Through the RISE program, mechanical engineering undergraduate students leverage our state-of-the-art facilities working alongside internationally-renowned faculty to tackle cutting edge projects that impact our society. The Mechanical Engineering Undergraduate Symposium (MEUS) is the accumulation of the students’ work.
During the day, seniors in ME 490 will present their RISE projects in 20 minute presentations. Everyone is welcome to attend these sessions and ask probing questions!
Sophomores and juniors conclude their projects with a poster session during the evening reception, where students will be available to discuss their projects in detail. The posters will also be on display during the day, if you are unable to attend the reception.
The public is invited to peruse the posters, attend presentations, and interact with the students throughout the day.
Complete poster and presentation schedule details can be found in the Symposium Schedule. RISE project technical abstracts can be viewed below.
Posters (ME 290 and 390)
A Fractional-Order Model-Free Predictor for Delay Compensation in Connected TestbedsStudent: Sicong Guo
Instructor: Tulga Ersal
Abstract: One major challenge in implementing remotely-integrated closed-loop systems is that the delays introduced through the communication network can deteriorate the performance of the system and even destabilize it. Predictive frameworks are required to compensate for communication delays, and ultimately maintain high-fidelity information exchange. In previous work, a proposed first-order model-free predictor with two degrees of freedom has been validated to be robust in delay compensation and more flexible in performance tune-ups, in both simulation and actual connected testbed environments. Our goal in this project is to further explore the potential of this predictor in delay compensation when fractional orders are used for the derivatives and integrals in the predictor dynamics. By studying the frequency response of this predictor as a function of the fractional orders, especially in the dominant frequency regions observed in our studied signals, we realize a potential improvement in its delay compensation performance. We also study the potential of a new framework with self-tuning parameters. Then, we test their transient performance in simulation and compare with our previous results. As two example applications, we use a networked motor-shaft-geartrain system and a driver-vehicle-engine system. Under the same performance metrics established in our previous work, we observe limited space for improvements in the transient performance of the proposed fractional and self-tuning predictors. Nevertheless, even if does not improve the performance, the self-tuning framework still offers the benefit of eliminating manual tuning.
Internally-Tiled Inflatable Morphing Windshield CowlingStudent: Tiantian Li
Instructor: Diann Brei
Abstract not yet available.
Medical AI Diagnosis Program Uncertainty QuantificationStudent: Don Guo
Instructor: Xun Huan
Abstract: Deep learning has emerged as a powerful and prevalent technique for building data-driven models in medical diagnosis. However, most models only report single-value predictions, and are not capable of providing prediction uncertainty resulting from, among other sources, noisy and limited training data. We thus seek to develop a computational capability for quantifying uncertainty in neural network (NN) models in a systematic manner. We focus on a particular residual convolutional NN model developed to predict isocitrate dehydrogenase (IDH) mutation status in gliomas from preoperative brain magnetic resonance imaging (MRI) [Chang et al., 2017]. Enabling uncertainty quantification for this model is crucial for supporting subsequent treatment decision making. As a first step, we analyze the sensitivity of model prediction to the trained weights. This is achieved through a Monte Carlo sampling approach, where random noise is added to the trained weights. Preliminary results indicate that a 5% perturbation to the weights can alter the prediction probability of IDH mutation up to 10%.
Keywords: uncertainty quantification; deep learning; neural network models; medical artificial intelligence
Optimal Distribution of Tasks in Human-Autonomy Teams Student: Peihang Li
Instructor: Bogdan Epureanu
Abstract not yet available.
Quantum Mechanical Calculations of Material Properties Student: Shukan Parekh
Instructor: Vikram Gavini
Abstract: The objective of this research was to help develop and test DFT-FE(Density Functional Theory Calculations with Finite-Elements). DFT-FE is a code developed by Professor Gavini’s group and is based on adaptive finite-element basis for ground-state pseudopotential and all-electron calculations. The main aspect of my work was to evaluate the a priori and a posteriori mesh adaptation capabilities of the code for a wide range of benchmark systems. I was also tasked with writing file-conversion algorithms to provide the pseudopotentials in various formats for the DFT-FE code and eliminate the need to perform these actions manually. To accomplish the aforementioned goals, I used the ‘libxml’ library in C++. My first project involved extracting the pseudopotential, local potential, and mesh spacing values from a .XML file into a format that was usable by the code. I also modified an existing .UPF file-converter and wrote a driver program to incorporate both file-conversion algorithms into a single executable in order to make the code easy to use. Before the open-source beta release of the code, I tested the accuracy and scalability of the code using the Flux HPC cluster. The code can handle non-periodic, semi-periodic and periodic geometries. For pseudopotential calculations, the code works with norm-conserving ONCV and Troullier-Martins pseudopotentials. The code incorporates all-electron and pseudopotential calculations in the same framework, accommodates arbitrary boundary conditions, and incorporates scalable and efficient solvers for the solution of the Kohn-Sham equations.
