Follow

• 

  Propeller Partial Ground Effect

  Jielong "Jacky" Cai

  We extend our recent propeller ground-effect study to consider a circular ground-plate, instead of planar-surface of assumed infinite extent. Parameter studies include propeller to plate diameter ratio, propeller diameter to ground-offset ratio, and propeller pitch to diameter ratio. As with classical ground effect, benefits of thrust-augmentation and/or power-reduction with proximity to the ground, depend on the propeller pitch to diameter ratio. Flow visualization suggests that for larger pitch to diameter ratio, the lack of conclusively large ground-effect benefits can be attributed to stalled flow about the blade, and spatially more diffuse tip-vortices. A circular ground-plate of half of the propeller diameter was found to have almost no distinction from that of an unimpeded free-stream, while when the plate and propeller have the same diameter, the resulting ground-effect already resembles that of the infinite-plate.

• 

  Resin Transfer Molding (RTM): Analytical and Numerical Approach Development and Verification for a Linear and Radial Flow Model using PAM-RTM

  Khalid Aldhahri

  Liquid Composite Molding (LCM) processes are among the most commonly used processes in composite manufacturing, including resin transfer modeling (RTM) and vacuum-assisted resin transfer molding (VARTM). These types of processes provide several advantages, such as reduced cost, faster cycle time, and efficient part fabrication compared to autoclave molding. RTM is increasingly used to produce composite materials for several applications, especially in the aerospace industry. It offers mass production of composite parts with simple and complex geometry and with small to medium sizes. RTM involved four essential steps to produce the composite part: loading the fiber preform into the mold, injecting the mold with resin, resin curing, and composite demolding. The final composite properties are affected by these steps of production, especially the mold filling and resin curing steps. Numerical process models offer potential benefits for use in LCM, such as improving the mold design, optimizing the location of resin injection gates and vents, controlling the position of the resin flow front, and improving part quality. The main thrust of this study was to demonstrate the effectiveness of simulation applied to RTM for understanding how to choose the right location of injection and vent ports, cure development, and monitoring the position of resin. Numerical modeling allows for initial viewing of the expected flow patterns and cure profiles before the actual resin injection experiment. The commercial PAM-RTM software is used to simulate key process variables, including resin velocity, pressure distribution, filling time, and process parameters. These results will be validated with analytical solutions using Darcy’s Law applied to linear and circular flow models. The model example will be used to compare results of these two approaches for flat panel molds and using two different resin injection strategies for each: constant pressure and constant flow rate.

• 

  Semi Supervised Learning for Accurate Segmentation of Roughly Labeled Pathological Data

  Rachel Rajan

  Recent advancements in medical imaging research have shown that digitized high-resolution microscopic images combined with deep learning architectures have been able to generate promising results better than pathologists in the field of pathology diagnosis. But, for supervised deep learning techniques, the unavailability of labeled data has limited applications for accurate medical image segmentation. Hence, we propose an enhanced adversarial learning approach in semi-supervised segmentation for incremental training of our deep learning-based model to utilize unlabeled data in achieving better learning performance. Studies reveal that unlabeled data combined with small amount of labeled data can improve the overall performance considerably. Since most of the existing methods use weakly labeled images, our proposed technique utilizes unlabeled instances to improve the segmentation model. Experiments on two publicly available datasets such as PASCAL VOC2012 and UCSB Bio-Segmentation Benchmark dataset demonstrate the effectiveness of the proposed method.

• 

  Soft Robot Actuator Design for Digital Light Processing

  Dillon Montgomery Balk

  This research involves the design, simulation and fabrication of novel soft robot actuators. Since the 1970s, robot design engineers have been experimenting soft materials in robotics components. Inspired by natural organisms, “soft robotics” involves the integration of a soft polymer material into a mechanism in order to achieve a variety of configurations. Pneumatically actuated by air through hollow channels within, a soft robotics component allows for very large, non-linear, displacements compared to classical rigid body components. These attributes allow soft robotics to have potential biomedical, industrial, and rescue applications. This research project involves designing and simulating various soft robotic actuators to mimic primitive motions, including twisting, bending, elongating, and angular displacement. The various actuators can be assembled to form serial and parallel chains to perform basic robotic tasks, such as search-and-retrieval or pick and place operations. Digital light processing (DLP) technology is an appealing fabrication technique because it is able to create very intricate parts with high resolution. Utilizing UDRI’s DLP capabilities, experiments with physical prototypes will calibrate and validate the simulation results.

• 

  The Potential of Biomass Derived Fusel Alcohol Mixtures for Improved Engine Performance

  Lily Carolyn Behnke

  The use of biomass derived fusel alcohol mixtures as a blending agent with gasoline has the potential to lower the greenhouse gas emissions from passenger vehicles. In this work, a computational approach based on experimental data was utilized to predict the properties of a multitude of fusel blends at various blending percentages with gasoline. Fusel blends composed of isobutanol, isopentanol, 2-phenyl ethanol, and ethanol are demonstrated to provide comparable efficiency gains, increased energy density, and lower Reid vapor pressure values than ethanol when blended with gasoline. The results were analyzed on the basis of the potential efficiency increase of blends at 10%, 20%, and 30% by volume into a gasoline base fuel, and demonstrate that fusel blends have the opportunity to be optimized to maximize efficiency gain while minimizing increases to blending vapor pressure and decreases to energy density. Random forest regression analysis was used to describe the relative importance of the fusel blends regarding properties of interest. Furthermore, ordinary least squares analysis was performed to indicate each fusel alcohol’s positive or negative impact on the merit score. With the objective of implementing renewable fuel sources into gasoline, fusel blends can be used as a solution to further improve upon the overall performance of gasolines.

