Oscillation period estimation system

Abstract

Even when the periods of multiple types of vibrations are close to each other, the period of each vibration is estimated. [Solution] The vibration period estimation system comprises: a vibration detection unit that detects vibrations and converts them into analog signals; a sampling unit that samples the analog signals at predetermined intervals and converts them into digital serial signals; a plurality of serial-parallel converters (21) that sequentially convert the serial signals into parallel signals composed of data for the number of output ports, and each having a different number of output ports; a signal accumulation unit (23) that accumulates the parallel signals and calculates each accumulated signal corresponding to each serial-parallel converter; an evaluation value calculation unit (25) that calculates an evaluation value for evaluating the magnitude of each accumulated signal based on each accumulated signal; and an estimation unit (27) that, if the evaluation value is greater than a threshold, estimates that vibrations with a period obtained by multiplying the number of output ports of the serial-parallel converter for which an evaluation value greater than the threshold was calculated are included.

Description

Technical Field

[0001] The present invention relates to a system for estimating the period of vibration.

Background Art

[0002] For example, there is a device that detects abnormalities in a bearing by attaching a vibration detection element to a bearing device, sampling the detected vibration signal, obtaining an envelope, and performing frequency analysis on the envelope by FFT (Fast Fourier Transform) (see Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] By the way, in an environment where vibrations of multiple machines overlap, when the frequencies (or periods, which are the reciprocals of the frequencies) of multiple types of vibrations are close to each other, it is difficult to estimate the frequency of each vibration by frequency analysis using FFT.

[0005] The present invention has been made to solve the above problems, and its main object is to provide a vibration period estimation system capable of estimating the period of each vibration even when the periods of multiple types of vibrations are close to each other.

Means for Solving the Problems

[0006] A first means for analyzing the above problems is a vibration period estimation system, a vibration detection unit that detects vibration and converts it into an analog signal, a sampling unit that samples the analog signal input from the vibration detection unit at a predetermined period and converts it into a digital serial signal, The serial signal output by the sampling unit is sequentially converted into a parallel signal composed of data for the number of output ports, and a plurality of serial-to-parallel converters with different numbers of output ports are used. A signal accumulation unit accumulates the parallel signals sequentially input from each serial-parallel converter and calculates the accumulated signal corresponding to each serial-parallel converter. An evaluation value calculation unit calculates an evaluation value for evaluating the magnitude of each accumulated signal based on each accumulated signal calculated by the signal accumulation unit, If the evaluation value calculated by the evaluation value calculation unit is greater than a threshold, the estimation unit estimates that the vibration detected by the vibration detection unit is included in the vibration with a period obtained by multiplying the number of output ports of the serial-parallel converter calculated based on the cumulative signal by the predetermined period, when the evaluation value greater than the threshold is greater than a threshold. It is equipped with.

[0007] According to the above configuration, the vibration detection unit detects vibrations and converts them into an analog signal. The sampling unit samples the analog signal input from the vibration detection unit at predetermined intervals and converts it into a digital serial signal. Then, the multiple serial-parallel converters successively convert the serial signal output by the sampling unit into a parallel signal composed of data for the number of output ports.

[0008] Let's assume that the vibration detection unit detects vibrations of a fixed period, and that the serial signal converted from the vibration is a repeating signal of a fixed period. In this case, if the length of the serial signal converted into a single parallel signal (composed of data for the number of output ports) by the serial-to-parallel converter matches the above fixed period, the parallel signals output sequentially by the serial-to-parallel converter will be the same each time. Therefore, the cumulative signal obtained by accumulating the parallel signals sequentially input to the serial-to-parallel converter will be a signal in which the sequentially input parallel signals have been amplified (more specifically, a signal in which each data point of the parallel signal has been amplified). On the other hand, if the length of the serial signal converted into a single parallel signal by the serial-to-parallel converter does not match the above fixed period, the parallel signals output sequentially by the serial-to-parallel converter will be different each time. Therefore, the cumulative signal obtained by accumulating the parallel signals sequentially input to the serial-to-parallel converter will not be a signal in which the sequentially input parallel signals have been amplified.

[0009] Here, since the number of output ports of the multiple serial-parallel converters differs from one another, the length of the serial signal that each of the multiple serial-parallel converters converts into a single parallel signal differs from one another. For this reason, in a serial-parallel converter where the sequentially output parallel signals are amplified and the cumulative signal becomes large, the length of the serial signal converted into a single parallel signal can be said to coincide with a constant period of vibration. In other words, it can be said that vibrations with a period obtained by multiplying the number of output ports of the serial-parallel converter with a large cumulative signal by the predetermined sampling period are included in the vibration detected by the vibration detection unit. From this principle, the period of vibrations included in the vibration detected by the vibration detection unit can be estimated.

