Abstracts Submitted from
Computer & Infromation Science, Electrical Engineering, Etc.

Undergraduate Summer Research Symposium August 12, 2009

Ordered alphabetically by student's last name

Davis, C Loughery Shall
Barber Enslen
Malik Varnell
Bose Habimana-Griffin Maxey White
Cobb Hopkins Pedersen

Investigating Data Models for Automatically Generating Tests  for Web Applications

Katie Baldwin1, Camille Cobb2, Carrie Hopkins2, Sara Sprenkle2, and Lori Pollock1
1University of Delaware, 2Washington and Lee University

Web applications must be dependable as the number and popularity of web applications increases, and people become more dependent on them. Web applications are difficult and expensive to test because of the large input space and frequent changes.  Thus, their characteristics demand an efficient and effective way of automating the test case generation process.  Current approaches to automatic test case generation for web applications do not attain all the goals of representing user behavior, maintaining good code coverage, and reducing the number of test cases.  This research is based on Sant et al.’s user-session-based test case generation approach, which applies statistical language learning algorithms to create control and data models, where a control model represents the possible URL sequences and the data model represents the possible parameter values.  Through analyzing user sessions, we identify factors that impact values in user sessions, and use these results to develop a set of data models for automatic test case generation.  This research is sponsored by the CRA-W.

Hybrid polymer-peptide hydrogels for vocal fold tissue engineering

Kathryn I. Barber1, Sarah E. Grieshaber2, and Xinqiao Jia2
1Department of Materials Science and Engineering, Pennsylvania State University and
2Department of Engineering Science and Mechanics,  University of Delaware

One of the challenges in the growing field of tissue engineering is the method of successfully imitating the extracellular matrix (ECM) in which the cells grow and live. The development of tunable scaffolds is needed to mimic the morphological, biological, and mechanical properties of the natural ECM for vocal fold tissue regeneration. The dual nature of hybrid natural-synthetic polymer hydrogels holds great potential for these applications  because the peptide segments allow control over structural assembly and biological properties, while the synthetic polymer domains provide greater tunability and improved mechanical properties.  RGD-containing elastin-mimetic peptides were synthesized and crosslinked with vinyl sulfone-functionalized Pluronic F-127 to serve as scaffolds for vocal fold fibroblasts (VFFs).

Synthesis and Characterization of Block Copolymer Nanoparticles
for Controlled Release of Cancer
Aditya Bose, Xiaoying Wang, and Xinqiao Jia
Department of Materials Science and Engineering

The focus of this research is to synthesize block copolymer nanoparticles that can be utilized as delivery vehicles for anticancer drugs. Amphiphilic block copolyesters carrying pendant cyclic ketal or ketone groups were synthesized by ring opening copolymerization of ε-caprolactone (CL) and 2-Oxepane-1,5-dione (OPD) using methoxy poly(ethylene glycol) (mPEG) as the initiator and stannous octoate (Sn(Oct)2) as the catalyst. Ureido-4-pyrimidones (UPY) was subsequently conjugated to the copolymers through the pendant ketone groups in the hydrophobic block. The polymer composition and molecular weight were characterized using proton nuclear magnetic resonance (1H NMR) and gel permeation chromatography (GPC), respectively.  Dexamethasone was encapsulated in the block copolymer nanoparticles by a nanoprecipitation procedure. Particle size and size distribution were analyzed by dynamic light scattering (DLS) and the drug encapsulation and release were monitored using high performance liquid chromatography (HPLC). Our preliminary results show that the custom-designed block copolymers exhibit relatively high drug encapsulation efficiency and tunable release kinetics.  We are evaluating the potentials of these nanoparticles in the treatment of childhood leukemia.

