JP2005274320A - Signal processing method and signal processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To precisely and quickly predict a convergence value of a sensor output. <P>SOLUTION: A sensor output signal y(t) is converted into a digital original signal sequence y(k) by over sampling (S1) to detect a signal sequence Δy(k) of an alternating current component included in the original signal sequence (S2). A phase and an amplitude of the original signal sequence is corrected to be conformed with those of the signal sequence of the alternating current component, the both corrected signal sequences are added or one thereof is subtracted from the other to detect a signal sequence z(k) of a direct current component in the original signal sequence, and the direct current component is thinned and high-frequency-shielded to find a convergence prediction value M' of the sensor output signal. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a technique for quickly and accurately predicting a convergence value from a signal output from a sensor in a transient state.

  Many sensors for detecting various physical quantities show a transient response to changes in the physical quantities.

  For example, a sensor for detecting the mass of an article, such as a load cell, is deformed by receiving the load of the article and outputs a voltage signal corresponding to the amount of deformation, but the load of the article on the sensor is abruptly performed. In this case, since the natural vibration mode of the sensor system is excited and transmitted to the sensor, the output signal y (t) exhibits non-linear vibration as shown in FIG.

  This non-linear vibration of the output signal is attenuated as time passes and finally converges to a constant value corresponding to the mass M of the article. However, when the mass inspection of the article is continuously performed using a line or the like, If this vibration is completely converged, efficient inspection cannot be performed.

  For this reason, conventionally, as shown in FIG. 9, the output signal y (t) from the sensor 1 is input to the low-pass filter 2 and the vibration component (transient fluctuation) included in the output signal y (t). Many methods have been employed in which the mass of the article is detected from the signal y (t) ′ output from the low-pass filter 2 as shown in FIG.

  However, since the vibration frequency of the transient response of the sensor 1 attached to the mechanical rigid body is as low as several Hz to several tens Hz, the high-frequency cutoff frequency of the low-pass filter 2 must be set very low accordingly. Rather, the time constant becomes very large.

  Therefore, as shown in FIG. 8B, the output signal y (t) ′ of the low-pass filter 2 does not reach the mass M unless a considerable time elapses from the load timing t0 of the article. The measurement time is limited by the time constant of the LPF 2, and it is not possible to cope with faster weighing.

  In addition to using a digital filter as a filter, a transfer function that represents a transient operation of the sensor is obtained, and a filter coefficient corresponding to a function that is the inverse of the transfer function is set in the digital filter, so that the vibration component is reduced. A method of suppressing has also been proposed (Patent Document 1).

JP-A-7-134057

  However, because actual sensors have nonlinear elements due to their structure, it is difficult to define a transfer function that accurately represents transient behavior, and errors due to nonlinear elements that cannot be defined by that function occur. . In addition, many sample values are required to suppress the vibration component, and the speed is not sufficient.

  An object of the present invention is to provide a signal processing method and a signal processing apparatus that can solve this problem and can predict a convergence value of a sensor output with high accuracy and high speed.

In order to achieve the above object, the signal processing method of the present invention comprises:
Converting the output signal of the sensor into a digital original signal sequence by oversampling (S1);
Detecting a signal sequence of AC components included in the original signal sequence (S2);
Matching the phase and amplitude of the original signal sequence and the extracted AC component signal sequence (S3);
Performing addition or subtraction of both signal sequences combined with the phase and amplitude to detect a signal sequence of a DC component of the original signal sequence (S4);
Performing a thinning process on the signal sequence of the DC component (S5);
Performing a high-frequency cutoff process on the thinned signal train (S6),
The signal sequence that has been subjected to the high-frequency cutoff processing is used as an expected convergence value of the output signal of the sensor.

The signal processing device according to claim 2 of the present invention is
A / D conversion means (21) for oversampling the output signal of the sensor (1) and converting it to a digital original signal sequence;
AC component detecting means (23) for detecting an AC component of the original signal sequence;
Correction means (24, 25) for matching the phase and amplitude of the signal sequence of the AC component output from the AC component detection means and the original signal sequence;
An arithmetic means (26) for performing addition or subtraction between the signal sequence of the AC component corrected by the correcting means and the original signal sequence, and detecting the signal sequence of the DC component of the original signal sequence;
Thinning means (30) for performing thinning processing on the signal sequence of the DC component output from the computing means;
A low-pass filter (31) that performs low-order high-frequency cutoff processing on the signal sequence of the DC component subjected to the thinning-out processing and outputs the processing result as a predicted convergence value of the output signal of the sensor.

