CHAPTER 3 ACOUSTIC EMISSION TECHNIQUE FOR DETECTION AND LOCATION OF PD

Transcription

1 63 CHAPTER 3 ACOUSTIC EMISSION TECHNIQUE FOR DETECTION AND LOCATION OF PD 3.1 INTRODUCTION PD measurements on high-voltage equipment, e.g. transformers, could be grouped into two major tasks. First, evidence of PD (detection process) is to be provided as sensitive as possible and to distinguish it from any potential disturbances at the same time. Secondly, in many respects important determination of the failure origin (location of the PD), classically referred to as Triangulation (Nieschwitz 1976; Lundgaard 1992) has to be performed. In other words, complete information about a PD activity in order to judge the insulation condition of a transformer comprise its level, maybe in terms of its apparent charge q, its type and location. Particularly for answering the question Where is the PD source?, which is of great interest for finding the reason for the PD activity and risk assessment, the inherent existent possibility of the acoustic method to geometrically localize the flaw by means of signal propagation delays were used very early. According to (Fuhr 1993) the first published discussion on acoustic PD location was in As the geometrical location of PD was based solely on electric measurements most of the time merely possible in an approximate manner, the acoustic location of PD stayed attractive until now, especially for transformers. Today commercial systems are available (Eleftherion 1995) and applied (Léger 2001).

2 PRINCIPLE OF AE TECHNIQUE PD is a localized breakdown in an insulation system with nearly instantaneous release of energy. A fraction of the released energy heats the material adjacent to it and can evaporate some of it, creating a small explosion. The resulting discharge manifests itself as an observable electrical, acoustic and sometimes optical signal. The discharge acts as a point of source for acoustics waves. As the discharge duration is very short, the emitted acoustic spectrum is very broad (several MHz). The intensity (db) of the emitted wave is proportional to the energy released in the discharge. Thus, the amplitude of the wave is proportional to the square root of the energy in the discharge. As the energy released is often proportional to the charge squared, a linear relationship between the amplitude of the acoustic wave and the discharge magnitude (in coulombs) is common. In detecting and locating the Acoustic Emissions generated by partial discharges, propagation path of the acoustic wave is important as the AE wave characteristics like frequency, signal amplitude and shape may change during wave journey to the transformer tank. Acoustic emission events produced by local PDs in transformers are individual signal bursts shown in below Figure 3.1, characterized by the following parameters: AE count (Ring Down Count) AE signal amplitude AE signal duration AE signal rise time Total events Amplitude distribution AE event energy

3 65 Figure 3.1 Typical AE burst type signal parameters Acoustic Emission produced by PDs in transformers are always associated with inevitable continuous background noise from transformer core and surroundings in the field. A threshold detection level is set above the background level and serves as a reference for the following waveform parameters. Acoustic Emission Count (or) Ring Down Count (RDC) is the number of times a signal crosses the preset threshold. The number of threshold crossings per unit time depends on sensor frequency, damping characteristics of the sensor and structure and threshold level. Peak Signal Amplitude can be directly related to the intensity of the source producing the acoustic emission. Peak amplitude measurements are generally performed using a log amplifier to provide accurate measurement of both large and small signals. Amplitude Distributions have been correlated with deformation mechanisms of the object under the test.

4 66 Acoustic Emission Event Energy can be related to the energy released and can be measured in several ways. The true energy is directly proportional to the area under the acoustic emission waveform. 3.3 PD DETECTION PRINCIPLE PDs in transformers cause high frequency pressure pulses to propagate through the insulating media. These are similar in character to stress waves propagated in solids during the crack formation (Acoustic Emission). The techniques involved in the detection of acoustic emissions from cracks are shown to be well suited to the detection of emissions from the partial discharges. AE discharge detection is based on the detection of the mechanical (pressure) signals emitted from the partial discharge. The discharge acts as a point source of acoustic wave that propagates spherically throughout the surrounding oil/insulating media and reach the transformer tank. High frequency piezo-electric sensors, mounted on the transformer tank, can pick up these pressure waves and convert them into electric signals, which would be analyzed using a data acquisition and processing system. The shape of the signal depends on the source, detection apparatus and AE sensors. The principle of AE detection technique is schematically shown in Figure 3.2 Figure 3.2 Principle of schematic AE technique

