Dutch utilities that have been using a new maintenance management methodology have saved more than 10% in operating and maintenance costs. Called MainManr, the diagnostic system provides a tool to change smoothly from time-based maintenance to condition-based maintenance methods with resultant savings and lifetime extension.

Because aging equipment can have higher failure rates, today's utilities have to increasingly consider the efficiency of their maintenance and replacement policies. In cooperation with the Dutch utilities, KEMA investigated how new maintenance methods should be implemented in the near future. For existing North American and European electrical infrastructure, more than US$3 billion (net present value over a 25-year period) could be saved on maintenance costs and postponed investments, most of it within the first 10 years of introducing the new policy.

About 10 years after the current average 25-year service life, aging processes will gradually increase the risk of faults of a wave of components, dependent on operational conditions. Reliability may decrease severely unless effective measures are taken using additional knowledge of the physical condition of the component and its failure causes.

Together with the utilities, KEMA formulated the following objectives: - Control and reduce maintenance costs. - Keep reliability and safety at an acceptable level. - Avoid failures due to aging preventively. - Obtain savings from lifetime extension if feasible. - Document the wealth of experience of retiring experts. - Improve the maintenance expertise.

Maintenance Management In many utilities most components are maintained by following a corrective (at failure) and time-based strategy. In the latter case, the time interval is usually chosen conservatively to ensure high reliability. On behalf of the Dutch utilities, KEMA proposed a smooth change into a new maintenance strategy, which is to be based more on the condition of the components and on what is functionally needed by the system. However, the maintenance actions are mainly organized by component. The need for maintenance action is analyzed on a basis of the loss of system functions by failure mode effects and criticality analysis (FMECA).

Condition-based maintenance (CBM) avoids the high costs of intensified time-driven tasks independent of condition. It addresses specific, effective tasks to the needs as indicated by diagnostic inspections. In comparison with time-based maintenance (TBM) the benefits of CBM result from maintenance savings and lifetime extension. In that respect, CBM can provide a more adequate management tool to meet the terminal behavior of the equipment.

Stepwise approach to implement strategy. On the way towards a dedicated maintenance concept between TBM and CBM, KEMA has guided the participating utilities through the following steps:

A. Preselection of components. - Return on investment analysis. B. System-functional analysis. - Failure statistics. - Failure mode effect and criticality analysis. C. Aging model. - Early indicators of deterioration. - Verification experiments. - Economic organization, bath-tub curve. D. Maintenance cycle. - Planning/inspection/diagnosis/revision. E. Expert support system. - Object structure. - Fixed data/inspection data. - Knowledge rules/maintenance activities. F. Feedback to improve aging model.

In Step A, a cost-benefit analysis is performed to rank the necessity of developing CBM protocols. A breakdown of the cost savings for substation components is shown in Fig. 1. Priority has been given to transformers in high-voltage substations because they constitute most of the attainable savings.

Reliability-centered maintenance approach and feedback. In Step B special attention has been given to analyzing the system-functions, which are sustained by a component. Specialized user groups, consisting of experienced engineers and managers from practice, worked together to explain their knowledge and expertise on dominant failure modes. Thereby, KEMA adopted largely the principles of reliability-centered maintenance (RCM). For RCM it is important to carry out a FMECA to prioritize the major failure causes (i.e. functional loss) and the feasibility of their early detection. Experience has led to "rules of thumb" that have been agreed upon by users and manufacturers. Steps B and C were carried out to derive from those insights and verification experiments a set of knowledge rules, which are used in an expert-support system for making diagnoses. Step F, feedback from practice, changes the knowledge rules, converging them into an optimal set.

Diagnosis of maintenance needs. To implement a new strategy, KEMA developed MainMan, the expert support system. Its specialized functional modules are related to the maintenance processes occurring at different organizational levels at the utilities. In Fig. 2a the central part of MainMan operates tools for diagnosis and optimization of the maintenance interval. Furthermore, services are available for inspection and trend analysis. Figure 2b shows how MainMan modules operate in a network with data storage in the central unit. On-line monitoring of equipment and on-site input by inspectors is enabled by calling in data from laptop computers in the field.