Adsorption and Reduction of 2,4-dinitroanisole (DNAN) using Graphitic Black Carbon

Craig Warren Davis and Pei C. Chiu

Nitro-aromatic compounds (NACs) such as DNAN (2,4-dinitroanisole) are common explosives and environmental contaminants, particularly in soil and groundwater systems.  Reduction of NACs in aqueous media is generally slow, due to the high activation energy involved.  However, recent work has shown that reduction of NACs can be greatly accelerated in the presence of graphitic black carbon (i.e., carbon left behind from combustion process that has undergone structural changes due to the high heat and pressure involved in combustion).  This study was undertaken to investigate the reduction of DNAN catalyzed by graphite (a model black carbon) in the presence of H2S (a common reductant in anaerobic environments).  We hypothesize that graphite serves a dual role in this process. First, graphite (surface area ≈ 200 m2/g) may act as a strong adsorbent and rapidly removes DNAN molecules from the aqueous phase. Secondly, graphite is conductive and hence may facilitate electron transfer from H2S to DNAN, effectively acting as a redox catalyst.  To date, we have evaluated the sorptive capability of graphite in artificial groundwater.  Reactors were set up and aqueous samples were taken at predetermined times.  Samples were analyzed using a high performance liquid chromatograph (HPLC) and quantified based on DNAN calibration standards. Next, we plan to evaluate the extent and rates of DNAN adsorption and reduction with graphite under reducing conditions.  The ultimate goals of this work are to understand the reaction pathway and mechanism and to develop kinetic and equilibrium models to describe the reduction and adsorption processes.

Mining Source Code to Automatically Split Identifiers for Software Analysis

Eric Enslen, Emily Hill, Lori Pollock, and K. Vijay-Shanker
Department of Computer and Information Science

Automated software analysis tools and recommendation systems increasingly rely on natural language information from comments and identifiers in code. The first step in analyzing words from identifiers requires splitting identifiers into their constituent words. Unlike natural languages, where space and punctuation are used to delineate words, identifiers cannot contain spaces. One common way to split identifiers is to follow programming language naming conventions. For example, Java programmers often use camel case, where words are delineated by uppercase letters or non-alphabetic characters. However, programmers also create identifiers by concatenating sequences of words together with no discernible delineation, which pose challenges to automatic identifier splitting. /  / In this paper, we present an algorithm to automatically split identifiers into sequences of words by mining the frequency of potential substrings from source code. With these word frequencies, our identifier splitter uses a scoring technique to automatically select the most appropriate partitioning for an identifier. In an evaluation of over 8000 identifiers from open source Java programs, our Samurai approach outperforms the existing state of the art techniques.

Preparation of HA-Pln DIV Conjugates Hydrogel

LeMoyne Habimana-Griffin1, Chao Liu2, Xinqiao Jia2
1Rose-Hulman Institute of Technology, 2Department of Materials Science and Engineering, University of Delaware

Photo-cross-linkable, mechano-responsive hydrogels have the potential to serve as a minimally invasive mean to regenerate vocal fold tissue in vivo. In particular, glycidyl methacrylate (GMA) functionalized hyaluronic acid(HA) –based hydrogels have shown great potential as vocal fold tissue scaffolds.  However, without infrastructure within the gel to support cell adhesion, the cells with merely fall of the gel. But the incorporation of an immunoglobulin sequence in perlecan domain IV peptide, which has recently been found to support cell adherence, will allow the cells to stick to the gel.  Nevertheless, highly toxic catalysts used in the synthesis of the gel, such as dimethylaminopyridine (DMAP), can interact with the HA-GMA and become trapped in the polymer during precipitation.  We present a novel method to purify the HA-GMA hydrogel.  Using this method we were able to eliminate most all of the DMAP, as confirmed by nuclear magnetic resonance spectroscopy.  This project was funded by the National Science Foundation.

