Learning to set up numerical optimizations of engineering designs

Abstract

Gradient-based numerical optimization of complex engineering designs offers the promise of rapidly producing better designs. However, such methods generally assume that the objective function and constraint functions are continuous, smooth, and defined everywhere. Unfortunately, realistic simulators tend to violate these assumptions, making optimization unreliable. Several decisions that need to be made in setting up an optimization, such as the choice of a starting prototype and the choice of a formulation of the search space, can make a difference in the reliability of the optimization. Machine learning can improve gradient-based methods by making these choices based on the results of previous optimizations. This paper demonstrates this idea by using machine learning for four parts of the optimization setup problem: selecting a starting prototype from a database of prototypes, synthesizing a new starting prototype, predicting which design goals are achievable, and selecting a formulation of the search space. We use standard tree-induction algorithms (C4.5 and CART). We present results in two realistic engineering domains: racing yachts and supersonic aircraft. Our experimental results show that using inductive learning to make setup decisions improves both the speed and the reliability of design optimization.

Adaptive evolutionary Monte Carlo algorithm for optimization with applications to sensor placement problems

“…Various statistical predictive methods have been used as the metamodels, according to the survey by Chen et al (2006), including neural networks, tree-based methods, Splines, and spatial correlation models. During the past few years, there have emerged a number of research developments, labeled as data-mining guided engineering designs (Guikema et al 2004;Huyet 2006;Liu and Igusa 2007;Kim and Ding 2005;Michalski 2000;Schwabacher et al 2001). The data-mining guided methods are basically one form of metamodel-based methods because they also use a statistical predictive model to guide the selection of design solutions.…”

“…The metamodel-based method originates from the research on computer experiments (Chen et al 2006;Fang et al 2006;Sacks et al 1989;Simpson et al 1997). This strategy is also called "data-mining" guided method, especially when the predictive model used therein is a classification tree model (Liu and Igusa 2007;Kim and Ding 2005;Schwabacher et al 2001) since the tree model is a typical "data-mining" tool. For the metamodel-based or data-mining guided methods, the major shortcoming is their ineffectiveness in handling complicated response surfaces, and as a result, they only look for local optima.…”

Tiltrotor CFD Part II - aerodynamic optimisation of tiltrotor blades

“…n−m (12) wherer is the nondimensional coordinate along the blade span, which has value 0 at the rotation axis, and 1 at the blade tip. The symbol K m,n denotes is the binomial coefficient, which is defined as…”

An introduction to swarming robotics: application development trends

“…Optimization and networks design applications are very broad, traversing from a specific area and ranging from engineering design (Schwabacher et al 1998;Coelho and Mariani 2008;Kashan 2011), process optimization (Egea et al 2010;Joshi and Pande 2011;Kwak and Kim 2012), scheduling system (Andersson et al 2007;Frantzénl et al 2011;Skobelev 2011), routing and flow control in networks and networking (Madan et al 2007;Shakkottai and Srikant 2007;Minoux 2010), to service oriented applications in finance (He et al 2008;Leibfritz and Maruhn 2009;Pennanen 2011), healthcare (Harrell and Lange 2001;Bagirov and Churilov 2003;José et al 2011), and bioinformatics (Hernandez and Kambhampati 2004;Nebro et al 2008;Arredondo et al 2011). For example, in the formulation of the scheduling system, optimization can be used to determine the course of vehicle systems to the various destinations, to determine the scheduling of jobs in the factory, scheduling lectures at universities, creating timetable, computer network design and planning strategies for finding an optimal decision.…”