal Dimensions of spherical agglomerates of different primary particles like clay, soot, sand etc using experimentally calculated values of permeability for these materials. Agglomerates are irregularlyshaped particles composed of several primary (spherical) particles having much smaller size. The structure of a fractal aggregate is characterized by its fractal dimension, Df, which is a measure of how particles fill the space in the considered aggregate. We apply genetic operators on a random population of Df values to arrive at an optimal value of Df for the corresponding material. In our case, the material in consideration is common clay. This optimum value of Df can then be used for accurate prediction of motion parameters and viscosity of the considered suspension.

INTRODUCTION Agglomerates are clusters of much smaller spherical primary particles. Agglomerates are normally formed by coagulation of primary particles, as for example flocks of solid sediments in liquids, soot aggregates discharged by motor engines, etc. Agglomerates’ structure information governs mass and size distributions, which are related to the macroscopic system behavior and properties, as for example suspension viscosity, mechanical, electrical, optical, physico-chemical, etc. Fluid flow relative to aggregates composed of small particles is of interest in aggregation, filtration and sedimentation processes. Most of the researchers till now have considered fractal agglomerates as homogeneous spherical particles. However there is now considerable evidence that the structure of aggregates is nonhomogeneous. In this light all predictions must incorporate the non-homogenous character for accurate prediction of the involved parameters. Taking into account the non-homogeneous character of agglomerates we need sizes and explicit positions of the primary particles within the aggregates for calculating the viscous drag and parameters of motion for fractal agglomerates suspended in a shear flow. Genetic Algorithms Term Project

Page 3

This can be done in two ways: 1. Numerical reconstruction using algorithms to fit microscopic (TEM) observations 2. Viewing the fractal agglomerates as continuum bodies permeable to the surrounding medium. The Numerical Reconstruction method does not fare well for agglomerates of low porosity and those with higher number of primary particles. The results from both the methods are comparable wherever both are applicable. The second method being applicable over large range of porosity and number of primary particles, is the method of our choice. This approach requires -i) Ns : the total number of the primary particles composing the agglomerate ii) Df: fractal dimension iii) kf :prefactor

The general approach for predicting the motion parameters involves the following approach : For a given aggregate the permeability K can be calculated experimentally. From this the fractal dimension can be calculated using some standard relations. Using these values of Df we can calculate the Brinkman factor which is used to calculate the various motion parameters of the suspension and the changed viscosity of the suspension. Till now all the work that has been done in this field using either the continuum or numerical method has led to only led to a range of values for Df. Actually the range of values of Df depends upon the mechanism of growth of the agglomerates. ClusterCluster mechanism leads to smaller values of Df whereas Particle-Cluster mechanism leads to higher values of Df. Considering all factors the overall range of Df is 1-3. As permeability is available in range we get the values of Df as a range. So without the actual values the predicted motion parameters often do not tally with those observed. So here we have optimized the value of Df using genetic algorithm. As a result the brinkman factor can be accurately calculated for any suspension of fractal agglomerates and its motion parameters can be accurately calculated.

Genetic Algorithms Term Project

Page 4

Mathematical Relations Consider a rigid porous agglomerate moving in an infinite linear shear flow U∞(x) (see Fig. 1). The agglomerate may generally have translation and rotational velocities relative to the fluid. Fluid inertia forces are considered negligible.

Here : r0 – Radius of larger fractal aggregate ds/ as-- Radius of primary particles

The following relations have been used:

1. Fractal Dimension is given by:

Ns = kf (ro/as) Df

2. Prefactor is given by:

kf = 0.414Df – 0.2111

Genetic Algorithms Term Project

Page 5

3. Local Volume Fraction in the Aggregate:

4. Average Internal Void Fraction:

γ = (Ns)1-3/ Df

γ1--(1-- o)e-- λ φ

5. Shielding Coefficient: 6. Permeability:

φ(r) = (kf Df/3) (ro/as) (Df –3)

where, o = 0.4, λ = 10

f(γ) = k/as2 = 2/(9 γ3)x(3—4.5 γ + 4.5 γ5 -- 3 γ6) )/( 3+ 2 γ5)

7. Brinkman’s parameter:

β= (γ/ kf)1/( Df—3) f-1/2(γ)

Working Principle Our target is to reach at a particular value of Df for a given fractal agglomerate. For the suspension of fractal agglomerates chosen we know the value of permeability and the range of values the Df can take. So using a random value of Df chosen from within its range we get a value of Ns for the agglomerate. In this, an initial population of Ns is created by generating different values of Df, within the range :– (1 – 3), using the random function. Then for each value of Ns, Df, γ andγare calculated to get the value of permeability k. Using the given value of permeability the fitness of each solution is calculated. Then selection, crossover and mutation are carried out on this generation to obtain the next generation of values and the cycle is repeated till the fitness value is minimized and Df value converges.