[0010] Based on the above, the evaluation value calculation unit calculates an evaluation value for evaluating the magnitude of each accumulated signal based on the accumulated signals calculated by the signal accumulation unit. The estimation unit then estimates that if the calculated evaluation value is greater than a threshold, the vibration detected by the vibration detection unit has a period obtained by multiplying the number of output ports of the serial-parallel converter calculated based on the accumulated signals by a predetermined period. Here, even if the periods of multiple types of vibrations are close to each other, the magnitude of the amplified accumulated signal obtained by sequentially inputting parallel signals is significantly different from the magnitude of the unamplified accumulated signal obtained by sequentially inputting parallel signals. Therefore, even if the periods of multiple types of vibrations are close to each other, the period of each vibration can be estimated.

[0011] To estimate the period of vibration, it is necessary to satisfy the condition that the parallel signals, when sequentially input from the serial-to-parallel converter, are amplified when accumulated. This condition is that the period obtained by multiplying the number of output ports of the serial-to-parallel converter by the predetermined sampling period for the analog signal (hereinafter also referred to as the "sampling period") matches the length of the serial signal converted into a single parallel signal (i.e., the period of vibration). Therefore, in order to satisfy the above condition for multiple types of vibration periods, it is necessary to increase the resolution of the serial signal conversion to a parallel signal, which may result in a very short sampling period. However, if the sampling period is shortened, the amount of data in the serial signal increases, which may increase the required memory capacity and the data transmission load.

[0012] In this regard, the second means includes an interpolation unit that inserts interpolation data between sample data in the serial signal output by the sampling unit and outputs an interpolated serial signal, which is the serial signal after the insertion of the interpolation data. The serial-to-parallel converter sequentially converts the interpolated serial signal output by the interpolation unit into a parallel signal composed of data for the number of output ports, instead of the serial signal output by the sampling unit. With this configuration, the serial-to-parallel converter sequentially converts the interpolated serial signal, which has interpolation data inserted between sample data in the serial signal, into a parallel signal. Therefore, when converting the interpolated serial signal into a parallel signal, not only the sample data but also the interpolation data is treated the same as the sample data, and a resolution similar to that of a sampling period in which the total number of sample data can be obtained can be realized. Thus, it is possible to improve the accuracy of estimating the period of vibration while suppressing an increase in the amount of data in the serial signal.

[0013] In the third method, the interpolated data is 0.

[0014] According to the above configuration, the interpolation unit inserts zeros as interpolation data between the sample data in the serial signal. Therefore, calculations to acquire the interpolation data are not required, and the accuracy of estimating the vibration period can be improved while reducing the computational load. Furthermore, the inventors of this application have confirmed through simulation that even when the interpolation data is set to zero regardless of the value of the sample data, the accuracy of estimating the vibration period can be improved.

[0015] In the fourth method, the interpolated data is a value obtained by linearly interpolating between consecutive sample data.

[0016] According to the above configuration, the interpolation unit inserts values ​​that linearly interpolate between consecutive sample data as interpolation data between the sample data in the serial signal. Therefore, interpolation data can be obtained by simple calculations. Furthermore, the inventors of this application have confirmed through simulation that linear interpolation between consecutive sample data can further improve the accuracy of estimating the period of vibration compared to interpolating between consecutive sample data with zero.

[0017] In the fifth method, the interpolated data is a value obtained by spline interpolation between consecutive sample data.

[0018] According to the above configuration, the interpolation unit inserts values ​​as interpolation data between consecutive sample data in the serial signal, by spline interpolating between the consecutive sample data. The inventors of this application have confirmed through simulation that when spline interpolation is performed between consecutive sample data, the accuracy of estimating the period of vibration can be further improved compared to when the consecutive sample data is interpolated with zero.

[0019] In the sixth method, the interpolation unit inserts between 2,000 to 32,000 interpolation data points in the serial signal output by the sampling unit.

[0020] The inventors of this application have confirmed through simulation that the accuracy of estimating the period of vibration can be significantly improved by inserting 2,000 to 32,000 interpolation data between sample data in the serial signal.

[0021] In the seventh method, the evaluation value calculation unit sets the evaluation value to the maximum absolute value among the absolute values ​​of each cumulative value of each data for the number of output ports in each cumulative signal.

