The goal: To expedite cleaning of maritime distress signals without manual filtering. Maritime distress transmissions received by the U.S. Coast Guard are subject to environmental noise and communication channel distortion, and must be filtered manually using current Rescue 21 software, which can be time consuming and imprecise. This project uses deep learning to improve signal filtration. The core of the software is a de-noising recurrent auto-encoder, a machine learning model that recognizes and removes distortion from input audio files. Noisy distress signals are input into the team’s software and automatically filtered by the auto-encoder, which generates a coherent, noiseless version of the input signal.Testing required the software to process a 5-minute audio segment in under a minute while using less than 2 GB of local memory. The software is compatible with all operating systems used by the Coast Guard, including the existing Rescue 21 software.

Project goal: To create a system of wireless transceivers that communicate with a fast, consistent stream to transmit acceptable quality audio. The designed system consists of a microphone, computer and two speakers,each with its own wireless transceiver. The transmission is secured by use of an Advanced Encryption Standard protocol compatible with the transceivers. The intended system application is simplification of a dispatch center setup by reducing the wiring used for audio equipment. With this system’s integrated encryption, sensitive information can be transmitted securely. To secure the transmission the design requires insertion of a physical encryption key to begin transmission. Using a physical rather than digital key eliminates wireless exchange of keys, reducing the likelihood of any transmission being compromised. Off-the-shelf components were modified to fit the requirements of the design. Testing of the system verified data transmission speeds and encryption methods. Additional testing quantified wireless transmission range and audio quality.

Project goal: To eliminate the United States Coast Guard’s reliance on expensive proprietary software for digital distress signaling. Digital selective calling is an international standard paging system used by the United States Coast Guard to automate distress and safety alerts sent over terrestrial medium-frequency, high-frequency and very-high-frequency marine radio systems. Most digital selective calling systems rely upon commercial off-the-shelf products with proprietary interfaces. The proprietary nature of these systems limits the ability for maintenance contractors to perform technology refreshes and improvements on Coast Guard systems. The system designed uses an open standard radio interface for processing packetized digital selective calling data. The design incorporates a web-based graphical user interface that can send, receive and acknowledge digital selective calling messages. Processed messages are displayed in real time with a sortable 30-day message history. Archiving capability is included to allow the storage and retrieval of messages through an automatically generated file format.

Project goal: To create and run an embedded web server and HTML5 host on a small cyber-hardened“internet of things”-style device. The proliferation of devices connected to the “internet of things” presents numerous security challenges. The team developed a web server and host on cyber-hardened device using components found in the Gumstix Autonomous Kit for Overo, including a Caspa VL camera, a Pre-GO GPS, an accelerometer, and a gyroscope. The designed system gathers data relevant to an unmanned aircraft system application and transmits it securely to the client over Wi-Fi using authentication procedures and a secure communication protocol. The device pushes data from the camera, GPS, accelerometer and gyroscope in real-time, without the browser polling for this information. The client can use the web page to remotely view the output and status of each sensor and individually turn on and off data communication from each sensor. The system has been penetration-tested for security.

Project goal: To develop a material that can absorb tissue stains at an intensity less variable than human tissue, and that can withstand the physical and chemical processes encountered in staining equipment. A limitation of histological staining technologies is the inherent variability in human tissue, which can cause deviation in staining results and introduce difficulties in cancer diagnoses.The team developed a material that can absorb stains more consistently than human tissue. The lower variability of the material means that it functions as a control slide to identify instrument-induced stain intensity variation in diagnostic assays. The design incorporates polymers and cellulose compounds tested to be compatible with the acidophilic and basophilic properties of hematoxylin and eosin stains. These materials are adhered to glass microscope slides in sections under 100 microns thick in order to allow cover-slipping. The assembled slides were tested in the HE 600 staining instrument to confirm their ability to withstand the various temperature and chemical conditions throughout the staining process. After staining, the material was subjected to image analysis to verify that the stained final product was less variable than the current control slides using human tissue.

