Acoustic Emission (AE) is a nondestructive testing method that has been successfully used in a variety of applications dating back to the early 1970s. It covers the detection of structural-borne sound (in the ultrasonic range) that has been generated by some source of distress. An application has been developed for in-service testing of power transformers to detect, locate and assess both thermal and electrical faults from the sounds originating inside a transformer.

In June 2001, the Electric Power Research Institute (EPRI; Palo Alto, California, U.S.) began a Tailored Collaboration to investigate the “Development of a New Acoustic Emission Technique for the Detection and Location of Gassing Sources in Power Transformers.” This program was motivated by the belief that AE could only be used to detect partial discharge (PD) and that field-testing evidence showed other sources of AE were being detected.

Since its initiation, this program has gone through a Phase I effort, is presently at the end of Phase II and will continue on a Phase III. This article discusses the test procedure, results and database that have been generated by Phase I and Phase II testing to date.

Laboratory Testing

At the start of the project, a series of laboratory tests was performed to determine whether AE could detect heating sources. Although several studies have shown AE can detect PD, none have looked at localized heating. The results of these lab studies indicated that detectable AE is generated when the localized temperature reaches about 120°C (248°F) and it increases as the temperature increases. This was the basis for the fieldwork.

Procedures for In-Service Testing

The overall procedures for collecting data are rather straightforward. The first step is to mount AE sensors on the walls of the transformer using vacuum grease and magnetic hold-downs. Core type, physical size of the transformer tank and the presence of accessories (such as load tap changer [LTC] compartment, radiators and pumps) determine the number and location of the sensors. Other types of sensors monitor load current, pump/fan current and LTC operation. Temperature sensors measure the surface temperature on the outside of the tank wall and LTC compartment. AE, current and temperature data are recorded and logged against time to be used in post-test data analysis.

After a test has been set up and calibration performed, data collection is initiated. This typically continues for 24 hours to note the behavior of the transformer through at least one complete load swing. A separate AE sensor is mounted on a structure near (but not connected to) the transformer so that there is a means of determining when rain and heavy winds are present during the data collection.

Post-Test Data Analysis

During the field test program, transformers that are gassing heavily at the time they are tested may produce sufficient data for evaluation, in a period of less than an hour. Less active gassing transformers produce small amounts of noise and usually require the full 24 hours of data acquisition to perform a proper assessment.

In post-test analysis, noise is removed from time periods when there is rain, snow, high winds, LTC operations or pump/fan operating noise. Several methods have been tried to identify or classify these signals so they can be removed from the original data set. When this happens, generally enough data are left to make a proper assessment.

At this stage of the EPRI TC (Phase II), there are several analysis tools and techniques that have led to the establishment of an in-service test procedure and a database of about 100 transformers tested. These include a unique mix of design, manufacturer, capacity, voltage and gassing history. The results of this development work have led to a grading system that integrates the AE test results with the dissolved gas analysis (DGA) and gassing trends. This grading system provides four grades from letter “A” (good) to letter “D” (bad).

The strength of the AE method is its ability to calculate the location of the fault in 3D and to assess the type of fault present at this location. When a thermal or electrical fault is present, it tends to generate AE with a signature unique to the type of fault. Electrical faults tend to produce AE signals that are well behaved in time and that correlate with the 60-Hz voltage in the transformer. In comparison, thermal faults tend to produce random AE that can cease completely when pumps start operating (due to reduction of temperature on the faulted area). Both types of faults tend to produce a unique set of AE signal features.

During data acquisition and post-test analysis, a graph can be generated showing the 3D location of each AE event. If the fault at this location is active during the 24 hours of data collection, it will generate a good number of signals that can be triangulated back to the same location. These “clusters” typically represent fault sites that can be traced back to critical locations either on the main tank (leads, coils, bushing-lead connections, core laminations, core-ground connections, DETC selector) or the LTC compartment (leads, contacts, supporting insulating bars/envelope).

In-Service Test Results

Several examples are available to demonstrate the benefits of this type of testing.

• Case 1. A Generator Step Up (GSU) transformer, 25/500 kV, 784 MVA, FOA, Westinghouse, three-phase, two pump groups, shell form (Fig. 1). This unit started gassing in December 2001 after being in service for more than 20 years. DGA indicated the existence of overheating by all seven methods applied. The gas generation rate increased periodically. Therefore, the decision was made to test this unit using AE to locate the gassing source(s) inside the transformer.

  This transformer was tested continuously for 68 hours (May 2002), while instrumented with 22 AE sensors. Many events were obtained over the course of the test with the event rate changing dramatically. Post-test analysis indicated the existence of two clusters of events: Cluster 1 located close to H1 high-voltage bushing and Cluster 2 located below the X1 low-voltage (LV) bushing. Most of the events and the highest-peak amplitude were detected in Cluster 2 (Fig. 2).