[0022] According to the above configuration, even if there are differences in the magnitudes of a plurality of data depending on the timing of sampling the analog signal, an evaluation value is calculated based on the data with the largest absolute value among the plurality of data. Therefore, it is possible to appropriately reflect in the evaluation value whether or not the parallel signals input sequentially are amplified when accumulated.

Brief Description of the Drawings

[0023] [Figure 1] Schematic diagram showing a serial signal, a parallel signal, and an accumulated signal when the number of samples per period matches the number of output ports of the SPC. [Figure 2] Schematic diagram showing a serial signal, a parallel signal, and an accumulated signal when the number of samples per period does not match the number of output ports of the SPC. [Figure 3] Block diagram of the signal analysis unit. [Figure 4] Block diagram of the vibration period estimation system. [Figure 5] Schematic diagram showing a mode in which a plurality of types of vibrations overlap. [Figure 6] Chart showing each vibration waveform before overlapping with each other. [Figure 7] Chart showing a waveform in which each vibration waveform in FIG. 6 overlaps and noise is added. [Figure 8] Diagram showing the frequency analysis result of the serial signal by FFT which is a comparative example. [Figure 9] Diagram showing the analysis result of the serial signal according to the present embodiment. [Figure 10] Time chart showing a signal waveform and sample values when the sampling frequency is a natural multiple of the signal frequency. [Figure 11] Time chart showing a signal waveform and sample values when the sampling frequency is not a natural multiple of the signal frequency. [Figure 12] Block diagram of a modified example of the vibration period estimation system. [Figure 13] Schematic diagram showing an example of the interpolation unit in FIG. 12. [Figure 14]A chart showing the sample number and sample value of the interpolated serial signal. [Figure 15] Figure 14 shows the analysis results of the interpolated serial signal. [Figure 16] This figure shows the relationship between the number of inserted data points and RMSE when the signal-to-noise ratio (SNR) is 20 dB. [Figure 17] This figure shows the relationship between the number of inserted data points and RMSE when the signal-to-noise ratio (SNR) is 10 dB. [Figure 18] This figure shows the relationship between the number of inserted data points and RMSE when the signal-to-noise ratio (SNR) is 0 dB. [Figure 19] This figure shows the relationship between the number of inserted data points and RMSE when the signal-to-noise ratio (SNR) is -10 dB. [Modes for carrying out the invention]

[0024] The following describes one embodiment of a vibration period estimation system for estimating the period of vibrations in machinery and other equipment, with reference to the drawings.

[0025] (Analysis principle) First, the principle of analyzing vibrations using the vibration period estimation system of this embodiment will be explained.

[0026] Figure 1 shows the case where a serial signal x[k] with 8 samples per period is converted to a parallel signal x8[g] by a serial-to-parallel converter 21A with 8 output ports. The parallel signal x8[g] consists of data corresponding to the number of output ports (8). Here, the number of samples (8) included in one period of the serial signal x[k] matches the number of output ports (8) of the serial-to-parallel converter 21A (SPC). In this case, the serial-to-parallel converter 21A outputs the same parallel signal x8[g] each time. Therefore, the cumulative signal x8a, which is the sum of the parallel signals x8[g] output sequentially by the serial-to-parallel converter 21A, is a signal that is an amplified version of the sequentially output parallel signal x8[g]. More specifically, each data output by each output port 22 is accumulated and amplified for each output port 22.

[0027] Figure 2 shows the case where a serial signal x[k] with 8 samples per period is converted to a parallel signal x7[g] by a serial-to-parallel converter 21B with 7 output ports. The parallel signal x7[g] consists of data corresponding to the number of output ports (7). Here, the number of samples (8) included in one period of the serial signal x[k] does not match the number of output ports (7) of the serial-to-parallel converter 21B. In this case, the serial-to-parallel converter 21B outputs a different parallel signal x7[g] each time. Therefore, the cumulative signal x7a, which is the sum of the parallel signals x7[g] output sequentially by the serial-to-parallel converter 21B, is not an amplified signal of the sequentially output parallel signals x7[g], but rather a signal in which the sequentially output parallel signals x7[g] weaken each other. More specifically, each data output by each output port 22 is accumulated for each output port 22 and weakens each other.

[0028] Therefore, as shown in Figure 3, serial signals x[k] are converted into parallel signals x1[g] to xn[g] by serial-parallel converters 21 (SPC#1 to SPC#n) with different numbers of output ports, and the number of output ports of the serial-parallel converter 21 whose accumulated signals x1a to xna, obtained by accumulating the parallel signals x1[g] to xn[g] by signal accumulation units 23 (#1 to #n), exceeds a threshold, matches the number of samples corresponding to one period of the periodic signal contained in serial signal x[k]. Thus, the period of the periodic signal (number of output ports × sampling period) can be estimated.