Genetic Algorithms Term Project

Page 6

Elements of GA used and their advantages: Coding: Real Coded In real coded GAs crossover and mutation operators are applied directly to real parameter values instead of using a via binary approach of the binary coded GAs. Advantages of Real coded GA  a) Binary coded GA face difficulties when handling continuous search space such as Hamming cliffs.  b) Binary coded GAs are unable to achieve arbitrary precision in the optimal solution. In binary GAs the string length has to be specified a priori to achieve a certain level of precision. The more the required precision the more the string length. For large string lengths the population size requirements also large and so increases the computational complexity manifold. Moreover the variable bounds must be known for the decision variables. But that information might not be available.  c) In Real coded GAs decision variables can be directly used for computing the fitness.  d) Since the selection operator uses only fitness values any selection operator used in Binary GAs can be used in the Real coded GA also.

Selection: Tournament selection In a Tournament selection tournaments are played between 2 solutions and the better solution is chosen and placed in the mating pool. Then two other solutions are chosen and same is done. If carried out systematically it will result in each solution being part of a selection twice. The best solution in a population will win both times and the worst will loose both times. Thus any solution in a population will have 2,1 or 0 copies of itself in the mating pool with better solutions standing a better chance of survival. Advantages of Tournament Selection  a) Simple to implement  b) Proven by Goldberg and Deb 1991 that it has better or equivalent convergence and time complexity properties when compared to any other reproduction operator.

Genetic Algorithms Term Project

Page 7

Crossover: Blend Crossover It is a self adaptive crossover scheme. The location of the offspring depends on the difference in the parent solutions. Thus if the difference between the parent solutions is small then the difference between the offspring and the parent solutions is small too and vice versa. This implies that if the diversity of the parent population is large then the diversity in the offspring population is large too and vice versa.

C(1,t+1)=(1- γi)*C(1,t)+ γi*C(2,t) Mutation: Non- Uniform Mutation In Non-Uniform Mutation the probability of creating a solution closer to the parent is more than the probability of creating one away from it. However as the generations proceed, this probability of creating solutions closer to the parents, increases. Thus this scheme behaves as a uniform distribution initially but later it converges rapidly and thus allows for a more focused search.

Mi(1,t+1)=Xi(1,t+1)+*(Xi(r1)-Xi(r2))*(1-r3(1-t/tmax)^b)

Genetic Algorithms Term Project

Page 8

Mathematica Code Clear[Df]; Clear[Ns]; Clear[]; Clear[]; Clear[]; Clear[k]; Clear[f]; Clear[S]; Clear[c]; Clear[c1]; Clear[M]; Array[Df,30]; Array[Ns,30]; Array[Kf,30]; Array[,30]; Array[,30]; Array[,30]; Array[k,30]; Array[f,{2,30}]; Array[c,{2,30}]; Array[c1,30]; Array[S,{3,30}]; Array[M,{2,30}]; Array[val,100]; Array[full,{30,100}]; Input["Give the value of r0"] ; Input["Give the value of as"] ; Input["Give the value of pb"] ; Pc = 0.8; Pm = 0.7; b=2; g=0; For[i=30,i1,i--,{S[3,i]=0,c[2,i]=0,M[2,i]=0}]

(*initializing generation index*)

(*generation of initial random population*) For[i=1, i30, i++,Df[i]=Random[Real,{1,3}]] For[i=1,i30,i++,Kf[i]=(0.414*Df[i]-0.2111)] For[i=1,i30,i++,Ns[i]=Round[Kf[i]*(r0/as)^(Df[i])]] For[i=1,i30,i++,[i]=(Ns[i])^(1-(3/Df[i]))] For[i=1,i30,i++,[i]=Kf[i]*(Df[i]/3)*(r0/as)^(Df[i]-3)] For[i=1,i30,i++,[i]=1-(1-0.4)*(^(-10*[i]))] Genetic Algorithms Term Project

