RU2331893C1 - Method of discrete component separation in signal spectre and device for its implementation - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: this group of inventions concerns instrument technology and can be applied in noise source detection in machines, mechanisms and systems, as well as construction elements located on vehicles during vehicle diagnostics. The method of discrete component separation in a signal spectre involves obtaining of the signal threshold level and separation of signal spectre maximums, so that before signal spectre maximum separation a smoothed continuous subspectrum is separated, then the spectre base level is defined as an amount of the said smoother continuous subspectrum and the spectre threshold level, and as the discrete signal spectre maximums those spectre readings with maximum level are taken, which belong to the groups of close spectral components exceeding the said base level. The device for separation of discrete signal components in a signal spectre includes narrowband spectre analyser, spectre maximum separation unit, threshold level obtaining unit, smoothed continuous subspectrum separation unit, and adder, interconnected in a proper pattern.

EFFECT: extended frequency range.

2 cl, 5 dwg

Description

The invention relates to the field of instrumentation and can be used to identify noise sources of machines, mechanisms and systems, as well as structural elements located, for example, in automobile or railway transport, as well as on ships for various purposes during their diagnostic examination.

The purpose of the diagnostic examination of complex technical structures is, in particular, the allocation of discrete components of the spectrum of noise or vibration, which are a diagnostic sign of the causes of increased vibration or noise, i.e. automated detection of discrete spectral components in the spectrum of an acoustic signal having a mixed character and determining their levels, frequencies and excesses over the solid part.

A known method for determining the average frequency of the spectrum, based on the relationship between the values of the instantaneous frequency of the signal and the frequencies of the spectrum (see Fink L.M. Signals, interference, errors ... - M .: Communication, 1978, pp. 96-103). The essence of the method consists in the fact that a narrow-band signal is formed in the studied acoustic signal by filtering in a frequency band including the studied discrete spectral component. After filtering the signal under study, an analytical signal is generated, which is used to estimate the instantaneous phase and then to estimate the instantaneous frequency by differentiating the instantaneous phase (see, for example, A. Sergienko, Digital Signal Processing. - St. Petersburg: Peter, 2003, p. 448-449). The average value of the instantaneous frequency corresponds to the frequency of the discrete spectral component, for which its level is determined. The disadvantage of this method is its low noise immunity due to the fact that, as a result of filtering, within the filter band, in addition to the useful signal, in the vicinity of the frequency of the discrete component under study, an interference signal is also observed due, for example, to the operation of extraneous mechanisms.

The closest in technical essence to the proposed solution is a method based on the allocation of the maximums of the “generalized momentum" - the method of selected points (see. Rosenberg V.Ya. Radio engineering methods for measuring parameters of processes and systems. - M.: Publishing house of the committee of standards, measures and measuring instruments ..., 1970, pp. 135-137). This method is applicable for impulse functions with an arbitrary argument, i.e. functions describing processes in the time, frequency or spatial domains. In the case under consideration, this method is used to analyze processes in the frequency or spectral region, i.e. to highlight discrete spectral components, and is considered as a prototype. The essence of the prototype method is reduced to the following basic operations:

- produce a narrow-band spectral analysis of the signal due to vibration or noise of the investigated source;

- set the threshold level, which determines the range of values of the argument (frequency band), within which there is a distinguished discrete spectral component;

- they search for the position of the maximum within a certain frequency band - determine the frequency value of the discrete spectral component (for simplicity of exposition of the essence of the prototype method, the operations of amplification, limitation, differentiation, etc., set forth in the primary source, are omitted as secondary operations).

The disadvantage of the prototype method is that its scope is limited to a narrow frequency band. This is due to the fact that for the implementation of the method it is necessary to first determine the threshold level, and therefore, to specify a narrow frequency band in which the discrete spectral component is concentrated. However, the spectrum of real acoustic processes is mixed, i.e. represents the sum of discrete spectral components (with different frequencies and levels) and the continuous part of the spectrum, the level of which depends on the frequency. Therefore, it is fundamentally impossible to set a single threshold level for all discrete components contained in the spectrum of the acoustic signal. Thus, the prototype method does not provide the allocation of discrete components of the spectrum of acoustic signals in a wide frequency range.

A device is known for determining the narrow-band spectrum of a signal (PULSE ™, Multi-analyzer System Type 3560, Sound & Vibration Cataloque, "Bruel & Kjaer", p.66-67, Naerum, Denmark, 1999), in which the discrete components of the spectrum are searched using a marker designed to determine the level of the discrete component and its frequency. The disadvantage of this device is that the definition of these parameters of the narrow-band spectrum of the signal is carried out by the operator visually, i.e. based on a subjective assessment of the excess of the discrete component over the continuous part of the narrowband spectrum, and, therefore, the described device also does not allow to select the discrete components of the signal spectrum.