[0029] (Oscillation period estimation system) Next, we will explain the vibration period estimation system that estimates the vibration period using the analysis principle described above.

[0030] As shown in Figure 4, the vibration period estimation system 10 includes a vibration detection unit 11, a sampling unit 13, a filter 15, and a signal analysis unit 20.

[0031] The vibration detection unit 11 detects vibrations, converts them into analog signals A, and outputs them to the sampling unit 13. The vibration detection unit 11 may detect vibrations as acceleration, may detect vibrations in a single axis, or it may detect vibrations in mutually orthogonal X, Y, and Z axis directions and output analog signals Ax, Ay, and Az representing the vibrations in the X, Y, and Z axis directions as analog signals A to the sampling unit 13.

[0032] Acceleration is one example of information representing vibration. The vibration detection unit 11 only needs to be able to detect vibration; it is not limited to detecting vibration by acceleration. For example, the vibration detection unit 11 may detect vibration by velocity or by displacement.

[0033] The sampling unit 13 samples the input analog signal A, converts it into a digital serial signal x[k], and outputs it to the filter 15. The sampling unit 13 is configured, for example, as an analog-to-digital converter (ADC). The sampling unit 13 samples the analog signal A at sampling periods T (corresponding to a predetermined period).

[0034] Filter 15 removes noise from the serial signal x[k]. For example, filter 15 removes frequency components that are clearly outside the frequency range of the vibration being analyzed, as well as DC components that are not part of the analysis, from the serial signal x[k]. Note that filter 15 is not mandatory and can be omitted.

[0035] The signal analysis unit 20 receives the serial signal x[k] that has passed through the filter 15 and estimates the period of the periodic signal contained in the serial signal x[k] using the above analysis principle.

[0036] (Signal analysis section) Next, the signal analysis unit 20 will be described in detail. As shown in Figure 3, the signal analysis unit 20 includes a serial-to-parallel converter 21 (SPC#1 to SPC#n), a signal accumulation unit 23 (#1 to #n), an evaluation value calculation unit 25, and an estimation unit 27. The signal analysis unit 20 is composed of electronic circuits, a microcontroller (including hardware and software), and the like.

[0037] The number of output ports of the multiple serial-parallel converters 21 (SPC#1 to SPC#n) differs from one another; for example, the number of output ports increases by one from serial-parallel converter 21 (SPC#1) to serial-parallel converter 21 (SPC#n). Each serial-parallel converter 21 (SPC#1 to SPC#n) converts the serial signal x[k] into parallel signals x1[g] to xn[g] and outputs them to the signal accumulation unit 23.

[0038] The signal accumulation units 23 (#1 to #n) correspond to the serial-to-parallel converters 21 (SPC#1 to SPC#n), respectively. The signal accumulation units 23 (#1 to #n) accumulate the input parallel signals x1[g] to xn[g], respectively, and output the accumulated signals x1a to xna to the evaluation value calculation unit 25, respectively.

[0039] The evaluation value calculation unit 25 calculates an evaluation value to evaluate the magnitude of each cumulative signal x1a to xna based on the input cumulative signals x1a to xna. Specifically, the evaluation value calculation unit 25 uses the absolute value of the maximum of the absolute values ​​of each cumulative value of each data for each output port as the evaluation value for each cumulative signal x1a to xna. For example, in the example shown in Figure 1, the evaluation value of cumulative signal x8a is the absolute value of cumulative value Pa8, which is the maximum of the absolute values ​​of each cumulative value Pa1 to Pa8 at each output port.

[0040] The estimation unit 27 estimates that if the evaluation value calculated by the evaluation value calculation unit 25 is greater than the threshold Th, then vibrations with a period obtained by multiplying the number of output ports of the serial-parallel converter (calculated based on the cumulative signal) by the sampling period T are included in the vibrations detected by the vibration detection unit 11. In the example shown in Figure 1, the evaluation value, which is the absolute value of the cumulative value Pa8, is greater than the threshold Th, so it is estimated that vibrations with a period obtained by multiplying the number of output ports of the serial-parallel converter 21A (8) by the sampling period T (8T) are included in the vibrations detected by the vibration detection unit 11. In the example shown in Figure 2, the evaluation value, which is the absolute value of the cumulative value Pb1, is less than the threshold Th, so it is estimated that vibrations with a period obtained by multiplying the number of output ports of the serial-parallel converter 21B (7) by the sampling period T (7T) are not included in the vibrations detected by the vibration detection unit 11.