Determining the variables that affect the yield of Aluminum when recycling chipsStudent: Ashvin Sharma
Instructor: Daniel Cooper
Abstract: Every year, over 400kt of Aluminum is lost in the U.S. during recycling because of dross generation: oxidation and entrapment of metal. The research objectives are to first determine the effect of material and geometric factors on metal recovery during aluminum sheet and machined chip recycling. These factors includes: surface roughness, thickness and alloy constituents. Secondly, we assess the viability of oxidation and wetting systems as the predominant mechanism in dross generation.
Melting experiments, to calculate melt loss, were conducted on sheet metals with varying thicknesses (.016” to 7/16”), alloys (1100 & 6061), surface roughness (0.8 to 6 μm) and on machined chips at a constant mass. A plot of melt loss against thickness demonstrates a significant difference in melt loss between sheet metal and machined chips. Our conclusions are that an increase in surface area available for oxidation results in increased melt losses and that different alloying constituents, like Mg, have varying oxidation affinities which also leads to differences in melt loss. This explains the decrease in aluminum recovery when thicknesses are decreased at a constant mass. I used this data to fit mechanistic models of wetting/entrapment and oxidation to describe the dominant cause of loss during melt recovery. The oxidation and wetting model fit well with the current results and indicate that oxidation is the more dominant mechanism at work; however, further testing is required, to increase reliability, in conjunction with chemical analysis of dross to determine the elements present after melt recovery.
Standardization of Microfluidic Devices Student: Ahmed Nafis Arafat
Instructor: Allen Liu
Abstract not yet available.
Localisation of dolphins to determine behaviors Student: Ritika Sibal
Instructor: Kira Barton
Abstract: The behaviors and conditions of animals living in the wild can be visually classified, laboriously, by highly trained experts, however it becomes significantly harder for the interested public, who may eager to learn from either the mechanics of an animal’s locomotion, or the conditions that lead to certain behaviors being exhibited, to classify the behaviors with a high degree of certainty. Moreover, it is time inefficient to closely monitor a wild animal, in its natural state, for extended periods of time (months/years) by simple visual inspection. This project was interested in developing automatic tools that enable efficient behavior monitoring and classification of animals, more specifically, bottlenose dolphins under managed care in Hawaii. Noninvasive tags with camera, IMU (accelerometers, gyroscopes and magnetometer) and pressure sensors were attached to the dolphins via suction cups. Data from the tags was then used to train an artificial neural network, which would be used to classify dolphins’ behaviors automatically providing new tag data. As such, the first step in this process was to create a way to efficiently record the groundtruth state of the dolphins’ behavior visually that can be used for model training. This step included developing a MATLAB program that allowed the labeling of behavior visually and then recording the matching tag data. Which was then used to create a testing and ground truth bench. Next, literature was researched and analyzed to highlight which neural network would have the most success within this application. Preliminary implementation and testing of a LSTM (Long Short-Term Memory) network was conducted, and will continue to be built upon in the upcoming semester to enhance its performance. Once completed, the LSTM network will be able to predict variable dolphin behavior with a high degree of certainty providing tag data of arbitrary length.
Soft Robotic Stroke Rehabilitation using a Coupled Joint DesignStudent: Melissa Brei
Instructor: Brent Gillespie
Abstract not yet available.
VR/AR Interface ProjectStudent: Zachary Taylor
Instructor: Shorya Awtar
Abstract: We are evaluating the possibility of utilizing the process of parallel kinematics for virtual reality. We are also evaluating the best approaches to minimizes signal delay and improve reliability and accuracy, while maintaining low costs and intuitiveness of controls. My primary focus has been on minimizing size and weight by looking at encoders under 15 mm in diameter and materials such as 3D printer plastics and silicone or polypropylene.