  Analysis indicated the number of events recorded was strongly related to pump operation. One pump group was running all the time, and the second group was operating when it reached a preset temperature value. It is believed these events were triggered by a high temperature value in the faulted area that reached its critical value just before the operation of pump group #2 was ordered. As soon as pump group #2 started running, this activity was significantly reduced. This is most likely the result of a drop in temperature at the fault location. This behavior, along with the DGA results, reinforced the belief that a thermal problem existed.

  This unit remained in operation with DGA values being monitored periodically. Twelve weeks after the AE test was completed, an internal inspection was performed. The internal inspection revealed carbon buildup caused by severe overheating on one of the cooper bus bars of the X1 LV bushing. This bar was also completely cracked with all of the current being carried by the other two bars (approximately 6000 A). This fault was detected near Cluster 2. Though it is only speculation, it is believed that had this transformer been left in this condition much longer, it might have failed catastrophically (Fig. 3).

  This utility estimates that about US$3 million was saved thanks to the location of this imminent fault.

  An internal inspection was performed on the area around Cluster 1 and no indication of degradation was found.

  The unit was put back in service after the repair of the bus bar, and a second AE test was performed in February 2003. No acoustic emissions were detected on the area where the bus bar fault was found. However, moderate acoustic activity was still detected in the area of Cluster 1. DGA values have remained steady, so no further action is required. If any future gassing develops, this area should be immediately checked with further acoustic emission testing.

  The sister unit of this transformer was tested, and this GSU did not indicate any gassing activity, and no acoustic emissions were detected using the same settings and sensor locations.

• Case 2. Three identical three-phase transformers, OA/FA/FA, core form, Westinghouse, built in 1977, 138/13.8 kV, 40 MVA. Transformers were generating considerable amounts of hydrogen. Two units (1 and 3) were tested in October 2001 during the first phase of the EPRI program. At that time, continuous acoustic activity was detected only in one channel (sensor 12) on Unit 1 (no events).

  Unit 1 failed in January 2003 (15 months after the AE test was performed). An internal inspection revealed the fault was located a couple of inches away from where sensor 12 was positioned (4 ft [1.2 m] from the floor). The fault occurred on the high-voltage winding (aluminum) in the lower part of the coil of phase C (Fig. 4).

  No AE test was performed on this unit between October 2001 and January 2003; therefore, it was not possible to determine the evolution of the fault using this technique.

  When Unit 3 was tested in 2001, a few events were detected on the upper and middle part of the phase C coil. When this unit was retested in February 2003, a significant amount of events was obtained, indicating a significant increase in the acoustic activity 2001, the same area where Unit 1 failed.

  Unit 2 was tested for the first time in February 2003. A strong acoustic source was located at the bottom of the coil of the phase A. Based on the location of the cluster, the fault could be located on the preventive autotransformer (LTC) or on the lower part of the phase A coil (same location, different phase than the sister units) (Fig. 5).

• Case 3. Single-phase autotransformer, 500/230/34.5 kV, 243 MVA, OA/FOA/FOA, shell form, Westinghouse. This unit was tested because it was exhibiting a positive trending in its gas concentration. Even when the TDCG did not exceed the values recommended by the IEEE C57.104, it represented a deviation from the baseline data.

  The AE test indicated a high amount of events at the top of the unit. Most of the events were detected when all the pumps were off. This can indicate a thermal origin for the acoustic activity. The hypothesis is that when all the pumps are off, the temperature on the fault starts going up until it reaches a value and starts producing acoustic emissions. When the pumps start running, the temperature goes down and the acoustic activity ends. This appears to be a repetitive cycle (Fig. 6).

  The area where the acoustic activity was found is where the connection between the windings leads and the bushings is located (Fig. 7). In previous experiences, the utility has found problems in the same area in similar units. This report helped the engineering department obtain a budget approval of US$2 million to replace the unit.

• Case 4. An ABB transformer, three-phase core-form, 230/13.8 kV, OA/FOA/FOA, 51/68/85 MVA that was tested using AE in a repair/refurbishment facility during an induced voltage test (Fig. 8). Simultaneously, the radio-influence voltage (RIV) test and an electrical method to detect PD were applied.

  In all tests (and in tests previously performed), RIV values and PD activity were higher in phases A and C (up to 2000 µV). Therefore, AE activity was expected to be detected and located in both of these phases. Four tests were performed, relocating the sensors accordingly to where the acoustic activity was being detected to get a more accurate location of the acoustic source. In all the tests, two clusters of events were detected during AE test on this unit.

  Cluster 1 was located in the lower part of the LV side at the bottom of the phase B coil. Most significant acoustic activity (higher energy, higher amplitude) and most of the events were located in this cluster. The origin of the acoustic activity was believed to be a LV or LTC lead(s). Events located in Cluster 2 could be a reflection of the activity detected in Cluster 1 because of propagation of the sound trough the lead (Fig. 9).

