(3) Condition-Based Predictive Maintenance for Mission Critical Systems with Probabilistic Knowledge Graph and Deep Learning

New tools and technologies are needed for modern US Navy surface and aviation fleets to augment current onboard condition monitoring and maintenance processes and help improve mission-critical systems availability, increase operational readiness, and reduce life cycle costs. The state-of-the-art condition-based predictive maintenance (CBPM) applies analytics to predict failures and recommends service only when needed. CBPM is preferred for naval assets because it provides a window into the future of each asset’s predictive performance.In this effort, IFT proposes to develop an integrated approach that includes both data-driven and physics-based modeling techniques in order to build reliable diagnostic and prognostic models. The proposed CBPM system is based on the state-of-the-art knowledge graph and deep learning framework. To explore streaming data from multiple sources, the knowledge graph will be constructed from collections of unstructured and structured data with natural language processing to extract entities, relationships, and events between them. The proposed cognitive-based decision support system is to support the operator to combine data, identify potential failures rapidly, and provide timely recommended proactive maintenance actions with increased efficiency in logistics and supply chain. This is particularly important for mission-critical systems to support sustained combat operations and readiness with minimum costs and unplanned downtime.