A list of projects and products GTC has developed
Automation and Autonomous Systems
Data Fusion for Improved Radiation Source Search
GTC presents a hardware-agnostic approach to locating radioactive materials that is transforming the field. The innovative model, which is grounded in principles of radiation physics and Bayesian statistics, enables confident estimates of source location and intensity to be made prior to arrival at the scene. This tool guides mobile detectors towards the sources, minimizing time required for threat response or containment of contamination.
The solution can be integrated directly with individual mobile detectors or hosted on a server such as TAK, through the advanced data fusion architecture. This architecture allows for the fusion of data from multiple networked detectors in the field, resulting in even faster and more accurate source location. Upgrade radiation detection capabilities with GTC’s cutting-edge solution.
ARADISS – Adaptive Real-time Anomaly Detection In Space Systems
ARADISS is a platform-agnostic and reusable framework for the rapid development of platform-optimized fault management tools. ARADISS is applicable to virtually all cyber-physical systems powered/supplemented by electrical batteries. Meaningful physical correlations from system operational variables can be mapped to battery voltage and current fluctuations by using computationally inexpensive machine learning models. This provides an approach that is transparent, explainable, and transferable to other systems making ARADISS ideal for on-board implementation on small space platforms to improve autonomous capabilities.
Automated Contingency Management of Water Recycling Systems
GTC has developed software for modeling and automated control of water recycling systems and its subcomponents. The software includes modeling of primary failure mechanisms: filter clogging and pump failures. Algorithms detect incipient failures and apply automated advanced controls (PID + Model Predictive Control) to optimize operations using a multi-objective cost function. GTC has deployed this technology with a cloud-based approach that ties into legacy PLC systems for remote system control, data management, and predictive maintenance and warning system. The origins of this technology are for NASA’s Space Habitats where processing power is limited; thus, the framework embeds event-based sampling to minimize the required CPU processing power. The software can be used for condition-based maintenance, remaining useful life predictions, and to satisfy strict operational and mission constraints, especially for space and remote applications.
Automated Resource Management Strategies for Resilient Habitat Systems
GTC has developed a framework utilizing machine learning algorithms to maintain the resilience of space habitat life support systems under subsystem failure conditions. In space and submarine operations, it becomes very critical to manage habitat conditions and resources like O2, N2, CO2 levels, drinking water, etc. to successfully complete missions. When a component or subsystem is compromised, the optimized control settings and schedule of operation of different systems on the habitat is not trivial. This framework allows machine learning algorithms to automate system control from data generated from a large number of previous simulations and optimizations which are not possible to implement in real-time during a mission due to computational resource constraints. This framework is applicable to complex systems comprised of many subsystems that operate together and interact with each other for resources like submarines, space habitats, ships, etc.
Hierarchical Fault Tolerant Control System for Unmanned Ground Vehicles (UGVs)
Global Technology Connection, Inc. developed rollover prediction/prevention and battery life monitoring algorithms for PackBots. Prevention of rollover in PackBots is critical as loss of robotic platforms can lead to failure of missions for soldiers on the field. Likewise, battery life monitoring is also important to missions where accurate state-of-charge and battery health aids in mission planning. Data collected from experiments on testbeds were used to design a fuzzy logic rollover risk assessment algorithm that provides normal, caution, and critical level notifications to the operator. When automatic rollover prevention is enabled, the robot will apply a brake when rollover risk is critical. The software package is ready for field deployment and experiments as an add-on package for robotic platforms.
Seismic-AI
RaNDL – RApid Nuclear & seismic events Discriminator & Locator using AI
RaNDL is a tool for seismic wave analysis with a suite of capabilities derived from multi-sensor data fusion and multiple deep learning algorithms. Phase detection and classification are performed simultaneously to identify the event type. A novel deep learning ensemble method is used for phase association. Lastly, RaNDL locates events spatially and temporally. These primary capabilities are powered by supporting algorithms, including an unsupervised multi-modal feature extraction process that automatically converts unstructured waveforms to usable structured data. While RaNDL was initially built as an earthquake and nuclear explosion analysis tool, GTC is transitioning this technology to monitor the micro-seismic events during fracking, waster water injection, mining activities for enhanced operational efficiency.
SMART FRAC – SeisMic AI in Real-Time for Informed Drilling and Fracking
SMART FRAC is a tool to improve fracking operations through better understanding of subsurface stress states through analyzing micro-seismic data. It is an innovative deep learning tool to turn ‘noisy’ sensor signals into actionable information during unconventional oil and gas extraction, such as horizontal drilling and hydraulic fracturing. Through our machine learning algorithms, SMART FRAC can automatically determine the focal mechanism and stress states of micro-seismic events. While stress state analyses are currently possible, they involve manual picking of the first arrivals and its polarities, which is cumbersome and time-consuming, and cannot be done in real-time or near-real-time.
