Historian Holds Key to Performance

Originally configured to gather a few hundred critical parameters across a 1,000-acre plant and display results on seven operator interface terminals, Aughinish Alumina's OSIsoft PI (pronounced pie) historian software now serves as the operational data repository for more than 25,000 measurements; interfaces with other applications; and provides results to operators, management, and laboratory ...

By Dave Harrold May 1, 2004


Data historian software

Enterprise integration

Key performance indicators

DCS and expert systems

Sidebars: About the PI data historian system How PI manages records, storage

Originally configured to gather a few hundred critical parameters across a 1,000-acre plant and display results on seven operator interface terminals, Aughinish Alumina’s OSIsoft PI (pronounced pie) historian software now serves as the operational data repository for more than 25,000 measurements; interfaces with other applications; and provides results to operators, management, and laboratory personnel on more than 200 user interfaces.

‘We first became aware of OSIsoft products during a new distributed control system [DCS] evaluation and selection project we were conducting in the late 1980s,’ says Aughinish Alumina’s director of IT, Justus de Hooge.

‘At that time, DCS suppliers offered their own limited functionality data historians. A couple of the DCS suppliers suggested we use OSIsoft and after examining the product and company, we decided to standardize on its PI data historian software, mainly because, at that time, it was able to connect to every new and legacy control system then on the market. We’ve since learned to appreciate the importance of having a single historian that’s independent of the control system manufacturer, and our engineers like only having to learn one interface, regardless of the current hardware platform,’ adds de Hooge.

Aughinish’s selection committee reviewed several data historians, specifically evaluating the manufacturer and its system’s ability to:

Support graphic and character terminals;

Interface with multiple DCS and programmable logic controller manufacturers;

Integrate with other business systems; and

Provide a long-term support relationship.

Producing aluminum year round

Aluminum (Al) is the most plentiful metallic element in the earth’s crust. First isolated in 1829 from aluminum chloride, aluminum started being commercially produced in 1886.

A subsidiary of Swiss-based natural resources group Glencore, Aughinish Alumina Ltd. is an alumina refinery situated on Aughinish Island on the south side of the Shannon estuary between Askeaton and Foynes, 20 miles downstream from Limerick City, Ireland.

Aughinish Alumina uses the Bayer process to treat bauxite ore to produce more than 1.5-million tons of alumina (Al203) per year. Alumina is a fine white granular powder that is exported to smelters in the UK, Scandinavia, and other European countries for further processing, eventually becoming aluminum metal. Bauxite is received via 65,000-ton bulk ore carriers from bauxite mines in West Africa, Brazil, Ghana, and Australia and is unloaded at the Aughinish deep-water marine terminal on the Shannon estuary.

Two tons of bauxite yield approximately one ton of alumina. The plant keeps more than 56 million gallons (250,000 m3) of process solution circulating through tanks, pressure vessels, and pipes. (See ‘Alumina refinery inputs and outputs’ diagram above.)

The Aughinish plant employs 450 persons and operates 24 hours per day, 365 days a year.

Business excellence, KPIs

Aughinish Alumina relies on the careful establishment, tracking, and reporting of key business process improvement indicators to ensure it is operating a stable, efficient, and profitable business. Within plant operations, this translates into appropriate key performance indicators (KPIs) that are displayed on digital dashboards using Microsoft Sharepoint portal technology as the user interface.

Because many KPIs are more complex than a single process measurement, Aughinish uses the PI-database and PI-advanced computing engine (PI-ACE) as the repository for calculating and assembling KPI data. Using a combination of software tools from Gensym, OSIsoft, and those developed in-house, Aughinish operators can accurately interpret digital dashboard content, make appropriate control adjustments, and communicate when and why process and equipment anomalies occurred. (See ‘Aughinish Alumina’s PI System Architecture’ diagram.)

For example, when equipment is intentionally stopped or stops on its own, the PI historian knows when it happens, but doesn’t know why. To complete the stopped equipment information record, operators select from a pull-down list of reasons why the equipment is stopped. Attaching consistent textual information to the equipment record gives engineers and maintenance personnel an easy and quick means of assembling related information for more accurate analysis. Also, Aughinish’s maintenance department uses actual running hours to determine maintenance schedules.

Another example of PI integration is with non-linear, multi-variable model-predictive controls implemented in the Honeywell TDC 3000 DCS.

Model inputs and outputs are read from and written to the PI database. However, because of the plant’s complexity, Gensym’s G2 expert system assists operators in understanding model predictions and plant behavior.

‘We intentionally chose not to close the loop on all the models because we’ve learned that if operators don’t stay directly involved in making control system decisions, their skills tend to diminish over time. Then, on those occasions when they need to be involved, they simply aren’t as comfortable as they would like or need to be. By keeping the loop open and providing expert system advice, we’ve found our operators’ skills remain quite good,’ says control system leader, George Troy.