  As described above, in the present invention, the AC component is detected from the original signal sequence obtained by oversampling the output signal of the sensor, and the addition and subtraction processing is performed by combining the phase and amplitude of the AC component and the original signal sequence. A DC component included in the original signal sequence is detected, and further, a thinning process and a high-frequency cutoff process are performed on the DC component to obtain a predicted convergence value of the output signal of the sensor.

  For this reason, it is possible to detect AC components with high accuracy from the original signal sequence with few errors obtained by oversampling, and to detect DC components with high accuracy by adding and subtracting the phase and amplitude according to the original signal sequence. can do. Then, the error fluctuation of the DC component can be removed in a short time by the low-order high-frequency cut-off process by the thinning process for the series component, and the convergence value of the sensor output signal can be obtained with high accuracy and high speed. Can be expected.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
First, the signal processing method of the present invention will be described based on the flowchart of FIG.

  As shown in FIG. 1, in the signal processing method of the present invention, an output signal y (t) from a sensor that receives a physical quantity M such as mass or pressure is converted into a digital original signal sequence y (k) by oversampling. The signal is converted (S1), and the AC component signal sequence Δy (k) included in the original signal sequence y (k) is detected (S2). This AC detection processing can be realized by difference calculation processing or discrete Hilbert transform.

  Next, the phases and amplitudes of the original signal sequence y (k) and the AC component signal sequence Δy (k) are matched (S3), and both signal sequences are added or subtracted to obtain the original signal sequence y (k ) Of the DC component signal sequence z (k) is detected (S4).

  Then, a thinning-out process (S5) and a high-frequency cutoff process (S6) are performed on the signal sequence z (k) of the direct current component to obtain a predicted convergence value of the sensor output signal y (t).

  As described above, it is possible to detect the AC component Δy (k) with high accuracy from the original signal sequence y (k) with a small error obtained by oversampling, and to convert the phase and amplitude into the original signal sequence y (k). By adding and subtracting together, the direct current component can be detected with high accuracy. Further, the error included in the direct current component is subjected to a thinning process on the signal sequence z (k) for the serial component, and the high-order cutoff processing of the lower order can be performed by the thinning effect. The fluctuation can be removed in a short time, and the convergence value of the sensor output signal y (t) can be predicted with high accuracy and at high speed.

  FIG. 2 shows a configuration of a signal processing device 20 according to an embodiment to which the signal processing method is applied.

  This signal processing device 20 receives an output signal y (t) of the sensor 1 in a transient state when receiving a physical quantity M (for example, weight), and predicts and detects its convergence value. A / D The converter 21 samples the analog signal y (t) output from the sensor 1 at a predetermined sampling cycle, converts the analog signal y (t) into a digital original signal sequence y (k), and outputs the digital signal to the vibration component removal unit 22.

  The sampling frequency of the A / D converter 21 is set sufficiently high (for example, several hundred Hz to several tens of kHz or more) with respect to the upper limit (for example, several tens of Hz) of the expected frequency band of the input signal y (t). . In general, the Nyquist frequency (twice the upper limit frequency of the signal) is sufficient for sampling the analog signal. However, in the signal processing device 20, as described above, the sampling frequency is more than twice the upper limit frequency of the analog signal. The quantization error is reduced by an oversampling method that performs sampling at a remarkably high frequency.

  The original signal sequence y (k) output from the A / D converter 21 is input to an AC component detecting unit 23 of the signal processing unit 22 and a phase correcting unit 24 described later.

  The AC component detecting means 23 is for detecting an AC component (transient vibration component) included in the original signal sequence y (k), and is obtained by differential calculation (differential calculation in a continuous signal space) or discrete Hilbert transform described later. Although it can be realized, here, a case of difference calculation will be described.

  In the case of the difference calculation, it can be realized by sequentially performing the next primary difference calculation for the input signal sequence y (k) and sequentially obtaining the difference signal sequence Δy (k).

Δy (1) = y (1) −y (0)
Δy (2) = y (2) −y (1)
Δy (3) = y (3) −y (2)
......
Δy (k) = y (k) −y (k−1)

  As described above, since the A / D converter 21 is an oversampling method and the sampling cycle is sufficiently short with respect to the change of the input signal y (t), the difference signal sequence Δy (k) is the input signal y (t ) Is almost exactly corresponding to the differentiated signal. The input signal y (t) is represented by the sum of a DC component corresponding to the physical quantity M and a sinusoidal and damped vibration component (AC component). The difference signal sequence Δy (k) is substantially equal to a 90-degree phase shift of the AC component, and therefore has a characteristic that does not include a DC component corresponding to the physical quantity M.

This difference signal sequence Δy (k) is input to the phase correction means 24.
The phase correction unit 24 performs a phase shift process on at least one signal sequence so that the phase of the difference signal sequence Δy (k) and the phase of the vibration component of the original signal sequence y (k) are in phase or in phase. .

