PDF Publication Title:
Text from PDF Page: 183
αf ∇fR sCf= k τf= f 7.5 Hybrid Filter Trust-region Algorithm in this section, we rely on the gradients of the objective and the constraints obtained from the reduced-order model to promote such descent. Formally, we evaluate Cauchy steps (steepest descent direction) for both objective function and infeasibility as below 0 if sC = 0 ∥∇fR∥ k fR(x )−fR(x +sC) kkkkf otherwise ifsCθ =0 otherwise (7.41) (7.42) αf ∥∇fR∥ k 0 sC= k τ= R R C αθ∇θR θ θ θ ( x ) − θ ( x + s ) ∥∇θR∥ kkkkθ k αθ ∥∇θR ∥ k for some αf , αθ ∈ (0, 1). Here fR is constructed with ZOC in equation (7.8), and θR is given kk by equation (7.35) with cR(x), and cR(x) given by equation (7.8). We note that since fR and EIk θR are based on the ROM for kth iteration, their gradients, and thus sC and sC can be cheaply kfθ evaluated for each trust-region iteration. Hence, with a little computational expense, we can determine if the reduced-order model can predict descent for fR or θR or both. If τf > 0 or τθ > 0, it can guarantee descent for the reduced objective function fR or reduced infeasibility k θR, respectively. Therefore, the normal subproblem with ZOC is solved only if τθ > 0, and k similarly, the tangential subproblem with ZOC is solved only when τf > 0. If both cannot be ensured, algorithm moves to Section II where FOC with exact gradients is used. In Section I, we also incorporate POD subspace augmentation which involves adding more POD basis functions to improve accuracy of the existing ROM for the kth iteration, and thus enhancing its ability to accurately predict the descent direction for fR or θR or both. To kk construct a reduced-order model, we choose the number of basis functions M by deciding on an error tolerance level λ∗. During the course of the algorithm, if we encounter a situation when both τf ≤ 0 and τθ ≤ 0, tolerance level λ∗ is reduced to increase the POD subspace dimension M and more basis functions are added to the ROM. This is repeated until either one of the τf or τθ or both become positive, or we hit the maximum allowable limit for POD subspace dimension Mmax. Once we reach Mmax, algorithm switches to Section II. kk Chapter 7. Trust-region Framework for ROM-based Optimization 169PDF Image | Design and Operation of Pressure Swing Adsorption Processes
PDF Search Title:
Design and Operation of Pressure Swing Adsorption ProcessesOriginal File Name Searched:
anshul_thesis.pdfDIY PDF Search: Google It | Yahoo | Bing
CO2 Organic Rankine Cycle Experimenter Platform The supercritical CO2 phase change system is both a heat pump and organic rankine cycle which can be used for those purposes and as a supercritical extractor for advanced subcritical and supercritical extraction technology. Uses include producing nanoparticles, precious metal CO2 extraction, lithium battery recycling, and other applications... More Info
Heat Pumps CO2 ORC Heat Pump System Platform More Info
CONTACT TEL: 608-238-6001 Email: greg@infinityturbine.com (Standard Web Page)