Title: Power/Efficiency Optimization of a Sorption Cooler Under Quantified Design Uncertainty
Abstract: While the design paradigm in engineering of searching for the optimum system has proven fruitful (and given a good model relatively straightforward, in principle), the desired end result of engineering development is rarely a model (even the optimum one), but a system. In this regard it has frequently been observed (generally with some disappointment) that what one can specify is not always what one gets. It is frequently the case that realized systems, no matter how carefully constructed according to specifications derived from verified and validated models, frequently depart from the designed-for behaviour, due to parametric incertitude. Given this not uncommon circumstance, a somewhat more useful question one might seek to answer during an optimization process is “what is the best system under the constraints which I can reasonably hope to build?” Design optimization under incertitude approaches based on intrusive modifications to the deterministic model, such as stochastic finite elements and chaos expansions, are tedious to apply, computationally expensive, and fraught with convergence issues. The simplest nonintrusive approach—direct Monte Carlo sampling— is far too slow to efficiently sample the joint response distribution of complex thermophysics transient models. The purpose of this paper is to address this topic by incorporating design uncertainty itself as a constraint during the optimization of a sorption cooler. In our method a Markov Chain Monte Carlo sampler is used as the means to develop a suitable ensemble from a practical set of computational results which circumscribe the power/efficiency characteristics of a cooler as a function of several dimensionless stochastic optimization parameters. The ensemble is used to estimate the covariance structure of the design uncertainty, which is then projected into the best low rank subspace where tests of hypothesis under the dominant generalized parameters can be formulated; growth in fluctuations of the generalized parameters along optimization trajectories becomes clearly evident and quantifiable. The method results in a classical power/efficiency diagram, with the addition of quantified design uncertainty. The utility of these diagrams is that they enable rapid-prototyping efforts to target the best cooler design that is most likely to function as expected.
Publication Year: 2007
Publication Date: 2007-01-01
Language: en
Type: article
Indexed In: ['crossref']
Access and Citation
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot