Measurement Uncertainty In Algorithms Using PCA Method

Propagation analysis of selected uncertainty sources in algorithms using the PCA method has been presented. The paper shows uncertainty analysis in algorithms, which use minimization of squared distances technique and maximizing the variance technique. On the basis of simulation tests the influence of the used signal sampling technique on the eigenvalues vector for the sinusoidal signal containing additive white noise, has been compared. Three applied sampling techniques have been analyzed: synchronization of the beginning of the data acquisition, for the successive sequences xi, immediately after a positive zero-level crossing of the analyzed waveform; separation of the NM-element sequence x to M of N-element sequences xi collected sequentially in subsequent rows of the matrix X; and application of sampling with the fractional delay of d=Ts/2, and X matrix construction from alternating strings: x2i-1 - sampled with a delay of d, and x2i - sampled with a zero delay from the moment of the zero crossing by analyzed waveform.