MainMan documents the equipment data as well as the knowledge rules. To support the maintenance department in decision making, a technical-alarm status and an economical interval optimization are provided for long- and short-term maintenance/replacement planning of the (sub)system considered.

Domain-specific knowledge and general-inference mechanism are separated, thus resembling the original concept of an expert system. The built-in flexibility of the knowledge rules makes it possible to change the maintenance strategy, for example, to a concept in between periodic- and condition-based maintenance. In this way, dedicated sets of rules are introduced that model the system's different functional needs. In addition, maintenance actions can be linked at the component level to meet technical and safety requirements.

Definition of system/component to maintain. The functionality covered by MainMan's "supervisor module" is made possible by its two unique features: a generic maintenance system concept and its diagnosis expert system.

The generic concept enables (re)specification of any maintainable system-function in terms of component structure (Fig. 3) and for each component in terms of its fixed data, inspection data, knowledge rules and activities (Fig. 4). This enablement is carried out without loss of data and without any interference of a programmer. So, extension of the system does not require any additional programming costs.

Functions Covered MainMan's "user module" covers registration of fixed data. The practical use of the methodology is closely related to registering data in MainMan's database. First, the data of the components comprising the maintenance system have to be filled in. Often the fixed data are already available in a central database containing an energy company's data of its technical infrastructure. MainMan contains a facility to download these data. Then a cycle of three activities starts, each activity supported by MainMan as follows: - Planning. MainMan generates an overview of inspections and revisions to take place within a specified period. Maintenance activities can then be grouped into tasks and allocated to maintenance units. MainMan supports both separate (for one component) and integrated (for more components) grouping of maintenance activities. A special function is the time-harmonization of maintenance schedules of different components. This way, maintenance activities can be carried out as efficiently as possible. - In-the-field registration of inspection data. For each maintenance unit, MainMan exports the tasks to the MainMech module. The maintenance unit can report its findings via laptop computer. When the tasks have been carried out, MainMan imports the results into its database. For the objective acquisition of qualitative aspects like "much corrosion" by various maintenance units, inspection classes are specified with the aid of digitized photographs. - Diagnosis. MainMan infers the condition of a component when the component inspection is registered and generates a best new-inspection or revision date, as shown in Fig. 5. The knowledge rules inferring the condition of the component also generate the specific maintenance tasks and add those to the work planner.

Additional Features MainMan supports various types of overviews, such as topographical and tabular overviews of diagnosed conditions and planned activities (Fig. 6). Trend analysis of inspection attributes of components is possible to develop or refine knowledge about the dominant aging processes. MainMan also generates management information about expenditures and economics of maintenance.

To facilitate the methodology's implementation in different utilities, MainMan meets several state-of-the-art and future technical requirements. It works with MS-Windows on a stand-alone computer or in a network environment. And, it can exchange data with other information systems, thereby supporting various standard formats.

Diagnostics and Inspections Since lifetime modeling often requires expensive verification experiments, condition performance indicators are used as an alternative approximation. Most information on older equipment's condition is obtained by man-made inspection and by applying available and cost-effective diagnostics. The scores of the diagnostic key quantities due to wear or functional loss, evaluated by the documented rules/criteria, provide an estimated stage of deterioration. MainMan collects the results of different diagnostic sources, transferring them into a system alarm status on behalf of the utility's decision-making process. The alarm status is linked to a maintenance action and interval.

Following are some practical examples of diagnostic key quantities: - Switchgear inspection, leakage currents. A dominant mode of medium-voltage substations' functional failure is the long-term deterioration of switch insulator surfaces caused by environmental pollution. The leakage current is a critical indicator in that respect. For extreme salt/fog conditions, an aged installation exceeds the level of functional failure within 1000 hrs of endurance testing.

The surface condition can be improved by revision. At normal operating conditions, corrosion inspections of the dielectric surface can be used to plan revision in time and at the right place. The information is transferred into a knowledge rule such as:

If leakage current >0.2 mA while relative humidity <75% then condition is urgent to system function: action within, one month; tasks: repair of insulator surface; expected costs: specified for labor hours and spare parts; optimized economic interval: 10 months.

- On-load tap changer diagnostics. Failure statistics show that more than 50% of the failure risk of power transformers is due to the on-load tap changer (OLTC). Among other priorities, FMECA confirmed the need to develop diagnostic tools for the resistance of the selector switch contacts, which can become carbonized on the long term.