Page 9

For[i=1,i30,i++, k[i]=((2*as^2)*(3-4.5*[i]+4.5*([i]^5)3*([i]^6))/(9*([i]^3)*[i]*(3+2*[i]^5)))] For[i=1,i30,i++,{f[1,i]=Abs[((pb)-k[i])*(10^11)],f[2,i]=0}] min=f[1,1] (*storing population with best or minimum fitness*) For[i=2,i<=30,i++,{If[f[1,i]min,min=f[1,i]]}] Array[Df,30] Array[Ns,30] Array[,30] Array[k,30] Array[f,{2,30}] For[z = 1,z<=100,z++, (*main iteration loop starts*) (*tournament selection*) { For[i=30,i1,i--,{ran=Random[Integer,{1,i}], {Df[30-(i-1)],Df[ran]}={Df[ran],Df[30-(i-1)]}, {f[1,30-(i-1)],f[1,ran]}={f[1,ran],f[1,30-(i-1)]}}], For[j=1,j<30,j=j+2,{If[f[1,j]<=f[1,j+1],{S[1,j]=Df[j],S[2,j]=f[1,j]},{S[1,j]=Df[j+1],S[2,j]=f[1,j+ 1]}]}], For[i=30,i1,i--,{ran=Random[Integer,{1,i}], {Df[30-(i-1)],Df[ran]}={Df[ran],Df[30-(i-1)]}, {f[1,30-(i-1)],f[1,ran]}={f[1,ran],f[1,30-(i-1)]}}], For[j=1,j<30,j=j+2,{If[f[1,j]f[1,j+1],{S[1,j+1]=Df[j],S[2,j+1]=f[1,j]},{S[1,j+1]=Df[j+1],S[2,j+ 1]=f[1,j+1]}]}], Array[S,{3,30}], (*crossover operation*) For[i=30,i1,i--,{ran=Random[Integer,{1,i}], {S[1,30-(i-1)],S[1,ran]}={S[1,ran],S[1,30-(i-1)]}, {S[2,30-(i-1)],S[2,ran]}={S[2,ran],S[2,30-(i-1)]}, {f[1,30-(i-1)],f[1,ran]}={f[1,ran],f[1,30-(i-1)]}}], For[i=1,i<30,i++,{c1[i]=(2*Random[]-0.5),If[Random[]< Pc ,c[1,i]=(1c1[i])*S[1,i]+c1[i]*S[1,i+1],c[1,i]=S[1,i]]}], Array[c,{2,30}], (*mutation operation*) For[i=30,i1,i--,{ran=Random[Integer,{1,i}], {c[1,30-(i-1)],c[1,ran]}={c[1,ran],c[1,30-(i-1)]}, {f[1,30-(i-1)],f[1,ran]}={f[1,ran],f[1,30-(i-1)]}}], For[i=1,i30,i++,{If[Random[]>Pm,{If[Random[]<= 0.5 ,=1,=1],M[1,i]=c[1,i]+*(c[1,Random[Integer,{1,30}]]-c[1,Random[Integer,{1,30}]])*(1Random[]^(1-M[2,i]/100)^b)},M[1,i]=c[1,i]]}], Array[M,{2,30}], (*updating generation index*) For[i=1,i30,i++,{Df[i]=M[1,i], c[2,i]=g, M[2,i]=g, S[3,i]=g}], Genetic Algorithms Term Project

Page 10

(*updating new generation population for next iteration*) For[i=1,i30,i++,Kf[i]=(0.414*Df[i]-0.2111)], For[i=1,i30,i++,Ns[i]=Round[Kf[i]*(r0/as)^(Df[i])]], For[i=1,i30,i++,[i]=(Ns[i])^(1-(3/Df[i]))], For[i=1,i30,i++,[i]=Kf[i]*(Df[i]/3)*(r0/as)^(Df[i]-3)], For[i=1,i30,i++,[i]=1-(1-0.4)*(^(-10*[i]))], For[i=1,i30,i++, k[i]=((2*as^2)*(3-4.5*[i]+4.5*([i]^5)3*([i]^6))/(9*([i]^3)*[i]*(3+2*[i]^5)))], For[i=1,i30,i++,{f[1,i]=Abs[((pb)-k[i])*(10^11)],f[2,i]=0}], min=f[1,1], For[i=2,i<=30,i++,{If[f[1,i]min,{min=f[1,i],val[z]=Df[i]}]}], min, val[z], Array[Df,30], Array[Ns,30], Array[,30], Array[k,30], Array[f,{2,30}], g++,

