Predictive maintenance for gas pipeline compressors
In 2010, Houston-based Columbia Pipeline Group (CPG) had a compression asset failure that interrupted service and had the potential to create customer dissatisfaction. CPG owns 15,000 miles of interstate pipeline across 16 states. They operate more than 100 stations with approximately 1.1 million horsepower of compression, delivering approximately 1.3 tcf of natural gas per year. The company needed a swift, reactive approach to resolve the problem.
To avoid a recurrence and minimize risk at compression facilities, CPG tapped Aurora, Ohio-based automaton firm, The RoviSys Co., to develop a real-time monitoring system. The new enterprise analytics program uses SharePoint, PI, and SQL technologies to analyze data and drive proactive and corrective actions. Its objectives are to avoid a facility shutdown and to minimize failures in compressor stations.
A new problem
As new sources of natural gas from the Marcellus and Utica shale formations in Pennsylvania, West Virginia, Ohio, and New York have made transporting gas a priority, demand to transport gas from its source to delivery points throughout the country adds flow-control stress and the potential for malfunction.
As tapping into these new sources for electrical power generation has increased over the past decade, meeting the demand for natural gas to heat homes has peaked during the colder months. Recently, demand has increased in the summer months to generate electricity for cooling due to the low price of using natural gas to generate electricity and the retirement of coal-fired power plants. When summer demand was lower, there was a large window for planned downtime and preventive maintenance (PM). Now that the demand is year round, the availability for planned downtime is drastically reduced.
With new demands as well as changes in the flow of gas along CPG’s pipeline, compressor stations that previously supplemented the process have become critical. Columbia needed to maximize throughput and reduce unplanned downtime of its compressors to keep up with the increasing market demands-especially in places where the compressor station is critical to transporting gas to key customers.
Preventive vs. predictive maintenance
Most major compressor units have alarms or fault conditions that protect the equipment from damage. At that point, it’s usually too late. The unit is already out of service and operations personnel must react to the situation. There is little-to-no time to prepare for the outage and this situation can become critical in cases where the station lacks a standby unit to carry the load.
Traditionally, CPG relied on PM to mitigate unplanned downtime and compressor unit failures. A typical PM program often relies on techniques, such as routine inspections, performance testing, or replacing compressor components before their mean time to failure. In many cases, these techniques require an outage of the unit.
Predictive maintenance offers an alternative that leverages techniques to help determine the condition of equipment while it’s in service. This offers operations and mechanics the option to choose when maintenance should be performed. Often, it’s beneficial to take a compressor out of service either before or after short-term peak demands, such as a period of very cold weather. CPG decided to leverage its knowledge about the expected performance of its compressors to create what they call "enterprise analytics," which uses predictive analytics for making data-driven decisions to improve the reliability and availability of the compressors in their fleet.
Initially, CPG brought in RoviSys for the automation company’s knowledge in implementing large-scale, redundant PI architectures and for its relationship with OSIsoft. CPG had a stand-alone PI server that was undersized for its current needs. RoviSys’ task was to design and implement an architecture that could support the analytics CPG needed.
The RoviSys engineering team set out to design a system with the following criteria:
- A redundant architecture was a priority, given that data-driven decisions require a system that’s available 24/7 with as little downtime as possible.
- The data in the system must be available to all authorized members of the organization, such as operators at a compressor station and the executive staff at corporate headquarters. The system must be scalable, because more than 1,000 users have access to the data. The data must be presented in formats compatible with a broad audience.
- CPG’s prior PI system lacked the tools necessary to do predictive analytics, present information in the form of a dashboard, and notify analysts of issues that require attention. The architecture needed significant expansion to support the desired tools.
- CPG wanted more than 10 years of valuable data transferred to the new architecture so the company could add more data points for analytics. CPG collected data at compressor stations and other operational points using dozens of interfaces to PI that needed to be migrated as well.
