Using an Aging Model for Determining Least Cost

Condition Monitoring Intervals

John E. Skog

Doble Consultant

Maintenance and Test Engineering Co.

January 21,2000

Introduction:

In my last Doble Exchange article, I illustrated how an aging model could be used for determining the least cost interval for performing intrusive time-based Preventive Maintenance. The example looked at breaker mechanism maintenance and used a three-parameter Weibull function to describe mechanism aging. It also assumed the preventive maintenance would rejuvenate the mechanism and return it to an “as-new” condition.

In this article, we will begin to examine condition monitoring activities that do not return the maintained device to an “as-new” condition. We will look at how condition monitoring differs from intrusive PM and begin developing a model for handling maintenance activities that do not renew an item but rather detect the on-set of a failure.

Condition Monitoring Tasks

Condition monitoring activities, whether they are continuous or periodic, are designed to identify a specific characteristic change that is indicative of a future functional failure. For a condition monitoring activity to be effective, the following must exist:

� The characteristic being monitored must be related to the function of interest.

� A change in the characteristic must be observable before loss of the function

� Changes in the monitored parameter must be measurable

� The time interval between condition monitoring activities must be appropriate and a function of the normal aging rate of the device being monitored.

In other words, condition monitoring activities can only be effective if the aging process is somewhat predictable. If the parameter being monitored does not provide advance notice of an imminent loss of function or does not provide maintenance personnel with adequate time to respond to a deterioration in functional performance then the monitoring task is not technically effective. Additionally, if the life cycle cost of performing a condition monitoring task is excessive then the task is not economically effective.

Maintenance strategies employing condition monitoring activities are a two-step process. First, the condition monitoring activity looks for a significant change in the parameter being monitored, and second, when a change is detected, a condition-directed maintenance task is initiated. The condition-directed maintenance task generally results in the replacement of a deteriorated sub component, adjustment or refurbishment of a major item. The exact maintenance path chosen is determined by cost, complexity and effectiveness.

Bushing Power Factor Example

Except for removing external contaminates from a bushing, the performance characteristics of a bushing is affected little by maintenance. In a classical sense, bushings may be considered a maintenance-free accessory of a transformer of breaker. In most cases, the expected life of a bushing exceeds that of the device they are installed in. However, there are a substantial number of bushing designs that do not meet these expectations.

This is not to say that bushing reliability should be taken for granted. We all know that the effects of a bushing insulation failure can be significant and the economic impacts can be orders of magnitude greater than the replacement cost of the bushing itself. We also know that periodic Power Factor testing or on-line monitoring of the bushing capacitance and insulation characteristics can detect incipient bushing faults.

Because the impacts of a bushing insulation failure can be so great and the probability of failure not small enough to ignore, condition monitoring is a necessary maintenance task for reliable apparatus operation. While condition monitoring activities do not increase the life of the bushing, they can increase the reliability of the apparatus by warning of an incipient failure. The questions that need to be answered are:

� Do we monitor on a continuous basis using on-line diagnostic techniques?

� Do we monitor periodically by performing Power Factor test on a de-energized device?

� If we monitor periodically, how often should we test?

Placing these questions in an economic/risk environment allows one to determine the most cost effective condition monitoring approach to maintenance.

The Model:

The economic/risk model developed, compares the costs of each of the possible responses to a bushing insulation failure problem. The model assigns a probability of detecting an incipient fault through Power Factor testing and a probability of the fault being catastrophic for each year of service. As the bushing ages, the probability of an elevated insulation Power Factor and subsequent failure also increases. The model assumes that a significant change in insulation Power Factor can be detected well before the bushing would fail. The main components of the model include:

� Bushing aging characteristics

� Relation between detection of a failure and actual failure

� Present age of the bushing population

� Customer outage impacts

� Maintenance costs and intervals

� Bushing replacement costs

� Associated transformer or breaker repair costs in the event of a bushing failure

� Interest and inflation rates

Aging Characteristics for Condition Monitoring

Power Factor test data for 6,000 bushings was obtained form the Doble Engineering files. This data was from bushings considered to be acceptable for service and all of the same family or general design. When the data was subjected to Weibull analysis, an aging equation was developed resulting in the following aging/cumulative failure equation:

Equation 1 F(t) = 1 – e ^–(t/^h⁾^b

Where:

F(t) = Cumulative Probability of the PF exceeding 0.5%

t = time in years

h = 42 years

b = 5.3

Changes in the bushing Power Factor have been considered by the industry to be the precursor to normal end-of-life. The above equation provided an estimate of the time when the bushing Power Factor would double indicating that the end of bushing life was near. End of bushing life would manifest itself in the catastrophic failure of the bushing and sever damage to the transformer or breaker.

Applying accelerated life techniques; the relation between the onset of an elevated Power Factor and end of bushing life was developed. This technique assumes that the failure mechanism for accelerated tests and normal operations is the same and that only the time scale changes. For the family of bushings analyzed, the above aging model was modified to predict the end-of-life. By using an acceleration factor of 1.5, the bushing end-of-life equation became:

Equation 2 F(t) = 1 – e ^–(t/^h⁾^b

Where:

F(t) = Cumulative Probability of the PF becoming exceeding 0.5%

t = time in years

h = 63 years

b = 5.3

Simply stated, these two models predict that for a large population of bushings, 63% would:

Exhibit elevated Power Factors in 42 years

Fail catastrophically in 63 years

The cumulative failure rate for both elevated Power Factor and catastrophic failure are depicted in figure 1 below.

Figure 1

Cumulative Failure Rates

Failure Detection by PF Test and Catastrophic Failure

Determining Failure Risk

Condition monitoring is about managing the risk of failure. In the case of bushing insulation, rarely does the insulation immediately change from a satisfactory condition to an unsatisfactory one. Except for insulation failures caused by severe voltage disturbances, bushing insulation tends to age gracefully. There does become a time when the aging process accelerates and the risk of near term failure becomes significant. Sometime before the onset of rapid insulation deterioration is when we want to remove the bushing from service. The question of when to remove the bushing from service is a question of risk.

Figure 1 illustrates the expected cumulative number of failures for a specific family bushings. The left curve is produced from equation 1 above and models the onset of failure and its detection by changes in insulation Power Factor. The right curve is produced from equation 2 above and models the actual cumulative failure rate of the bushing. Power Factor changes are assumed to be a pre-cursor to the actual bushing failure and can be detected at an early stage. When the probability of detecting a significant change in Power Factor via diagnostic methods exceeds 50%, bushing failure is assumed possible. The model predicts that a 50% probability of the bushing having a significant increase in Power Factor occurs at 40 years of service, at that time, the risk of bushing failure is approximately 8%.

In the Next Doble Exchange:

We have used actual test data to develop a model for detecting the onset of a failure through insulation testing. From this model we have developed a second model that predicts the risk of bushing failure. In the next Doble Exchange we will perform an economic evaluation and determine the most appropriate condition monitoring response for this particular family of bushings.

Editors Note: John Skog is a Doble Consultant in the area of Maintenance Management. He introduced the Client Group to RCM in 1993. John performs RCM training and consulting services through Doble and is available to assist clients in the refinement of their current maintenance programs. Previous RCM articles published in the Exchange may be found at: www.mtec2000.com