If[Mod[z,10]0,{ListPlot[{{k[1],Df[1]},{k[2],Df[2]},{k[3],Df[3]},{k[4],Df[4]},{k[5],Df[5]},{k [6],Df[6]},{k[7],Df[7]},{k[8],Df[8]},{k[9],Df[9]},{k[10],Df[10]},{k[11],Df[11]},{k[12],Df[12] },{k[13],Df[13]},{k[14],Df[14]},{k[15],Df[15]},{k[16],Df[16]},{k[17],Df[17]},{k[18],Df[18]}, {k[19],Df[19]},{k[20],Df[20]},{k[21],Df[21]},{k[22],Df[22]},{k[23],Df[23]},{k[24],Df[24]},{ k[25],Df[25]},{k[26],Df[26]},{k[27],Df[27]},{k[28],Df[28]},{k[29],Df[29]},{k[30],Df[30]}},Ti cks{{1.5*10^-11,1.75*10^-11,1.85*10^-11,1.9*10^-11,1.95*10^-11,2*10^11},Automatic},AxesLabelTraditionalForm/@{k values,Df values},TextStyle{FontFamily"Times",FontSize12},GridLinesAutomatic,Frame>True,FrameStyle{{RGBColor[0,1,0]},{RGBColor[0,1,1]}},TextStyle{FontSlant"Plain",F ontSize11},AxesStyle{RGBColor[0,0,1]},PlotStyle{Hue[0.9],PointSize[0.02]}]}] For[q=1,q<=30,q++,full[q,z]=Df[q]] }]

(*main loop ends*)

Array[f,{2,30}] min Array[val,100] Array[k,30] Array[Ns,30] Array[Df,30]

Genetic Algorithms Term Project

Page 11

Results Input:

r0 = 0.5*10^-6 ; as = 25*10^-9 ; pb =2*10^-11

Output: Initial Random Population Values: a) Ns values

{6689,117,316,179,841,14,46,7321,350,407,394,905,17,253,5,30,4,113,37,1069,110,415 9,763,43,139,247,13,2269,1500,93} b) Df values

{2.93847,1.79869,2.06781,1.91345,2.34043,1.26178,1.55973,2.96495,2.09594,2.13775, 2.12871,2.36131,1.30974,2.00709,1.02864,1.44933,1.01504,1.79028,1.49986,2.40859,1 .78397,2.79949,2.31312,1.53731,1.8455,2.0008,1.24531,2.62386,2.50507,1.73902} c) K values

3.52099  10 , 4.02864  10 , 5.4561  10 , 1.83505  10 , 4.34727  10 , 1.79948  10 18

12

19

13

11

12

11

12

13

13

12

13

14

11

14

5.84068  10 , 4.29145  10 , 2.97223  10 , 3.22146  10 , 3.52103  10 , 1.91142  10 12

11

14

12

1.79009  10 , 1.06135  10 , 4.2187  10 , 1.63549  10 , 2.17102  10 , 4.35791  10 13

11

15

d) Fitness values

,

13

, 3.36582  10

, 8.9821  10

16

, 1.79766  10

15

1.47026  10 , 2.95191  10 , 9.43078  10 , 1.68745  10 , 2.06033  10 , 7.83781  10

11

, 1.28372  10

11

e) Min fitness

0.513911

Page 12



14

, 5.7024  10

12

, 5.71597  10

{{2.,1.59714,1.94544,1.8165,1.99565,0.200517,0.716277,2.,1.95709,1.97028,1.96779,1. 99648,0.0885836,1.91018,1.66342,0.209914,1.89386,1.57813,0.364508,1.99783,1.5642 1,1.99998,1.9943,0.529741,1.70481,1.90569,0.312546,1.99979,1.99922,1.4284}

Genetic Algorithms Term Project

12

,

,

Graphs generated after every 10 generations :

1.4 1.38 1.36 1.34

k values  11

 11

 11

 11

 11

 11 11

1.7 1.75 10 1.8 10 1.85 10 1.9 10 1.95 10  10 2 10 2.05  10 11

Df values 1.39 1.38 1.37 1.36 1.35  11  11  11 11 1.8  101.85  101.9  101.95  10211 102.05  1011

