Klaus Deckelnick,Philip J. Herbert,Michael Hinze
Abstract:We present a general shape optimisation framework based on the method of mappings in the $W^{1,\infty}$ topology. We propose steepest descent and Newton-like minimisation algorithms for the numerical solution of the respective shape optimisation problems. Our work is built upon previous work of the authors in (Deckelnick, Herbert, and Hinze, ESAIM: COCV 28 (2022)), where a $W^{1,\infty}$ framework for star-shaped domains is proposed. To illustrate our approach we present a selection of PDE constrained shape optimisation problems and compare our findings to results from so far classical Hilbert space methods and recent $p$-approximations.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
This paper attempts to solve the problem of using methods based on \(W^{1,\infty}\) topology to find the optimal shape in shape optimization problems. Specifically, the paper proposes first - order and Newton - type minimization algorithms under \(W^{1,\infty}\) topology and applies them to shape optimization problems constrained by partial differential equations (PDEs). The main contributions of the paper are as follows:
1. **Method framework**: A shape optimization framework based on the mapping method is proposed, which is carried out under \(W^{1,\infty}\) topology. This is different from the traditional Hilbert space methods, which may encounter regularity problems when dealing with certain problems.
2. **Algorithm implementation**: Gradient - descent and Newton - type minimization algorithms are proposed for the numerical solution of shape optimization problems. These algorithms have good convergence and are easy to implement under \(W^{1,\infty}\) topology.
3. **Problem extension**: The previous research limited to star - shaped domains is extended to more general domains, and the application of Newton - type methods is considered. This makes the method more universal and practical.
4. **Numerical experiments**: The effectiveness of the proposed method is verified through a series of numerical experiments, and it is compared with the classical Hilbert space method and the recent \(p\)- approximation method.
### Formula summary
- **Objective function**:
\[
\min J(\Omega), \quad \Omega \in S
\]
where \(S\) is a set of feasible domains and \(J\) is the objective functional.
- **First - order derivative**:
\[
J'(\Omega)[V]=\int_{\hat{\Omega}}\left(j_x(\cdot, \hat{y})\cdot V + A[V]\nabla\hat{y}\cdot\nabla\hat{p}-\hat{p}\text{div}(FV)\right)d\hat{x}
\]
- **Second - order derivative**:
\[
J''(\Omega)[V, W]=\langle h_1, y'(0)[W]\rangle+\langle h_2, W\rangle
\]
where:
\[
\langle h_1, y\rangle = L_{yy}(\hat{y}, 0, \hat{p})[y'(0)[V], y]+L_{yV}(\hat{y}, 0, \hat{p})[V, y]
\]
\[
\langle h_2, W\rangle = L_{Vy}(\hat{y}, 0, \hat{p})[y'(0)[V], W]+L_{VV}(\hat{y}, 0, \hat{p})[V, W]
\]
- **Lagrangian function**:
\[
L(y, V, p)=J(V, y)+\langle p, e(V, y)\rangle
\]
- **Discretization direction selection**:
\[
V_h^*\in\arg\min\left\{\frac{t}{2}J''(\hat{\Omega})_h[V_h, V_h]+J'(\hat{\Omega})_h[V_h]:V_h\in V_h, |DV_h|\leq1\right\}
\]
### Conclusion
The paper proposes a new shape optimization algorithm through the method framework under \(W^{1,\infty}\) topology and verifies its effectiveness and superiority through numerical experiments. This method not only avoids the regularity problems in traditional methods but also is more flexible and efficient in practical applications.