Engineering 2017

Engineering Index Poster #:

A paper-based lateral flow strip for detection of traumatic brain injury protein biomarker | Devan J. Shell, Xuefei Gao2 and Nianquiang (Nick) Wu

Department of Chemistry, Catawba College, Salisbury, NC 28144 and Department of Mechanical and Aerospace Engineering, West Virginia University, Morgantown, WV 26506

Protein biomarkers are becoming increasingly important in the diagnosis of various diseases such as traumatic brain injury (TBI). Conventional methods, such as enzyme-linked immunosorbent assay (ELISA), Western blot, and radio-immunoassay offer high accuracy and sensitivity for protein measurements. However, most of these methods require time consuming procedures and high cost, which is not suitable for efficient point-of-care use. In our study, we developed paper-based lateral flow test strips (LFTS) for protein biomarker detection in blood plasma samples. Both colorimetry and Surface-Enhanced Raman Spectroscopy (SERS) methods were integrated as signal transducers into PLFS. Detection sensitivity, anti-interference ability, and efficient point-of-care use were taken into careful consideration. The results indicate that colorimetric LFTS have advantages for protein biomarker such as timely analysis (within 20 min.), cost-effectiveness, and efficient point-of-care use. However, disadvantages include low detection sensitivity and signal interference from blood plasma sample matrices. Compared with colorimetric LFTSs, our SERS-based LFTS exhibits further advantages, including high sensitivity and anti-interference ability towards blood plasma samples for protein biomarker detection, which shows great promise for biomedical diagnosis of diseases.
Towards virtualized transradial prosthesis with simulated dynamics and surface EMG interface | Amanda Barbarossa, Matthew Boots, Anton Sobinov, and Sergiy Yakovenko

Department of Chemical and Biomedical Engineering, Center for Neuroscience,
Mechanical and Aerospace Engineering Department, Exercise Physiology,
West Virginia University, Morgantown, WV 26506

Highly advanced upper limb prosthetics often require an extensive trial and error process that can be frustrating for amputees. Diminished residual ability necessitates a personalized approach, accounting for the individual’s capability to control prosthetics of different complexity. This problem may be solved by “test-driving” virtualized prosthesis. This study is driven by the rationale to develop the framework for testing the real-time control of a virtual hand model with 18 degrees of freedom (DoF) in a custom virtual reality (Multi-Joint dynamics with Contact, MuJoCo). The biological control signals were captured from the transradial forearm with a low-cost, user-friendly surface EMG system (Myoband). Able-bodied subjects were asked to follow behavioral tasks including hand flexion and extension. The arm and hand motions were captured by the custom motion capture system based on multiple six-DoF tracking devices (HTC Vive). In the developed real-time system, users control a realistic biomimetic model. We expect to solve the problem of kinematic tracking for verifying correct control of the model. Future work will use these systems to create virtual prosthetic testing for amputees.
Nature inspired advanced ceramic 3D structures with high strength density ratio | Megan Schmitz, Vincent Spada, Kodey Jones and Edward M. Sabolsky

Department of Mechanical and Aerospace Engineering,
West Virginia University, Morgantown, WV 26506

Advanced ceramics have many desired properties for industries since they can withstand extremely hot temperatures, they are lightweight and they have high strength compared to other materials. The main problem with using ceramics in industry is that they are difficult to process and almost impossible to create complex structures. The microstructures that have been designed for mechanical testing are inspired by nature. The solution uses a 3D extruder to create the desired detailed shapes. The 3D extruder utilizes a syringe filled with a silicone paste, where upon ejection the paste is set with a UV light. After the complete part has been extruded the object will then be sintered in nitrogen gas. When silicone is introduced to nitrogen during sintering the object becomes a very strong ceramic. Once the object is extruded and sintered it is tested for mechanical strength. After testing the mechanical strength, the expected result is that the strength to density ratio of the natural microstructure will be higher than alternative materials.
Evaluating sediment loss from reclaimed roadside locations | Katelyn Kosar, Marissa Poultney and Leslie Hopkinson