1.3711 1.371 1.3709 1.3708 1.3707 1.3706

k values

 11 11 11 11  11  11 1.9975 210 2.0025 10 2.005  102.0075  10 2.01  102.0125  10 10 11

1.37099 1.37098 1.37098 1.37098 1.37098 1.37098 1.37098 11 1.99998 1.99999 1011  10211  102.00001 2.00002 10 11  10 11

Genetic Algorithms Term Project

Page 13

1.37098 1.37098 1.37098 1.37098 1.37098

k values 2  10 11 2  10211 1011 2  10 11 2  10 11

1.37098 1.37098 1.37098 1.37098 1.37098 2  10112 10112  10 112  10 112  10112 1011

1.37098 1.37098 1.37098 1.37098 1.37098 1.37098 1.37098 2 10112 10112 10112 10 112  10 11 2.5 2 1.5 1 0.5 0 0

1  10 11

2  10 11

3 10 11

4 1011

1.37098 1.37098

1.37098 1.37098 1.37098 1.37098 2  10211  10 11 2 1011 2  10 11 2 10211  10 11 2 1011

Genetic Algorithms Term Project

Page 14

1.37098 1.37098

1.37098 1.37098 1.37098 1.37098 2  10211  10 11 2 1011 2  10 11 2 10211  10 11 2 1011

….100 generations later :

Final Optimal Values: a) Ns values

{23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23,23, 23,23} b) Df values

{1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156, 1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1 .38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156,1.38156} c) K values

2.  10

11

11

11

11

11

11

11

11

11

11

11

11

11

11

11

11

11

11

2.  10 2.  10

, 2.  10 , 2.  10 , 2.  10

d) Fitness values 1.68021  10

, 2.  10 , 2.  10 , 2.  10

, 2.  10

, 2.  10

, 2.  10

, 2.  10

, 2.  10

, 2.  10

, 2.  10 , 2.  10 , 2.  10

11

, 2.  10

11

, 2.  10

11

, 2.  10

11

, 2.  10

11

, 2.  10

, 2.  10 , 2.  10

11

, 2.  10

15

15

15

15

15

15

15

15

14

15

15

8.40105  10

, 9.04729  10

, 8.40105  10 , 8.40105  10

, 9.04729  10 , 1.68021  10

, 9.04729  10 , 9.04729  10

,

11

, 2.  10

11

, 2.  10

14

, 8.40105  10

11

, 9.04729  10 , 8.40105  10

11

,

11

, 2.  10

, 2.  10 15

,

14

,

, 9.04729  10 , 1.68021  10

1

8.40105  1015, 8.40105  1015, 8.40105  1015, 9.04729  1015, 8.40105  1015, 8.40105  1015 , 8.40105  1015 , 9.04729  1015 15

8.40105  10

e) Min fitness

15

, 9.04729  10

15

, 8.40105  10

15

, 8.40105  10

15

, 8.40105  10

15

, 8.40105  10

15

, 9.04729  10

2.58494  1015 Thus we see that after a 100 iterations of genetic operations we reach accuracy of the order of 10-15.

Genetic Algorithms Term Project

Page 15

14

, 1.68021  10

Practical Applications Using this program we can obtain a definite value of Df for any suspension of fractal agglomerates. Using that we can calculate Brinkman factor β. Using β we can calculate the following things: 1. Drag force experienced by the Fractal agglomerates. 2. Perturbation of the velocity field of the suspension 3. Translational and Rotational velocities of the agglomerates 4. Effective viscosity of medium of suspension.

Using these above results we can predict: 1. Complete flow characteristics of the fractal agglomerates. 2. Fractal distribution and a measure of porosity. 3. Behavior of the agglomerates in a reaction 4. Rate of coagulation of the agglomerates

Genetic Algorithms Term Project

Page 16

References A. Veerapaneni, S., & Wiesner, M.R. (1996). Hydrodynamics of fractal aggregates with radially varying permeability. Journal of Colloid and Interface Science, 177, 45-57. B. Marco Vanni (1999). Creeping flow over spherical permeable aggregates. Chemical Engineering Science, 55(2000), 685-698. C. P. Vainshtein, M. Shapiro, (2004). Mobility of permeable fractal agglomerates in slip regime. Journal of Colloid and Interface Science, 284(2005),501–509. D. P. Vainshtein, M. Shapiro (2005). Porous agglomerates in the general linear flow field. Journal of Colloid and Interface Science ,298(2006),183–191.

