🔒 Locked 🎮 Unity

Drag and Lift per triangle (boat simulation)

Started by Jareso May 17, 2005 at 6:05 AM 2 replies 3k views

Original Post

May 17, 2005 at 6:05 AM

Hi I am doing high speed boat simulation (as i wrote in some threads before). For bouyancy forces calculation i use spherical approximation it works OK. But for drag and lift I found out that spherical approx. in not enough... I decided to use triangular convex hull for lift and drag calculation. For now i do it like this: Every frame i cut predefined triangular approximation hull(boat hull) by water plane so i get new hull that is completly under water and store each triangle in array pWSubTriHull. This also works OK. Calculation of lift and drag:


//--------------------------------------------------------------------------- 
void JrsBoat::UpdateLiftAndDrag(JrsWater *pWater) 
{ 
   // Loop through every submerged triangle...
   for(int i=0; i<iNumOfSubTri; i++) 
   { 
      // I need to find out if boat "vVelocity" vector is in positive half space defined
      // by triangle plane. If is than triangle will be used for calculation.
      Plane pPlane(pWSubTriHull.vA,pWSubTriHull.vB,pWSubTriHull.vC); 
      if (pPlane.WhichSide(vVelocity) > 0) 
      { 
         // Get area of this submerged triangle...
         float fArea = AreaOfTri(pWSubTriHull.vA,pWSubTriHull.vB,pWSubTriHull.vC); 

         // Direction of lift force is (vVelocity X vN) X vVelocity   
         Vector3 vDirrOfLift = Cross(Cross(vVelocity,pPlane.Normal()),vVelocity); 
         vDirrOfLift.Normalize(); 
         // Drag acts opposite of boat velocity vector. 
         Vector3 vDirrOfDrag = -vVelocity; 
         vDirrOfDrag.Normalize(); 
          
         // Angle between lift and drag vector
         float fDot = Dot(vDirrOfLift,pPlane.Normal()); 
         float fAngOfAttack = RadToDeg((float)acos(fDot)); 

         // Formula: F = 1/2 * Area * Ang * Coeff * Density * speed^2 
         // fLCoeff is lift coef, fDCoeff is drag coeff
         float fSPwTwo = vVelocity.Length() * vVelocity.Length(); 
         float fFl = 0.5f * fArea * fAngOfAttack * fLCoeff * pWater->fDensity * fSPwTwo;        
         float fFd = 0.5f * fArea * fAngOfAttack * fDCoeff * pWater->fDensity * fSPwTwo;        

         // The force acts through this point. Average of all triangle vertexes???
         Vector3 vAppPoint = (pWSubTriHull.vA+pWSubTriHull.vB+pWSubTriHull.vC)/3.0f; 

         // Lift 
         this->AddWorldForce(vDirrOfLift*fFl,vAppPoint); 
         // Drag 
         this->AddWorldForce(vDirrOfDrag*fFd,vAppPoint); 
      } 
   } 
}

This seems not to work... So questions: Can somebody explain me how to calculate drag and lift per triangle? (best on this example...) And to calculate angle of attack for lift of boat in 3D? How to make this work? A searched internet for this topic but found allmost nothing...or nothing realy usable... Thanks

John Schultz

May 17, 2005 at 3:02 PM

1. Compute velocity vector at FluidSample point: vel
2. Angle of attack (angle) is: velNormalized dot surfaceNormal. This value should be positive, if not, skip this fluid sample. You'll get velSquaredScalar when you normalize vel.
3. Flat plate lift + induced drag is: force = (angle*area*density*.5*velSquaredScalar)*surfaceNormal. Note the .5 can be precomputed into the density constant, and area*density*.5 could also be precomputed as scale.
4. addWorldForce(force,sampleLocation). sampleLocation can be the center of the triangle.

Once you have 1-4 working (note no cross products, just one dot product), you can project the velocity vector to the plane defined by surfaceNormal and sampleLocation (this involves dot product, not cross product) to compute skin drag.

To make the surface more "efficient" (less induced drag), you can use a lookup table based on angle, or subtract away part of the induced drag vector component (at very high angles of attack, you can also simulate stall (drop in total force due to energy loss through turbulence). You don't need this step to get the sytem working (you can even skip it for a boat simulation), though. Get 1-4 above working first, then try improving/tweaking behavior.

If you are still having problems after trying to implement 1-4, draw real-time lines showing the vectors for your normals, sampleLocations, computed velocity vectors, and resulting force magnitudes. Any errors in sign/direction/location will become apparent when visualizing the process real-time.

Jareso

May 18, 2005 at 10:24 AM

Hi

Thank you very much.
I have done those steps 1-4, and it works great.

I am also interested in step 5 (simulate stall,...)
Can you please tell me more about that.

Once again, thank you very much.

John Schultz

May 18, 2005 at 1:17 PM

Quote:
Original post by Jareso
Hi

Thank you very much.
I have done those steps 1-4, and it works great.

I am also interested in step 5 (simulate stall,...)
Can you please tell me more about that.

Once again, thank you very much.

This can be applied to the rudder/wings (as opposed to the boat hull): as the angle of attack increases, so does lift/force. At some maximum angle, the lift will begin to drop, and if the angle of attack increases further, the fluid will break up and become turbulent, resulting in a loss of net force against the surface. This is a stall condition. This simplest implementation cuts force when the stall condition occurs. See google and search for "flat plate stall". One example here.

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.