Math 2.0

Calculus of variations and optimal control; optimization Definition

Shape derivative

Builds on FunctionalFréchet derivativeNormal vector

Definition

Let d≥1d \ge 1 and let W=W1,∞(ℝd;ℝd)W = W^{1,\infty}(\R^d; \R^d) be the Banach space of bounded Lipschitz vector fields θ ⁣:ℝd→ℝd\theta \colon \R^d \to \R^d, with norm ∥θ∥W=∥θ∥L∞+∥∇θ∥L∞\norm{\theta}_W = \norm{\theta}_{L^\infty} + \norm{\nabla \theta}_{L^\infty}. Let 𝒪\mathcal{O} be a class of open subsets of ℝd\R^d, and let J ⁣:𝒪→ℝJ \colon \mathcal{O} \to \R be a shape functional. Let Ω∈𝒪\Omega \in \mathcal{O} be such that (Id+θ)(Ω)∈𝒪(\mathrm{Id} + \theta)(\Omega) \in \mathcal{O} whenever ∥θ∥W\norm{\theta}_W is small enough. Then JJ is shape differentiable at Ω\Omega if the map θ↦J((Id+θ)(Ω))\theta \mapsto J((\mathrm{Id} + \theta)(\Omega)) is Fréchet differentiable at θ=0\theta = 0. Its Fréchet derivative, a bounded linear map J′(Ω) ⁣:W→ℝJ'(\Omega) \colon W \to \R, is the shape derivative of JJ at Ω\Omega.

Remarks

For ∥θ∥W<1\norm{\theta}_W < 1 the map Id+θ\mathrm{Id} + \theta is a bi-Lipschitz homeomorphism of ℝd\R^d, so (Id+θ)(Ω)(\mathrm{Id} + \theta)(\Omega) is again open. Example: the volume J(Ω)=∣Ω∣J(\Omega) = \abs{\Omega} of a bounded open set. By the change of variables formula J((Id+θ)(Ω))=∫Ωdet⁡(I+∇θ) dxJ((\mathrm{Id} + \theta)(\Omega)) = \int_\Omega \det(I + \nabla \theta) \, dx. Since det⁡(I+∇θ)=1+div⁡θ+O(∥∇θ∥2)\det(I + \nabla \theta) = 1 + \operatorname{div} \theta + O(\norm{\nabla\theta}^2), the shape derivative is J′(Ω)(θ)=∫Ωdiv⁡θ dxJ'(\Omega)(\theta) = \int_\Omega \operatorname{div} \theta \, dx. If Ω\Omega has a C1C^1 boundary with outward unit normal vector nn, the divergence theorem gives J′(Ω)(θ)=∫∂Ωθ⋅n dsJ'(\Omega)(\theta) = \int_{\partial \Omega} \theta \cdot n \, ds. Non-example: J(Ω)=∣∣Ω∣−1∣J(\Omega) = \abs{\abs{\Omega} - 1} at a set Ω\Omega with ∣Ω∣=1\abs{\Omega} = 1 is not shape differentiable. Take Ω\Omega bounded and a smooth function χ\chi with compact support that equals 11 on a neighbourhood of Ω‾\overline{\Omega}. Along θs(x)=sχ(x)x\theta_s(x) = s \chi(x) x one gets (Id+θs)(Ω)=(1+s)Ω(\mathrm{Id} + \theta_s)(\Omega) = (1 + s) \Omega and J=∣(1+s)d−1∣J = \abs{(1 + s)^d - 1}, which has a corner at s=0s = 0. For smooth Ω\Omega and smooth JJ, the Hadamard structure theorem gives J′(Ω)(θ)=∫∂Ωg (θ⋅n) dsJ'(\Omega)(\theta) = \int_{\partial \Omega} g \, (\theta \cdot n) \, ds for a function gg on ∂Ω\partial \Omega, the shape gradient.

Reading path

To understand this page from the ground up, read these 11 pages in order. The foundations come first.

Show the path as a graph
Each page uses the pages above it. The colour of the bar on a box shows how strong the evidence is, and the box also gives it as text. The list has the same pages.
  • no evidence level
  1. Angle between two vectors concept
  2. Euclidean space object
  3. Functional concept
  4. Length of a vector concept
  5. Line concept
  6. Plane concept
  7. Fréchet derivative definition
  8. Geometric formula for the dot product statement
  9. Vectors are perpendicular if and only if their dot product is zero statement
  10. Normal vector definition
  11. Shape derivative definition

Further reading

  1. Web https://en.wikipedia.org/wiki/Shape_optimization
About this page
TagRCSXGDTTET
URIm2://RCSXGDTTET
Kinddefinition
WritingW0 stub
MSC49Q10, 49Q12
DefinesShape derivative
UsesFunctional, Fréchet derivative, Normal vector

Markdown · JSON · Edit on GitHub

What this page needs for W1

  • an inventory row: the row of the inventory that the node accounts for (0 of 1)
  • a source with a locator (0 of 1)

The writing levels say how complete the text of a page is. They do not measure its truth.

What links here

No page links to this page yet.