SU390535A1 - METHOD OF ANALYSIS OF THE STATIONARY RANDOM PROCESS - Google Patents

Description

The invention relates to the field of automation and computer technology and can be used for the spectral analysis of stationary random processes, which can be, for example, the noise of cyclically working mechanisms of machines.

A known method for analyzing the stationary case of the IrOCESS, “is an example of noise cyclically operating in their mechanisms, which consists in determining the spectrum by the Fourier transform of the process implementation, represented by a discrete choice: which amplitude is determined.

However, TaiK as a diocet sample of the amplitudes of the sipnals is 1c of correlated values, there is a considerable amount of information redundancy in the sample. As a result of the analysis, a significantly larger amount of computations is required, moreover, in the case of a TaiKOM survey, the implementation of measuring equipment and a 1СL calculator: the ITU process processing devices 1ply require all the frequency properties, measurement errors, complexity, and so on. .

The proposed method eliminates the noted deficiency and is characterized in that the amplitude signal residuals are memorized at the time points counted in each cycle from a single phase response and in the memory time interval, which is proportional to

they regulate the phases of the harmonic components of HH .HX of the Fourier transform of the thus-corrected signal sampling.

The essence of the proposed method lies in the sweet.

Fourier transform spectral analysis for discrete selection of the random process implementation is performed at settling and the polling interval in accordance with the Koteliiiuova theorem. Many sushi; there is a shortage, which is determined by such a sample, it can be eliminated by applying a sample of uncorrected values of signal widths. This means that the survey of the implementation of the process should be made at an interval longer than the correlation interval. Such a survey can be done with the help of known devices. In this case, a random process should possess the marketing properties of a stichuniness and ergodicity.

For the Fourier transform of a Correlated Discrete Sample, it is necessary to produce in 1 the product of the a1 MLF of the values obtained at a certain time, and the cosine and sine values of the same time for the frequencies of the entire computed range.

Practically, this means that in the spectral analyzers, the correlated sample of the implementation must have a harmonic pseudon-generator, cosine-like, and O-pause signals of which are sent to multiplication systems simultaneously with the distribution of the discrete values of the prlitose amludus.

With uncorrelated samples, the readings of such a generator, operating regardless of the cycle of operation of the mechanism, will contribute to the error when changing the cycle. The later the onros is produced, the more the instability of the cycle of the mechanism l of the generator, especially in the higher frequency component, is affected.

Reduce computational errors for an uncorrelated MOJO sample, if the onipoca time of the polling amplitude value is counted not from the beginning of the process analysis, but, for example, from the beginning of the mechanism's cycle of operation or from another moment in the cycle, but the same in all cycles. Such moments can be obtained, for example, using the engine angle sensor.

The proposed representation of the sampling time of the croceos amplitudes means that all amplitude values can be presented on a time interval equal to the mechanism operation, in the order of increasing polling time values, counted from the same moment in different cycles of the mechanism operation. As a result, a discrete uncorrelated sample of amplitudes is already covered by a sample of correlated amplitudes (more precisely, quasi correlated). The position of each amplitude on a time interval equal to the cycle of the mechanism, determines the time value in the harmonic rest (cosine and sine) argument of the Fourier transform.

The amplitudes of uncorrelated samples and the time of their interrogation with respect to one, and the same moment in each cycle of the mechanism can be practically obtained in B (the ideals of signals recorded, for example, by a binary counter of time stamp pulses for the values of interrogation time signals and a similar counter for the amplitudes of the process sensor signals amplified and fed to the counter via the H1 converter “analog-code.

Subject invention

A method for analyzing a stationary random process, for example, the Shchuma cyclically operating mechanisms, based on an up-to-date Fourier transform spectrum, represented by a discrete sampling, characterized in that, in order to reduce the amount of computation, the amplitude values of the sipnal are remembered at the time points counted in each cycle from the ship's phase phase, in the order of increasing the memory time interval, which is proportional to the harmonic components of the Fourier transform signal azovi