Technical Article

High-Precision Winding Testing with a New Type of Impulse Winding Tester

December 02, 2019 by Yuki Maita

This article introduces the characteristics of Hioki’s Impulse Winding Tester ST4030A along with examples of actual measurement.

An Innovative Method for Evaluating Winding Reliability Using Characteristic Distribution of LC and RC

Impulse testing, also known as surge testing, is regarded as one of the most effective ways to evaluate winding reliability. At the same time, because it relies on comparing test waveforms with a reference waveform from a known-good winding, it struggles to test minuscule waveform variations with an acceptably high degree of precision.

With conventional testing methods, impulse testers infer the presence of defects in a winding by applying an impulse voltage across the windings’ terminals and observing the waveforms that result. The method offers an exceptional technique for detecting wire breaks, shorts between phases, and variations in inductance in windings.

Recently, the space factor 1 of stator windings has been increasing as manufacturers look to reduce the size of electric motors such as those used to drive electric vehicles. Partial discharge (PD) occurs in such motors when insulation performance between wires is compromised by increasingly high drive voltages and surge voltages caused by inverter control, posing the risk of significantly lower reliability during the motors’ use.

Hioki developed the Impulse Winding Tester ST4030A to resolve these issues. This article describes the characteristics of the ST4030A and offers some measurement results obtained using the instrument.

 

Pass/Fail Decision Based on Characteristic Distribution

The conventional method of identifying windings as either defective or non-defective based on differences in waveform area suffers from a number of problems:

  • Difficulty distinguishing between defective and non-defective windings when differences in the area are minuscule [1]. The percentage of the coil’s cross-sectional area that is occupied by wires
  • Difficulty assigning physical and quantitative significance to calculation results that consist of differences in area
  • Difficulty identifying threshold values
  • Difficulty managing enormous volumes of waveform data in applications where it is desirable to retain such data

On the other hand, the ST4030A uses characteristic distribution,2 a parameter identification technique proposed by Toenec Corporation, as one of its evaluation standards. The characteristic distribution method identifies two parameters of key importance for characterizing attenuating oscillation waveforms obtained from impulse testing (elements of oscillation frequency and attenuation).

In line with previous studies(1), this article will describe the two parameters used in characteristic distribution as LC and RC characteristics. These characteristics are the products of the resistance component R, the capacitance component C, and the inductance component L of the equivalent circuit shown in Figure 1 for the instrument and workpiece under test, in the sense of L × C (LC) and R × C (RC). The former corresponds to oscillation frequency and the latter to attenuation in the attenuating oscillation waveform. The ST4030A provides functionality for displaying the LC and RC characteristics obtained from test oscillation waveforms as a scatter diagram and for making judgments based on whether the characteristics obtained during testing fall within a user-defined pass area (Figure 2).

Compared to the conventional method of screening windings based on differences in waveform area, evaluation based on characteristic distribution offers the following advantages:

  • It allows the configuration of a visually clear judgment area.
  • It is possible to sort windings in motors already installed with a rotor - a task that is considered difficult to accomplish using the conventional method. This is possible because tested windings can be sorted into defective and non-defective categories by setting a pass area as appropriate relative to a distribution that reflects changes in the rotor position.
  • It allows the nature of workpiece defects to be inferred based on the characteristic distribution(2).

Another advantage of pass/fail decisions based on characteristics lies in the technique’s high sorting capability. Consider an example of the difference between the conventional method of waveform comparison and the sorting capability of an approach based on LC and RC characteristics. Fig. 2.3 illustrates the changes in the test voltage waveform and in LC and RC characteristics when a slight change is applied to a workpiece under test. An impulse test was repeated 100 times before the slight change (“Before” in the legend) and after the slight change (“After” in the legend) and variability in measured values was checked under the same conditions using LC and RC characteristic distribution. (The figure includes only voltage oscillation waveform values for 1 cycle before and after the slight change.)

 

Figure 1: Equivalent circuit model for identifying LC and RC characteristics
Figure 1: Equivalent circuit model for identifying LC and RC characteristics

 

An examination of Figure 3 reveals that although no significant difference can be observed in the outline of the before and after waveforms using the conventional method of waveform comparison, a clear difference can be seen in the distribution of LC and RC characteristics before and after the change. Those results indicate the viability of sorting, even when variability in measured values is taken into account. The high sorting capability of LC and RC characteristics offers advantages for detection of minuscule changes such as would be introduced by a single-fault turn in a winding, a defect that would be difficult to identify with conventional waveform comparison.

 

Figure 2: LC and RC characteristic distribution and pass area (shown in gray) on the ST4030A
Figure 2: LC and RC characteristic distribution and pass area (shown in gray) on the ST4030A
Figure 3: Comparison of waveforms (top) and characteristics (bottom) when slight changes are applied to a winding
Figure 3: Comparison of waveforms (top) and characteristics (bottom) when slight changes are applied to a winding
Figure 3: Comparison of waveforms (top) and characteristics (bottom) when slight changes are applied to a winding

 

Partial Discharge Detection Option ST9000

Development Process

Researchers have proposed a number of testing techniques for identifying changes in impulse oscillation waveforms caused by partial discharges. Typical examples of such numerical calculation techniques include Flutter and Laplacian calculation. For convenience’ sake, this article will use the abbreviations FLTR and LAPC to refer to the results of those calculations. The ST4030A implements these two calculations as follows:

Both calculations are based on differential calculation as practiced in the field of discrete mathematics. Although they offer the advantage of being simple enough in terms of processing that a single impulse tester can estimate partial discharge magnitude while carrying out impulse testing, they pose the following problems:

  • Calculation results take the form of the sum of voltage differential values such as those shown above, making it difficult to assign meaning to threshold values.
  • The effects of the noise component contained in the oscillation waveform cannot be ignored.

