Computation of the Hessian matrix in aerodynamic inverse design using continuous adjoint formulations

Continuous adjoint formulations for the computation of (first and) second order derivatives of the objective function governing inverse design problems in 2D inviscid flows are presented. These are prerequisites for the use of the very efficient exact Newton method. Four new formulations based on all possible combinations of the direct differentiation method and the continuous adjoint approach to compute the sensitivity derivatives of objective functions, constrained by the flow equations, are presented. They are compared in terms of the expected CPU cost to compute the Hessian of the objective function used in single-objective optimization problems with N degrees of freedom. The less costly among them was selected for further study and tested in inverse design problems solved by means of the Newton method. The selected approach, which will be referred to as the direct-adjoint one, since it performs direct differentiation for the gradient and, then, uses the adjoint approach to compute the Hessian, requires as many as N+2 equivalent flow solutions for each Newton step. The major part of the CPU cost (N equivalent flow solutions) is for the computation of the gradient but, fortunately, this task is directly amenable to parallelization. The method is used to reconstruct ducts or cascade airfoils for a known pressure distribution along their solid boundaries, at inviscid flow conditions. The examined cases aim at demonstrating the accuracy of the proposed method in computing the exact Hessian matrix as well as the efficiency of the exact Newton method as an optimization tool in aerodynamic design.     

Structural optimization using global stress-deviation objective function via the level-set method

The paper deals with minimum stress design using a novel stress-related objective function based on the global stress-deviation measure. The shape derivative, representing the shape sensitivity analysis of the structure domain, is determined for the generalized form of the global stress-related objective function. The optimization procedure is based on the domain boundary evolution via the level-set method. The elasticity equations are, instead of using the usual ersatz material approach, solved by the extended finite element method. The Hamilton-Jacobi equation is solved using the streamline diffusion finite element method. The use of finite element based methods allows a unified numerical approach with only one numerical framework for the mechanical problem as also for the boundary evolution stage. The numerical examples for the L-beam benchmark and the notched beam are given. The results of the structural optimization problem, in terms of maximum von Mises stress corresponding to the obtained optimal shapes, are compared for the commonly used global stress measure and the novel global stress-deviation measure, used as the stress-related objective functions.     

Introducing the sequential linear programming level-set method for topology optimization

This paper introduces an approach to level-set topology optimization that can handle multiple constraints and simultaneously optimize non-level-set design variables. The key features of the new method are discretized boundary integrals to estimate function changes and the formulation of an optimization sub-problem to attain the velocity function. The sub-problem is solved using sequential linear programming (SLP) and the new method is called the SLP level-set method. The new approach is developed in the context of the Hamilton-Jacobi type level-set method, where shape derivatives are employed to optimize a structure represented by an implicit level-set function. This approach is sometimes referred to as the conventional level-set method. The SLP level-set method is demonstrated via a range of problems that include volume, compliance, eigenvalue and displacement constraints and simultaneous optimization of non-level-set design variables.     

Efficient symmetric Hessian propagation for direct optimal control

Direct optimal control algorithms first discretize the continuous-time optimal control problem and then solve the resulting finite dimensional optimization problem. If Newton type optimization algorithms are used for solving the discretized problem, accurate first as well as second order sensitivity information needs to be computed. This article develops a novel approach for computing Hessian matrices which is tailored for optimal control. Algorithmic differentiation based schemes are proposed for both discrete- and continuous-time sensitivity propagation, including explicit as well as implicit systems of equations. The presented method exploits the symmetry of Hessian matrices, which typically results in a computational speedup of about factor 2 over standard differentiation techniques. These symmetric sensitivity equations additionally allow for a three-sweep propagation technique that can significantly reduce the memory requirements, by avoiding the need to store a trajectory of forward sensitivities. The performance of this symmetric sensitivity propagation is demonstrated for the benchmark case study of the economic optimal control of a nonlinear biochemical reactor, based on the open-source software implementation in the ACADO Toolkit.     

Aerodynamic Shape Optimization Using First and Second Order Adjoint and Direct Approaches

This paper focuses on discrete and continuous adjoint approaches and direct differentiation methods that can efficiently be used in aerodynamic shape optimization problems. The advantage of the adjoint approach is the computation of the gradient of the objective function at cost which does not depend upon the number of design variables. An extra advantage of the formulation presented below, for the computation of either first or second order sensitivities, is that the resulting sensitivity expressions are free of field integrals even if the objective function is a field integral. This is demonstrated using three possible objective functions for use in internal aerodynamic problems; the first objective is for inverse design problems where a target pressure distribution along the solid walls must be reproduced; the other two quantify viscous losses in duct or cascade flows, cast as either the reduction in total pressure between the inlet and outlet or the field integral of entropy generation. From the mathematical point of view, the three functions are defined over different parts of the domain or its boundaries, and this strongly affects the adjoint formulation. In the second part of this paper, the same discrete and continuous adjoint formulations are combined with direct differentiation methods to compute the Hessian matrix of the objective function. Although the direct differentiation for the computation of the gradient is time consuming, it may support the adjoint method to calculate the exact Hessian matrix components with the minimum CPU cost. Since, however, the CPU cost is proportional to the number of design variables, a well performing optimization scheme, based on the exactly computed Hessian during the starting cycle and a quasi Newton (BFGS) scheme during the next cycles, is proposed.     

