private static double CalcCompliance(int WX, int WY, int penal, KXMSolvers kx, double[] Rd1, double[] U)
{
var ke = kx.KE;
double c = 0.0;
for (int elx = 0; elx < WX; elx++)
{
for (int ely = 0; ely < WY; ely++)
{
int n1 = (WY + 1) * elx + ely;
int n2 = (WY + 1) * (elx + 1) + ely;
var edof = np.array(2 * n1, 2 * n1 + 1, 2 * n2, 2 * n2 + 1, 2 * n2 + 2, 2 * n2 + 3, 2 * n1 + 2, 2 * n1 + 3);
double[] Ue = new double[8];
for (int i = 0; i < edof.size; i++)
{
Ue[i] = U[edof[i]];
}
double xp = Math.Pow(Rd1[ely + elx * WY], penal);
double[] KEUe = MatrixMath.MatVecProd(ke, Ue);
double UEKUE = 0;
for (int i = 0; i < KEUe.Length; i++)
{
UEKUE += Ue[i] * KEUe[i];
}
c += xp * UEKUE;
}
}
return(c);
}
private static double CalcTree(int WX, int WY, double volfraq, double[,] retval, int penal)
{
KXMSolvers kx = new KXMSolvers();
var Rd1 = MatrixMath.To1D(retval);
var U = kx.FE(WX, WY, Rd1, penal);
/* double[,] V;
* var lambdas = kx.Modal(WX, WY, Rd1, penal, 1, out V);
*
* List<double> womega = new List<double>();
* int iwReal = 0;
* for (int i = 0; i < lambdas.Length; i++)
* {
* double freq = Math.Sqrt(lambdas[i]) / (2 * Math.PI);
* womega.Add(freq);
* if (freq <= 1.0 || double.IsNaN(freq))
* {
* iwReal++;
* }
* }
* if (iwReal >= womega.Count && womega.Count > 0)
* {
* iwReal = womega.Count - 1;
* }
* else if (womega.Count == 0)
* {
* womega.Add(10000);
* iwReal = 0;
* }
*
* double sumV = Rd1.Sum();
* sumV = sumV - (double)(WX * WY) * volfraq;
* double F = 0;
* double F1 = 50.0;
* double F2 = 50.05;
* double F3 = 75.382;
* if (womega.Count < iwReal + 3)
* {
* F = 1e6;
* }
* else
* {
* double d1 = womega[iwReal] - F1;
* double d2 = womega[iwReal + 1] - F2;
* double d3 = womega[iwReal + 2] - F3;
* F = (d1 * d1 + d2 * d2 + d3 * d3) * 1000.0;
* }
* /**/
double c = CalcCompliance(WX, WY, penal, kx, Rd1, U);
double sumV = Rd1.Sum();
sumV = sumV - (double)(WX * WY) * volfraq;
double F = c * c + sumV * sumV;
return(F);
}
public static double[] Test_eval_weave_Modal(Genome <int> g, int WX, int WY, double volfraq, double F1, double F2, double F3)
{
double[] retval = new double[1];
int penal = 3;
var ge = convertWeaverToField(g, WX, WY);
KXMSolvers skx = new KXMSolvers();
double[,] U;
var lambdas = skx.Modal(WX, WY, ge, penal, penal, out U);
List <double> womega = new List <double>();
int iwReal = 0;
for (int i = 0; i < lambdas.Length; i++)
{
double freq = Math.Sqrt(lambdas[i]) / (2 * Math.PI);
womega.Add(freq);
if (freq <= 1.0 || double.IsNaN(freq))
{
iwReal++;
}
}
if (iwReal >= womega.Count && womega.Count > 0)
{
iwReal = womega.Count - 1;
}
else if (womega.Count == 0)
{
womega.Add(10000);
iwReal = 0;
}
if (womega.Count > iwReal + 3)
{
g.Addata = "";
for (int i = iwReal; i < iwReal + 3; i++)
{
g.Addata += "\t" + womega[i];
}
g.defs = U;
}
double sumV = ge.Sum();
sumV = sumV - (double)(WX * WY) * volfraq;
if (womega.Count < iwReal + 3)
{
retval[0] = sumV * sumV + Math.Pow(womega[iwReal] - 28.16, 2) * 1000.0;
}
else
{
double d1 = womega[iwReal] - F1;
double d2 = womega[iwReal + 1] - F2;