‘We have over 150 people per day, working in ‘leaderless’ teams, relying on digital dashboards and KPIs to keep them focused on what’s important. Only a robust data historian, such as the PI system, is capable of supporting our ongoing development and deployment of KPI indicators designed to keep our business on the straight and narrow. This is how Aughinish is able to generate additional efficiencies and profits without increasing head count,’ adds de Hooge.

Lasting relationships, security

OSIsoft’s founder, J. Patrick Kennedy, is a chemical engineer and a registered professional engineer in control system engineering who was once an end-user working in the process industry. He understands the importance of having a focus on a specific market need, and uses this knowledge to direct OSIsoft’s product development and way of doing business.

Both de Hooge and Troy attend a number of user group meetings each year. They say they appreciate interaction of software developers and users at OSIsoft meetings, and like OSIsoft customer service. A call can result in a conversation with a software developer within minutes, Troy said. Aughinish’s initial (1989) PI system computer was a DEC Vax 6310 running the Open VMS operating system. Today Aughinish is running PI on a Microsoft Windows 2000 Intel Xeon-based server. ‘On the old [VMS operating system] system, PI was down for patches about once every six months. Now we have to take PI down every month to install patches and critical updates to the Microsoft Windows operating system. As this system is the main process interface for all users, its connection to the business network includes robust security and protection. Our need to install patches and upgrades to OSIsoft’s software remains about the same as our original system [about twice a year] and OSIsoft continues to do a good job of testing to ensure installation goes smoothly and fixes are correct the first time,’ says de Hooge.

More to come: models, integration

One test of a software application’s value is how much the software users are willing to push the software’s performance and capabilities envelope.

Many of us tend to learn those parts of a software application we need to get a job done, and are seldom willing to invest additional time and energy to become truly proficient in using the software.

Aughinish Alumina’s PI users seem an exception by aggressively expanding and extending use of:

PI DataLink to populate Microsoft Excel spreadsheets;

PI ActiveView to provide secure Web views of PI ProcessBook screens;

Gensym G2 expert system information with the PI database;

Additional DCS-based predictive models with G2 advisories; and

In-house developed .Net-based solutions that integrate PI into the business systems.

‘When we first committed to use the PI data historian, we grossly underestimated its importance and positive impact to plant operations. We originally sent four people to training—two to become system managers and two to become our onsite trainers. We thought seven user interface terminals would be all we would ever want or need. Within the first six months, we doubled the number of user interfaces and now have over 200 people trained to use PI. We aren’t really sure how many there will be once we extend use of PI portal displays across the enterprise, but we do know we can’t run this plant very efficiently for very long without it,’ says de Hooge.

About the PI data historian system

OSIsoft’s PI data historian is a suite of integrated applications designed to meet the needs of continuous and batch applications in process industries.

The PI suite includes the following.

Universal Data Server (UDS) is the heart of the PI System, acquiring and routing real-time data throughout the PI information infrastructure.

PI ActiveView forms the Internet portion of a data delivery system and gives users ability to view full function, interactive PI ProcessBook displays on the Web.

PI BatchView provides desktop viewing of historical and real-time batch trends and permits comparing multiple batches in a single view.

PI ControlMonitor ensures that control systems are functioning properly by monitoring and displaying quality variables as dependent variables, and the control loop’s setpoint, measured variable, and control output as independent variables.

PI DataAccess Package is a set of tools that helps advanced users review, structure, and access data within the PI System. For example, the software development kit provides ActiveX programming tools for accessing PI servers and related subsystems.

PI DataLink creates a bi-directional connection between spreadsheet programs and the UDS, providing easy access to process data and the analysis tools of Microsoft Excel or Lotus 1-2-3.

PI ProcessBook is the graphical display front-end of the PI system. It helps transform raw data into useful, dynamic, real-time displays available from any desktop PC or via an Internet browser using PI ActiveView.

PI ACE is an advanced computing engine designed to take the guesswork out of building calculations, allowing users to focus on turning real-time data into useful information. ACE equations are written with a minimal amount of code and are transportable across UDS installations.

For more information about PI data historian, visit

How PI manages records, storage

PI DataStorage, working with the Universal Data Server, handles the collection, storage, and retrieval of numerical and string data. It uses a ‘swinging-door compression’ process that discards data that fall within an acceptable range of user defined values, or within a ‘compression deviation blanket,’ thus eliminating gathering and storing like values.

The compression deviation blanket is a parallelogram extending between the last recorded value and the new value with a width equal to twice the compression deviation specification.

When a new data point is received, the previous data are only recorded and held in DataStorage if they fall outside the compression deviation blanket. This permits PI Data-Storage to efficiently accommodate small and large amounts of data, at its original resolution, for extended periods of time.