precise prediction of mass loss is needed. To achieve this, the time domain chromatic parameters Lt,lw and 1-St,yw may be combined on a single map(Figure 5.14). The
variation of CuW contact mass loss is clearly apparent on this map. In general, as the mass loss increases, the values of Lt,lw increases and 1-St,yw decreases. As an example,
for tests at 35kAp (i.e. tests l1-l5), the position of test l1 on the graph is distinctively
away from the rest of the tests at the same current amplitude (within the dash circle). A similar phenomenon is observed for tests at 40kAp (m1 is away from the gathering
of m2-m5). It is indicating that, by using the time domain chromatic parameters, the essential process of contact erosion is addressed and the accuracy of prediction of contact mass loss may be improved accordingly.
Fig. 5.14 Time domain chromatic parameters Lt,lw versus St,yw of CuW arcing contact
in SF6.
5.3
Interpretation of Chromatic Analysis Results
In this section, the prediction of mass loss of arcing contact using the square of peak current I2
peak and chromatic parameters (wavelength and time domains) are interpreted
5.3.1
Wavelength Domain Parameter Analysis
For tests of Cu contact in N2, it is found that parameters xw(tp) and yw(tp) both
have significant linear correlation with the arcing contact mass loss and are therefore selected as predictors P0 and P1 (Figure 5.5 (a) and (b)). Using the linear regression
technique, the regression coefficients are derived as listed in Table 5.2. Substituting the coefficients in Equation 5.2, the mass loss of Cu arcing contact can be obtained. The standard error of the estimate (SEE) may be determined by comparing with the measured mass loss (Chapter 3). The error from the chromatic results is 54.1mg which is slightly larger than the value (48.3mg) obtained using the square of the peak current
Ipeak2 as an predictor of contact mass loss.
For tests of CuW contact in SF6, by using the chromatic graphs obtained in the
previous section, the mass loss for an arcing contact can be evaluated using Linear Regression methods. Figure 5.8 indicates that the parameters Lw(tp) and 1-Sw(tp)
have relatively strong linear correlation with the arcing contact mass loss. These are therefore selected as the predictors P0 and P1. The constant C in Equation (5.2)
is excluded in this case. The derived regression coefficients are listed in Table 5.2. Substituting these coefficients into Equation 5.2 yields the predicted mass loss of CuW arcing contact and the SEE value of 51.3mg. Using the square of the peak current (I2
peak) as a predictor of contact mass loss, the SEE value is 66.0mg. Thus, the
wavelength domain chromatic parameters improve the accuracy of the CuW contact mass loss by 22.3%. The improvement is due to the weak effect of droplet ejection on erosion and elimination of spectral emissions from the anode region.
Table 5.2 Regression coefficients of mass loss prediction using wavelength domain chromatic parameters.
Test Code β0 β1 C
CuN2-S -1238.4 (P0:xw(tp)) -1211.6 (P1:yw(tp)) 974.8
CuWSF6-S 1.26×10−7 (P
0:Lw(tp)) -268.5 (P1:1-Sw(tp)) n/a
For now, the wavelength domain chromatic parameters not only improve the accuracy of mass loss prediction of arcing contact, but also contain valuable information regarding the state of the arcing system. For instance, the value of Lw(tp) represents
5.3 Interpretation of Chromatic Analysis Results
indicator of current amplitude; the value of Hw(tp) represents the dominant wavelength
which can be treated as an indicator of the onset of contact erosion etc. To summary, by using the six wavelength domain chromatic parameters, the characteristics of the spectral arc emission can be deduced, the trend of mass loss increasing can be intuitively traced on the x-y, H-S and H-L plots, and the mass loss values can be estimated using linear regression method.
5.3.2
Time Domain Parameter Analysis
The time domain chromatic parameters represent the time variation of wavelength domain chromatic parameters which is expected to be more reliable for mass loss prediction. For tests of Cu contact in N2, it is found that xt,xw and yt,xw have a
strong linear correlation with the Cu arcing contact mass loss (Figure 5.10). The time domain parameter yt,xw is selected as the only predictor P0. The derived regression
coefficient and the constant are listed in Table 5.3. By substituting these coefficients in Equation 5.2, the mass loss of the Cu arcing contact can be obtained and the SEE value is 40.2mg which is much improved in comparison with the SEE value of 54.1mg when using the wavelength domain parameters as predictors. By using the time domain chromatic parameter yt,xw, the accuracy of the prediction of Cu contact mass loss is
improved by 16.8% comparing with arc energy method.
For tests of CuW contact in SF6, the time domain chromatic parameters Lt,lw and
1-St,yw are selected as the predictors of CuW contact mass loss (i.e. P0 and P1) and
the constant C is omitted. The derived regression coefficients are listed in Table 5.3. Substituting the coefficients in Equation 5.2, the mass loss of CuW arcing contact can be obtained and the SEE value is reduced to 40.2mg. The accuracy of mass loss prediction is improved by 39.1% comparing with arc energy method.
Table 5.3 Regression coefficients of mass loss prediction using time domain chromatic parameters.
Test Code β0 β1 C
CuN2-S -5502.4 (P0:yt,xw) n/a 1910.6
CuWSF6-S 5.25×10−8 (P
5.3.3
Analysis of Accuracy of Mass Loss Prediction
Table 5.4 compares the the SEE values of mass loss prediction using different predictors (i.e. the square of peak current, the wavelength domain chromatic parameters and the
time domain chromatic parameters).
Table 5.4 Comparison of Standard Error of Estimate (SEE) of Mass Loss Prediction using Various Predictors.
Predictor CuN2-S CuWSF6-S
I2
peak 48.3mg 66.0mg
Wavelength Domain Parameters 54.1mg 51.3mg Time Domain Parameters 40.2mg 40.2mg
For Cu arcing contact in N2, the SEE value of mass loss prediction using Ipeak2 is
48.3mg and increases to 54.1mg when using the selected wavelength domain chromatic parameters (xw(tp) and yw(tp)). In this case, the lower accuracy of the prediction using
the wavelength domain chromatic parameters is probably due to three reasons. First, the melting and boiling temperature of copper are relatively low so that molten droplet ejection is expected to be severe during the period of arcing. Overall the amount of mass loss caused by molten droplet ejection (the portion which is not vaporized in the arc column) can be significant which is not effectively represented by the arc spectra. Second, the install position of the optical sensor is at the centre of the observing window (Figure 2.6), thus the spectral emission from the anode (tulip contact) region is superimposed upon that of cathode (plug contact) which will inevitably result in larger errors especially at high currents. Third, since the arc current is sinusoidal, the arc spectra are always mid transient without reaching a steady state. The analysis of the arc spectra at a specific moment overlooks information which might exist in the time domain.
For CuW arcing contact in SF6, the SEE value of mass loss prediction using Ipeak2 is
66.0mg and decreases to 51.3mg when using the selected wavelength domain chromatic parameters (Lw(tp) and 1-Sw(tp)).