5 AE System Instrumentation be suitable to Any typical AE instrument used for PD detection and location must Detect Acoustic Emissions Locate the sources of Acoustic Emissions Provide a means of discriminating between signals of interest and noise sources such as friction, impact and electromagnetic interference. Depending on the applications, AE instruments commercially available vary widely in form, function and price. In multi-channel system, measurement of the signal proceeds simultaneously on every channel that detects the AE wave. The stream of event description is passed through a central processor that coordinates the task of data storage, display and operator communications Test Standards for Acoustic Emission Many different types of instrumentation are available for detection and location of AE. Several typical systems are described below, as per IEEE PC57.127/D1.0, March 3, 2003, which have been shown effective in certain transformer arrangements; however, other systems may be equally or more effective, depending on the transformer physical parameters and the location of the PD. AE system are: As per IEEE Standard C , the components of typical

6 68 Sensing transducer, piezo-electric displacement transducer of resonant frequency between khz (Differential or shielded type to minimize the effect of electromagnetic fields). Preamplifier can be an integral part of sensor system or a separate unit and it should accept high impedance (10000 Ohms), low amplitude (<100µV) signals, provide a gain of 40dB and is capable of working into a 50 ohm load Filter, negates vibrations caused by the magnetostrictive action of the core (Barkhausen noise), pumps and fans. Many of these fall below 30 khz, but Barkhausen noise emanating from the core has been found to be in the range of 50 khz to 300 khz. So 100 khz high pass section with a rapid, roll-off response characteristic is needed. The reasonably generous band-pass (200 khz) allows for variations between different transducers, in so far as their resonant frequencies are concerned. Power Amplifier has a flat frequency response (± 5%) between 50 khz and 300 khz. Its gain must be adjustable in at least 2 db increments over the range of 20 to 70 db. The maximum inherent noise level produced by the amplifier should not exceed 100 µv referred to its input. Counter, the counter circuit is capable of counting the number of individual oscillations contained in the AE signal during a time interval, commonly 1 s, or a set number of cycles. Display, the output from the counting circuit, covers a wide dynamic range. Oscillation rates from 0 to 10 6 counts/s should be capable of being displayed. To accomplish this, a logarithmic or digital display is recommended. The physical display should be convenient and easy to read.

7 69 Power Supply; it should be powered by an isolated supply. This is necessary because the power-supply ground is unlikely to be the same physical ground as the transformer tank. Ground loops are to be avoided for both safety and noise reduction. When size and transportation limitations do not preclude an adequate supply of power for the desired time period of monitoring, batteries may be used. 3.4 AE SENSORS The main advantages of sensors are that they do not need any electrical connection to the high voltage supply. The current PD detection normally consists of converting the PD pulse current from the sensor to a voltage signal, since most of the measurement instrumentation is sensitive to voltage rather than current. The measurements of acoustic signals have distinct advantage of being free of electrical interference and simple to implement compared with electric PD measurement method. Various types of commercially available AE sensors are as shown in Figure 3.3. WD Sensor R15-I sensor R15 Sensor Figure 3.3 WD, R15-I and R15acoustic sensors

8 Preamplifier Gain and System Threshold Setting of Acoustic Emission Sensor AE produced by PD in transformers are always associated with inevitable continuous background noise from the transformer core and surroundings in the field. Hence, a threshold detection level and pre-amplifier are set so as to eliminate background level and increase the measured PD signal strength. Hence, to set the threshold detection level, pre-amplifier gain and atmospheric correction factor experimental studies are performed. Figure 3.4 Pencil break method for AE sensors calibration

