Robust algorithms for spectral analysis of the technological parameters ofindustrial plants

It is shown that, in spectral analysis, different types of noise in the tec hnological parameters produce significant distortions in the values of thes e: parameters. The reason for this circumstance is that in the course: of p rocessing the data, there arises a difference between the microscopic error s of the positive and negative paired products of the readings of the signa ls which are being analyzed, on the one hand, and the readings of the cosin e and sine curves. As a consequence, the estimators of the coefficients of the Fourier series will vary outside the range of variation of the noise, a result that makes it difficult to apply well-known algorithms for solving numerous problems. Robust algorithms are proposed that eliminate any depend ence of the desired estimators on the parameters of the noise.