Modeling and Simulation of Turbomachinery DampersStudent: Justin Shim
Instructor: Bogdan Epureanu
Abstract: Bladed disks, also known as blisks or rotors, are critical components in modern turbomachinery. These bladed disks are at the core of aircraft engines and gas turbines. Blisks operate at high rotational speeds and undergo severe mechanical stresses. Furthermore, they are susceptible to vibrations due to the high sensitivity of the system to mistuning, i.e., any variation from nominal property or geometry. Under such conditions, blisks can suffer from structural fatigue over a large number of load cycles, possibly leading to cracks or even system failure. With advancements in manufacturing technologies, turbomachinery design is moving towards higher performing integrally bladed rotors (IBRs) and away from conventional rotors. However, the joints in conventional rotors between the disk and blades naturally provided damping to the system. In order to combat high cycle fatigue in modern-day IBRs, dampers have been used. This project aims to explore the effectiveness of a tuned mass damper (TMD) that shifts the resonant frequency of the original blisk system and dissipates energy through frictional nonlinearities. Parametric studies and modal and harmonic analyses through ANSYS Workbench software were used to design a TMD able to reduce vibration on a cantilever beam, which represents the blade of a blisk. Future work will focus on the simulation of frictional nonlinearities and testing of the system.
Keywords: turbomachinery, blisk, tuned mass damper, modal analysis, parametric study.
Paper Sessions (ME 490)
1. Processing, Manufacturing, & Filtration
Investigation into the effects of rapid cooling on the strength and quality of a heatstake Student: Aditya Iyer
Instructor: Chinedum Okwudire
Abstract not yet available.
Green Manufacturing GroupStudent: Rana Dabaja
Instructor: Mihaela Banu
Abstract not yet available.
Characterization of Lightweight Materials Using Molecular Dynamic SimulationStudent: Yi Qiu
Instructor: Mihaela Banu
Abstract not yet available.
MDP Amway Air Filtration: CFD & Acoustic Model DevelopmentStudent: Peter Nguyen
Instructor: David Dowling
Abstract not yet available.
Experimental and numerical characteristics of thermoplastic fiber composites including interfaces.Student: Zonta Owens
Instructor: Mihaela Banu
Abstract not yet available.
2. Energy Transmission and Sensing
Advanced Catheter Assembly for Traumatic Brain Injury (TBI) MonitoringStudent: James Kennan
Instructor: Kenn Oldham
Abstract: Traumatic Brain Injury (TBI) is a serious health condition that results from blunt force trauma to the head. Symptoms of a Traumatic Brain Injury may include swelling of the brain, which causes an increase in intracranial pressure (ICP), and cerebral ischemia, a condition in which the blood flow to the brain is insufficient and results in poor oxygen supply to the brain tissue. It is essential to monitor both the intracranial pressure and blood volume flow within the brain in order to prevent permanent damage. In order to track this condition, a cerebral catheter can be used. Current cerebral catheters monitor ICP using piezo electrics; however, this does not provide information on the volume of blood reaching the brain. During this research, fiber optic reflectance was used to obtain a PPG signal which can help with monitoring blood flow and piezoelectric (PVDF) was used to monitor changes in pressure. When combined, these signals can give insight into the amount of blood reaching the brain. This project explores methods of assembling this device and assessing the components necessary to achieve a significant PPG signal while being small enough to pass through a cerebral bolt into the brain. The objective is to design a device that can assist doctors in monitoring their patients by identifying changes in blood volume flow, and therefore reduce permanent brain damage caused by TBI.
EGR Cooler FoulingStudent: Jason Hebert
Instructor: Andre Boehman
Abstract not yet available.
Gas Chromatography Temperature Calibration for Rapid Compression Facility Research UseStudent: Caitlin Recker
Instructor: Margaret Wooldridge
Abstract: This paper presents the gas chromatography (GC) machine calibration results for methanol blends. These calibrations are critical in making the quantitative measurements which characterize fuel properties in transient systems, and will be utilized in the research of ignition delay times being conducted by doctoral candidates in the Wooldridge Combustion Lab.