9 71 In general as per the patent (Hsu 1976), of acoustic emission simulator (Figure 3.4) is used to calibrate the sensors before every test to capture the effect of change of atmospheric and physical conditions of the test object. in Table 3.1. The gain and threshold values set for different sensors are as given Table 3.1 Sensor settings Type of Sensor Pre Amplifier Gain (db) Threshold (db) Type A Type B Type C Frequency Characteristics of Acoustic Sensor AEA system was operated in transient mode to analyze the frequency component of acoustic signals acquired by all the three sensors placed at 20 cm (in Z plane) away from the PD source. Markings were made on the transformer tank to facilitate fixing of sensors with respect to PD source. The captured acoustic signals were processed using digital filters (in-built) and Fast Fourier Transform (FFT). The measured responses of the PD were digitized into discrete samples. The measured responses were plotted using MATLAB functions (M-file) utilizing the discrete sample data obtained from experimental study. Figure 3.5 show the time and frequency domain responses of the PD due to plane needle electrode arrangements, respectively.

10 72 Figure 3.5 Time and frequency domain plots of captured AE signals Three different sensors were characterized based on the signals captured and shown in Table 3.2. Table 3.2 Characteristics of the Acoustic sensor Type of Sensor Sensitivity (in mv max) Predominant frequencies Type A 200 ~ 150kHz and ~ khz Type B 600 ~150 khz Type C 300 ~ 150kHz Type A sensor picked up switching surges while applying each incremental voltage step. Types B and C showed similar frequency response. Also type C having in-built PA was less prone to electro-magnetic

11 73 interferences by nearby operations of other control circuits. Hence, type C sensor is found more suitable for field applications. 3.5 EXPERIMENTAL STUDY TO EVALUATE AE METHOD OF PD DETECTION Further experiments were conducted to examine the effect of barriers i.e., windings and other structural components which are seen in the actual transformer. Hence, a model was designed to examine such effect and experiments were carried out Experimental test Setup for PD Measurement A transformer tank of 1045mm X 400mm X 925mm dimensions was designed to study AE PD measurement. To simulate the PD in transformer tank, plane-needle electrode configuration is used (Figure 3.6(a)). 1.5mil Mylar paper was sandwiched between two 1.5 mm pressboard to avoid breakdown at higher electrical stresses (Figure 3.6(b)). Floating objects in the form of pins were inserted in the pressboard insulation to incept higher pc levels. The electrode arrangements were suspended in oil filled transformer tank with the facility to vary the position of the designed PD source model. To study the attenuation of AE waves due to transformer insulation like oil and windings, three identical Rod-plane electrode systems were arranged inside the tank. An acrylic board (transparent) was placed on one side of the transformer tank along the length to observe any visual discharges/activities inside the tank during experimentation.

12 74 Attenuation due to barriers I - No barriers around the PD source II - PD source surrounded with two windings (1.5 X 4 mm paper covered conductor used for winding. III - PD source, surrounded with one winding Figure 3.6(a) Plane needle electrode configuration with solid insulation Figure 3.6(b) Extended view of solid insulation between the electrodes Experiments were carried out to find the effect of barrier and the results are tabulated in Table 3.3. Table 3.3 AE parameters with respect to barrier variation Configuration PD Source position Sensor (R15) Max db Without barriers (85, 47.5, -20) (85,47.5,0) 76 With 1 barrier layer (21, 47.5, -20) (21,47.5, 0) 70 With 2 barrier layers (57.5,47.5,-20) (57.5,47.5,0) 67

13 75 It is found that the db level sensed reduced with inclusion of barriers. The attenuation of AE signal through oil and solid barriers needed more extensive experimentation with more geometrical variation Effect of Sensor Location The experimental studies were performed to study the effect of sensor location with respect to PD source captured. To facilitate this, markings were made on tank surface. R15 sensor was placed at various positions with respect to PD source in vertical and horizontal planes. Figure 3.7 shows the position of R15 sensor positions P1 (85, 49.5, 0), P2 (85, 79.5, 0), P3 (85, 19.5, 0), P4 (57.5, 49.5, 0) and P5 (21, 49.5, 0) with respect to PD source at (85, 49.5, -20). Figure 3.7 Tank marking showing sensor positions (P1 to P5)