double d3 = womega[iwReal + 2] - F3;
retval[0] = (d1 * d1 + d2 * d2 + d3 * d3) * 1000.0;
}
return(retval);
}
public static double[] Harmonic_eval_simp(Genome <double> g, int WX, int WY, double volfraq, double f1, double f2, double f3)
{
double[] retval = new double[1];
int penal = 3;
int qenal = 1;
KXMSolvers skx = new KXMSolvers();
Dictionary <int, double> SelectedUs = new Dictionary <int, double>();
List <double> AFC = new List <double>();
for (double w = f1; w <= f2; w += f3)
{
var U = skx.Harmonic(WX, WY, g.Genom, penal, qenal, w);
double multW = Math.Pow(2 * Math.PI * w, 2);
double CornerDeform = Math.Sqrt(U[2 * (WY + 1) * (WX + 1) - 1] * U[2 * (WY + 1) * (WX + 1) - 1] + U[2 * (WY + 1) * (WX + 1) - 2] * U[2 * (WY + 1) * (WX + 1) - 2]);
AFC.Add(CornerDeform);
for (int elx = 0; elx <= WX; elx++)
{
for (int ely = 0; ely <= WY; ely++)
{
if (ely == WY / 2)
{
double Mass = g.Genom[Math.Min(ely, WY - 1) + Math.Min(elx, WX - 1) * WY];
int ind = 2 * ((WY + 1) * elx + ely);
if (SelectedUs.ContainsKey(ind))
{
SelectedUs[ind] += multW * U[2 * ((WY + 1) * elx + ely)] * Mass;
}
else
{
SelectedUs.Add(ind, multW * U[2 * ((WY + 1) * elx + ely)] * Mass);
}
if (SelectedUs.ContainsKey(ind + 1))
{
SelectedUs[ind] += multW * U[2 * ((WY + 1) * elx + ely) + 1] * Mass;
}
else
{
SelectedUs.Add(ind + 1, multW * U[2 * ((WY + 1) * elx + ely) + 1] * Mass);
}
}
}
}
}
foreach (var q in SelectedUs.Keys.ToList())
{
SelectedUs[q] = SelectedUs[q] / ((f2 - f1) / f3);
}/**/
}/**/
for (int inner = 0; inner < 1; inner++)
{
var Ukx = skx.FE_inded(WX, WY, g.Genom, SelectedUs, penal);
var ke = skx.KE;
double c = 0.0;
double[] dc = new double[WX * WY];
for (int elx = 0; elx < WX; elx++)
{
for (int ely = 0; ely < WY; ely++)
{
int n1 = (WY + 1) * elx + ely;
int n2 = (WY + 1) * (elx + 1) + ely;
var edof = np.array(2 * n1, 2 * n1 + 1, 2 * n2, 2 * n2 + 1, 2 * n2 + 2, 2 * n2 + 3, 2 * n1 + 2, 2 * n1 + 3);
double[] Ue = new double[8];
for (int i = 0; i < edof.size; i++)
{
Ue[i] = Ukx[edof[i]];
}
double xp = Math.Pow(g.Genom[ely + elx * WY], penal);
double[] KEUe = MatrixMath.MatVecProd(ke, Ue);
double UEKUE = 0;
for (int i = 0; i < KEUe.Length; i++)
{
UEKUE += Ue[i] * KEUe[i];
}
c += xp * UEKUE;
dc[ely + elx * WY] = -penal *Math.Pow(g.Genom[ely + elx * WY], penal - 1) * UEKUE;
}
}
dc = skx.Check(WX, WY, 1.5, g.Genom, dc);
g.Genom = skx.OC(WX, WY, g.Genom, volfraq, dc);
}
return(AFC.ToArray());
}
public static double[] Test_eval_simp(Genome <double> g, int WX, int WY, double volfraq, double f1, double f2, double f3)
{
double[] retval = new double[1];
int penal = 3;
KXMSolvers skx = new KXMSolvers();
var U = skx.FE(WX, WY, g.Genom, penal);
var ke = skx.KE;
double c = 0.0;
double[] dc = new double[WX * WY];
for (int elx = 0; elx < WX; elx++)
{
for (int ely = 0; ely < WY; ely++)
{
int n1 = (WY + 1) * elx + ely;
int n2 = (WY + 1) * (elx + 1) + ely;
var edof = np.array(2 * n1, 2 * n1 + 1, 2 * n2, 2 * n2 + 1, 2 * n2 + 2, 2 * n2 + 3, 2 * n1 + 2, 2 * n1 + 3);
double[] Ue = new double[8];
