Research

Research Areas

• Industrial Predictive Analytics and Machine Learning
  • Industrial predictive analytics focuses on fundamental developments in statistical machine learning and data modeling aimed at delivering novel methodologies and algorithms specifically tailored to address the challenges of industrial data. Industrial data is generated from a variety of engineering systems and machinery that are highly dynamic and complex — often comprised of subsystems and components that possess intricate dependencies. Unlike social media, retail, and other applications, industrial data are also measured under extremely harsh environments resulting in numerous data quality challenges.  Unique research topics relate to systems exhibiting multiple fault/failure modes, systems comprised of multiple components with interactive degradation processes, and systems with high levels of data censoring and poor data quality.  The goal is to ensure the reliability and accuracy of the outputs generated by these novel predictive analytic algorithms to enhance operator confidence and ensure lean operations of critical capital-intensive assets. Focus applications have been in the manufacturing and power sectors.

• The HOME Project
  • Habitats Optimized for Missions of Exploration is a multi-university Space Technology Research Institute intended to execute NASA’s strategic vision around deep space human-exploration missions.  To this end, the design of deep-space habitats requires a fundamentally disruptive approach – one that relies not only on traditional subsystem reliability engineering and probabilistic risk analysis, but on emergent technologies in autonomous systems, failure-tolerant design, human/automation teaming, dense sensor populations, data science, Machine Learning (ML), robotic maintenance, and on-board manufacturing.  HOME is tasked with two primary operational requirements for NASA’s deep-space habitats, (1) Keep humans alive while they are resident, and (2) Keep the vehicle/habitat alive (operational) while they are not.

• • Research in this project revolves around several topics:

1. 1. • Autonomous predictive analytics for Self-Awareness: Self-Awareness (SA)refers to the habitat’s ability to utilize onboard information and data to maintain constant situational awareness and assessment. Autonomy is a key component of SA.  This requires revisiting conventional ML/AI and data analytic pipelines, identifying all explicit and implicit human-in-the-loop components, and and redesigning these frameworks to be “Earth/Human Independent”.

• • • MRO Optimization: In the context of SA, the theme of this topic centers on generating optimal maintenance/repair schedules. The state-of-the-art in preventive maintenance (PM) and condition-based maintenance (CBM) is very rich and has been studied extensively. However, one of the major limitations of actually implementing many of these models is a byproduct of mathematical convenience/elegance. The resulting maintenance schedules are often rigid and unrealistic. In reality, maintenance schedules need to be adaptable.  This requires a streamlined integration between diagnostics and prognostics with MRO optimization models – optimization models that consider system/component stochasticity and uncertainty as well as the variety of strategies spanning classic corrective replacement to opportunistic repairs and component cannibalization.

• • • Autonomous Logistics for Spare Parts: Separately optimizing MRO and spare part logistics always yields either excessive overstocking and obsolescence or mission-critical stockouts and shortages. It is, therefore, necessary to jointly optimize both spare parts logistics in light of planned MRO activities. Joint maintenance and inventory models a limited to PM policies, often incorporating rigid assumptions regarding demand (e.g. failures are random) and do not account for component reliability.  to ensure autonomy, spare parts provisioning policies need  to leverage long-term life predictions and as well as reliability estimates. This requires a tight coupling between maintenance scheduling, spare parts provisioning and operational requirements, such as derating operations to delay maintenance (life extension) and compensate for longer than usual lead-times.

• Cyber Security and Data Privacy in IoT Systems
  • Affordable sensor technologies coupled with wireless communications have given rise to a growing wave of industrial digitization. Digitization of this kind comes with an increased level of automation and digital control components that are collectively referred to as Industrial Control Systems (ICS). The growing integration of IoT devices has increased the vulnerability of ICS components to various kinds of cyberattacks.  This research area focuses on utilizing data analytics and model-based frameworks to detect cyberattacks that target ICS components with the intent to cause unexpected breakdowns and/or increase system inefficiency and degradation.