European Business School, Diss.--Oestrich-Winkel, 2006.

This thesis investigates several issues in data and computation modeling for scientific problem solving environments (PSEs). A PSE is viewed as a software system that provides (i) a library of simulation components, (ii) experiment management, (iii) reasoning about simulations and data, and (iv) problem solving abstractions. Three specific ideas, in functionalities (ii)-(iv), form the contributions of this thesis. These include the EMDAG system for experiment management, the BSML markup language for data interchange, and the use of data mining for conducting non-trivial parameter studies. This work emphasizes data modeling and management, two important aspects that have been largely neglected in modern PSE research. All studies are performed in the context of S4W, a sophisticated PSE for wireless system design. / Master of Science

This thesis describes a computing environment WBCSim and its experiment management component. WBCSim is a web-based simulation system used to increase the productivity of wood scientists conducting research on wood-based composite and material manufacturing processes. This experiment management component integrates a web-based graphical front end, server scripts, and a database management system to allow scientists to easily save, retrieve, and perform customized operations on experimental data. A detailed description of the system architecture and the experiment management component is presented, along with a typical scenario of usage. / Master of Science

A problem solving environment (PSE) is a computational system that provides a complete and convenient set of high level tools for solving problems from a specific domain. This thesis takes an in-depth look at the experiment management aspect of PSEs, which can be divided into three levels: 1) data management, 2) change management, and 3) execution management. At the data management level, anything related to an experiment (computer simulation) should be stored and documented. A database management system can be used to store the simulation runs for a PSE. Then various high level interfaces can be provided to allow users to save, retrieve, search, and compare these simulation runs. At the change management level, a scientist should only focus on how to solve a problem in the experiment domain. Aside from running experiments, a scientist may only consider how to define a new model, how to modify an existing model, and how to interpret an experiment result. By using XML to describe a simulation model and unify various implementation layers, changing an existing model in a PSE can be intuitive and fast. At the execution management level, how an experiment is executed is the main concern. By providing a computational steering capability, a scientist can pause, examine, and compare the intermediate results from a simulation. Contrasted with the traditional way of running a lengthy simulation to see the result at the end, computational steering can leverage the user's expert knowledge on the fly (during the simulation run) and provide new insights and new product design opportunities. This thesis illustrates these concepts and implementation by using WBCSim as an example. WBCSim is a PSE that increases the productivity of wood scientists conducting research on wood-based composite materials and manufacturing processes. It integrates Fortran 90 simulation codes with a Web based graphical front end, an optimization tool, and various visualization tools. The WBCSim project was begun in 1997 with support from United States Department of Agriculture, Department of Energy, and Virginia Tech. It has since been used by students in several wood science classes, by graduate students and faculty, and by researchers at several forest products companies. WBCSim also serves as a test bed for the design, construction, and evaluation of useful, production quality PSEs. / Ph. D.

Large scale, multidisciplinary, engineering designs are always difficult due to the complexity and dimensionality of these problems. Direct coupling between the analysis codes and the optimization routines can be prohibitively time consuming. One way of tackling this problem is by constructing computationally cheap(er) approximations of the expensive simulations, that mimic the behavior of the simulation model as closely as possible. This paper presents a data driven, surrogate based optimization algorithm that uses a trust region based sequential approximate optimization (SAO) framework and a statistical sampling approach based on design of experiment (DOE) arrays. The algorithm is implemented using techniques from the two packages SURFPACK and SHEPPACK that provide a collection of approximation algorithms to build the surrogates and three different DOE techniques: full factorial (FF), Latin hypercube sampling (LHS), and central composite design (CCD) are used to train the surrogates. The biggest concern in using the proposed methodology is the generation of the required database. This thesis proposes a data driven approach where an expensive simulation run is required if and only if a nearby data point does not exist in the cumulatively growing database. Over time the database matures and is enriched as more and more optimizations are performed. Results show that the response surface approximations constructed using design of experiments can be effectively managed by a SAO framework based on a trust region strategy. An interesting result is the significant reduction in the number of simulations for the subsequent runs of the optimization algorithm with a cumulatively growing simulation database. / Master of Science

Experimentieren in der vorliegenden Arbeit bedeutet, Experimente auf der Basis von computerbasierten Modellen durchzuführen, wobei diese Modelle Struktur, Verhalten und Umgebung eines Systems abstrahiert beschreiben. Aus verschiedenen Gründen untersucht man stellvertretend für das System ein Modell dieses Systems. Systematisches Experimentieren bei Variation der Modelleingabeparameterbelegung führt in der Regel zu sehr vielen, potentiell lang andauernden Experimenten, die geplant, dokumentiert, automatisiert ausgeführt, überwacht und ausgewertet werden müssen. Häufig besteht dabei das Problem, dass dem Experimentator (der üblicherweise kein Informatiker ist) adäquate Ausdrucksmittel fehlen, um seine Experimentier-Prozesse formal zu beschreiben, so dass sie von einem Computersystem automatisiert ausgeführt werden können. Dabei müssen Verständlichkeit, Nachnutzbarkeit und Reproduzierbarkeit gewahrt werden. Der neue Ansatz besteht darin, generelle Experimentier-Workflow-Konzepte als Spezialisierung von Scientific-Workflows zu identifizieren und diese als eine metamodellbasierte Domain-Specific-Language (DSL) zu formalisieren, die hier als Experimentation-Language (ExpL) bezeichnet wird. ExpL beinhaltet allgemeine Workflow-Konzepte und erlaubt das Modellieren von Experimentier-Workflows auf einer frameworkunabhängigen, konzeptuellen Ebene. Dadurch werden die Nachnutzbarkeit und das Publizieren von Experimentier-Workflows nicht mehr durch die Gebundenheit an ein spezielles Framework behindert. ExpL wird immer in einer konkreten Experimentierdomäne benutzt, die spezifische Anforderungen an Konfigurations- und Auswertemethoden aufweist. Um mit dieser Domänenspezifik umzugehen, wird in dieser Arbeit gezeigt, diese beiden Aspekte separat in zwei weiteren, abhängigen Domain-Specific-Languages (DSLs) zu behandeln: für Konfiguration und Auswertung. / Experimentation in my work means performing experiments based on computer-based models, which describe system structure and behaviour abstractly. Instead of the system itself models of the system will be explored due to several reasons. Systematic experimentation using model input parameter variation assignments leads to lots of possibly long-running experiments that must be planned, documented, automated executed, monitored and evaluated. The problem is, that experimenters (who are usually not computer scientists) miss the proper means of expressions (e. g., to express variations of parameter assignments) to describe experimentation processes formally in a way, that allows their automatic execution by a computer system while preserving reproducibility, re-usability and comprehension. My approach is to identify general experimentation workflow concepts as a specialization of a scientific workflow and formalize them as a meta-model-based domain-specific language (DSL) that I call experimentation language (ExpL). experimentation language (ExpL) includes general workflow concepts like control flow and the composition of activities, and some new declarative language elements. It allows modeling of experimentation workflows on a framework-independent, conceptional level. Hence, re-using and sharing the experimentation workflow with other scientists is not limited to a particular framework anymore. ExpL is always being used in a specific experimentation domain that has certain specifics in configuration and evaluation methods. Addressing this, I propose to separate the concerns and use two other, dependent domain-specific languages (DSLs) additionally for configuration and evaluation.