The objective of the proposed method for separating discrete signal and device remaining in the spectrum for its implementation is to expand the frequency range in which the maximums of the spectrum are extracted.

The specified technical result is achieved by the fact that in the method for extracting discrete components in the signal spectrum, including narrow-band spectral analysis of the signal, generating a threshold level of the spectrum and extracting the maximums of the signal spectrum, before highlighting the maximums of the signal spectrum, a smoothed solid part of the spectrum is preliminarily selected, which is then summed with the threshold level spectrum.

As a result of the summation, a reference level is formed, which is used to search for the maximum values of the spectrum of spectrum values that exceed the found reference level, i.e. to determine the frequencies of detected discrete spectral components and determine their levels.

A device for extracting discrete components in the spectrum of the signal, including a narrow-band spectrum analyzer, the signal output of which is connected to the input of the spectrum maximum extraction unit, and the control signal output of the spectrum analyzer is connected to the input of the threshold level generating unit, a smoothing solid spectrum extraction unit is introduced, the input of which connected to the signal output of the spectrum analyzer and the input of the unit for selecting the maximums of the spectrum, as well as an adder having at least two inputs. One of the inputs of the adder is connected to the output of the threshold level generating unit, and the other input of the adder is connected to the output of the smoothing solid spectrum extraction unit, while the adder output is connected to the control input of the spectrum maximum selection unit.

Selecting the smoothed solid part of the signal spectrum and summing it with a threshold level allows you to create a comparison level necessary to determine the parameters of the discrete components to be distinguished.

Introduction to the device for extracting discrete components of the signal spectrum of a series-connected smoothed solid part of the spectrum and the adder, provides an automated search for the levels and frequencies of the narrow-band spectrum that exceed the comparison level obtained by summing the smoothed solid part of the spectrum and the threshold level.

The invention is illustrated by drawings, where

figure 1 presents a block diagram of a device for separating discrete components in the spectrum of a signal that implements the proposed method;

figure 2 presents the spectrogram of the signal, the spectrum of which is mixed;

figure 3 presents a spectrogram of the signal explaining the principle of separation of discrete spectral components in the narrow-band spectrum of the signal;

figure 4 presents a spectrogram of the vibration of the mechanism;

figure 5 presents a table with a list of selected discrete components and an indication of their parameters - frequency, level and excess over the solid part of the spectrum.

A device for extracting discrete components in the signal spectrum includes a narrow-band spectrum analyzer 1, the signal output of which is connected to the input of the spectrum maximum pickup unit 2, and the control signal output of the spectrum analyzer 1 is connected to the input of the threshold level generating unit 3. The input of the smoothing solid spectrum selection unit 3 connected to the signal output of the spectrum analyzer 1 and the input of the maximum spectrum allocation unit 2.

One of the inputs of the adder 5, which has at least two inputs, is connected to the output of the threshold level forming unit 3, and the other input of the adder 5 is connected to the output of the smoothing solid spectrum extraction unit 4. The output of adder 5 is connected to the control input of the spectrum maximizing unit 2. The output of the maximum spectrum allocation unit 2 is the output of the device as a whole.

In accordance with the described device, the proposed method is implemented as follows.

, где N - число усреднений спектра и, следовательно, обуславливает надежность выделения дискретных спектральных составляющих на фоне сплошной части спектра. При этом снижается дисперсия сплошной части спектра.The input signal, due, for example, to the vibration of the mechanism, is fed to the input of a narrow-band spectrum analyzer 1 (see figure 1). As the source of the input signal, any measuring transducer can be considered - a microphone, hydrophone, vibration acceleration sensor, etc. The input signal is the sum of tonal (quasitonal) signals at the frequencies of discrete spectral components, due to the operation of vibration or noise sources, and an interference signal with a continuous spectrum. The broadening of the spectrum at frequencies of discrete spectral components due to the intrinsic instability of their sources is assumed to be insignificant. An example of an input signal spectrum is shown in FIG. When implementing the proposed method, a constant signal is generated in the threshold level 3 generation unit, the level of which is determined by the number of averaging spectra generated by the narrow-band spectrum analyzer 1. Information on the selected number of spectrum averaging is supplied to the input of the threshold level 3 forming unit from the control signal output of the narrow-band spectrum analyzer 1. When forming a threshold level, this information determines a random error in the measurement of the spectrum

, where N is the number of averaging of the spectrum and, therefore, determines the reliability of the selection of discrete spectral components against the background of the solid part of the spectrum. In this case, the dispersion of the continuous part of the spectrum decreases.