Total pressure losses minimization in turbomachinery cascades using the exact Hessian

A method for the design of turbomachinery cascades with minimum total pressure losses, subject to constraints on the minimum blade thickness and flow turning, is presented. It is based on the Newton–Lagrange method which requires the computation of first- and second-order sensitivities of the objective function and the constraints, with respect to the design variables. The computation of the exact Hessian of the function which expresses the difference in total pressure between the inlet to and the outlet from the cascade, is new in the literature. To compute the Hessian, the direct differentiation of the viscous flow equations is used for the first-order sensitivities of the functional and the flow-related constraints, followed by the discrete adjoint method. Since the objective function is defined along boundaries other than those controlled by the design variables, it is challenging to investigate the significance of all terms comprising the exact second-order sensitivity expressions. All these terms were temporarily computed using automatic differentiation and those which proved to be significant are hand-differentiated to minimize CPU cost and memory requirements. Insignificant terms are eliminated, giving rise to the so-called “exact” Hessian matrix. An “exactly” initialized quasi-Newton method was also programmed and tested. In the latter, at the first cycle, the exact gradients and Hessians are computed and used; during the subsequent optimization cycles, the discrete adjoint method provides the exact gradient whereas the Hessian is updated as in quasi-Newton methods. The comparison of the efficiency of the aforementioned methods depends on the number of design variables used; the “exactly” initialized quasi-Newton method constantly outperforms its conventional variant in terms of CPU cost, particularly in non-convex and/or constrained optimization problems.     

Solution to boundary shape optimization problem of linear elastic continua with prescribed natural vibration mode shapes

This paper presents a practical method of numerical analysis for boundary shape optimization problems of linear elastic continua in which natural vibration modes approach prescribed modes on specified sub-boundaries. The shape gradient for the boundary shape optimization problem is evaluated with optimality conditions obtained by the adjoint variable method, the Lagrange multiplier method, and the formula for the material derivative. Reshaping is accomplished by the traction method, which has been proposed as a solution to boundary shape optimization problems of domains in which boundary value problems of partial differential equations are defined. The validity of the presented method is confirmed by numerical results of three-dimensional beam-like and plate-like continua.     

非结构网格上解二维Hamilton-Jacobi方程的一种有限体积方法

本文利用最小二乘插值的思想,发展了一类在非结构网格上解Hamilton-Jacobi方程的方法．此方法通过确定超定线性方程组来得到所求单元上的二次插值多项式,并利用极值原理的思想,保证其数值解的导数不出现新的极值．典型算例表明此方法计算速度快,对间断有很好的分辨能力．     

Topological shape optimization of geometrically nonlinear structures using level set method

Using the level set method, a topological shape optimization method is developed for geometrically nonlinear structures in total Lagrangian formulation. The structural boundaries are implicitly represented by the level set function, obtainable from “Hamilton-Jacobi type” equation with “up-wind scheme,” embedded into a fixed initial domain. The method minimizes the compliance through the variations of implicit boundary, satisfying an allowable volume requirement. The required velocity field to solve the Hamilton-Jacobi equation is determined by the descent direction of Lagrangian derived from an optimality condition. Since the homogeneous material property and implicit boundary are utilized, the convergence difficulty is significantly relieved.     

A level-set method for vibration and multiple loads structural optimization

We extend the level-set method for shape and topology optimization to new objective functions such as eigenfrequencies and multiple loads. This method is based on a combination of the classical shape derivative and of the Osher–Sethian level-set algorithm for front propagation. In two and three space dimensions we maximize the first eigenfrequency or we minimize a weighted sum of compliances associated to different loading configurations. The shape derivative is used as an advection velocity in a Hamilton–Jacobi equation for changing the shape. This level-set method is a low-cost shape capturing algorithm working on a fixed Eulerian mesh and it can easily handle topology changes.     

On reachability and minimum cost optimal control

Questions of reachability for continuous and hybrid systems can be formulated as optimal control or game theory problems, whose solution can be characterized using variants of the Hamilton-Jacobi-Bellman or Isaacs partial differential equations. The formal link between the solution to the partial differential equation and the reachability problem is usually established in the framework of viscosity solutions. This paper establishes such a link between reachability, viability and invariance problems and viscosity solutions of a special form of the Hamilton-Jacobi equation. This equation is developed to address optimal control problems where the cost function is the minimum of a function of the state over a specified horizon. The main advantage of the proposed approach is that the properties of the value function (uniform continuity) and the form of the partial differential equation (standard Hamilton-Jacobi form, continuity of the Hamiltonian and simple boundary conditions) make the numerical solution of the problem much simpler than other approaches proposed in the literature. This fact is demonstrated by applying our approach to a reachability problem that arises in flight control and using numerical tools to compute the solution.     