[0041] (Analysis example) Next, a specific analysis example will be explained. Figure 5 is a schematic diagram showing a configuration in which multiple types of vibrations overlap. Machines M1, M2, and M3, such as rotating machinery, each generate vibrations S1(t), S2(t), and S3(t) with their natural periods, respectively. In this work environment, vibrations S1(t), S2(t), and S3(t) overlap, and noise is added to form vibration x(t).

[0042] Figure 6 is a chart showing each vibration waveform before they overlap. The horizontal axis represents discrete time, which corresponds to the sample number. The vertical axis represents the digital serial signals S1[k], S2[k], S3[k] obtained by detecting vibrations S1(t), S2(t), and S3(t) respectively by the vibration detection unit 11 and converting them into analog signals A, and then sampling the analog signals A at sampling periods T by the sampling unit 13. In this example, the periods of serial signals S1[k], S2[k], and S3[k] are 30, 31, and 32 discrete times, respectively.

[0043] Figure 7 is a chart showing the waveforms obtained by overlapping the vibration waveforms in Figure 6 and adding noise. The horizontal axis represents discrete time, corresponding to the sample number. The vertical axis represents the digital serial signal x[k] obtained by detecting the vibration x(t) by the vibration detection unit 11 and converting it into an analog signal A, and then sampling the analog signal A at sampling periods T by the sampling unit 13 and converting it.

[0044] Figure 8 shows the frequency analysis results of a serial signal x[k] using FFT (Fast Fourier Transform) as a comparative example. No characteristic frequency peaks were detected, indicating that with FFT frequency analysis, it is difficult to estimate the period of each vibration when the periods (or reciprocal frequencies) of multiple types of vibrations are close to each other.

[0045] Figure 9 shows the analysis results of the serial signal x[k] according to this embodiment. The horizontal axis represents the number of output ports of the serial-parallel converter 21. The vertical axis represents the evaluation values ​​corresponding to each serial-parallel converter 21 with a different number of output ports. The magnitude of the evaluation values ​​is normalized. The evaluation values ​​corresponding to the serial-parallel converters 21 with 30, 31, and 32 output ports, respectively, are all greater than the threshold Th. Therefore, the estimation unit 27 estimates that the vibrations detected by the vibration detection unit 11 include vibrations with a period obtained by multiplying the number of output ports (30) by the sampling period T, vibrations with a period obtained by multiplying the number of output ports (31) by the sampling period T, and vibrations with a period obtained by multiplying the number of output ports (32) by the sampling period T.

[0046] The embodiment described in detail above has the following advantages.

[0047] Since the number of output ports of the multiple serial-parallel converters 21 (SPC#1~SPC#n) are different from each other, the lengths of the serial signals x[k] that each serial-parallel converter 21 (SPC#1~SPC#n) converts to parallel signals x1[g]~xn[g] are different from each other. For this reason, in a serial-parallel converter 21 where the sequentially output parallel signals x1[g]~xn[g] are amplified and the cumulative signals x1a~xna become large, the length of the serial signal x[k] that is converted to a single parallel signal can be said to match the constant period of the vibration. In other words, it can be said that the vibration x(t) detected by the vibration detection unit 11 has a period of vibration obtained by multiplying the number of output ports of the serial-parallel converter 21 where the cumulative signals x1a~xna have become large by the sampling period T. From this principle, the periods of vibrations S1(t), S2(t), and S3(t) included in the vibration x(t) detected by the vibration detection unit 11 can be estimated.

[0048] The evaluation value calculation unit 25 calculates an evaluation value to evaluate the magnitude of each accumulated signal x1a to xna based on each accumulated signal x1a to xna calculated by the signal accumulation unit 23 (#1 to #n). Then, the estimation unit 27 estimates that if the calculated evaluation value is greater than the threshold Th, vibrations with periods 30T, 31T, and 32T, calculated by multiplying the number of output ports (30, 31, 32) of the serial-parallel converter 21 calculated based on the accumulated signals by the sampling period T, are included in the vibration x(t) detected by the vibration detection unit 11. Here, even if the periods of multiple types of vibrations S1(t), S2(t), and S3(t) are close to each other, the magnitude of the accumulated signal when the sequentially input parallel signals x1[g] to xn[g] are amplified is significantly different from the magnitude of the accumulated signal when the sequentially input parallel signals x1[g] to xn[g] are not amplified. Therefore, even when the periods of multiple types of vibrations S1(t), S2(t), and S3(t) are close to each other, it is possible to estimate the periods 30T, 31T, and 32T of each vibration S1(t), S2(t), and S3(t).

