《建筑几何》Chapter 18:形状优化问题

Shape optimization problems

  • The realization of a design idea in an optimal way, both from the aesthetic and functional perspective, is a difficult and complex problem area.

Remarks on mathematical optimization 数学优化注解

  • Any optimum is only an optimum within the conceptualization of the problem space and the boundary conditions applied.
  • Gradient descent.
    We form the gradient ∇f = (fx ,fy), which is the vector whose coordinates are the partial derivatives of f with respect to x and y.
    To move “downhill” toward a minimum one may always move in the direction of the negative gradient –∇f.

Geometric optimization 几何优化

  • Minimal surfaces
    Given a smooth closed curve c without self-intersections, find the surface patch with boundary curve c that has the smallest surface area A.
  • A minimal surface has vanishing mean curvature in each of its points. 极小曲面的平均曲率为0
  • a generic surface point p of a minimal surface must be a hyperbolic point (saddle-like) 极小曲面点是双曲点
  • This is a hyperbolic paraboloid whose rulings in the plane z = 0 are the two orthogonal straight lines x + y = 0 and x – y = 0. These straight lines are the asymptotic directions (directions of vanishing normal curvature). 平面z=0与双曲抛物面相交的是两条正交直线rulings,是渐近线
  • in each point of a minimal surface the asymptotic directions are orthogonal. 极小曲面的每个点,渐近方向正交
  • the bisecting lines of the (always orthogonal) principal directions are the asymptotic directions. 主法曲率线的对角线是渐近方向
  • on a minimal surface the asymptotic curves form an orthogonal curve network. It is the bisecting curve network of the network of principal curvature lines. 极小曲面,渐近线网是正交网,也是对分正交的主法曲率网得到的
  • For a minimal surface, the asymptotic curve network is also mapped to an orthogonal curve network of S. This is a result of the remarkable property that the Gaussian spherical mapping of a minimal surface S to the unit sphere S is conformal (angle preserving). 极小曲面的高斯球像是共形映射(保角),因此正交的渐近网的高斯像也是正交的

enneper 曲面、 helicoid曲面、catenoid曲面、Scherk极小曲面

  • Enneper.
  • Helicoid. 唯一的直纹极小曲面
  • Catenoid. 悬链面:唯一的旋转极小曲面
  • Helical minimal surface. 由catenoid剪开变化得到,每个过程都是极小曲面
  • Scherk’s minimal surface.
  • Surfaces with constant mean curvature
  • Willmore energy.
    Willmore energy is the surface integral of the function $(k_1 – k_2)^2$: \(E_W = ∫S(k_1 – k_2)^2dA.\)
  • Willmore energy tries to make surfaces as spherical as possible.
  • An important advantage of the Willmore energy is that one can prescribe a boundary strip (i.e., a boundary curve plus tangent planes along it) and compute an energy minimizer that interpolates these data.
  • Fair curves and polygons.
    The bending energy of an idealized thin elastic beam is proportional to the integral of the squared curvature, expressed as $E=∫c k^2 ds.$ 用$Es = (1/4) ∑(v_{k-1} + v_{k+1} – 2v_k)^2$替代求解.
  • Fair webs and mesh beautification.
  • Geometric constraints.

Functional optimization 函数优化

References and further reading