A credible Pacific deterrence posture for the U.S. Navy requires that the fleet of ships, submarines, and aircraft be available to the combatant commander at a rate that outpaces potential adversaries, in order to maintain control of strategic geographic areas and vital supply chains. A new, AI-enabled approach to predictive maintenance can help achieve this goal, and increase operational availability across the INDOPACOM AOR and elsewhere.
With this approach, AI looks for patterns in vast amounts of maintenance sensor data to predict when parts or systems might fail— and can often find potential problems long before they show up on watchstanders’ consoles. At the same time, the AI helps supply-chain personnel deliver the necessary parts and repair crews with just-in-time logistics. These two components—diagnostic and supply chain— together make up what is known as AI-enabled predictive maintenance.
One way that AI-enabled predictive maintenance helps keep Naval forces forward deployed is by lowering the risk that a key propulsion, weapon or other system will fail during operations, potentially taking the vessel or aircraft out of action. It also reduces the need to bring ships and submarines into port for lengthy planned-maintenance work.
AI-enabled predictive maintenance is not so much a revolution as an evolution, building on the Navy’s rapid progress in sensor technologies, advanced analytics, secure satellite communications, cloud computing and a host of other areas.
PREDICTIVE DIAGNOSTIC ENGINEERING
A key aspect of the process is predictive diagnostic engineering. Currently, sensors on propulsion, auxiliary and combat systems feed data to watchstanders’ consoles, prompting alerts whenever readings, such as engine speeds or fuel-oil temperatures, exceed safe operating limits. Predictive diagnostic engineering—which can be conducted either onboard or through a common data network—brings together and analyzes such sensor data from across the Navy. It looks not just at a fuel pump on a single ship, for example, but at all similar fuel pumps currently or formerly in use across a ship class or fleetwide. What emerges in the data is a predictable pattern of decay—essentially, the normal lifecycle of that type of pump.
The AI then compares the data from an individual ship with the overall patterns, looking for anomalies. It may find, for example, that the decay pattern of a particular fuel pump is moving much faster than might be expected—even though the sensor readings on the consoles aren’t yet changing. The AI might also look at what happened to other fuel pumps with similarly accelerated decays, to provide an estimate of when the fuel pump in question will ultimately fail.
In addition to the maintenance-sensor data, the AI brings in contextual data to provide a higher fidelity estimate. It might look at atmospheric conditions affecting the ship, such as temperature and humidity, and evaluate how those conditions have historically sped up or slowed down decay patterns. The AI might also consider a ship’s maintenance records—factoring in, for example, repairs previously made to the fuel-oil system, and the historical impact of those repairs on similar fuel-oil systems.
A SECURE COMMON DATA NETWORK
To determine the larger data patterns of parts and systems, predictive diagnostic engineering brings together data from across the Navy through a common network. Maintenance and other data is transmitted from ships, submarines and aircraft via satellite to the network, and then integrated with historic data. Thanks to advances in cybersecurity, this data transmission can be done securely, using the same protocols now in place for communications, navigation, logistics, and other types of data.
The network is designed with open frameworks and other architectures, making it vendor-agnostic and able to accept data from any of the Navy’s different types of propulsion, auxiliary and combat systems. This ability to bring data together is critical, because the more maintenance data that is collected across the Navy, the more accurate the AI becomes. Data transmitted from a ship not only helps diagnose specific problems on that ship, it also adds to the larger pool of data about those systems—which in turn helps the AI to better diagnose problems on other ships.
When the AI predicts that parts or systems are heading toward failure, it identifies what maintenance and repairs will be needed, and when.
For example, by looking at the pool of data on a particular type of engine— including problems and repair histories—the AI can determine which actions, taken at which times, have proven most effective in keeping the engine operational.
Once the AI has identified a potential failure, it can help get parts—and if necessary, specialized maintenance crews—to the ship or submarine in time for repairs. The AI can look across the entire supply chain, pinpointing where the parts and maintenance crews are, when they can become available, and how they can best get to the vessel.
The AI does this by analyzing a wide range of databases related to Navy supply chains and logistics. In some cases, the AI may recommend sending the parts and crews to a forward port that the ship or submarine is expected to visit, while in more urgent cases the AI may recommend delivering the parts and crews to a certain location at sea.
By running simulations, the AI works out the logistics of getting the resources where they the need to be, and at the optimal time. The AI can also put in place alternative plans if conditions change, for example if it detects that the decay pattern of a part is suddenly accelerating, or if a forward port is no longer available.
STRENGTHENING PACIFIC DETERRENCE
AI-enabled predictive maintenance is not a single, overarching system, but rather a system of systems that integrates many of the advanced technologies the Navy is currently developing.
These include machine learning and other forms of AI, as well as open architectures and other technologies that make it possible to analyze large amounts of disparate data. In addition, new sensor technologies, data links, and communications networks are enabling increasingly sophisticated diagnostic engineering, and the Navy’s advances in cyber and electronic warfare are making the transmission and storage of maintenance data more secure.
The Navy now has an opportunity to bring these and other capabilities together to strengthen deterrence activities in the Pacific, by increasing the operational availability of forward deployed ships, submarines and aircraft.
CAPTAIN STEVE SOULES
[email protected], a Booz Allen executive vice president who leads the firm’s Joint Combatant Command account, retired from the U.S Navy after serving 27 years, including in six Western Pacific/Indian Ocean deployments.
CAPTAIN JEFF JAMES
[email protected], a retired Surface Warfare Officer whose commands included the USS PIONEER (MCM 9), USS HOPPER (DDG 70), and Joint Base Pearl Harbor-Hickam, leads Booz Allen’s infrastructure, energy, and environmental business across the PACRIM, delivering technical solutions leveraging AI and ML to Navy, Marine Corps, Air Force, and Joint clients in the region.
[email protected], leads Booz Allen’s develop- ment of AI-enabled predictive maintenance and supply-chain capabilities for clients throughout the DoD and other federal agencies.
AARON VAN BLARCOM
[email protected], a solution architect on Booz Allen’s PACRIM analytics team in Hawaii, develops a broad range of AI and machine learning solutions for Navy clients.