Department of Civil and Environmental Engineering,
West Virginia University, Morgantown, WV 26506

Establishing vegetation cover following road construction is a common restoration practice to reduce sediment loss. This study evaluated the sediment yield from a roadside location in Grant County, West Virginia along US 48. Sediment yield was both measured in the field and predicted using the Modified Universal Soil Loss Equation (MUSLE). This study tested the ability of MUSLE to approximate sediment yield for a single storm. Four sediment traps were installed on four test plots (4.5 ft x 4.5 ft). Runoff samples were gathered from storms events, the samples were analyzed for total suspended solids, and sediment yield was calculated. Sediment yield was predicted using the geographic factors of the Moorefield area. Predicted erosion rates were over 18,000 times less than the measured values. These preliminary results suggest that MUSLE does not predict erosion rates adequately for small, roadside sites. Limitations included small plot size, limited rainfall, and plot distance. Data collections for additional storm events continue.
Blind search of isolated astrophysical pulses in phased array feed data | Morgan R. Menke and Natalia A. Schmid

Benjamin M. Statler College of Engineering and Mineral Resources, West Virginia University, Morgantown, WV 26506

In Radio Astronomy, the hardware has changed from using a single pixel receiver to using multiple sensors to widen the view angle of a telescope allowing for larger data collections in a single pointing. Due to this change in hardware, new signal detection algorithms, which optimally combine information “seeing” from each sensor, need to be developed. We have developed a mathematical description of an existing Phased Array Feed’s (PAF) geometry and related it to an array manifold. We have also developed a model for noise correlation created by the array. Raw data acquired by the PAF was simulated using the array manifold and the noise correlation matrix. Three different optimal beamforming algorithms were implemented and used to search for an astrophysical pulse and to estimate the direction of its arrival. A blind search, for pulse detection, was performed by forming multiple beams with varying angles and calculating the value of the conventional Signal-to-Noise ratio (SNR). We present the developed algorithm for raw data simulation and report the SNR values for each optimal beamforming approach.
Hand-held automatic diagnostic tool for the purposes of ear infection detection | Olivia Santee, James Smit h, James Hunsucker III, Matthew Smith and Andrew Smith

Benjamin M. Statler College of Engineering and Mineral Resources and West Virginia University Hospital, West Virginia University, Morgantown, WV 26506

Ear infections [acute otitis media (AOM)] primarily affect young children between 6 and 24 months old: a demographic who are unable to adequately communicate their symptoms. Billions of dollars are spent between doctor’s visits and treatments; however, not all ear infections require treatment from a doctor, often leading to overprescribed antibiotics. In order to more accurately and efficiently determine the presence of an ear infection, a hand-held automatic diagnostic tool is being developed. This research focuses on building a diagnostic tool that can be operated by an unskilled user to make a diagnosis. The initial prototype will be able to translate the qualitative parameters such as the color, position, and translucency of the tympanic membrane into a quantitative value that will be used to make the diagnosis. This technology has the capacity provide a significant benefit to the health of children in the United States as well as in developing countries.
Modeling and simulation of pollution control units for improved sustainability | Yacine Feliachi, Shuyun Li and Fernando V. Lima

Department of Chemical and Biomedical Engineering,
West Virginia University Morgantown, WV 26506

The United States Environmental Protection Agency, U.S. EPA, has long pushed for regulations on chemical processes to reduce the impact the processes have on the environment. In order to cope with the growing problem of pollution by the release of various chemicals into the environment, it is necessary to develop pieces of equipment to purify and clean waste streams from chemical plants. As part of this project, four pollution control units have been modeled in Microsoft Excel using Microsoft VBA. In order to test the units, a chemical process model for the simulation of the manufacture of acetic acid from methanol was developed in CHEMCAD. The output from the gaseous waste streams of the acetic acid production process are connected to the units, where the waste products are treated to minimize the environmental releases. Preliminary results show that the pollution control units are capable of treating two process output streams, thus increasing process feasibility and sustainability.
Exploring undergraduate transportation engineering student attitudes towards a novel research topic identification technique | J.D. Gatlin, V.D. Pyrialakou, H.A. Diefes-Dux, K. Gkritza and D. Martinelli