for (int i = 0; i < edof.size; i++)
{
Ue[i] = U[edof[i]];
}
double xp = Math.Pow(g.Genom[ely + elx * WY], penal);
double[] KEUe = MatrixMath.MatVecProd(ke, Ue);
double UEKUE = 0;
for (int i = 0; i < KEUe.Length; i++)
{
UEKUE += Ue[i] * KEUe[i];
}
c += xp * UEKUE;
dc[ely + elx * WY] = -penal *Math.Pow(g.Genom[ely + elx *WY], penal - 1) * UEKUE;
}
}
dc = skx.Check(WX, WY, 1.5, g.Genom, dc);
g.Genom = skx.OC(WX, WY, g.Genom, volfraq, dc);
double sumV = 0;
foreach (var gg in g.Genom)
{
sumV += (double)gg;
}
sumV = sumV - WX * WY * volfraq;
retval[0] = c;
return(retval);
}
public static double Test_eval(Genome <double> g, int WX, int WY, double volfraq)
{
double retval = 0.0f;
int penal = 3;
KXMSolvers skx = new KXMSolvers();
var U = skx.FE(WX, WY, g.Genom, penal);
var ke = skx.KE;
double c = 0.0;
for (int elx = 0; elx < WX; elx++)
{
for (int ely = 0; ely < WY; ely++)
{
int n1 = (WY + 1) * elx + ely;
int n2 = (WY + 1) * (elx + 1) + ely;
var edof = np.array(2 * n1, 2 * n1 + 1, 2 * n2, 2 * n2 + 1, 2 * n2 + 2, 2 * n2 + 3, 2 * n1 + 2, 2 * n1 + 3);
double[] Ue = new double[8];
for (int i = 0; i < edof.size; i++)
{
Ue[i] = U[edof[i]];
}
double xp = Math.Pow(g.Genom[ely + elx * WY], penal);
double[] KEUe = MatrixMath.MatVecProd(ke, Ue);
double UEKUE = 0;
for (int i = 0; i < KEUe.Length; i++)
{
UEKUE += Ue[i] * KEUe[i];
}
c += xp * UEKUE;
}
}
// g.Genom = skx.Check(WX, WY, 1.5, g.Genom);
double sumV = 0;
foreach (var gg in g.Genom)
{
sumV += (double)gg;
}
sumV = sumV - WX * WY * volfraq;
retval = c * c + sumV * sumV * 2.0;
return(retval);
}
/**/
public static double Test_eval_weave(Genome <int> g, int WX, int WY, double volfraq)
{
double retval = 0;
int penal = 3;
var ge = convertWeaverToField(g, WX, WY);
KXMSolvers skx = new KXMSolvers();
var U = skx.FE(WX, WY, ge, penal);
var ke = skx.KE;
double c = 0.0;
for (int elx = 0; elx < WX; elx++)
{
for (int ely = 0; ely < WY; ely++)
{
int n1 = (WY + 1) * elx + ely;
int n2 = (WY + 1) * (elx + 1) + ely;
var edof = np.array(2 * n1, 2 * n1 + 1, 2 * n2, 2 * n2 + 1, 2 * n2 + 2, 2 * n2 + 3, 2 * n1 + 2, 2 * n1 + 3);
double[] Ue = new double[8];
for (int i = 0; i < edof.size; i++)
{
Ue[i] = U[edof[i]];
}
double xp = Math.Pow(ge[ely + elx * WY], penal);
double[] KEUe = MatrixMath.MatVecProd(ke, Ue);
double UEKUE = 0;
for (int i = 0; i < KEUe.Length; i++)
{
UEKUE += Ue[i] * KEUe[i];
}
c += xp * UEKUE;
}
}
double sumV = ge.Sum();
sumV = sumV - (double)(WX * WY) * volfraq;
retval = c * c * 1.0 + sumV * sumV * 0.0;
return(retval);
}
public static double[,] GoSearch(IProgress <Image> progress, int WX, int WY, double volfraq, IProgress <double> prog)
{
Random r = new Random();
double[,] retval = new double[WX, WY];
QuadTree <double> SearchTree = new QuadTree <double>();
SearchTree.Root.BB = new System.Windows.Rect(0, 0, WX, WY);
SearchTree.Root.NodeValue = 1.0;
List <double> Graph = new List <double>();
float MultySize = 20.0f;
Image MyImage = new Bitmap((int)((WX * 4) * MultySize), (int)((WY * 4) * MultySize));