The threshold level in discrete spectral components can be taken into account only after determining the levels of spectrum fragments in the vicinity of the discrete spectral components. In this case, discrete spectral components are excluded from the mixed spectrum of the signal under study. This operation is performed in the block selection smoothed solid part of the spectrum 4, the input of which from the output of the spectrum analyzer 1 receives a signal corresponding to the spectrum of the input signal. Actually, the solid part of the spectrum is selected by filtering (smoothing) the signal corresponding to the spectrum of the input signal.

The signal proportional to the solid part of the spectrum, taken from the output of the block for separating the smoothed solid part of the spectrum 4, and the threshold level signal generated in the block for generating the threshold level 3, are fed to the corresponding inputs of the adder 5, at the output of which a reference level is formed. The reference level supplied to the control input of the unit for selecting the maximums of spectrum 2 is used for comparison with the levels of detected spectral components. If the spectral levels of the studied signal of the reference level are exceeded, a spectral count with a maximum level is determined from several closely spaced spectral components. It should be noted that in digital spectral analysis, each discrete spectral component can be represented by several spectral samples, and the number of spectral samples is determined by the weight function used. In the block for extracting the maxima of spectrum 2, taking into account the sampling frequency of the input signal input to the narrowband analyzer of spectrum 1, the frequencies of the selected discrete spectral components are also determined. Figure 3 illustrates the principle of separation of the solid part of the spectrum for the sum of tonal signals forming the corresponding discrete components of different levels and colored noise, the level of which depends on the frequency. It also shows the nature of the formed reference level, relative to which the discrete spectral components are extracted.

An example of a specific implementation.

The reliability of the proposed method is confirmed by the results of the following model experiment.

A signal was formed, which is the sum of 5 unequal in amplitude signals of tonal signals at frequencies of 19, 35, 192, 207 and 256 Hz. This signal was summed with noise having a normal distribution. Previously, the noise was subjected to random amplitude weighing, which leads to the dependence of the noise level on frequency. Thus, a signal was simulated whose spectrum contained discrete spectral components with parameters known a priori. The memory size of the narrow-band spectrum analyzer was 2048, and the sampling frequency was 600 Hz. Comparison of the parameters of the selected discrete spectral with a priori given parameters of the tonal signals showed a confident selection of discrete components in the spectrum of the model signal when implementing the proposed method.

An example of the application of the method for separating discrete spectral components in the narrow-band spectrum of a natural signal is shown in Fig. 4 in the form of a spectrogram of a vibration signal, in which the numbers indicate the discrete components highlighted in the narrow-band spectrum. Figure 5 shows a table in which, in accordance with the numbers of the selected discrete components of the spectrum, their parameters are given - frequencies, levels and excesses over the solid part of the spectrum.

When implementing the proposed method, the vibration signal of the studied mechanism in the frequency range of 5-1000 Hz was considered as the signal under investigation. Spectral analysis was performed with a resolution of 0.5 Hz. The number of spectrum averagings was 11. To increase the reliability of discrete spectral components separation, the threshold level for comparing the comparison signal was chosen equal to 6 dB.

The measurements and modeling showed that the use of new operations in accordance with the proposed method allowed us to detect discrete spectral components and determine their parameters in the entire frequency range in which the acoustic signal is analyzed. Note that in the prototype method, the detection of discrete components is practically carried out visually by the results of narrow-band spectral analysis, and the determination of their parameters is only in the vicinity of the detected visually discrete spectral components.

Compared with the prototype method, the expansion of the working frequency band allows you to automate the process of selecting discrete spectral components and determining their parameters. Automation of this process reduces the processing time of the measurement data and, therefore, reduces labor costs when conducting field tests of complex objects - workshops of industrial enterprises, vehicles, etc.

Claims (2)

1. A method of extracting discrete components in the signal spectrum, including narrow-band spectral analysis of the signal, generating a threshold level of the spectrum and extracting the maximums of the spectrum of the signal, characterized in that before highlighting the maximums of the spectrum of the signal, a smoothed solid portion of the spectrum is preliminarily selected, then the reference spectrum level is determined as the sum of the indicated the smoothed solid part of the spectrum with a threshold level of the spectrum, and those spectral samples with maxima are selected as discrete maxima of the signal spectrum level, which are part of groups of closely spaced spectral components that exceed the specified reference level. 2. A device for extracting discrete components in the spectrum of the signal, including a narrow-band spectrum analyzer, the signal output of which is connected to the input of the spectrum maximum extraction unit, and the output of the control signal of the spectrum analyzer is connected to the input of the threshold level generating unit, characterized in that the extraction unit is introduced into it smoothed solid part of the spectrum, the input of which is connected to the signal output of the spectrum analyzer and the input of the spectrum maximum extraction unit, as well as an adder having at least two inputs s, one of which is connected to the output unit for generating a threshold level and the other input of the adder coupled to the output selection unit smoothed continuous part of the spectrum, the output of the adder is connected to the control input allocating spectrum maxima block.

Images