Extraction of Lipids from Wastewater Microorganisms for Biodiesel Fuel Production

Scott R. Loughery, Nathan S. Kiracofe, and Daniel K. Cha
Department of Civil and Environmental Engineering

Natural fatty acids from wastewater activated sludge and Nocardia Amarae foam were extracted and analyzed for the purpose of converting these lipids into biodiesel fuel.  Special attention was paid to the Nocardia bacterial foam.  Because the Nocardia species has a filamentous cell structure, a higher oil yield was expected.  Lipids from microorganisms can be converted into fuel sources by a process known as transesterification.  This process involves converting triglyceride molecules found in cellular lipids into fatty acid methyl esters (FAME).  The FAME can be burned as a diesel fuel source.  The conversion of a triglyceride molecule into FAME is triggered by the addition of an organic carbon source at a moderately high temperature.  A hexane/methanol solvent was used as the organic carbon source.  Activated sludge and Nocardia foam were retrieved from several different aeration basins from the Wilmington Wastewater Treatment Plant, where they were then oven-dried to remove moisture and stored in preparation for extraction.  The FAME process involved transesterifying the samples with methanol/sulfuric acid, extracting the samples with hexane, separation of the biomass and organic solvent layers by centrifuge, and removal and evaporation of the organic solvent.  The composition of the samples after undergoing this process was a dark oil.  The specific oil yield of our samples were typically between 10 and 20 percent by mass.  These yields compare favorably to other common biodiesel sources such as soybean oil, peanut oil, and algae and help to show that wastewater bacteria can be a viable source of diesel fuel.  Research funded by the Science and Engineering Scholars Program and the University of Delaware Undergraduate Research Program.

Applying a Software Word Usage Model (SWUM) to Other Languages

Sana Malik, Emily Hill, Lori Pollock, and K. Vijay-Shanker
Computer and Information Sciences

With program code growing larger and larger, software developers need more support, especially for software maintenance. Currently, there are many automatic and semi-automatic tools meant to expedite software maintenance; however, most of these tools rely solely on the structural model of the program, while disregarding any semantic information from the natural language used by programmer. Our previous work towards solving this problem involves a general Software Word Usage Model (SWUM). Unlike other software maintenance tools that use lexical concepts (the function or meaning of a word), SWUM looks for and applies linguistic relations between these lexical concepts to form a more complete interpretation of the program.   Although SWUM is capable of representing all programming languages, its current implementation is limited to Java. The potential structural, semantic, and syntactic differences with other languages must be examined to generalize SWUM beyond a single language. In this paper we analyze the differences between Java and C++, modify the SWUM construction algorithm to work for C++, and evaluate the accuracy of the phrases it generates for C++.   This project is sponsored by the CRA-W DREU program.

Detection of Structural Damage using Thermal Imaging

Joseph Maxey,  Mary Marchegiano, Robert HunspergerMichael Chajes, and Erik G. Kunz
Departments of Electrical & Computer Engineering  and Civil & Environmental Engineering

Major problems in bridge collapses are due to flaws in its structure. NASA’s space exploration and aeronautics also can have difficulties due to material flaws. This paper presents a method for detecting hidden flaws in structures and materials by thermal imaging. The idea is to use a thermal imaging camera to detect flaws in some samples of materials, including steel strands, and a composite piece made of e-glass with epoxy resin. A crack was put into the steel strand and a cut was put into the composite piece. Natural corrosion was present on another steel strand. The flawed regions should heat faster because their resistance is more than that of the non-flawed region, which follows from a Law of Heat Transfer. Electrical current was used in the steel strands to heat them and to detect the cut on one strand, and the corrosion on the other strand, while thermal induction was used to heat the composite piece. Results from the thermal camera did indeed show the flaws having a higher temperature than their surroundings. This may means that structure damage can be detected by thermal imaging early enough to prevent catastrophes. The NASA Space Program sponsored this work.

Implementation of a Correlation Algorithm on the Cyclops-64 Architecture

Jeremy Pedersen, Juergen Ributzka, Guang R. Gao
Computer Architecture and Parallel Systems Laboratory (CAPSL)