Graphics g = Graphics.FromImage(MyImage);
for (int h = 0; h < 10000; h++)
{
var SearchRootList = SearchTree.getLeafs();
double step = 0.3 * r.NextDouble() - 0.1;
List <double> Gradient = new List <double>();
retval = convertTree2Mesh(SearchTree, WX, WY);
int penal = 3;
double Fzero = CalcTree(WX, WY, volfraq, retval, penal);
Graph.Insert(0, Fzero);
foreach (var q in SearchRootList)
{
var oldVal = q.NodeValue;
q.NodeValue = Math.Min(1.0, Math.Max(0.001, q.NodeValue + step));
retval = convertTree2Mesh(SearchTree, WX, WY);
double F = CalcTree(WX, WY, volfraq, retval, penal);
double Sum = MatrixMath.To1D(retval).Sum();
Gradient.Add((F - Fzero) / step);
q.NodeValue = oldVal;
///
for (int ix = 0; ix < WX; ix++)
{
for (int iy = 0; iy < WY; iy++)
{
double H = 1.0 - (float)retval[ix, iy];
H = Math.Max(0, Math.Min(1.0, H));
if (double.IsNaN(H))
{
H = 0;
}
H = H * 255;
Color customColor = Color.FromArgb(255, (int)H, (int)H, (int)H);
SolidBrush shadowBrush = new SolidBrush(customColor);
g.FillRectangle(shadowBrush, ix * MultySize, iy * MultySize, MultySize, MultySize);
}
}
for (int ix = 0; ix < WX; ix++)
{
for (int iy = 0; iy < WY; iy++)
{
double H = 1.0 - (float)retval[ix, iy];
H = Math.Max(0, Math.Min(1.0, H));
if (double.IsNaN(H))
{
H = 0;
}
H = H * 255;
Color customColor = Color.FromArgb(255, (int)H, (int)H, (int)H);
SolidBrush shadowBrush = new SolidBrush(customColor);
g.FillRectangle(shadowBrush, ix * MultySize + MultySize * WX + 5, iy * MultySize, MultySize, MultySize);
}
}
progress.Report(MyImage);
}
float dY = 500;
g.FillRectangle(new SolidBrush(Color.White), 0, dY - 150, 500, dY + 150);
g.DrawLine(new Pen(Color.Black, 2), 0, dY, 500, dY);
double data = Graph[Graph.Count - 1];
double olddata = Graph[Graph.Count - 1];
float MaxG = (float)(Graph.Max() + 0.00001f);
Debug.WriteLine(Graph.First());
for (int i = 0; i < Math.Min(500, Graph.Count); i++)
{
if (!double.IsNaN(Graph[i]))
{
data = Graph[i];
g.DrawLine(new Pen(Color.Blue, 2), Math.Min(500, Graph.Count) - i, dY - 150.0f * (float)olddata / MaxG, Math.Min(500, Graph.Count) - i - 1.0f, dY - 150.0f * (float)data / MaxG);
olddata = data;
progress.Report(MyImage);
}
}
SearchRootList = SearchRootList.OrderBy(d => (Gradient[SearchRootList.IndexOf(d)])).ToList();
KXMSolvers kx = new KXMSolvers();
var xnew = kx.OC(Gradient.Count, 1, SearchTree.GetLeafsValues().ToArray(), 0.5, Gradient.ToArray());
double gMx = Gradient.Max();
double gMn = Gradient.Min();
foreach (var q in SearchRootList)
{
double delta = -Gradient[SearchRootList.IndexOf(q)];
if (!q.IsLast)
{
//q.NodeValue = Math.Min(1.0, Math.Max(0.001, xnew[SearchRootList.IndexOf(q)]));
//q.NodeValue = Math.Min(1.0, Math.Max(0.001, q.NodeValue*Math.Sqrt(delta)));
q.NodeValue = Math.Min(1.0, Math.Max(0.001, q.NodeValue + step));
if (Math.Abs(q.NodeValue - 1.0) < 0.1)
{
q.SubDivide(q.NodeValue / 2);
}
break;
}
else
{
q.NodeValue = Math.Min(1.0, Math.Max(0.001, q.NodeValue + step));
}
}
}
g.Dispose();
retval = convertTree2Mesh(SearchTree, WX, WY);
return(retval);
}