14 76 Configuration-I was energized to 10 kv to give a consistent pc in the range of 250 to 280pC at a confined location. Signals were captured in parametric mode and sensitive parameters with position and tabulated in Table 3.4. The sensor output was analyzed in parametric mode at each position and the results indicated maximum hits/sec at position P1 and db to be varying with sensor position non-linearly which needed extensive experimentation to arrive at conclusion. Table 3.4 AE parameters with respect to sensor position Source position Sensor (type C) Position Max db (85, 47.5, -20) (85, 47.5, 0) 70 (85, 47.5, -20) (57.5, 47.5, 0) 64 (85, 47.5, -20) (21, 47.5, 0) 60 (85, 47.5, -20) (85, 77.5, 0) 70 (85, 47.5, -20) (85, 17.5, 0) 68 The configurations I, II and III of the experimental model were designed to study the attenuation of Acoustic Signal due to winding barriers Different Types of PD Models The test setup described in section (Figure 2.5) is used to analysis the acoustic emission methods with respect to PD intensity level, sensitivity, etc. The tank is marked for the coordinates to easily understand and identify the internal location of PD source model. Markings were provided to place piezo-electric acoustic sensors. The complete assembly of transformer insulation model is shown in Figure 3.8. Markings were made on the transformer tank to facilitate fixing of

15 77 sensors with respect to PD source. Acoustic sensors were placed on the outer tank wall to detect pressure waves generated by the partial discharge process. The acoustic impedance between a PD source and the detection point is complicated as the acoustic wave will propagate along multiple pathways and is distorted by attenuation. Figure 3.8 Complete assemblies with bushings Figure 3.9 shows placement of both acoustic and HFCT sensors. An acoustic sensor PAC WDU is attached on the transformer tank. To improve the acoustic medium, there is coupling gel is applied between acoustic sensor and transformer tank. HFCT (IPEC CAE 140/100) is connected via ground terminal of the transformer model. The response and sensitivity of AE sensor with respect to the PD inception levels for the three PD models is shown in Table 3.5.

16 78 Figure 3.9 Acoustic sensor mounted on the tank wall Table 3.5 Sensing characteristics of AE sensor in different arrangements Arrangement PD inception voltage (kv) AE response kv pc Rod Plane Cavity model Floating electrode PD SOURCE LOCATION PRINCIPLE The possibility of PD location is one of the major features of Acoustic discharge detection. Location can be based on either measurement of the time at which signal arrives the sensor or on measurement of signal amplitude. In practical situations, a location based on a time-of-flight measurement requires two or more simultaneous measurements in order to facilitate triangulation to determine the source location. The simplest approach is to measure the electrical signal simultaneously with the acoustic signal. If the acoustic propagation velocity is known, then calculation of the

17 79 source location becomes simple. However, the fact that different wave components travel along different paths in a structure is a complicating factor. If the electrical signal cannot be detected, a triangulation can be carried out as a simultaneous measurement with several acoustic sensors. In a continuum, the source must be located on a hyperboloid between the two sensors, which can be determined from analyses of time lag of the signal. All locations on such a hyperboloid will result in the same time lag between the signal arrivals at the two sensors. Different types of algorithms were developed based on the above facts namely, Linear Triangulation and Spherical, etc When external mounted sensors are used for location purposes, a low pass filter is employed to remove all frequency components that travel faster in the tank than in oil. All distances can then be calculated based on the propagation in oil alone, as the direct wave from the oil arrives first at the sensor Location Algorithms Principle of one dimensional view for PD location In 1-dimensional PD source localization as shown in Figure 3.10, where S1 and S2 acoustic sensors are placed in same plane. Here differences in distances from the PD source to sensors S1 and S2 is proportional to the difference in times to reach the sensors.

