Maintaining fleet readiness for the US Navy has always required high-level planning involving mountains of data of diverse types and quality. Unfortunately, data alone cannot point the way to a better approach to maintenance and sustainment. The ability to collect appropriate data sets, integrate and analyze them with incorporation of domain knowledge, and produce an executable action plan is what’s needed to make sustainment operations more effective.
The Department of Defense has long recognized that digital processing and data science holds the key to enabling a condition-based maintenance (CBM) strategy. Such a system would allow rapid assessment and more understanding of current and evolving equipment conditions. Furthermore, it would deliver faster and more accurate detection and correction of anomalies, service life predictions, usage driven equipment assessment, and performance risk management. The advent of digital twin technology has brought forth a highly effective CBM approach. It enables a virtual model of an asset to be fed real-time data from sensors on the physical asset, and to then use simulation, machine learning and reasoning to help predict how that asset will perform throughout its lifecycle and be impacted by hardware and operation changes.
A current CTMA project is working to develop just such a condition-based maintenance program. As a testbed, it has targeted the major components of the propulsion drivetrain for a Navy LSD 49. With this small, discrete set of components, the project’s industry partner, American Bureau of Shipping (ABS), is establishing best practices in data collection and pre-processing, modelling and demonstrating how a digital twin can be developed and implemented for machinery health monitoring. In addition, ABS has defined the key dimensions for building a digital twin for machinery equipment. This demonstration could then point the way for other ships in its class as well as other classes of Navy surface ships.
At the start of the project, the Navy provided to ABS its existing data sets for the LSD 49 propulsion drivetrain. This included transactional data (e.g., planned maintenance records, failure events, etc.) and sensor time-series data (e.g., the Navy’s integrated condition assessment system (ICAS) data, etc.). ABS used these datasets for developing a methodology to perform a thorough data quality assessment, models to detect incipient anomalies and framework for implementing a digital twin. They recommended using a combination of open source and commercial platforms such as the Microsoft Azure Government Cloud. It then demonstrated how proper methods of pre-processing the data can ensure data fidelity and completeness, which enables accurate follow-on analytics.
Finally, ABS developed a suite of algorithms to detect early indications of incipient failures within the drivetrain. This work has formed the foundation of a machinery digital twin. It required high-level analysis using physical knowledge about the drivetrain, input from sensors that detect anomalies, and machine learning to train the system on what to look for. A blueprint for further development and implementation of the machinery digital twin was delivered at the completion of the project.
“Working in collaboration with the Navy, we have identified key elements for building an accurate condition-based monitoring system that, when it’s complete, will allow for identifying predictive compliance risk factors, with real time detection of incipient anomalies that have the potential to cause an unplanned breakdown of a propulsion drivetrain,” says Subrat Nanda, of ABS. “Gaining detailed, actionable information ahead of time is everyone’s goal, as it could mean the difference between completing a scheduled mission and being stranded motionless at sea.”
Planning for new project phases involves exploring the feasibility of a shipboard digital twin deployment that enables automated condition-based maintenance activity scheduling.