Macsea Improves Navy Ship Reliability
In the popular motion picture The Matrix, Agent Smith is a software object sent out to monitor and correct problems with humans that the Matrix relies on for electric power. In real life, software developer Macsea has developed a software product called Dexter that allows its customers to build and deploy machinery-health-monitoring software agents designed to automate the majority of data gath...
In the popular motion picture The Matrix, Agent Smith is a software object sent out to monitor and correct problems with humans that the Matrix relies on for electric power. In real life, software developer Macsea has developed a software product called Dexter that allows its customers to build and deploy machinery-health-monitoring software agents designed to automate the majority of data gathering and analytics associated with condition-based maintenance (CBM).
These software agents are deployed in machinery data networks and work unobtrusively in the background, alerting maintenance and operations personnel when they detect an equipment problem or predict an impending problem. This allows expeditious isolation and repair of existing problems and, through predictive analytics, helps avoid problems that could degrade machinery reliability.
Military Sealift Command (MSC), the part of the U.S. Navy charged with providing ocean transportation for U.S. military forces worldwide, has been an early adopter of Macsea’s intelligent software agent technology to support cost-effective CBM on ships with few personnel. Dexter agents have been customized and deployed aboard MSC vessels for several years, primarily for diesel and gas turbine engine health monitoring.
Diagnostic agents interact through the local area network.
As part of the U.S. Navy, MSC provides strategic sealift and ocean transportation for U.S. military forces. Its mission is to transport equipment, fuel, supplies, and ammunition to sustain U.S. forces worldwide. MSC operates ships that provide combat logistics support to U.S. Navy ships at sea; special mission support to U.S. government agencies; pre-positioning of U.S. military supplies and equipment at sea; and ocean transport of DOD cargo. The agency operates more than 120 ships crewed by civilian mariners.
As the United vStates continues to reduce the size of its military presence overseas, military readiness and rapid response capabilities depend increasingly on MSC. It is a matter of national priority that MSC ships be available at a moment’s notice and perform reliably when called upon. Recent crises have reinforced MSC’s vital role as a major contributor in the execution of U.S. national strategy.
Low asset costs, high reliability
The Navy places strong emphasis on reducing the total lifecycle ownership costs of its assets. Studies indicate that costs for personnel and maintenance consume the largest part of the service’s operations and support budget and continue to grow most rapidly. Therefore, cost reduction efforts are focused on reducing personnel and implementing CBM on future Navy ships.
Cost/benefit analyses of CBM strategies for machinery-health monitoring in many industries have clearly established their value. An effective CBM program, however, can’t be implemented without expense. Machinery condition must be measured via sensor instrumentation and assessed using special analytical methods, which often require special expertise. Specific CBM work tasks must be diligently performed and can generate significant crew-workload requirements. There simply are not enough people to gather and analyze performance data on the multitude of machinery and equipment aboard Navy ships. This situation has caused a needs-gap within the Navy maintenance community.
Existing shipboard machinery automation systems are primarily designed for safety and protection. They also provide large amounts of data for equipment health monitoring. Transforming data into actionable information for effective CBM, however, remains an arduous task. The situation will be exacerbated on future all-electric Navy ships, as distributed, integrated power systems will include orders of magnitude more sensors than today’s ships and present increasingly complex maintenance requirements. Machinery performance monitoring and health assessment are areas where the exploitation of software-agent technology will yield substantial near-term economic benefits.
Failure prediction process mirrors the failure recovery process.
Software agents can be used to clone human intelligence, perform human-like reasoning, and interact with human clients. Agents can perform tedious, repetitive, time-consuming, and analytically complex tasks more accurately and reliably than humans. Software agents also can serve as expert assistants in monitoring, troubleshooting, and predicting failures in complex machinery processes. Imparting intelligent processing functions into software agents will allow the Navy to leverage valuable knowledge across a geographically distributed ship fleet. Agents can be distributed when and where needed to enhance fleet operations, performance, and readiness. Their intelligence can be upgraded remotely. Human-software agent teams can provide higher levels of platform readiness and reliability at far less cost than equivalent human-only resources.
Intelligent software agents automate the bulk of the work necessary to continuously monitor machinery health. They autonomously perform complex information processing tasks to identify impending failures and accurately predict remaining useful equipment life. Software agents can be deployed to monitor and analyze hundreds of thousands of data points automatically, while being integrated into existing automation system environments at relatively low cost.
“Division of labor for CMB tasks” graphic illustrates the division of labor between software agents and humans for typical equipment health monitoring tasks. The six main CBM processes include data acquisition, equipment performance analysis, condition assessment, fault diagnosis and isolation, problem verification, and maintenance/repair action. Agents automate the majority of these processes, reducing humans’ roles to fault verification and corrective action. Furthermore, these human activities only become necessary after a problem has been identified by an agent, resulting in very significant time savings for CBM implementation.
Dexter agents can be created for any type of machinery system for which sensor measurements are available. This technology allows companies that have already invested in plant automation systems and process control software to leverage these investments further with advanced agent-based analytics.
Software agents for CBM divide into two classes:
CBM labor is divided as shown.