18 80 Figure 3.10 Linear PD source location in 1-dimension Principle of two dimensional view for PD location Similarly in 2-dimensional PD source localization as shown in Figure 3.11, where S1, S2 and S3 are acoustic sensors placed in the same plane along the tank surface. Figure 3.11 PD source location model in 2 dimensions Z = R sin (3.1) Z 2 = r 12 - (D-R Cos ) 2 (3.2) R 2 = r 12 - D 2 +2D R Cos (3.3) R 2 = Sin 2 + r 12 - (D- R Cos ) 2 (3.4) R 2 D 2( tv 2 2 t v Dcos ) (3.5)

19 81 In this case, AE source can be any point on the locus of hyperbola given by the above equation. In order to locate the exact source point on the locus of hyperbola, one more sensor is required and is called triangulation algorithm. Figure 3.12 Triangulation method of location in 2-dimension Figure 3.12 shows the contours drawn on the xy-plane to locate the PD source using 3 sensors in same plane. Solution for the model for triangulation location (x-y) coordinates is given by: t 1 = r 1. R (3.6) Z = R sin (3.7) Z 2 r 2 1 (D x R cos ) 2 (3.8) From above equations: R 2( t (D 1 v 2 1 D 1 t 2 1 v cos( 2 ) 1 ) (3.9) Similarly between S 1 and S 3 R 2( t (D 2 v 2 1 D 2 t 2 2 v cos( 2 ) 3 ) (3.10)

20 82 where, V, velocity of sound t1 and signals arrival time. t2, time differences between S 1 -S 2 and S 1 -S 3 sensor Triangulation method (3 dimensional location) Different PD sites have different components; successfully locating a PD source can help identify the problem in the transformer. Also, precisely locating a PD source is very useful for transformer maintenance and repair. However, accurate PD location is not easy due to complex transformer structures, which may obstruct the path between the PD sites and the sensor terminals, or distort the PD waveforms. The positioning algorithm employed to find the acoustic signal in a transformer tank is based on the range equations from a known point to the position of the source. The range between the source and sensor can be expressed in terms of the sides of a right triangle using the Pythagorean Theorem. AE source (X, Y, Z) And the sensors are located at x i, y i and z i where i = 1,2,3 and 4 T is the propagation time from the Pd source to the first hit sensor (s i ) and the time delays between S 1 to the other sensors are 12, 13 and 14 D 1 = V s T D 2 = V s (T+ 12 ) D 3 = V s (T+ 13 ) D 4 = V s (T+ 14 ) (X-x 1 ) 2 + (Y-y 1 ) 2 + (Z-z 1 ) 2 = (V s T) 2

21 83 (X-x 2 ) 2 + (Y-y 2 ) 2 + (Z-z 2 ) 2 = (V s (T+ 12 )) 2 (X-x 3 ) 2 + (Y-y 3 ) 2 + (Z-z 3 ) 2 = (V s (T+ 13 )) 2 (X-x 4 ) 2 + (Y-y 4 ) 2 + (Z-z 4 ) 2 = (V s (T+ 14 )) 2 By using least square algorithm, it can be solved for unknown source location X,Y and Z. Where, (x,y,z) are the coordinates of the source and (x 1,y 1,z 1 ) are the coordinates of the sensor. For each sensor in an array, one of these spherical equations can be written and a system of nonlinear equations is created. Many methods have been developed to solve systems of nonlinear equations. A common example is the Newton-Raphson method, which uses a Taylor Series Expansion to linearize the set of equations and an iterative algorithm to find the solution given an initial guess. However, this method requires that the guess be close to the true position of the source and can be computationally expensive. An alternative to the Newton-Raphson Method is to form a difference equation using the range equations from two sensors If the range is written as the speed of sound in the medium times the signal propagation time then the difference equation can be written in terms of time difference between two sensors (Sacha Markalous,2006). This leads to a system of linear equations that relates the source coordinates to the unknown range from the first sensor to the source. Where (x i1,y i1,z i1 ) is the difference between the coordinates of the ith sensor and the first sensor, r i1 is the speed of sound in the medium times the time difference found between the first and the ith sensor, r 1 is the range from sensor 1 to the source, and K i = x 2 i + y 2 i + z 2 i. If each of the source coordinates (x,y,z) are solved in terms of r 1, the resulting expressions can be inserted in the range equation for sensor 1 in order to solve for r 1. Once the positive root of the quadratic equation is determined, the value of r 1 can be used to determine the source coordinates (Raja.K and Floribert.T, 2002).

