Annotated List of Publications
1.
Schenk, J.R., Huang, D., Zheng, N., and Allen, T.T., “Multiple Fidelity Simulation
Optimization of Hospital Performance under High Consequence Event Scenarios”,
submitted to the 2005 Winter Simulation Conference
Proceeding.
In optimizing systems,
experimental models are often available with different levels of cost and
different levels of “fidelity” or trustworthiness, a fact that can be
exploited. For example, a highly detailed
model might be made for a few possible configurations, supplemented by a large
number of rough models that are less expensive to construct. The purpose of this paper is to illustrate
the application of a recently proposed Multiple Fidelity Sequential Kriging
Optimization (MFSKO) method to derive the optimal re-source allocation for
disaster preparedness of a hospital. The
system is evaluated via discrete event simulations of two sophistication
levels. The MFSKO method integrates
multiple fidelity data in search for the global optima with less total
evaluation cost. Kriging meta-models are
generated as by-products of the optimization.
2.
When cost-per-evaluation on a
system of interest is high, surrogate systems can provide cheaper but lower-fidelity
information. In the proposed extension
of the Sequential Kriging Optimization method, surrogate systems are exploited
to reduce the total evaluation cost. The
method utilizes data on all systems to build a kriging meta-model that provides
a global prediction of the objective function and a measure of prediction
uncertainty. The location and fidelity
level of the next evaluation are selected by maximizing an augmented expected
improvement function, which is connected with the evaluation costs. The proposed method was applied to test
functions from the literature and a metal-forming process design problem via
Finite Element simulations. The method
manifests sensible search patterns, robust performance, and appreciable
reduction in total evaluation cost as compared to the original method.
3.
Huang,
D., Allen, T. T., Notz, W. I., and Zheng, N., “Global Optimization of
Stochastic Black-Box Systems via Sequential Kriging Meta-Models”, accepted
by the Journal of Global Optimization.
This paper proposes
a new method that extends the Efficient Global Optimization (EGO) to address
stochastic black-box systems. The method is based on a kriging meta-model that
provides a global prediction of the objective values and a measure of
prediction uncertainty at every point. The criterion for the in-fill sample
selection is an augmented Expected Improvement (EI) function with desirable
properties for stochastic responses. The proposed method is empirically
compared with the Revised Simplex Search (RSS) and Simultaneous Perturbation
Stochastic Approximation (SPSA) methods using six test problems from the
literature. The results suggest that the proposed method is consistent and
relatively efficient in finding global optimal solutions.
4.
Huang,
D., Allen T. T., “Design
and Analysis of Variable Fidelity Experimentation Applied to Engine Valve Heat
Treatment Process Design”, Journal of
Royal Statistics, Series C. 54, Part 2, p. 443-463, (2005).
When experimentation
on the real system is expensive, data are often collected using cheaper,
lower-fidelity surrogate systems. This paper concerns response surface methods
in the context of variable fidelity experimentation. We propose the use of
Generalized Least Squares (GLS) to generate the predictions. We also present
perhaps the first optimal designs for variable fidelity experimentation, using
an extension of the Expected Integrated Mean Squared Error (EIMSE) criterion.
Numerical tests are used to compare the method performance with alternatives
and to investigate the robustness to incorporated assumptions. The method is
applied to automotive engine valve heat treatment process design in which real
world data were mixed with data from two types of computer simulations.
5.
Huang,
D., Drummond, C. H., III, Wang, J., Blume, R. D., “Incorporation of
Chromium (III & VI) Oxides in A Simulated Basaltic Industrial Waster
Glass-Ceramic”, Journal of American
Ceramic Society, Vol. 87, Issue 11,
p. 2047-2052 (2004).
Chromium, an EPA
listed toxic element concentrated in many industrial wastes, was stabilized
utilizing waste vitrification. Cr2O3 and CrO3 were loaded into a simulated basaltic base
composition, vitrified, and cooled at various rates. Chromium incorporation
mechanisms, vitrification processability, effect of initial Cr oxidation state,
and product performance were investigated. At 1500°C, Cr2O3 has a low solubility limit (0.54wt%) in the base
composition, and crystallized as Cr-rich primary spinel (Mg, Fe)(Fe, Al, Cr)2O4. Upon cooling,
Cr-depleted secondary spinel and augite (Na,Ca)(Mg,Fe2+,Al)(Si,Al)2O6 crystallized.
Cr(VI) was converted into Cr(III) upon vitrification. The apparent viscosity of
the melts was estimated using the Bottinga-Weill model as corrected by Roscoe’s
equation. The end products showed Cr2O3 loading capacities as high as 16.7wt% without
exceeding the toxicity-leaching limit defined by EPA. The annealed products had
Vicker’s hardness of about 800 KgF/mm2 and can be classified as medium grade abrasives.
6.
Chantarat,
N., N. Zheng, T. T. Allen, and Huang, D., “Optimal
Experimental Design for Systems Involving Both Quantitative and Qualitative
Factors,” Proceedings of the Winter
Simulation Conference, R. D. M. Ferrin and P. Sanchez (www.wintersim.org),
(2003).
Often in discrete-event simulation, factors being
considered are qualitative such as machine type, production method, job release
policy, and factory layout type. It is also often of interest to create a
Response Surface (RS) metamodel for visualization of input-output
relationships. Several methods have been
proposed in the literature for RS metamodeling with qualitative factors but the
resulting metamodels may be expected to predict poorly because of sensitivity
to misspecification or bias. This paper proposes the use of the Expected
Integrated Mean Squared Error (EIMSE) criterion to construct alternative
optimal experimental designs. This approach explicitly takes bias into account.