Department of Civil and Environmental Engineering,
West Virginia University, Morgantown, WV 26505, School of Engineering Education and Lyles School of Civil Engineering, Purdue University, West Lafayette, IN 47907

The expansion of active and project-based learning highlights new issues in and opportunities for engineering education. Some introductory transportation engineering instructors have begun to incorporate collaborative, semester-long research and synthesis projects to address the increasingly complex and interdisciplinary nature of their industry. Open topic selection, while praised for its ability to increase student ownership, appears to introduce problems of its own. At Purdue and West Virginia University, 141 students were surveyed about their use of a moderated research needs database in their topic identification process. Accounting for demographics and other control variables, and combining the survey responses with other datasets generated from student citations, products, and performance, this research will use quantitative analyses to test hypothesized singular and composite indicators associated with attitudes in use of the database: high-performing students and groups, exposure, ease of use, content availability, and student strategies. Additionally, a set of descriptive statistics and a qualitative thematic analysis of the open-ended questions will begin to build a feedback mechanism for future use and improvement of the database in undergraduate academia.
Deep learning for supervised domain adaptation and generalization | Timothy Scott, Saeid Motiian and Gianfranco Doretto

Lane Department of Computer Science and Electrical Engineering,
West Virginia University, Morgantown, West Virginia, 26506

Deep learning has shown great success in many fields, especially computer vision. One of the main limitations of deep learning is availability of large datasets. One solution to this problem is supervised domain adaptation, in which a deep learning model is trained on one large labeled dataset, and then adapted to work on a different task, using a very small, labeled dataset. In this work, we investigate the effectiveness of several different domain adaptation techniques, including alignment, separation, contrastive loss, and triplet loss. We used the MNIST and USPS datasets, which are two slightly different sets of images of handwritten digits. We show that contrastive loss and triplet loss perform significantly better than a baseline experiment with no explicit domain adaptation. Specifically, on MNIST to USPS adaptation with one sample per class, contrastive loss achieves 88.6% accuracy vs. 77.6% for the baseline. On USPS to MNIST adaptation, contrastive loss achieves 91.7% accuracy vs. 54.9% for the baseline. The triplet loss achieved accuracy not significantly different from the contrastive loss.
Robust empirical modeling of stream Chlorophyll-a across time and space: scaling by a single reference observation | Brittany I. Brush, Omar I. Abdul-Aziz and Mohammad A. Z. Siddik

Department of Civil and Environmental Engineering,
West Virginia University, Morgantown, WV 26505

Chlorophyll-a (Chl-a) is a green pigment in plants, and is a powerful indicator of stream water quality and ecosystem health. Chl-a concentrations typically represent a diurnal cycle, which can vary across sites, seasons and years. However, continuous observations of Chl-a are rarely available. A[u1] [OAA2] robust empirical model was developed to solve this problem by using continuous (hourly) observations recorded in 26 streams across the continental United States. A single reference observation from each diurnal cycle was considered as a scaling parameter to normalize different cycles into a general dimensionless cycle. The scaled cycle was then estimated by employing an extended stochastic harmonic algorithm. Model calibrations and validations indicated good predictions. Estimated model parameters did not noticeably vary in time and space, leading to a single parameter set. The model can predict the entire diurnal cycle of hourly Chl-a from the corresponding single reference observation for a chosen stream. The model can, therefore, be used as an ecological engineering tool to robustly estimate missing data, and make a dynamic assessment of stream health.
Effect of obstacles on premixed flame oscillation in narrow, fully open channels | Elizabeth Ridgeway, Abdulafeez Adebiyi and V’yacheslav Akkerman