There has been a shift in radio astronomy away from the use of large dishes and towards the use of  arrays of smaller antennas. This requires that large amounts of sample data be cross correlated to generate useful signals. Cross correlation is an I/O intensive operation at 1 FLOP per byte loaded from memory. Since large antenna arrays like SKA and LOFAR will generate massive amounts of data (in the case of LOFAR it is 100 TB per day), an efficient means of processing this data in real-time is required. Computer architectures with large amounts of on-chip memory and enough parallelism to truly take advantage of the embarrassingly parallel nature of the correlation algorithm are needed. Cyclops-64, with its 5 MB of on chip SRAM and 160 Thread Units (TU) is a great candidate for this task. Because Cyclops-64's on-chip SRAM is not a cache, but is user managed, clever use of this space can overlap memory loads and stores with computation, and increase the number of arithmetic operations performed per byte loaded from memory. We look first at a naíve implementation of the correlation algorithm, and then at a more sophisticated implementation in which a ring buffer is set up in SRAM for storing sample data to be correlated, allowing correlations to be performed by some threads while others do fetches from main memory into on-chip memory.

Towards Generating Useful Comments for Program Code

Jonathan Schall, Giriprasad Sridhara, Emily Gibson Hill, Lori Pollock, K. Vijay Shanker
Department of Computer and Information Science

With the use of 60-90% of software cycle resources being used on program maintenance, there is a call for automated software tools that help software engineers explore and comprehend today’s large and complex computer programs.  Modern software tools often ignore the lexical information contained in comments, identifiers, and program structure.  My work involves identifying the useful information in computer programs (both structure and naming conventions) that can be used to automatically generate useful English-like comments.  To this end, I analyzed program code information for appropriateness for generating comments. I compared this code information with real programmer-written comments and identified the program structure information that helps to correct discrepancies between the programmer-written comment and the automatically extracted phrases.   I documented the relevant program code information, natural language comment, along with the parts of the program structure I found useful.  My research suggests that the assignment operator, looping structures, and the naming of classes and methods are useful in automatically generating useful natural language comments for software maintainers.

Fiber Optic Sensors

Deborah M. Varnell, Robert G. Hunsperger, Eric Kunz, and Michael J. Chajes

Fiber optic cables are used to create simple sensors that can detect problems in hidden places like inside bridge supports or landfills.  There are several reasons why fiber optic sensors are different than other sensors.  They are less expensive than other systems because the materials they are made from (glass or plastic) are readily available.  Also, fiber optics are not effected by electronic interference (like lightning). To make the sensor that we have investigated, the fiber optic cable is shaped into several tight loops. If the loops are tight enough, some of the light traveling through the fiber will escape.  The loops must be tight enough to let light out so that it can sense the surroundings, but not too tight or almost all of the light will escape.  The amount of light that escapes from the loops depends on what is  around the fiber, which allows the sensor to detect the amount of water or chloride present.  This information would be used to determine if the correct amount of water is in a landfill or how much water and chloride are present inside a bridge support.  The presence of water or chloride would indicate that there is a danger of corrosion.   Funded by the Undergraduate Research Program at University of Delaware

Synthesis and Characterization of New Ceramic Oxygen Conductors

James L. White and Joshua L. Hertz

In recent years, cost-effective and environmentally friendly sources of energy have become increasingly important for research and development.  Solid oxide fuel cells (SOFCs), which release energy from a fuel using oxygen as the oxidant, have great potential for efficiency and minimal carbon emissions.  Current SOFCs utilize ceramic electrolytes, such as yttria-stabilized zirconia (YSZ) and gadolinium-doped ceria (GDC), that conduct oxygen at high temperatures (over 600°C) through vacancies in their crystal structures.  Solid-state electrolytes that conduct oxygen as crystalline interstitials instead of vacancies were synthesized and characterized in an effort to produce improved ceramics for use in SOFCs.  The K2NiF4 structure, known for its high concentration of interstitial defects, is the target configuration for the new ceramics.  X-ray powder diffractometry has revealed moderate success in obtaining the structure for several of the different ceramics. 

Links: Summer 2009 Undergraduate Research Symposium, Symposium Abstracts from other Colleges and Departments,
2009 Undergraduate Research Summer Enrichment ProgramUnversity of Delaware Undergraduate Research Program, Howard Hughes Undergraduate Program.
Created  7 August 2009. Last up dated 30 August 2009 by Hal White
Copyright 2009, University of Delaware