  For example, with respect to the vibration component of the original signal sequence y (k) indicated by the solid line in FIG. 3A, the phase of the difference signal sequence Δy (k) obtained by the difference calculation process of the AC component detecting means 23 is as shown in FIG. As shown in (b) of FIG. 3, if it is assumed that there is no processing delay of the AC component detecting means 23, the difference signal sequence Δy (k) indicated by the dotted line in (b) of FIG. By delaying by 90 degrees as in the signal sequence Δy ′ (k), the phase of the vibration component of the original signal sequence y (k) can be made in phase.

  Conversely, by delaying the original signal sequence y (k) by 90 degrees as in the original signal sequence y ′ (k) indicated by the dotted line in FIG. 3A, the phases of both signal sequences are shifted by 180 degrees. The phase can be reversed.

  The amplitude correction means 25 constitutes the correction means of this embodiment together with the phase correction means 24, and the phase correction is performed by multiplying the phase-corrected difference signal sequence Δy ′ (k) by, for example, a predetermined correction coefficient. The amplitude of the vibration component of the original signal sequence y (k) ′ is matched.

  In this way, if the phase of the difference signal sequence and the vibration component of the original signal sequence are in phase or opposite phase and the amplitude is matched, the AC component can be removed (cancelled) by subtraction or addition of both signal sequences. Minute detection is possible.

  However, since there is actually a delay for processing of the AC detection means 23 and a delay for amplitude correction by at least one clock, the phase correction means 24 performs the phase shift process in anticipation of these processing delay times. The amplitude-corrected difference signal sequence Δy ″ (k) and the phase-corrected original signal sequence y (k) ′ are input to the computing means 26 in the same phase or in reverse phase.

  The phase shift amount for the phase correction and the coefficient for the amplitude correction are set by obtaining an optimum value for the sensor 1 in advance.

  The computing means 26 performs subtraction (in the case of in-phase) or addition (reverse phase) of the differential signal sequence Δy ″ (k) subjected to phase correction and amplitude correction as described above and the original signal sequence y ′ (k) subjected to phase correction. In this case, the vibration component is removed from the original signal sequence y ′ (k) and a signal sequence z (k) for direct current is output.

  The signal sequence z (k) output from the computing means 26 corresponds to the magnitude of the physical quantity M received by the sensor 1 and its change.

  However, the signal sequence z (k) for direct current obtained by the addition / subtraction process includes an error due to the nonlinear element of the sensor 1, and the required accuracy is, for example, about 1/100 (1 percent). If so, the prediction based on the signal sequence z (k) is often sufficient, but it is not sufficient when accuracy of 1/1000 or less (0.1 percent or less) is required.

  This error has the same frequency component as the frequency of the main mode of the non-linear vibration, and the suppression can be realized by using an FIR type low-pass filter, but as described above, the input signal y (t) On the other hand, in order to obtain a DC component (including a very low frequency band) with high accuracy from a signal sequence obtained by the oversampling method, a large filter of a very high order (for example, 500th order) is required. In this case, a large number of multiplication operations of the filter coefficients corresponding to the orders are required, and more time is required to obtain a result with less error due to processing delay.

  In order to prevent this, the signal processing device 20 performs a thinning (decimation) process and a low-order high-frequency cutoff process on the DC signal sequence z (k) obtained by the computing means 26 to reduce the error fluctuation. The predicted convergence value M ′ is obtained at high speed and with high accuracy.

  That is, a signal sequence z (k) for direct current output from the calculation means 26 is input to the thinning means 30 to perform a thinning process of 1 / N (N is, for example, 10), and a signal string obtained by the thinning process. z (m · N) (m is a positive integer) is input to a low-order (for example, 30th order) FIR type low-pass filter 31 to suppress an error variation.

  By this thinning-out processing and low-order high-frequency cutoff processing, it is possible to obtain a predicted convergence value that is much faster and more accurate with a simple configuration than the method of suppressing error fluctuations using only a high-order filter.

  FIG. 4 shows a simulation result at the sampling frequency of about 10 kHz of the signal processing apparatus 20 described above.

  FIG. 4A shows an original signal sequence y (k) obtained by oversampling the output signal y (t) when a load is applied to the sensor 1 for a fixed time. Fluctuates following changes in load with vibration. As described above, since the sampling period of the A / D converter 21 is sufficiently short, the original signal sequence y (k) follows the sensor output signal y (t) with a small error (quantization error). Yes.

  FIG. 4B shows the result of high-frequency cutoff processing using a high-order digital filter for the original signal sequence y (k) in FIG. 4A (result of the conventional method). Yes.