[0049] The evaluation value calculation unit 25 uses the absolute value of the maximum among the absolute values ​​of each cumulative value (e.g., Pa1 to Pa8) of each data for each output port in each cumulative signal x1a to xna as the evaluation value. With the above configuration, even if there is a difference in the magnitude of multiple data depending on the timing of sampling the analog signal A, the evaluation value is calculated based on the data with the largest absolute value among the multiple data. Therefore, it is possible to appropriately reflect in the evaluation value whether or not amplification occurs when the sequentially input parallel signals x1[g] to xn[g] are accumulated.

[0050] Furthermore, the above embodiment can also be implemented with the following modifications. Parts identical to those in the above embodiment are denoted by the same reference numerals, and their descriptions are used accordingly.

[0051] To estimate the periods of vibrations S1(t), S2(t), and S3(t) contained in vibration x(t), it is necessary to satisfy the condition that when the parallel signals x1[g] to xn[g] sequentially input from the serial-parallel converter 21 (SPC#1 to SPC#n) are accumulated, these parallel signals x1[g] to xn[g] are amplified. This condition is that the period obtained by multiplying the number of output ports of the serial-parallel converter 21 (SPC#1 to SPC#n) by the sampling period T matches the length of the serial signal x[k] converted into each parallel signal x1[g] to xn[g] (i.e., the periods of vibrations S1(t), S2(t), and S3(t)).

[0052] Figure 10 is a time chart showing the signal waveform and sample values ​​when the sampling frequency (4 Hz) is a natural number multiple of the signal frequency (1 Hz). In this case, if the number of output ports is 4, the period (1 second) obtained by multiplying the number of output ports (4) by the sampling period T (1 / 4 second) matches the length (1 second) of the serial signal x[k] converted into a single parallel signal, i.e., the signal period.

[0053] On the other hand, Figure 11 is a time chart showing the signal waveform and sample values ​​when the sampling frequency (4 Hz) is not a natural number multiple of the signal frequency (1.2308 Hz). In this case, for example, the period (1 second) obtained by multiplying the number of output ports (4) by the sampling period T (1 / 4 second) does not match the length of the serial signal x[k] converted into a single parallel signal (13 / 16 seconds), i.e., the period of the signal.

[0054] Therefore, in order to satisfy the above conditions for the periods of multiple types of vibrations S1(t), S2(t), and S3(t), it is necessary to increase the resolution of converting the serial signal x[k] to parallel signals x1[g] to xn[g], which may result in a very short sampling period T. However, if the sampling period T is shortened, the amount of data in the serial signal x[k] will increase, which may increase the required memory capacity and the data transmission load.

[0055] Therefore, as shown in Figure 12, the vibration period estimation system 10 may include an interpolation unit 17 that inserts interpolation data between sample data in the serial signal x[k] output by the sampling unit 13 and outputs an interpolated serial signal xi[k], which is the serial signal after the insertion of the interpolation data. The serial-to-parallel converter 21 (SPC#1~SPC#n) then sequentially converts the interpolated serial signal xi[k] output by the interpolation unit 17 into parallel signals x1[g]~xn[g] composed of data for the number of output ports, instead of the serial signal x[k] output by the sampling unit 13. Specifically, as shown in Figure 13, an interpolation unit 17A can be adopted as the interpolation unit 17, which inserts 0 as interpolation data di (white circle) between sample data d[k] (black circle) in the serial signal x[k].

[0056] According to the above configuration, the serial-to-parallel converter 21 (SPC#1~SPC#n) sequentially converts the interpolated serial signal xi[k], obtained by inserting interpolation data di between sample data d[k] in the serial signal x[k], into parallel signals x1[g]~xn[g]. Figure 14 is a chart showing the sample number (k) and sample value of the interpolated serial signal. In this example, three zeros are inserted as interpolation data di between sample data d[k]. Therefore, when converting the interpolated serial signal xi[k] into parallel signals x1[g]~xn[g], not only the sample data d[k] but also the interpolation data di is treated the same as the sample data d[k], and a resolution similar to the sampling period T (1 / 16 second) can be achieved, which yields the total number of sample data. In this example, the sampling frequency (16Hz) is a natural number multiple (13 times) of the signal frequency (16 / 13Hz). Therefore, when the number of output ports is 13, the period (13 / 16 seconds) obtained by multiplying the number of output ports (13) by the sampling period T (1 / 16 seconds) matches the length (13 / 16 seconds) of the serial signal x[k] converted into a single parallel signal, i.e., the period of the signal (13 / 16 seconds). As a result, as shown in Figure 15, the evaluation value corresponding to the serial-parallel converter 21 with 13 output ports becomes larger than the threshold Th, and the estimation unit 27 can estimate that vibrations with a period (13 / 16 seconds) obtained by multiplying the number of output ports (13) by the sampling period T (1 / 16 seconds) are included in the vibrations detected by the vibration detection unit 11. Thus, the accuracy of estimating the vibration period can be improved while suppressing an increase in the amount of data in the serial signal x[k].