A Newton method for Bernoulli’s free boundary problem in three dimensions

Summary The present paper is dedicated to the numerical solution of Bernoulli’s free boundary problem in three dimensions. We reformulate the given free boundary problem as a shape optimization problem and compute the shape gradient and Hessian of the given shape functional. To approximate the shape problem we apply a Ritz–Galerkin discretization. The necessary optimality condition is resolved by Newton’s method. All information of the state equation, required for the optimization algorithm, are derived by boundary integral equations which we solve numerically by a fast wavelet Galerkin scheme. Numerical results confirm that the proposed Newton method yields an efficient algorithm to treat the considered class of problems.      

Shape design sensitivity analysis of nonlinear structural systems

A unified approach is presented for shape design sensitivity analysis of nonlinear structural systems that include trusses and beams. Both geometric and material nonlinearities are considered. Design variables that specify the shape of components of built-up structures are treated, using the continuum equilibrium equations and the material derivative concept. To best utilize the basic character of the finite element method, shape design sensitivity information is expressed as domain integrals. For numerical evaluation of shape design sensitivity expressions, two alternative methods are presented: the adjoint variable and direct differentiation methods. Advantages and disadvantages of each method are discussed. Using the domain formulation of shape design sensitivity analysis, and the adjoint variable and direct differentiation methods, design sensitivity expressions are derived in the continuous setting in terms of shape design variations. A numerical method to implement the shape design sensitivity analysis, using established finite element codes, is discussed. Unlike conventional methods, the current approach does not require differentiation of finite element stiffness and mass matrices.     

Single scale wavelet approximations in layout optimization

The standard structural layout optimization problem in 2D elasticity is solved using a wavelet based discretization of the displacement field and of the spatial distribution of material. A fictitious domain approach is used to embed the original design domain within a simpler domain of regular geometry. A Galerkin method is used to derive discretized equations, which are solved iteratively using a preconditioned conjugate gradient algorithm. A special preconditioner is derived for this purpose. The method is shown to converge at rates that are essentially independent of discretization size, an advantage over standard finite element methods, whose convergence rate decays as the mesh is refined. This new approach may replace finite element methods in very large scale problems, where a very fine resolution of the shape is needed. The derivation and examples focus on 2D-problems but extensions to 3D should involve only few changes in the essential features of the procedure.     

Global regulation of flexible joint robots using approximatedifferentiation

Tomei (1991) presented a globally asymptotically stable PD regulator for robots with flexible joints. A drawback of this scheme is that, as is well known, noise in velocity measurements degrades performance, and numerical differentiation is inaccurate for low and high speeds. On the other hand, approximate differentiation, replacing the derivative operator p by the high pass filter (bp/p+a), is commonly used in applications since it yields good behavior for regulation tasks. In this paper, we show that velocity measurement in Tomei's scheme can be replaced by approximate differentiation preserving global asymptotic stability for all positive values of b and a. Simulations that illustrate our result are also presented     

A new high accuracy method for two-dimensional biharmonic equation with nonlinear third derivative terms: application to Navier–Stokes equations of motion

In this paper, we propose a new compact fourth-order accurate method for solving the two-dimensional fourth-order elliptic boundary value problem with third-order nonlinear derivative terms. We use only 9-point single computational cell in the scheme. The proposed method is then employed to solve Navier–Stokes equations of motion in terms of streamfunction–velocity formulation, and the lid-driven square cavity problem. We describe the derivation of the method in details and also discuss how our streamfunction–velocity formulation is able to handle boundary conditions in terms of normal derivatives. Numerical results show that the proposed method enables us to obtain oscillation-free high accuracy solution.     

A new two-step gradient-type method for large-scale unconstrained optimization

In this paper, we propose some improvements on a new gradient-type method for solving large-scale unconstrained optimization problems, in which we use data from two previous steps to revise the current approximate Hessian. The new method which we considered, resembles to that of Barzilai and Borwein (BB) method. The innovation features of this approach consist in using approximation of the Hessian in diagonal matrix form based on the modified weak secant equation rather than the multiple of the identity matrix in the BB method. Using this approach, we can obtain a higher order accuracy of Hessian approximation when compares to other existing BB-type method. By incorporating a simple monotone strategy, the global convergence of the new method is achieved. Practical insights into the effectiveness of the proposed method are given by numerical comparison with the BB method and its variant.     

COOPT — a software package for optimal control of large-scale differential–algebraic equation systems

This paper describes the functionality and implementation of COOPT. This software package implements a direct method with modified multiple shooting type techniques for solving optimal control problems of large-scale differential–algebraic equation (DAE) systems. The basic approach in COOPT is to divide the original time interval into multiple shooting intervals, with the DAEs solved numerically on the subintervals at each optimization iteration. Continuity constraints are imposed across the subintervals. The resulting optimization problem is solved by sparse sequential quadratic programming (SQP) methods. Partial derivative matrices needed for the optimization are generated by DAE sensitivity software. The sensitivity equations to be solved are generated via automatic differentiation.COOPT has been successfully used in solving optimal control problems arising from a wide variety of applications, such as chemical vapor deposition of superconducting thin films, spacecraft trajectory design and contingency/recovery problems, and computation of cell traction forces in tissue engineering.