by Arne Oas

Systems and equipment often show repeatable patterns of operation and performance that you can use to identify and predict problems. The analysis used is called, obviously enough, pattern recognition.

Pattern recognition analysis is usually shown mathematically (trending), by deviations from the norm (exceptions’ reports) or through a visual presentation (photographic, video).

Consider the trend of the heat exchanger example from last issue. Heat exchanger fouling changes during the year due to organism or plant growth (see graph on top of next page). Gradually, growth blocks the passage of cooling water.

Starting with a clean system in the winter, it has full flow. Left untreated, there is no reduction in flow until spring. At that time, growth and bio-fouling start to happen, accelerating through the spring and summer, then tapering off in the fall as the water temperature drops. Finally, growth stops in the winter.

Visual inspections would confirm the cause of the fouling pattern. In the spring, you would see small plants and organisms starting to appear. Subsequent inspections would show growth through the spring and summer, tapering off in the fall as the water temp drops. In the winter, many of the organisms are dead and washed away.

Bio-fouling is an equipment problem with relatively short (yearly or less) cycles. These patterns can be complex, like the heat exchanger, or as simple as a cylinder failing every other month. Patterns are usually easy to recognize when comparing one year or period to the next. If the pattern follows the normal progression or behavior, related problems are easy to predict. The more periods/ cycles analyzed, the stronger the prediction.

If systems are steady for a long period and a new pattern or trend starts to develop in the cycle, immediately look for unusual or destabilizing conditions. Once you determine the conditions, determine the root cause of the change. Finding this often requires additional data collection and analysis.

The change graph below shows the decrease in flow to our heat exchanger accelerating and the new forecast trend. This means we need to service it before the scheduled winter cleaning. Investigating the change may reveal an exceptionally warm spring or that the exchanger was idled for a time. Either could cause increased growth rates in plants and organisms. Prevention may be as easy as adding a step to drain the exchanger when idled.

Starting in areas where corrosion, erosion, leakage or mechanical failure of a repeatable nature is occurring, you can develop patterns for many systems and components. Look to production. Many problems occurring may be seasonal, or product or operations related.

Establishing knowledge about patterns requires you to:
• Decide what data to record.
• Decide how to record the data and in what form.
• Determine reliable access to the data and past analysis.
• Document the procedures used for collecting, recording and analyzing the data.

Pattern recognition often requires a long time to develop the periods and determine repetitive cycles or interval between events. Continuity of personnel throughout the process ensures consistency of information. Pattern recognition, given patience and extra effort to implement, guarantees a payback.

Arne Oas is the senior maintenance consultant at Management Resources Group. If you have a maintenance management software question, contact Coach Oas at 215-918-2165, or e-mail oasa@mrginc.net.

This article appeared in the December 2001/January 2002 issue of MRO Today magazine. Copyright 2002.

Back to top Back to MRO Coach archives