Project goal: To modify an existing low-profile gastrostomy tube system to include a reflux valve. The primary function of a gastrostomy tube is to transfer liquid food and nutrients directly into the stomach through a stoma. A new injection mold plate, core pins and cap molds were designed in SolidWorks and fabricated using silicon injection molding. Durometer analysis determined the optimal hardness of silicon needed to maintain the correct flexibility, sturdiness and bio-compatibility that would ensure complete medical functionality. The team designed a hub to house the sponsor-provided cross-slit petal reset valve, a cap that securely closes access to the stoma, and a set of core pins used to fabricate gastrostomy tubes of varying internal diameters. Pressurized leak tests and tensile strength tests were performed in accordance with Federal Drug Administration and medical device regulatory compliance standards. The testing was designed to verify that the new device is substantially equivalent to the existing device for future FDA regulatory approval and commercial viability.

Project goal: To develop a lung-on-a-chip that mimics the blood-gas exchange that occurs in the alveoli found in human lungs. The microfluidic-based lung-on-chip provides researchers with a cost-effective device that enables real-time ex vivo assays of foreign agent effects on human lung tissue with an air-liquid interface. Micro-fabrication techniques were used to create a device made of a clear, flexible and inert polydimethylsiloxane polymer. The device is composed of two stacked flow chambers separated by a porous membrane, which creates an air-liquid interface that mimics the lung’s alveoli. Epithelial lung cells on the top channel of the device are subjected to air flow, which recreates breathing motions, while the endothelial cells opposite the membrane are subjected to nutrient flow (that is, blood flow). A system that achieved real-time monitoring of the lung tissue was tested and demonstrated an application of the lung-on-chip. This involved the introduction of e-cigarette aerosol particles to the air chambers and subsequent analysis of the cell characteristics. It was confirmed that the system design sustains a tissue mono-layer in the micro-channel under regular air and liquid flow for an extended period.

Project goal: To design and implement a tool for measuring the thickness of histological tissue sections on a microscope slide. Tissue samples in a given protocol are processed in an identical fashion, but their thickness might vary. This inconsistency could influence the final staining results and, therefore, the resulting diagnosis. Mitigating this inconsistency requires determining variations in thicknesses by directly measuring the samples before they are processed. The design uses near-infrared reflectance confocal microscopy. A near-infrared laser is focused on the tissue section as it moves through the beam. Movement is tracked using a differential micrometer paired with a rotary encoder. As the index of refraction changes, particularly at the top and bottom interfaces of the section, the laser is reflected onto a photo-diode. Microscope objectives and a spatial filter are used to reduce noise, increasing sensitivity and signal-to-noise ratio. The thickness of the section is calculated by tracking how far the stage has moved between the strong signals corresponding to the bottom and top of the section.

Project goal: To design and build a more efficient and cost-effective system for aerating natural gas from saturated soil.Underground natural gas leaks cause the surrounding soil to become saturated with gas. If a gas leak occurs at a facility owned by the sponsor or one of its customers, the sponsor can remove the natural gas safely with aeration equipment. The aeration ejector uses the Venturi effect to force compressed air through an inlet nozzle designed to choke the flow and cause a pressure drop, which draws the natural gas out of the ground and into the second-stage nozzle before it is ejected through the diffuser to the atmosphere. To ensure the new design improved the efficiency of the old design, the current system was modeled in SolidWorks and complex flow simulations were conducted. Pressure gauges were used on the current system to validate the accuracy of the simulations. Simulations of the new design were used to optimize the internal geometry. The new system will save time, money and energy because it extracts the natural gas more efficiently.

Project goal: To create software to read and manipulate data from two commercial off-the-shelf unmanned aircraft. The unmanned aircraft follow a flight path and gather photographs of the surrounding area. Open source flight-control software ArduPilot is used to program waypoints into the unmanned aircraft using GPS coordinates. The team edited the software to handle directions for two unmanned aircraft, and modified it to output a data log. ArduPilot transfers the gathered imagery and unmanned aircraft statistical data to a mission planner software program coded in C++. The mission planner software displays aircraft velocity, elevation, battery life and photos via a graphical user interface, and generates a report for subsequent analysis and processing by proprietary software provided by the sponsor.