We use a discrete-event simulation example from the literature, coded in ARENATM,
to illustrate the proposed method and to compare metamodeling accuracy of
alternative approaches computationally.
7.
Huang,
D., Wu, W. T., Lambert, D., and Semiatin, S. L., "Computer Simulation of
Microstructure Evolution During Hot Forging of Waspaloy and Nickel Alloy 718",
Microstructure Modeling and Prediction
During Thermalmechanical Processing, TMS, p.137-147, (2001).
Computer simulation
of microstructure evolution during hot forging of superalloys is of great
interest, particularly for the manufacture of critical components for aerospace
applications. To this end, recrystallization and grain growth were modeled
using a phenomenological approach and implemented in the commercial metal
forming code DEFORMTM that establishes the thermomechanical
history. Necessary material parameters
used in the analysis were collected from available literature. To validate the
model with Waspaloy, an industrial disk forging process was simulated.
Predictions of recrystallized volume fraction and grain size were in good
agreement with the measurements. For nickel alloy 718, simple upsetting and
double-cone forging were simulated and compared with experimental data. A
three-dimensional cogging simulation was also performed for as-cast 718 ingot
material to demonstrate the capability of the model.
8.
Huang,
D., LaCount, B. J., Castro, J. M., Ignatz-Hoover, F., “Development of a
service-simulating aging test method for exterior tire rubber compounds part I:
Cyclic Aging”, Polymer Degradation and
Stability, 74, p. 353, (2001).
Two conventional standard aging tests – thermal aging
and dynamic ozone aging and a novel multiple-factor test – cyclic aging, are
compared to an outdoor aging test, which is assumed close to real service
aging. The cyclic aging test consists of
four single-factor sub-tests, including oxygen bomb, dynamic ozone, ultraviolet
light, and water solution attack. Four
formulations of tire rubber compound are used as experimental materials. The degradation of material properties as a
function of aging is measured in order to compare the aging test methods. The experimental results show that the cyclic
aging test is closer to the outdoor aging than the conventional dynamic ozone
and thermal aging tests. The static
modulus increases with thermal aging, and decreases with dynamic ozone aging
time. For outdoor and cyclic aging, the
modulus initially increases, but then later decreases, showing traces of both
of the previously mentioned results. It
is also found that the size of the specimen has a significant acceleration
affect on aging. Such effect can be
utilized to accelerate the artificial aging without elevating the temperature.
9.
Huang,
D., Arimoto, K., Lee, K., Lambert, D., Narazaki, M., "Prediction of Quench Distortion on
Steel Shaft with Keyway by Computer Simulation”, Heat Treating Conference & Exposition Proceeding, ASM, (2000).
During quenching, steel shafts tend to distort much
more if they contain a keyway. Non-uniform cooling due to the non-axisymmetric
shaft geometry may induce the additional distortion. Distortion measurements
were obtained from a number of shaft steels and quenchants to compare alloy and
cooling effects. The experiments were simulated using the Finite Element Method
(FEM) system DEFORM™-HT, with estimated surface heat transfer coefficients.
10. Arimoto, K.,
Huang, D., Lambert, D., and Wu, W. T., "Computer Prediction and Evaluation of
Inverse Quench-Hardening of Steel", Heat Treating Conference & Exposition Proceeding, ASM, (2000).
In oil quenched bearing grade steels, hardness
distributions have occasionally exhibited lower hardness values at the surface
that at the core. This phenomenon is
referred to as “inverse quench-hardening”.
The Finite Element Method (FEM) system DEFORMTM-HT was
utilized to predict the volume fractions of metallic phases resulting form the
inverse quench-hardening phenomenon. The
simulated results have clarified the reasons for why the phenomenon occurs.
Proposals
Huang, D., Allen, T. T., Notz, W. I.,
2004 NSF Proposal “Efficient Global Optimization Using Variable Fidelity Data”.
The needs of U.S. manufacturing for optimization
continue to grow in the face of increasing global competition. Shop floor engineers are demanding
optimization tools combined with simulation to increase quality and reduce
production costs. This type of
optimization problem often has expensive function evaluations with noise. The efficient global optimization (EGO)
method based on the Design and Analysis of Computer Experiment (DACE)
metamodeling framework has been shown to be promising from an efficiency
standpoint for the associated nonlinear stochastic optimization problems, i.e.,
using only a limited number of objective function evaluations or “data,” the
EGO methods consistently found the global minima in test problems.
“Fidelity”
refers to the degree of similarity between outputs of the system of interest
and of surrogate systems that are trying to emulate the system of
interest. Data of varying levels of
fidelity are often available in practice, e.g., production data, data from
laboratory tests, and computer simulations of various types all might be trying
to emulate shop floor performance.
Because lower fidelity function evaluations may be orders of magnitude
less expensive than highest fidelity calls (running new parts on the shop floor),
solvers that utilize variable fidelity data promise dramatic increases in
efficiency. In this context, adapting
the DACE/EGO framework has an obvious advantage compared with alternative
optimization methods that do not involve building global meta-models. This follows because meta-modeling permits
the solver to explicitly address the systematic errors associated with the data
from the lower fidelity systems. A major
subtask of the research is to explore design and analysis of variable fidelity
experiments for accurate prediction of the highest fidelity surface.