For this research, synthetic methanol-nitrogen blends of varying concentrations were created and analyzed by the four separate GC machines present in the Wooldridge Combustion Lab. The mixture was analyzed by each GC machine several times at multiple dilution conditions to ensure accuracy and repeatability of the data. Through MATLAB analysis of the resulting GC data, usable voltage readouts were collected for each experiment. Conversion of these voltage readouts correspond to the known concentration values used for each experiment conducted. Each of the data points for all four GC machines were plotted on separate graphs, and through linear regression the calibration curves were determined.
The unique calibration curves for each GC are necessary for the separation and quantification of species present in gas samples of unknown concentration at varying temperatures, pressures, and time intervals of the combustion process. The resulting knowledge of the combustion chemistry from these experiments will provide useful information in the making of future transportation policy.
GM TransmissionStudent: Perry Francois-Edwards
Instructor: Greg Hulbert
Abstract: General Motors (GM) has requested that the University of Michigan Multidisciplinary Design Project Team investigate how to most accurately predict fuel economy for 9 and 10 speed transmission applications, with minimal temperature points and a specific transmission loss test. Initially, Design of Experiment (DOE) test matrices were created to obtain raw data for the 9 and 10 speed transmissions, based on the transmission loss tests. The two transmission loss tests, Spinloss and Efficiency, were compared and analyzed to finally give GM its final recommendations to accurately predict fuel economy.
To clearly understand how the transmissions performed, the transmissions functionalities need to be investigated. The basis of the additional work involves investigating the 9 and 10 speed transmission properties and how they could affect the performance of the transmissions. To fully understand the effects of the transmissions, the 9 and 10 speed operations and torque loss, at different temperatures and input conditions, will be discussed. Along with its specific properties, which include parts, fuels and materials. General differences between the GM given transmissions and a lower gear state transmission will be discussed.
GM Transmission LossesStudent: Cullen Hoffman
Instructor: Greg Hulbert
Abstract not yet available.
3. Robotics and Legged Locomotion
Three Leg Leaf Spring Hopper Student: Taylor McLaughlin
Instructor: Ram Vasudevan
Abstract: In recent years, interest in legged robotics has piqued within the engineering community. However, in practice few of these platforms display the capacity to recover a significant fraction of their mechanical energy step-to-step. Energy recovery has been observed in humans and kangaroos, where the elastic energy stored can range from 20-50%. In this contribution I discuss the design, fabrication, and testing of an oscillatory hopping tripod, with elastic energy storage. When repeatedly jumping, some of the energy from previous jump can be reused in the next jump. For this robot, the legs are made of spring steel, which acts as the elastic energy storage mechanism. I present the iterative design of the energy storage mechanism, the tests done to measure passive energy storage and spring stiffness, and the initial work on an adaptive Hopf oscillator for leg phase synchronization.
Dynamics on Large Beam Deflection for Robot GaitsStudent: Ki Yun Lee
Instructor: Kenn Oldham
Abstract not yet available.
The VSO: A Quasipassive Ankle Foot Orthosis with Continuously Variable StiffnessStudent: Marcos Cavallin
Instructor: Elliott Rouse
Abstract: Most orthotic devices improperly address pathological mobility deficits because they have a static stiffness throughout gait and neglect the variable stiffness exhibited by a healthy human ankle across and within ambulatory tasks. Devices that mitigate this issue are not commercially available and usually consist of powered ankle foot orthoses with large actuators and cabling that makes transport outside of a laboratory setting difficult. This project aims to develop a quasi-passive variable stiffness ankle foot orthosis (VSO), with a rolling cam transmission that allows the prescription of human torque data to the user’s ankle without the need for large actuators. The VSO has a customizable nonlinear torque angle profile and the ability to modulate its overall stiffness by varying the location of a spring support in leg swing. Modulating this support allows the device to span many different mobility tasks and patient populations that require different levels of overall stiffness and augmentation. In this work we introduce the design of the VSO as a possible solution to pathological mobility deficits and characterize the available torque-angle curves of the device.
Educational Platform For The First Open Sourced Robotic Ankle and Knee for ResearchersStudent: Enis Habib
Instructor: Elliott Rouse
Abstract: The objective of my project is to help build a website to disseminate how to acquire parts for an open sourced, inexpensive and high-performance robotic ankle and knee system for researchers so that they can work on prostheses without developing their own hardware, which is both time-consuming and expensive. Through standardizing the hardware, the robotic ankle and knee system aims researchers to collaborate with one another and create reproducible results. The results of my paper and the assembly videos I created will help researchers to understand how to assemble this complex electromechanical system that consists of over 100 parts. Additionally, researchers can also understand the design decisions and compromises that were made in order to make the system inexpensive, simple and high performance.