Center for Innovation in Gas Research and Utilization (CIGRU); Center for Alternative Fuels, Engines and Emission (CAFEE), Department of Mechanical and Aerospace Engineering, West Virginia University Morgantown, West Virginia 26506-6106, USA

Studying combustion in channels/tubes is important in today’s fuel-reliant society. The more we know about these catastrophic events, the more likely it is that detonation could be prevented or mitigated. This research aimed to understand how varying different parameters affected oscillation patterns of flame velocity in narrow channels with evenly spaced obstacles along the walls, similar to a comb. The parameters varied included: obstacle size, obstacle spacing, thermal expansion coefficient (different expansion coefficient signifies different fuel mixtures, and thus burning temperature), and resolution. The work was conducted using a program that calculated flame specifications when different variables were altered. Then, the data was analyzed using graphing tools and video simulations. It was found that period of oscillation increased with larger obstacle size, larger spaces between the obstacles, and smaller expansion coefficient. Amplitude increased with smaller thermal expansion coefficient and larger spaces between obstacles, and average burning rate increased with smaller expansion coefficient and smaller spaces between obstacles.
Musculoskeletal modeling of the lower-limb: a novel approach for locomotor rehabilitation | Abby Williamson , Matthew Boots , Anton Sobinov and Sergiy Yakovenko ,

Hajim School of Engineering, University of Rochester, Rochester, NY 14627 and Center for Neuroscience, West Virginia University, Morgantown, WV 26506

Locomotor rehabilitation is an effective method for improving basic activities of daily living after stroke, spinal cord injury, or amputation. However, multiple subject-specific factors can hinder improvements during rehabilitation. These problems could be addressed with assistive technologies (i.e. prostheses or the functional electrical stimulation operated by biomimetic controllers). This project aimed to develop the basis for all biomimetic controllers: a biomechanical real‐time model of the human lower‐limb. I developed a musculoskeletal model with 46 main leg muscles spanning the hip, knee, and ankle joints. The constraints on muscle path geometry were adjusted interactively until the simulated multi-dimensional relationship matched the experimental dataset collected from over 15 published sources. We validated muscle models using both subjective and analytical measures. Model function was than evaluated in a real-time experiment to control virtualized leg movement using 16 EMG signals streamed from main leg muscles. The model simulated a cycling motion similar to the desired movement. The lower-limb model is a novel tool for real-time prosthetic control. Moreover, this model can be used to evaluate residual functions in lower-limb amputees and in development of innovative quantitative methods for their locomotor rehabilitation.
Learning basic concepts of self-driving cars using an NXP cup car | Gregory Hughes and Powsiri Klinkhachorn

Lane Department of Computer Science and Electrical Engineering, West Virginia University, Morgantown, WV 26506

The feasibility of self-driving cars is a heavily researched topic in the field of Computer Science. This led the NXP Semiconductors company to start the NXP Cup, a competition where teams of students make and program little model cars to autonomously navigate a series of tracks as fast as possible, to help supplement their own research on the subject. To aid in the understanding of the basic concepts of self-driving cars, a partially assembled NXP cup car was acquired and tinkered with. The car came with a line-scan camera and a Freescale-brand microcontroller. Using the C language and the mbed online compiler, programs were written and transferred to the microcontroller to interface with the motor, the steering, and the camera all separately. I expect to be able to get those three components to work together for an extremely simplified demonstration of how self-driving cars function. While not groundbreaking, this project has broadened my understanding of microcontrollers and image processing.
Robotics research test platform | Conner Castle, Dylan Reynolds and Yu Gu

Benjamin M. Statler College of Engineering and Mineral Resources, Department of Mechanical and Aerospace Engineering, West Virginia University, Morgantown, WV 26506