Onset Determination
a phase onset time determination algorithm using deep-learning probabilistic regression. This approach trains on analyst picks, but does not require the analyst picks to be accurate. Rather the algorithm learns the probability spread of the picks through the whole training set. The underlying assumption is the spread of analyst picks throughout the dataset is centered around the ideal picks. The result is that the algorithm can out pick the analyst and produce a realistic probability distribution, which represents the algorithms’ confidence.
Modeling & Simulation
Enterprise Architecture using MBSE
An Enterprise Architecture model designed to replace traditional, document-based methods of facilities management by using model-based systems engineering (MBSE). The model connects master planning requirements, facility condition score, building information systems, work-order systems, standards, and more, and links them to stakeholders’ needs. This model acts as a sole source of truth, allows for easier access to information at relevant granularity, and enables better connectivity and traceability.
By connecting all of an organization’s facilities data with the buildings’ capabilities, condition, and stakeholder needs, EA-MBSE reduces cost, risk, and enables better decision-making for planning and resource allocation.
TRAIT – Transition-pipeline Recommender & Attrition Identification Tool
TRAIT is a data fusion and predictive analytics tool that provides insights to reduce costs of pilot training. It ingests student aviator training performance data and applies AI/ML techniques to estimate and monitor students’ training progression and proficiency in different key areas. Further, it estimates students’ risk of attrition and their suitability to different aircraft pipelines to guide pilot training operations.
TRACAIR – Trajectory Course Analyzer & Anomaly Identifier
TRACAIR is a technology developed and funded by NASA to improve aviation safety and operations. TRACAIR fuses disparate aircraft trajectory, air traffic congestion, and weather data to generate contextual trajectory baselines and to identify anomalous safety risk events. TRACAIR also has the capability to analyze individual flights and compare them with pre-generated baselines. Using the already developed data pipelines, a pilot or analyst could analyze, monitor, and compare a flight or segment of it with others as well as quantify pilot proficiency which has the potential to prevent high-risk flights and maneuvers. These features combined with our visual analytics dashboard improve knowledge of operations and enable agile decision-making to improve safety and operations.
Prognostics & Health Management
Corrosion Modeling & Life Cycle Analysis of Metal Canisters – CaskADES
CaskADES is a physics-based life cycle analysis modeling tool for steel canisters subject to chloride-induced stress corrosion cracking (CISCC). The tool applies environmental conditions based on input locations and site data, and updates canister states by using Bayesian inference based on visual inspections and any crack mitigation operations performed like cold spraying, shot peening, and laser peening operations. CaskADES is designed for the management of spent nuclear fuel holding steel canisters, license extension applications, but is applicable for all steel canisters undergoing CISCC.
Aircraft Generator Diagnostics and Health Management Tool
Global Technology Connection, Inc. (GTC) has developed aircraft generator diagnostic algorithms to differentiate between failure modes (bearings, winding failures, etc.), normal aircraft operational modes (speed, load, etc.), and environments. The algorithms correlate electrical signatures to the state-of-health (failure type, severity, and location) of low-hour/healthy and degraded generators. The algorithms are embedded into a walkup tester (portable PC, data acquisition unit, sensors) and an aircraft power generation system to manage the health of the system. The walkup tester can be used on generators during maintenance or pre/post-flight checks of generators onboard aircraft on the ground.
Dynamic Maintenance Scheduling Algorithm for Military Aircraft Fleet
GTC developed a framework to utilize prognostic health information from aircraft systems like fuel pumps to improve aircraft fleet maintenance schedules. The algorithms consider resource (parts, depot availability, tools) and personnel availability constraints and rules for maintenance operations (order of operations). The framework also allows for unscheduled maintenance operations to be dynamically included to minimize overall delay and schedule related costs. In order to achieve an optimal schedule solution, costs are developed for delays in maintenance, delays or unavailability of aircraft for a mission, etc, and the schedule with the lowest cost is arrived at using automated optimization algorithms without requiring any human effort.
Valve Prognostic Health Monitoring System
Global Technology Connection, Inc. in collaboration with academic partners developed a Valve Prognostic Health Monitoring System (VPHMS) that detects and identifies valve incipient failure modes and estimates valve remaining useful life from sensor data. Hydraulically actuated butterfly valves are considered as they are commonly used in the industry. Valve failure modes that are primarily examined are seal failures, sensor failures, loose linkages, worn bearings, and degraded/contaminated hydraulic fluids. Valve failures will be identified through a set of condition indicator features and a fuzzy logic expert rule-base. The prognostics approach utilizes a trended anomaly detection output along with remaining useful life (RUL) threshold governed by the identified failure to predict the time-to-maintenance for a valve