22 EXPERIMENTAL FOR PD LOCATION BY AE METHOD Two Dimensional Source Location using real PD Discharges Figure 3.13 shows the placement of 2 AE sensors on the transformer tank. To improve the acoustic medium, coupling gel is applied between acoustic sensor and transformer. Results of this 2 dimensional PD experiment are given in the Table 3.6. Figure 3.13 Experimental setup for 2-dimensional PD location Table 3.6 Two-dimensional source location S. No Source position (x, y) in mm 1 160, , ,200 TDOA t 12 =57.1, t 13 =73.2 t 12 =56.1, t 13 =74.4 t 12 =55.3, t 13 =72.3 Computed Position (x c, y c ) in mm Error x (%) Error y (%) (163.3,200.9) (162.7,201.0) (163.4, 201.4)

23 85 Table 3.6 (Continued) S. No Source position (x, y) in mm 4 160, , ,200 TDOA t 12 =54.1, t 13 =71.1 t 12 =57.1, t 13 =72.2 t 12 =57.1, t 13 =75.2 Computed Position (x c, y c ) in mm Error x (%) Error y (%) (163.3,200.5) (162.3,197.5) (162.0,201.7) Mean/Standard deviation Three-dimensional Source Location using Real PD Signals Different PD sites have different components; successfully locating a PD source can help identify the problem in the transformer. Also, precisely locating a PD source is very useful for transformer maintenance and repair. However, accurate PD location is not easy due to complex transformer structures, which may obstruct the path between the PD sites and the sensor terminals, or distort the PD waveforms. The positioning algorithm employed to find the acoustic signal in a transformer tank is based on the range equations from a known point to the position of the source. The experimental setup with four AE sensors were mounted on the tank. Figure 3.14 shows the coordinate systems of the four sensors and the acoustic source, with x, y and z values (in mm) for sensor 1 of (90,150,0), sensor 2 of (90,150,0), sensor 3 of (90,150 0), sensor 4 of (90, 150, 0) and the acoustic source of (95,150,150). Figure 3.15 shows the samples of the oscilloscope waveforms from the four sensors.

24 86 Figure 3.14 Sensor positions for three dimensional source location Figure 3.15 Waveforms of the signals captured by 4 AE sensors Sensor Coordinates S 1 (0,150,90), S 2 (120,0,90), S 3 (150,310,90),S 4 (305,150,90) x-axis: time in 50 µv/div; y-axis: voltage in 20 mv/div. The PD location was estimated by using time difference of arrival (TDOA) method on the sample tank shown in Figure The result of this method is as shown in Table 3.7.

25 87 Table 3.7 Source location results for sample tank measurement Sl. No Source position (x,y,z) in mm Computed position (x c,y c,z c ) in mm Error x (%) Error y (%) Error z (%) 1 (90,160,190) (88.21,155.12,188.92) (90,160,190) (86.32,153.32,187.62) (90,160,190) (89.24,156.64,189.02) (90,160,190) (88.31,158.24,187.52) CONCLUSION For online detection and location of PD in transformer using acoustic emission PD method, acoustic emission characteristics such as frequency, signal amplitude and shape are utilized. Wide-band (WD), resonant tuned sensor with external amplifier (R15) and resonant tuned sensor with in-built pre-amplifier (R15-I) types of acoustic sensors were for experiments, and it is found that resonant tuned sensors with pre-amplifier was most suitable for online measurement applications due to its immunity to external electromagnetic inferences. To locate the PD based on response of the acoustic signals experiments were conducted on 1-dimension, 2-dimension and 3-dimension. It was found that this method gave consistent results in all experiments. The measured AE signals from three AE sensors are utilized for PD location using triangulation method. The triangulation method locates the PD sources successfully for all the live discharges with the accuracy of 10%. The sensor is sensitive only for the signals of 100pC and above (consistently). Lower intensity values are sensed only after prolonged measurement.

26 88 Acoustic PD measurement based condition assessment of site transformers is suitable for units having PD levels higher than 100pC, which is considered as faulty transformer as per the new revised acceptable limits. Hence, it is not suitable for assessment of transformers below the limit, for planning the maintenance. However, based on the practical experience of the author during site assessment of transformers it is formed that there are many transformers on-site with PD levels higher than 100 pc and are operating time.