[0057] Furthermore, the interpolation unit 17A inserts 0 as interpolation data di between the sample data d[k] in the serial signal x[k]. Therefore, calculations to obtain the interpolation data di are not required, and the accuracy of estimating the vibration period can be improved while reducing the computational load. The inventors of this application have confirmed through simulations described later that even when the interpolation data di is set to 0 regardless of the value of the sample data d[k], the accuracy of estimating the vibration period can be improved. It should be noted that a constant other than 0 can also be used as the interpolation data di.

[0058] As shown in Figure 13, an interpolation unit 17B can also be adopted as the interpolation unit 17, which inserts a value that linearly interpolates between consecutive sample data d[k] and d[k+1] as interpolation data di (white circle) between sample data (black circle) in the serial signal x[k]. That is, the interpolation data di is a point on the straight line connecting sample data d[k] and sample data d[k+1]. With the above configuration, the interpolation data di can be obtained by simple calculation. Furthermore, the inventors of this application have confirmed through simulations described later that linear interpolation between consecutive sample data d[k] and d[k+1] can further improve the accuracy of estimating the period of vibration compared to interpolating between consecutive sample data d[k] and d[k+1] with 0.

[0059] As shown in Figure 13, an interpolation unit 17C can also be adopted as the interpolation unit 17, which inserts a value as interpolation data di (white circle) between sample data (black circles) in the serial signal x[k], which is a value that spline-interpolates between consecutive sample data d[k] and d[k+1]. That is, the interpolation data di is a point on the spline curve connecting sample data d[k] and sample data d[k+1]. The inventors of this application have confirmed through simulations described later that linear interpolation between consecutive sample data d[k] and d[k+1] can further improve the accuracy of estimating the period of vibration compared to interpolating between consecutive sample data d[k] and d[k+1] with 0.

[0060] Figures 16-19 show the relationship between the number of interpolated data points di inserted between consecutive sample data points d[k] and d[k+1] and the Root Mean Squared Error (RMSE) for signal-to-noise ratios (SNR) of 20dB, 10dB, 0dB, and -10dB, respectively. The Root Mean Squared Error is the square root of the mean square of the difference between the estimated and true values ​​of the vibration period. For SNRs of 20dB, 10dB, 0dB, and -10dB, the magnitude of the vibration is 100 times, 10 times, 1 time, and 1 / 10 times, respectively, compared to the magnitude of the noise.

[0061] Here, the simulation is performed under the following conditions: sampling frequency = 50Hz, serial signal x[k] frequency = 1.7Hz, serial signal x[k] waveform = sine wave, number of sample data constituting the serial signal x[k] = 1024, and number of oscillation period estimations when calculating the mean square error = 100. Zero-insertion uses interpolation unit 17A, Linear-interpolation uses interpolation unit 17B, Spline-interpolation uses interpolation unit 17C, and FFT uses frequency analysis by Fast Fourier Transform. Note that in FFT, instead of inserting interpolation data di between consecutive sample data d[k], d[k+1], a common technique to improve the resolution of the frequency spectrum is to add the same number of zeros as the interpolation data di to the end of the serial signal x[k] (Zero Padding).

[0062] As shown in Figures 16-19, for all signal-to-noise ratios (SNR) = 20dB, 10dB, 0dB, and -10dB, when the number of inserted data points is 0, i.e., when interpolation data is not inserted, the vibration period estimation system 10 exhibits a smaller mean square error than frequency analysis using FFT. Furthermore, for all SNRs = 20dB, 10dB, 0dB, and -10dB, when the number of inserted data points is 8000 or less, the vibration period estimation system 10 exhibits a smaller mean square error than frequency analysis using FFT. In addition, for all interpolation units 17A-17C, when the number of inserted data points is between 2000 and 32000, the mean square error is significantly reduced compared to when no interpolation data is inserted (when the number of inserted data points is 0). In particular, for signal-to-noise ratios of 20 dB and 10 dB as shown in Figures 16 and 17, when the number of inserted data points is between 2,000 and 32,000, the mean square error is smaller in all interpolation units 17A to 17C than in frequency analysis using FFT.

