请问：我该用什么软件呢？用optimus, 还是isight或者comsol, hyperstudy?

淮南子

请教诸位：

我要做一些固体力学的优化问题和一个流固耦合的优化问题，主要是与有限元有关的优化．
不过我的问题对优化程序要求很高，对用户自己对模型的控制也要求很高。不是有限元软件里常见的质量，刚度，形状优化．象ansys的优化功能是绝对不行的．

不知道哪个程序适合我呢？COMSOL，isight, optimus,　还是hyperstudy?
听说optimus很厉害，和各个有限元软件都可以结合起来，而COMSOL的特点是它和matlab结合紧密，可以使用matlab的资源和工具箱．

导师要我自己做个调查，赶快决定用什么软件，决定后他会去购买软件，所以我一旦决定了以后就很难再改了．
顺便说一下，我要使用的有限元软件可以是ansys, abaqus, 甚至是自己的一些Ｃ或者matlab程序．

谁可以给我一个推荐呢？

淮南子

谢谢yekai大侠的帮助，能不能再问一个问题呢？

现在我还想请问诸位大拿：能不能介绍一些这些优化软件里具有哪些优化算法呢？

我知道ANSYS有一个简单的response surface的优化算法，一些DOE的程序。不知道optimus, isight， COMSOL等拥有哪些算法？

对有限元优化，我以为最利害的大概是SQP（sequencial response programming) 算法，遗传算法则最强大。
如果这个程序能够使用用户自己的优化程序，比如我自己编写的matlab或者C的程序，那就更好。

还有，如果我建立自己的模型，推导自己的gradient公式，能不能把自己的gradient公式在程序中建立起来，这可能需要用户子程序，或用户自己的有限元程序。

谢谢各位啦。

yekai

在isight中可以导出梯度。优化算法很多有20多种，其中常用模拟退火、遗传算法和一些基于梯度的。试验设计、近似建模等有多种算法，你可以在板块里面搜索一下。

molen

For your reference:

There are two popular multidisciplinary computer-aided-optimization (CAO) software that can be used for FE-model optimization: LMS-Optimus and iSIGHT.

1.        OPTIMUS

Optimus has the following optimization algorithms:

        Gradient-based nonlinear programming techniques (local methods)
o        Sequential quadratic programming
o        Conjugate gradient methods

        Genetic algorithms (global methods)
o        Self-adaptive evolutionary algorithm
o        Differential evolutionary algorithm
o        Simulated annealing

        Response surface-based methods
o        Linear, quadratic, tri
o        O-RSM model
o        User-defined surrogates

        Random search

        Feasible direction method


        Design-of-experiment (DOE)
o        Quarter Fractional Factorial
o        Latin hypercube
o        Taguchi, Plackett and Burman
o        User-difened DOE

        Multi-objective optimization
        Numerical sensitivity analysis
        Integer programming
        User-defined optimization algorithms

        Probabilistic optimization using Monte Carlo simulation
o        Design variables can be defined as Gaussian, Exponential, Rayleigh distributions
o        ser-defined probabilistic distribution

OPTIMUS has a GUI, and a script language for users to write scripts controlling the optimization process. Users can also write JAVA codes to control the optimization.

OPTIMUS is said can be connected to any FEA software and programs, including user-written codes in MATLAB and C.  


2.        ISIGHT

The optimization techniques in iSIGHT can be divided into three main categories:

        Numerical Optimization Techniques
        Exploratory Techniques
        Expert System Techniques


Numerical optimization techniques generally assume the parameter space is unimodal, convex, and continuous. The techniques including in iSIGHT are:
        ADS-based Techniques
        Exterior Penalty
        Modified Method of Feasible Directions
        Sequential Linear Programming
        Generalized Reduced Gradient - LSGRG2
        Hooke-Jeeves Direct Search Method
        Method of Feasible Directions - CONMIN
        Mixed Integer Optimization - MOST
        Sequential Quadratic Programming - DONLP
        Sequential Quadratic Programming - NLPQL
        Successive Approximation Method

The numerical optimization techniques can be further divided into the following two categories:
        direct methods
        penalty methods


        Direct Methods                                                                          
o        Generalized Reduced Gradient - LSGRG2                              
o        Method of Feasible Directions - CONMIN                              
o        Mixed Integer Optimization - MOST
o        Modified Method of Feasible Directions - ADS
o        Sequential Linear Programming - ADS
o        Sequential Quadratic Programming - DONLP
o        Sequential Quadratic Programming - NLPQL
o        Successive Approximation Method

        Penalty methods
o        Exterior penalty
o        Hooke-Jeeves Direct Search


Exploratory techniques avoid focusing only on a local region. They generally evaluate designs throughout the parameter space in search of the global optimum. The techniques included in iSIGHT are:
        Adaptive Simulated Annealing
        Multi-Island Genetic Algorithm

Expert system techniques follow user defined directions on what to change, how to change it, and when to change it. The technique including in iSIGHT is
        Directed Heuristic Search (DHS) .

iSIGHT also has a GUI, and can be connected to most of the FEA programs.



3.        COMSOL

Comsol is actually implemented in Matlab as a toolbox. Therefore, Comsol is able to output exactly what it having Matlab do as a series of commands in an m file (a Matlab script). Each major action (usually involving hitting ok and a statement at the bottom of the main window) performed in Comsol maps to a statement written in the m file, in the order they are performed. For example, if you hit the "Refine Mesh" button, it corresponds to a line in the m file.

The advantage of COMSOL is that it can naturally be connected into MATLAB, and all the MATLAB optimization resources can be used to do model-update with COMSOL.

yekai

iSIGHT的算法模块分为4大类：试验设计、优化方法、近似建模、质量工程。楼上写的只是iSIGHT功能的一部分。