• 

  Thermal Impact of BFRP and GFRP Reinforced Concrete Beams

  Hao Dong

  This study describes a series of experiments examining the thermal damage of twelve concrete beams reinforced with Basalt and Glass FRP rebars. The 6-in by 6-in by 24-in BFRP/GFRP reinforced beams (single layered, two rebars) were molded with a 1.5-in concrete cover and exposed to heat for 30 minutes using a radiant heater with 54 kW/m² heat flux. Thermocouples were used to measure temperatures across the thickness of the beams. A three-point bending test was conducted 24 hours after heat exposure, and digital image correlation (DIC) system was used to analyze the displacements, strains’ distributions and crack propagations. There was no certain evidence to show that moment capacity reduction would occur after heat exposure under this scenario. A 1.5- in concrete cover is sufficient to protect the FRP rebars, and 30 minutes thermal exposure is recommended as incremental period for future research.

• 

  The Safety Impact of Raising Speed Limit on Rural Freeways in Ohio

  Oluwaseun Olufowobi

  The impacts of raised speed limits on traffic safety is an area that has generated much research, although a strong consensus has not emerged on the relationship between speed and safety. In response to the nationwide ongoing trend of raising speed limits, led to the implementation of the 70-mph speed limit on 570 miles of rural freeways in Ohio on July 1, 2013 and an additional 398 miles of rural freeways starting on September 29, 2013. The primary goal of the research detailed in this study is to investigate the safety impacts of this new speed limit using available crash, roadway, and traffic characteristics data. Statewide crash data from January 1, 2010 to December 31, 2018 were obtained from the Highway Safety Information System (HSIS). The study utilizes the empirical Bayes before-after study method. The intent of this method is to estimate the actual performance (in terms of crash frequency and severity) following the speed limit Increase and what the performance would have been if the increase in speed limit had not been applied.

• 

  Understanding Deep Neural Network Predictions for Medical Imaging Applications

  Redha Ali, Supun Samudika De Silva, Nathan Kremer Kueterman

  Computer-aided detection has been a research area attracting great interest in the past decade. Machine learning algorithms have been utilized extensively for this application as they provide a valuable second opinion to the doctors. Despite several machine learning models being available for medical imaging applications, not many have been implemented in the real-world due to the uninterpretable nature of the decisions made by the network. In this paper, we investigate the results provided by deep neural networks for the detection of malaria, diabetic retinopathy, brain tumor, and tuberculosis in different imaging modalities. We visualize the class activation mappings for all the applications in order to enhance the understanding of these networks. This type of visualization, along with the corresponding network performance metrics, would aid the data science experts in better understanding of their models as well as assisting doctors in their decision-making process.

• 

  United Rehabilitation Services Walker

  Christina Louise Diersing

  The United Rehabilitation Services (URS) walker is a walker modified for a young girl who is blind and autistic. It reminds her to hold onto her walker with both hands by playing Disney music or vibrating in a pattern she likes when she has both hands on the walker. When she takes her hands off of the walker, the music and/or vibrations will stop.

• 

  Unsupervised Real-Time Network Intrusion and Anomaly Detection by Memristor Based Autoencoder

  Md. Shahanur Alam

  Custom low power hardware systems for real-time network security and anomaly detection are in high demand, as these would allow for adequate protection in battery-powered network devices, such as edge devices and the internet of the things. This paper presents a memristor based system for real-time intrusion detection, as well as an anomaly detection based on autoencoders. Intrusion detection is performed by training only on a single autoencoder, and the overall detection accuracy of this system is 92.91%, with a malicious packet detection accuracy of 98.89%. The system described in this paper is also capable of using two autoencoders to perform anomaly detection using real-time online learning. Using this system, we show that the system flags anomalous data, but over time the system stops flagging a particular datatype if its presence is abundant. Utilizing memristors in these designs allows us to present extremely low power systems for intrusion and anomaly detection while sacrificing little accuracy.

• 

  Validation of Center of Mass Estimation in Humans

  Colin R. Theis

  Computational simulations of human movement require precise knowledge of the location of the body's center of mass (CoM). The CoM is a quantity that is impossible to directly measure experimentally. The location varies on a subject-to-subject basis and is essential for the generation of accurate simulations. OpenSim, a musculoskeletal modeling software, utilizes a built-in software function to estimate the CoM based off the geometry of the model. However, this estimation technique can be imprecise because the estimation is based solely off generic mass and geometry distributions. Since every individual is different with respect to body segment length and mass distributions, it is likely this estimation is inaccurate for all individuals. Therefore, using a new technique developed in UD’s DIMLab, we can estimate an individual's CoM more accurately than OpenSim. This estimation technique uses the statically equivalent serial chain model (SESC). The technique is based on a virtual chain, identified from a minimal amount of experimental kinematic data to be accurate. The system does not require knowledge of the total mass, or any of the individual segment mass or length properties. The SESC model is a function of the anatomical joint angles measured experimentally from the subject and terminates at the CoM. This project explores the feasibility of combining experimental CoM estimation methods with simulation based estimates of CoM. We aim to find a method to validate CoM estimates applied in simulations and improve simulation accuracy. We aim to integrate the SESC model into the OpenSim software package as the main mechanism for locating the CoM.