Keywords: robotics, rehabilitation, prosthetics, ankle, knee, assembly, design, open sourced
Web Based Documentation of Open Source Robotic LegStudent: Josiah Humphrey
Instructor: Elliott Rouse
Abstract: “Web Based Documentation of Open Source Robotic Leg”
Josiah Humphrey, Mechanical Engineering
In the United States alone, hundreds of thousands of leg amputees have a reduced quality of life due to persistent mobility limitations. In addition, advancement in the control of robotic leg prostheses has been hindered by lack of a uniform hardware system and a high barrier to entry. With the goal of unifying the research field of prosthetic leg controls and lowering barrier to entry, an open source robotic prosthetic leg (OSL) was proposed by the Neurobionics Lab at UM. The robotic leg has been designed and built, however, online documentation was needed to effectively disseminate information on part files, assembly, and current control strategies. To this end, a website was created to provide researchers with everything they need to build and control the robotic leg. The website contains downloadable part files that can be sent to a manufacturer, a bill of materials that lays out all parts that need to be ordered, videos on how to assemble the leg once the parts have been obtained, and control files to be run using a microcomputer to enable the leg to function properly. As a result, there will be no need for subsequent research groups to spend copious amounts of money and time designing a leg prior to studying control strategies. Ultimately, having a ubiquitous robotic leg will help facilitate comparison between control strategies, potentially streamlining the field towards highly functional robotic prosthetic legs. This paper will detail how and why the website was implemented.
Keywords: ambulation, robotics, prostheses, controls, feedback, impedance, position
4. Novel Materials and Actuation
Multifunctional Active Surface with rigidization capabilityStudent: Sarah Sober
Instructor: Diann Brei
Abstract not yet available.
Reconfigurable Inflatable Head Restraint SystemStudent: Zachary Boehm
Instructor: Diann Brei
Abstract not yet available.
Presentable Vehicle Accessory Controls using Multi-Layer Inflatable Membranes Student: William O'Neal
Instructor: Diann Brei
Abstract not yet available.
Developing Machine Learning Methods to Generate Analytic Descriptions of Data SetsStudent: Rik Banerjee
Instructor: Krishna Garikipati
Abstract not yet available.
Cold Acetone Vapor Chemical Polishing of ABS Surfaces Fabricated by Fused Deposition ModelingStudent: Byungwook Choi
Instructor: Albert Shih
Abstract: This study investigates the use of cold acetone vapor in surface polishing of acrylonitrile butadiene styrene (ABS) surfaces fabricated by fused deposition modeling (FDM). Prior works utilizing hot acetone vapor to polish ABS surfaces show the polishing process results in inconsistent surface finishes and poses a safety risk. The design of a previously developed enclosed test chamber is optimized to allow ABS FDM parts to be polished by cold acetone vapor. The factors of saturation or pre-processing time, part orientation, and polishing or processing time are investigated to determine the ideal conditions for ABS FDM surfaces to be polished. Results show that the optimized design of the test chamber can be fully saturated with acetone vapor within 45 minutes and can polish a 20 mm x 20 mm ABS FDM cube within 30 to 40 minutes in the optimal vertical orientation to achieve the desired surface finish. Dynamic movement with the intermittent rolling of the polishing part is also found to be able to polish all faces on a given cube.
Characterizing Fatigue Properties of Metastructures Student: Srikar Srivatsa
Instructor: Kon-Well Wang
Abstract: There has been recent work surrounding the development of a new passive vibration-suppression method called metastructures. Metastructures are a type of metamaterial: an object arranged or constructed in such a manner that it has traits unachievable by its constituent material. Metastructures, specifically, are structural elements containing masses on damped springs-- all homogeneously part of the structure. Tuning the structure to damp vibrations at a specific frequency is achieved by changing the size of mass dampers. These integrated mass dampers are used to redirect the transmission of energy from the rest of the structure for vibration suppression. However, this redirection of energy through the mass dampers indicates that they are prone to sustaining failure by fatigue. Characterizing the relationship between the mass dampers’ geometric parameters and their fatigue performance would inform the design and implementation of metastructures. The presence of multiple geometric parameters that might result in structures with the same natural frequency but different stresses distributions indicates an inverse problem. A sensitivity analysis was performed to understand the input parameter with the largest impact on natural frequency and stress concentration. Analytical models for natural frequency and maximum stress were developed for the mass dampers, and compared to a finite element model. The results indicate that the mass’ parameters have a larger impact on stress concentration and natural frequency than those of the spring-damper. The stress concentration results appear to agree with theory, but the natural frequency do not entirely appear to agree.