No process interruption
Upon implementation, the process for cutting over from the old PI system to the new system was seamless. Service to end users was not disrupted and data were available during the entire cutover. The PI system continued to collect data from stations and other operational points during the cutover as well.
The PI system architecture went from a single server with two types of clients to a multiple server environment with several applications, data storage, Web, and analytic servers running a vast majority of the available OSIsoft PI software tools. This setup provided a place for analytics to run, tens of thousands of data points to be organized by physical asset, analysts to receive e-mail notifications, and users to view data in a Web portal. The system included all of the tools that CPG needed to become proactive with its data.
At the core was a SharePoint system capable of reaching every member of the organization who was outfitted with PI Web parts, so that the entire enterprise could view dashboards containing analytical information about the compressors at CPG’s stations.
Also, two additional PI environments were designed and implemented: one for CPG to develop new analytic concepts and the other for early adopters to test those concepts before deploying the technology to the entire enterprise. These environments allow CPG to grow its ideas without disrupting day-to-day operations.
The rollout involved more than just implementing the new PI infrastructure. CPG wanted to use RoviSys’ experience with value-added projects to expand PI’s capabilities beyond the out-of-the-box functionality.
Building the analytical dashboards was a cooperative effort that coupled RoviSys’ technical knowledge of PI with CPG engineers having detailed knowledge of compressor equipment and its history of issues they hoped to identify with predictive analytics. Results of this effort include:
- RoviSys integrated all of the PI components so that CPG had a single experience through a set of simple-to-use dashboards, created custom interfaces between components of the systems, and created tools that extended the functionality of the off-the-shelf PI components to meet the specific needs of predictive analytics and customer-use cases.
- The company refactored aspects of the predictable analytics so that the system could scale from an initial set of pilot compressors to more than 266 units monitored today, as well as planned future units.
- It implemented a software development process at CPG for developing, testing, and deploying dashboards and notifications for more than 266 compressors of different types (reciprocating, centrifugal, and electric).
- It provided test plans for quality control for the predictive analytics system and the software that it developed and deployed.
- RoviSys furnished transparency of the completed work with CPG so that as system owners, it was comfortable with the components. This was accomplished through strong working relationships and documented processes.
The use of new technology has prevented recurrences of the incidents that almost caused compressor failures at the same station. Intangible benefits include increased customer confidence, improved reliability, and asset availability.
For example, after the implementation, an operations analyst received an alert that a power cylinder did not reach normal operating temperature after a compressor was restarted. After further investigation, the analyst noticed that there was a 500 F difference between the exhaust temperate for cylinder 1 and the other cylinders. The history for the unit showed that cylinder 1 typically operated near the same temperature as the other cylinders and that the unit was currently operating at 76% of its rated brake horsepower load. This information prompted the analyst to contact the team at the station. Conditions on the pipeline demanded that the unit continue to run to meet a critical need for storage injections. After pipeline conditions improved, the gas control team approved taking the unit offline. The team at the station found a bad fuel valve, and promptly replaced it. The unit was returned to service allowing gas control to resume injections into the storage field.
In another situation, just prior to the cold season, an operations analyst received an alert that one of the compressors had a 400% increase in vibration. During the investigation, the compression engineer and equipment analyst noted that the turbocharger for this unit was recently overhauled during the off-peak season. The operations analyst dug deeper into the data and discovered that the turbo oil pressure was lower than usual. This prompted the team at the station to take a closer look at the vibration components on the turbocharger. The equipment analyst connected a vibration analyzer to the turbocharger and discovered a balancing problem. The turbocharger was removed and sent back to the company that performed the overhaul.
– Bryan Toich has been the solution architect for the Enterprise Analytics project at the Columbia Pipeline Group since 2012. He is a lead software engineer with experience designing and implementing software solutions for natural gas pipelines. He also provides input to project planning activities and business case development for his customers. Edited by Eric R. Eissler, editor-in-chief, Oil & Gas Engineering, firstname.lastname@example.org.