New Navy and Coast Guard vessels have sensors monitoring every pump, engine cylinder, scuttle, and generator winding. Thousands of sensors feed one computer that aggregates data, verifies operation against set parameters, and prioritizes fault indicators to alert watchstanders. While this has worked reasonably well and allowed new vessels, such as the Coast Guard national security cutter, to significantly reduce manning, these ships are expensive to build. In addition, machinery control software (MCS) frequently has flaws that can be dangerous and occasionally requires maintenance procedures, making crews rely on increasingly complex data acquisition and processing units. This frustrates engineering departments, which are undermanned and overly reliant on “black box” machines. Yet, while it sounds counterintuitive, more technology can solve this problem.
Current MCS systems are marginalizing afloat engineers, while presenting redundancy issues with machinery control and increasing vessel construction costs. Today, MCS is employed in a most limited sense—a computer screen using a simple color schematic that indicates which equipment is running. It can show the engineer on duty a tank’s levels or if a fan is on. In some cases, it can even energize and secure equipment.
This all happens with a data network—thousands of sensors feeding central data acquisition and processing units. However, data is frequently discarded or copied onto a hard drive and then ignored. The amount of data is hard to overstate—terabytes from a single deployment detailing the entire machinery plant’s status, fluid pressures, engine pyrometer readings, and lube oil and jacket water temperatures. This data is a gold mine for identifying equipment degradation and could be used to flag which pieces of equipment require maintenance.1 For example, instead of cleaning a random portion of the heating, ventilation, and air conditioning (HVAC) system every 18 months and paying a contractor thousands of dollars to wipe down clean ventilation ducts, engineers could analyze the recorded data and identify which portion of the ventilation system needs cleaning. Indeed, targeted maintenance, also known as condition-based maintenance (CBM), can save tens of millions of dollars across the Sea Services each year.2
CBM and Smart Data Collection
Current equipment maintenance schedules frequently call for preventive maintenance, in which valuable man hours, consumables, and spare parts are used to prevent failures. The Sea Services have favored this conservative method of maintenance for decades, as it incurs the lowest risk of equipment failure and is more cost-effective than casualty response.3 However, the massive amount of equipment data now collected could enable a CBM regimen that would optimally allocate maintenance resources to equipment that requires it.
Preventive maintenance would not need to be discarded in all cases, as there are plenty of scenarios in which there will be no historical data to properly inform decisions. But in cases in which CBM can be implemented, the benefits far outweigh the risks. With CBM, instead of greasing the same zerk fitting, rinsing already clean filters, or replacing perfectly good transmission belts, engineers do work that translates directly to vessel readiness.
In conjunction with driving down maintenance costs through CBM, the Navy and Coast Guard also should investigate inexpensive alternatives to putting a sensor on everything. A well-researched concept gaining traction is the nonintrusive load monitor (NILM). A NILM is a device that monitors an aggregate electrical power stream at a central location on a ship’s microgrid. Because of the high sampling frequency (8 KHz) and machine-learning techniques, the NILM can detect on/off signatures of equipment on the grid, and then feed MCS with equipment status.4 A NILM provides two major advantages over a fleet of sensors—cost and functionality.
A NILM uses three sensors and a voltage tap, for a total cost on the order of thousands of dollars for the entire system.5 This includes a processing unit, graphical user interface, and data acquisition unit, among other features.
Putting sensors on every electrically powered piece of equipment to determine status and feed the main MCS costs hundreds of thousands of dollars, not including replacement sensors or backup servers. The NILM advantage is compounded when it comes to serviceability and redundancy. A fleet of sensors is susceptible to slow failure over time as sensors reach end of life or get damaged. The system slowly degrades, requiring more maintenance work. A NILM is centrally located and not susceptible to the same degrading pressures.6 This increases robustness at lower cost.
Second, and more important, is the added functionality of a NILM. Because of its ultra-high sampling frequency, a NILM can detect, identify, and catalogue equipment electrical signatures. These signatures are unique identifiers that alert a NILM to specific equipment operation in the aggregate power stream.7 As equipment operation changes mechanically, it frequently manifests itself as a shift in an electrical signature, enabling a NILM to detect soft faults.8 One example is a motor pump coupled with a flexible Lovejoy coupling; as the coupling wears, becoming stiffer, the turn-on electrical signature of the motor changes slightly, manifesting as a frequency bubble.9 A NILM can detect such events, alerting watchstanders to degraded couplings and enabling maintainers to service equipment before catastrophic failures occur.10
In another example, as HVAC systems are run, a NILM can monitor the fan motor current using harmonic content to determine the fan’s speed.11 Knowing the fan’s speed coupled with historic fan speed operation rates allows watchstanders to determine which ventilation ducts require cleaning. Similarly, rotor slot harmonics in the current data can be used as tachometers on the ship’s induction motors, potentially allowing watchstanders to monitor every motor speed in real time from a single sensor.
These two examples have been proven to work and demonstrate the untapped potential of advanced signals processing applied to equipment electrical signatures. A NILM, when paired with a traditional MCS system, can drastically improve a vessel’s maintenance regimen, shifting scarce resources from preventive maintenance to targeted CBM. This could enable ships to operate with smaller crews by increasing the automation.
As vessel construction trends continue to favor automation, it makes sense to explore alternatives to traditional machinery control and monitoring software. NILM systems could lower automation costs while adding significant functionality through soft-fault detection. In addition to exploring NILM technology as a secondary MCS system, the Navy and Coast Guard should leverage the collection and analysis of data by shoreside engineering support to best develop an optimized and targeted maintenance plan. This could save the Sea Services money and address some current MCS shortfalls.
1. James Paris, John S. Donnal, and Steven B. Leeb, “NilmDB: The Non-Intrusive Load Monitor Database,” IEEE Transactions on Smart Grid 5, no. 5 (2014): 24, 59–67.
2. Congressional Budget Office, “Comparing a 355 Fleet Ship with Smaller Naval Forces,” CBO, March 2018.
3. Stuart Smith, “The Advantages of Preventative Maintenance,” Transcendent Magazine, 26 September 2012.
4. Daisy Green, et al, “Dashboard: Nonintrusive Electromechanical Fault Detection and Diagnostics,” 2019 IEEE AUTOTESTCON, August 2019, 1–9.
5. Stephen Kidwell et al, “NILM Dashboard: Power Monitoring for Condition-Based Maintenance,” American Society of Naval Engineers Technology, Systems, and Ships Symposium 2019.
6. Joshua C. Nation et al, “Nonintrusive Monitoring for Shipboard Fault Detection,” in Sensors Applications Symposium, IEEE (2017): 1–5.
7. Green, et al., “Dashboard: Nonintrusive Electromechanical Fault Detection and Diagnostics.”
8. Kidwell et al., “NILM Dashboard: Power Monitoring for Condition-Based Maintenance.”
9. CDR Thomas DeNucci et al., “Diagnostic Indicators for Shipboard Systems Using Non-Intrusive Load Monitoring,” IEEE Electric Ship Technologies Symposium (Philadelphia, PA: 2005), 413–20.
10. Nation et al., “Nonintrusive Monitoring for Shipboard Fault Detection.”
11. Kayhun Lee, Lukasz Huchel, Daisy H. Green, and Steven B. Leeb, “Automatic Power Frequency Rejection Instrumentation for Nonintrusive Frequency Signature Tracking,” IEEE Transactions on Instrumentation and Measurement 70 (2021): 1–11.