Keywords: metamaterial, fatigue, vibration suppression, mass damper
5. Product and Process Invention
Morphing Slit Skins with Distributed Pneumatic Delivery SchemesStudent: Paul Powers
Instructor: Diann Brei
Abstract not yet available.
An Investigation of the Effect of Varying Print Parameters on Print Quality for Silicone 3D PrintingStudent: Naman Shah
Instructor: Diann Brei
Abstract not yet available.
Prediction of Transverse Weld Characteristics in the Continuous Extrusion Process of AluminumStudent: Anselm Adrian Anak Christopher
Instructor: Daniel Cooper
Abstract: With an increase demand for aluminum in processing, there is a growing need for processes to be efficient. Currently, extrusion has a low material efficiency that leads to excess scraps. The source of this scrap comes from the formation of a weld between consecutively extruded billets which has to be removed as integrity cannot be guarantied. Thus, 2 possible approaches exist to reduce waste; reduce the length of the weld so less scrap is created or strengthen the weld so it need not be removed. This research aims to identify the conditions that affect the transverse weld length in direct extrusion. Through experimentation with clay, the weld length was observed and correlated with extrusion conditions; lubrication, die angle, extrusion ratio, die profile/shape and ram profile. An analytical upper-bound method was developed that predicts the weld length under plane strain extrusion conditions.
Pace BuddyStudent: Sarah Corbe
Instructor: Noel Perkins
Abstract: The objective of this project is to create an autonomous line-following robot that paces a runner at a specific speed and distance on a track. The project was inspired by existing autonomous robotic technologies, and the desire to expand and enhance functionality to amateur, professional, and military training. Using our previous mechanical prototype, the robot control system was designed to accurately follow the lines of the track using IR sensors that detect reflectivity of the track's surface. This prototype ultimately relies on a PD control loop and communication from the optical encoders and our IR sensors. This prototype proves the functionality of our design, but our next effort will focus on optimization of the prototype and user experience.
Pace BuddyStudent: Kristjan Cenko
Instructor: Noel Perkins
Abstract: The objective of this project is to create an autonomous line-following robot that paces a runner at a specific speed and distance on a track. The project was inspired by existing autonomous robotic technologies, and the desire to expand and enhance functionality to amateur, professional, and military training. Using our previous mechanical prototype, the robot control system was designed to accurately follow the lines of the track using IR sensors that detect reflectivity of the track's surface. This prototype ultimately relies on a PD control loop to reduce error and the amount of overshoot, relying on communication from the optical encoders and our IR sensors. This prototype proves the functionality of our design, but our next effort will focus on optimization of the prototype and user experience.
Reducing emissions by using cars for longer and more intensivelyStudent: Jiankan Liao
Instructor: Daniel Cooper
Abstract: Material production must be decreased to meet the recommended greenhouse gas emission cuts. This can be achieved by increasing intensity of use (IOU), which would reduce the lifespan of a product and increase the use-phase emission at the same time. Previous studies on increasing vehicle IOU only considered the use-phase impact and fail to consider impact of material production. The project mainly focuses on reducing material production and use-phase emission from light duty vehicles (LDV) by increasing IOU providing a temporal analysis on overall emission based on a given service demand for LDV, which is expressed in annual passenger mileage. The project uses several models including vehicle lifespan model (VLSM), vehicle emission model (VEM) and dynamic product flow analysis (DPFA) to quantitatively analyze the material production and use-phase emission. The analyses are applied to several custom scenarios, both retrospective and projective, based on actual situation. The result shows reduction of material production in short term and total emission reduction in long term.
Keywords: dematerialization, material efficiency, car sharing, car pooling
