Partial Least Squares (PLS) Integrated Fourier Transform Infrared (FTIR) Approach for Prediction of Moisture in Transformer Oil and Lubricating Oil

Table 7

Prediction of moisture blind samples of transformer oil using PLSR, verified with Karl Fischer method.

Expected concentration (ppm)	Predicted concentration in ppm (% error/CV)
	PLS regression (n = 5)	Karl Fischer (n = 2)
0	131 ± 208 (NA/158%)	3.74 ± 0.49 (NA/13%)
600	771 ± 443 (29%/57%)	342 ± 214 (43%/63%)
900	1493 ± 688 (66%/46%)	501 ± 140 (44%/28%)
1300	1233 ± 673 (5%/55%)	672 ± 105 (48%/16%)
2100	1798 ± 306 (14%/17%)	2055 ± 258 (2%/13%)
2400	2712 ± 849 (13%/31%)	2421 ± 2034 (1%/84%)
3200	2824 ± 745 (12%/26%)	2037 ± 411 (36%/20%)
3900	3779 ± 847 (3%/22%)	3263 ± 1600 (16%/49%)
4600	4182 ± 495 (9%/12%)	3722 ± 832 (19%/22%)

% error: percentage error; CV: coefficient of variation.