(Huntsville, Ala. | December 19, 2019) – CFD Research Corporation today announced the award of an Army SBIR Phase II project to develop a novel machine learning (ML) capability for real-time monitoring, prognostics, and control of complex mechanical systems. There is considerable economic pressure to extend the in-service life of existing complex mechanical systems in Army vehicles due to increasing costs and shrinking budgets. CFD Research’s intelligent analysis of sensor readings and Health and Usage Monitoring Systems (HUMS) data for real-time situational awareness and response can offer early fault detection for all significant moving parts (e.g. engines, gearboxes, shafts, fans, rotor systems) in Army vehicles and facilitate mission success.
Current HUMS devices provide operators, both onboard and ground-stationed, with critical component health reports based on component sensor readings, but the Army recognized that CFD Research’s real-time, autonomous situational response system could reduce operator dependence and provide immediate and automatic remedies when a vehicle is operating in a degraded state.
“The end-product arising from this effort will be an innovative, commercial-quality, plug-and-play system that can be integrated into any vehicle’s HUMS interface and provide real-time data analytics, adapt to the changing dynamics of the vehicle, and perform autonomous corrective control when vehicle operation is degraded,” said Jackson Cornelius, CFD Research Principal Investigator.
In prior work with Dr. Yi Wang at the University of South Carolina (USC), CFD Research was able to demonstrate real-time HUMS data analysis including “around-the-clock” neural network learning and model predictive control on a resource-limited embedded computing platform. Evolutionary optimization was used to automate feature/sensor selection and neural network design choices to ensure that high-performing, efficient system models were generated while removing the need for prior prognostics expertise. Using extensive US Army relevant case studies, demonstrations showed exceptional in-process vehicle dynamics tracking, health characterization, and control reconfiguration capabilities that enabled mission completion under degraded system performance.
Through continued collaboration with USC, this new effort will optimize performance and functionality of the plug-and-play system by refining the neural network model to increase efficiency and reliability. Already, fault detection and identification algorithms have been identified to accelerate situational response, and an embedded AI computing platform with hybrid CPU/GPU architecture will be utilized for enhanced deployment efficiency and integration with Army-designated workflows for HUMS data analysis. Both the software and the AI-enabled hardware will be extensively validated and demonstrated via selected case studies of Army interest including integration into an unmanned aerial vehicle and ground-based robot.
The proposed technology will be of tremendous commercial value in DoD, DHS, NASA, and civilian sectors. In the civilian sector, the framework and software platform will find widespread use in various engineering areas due to its robust and reliable data analytics; online ML and system identification capabilities; and real-time situational response, including aerospace, aviation, navigation, environmental monitoring, and earth observation.
About CFD Research: Since its inception in 1987, CFD Research has worked with government agencies, businesses and academia to provide innovative solutions within the Aerospace & Defense, Biomedical & Life Sciences, and Energy & Materials industries. Over the years CFD Research has earned multiple national awards for successful application and commercialization of innovative component/system technology prototypes, multi-physics simulation software, multi-disciplinary analyses, and expert support services. CFD Research is an ISO9001 and AS9100 registered company and is appraised at CMMI Level II for services. Learn more at www.cfdrc.com.