[0063] The evaluation value calculation unit 25 can also use the sum of the squares of the cumulative values ​​of each data for each output port in each cumulative signal x1a to xna as the evaluation value. For example, in the example shown in Figure 1, the sum of Pa1^2 to Pa8^2 is used as the evaluation value. Pa1^2 is the square of Pa1. With this configuration, even if there are differences in the magnitude of multiple data due to the timing of sampling the analog signal A, the effect of these differences can be further suppressed. In addition, the evaluation value can also be the value that is the largest among the variances of each cumulative value, the value that is the largest among the mean of the absolute values ​​of each cumulative value, the value that is the largest among the standard deviations of each cumulative value, the value that is the largest among the median of the absolute values ​​of each cumulative value, etc.

[0064] Alternatively, instead of the signal accumulation units 23 (#1~#n), a single signal accumulation unit can be used that accumulates the parallel signals x1[g]~xn[g] output by the serial-parallel converters 21 (SPC#1~SPC#n) and calculates the respective accumulated signals x1a~xna corresponding to each serial-parallel converter 21 (SPC#1~SPC#n).

[0065] The vibrations analyzed by the vibration period estimation system 10 are not limited to standing waves; they may also be non-standing waves. When analyzing non-standing wave vibrations, the serial signal x[k] can be divided into short time intervals (windows), and the vibration period can be estimated for each interval.

[0066] The vibrations to be analyzed are not limited to vibrations generated by machinery such as rotating equipment, but can be any periodic vibration, such as vibrations generated by a robot arm, human heartbeat vibrations, brainwave vibrations, sound vibrations, air pressure vibrations, electromagnetic wave vibrations, and light vibrations as waves.

[0067] Furthermore, the above embodiments and their respective modifications can be combined and implemented to the extent possible. [Explanation of symbols]

[0068] 10...Vibration period estimation system, 11...Vibration detection unit, 13...Sampling unit, 17...Interpolation unit, 17A...Interpolation unit, 17B...Interpolation unit, 17C...Interpolation unit, 20...Signal analysis unit, 21...Serial-to-parallel converter, 22...Output port, 23...Signal accumulation unit, 25...Evaluation value calculation unit, 27...Estimation unit.

Claims

1. A vibration detection unit that detects vibrations and converts them into analog signals, A sampling unit that samples the analog signal input from the vibration detection unit at predetermined intervals and converts it into a digital serial signal, The serial signal output by the sampling unit is sequentially converted into a parallel signal composed of data for the number of output ports, and a plurality of serial-to-parallel converters with different numbers of output ports are used. A signal accumulation unit accumulates the parallel signals sequentially input from each serial-parallel converter and calculates the accumulated signal corresponding to each serial-parallel converter. An evaluation value calculation unit calculates an evaluation value for evaluating the magnitude of each accumulated signal based on each accumulated signal calculated by the signal accumulation unit, If the evaluation value calculated by the evaluation value calculation unit is greater than a threshold, the estimation unit estimates that the vibration detected by the vibration detection unit is included in the vibration with a period obtained by multiplying the number of output ports of the serial-parallel converter calculated based on the cumulative signal by the predetermined period, when the evaluation value greater than the threshold is greater than a threshold. A vibration period estimation system equipped with the following features.

2. The system includes an interpolation unit that inserts interpolation data between sample data in the serial signal output by the sampling unit and outputs an interpolated serial signal, which is the serial signal after the insertion of the interpolation data. The vibration period estimation system according to claim 1, wherein the serial-to-parallel converter sequentially converts the interpolated serial signal output by the interpolation unit into a parallel signal composed of data for the number of output ports, instead of the serial signal output by the sampling unit.

3. The vibration period estimation system according to claim 2, wherein the interpolation data is 0.

4. The vibration period estimation system according to claim 2, wherein the interpolated data is a value obtained by linearly interpolating between consecutive sample data.

5. The vibration period estimation system according to claim 2, wherein the interpolated data is a value obtained by spline interpolation between consecutive sample data.

6. The vibration period estimation system according to any one of claims 2 to 5, wherein the interpolation unit inserts 2,000 to 32,000 interpolation data between the sample data in the serial signal output by the sampling unit.

7. The vibration period estimation system according to any one of claims 2 to 5, wherein the evaluation value calculation unit takes the absolute value that is the maximum of the absolute values ​​of the cumulative values ​​of each data for the number of output ports as the evaluation value for each cumulative signal.

Images