ELECTRICAL WORLD, December 1985

BAD INSULATORS POSE HIDDEN THREAT

A common assumption is that defective suspension insulators on a transmission line can readily be identified either visually or by visual inspection coupled with a knowledge of local conditions. However, the, results of a systematic program of testing energized suspension insulators in situ, conducted over a period of almost 10 years by West Kootenay Power & Light Co. Ltd., British Columbia, Canada, shows that the majority of defective insulators of this type reveal no externally visible symptoms.

West Kootenay's current test program began when a 55-mi, 60-kV line that had previously exhibited good reliability began to experience unexplainable trips, usually after precipitation had occurred. Reclosing was successful, and patrols on foot and by helicopter failed to detect any causes. Finally, in 1975, the circuit failed and locked out, and though patrols again failed to uncover a cause, a rural customer detected a flashover of the center string on a vertical running corner. This string was replaced and the line successfully reclosed. The insulators in the string appeared to be perfectly normal but tested defective, so the company became convinced that invisible insulator failures were causing the trips.

Because every string on the line had been closely inspected visually, the company tried several other methods to detect these invisible failures. Each string was scanned by a large parabolic microphone with a frequency divider. Unfortunately, every string was found to be noisy.

Because every string on the line had been closely inspected visually, the company tried several other methods to detect these invisible failures. Each string was scanned by a large parabolic microphone with a frequency divider. Unfortunately, every string was found to be noisy.

Every string was then examined using an infrared detector, looking for signs of heat generated by leakage current. None was found.

Linemen measured voltage drop across each insulator in a string, using phasing sticks, and the results were plotted on the assumption that a voltage drop would fall within predictable limits, depending on where in the string the insulator was located. The results were inconclusive.

Working with a prototype of the Hi-Test Detection Instrument, West Kootenay developed a tester designed to duplicate, on an energized line, the conventional shop test. That test "meggered" resistance at 10 kV and rejected any unit with less-than-infinite resistance. The tester that has evolved from that prototype is a self-contained, HVDC power source that is unaffected by ac voltages of up to 30 kV across its test probes. It imposes a 10-kV test voltage across the test points and measures the resultant current flow. It can be powered either by a 120-V-ac or a 12-V-dc source, and is light enough to be used by a single lineman at the end of a standard hot stick. A current-limiting feature prevents shock to personnel even at the 10 kV across the probes.

When the early prototype of the tester was put into service on the line, the results were startling. Over a third of the insulators tested on nontangent structures showed as defective, and subsequent laboratory tests confirmed the results. By 1978, all insulators shown as defective by the field test had been replaced. Since that time, not one trip attributable to a suspension insulator has been experienced.

Recent experience

Live-line crews recently were called upon to replace a dead-end string believed to have been damaged by rifle fire. The string was on a 170kV line, 88 ml long, constructed in 1965 using 15.kip insulators supplied by several manufacturers. The insulators were installed in strings of twelve and supported 477-kcmil Hawk conductor with a normal tension of 4,000 lb.

Examination of the broken shells of the seven insulators that lay on the ground revealed that they had not been shot, but were broken almost exactly in half. Using the tester, the crew determined that only three of the remaining five insulators in the string remained electrically effective, so both ends of the string had to be considered hot during changeout.

West Kootenay now routinely tests suspension insulators on nontangent structures to determine that they can be safely hot-sticked before any live-line procedure is undertaken. Frequently, insulator conditions discovered are such that routine live-line procedures would place the line men at risk.

Each string is tested by a pair of line men, one who operates the tester and one who records the data on a standard form, starting at the structure and proceeding towards the conductor. If an insulator tests defective, then a second reading is taken across the porcelain to measure surface contamination. With the exception of some insulators in an area of known contamination, all insulators that have tested defective registered zero deflection across the skirt and full deflection from pin to cap, Average test time is 13 minutes per structure. Previous testing has convinced the company that failures on tangent structures are sufficiently rare so as not to justify testing of those structures at this time.

The test procedure stipulates that testing of a particular string stop when a sufficient number of defective units had been found so that the voltage to ground divided by the number of remaining untested insulators would yield 30 kV or higher.

Tests are not conducted on noisy strings, which indicate high electrical stress. These measures are designed to protect the linemen and to avoid triggering a flashover.

Test results

A total of 84 structures on the aforementioned 88-mi line were tested, including 30 running corners and 54 dead-ends. Corrective action was needed on 42 of the structures. The data from these tests support the conclusions discussed below.