  Acoustic activity was first detected by zonal location only (no events) when 50% of the nominal voltage was applied. Most of the events were obtained when the voltage was 100% of the nominal value or higher. The minimum RIV level when AE was detected was about 100 µV.

  An internal inspection was performed on this unit. Discolored insulated paper was found at phase B in both LV and LTC leads, both of which were not properly separated and insulated.

  As these leads come out of the winding and are bent vertically upward, both LV and LTC leads were pushed up right against each other and generated PD when the induced test was performed.

  The same discoloration was found on phase A in both LV and LTC leads but not to the same degree as phase B. All of phase C insulation was clean. The LV/LTC leads for phases A and B were separated and the connections re-insulated.

  The unit was tested after the repair and monitored only with RIV and electric PD detection techniques (no AE). Levels obtained were well below the values obtained before the repair, and the transformer was delivered to the customer to be put back in service. AE tests allow the manufacturer to have a quick indication of the position of the fault and to focus all the internal inspection on this area.

• Case 5. A continuous monitoring of a single-phase, shell-form, GSU transformer was performed during a solar storm. Along with the AE sensors, the current on the ground terminal was monitored to detect geomagnetically induced ground currents (GIC) and correlate this parameter with any AE detected.

  A sudden increase was detected in the ground current (Fig. 10) at the same time acoustic activity was detected, as well as a few events located in the area where the core-ground connection exits. Twelve hours later, the on-line gas monitoring system indicated increases on ethane and methane gases in the unit.

• Case 6. The last example is a three-phase McGraw Edison core-form transformer, 26.4/13.8 kV, OA/FA, 7 MVA with a 550-BLS load tap changer. Both the main tank and the LTC compartment were tested simultaneously using AE. The LTC compartment was observed to have a higher temperature than the main tank.

After 20 hours and 15 minutes of continuous monitoring, an active AE source was located inside the LTC compartment (Fig. 11). At the completion of the test, it was determined that the difference of temperature between the main tank and LTC compartment was 30°C (86°F) (Fig. 12). No acoustic sources were detected inside the main tank.

Conclusions and Considerations

AE has shown to be effective in detecting and locating faults when applied on power transformers in service. Assessment cannot be performed without extensive post-test analysis, which includes comparison with data from other sensors as well as external events, such as rain and high wind. The accuracy of the location is sufficient to direct internal inspection to a specific location.

Only a handful of cases have internal inspection with accuracy in the location of the problem. Without more feedback, it is difficult to completely evaluate the AE method and focus on areas where improvements can be made.

At the present stage, however, we have an effective in-service test method that correlates well with the DGA results and trends, and allows the detection of electrical, thermal and mechanical faults. As other post-test analysis tools evolve, expect to see a significant improvement in the fault assessment effort. Overall, it is anticipated that there will be a gradual acceptance of AE as a means to generate additional information regarding the state and integrity of operating power transformers.

Acknowledgments

The authors would like to express their gratitude to the sponsors of the EPRI Tailored Collaboration project.

Arturo Núñez is the manager of the Power Equipment Division at Quality Services Laboratories. His division is responsible for the application/development of nondestructive testing techniques for the condition assessment of power equipment. Prior to assuming this position, Núñez was a research engineer at the Instituto de Investigaciones Eléctricas of Mexico. He earned a BSEE degree in electric power systems from the Instituto Politécnico Nacional of Mexico.
ANunez@qslplus.com

Samuel Ternowchek is the vice president of operations at Quality Services Laboratories. He is responsible for the overall operations of the Advanced NDT Group of QSL. Prior to joining QSL, he held several positions at Physical Acoustics Corp., and also spent several years at Lucent Technologies in the area of failure mode analysis and process control. He has a degree in engineering technology from Penn State University, and is a fellow of ASNT.
STernowchek@qslplus.com

Ronnie K. Miller is the group technical director for the MISTRAS Holdings Corp. Before joining MISTRAS Holdings, he had a varied career in the aerospace industry including Hughes Aircraft, General Dynamics and United Technologies. In addition, his NDT experience includes working for Dunegan/Endevco, MQS/Dunegan Testing and United Technologies Research Center. Miller received his BS, MS and PhD degrees from Purdue University and the School of Aeronautical and Aerospace Engineering.
RMiller@pacndt.com

Barry Ward is project manager for Power Transformers and High-Voltage Instrument Transformers in the Transmission & Substations Business Area of the Science & Technology Development Division of EPRI. Prior to joining EPRI in 1997, he was employed by AVO International for 19 years. Ward is a registered professional engineer and is a member of the IEEE serving on the Transformers Committee of the Power Engineering Society. He received his BSEE degree from The University of Bradford, England.
Baward@epri.com