Robotics research tends to involve many separate devices on which a certain experiment or research can be tested. These devices take up physical space and may not be adaptable as time progresses and research changes. This project proposes a single testing platform with the robustness and customizability to perform a multitude of differing types of robotics research; which can be updated and changed to the needs of the researcher as the research progresses. The proposed platform consists of a test bed, cameras, projector, computer, accompanying “base” robots, all built for a custom-made T-slot aluminum structure. The software operating the platform is the robotics operating system (ROS); which allows for a variety of different programs to be compatible, and the T-slot aluminum allows for customization as time progresses. The structure has been erected and one camera has been integrated with the ROS software. Once the system is entirely put together, a researcher will be afforded the ability to perform many types of research operating from a single convenient platform instead of many separate devices.
Autonomous pollination robot for use on blackberry and raspberry plants | Michael J. Lee, Cagri Kilic, Jared Strader, Jason Gross and Yu Gu

Department of Mechanical and Aerospace, West Virginia University, Morgantown, WV, 26506-6106

Due to the decline of native pollinators in recent years, the cost of farmers to rent these pollinators for their crops has increased. In order to alleviate the cost burden for farmers and offer a reliable alternative for pollination to increase productivity, a team at WVU has been tasked with designing a robot that will autonomously pollinate blackberry and raspberry plants in a controlled greenhouse environment. Machine learning is being used to “teach” the robot what a flower is, flowers' poses, and whether they are ready to pollinate. Inception v3's (Google's machine learning open source code) final layer is being retrained using TensorFlow to achieve this. The results currently achieve approximately 80% accuracy, but the goal is for over 95%. Other key programs are in the process of being finished so the accuracy can reach its desired level. After every part of the robot is complete, it will be tested on brambles to determine its efficiency compared to native pollinators.
Modeling and optimization of ion transport membrane for oxygen separation from air | Allyson M. Brezler, Gaurav V. Mirlekar, and Fernando V. Lima

Department of Chemical and Biomedical Engineering,
West Virginia University, Morgantown, WV 26506

The use of high purity oxygen instead of air in fossil fuel combustion processes can increase conversion and aid in the decrease of CO2 and NOx emissions. Cryogenic air separation is the current technology implemented to produce oxygen, but operation requires large energy inputs that result in high costs. Ion transport membranes (ITM) for oxygen separation is a new technology that can be used in place of the cryogenic process to decrease costs. An ITM model was designed by applying a permeation flux equation to a plug flow reactor design equation. Preliminary results show that after optimizations the cost of a membrane required to produce oxygen for an integrated gasification combined cycle (IGCC) power plant decreased from the base case design cost. Further optimizations will be discussed to improve cost reductions. By generating a model for the membrane that is overall less expensive than current air separation units will help push this technology further towards implementation into chemical processes.
High-pressure analysis for fluid flow properties of shale core samples | Steven H. Black, Mohamed Elsaig and Kashy Aminian

Benjamin M. Statler College of Engineering and Mineral Resources,
West Virginia University, Morgantown, WV, 26506

The petroleum and natural gas industry has been expanding in the recent years. There is an increasing desire to help the industry become more accurate, and efficient with its measurements, in order to expedite discovery and extraction of petroleum and natural gas. Testing samples at reservoir conditions will help complete this as well as greatly increase the understanding and accuracy of potential production amounts and rates. The process placed shale rock core samples in a controlled system so that analysis of fluid flow and potential productivity at reservoir conditions could be estimated. The system involves two tanks at set pressures that transfer gas from one to the other, then based on the gas flow rate it is possible to calculate permeability or porosity. The findings show an exponential relationship between the reservoir confining pressure and permeability, or porosity, as well as an even stronger relationship between pore pressure and these properties. The raw data collected can then be used with equations and corrections to give more information about the reservoir in absolute terms.

Engineering 2017

Engineering Index Poster #:

Follow Us