• Suspension insulators fail in service in ways that give no visual indication. A total of 5,681 insulators tested produced 295 defective units. This does not include those with defects capable of visual identification, such as broken skirts, flash marks or cracks. In other words, all insulators identified as defective by the tester bore no visible signs that they were defective. This is consistent with the results of all previous testing programs within the company. Sectioning the defective insulators invariably showed cracks that penetrated through the porcelain under the cap (Fig I).
  
  
• Failure is random with respect to position in a string. A common belief in the industry is that the insulator closest to the conductor is the one most likely to fail. This is not supported by the test data. (Table I). A Chi Square analysis of these data gives the value of X² as 20.0638. At 11 deg of freedom and 0⁰=0.02, X² =22.618. These values indicate that the occurrence of defects in a string is random with respect to position relative to the conductor.
  
  
• Defective insulators are not randomly distributed throughout the line. Tangent structures rarely have defective insulators. The test data suggest several additional ways in which defective units seem to be concentrated (Table II). Although the demonstrated average probability of finding a defect is P = 0.05 for all insulators tested, the probability varies by type of structure from a low of P = 0.00 on down hill dead-ends to a high of P = 0.28 on up hill dead-ends.
  
  The following hypotheses are advanced for possible consideration.
  • First, the greater the exposure of the cement bonding to moisture and contamination, the higher the incidence of failure. Note that there is an extremely low incidence of failure on tangent structure where the insulator strings hang vertically and varying higher probabilities of failures for different configurations having varying degrees of exposure.

    Note also the data on uphill and down hill dead-end strings in Table II. The probability of finding failures in the uphill strings is a high 0.28, compared to essentially zero for downhill strings. The structures represented by these data are located on mountainsides such that the uphill strings have a high exposure and the downhill strings a low exposure.
    
    

  • A second hypothesis is that the greater the mechanical stress, the higher the incidence of failure. The test program produced some interesting information regarding stress to which the insulators might be exposed-but not stress of the kind imposed by operating conditions. Three anomalies appeared in the test data. First, all nine of the defective units on running corners were found on one particular corner out of 30 tested. Also, of the 20 defective units found in 84 jumper strings, four were in one particular string, and three in another.

    Mechanical stress that could damage the insulators might be imposed by:

    1. Missing or improperly located vibration dampers
    2. heavy snow loading, or
    3. improper handling during installation.

    Discussions with the crew that had constructed the line revealed that the strings in question might have been hoisted into position by a hitch secured at the center of the string. During the hoist, the line occasionally slipped, letting the string fall free for several feet before it was jerked to a stop. It is not unreasonable to suspect that this shock load might cause pin-to-cap cracking, although the data are not adequate to support this hypothesis.

• The nonrandom distribution of failures poses a significant hazard to both reliability and personnel safety. A cursory review of the data could lead to the erroneous conclusion that the overall incidence of insulator failure is so low as to pose no hazard to system reliability or the safety of personnel. With a probability of failure of any insulator being only 0.05, and the situation with the worst probability of failure-that occurring at dead-ends--affecting only seven structures out of 84, there doesn't appear to be a significant threat.

  Further analysis, however, disproves this. Using the full string as the unit of analysis, there appears to be a "cascading" effect of failure within a string (Table III).

  Using the string as the basic unit of analysis, the probability of finding at least one faulty insulator in that string is P = 0.28. This is based on the test of 414 strings (excluding the running corners that may have been overstressed during installation - see above), of which 115 had at least one defective unit. This is a very large increase over the probability calculated with the individual insulator as the basic unit. What is significant with respect to the reliability of the line and the safety of personnel is what happens to the probability of finding additional defective units in a string once the first one has been found.

  Given that at least one in the string is defective, the probability that there are more rises to P = 0.65.

  Given that at least two defective units have been found in a string, the probability that there are more is P = 0.59.

  Table III lists the probabilities associated with the higher incidences of defective units.

  These results demonstrate the cascading effect that says that once one defective insulator has been found in a string, the probability of finding more in that string increases significantly and remains high right up to the point where enough are defective to pose a hazard to reliability and safety.

  To illustrate the potential loss of reliability, the data show that, on the 88-MI line in this analysis, there were 31 strings where the insulation level had been reduced to 67% or less of installed values, and, of these, 13 were at less than 50% of installed value.

  From a safety viewpoint, standard operating procedures to service the insulators on the same line would have created a situation where the level of insulation was reduced by 50% or more on 24 strings and by 60% or more on 13 of those strings. The defective insulators show no visible signs of defects, and would have appeared to the line crew to be functioning normally.

Go to the IT-4 Insulator Tester product page

<top of article>

Back to Library / Archive Index