Diagnostic agents perform real-time assessment of existing alarm conditions. Automatic fault diagnostics provide troubleshooting assistance to maintenance personnel, directing them to the most likely problems causing the alarms. The agents reduce troubleshooting time and help restore normal operations as quickly as possible, minimizing the cost of any process disruption. Diagnostic agents also log all diagnostic events, allowing a management review of equipment reliability over any operating time interval. One distinguishing diagnostic agent feature is real-time assessment of behavioral anomalies in plant machinery. Probabilistic assessment of equipment faults can be particularly useful to new maintenance personnel that may be unfamiliar with plant operations. Personnel job performance and, in turn, plant reliability, stands to benefit from the knowledge and experience of the team of experts that develop the diagnostic knowledge base. Their cumulative intelligence becomes embedded in the software agents for deployment throughout the fleet or worldwide plant facilities.
Prognostic agents predict machinery problems at their earliest stage of development. Evolving equipment problems can often be discovered by following degrading performance trends in historical data. Prognostic agents automatically perform statistical trending analysis to detect abnormal machinery performance, which is at the heart of effective CBM. Predictions of future machinery faults include estimated time to failure and can help determine when maintenance should be carried out. By predicting machinery problems before they occur, unexpected breakdowns can be avoided. In the absence of significant trends, equipment overhaul periods may be extended rationally, thereby eliminating unnecessary maintenance work. The ability to predict future maintenance requirements improves maintenance planning, cost management, and plant reliability. Maintenance and repair decisions can be tied to actual plant operating conditions based on the severity of degrading trends and predicted plant problems.
Neural network diagnostics
During the past several decades, scientists developed computer models of biological neural networks that can learn and perform brain-like functions. These models, referred to as artificial neural networks, are able to learn from examples and are particularly useful for certain tasks, such as pattern recognition. Dexter agents use neural networks for diagnostic and prognostic reasoning about machinery faults. The software agent’s neural network automatically learns to associate patterns of alarm conditions with machinery faults.
Missed diagnostic calls and false calls translate directly into added maintenance costs, either from unexpected machinery failures or unnecessary maintenance activities. A diagnostic system’s robustness, therefore, directly impacts maintenance expenditures, as well as equipment reliability. Dexter’s neural networks tolerate noisy or incomplete input patterns, making their diagnostics more robust than those developed from logic or rule-based approaches. Even if one or more symptoms are missing, Dexter still identifies the most probable faults based on all available evidence.
A diagnostic system that monitors machine reliability has to be extremely reliable. A typical shipboard environment is about as harsh as it gets for a computer system. Corrosion, high vibration, low power quality, frequent power interruptions, high temperature, dust and dirt are some of the factors that can destroy most electronic equipment. After investigating commercial industrial grade computing platforms and Macsea selected the Advantech IPC 6806 series computer to host its shipboard diagnostic software, now used on more than 40 ship installations. “The Advantech IPC computers have worked reliably on some ships for over ten years without a problem,” reports Robin Osmer, Macsea field service technical manager. “The locations where we install them aren’t the best for computer equipment, which is, in most cases, a small enclosed area inside a control console.” Reliability is good even with poor air flow and high temperatures, Osmer says.
“Distributed machinery diagnostic agents” illustrates a typical shipboard configuration for deploying Dexter machinery diagnostics. The Dexter Server (IPC 6806) interfaces with existing machinery alarm monitoring and control systems through a serial or Ethernet connection. This allows real-time machinery performance data to be acquired, archived, and analyzed by the diagnostic software agents. The server is also connected to the ship’s high-speed, fiber-optic LAN, which serves as the ship’s data highway. Diagnostic workstations running Dexter client software are distributed throughout the ship and allow engineers real-time access to plant data and diagnostic-agent results. By transmitting real-time machinery performance data across the LAN, Dexter provides a very convenient and inexpensive means for engineers to keep close watch on machinery from the comfort of their staterooms, which are typically several decks away from the control and engine rooms. When an abnormal event occurs in the middle of the night, they can check it out a few steps away on their workstation, instead of having to walk down flights of stairs to the control room.
On one particular class of Dexter-installed ships, MSC engineers wanted to enhance monitoring of the main gas turbine engines by measuring inlet air-filter differential pressure. Inlet air has a large impact on gas turbine performance, but was not made part of the normal automation system during ship construction. After receiving a high cost estimate from a control system vendor, MSC approached Macsea, which provided a simple design composed of an Adam 4017 8-channel analog input module, small power supply, and enclosure. The system was subsequently installed and tested aboard two gas turbine-powered ships. The module’s signals were fed into the Dexter Server for display and analysis, an economical solution being installed on two more MSC gas turbine ships.
Kevin Logan is president, Macsea, and Chuck Harrell is a manager at Advantech eAutomation Group.
Case Study Database
Get more exposure for your case study by uploading it to the Control Engineering case study database, where end-users can identify relevant solutions and explore what the experts are doing to effectively implement a variety of technology and productivity related projects.
These case studies provide examples of how knowledgeable solution providers have used technology, processes and people to create effective and successful implementations in real-world situations. Case studies can be completed by filling out a simple online form where you can outline the project title, abstract, and full story in 1500 words or less; upload photos, videos and a logo.
Click here to visit the Case Study Database and upload your case study.