As the switch is in series with the transformer's high-voltage coil, it takes a long time to measure the resistance by conventional methods. KEMA has developed a patented method with which the contact resistances can be measured quickly. In this method the low-voltage side of the transformer is short circuited at the terminals, and a voltage source is connected to the high-voltage side.

The correlation between dissolved gases in oil and the OLTC condition has been investigated, showing that trends for specific gases provide a warning only in severe cases. Another dominant failure mode concerns the OLTC mechanical part. Therefore, motor characteristics have been used for diagnosis. Within the set of knowledge rules of MainMan, all different performance indicators, including redundant control parameters, can be added together to diagnose the system's condition. A weighted score, along with links to probable actions, results in the alarm level for the functional status of the OLTC system.

- Circuit breaker diagnostics and vibration measurements. A CIGRE survey shows that the circuit breaker's mechanical system is responsible for most of the failures. Measurement of the contact travel curve and the vibration measurements can be reduced to key numbers for the mechanical conditions. Deviations in the starting times of the vibration events can be used to recognize common faults like leakage in the damper, reduced SF6 gas pressure, lengthy delay time or incorrect assembling of the drive unit.

- Diagnostics of minimum-oil-type 50-kV circuit breakers. In addition to regular inspections, easily applicable diagnostics have been implemented, such as contact velocity measurements that effectively trace deviations in the mechanical system. Also a non-contacting high-frequency discharge capacitive probe enabled a first location of a few out of 200 inspected circuit breakers in service. Discharges have been located in some bushings and cable terminations, which have been replaced.

An interesting future option is application of advanced partial-discharge analyses and the differences in their phase-resolved distribution. Depending on the aging processes, it is possible to recognize early stages of aging in the partial-discharge fingerprints so warnings could be generated and precautions could be taken before the end of the lifetime.

- Feedback from practice. KEMA has performed several MainMan implementations in the Netherlands, changing the organization and the technical procedures of the maintenance department. After a year of field experience, one of the utilities calculated a 13% reduction in maintenance expenditures by applying the condition-dependent maintenance planning with MainMan. For distribution substations in the coastal area, the expert system conclusion was that approximately 10% of the installations were in urgent need of selected revision tasks. The tasks were all reconfirmed by the maintenance department.

Conclusion Utilities are faced with a growing fraction of equipment coming closer to the end of its lifetime. A higher failure rate results unless adapted condition-dependent and updating strategies are implemented in time. To keep the electric infrastructure well in service for the coming decade(s), costs, reliability and safety are more of a concern. The need for cost reduction and efficiency improvement requires a review of the existing maintenance policies.

The close cooperation with the utilities was important since they were participants in support groups for functional analysis of the system. Support groups were formed by component in the areas of medium- and high-voltage substations and overhead lines. The methodology has been based on quantitative information about statistics and costs and on existing expertise and empirical data on the system's failure modes. During practical evaluations of the MainMan methodology, OM benefits of over 10% have been reported by utilities in making the transition to more condition-based maintenance. The methodology, developed by KEMA, facilitates a smooth transition from present to future maintenance strategies. Inspections and diagnostics for assessing the need for maintenance are an integral part of the methodology.

Editor's Note: This article was adapted from the authors' presentation in paper number 311 at the Conference of the Electricity Power Supply Industry (CEPSI), October 21-25, 1996.

Kees Ackerman is section manager at KEMA Transmission & Distribution, Arnhem, The Netherlands, which he joined in 1995. He has the professional engineer's degree from Delft University of Technology. His responsibilities include the management of the consultant group on Maintenance Management and Diagnostics. He has many years of experience within the industry with larger maintenance organizations dealing with policy, information technology and organization set-ups.

Johan Smit is senior consultant with KEMA Transmission & Distribution, Arnhem, The Netherlands, which he joined in 1979. He has the PhD in physics from Leiden State University, The Netherlands. In 1996 he became professor of high voltage technology at Delft University of Technology. He is secretary of CIGRE study committee 15, Materials for Electrotechnology, and member of several CIGRE and IEC working groups dealing with diagnosis and maintenance of electrical systems.