Hioki addressed the above potential problems with the ST4030A by combining high-precision waveform detection capability—one aspect of the instrument’s fundamental performance—with newly developed digital signal processing technology in a sophisticated manner to develop a proprietary solution in the Partial Discharge Function ST9000.

 

Characteristics of the ST9000

The ST9000’s most distinguishing characteristic is its ability to isolate the partial discharge component from the Gaussian noise component that exists to a certain extent in test waveforms. The function uses the following calculation process to detect the partial discharge component:

  • It extracts only the noise component from the test voltage waveform using digital signal processing.
  • It standardizes the magnitude of the extracted noise component using its standard deviation.
  • In the event of partial discharge, the amount of deviation will exceed that observed in a non-defective winding.

Compared to conventional methods such as Flutter and Laplacian calculation, partial discharge detection processing as implemented by the ST9000 offers the following advantages:

  • Since the process yields a discharge waveform that is synchronized with the time axis of the test voltage waveform, locations of waveform fluctuation that correspond to partial discharges can be clearly identified.
  • Since calculation results are standardized, there is no need to set judgment levels (although they can be adjusted).

 

PDIV Testing with the ST9000

Hioki designed the ST4030A with a breakdown voltage (BDV) function to allow operators to take further advantage of the ST9000’s partial discharge detection capability. This function evaluates voltages that trigger fail decisions while gradually increasing the voltage applied to the workpiece under test. IEC/TS 61934 defines the partial discharge inception voltage (PDIV) as the voltage at which a partial discharge is first detected while increasing the voltage.

As illustrated in Figure 5, an experimental setup was prepared in which a twisted pair consisting of tinned wire and enameled wire is connected to the terminals of a standard winding while the ST4030A was configured to display the test screen while using the BDV function (Figure 4). The screenshot shows a fail result for partial discharge detection judgment (“DCHG” on the screen) at 700 V, indicating that the PDIV for the workpiece under test is 700 V.

 

Figure 4: The ST4030A’s BDV function
Figure 4: The ST4030A’s BDV function
Figure 5: Partial discharge detection experimental setup
Figure 5: Partial discharge detection experimental setup
ST9000 Decision FLTR LAPC
No discharge 1393 872
No discharge 1403 891
No discharge 1397 852
Discharge 1409 890
Table 1 Occurrence of discharge and changes FLTR/LAPC values

 

Comparison of the ST9000 with Conventional Methods

This section offers an example of how the ST9000 improves partial discharge detection capability. When four impulse tests were conducted after setting the ST4030A’s application voltage to 720 V using the setup shown in Figure 5, the ST9000 detected discharge during only one of the tests. The figure and the table provides FLTR and LAPC values as well as the oscillation waveform obtained during discharge testing. The calculation of FLTR and LAPC values was limited to the interval indicated in Figure 6.

An examination of Table 1 reveals no clear, significant differences in FLTR and LAPC values, regardless of whether discharge occurred, making it difficult to check for partial discharge phenomena based on these results.

 

Figure 6: Test oscillation waveform at partial discharge detection and enlarged view of area in question with partial discharge magnitude
Figure 6: Test oscillation waveform at partial discharge detection and enlarged view of area in question with partial discharge magnitude

 

By contrast, an examination of the partial discharge detection results from the ST9000 reveals a clear increase in the amount of deviation, which indicates partial discharge, at the same locations as the high-frequency components that can be observed in the oscillation waveform. Furthermore, detection precision can be maintained at approximately the same level even if the calculation interval is set to the entire waveform.

This article has introduced the characteristics of Hioki’s Impulse Winding Tester ST4030A along with examples of actual measurement. It reviewed the many advantages of the characteristic distribution judgment capability provided by the ST4030A compared to the conventional approach of waveform area comparison judgment. It also described how partial discharge components that cannot be detected using conventional numerical calculation methods can be clearly extracted by using the Partial Discharge Function ST9000. It is the author’s hope that this article will prove to be useful in customers’ testing of windings and motors.

 

About the Author

Yuki Maita worked at Hioki E.E. Corporation that is a manufacturer of electrical measuring instruments supporting technological advancement. These electrical measuring instruments are used in a broad range of industries and fields, from maintenance and inspection of electrical work and equipment to the testing of electronic components used in smartphones and computers and development of electric vehicles and solar power generation technologies.

 

References

  1. Hisahide Nakamura: “Use of Impulse Testing to Diagnose Shorts in Stator Windings in Low-voltage Induction Motors,” Transactions on Electrical and Electronic Engineering D, Vol. 132 No. 9 pp.915-921.
  2. Hisahide Nakamura and Yukio Mizuno: “Efficient Diagnosis of Shorts and Insulation Degradation in Electric Motor Stator Windings,” Transactions on Electrical and Electronic Engineering D, Vol. 132 No. 2 pp.258-267.

 

This article originally appeared in the Bodo’s Power Systems magazine.