Acknowledgements We would like to acknowledge our heartfelt gratitude to Prof. N. Chakraborty for letting us work on the application of genetic algorithms in the field of computational fluid dynamics which has helped us immensely in gaining valuable practical knowledge of the subject and has aroused our interest in this field. We would also like to thank Prof. G.P. Rajashekhar for his invaluable help and advice in the formulation, development and improvement of our ideas.

Genetic Algorithms Term Project

Page 17

Application of Genetic Algorithms in Fluid Flow through ...

For[i=30,i 1,i--,{ran=Random[Integer,{1,i}],. {Df[30-(i-1)],Df[ran]}={Df[ran],Df[30-(i-1)]},. {f[1,30-(i-1)],f[1,ran]}={f[1,ran],f[1,30-(i-1)]}}],. For[j=1,j<30,j=j+2,{If[f[1,j]<=f[1,j+1],{S[1,j]=Df[j],S[2,j]=f[1,j]},{S[1,j]=Df[j+1],S[2,j]=f[1,j+. 1]}]}],. For[i=30,i 1,i--,{ran=Random[Integer,{1,i}],. {Df[30-(i-1)],Df[ran]}={Df[ran],Df[30-(i-1)]},. {f[1,30-(i-1)],f[1 ...

565KB Sizes 1 Downloads 209 Views

Recommend Documents

Application of Genetic Algorithms in Optimization and ...
dynamic plants, with aid of real-coded GAs for neural network (NN) training. ... trailer-truck verify that the adaptive design can guide the actual plant to desired ... my academic growth and social adjustment since I came to the University of the.

On Application of the Local Search and the Genetic Algorithms ...
Apr 29, 2010 - to the table of the individual MSC a column y0 consisting of zeroes. Since the added ... individual MSC problem. Now we will ..... MIT Press,.

On Application of the Local Search and the Genetic Algorithms ...
Apr 29, 2010 - j=0 cj log2 cj, where cj. - is the 'discrete' ..... Therefore, we propose a criterion that would reflect the degree of identification of the set L of events.

Genetic Algorithms and Artificial Life
In the 1950s and 1960s several computer scientists independently studied .... individual learning and species evolution a ect one another (e.g., 1, 2, 13, 37 ... In recent years, algorithms that have been termed \genetic algorithms" have ..... Bedau

Genetic Algorithms and Artificial Life
... in the population. 3. Apply selection and genetic operators (crossover and mutation) to the population to .... an environment|aspects that change too quickly for evolution to track genetically. Although ...... Princeton University Press, Princeto

Genetic Algorithms and Artificial Life
In the 1950s and 1960s several computer scientists independently studied .... logical arms races, host-parasite co-evolution, symbiosis, and resource ow in ...

Genetic Algorithms and Artificial Life
In the 1950s and 1960s several computer scientists independently studied ... ther developed by Holland and his students and colleagues at the University of .... If if the environment is stable so that the best things to learn remain constant, then th

pdf-1843\nanoparticle-heat-transfer-and-fluid-flow-advances-in ...
... the apps below to open or edit this item. pdf-1843\nanoparticle-heat-transfer-and-fluid-flow-ad ... nal-and-physical-processes-in-mechanics-and-therm.pdf.

Compartmental Fluid-Flow Modelling in Packet ...
the tutorial paper [45] and also, for instance, [3], [17], [24], [25], [27],. [39], [46], [60] ..... 28. Chapter 2. Fluid flow modelling of a FIFO buffer. Equilibrium λ r(x) [pps].

Genetic Algorithms in Search, Optimization, and ...
Book sinopsis. Genetic Algorithms in Search, Optimization and Machine Learning This book describes the theory, operation, and application of genetic ...

Online PDF Viscous Fluid Flow
... is an open access online peer reviewed international journal that publishes research Composition M P 176 C S L Toxic Eutectic Comments Sn Pb Ag Cu Sb Bi ...

17426-Fluid Flow Operation.pdf
a) Draw the diagram and state the application of (a) Tee (b) plug. b) What is NPSH? Give the formula to calculate it. c) Explain the construction of Rotameter.