Case Study: Knowledge Fusion

Case Study: Knowledge Fusion

For most companies, improper machine maintenance is probably a matter of lost profits: A little downtime, a screwed-up order, an unhappy customer. The Tennessee Valley Authority's nuclear division must play for higher stakes, however. For them, maintenance could be a matter of life and death.

That's why the TVA, the largest public power supplier in the US, overhauled its nuclear division's maintenance management system, creating a centralized knowledge management system that cuts time and errors out of the cycle.

Serving nearly 8 million customers, TVA uses a mixture of fossil, hydroelectric and nuclear energy. TVA Nuclear's three facilities represent about 20 per cent of TVA's generating capacity. After analyzing TVA's major business processes, a group of TVAN senior executives targeted the maintenance management process at TVAN as a prime candidate for improvement. For one thing, the tools currently used were not Y2K compliant; for another, maintenance work relies heavily on documents such as vendor manuals, drawings and formally regulated work instructions.

"It's document management, which requires so much paper, that it's always a target," says Karl Singer, senior vice president of nuclear operations at TVA. TVAN estimates that nearly US$48.7 million was spent annually generating, planning and performing maintenance work orders in this heavily regulated and safety conscious industry. The executives tapped individuals from various groups at the Browns Ferry Nuclear Plant to develop and pilot the program, with input from the folks at the two other nuclear plants in the TVA network. The group started by analyzing the current state of affairs, looking for places to consolidate work and deciding what performance metrics they most wanted to improve, says Robert Rupe, the process improvement manager at TVAN and the person responsible for coordinating TVAN's various process improvement efforts. Since more than 14,000 work orders are written annually at Browns Ferry alone, there was room for significant savings through productivity improvements. They then charted the existing work-order and procedures-management processes, conducting more than 350 hours of interviews with plant employees and also benchmarking the maintenance processes of 15 utilities and six other companies.

One of the big "aha!'s" says Rupe, was just how closely the planning and execution of maintenance work orders was linked to procedures management and document workflow. For example, a typical handwritten work order came attached to various paper copies of machine diagrams, documentation and procedural instructions. But the work orders lagged behind document and procedural revisions. "Most of our work is planned two months ahead of time," says Rupe, and some orders are planned six months to a year in advance. Meanwhile, "We're constantly getting new procedural updates from the outside, so stuff gets revised fairly frequently," he says.

After six months of full-time work by seven people, the group produced a work process that combined maintenance order workflow with procedural document management. They built two integrated systems that would support the new processes by linking procedural documents electronically to the work management system. That process would eliminate five or six separate software packages that did not speak to each other or share a common interface.

There was no off-the-shelf software available, so the team partnered with a vendor (The System Works) in an arrangement that gave process requirements to the vendor, who, in turn, built a commercial product around those specifications. In essence, says Rupe, "We helped define a new off-the-shelf product."

The resultant new system took 28 months from first brainstorm to rollout, says Rupe. It gives workers desktop access to documentation on company assets, records, equipment and parts information, and work instructions for all three sites. Now, a work order is generated electronically, routed for approval and attached, also electronically, to the most recent batch of pertinent drawings and documentation.

Moreover, the electronic documents are indexed down to each piece of equipment in the plant. All three nuclear plants are now standardized on the systems, and the maintenance groups can draw on the repository of previously completed work orders to analyse and manage future activity more effectively.


After spending so much time on the analytical and design side, Rupe says, "We butchered change management." The worst mistake was grossly underestimating the computer literacy of the workforce, he says. There were people who'd never used a keyboard, and people who thought the CD-ROM drive was a coffee cup holder. The lesson that TVAN took away, says Rupe, was twofold.

First, the rollout team needed to get basic computer skills up to a standard level. But more important, the group needed to couch the new system in terms of how it improved the overall process rather than what each individual would see. "Because we went to distributed computers, some of the response times were actually slower, and people saw that. But we didn't explain that the overall process would result in huge time savings. We needed to manage expectations better."

The system wasn't cheap: It cost about US$5.1 million initially, and more upgrades are planned. But the savings are equally impressive: The time it took to process a work order dropped from 39.8 to 23.3 person-hours per order, and TVAN saved an estimated $8.4 million annually in labor costs. And, Rupe says, in an industry driven by cost savings, that's a big payback.

More important, however, the system helps maintenance workers learn from captured data. It might be something as mundane as noticing that the label on a valve doesn't match the label on a drawing; but in the end, such a mundane thing can add up to safety for millions. As Rupe says, "This is a business that requires precision."


TVA's is a fine example of information systems practice, replete with the typical change-management problems and a pretty decent financial payoff. My question, however, is whether it is really knowledge management. A few years back, it would clearly have been called a document management and workflow system. Why adopt the newfangled terminology? Maintenance documents such as those captured and managed by the TVA system can obviously contain knowledge. But accessing and distributing documents electronically was possible long before knowledge management and can be accomplished through a variety of means--not only workflow, but intranets, e-mail, electronic publishing and so on. Are all of these knowledge management applications?

To my conservative mind, whether this is a knowledge management application rises and falls on one factor: that is whether the knowledge flows to and from the maintenance workers. A system that distributes documents along the maintenance process is a document management system. A system that allows maintenance workers to record their own observations about the documents, the process and their day-to-day use on the job is a knowledge management system.

When I talked to Robert Rupe at TVA, he said that maintenance workers have some ability to modify the system's knowledge. They can note that the work plan is inaccurate--for example, they can see that the plan didn't say if insulation had to be taken off before changing a valve. Or they can determine if scaffolding is necessary to do the job safely. These workers note such issues on paper, which later gets scanned into the system for planners to notice and use. What's particularly reassuring is that, while Rupe says they didn't call it knowledge management when they built the system, sharing knowledge within nuclear plants and across the industry is an extremely important goal in the post-Three Mile Island world. When we're talking about nuclear power, I personally would like to see everyone using their brains and sharing knowledge to the maximum degree possible.

So let's say that the TVA maintenance application is at least a knowledge-oriented system. Although it's not the sexiest application around, its existence is good news for knowledge management. It suggests that knowledge management is making its way into the down-and-dirty parts of organisations. It points out that knowledge is an important resource not just for white-collar knowledge workers but for people who use wrenches and valves and flanges as well. The system also indicates that we are finally starting to embed knowledge into jobs rather than adding it on top of them.

The TVA analysts were diligent about ensuring that the flow of knowledge was integrated with the flow of the maintenance process. I hope that we'll see more hybrids of process redesign and knowledge management in the future.

- Tom Davenport

Join the CIO Australia group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.

Join the newsletter!

Error: Please check your email address.

More about Island

Show Comments

Market Place