  FIG. 4C shows the difference obtained by performing the phase correction (in-phase) and the amplitude correction on the difference signal sequence Δy (k) obtained by the difference calculation process of the original signal sequence y (k). 4 represents a signal sequence Δy ″ (k), and a signal sequence z (k) for direct current obtained by subtracting the difference signal sequence from the original signal sequence y (k) in FIG. (D).

  This signal sequence z (k) shows a much faster follow-up performance with respect to a change in load than the result of the conventional method shown in FIG. 4B. include.

  FIG. 5A is an enlarged view of this error variation, and is a size that cannot be ignored when accuracy of 1/1000 or less is required as described above.

  However, by performing the above-described thinning-out processing and low-order high-frequency cutoff processing, the fluctuation due to this error is suppressed to an extent that can be almost ignored in the enlarged views of FIGS. 4D and 5B. It can be seen that the number of computations for the above-described operation is small and the delay time is relatively short, and the high-speed performance with respect to the result of the conventional method of FIG. 4B is not lost.

  From the above simulation results, it can be seen that the signal processing apparatus 20 according to the embodiment can predict the convergence value of the output signal of the sensor 1 at a much higher speed and higher accuracy than the conventional one.

  As described above, the case where the AC component detecting unit 23 detects the AC component of the original signal sequence y (k) by the difference calculation (differential calculation in the continuous signal space) has been described. As described above, the discrete Hilbert transform is used. It is also possible to detect the AC component. The discrete Hilbert transform has a function of shifting the AC component of the input signal sequence by 90 degrees and outputting it, and can detect the AC component of the original signal sequence y (k) as in the difference calculation. It is.

  In this case, as shown in FIG. 6, if the AC component detecting means 23 is configured by cascading two discrete Hilbert transformers 41 and 42, 180 degrees with respect to the vibration component of the original signal string y (k). A signal sequence with a delayed phase can be obtained, and the phase correction means 24 can be simplified. The latter-stage Hilbert transformer 42 can also be interpreted as a part of the phase correction means 24.

  FIG. 7 shows a simulation result when the two Hilbert transformers 41 and 42 are used as described above.

  7A shows the original signal sequence y (k) obtained by oversampling the output signal y (k) of the sensor 1, and FIG. 7B shows the result of the conventional method. FIG. 7 shows an AC signal sequence Δy ″ (k) obtained by performing two-stage Hilbel transform processing on the original signal sequence y (k), phase correction (correction regarding processing delay), and amplitude correction. (C).

  A signal sequence z (k) for direct current obtained by adding the signal sequence Δy ″ (k) to the original signal sequence y (k) is as shown in FIG.

  The signal sequence z (k) for the direct current also shows extremely high tracking performance with respect to the load as compared with the result of the conventional method, but also includes error fluctuations that cannot be ignored as described above.

  This error fluctuation is suppressed to a negligible level as shown in FIG. 7D by performing the thinning process by the thinning means 30 and the low-order high-frequency cutoff process by the low-pass filter 31 as described above. It can be seen that high speed is not lost.

  In the above description, the amplitude correction is performed after the phase correction is performed on the AC signal sequence. However, the phase correction may be performed after the amplitude correction is performed first.

  Moreover, you may employ | adopt the polyphase structure which makes a calculation slow down as a low-pass filter used for a high region cutoff process.

  In the above description, the signal processing for the output signal of the sensor for weight measurement has been described. However, the present invention can be similarly applied to the output signal of another sensor that shows a transient response to a physical quantity load. .

The flowchart which shows the procedure of the signal processing method of this invention The figure which shows the structure of embodiment of this invention Waveform diagram for explaining the processing of the main part of the embodiment Waveform diagram showing simulation results of an embodiment using differential processing Waveform diagram enlarging a part of FIG. The figure which shows the example which comprised AC component detection means with the Hilbert converter Waveform diagram showing simulation results of an embodiment using Hilbert transform processing Waveform diagram of sensor output signal Configuration diagram of conventional equipment

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Sensor, 20 ... Signal processing apparatus, 21 ... A / D converter, 23 ... AC component detection means, 24 ... Phase correction means, 25 ... Amplitude correction means, 26 ... Calculation means, 30 .... thinning means, 31 ... low-pass filter, 41, 42 ... Hilbert transformer

Claims (2)

Converting the output signal of the sensor into a digital original signal sequence by oversampling (S1);
Detecting a signal sequence of AC components included in the original signal sequence (S2);
Matching the phase and amplitude of the original signal sequence and the extracted AC component signal sequence (S3);
Performing addition or subtraction of both signal sequences combined with the phase and amplitude to detect a signal sequence of a DC component of the original signal sequence (S4);
Performing a thinning process on the signal sequence of the DC component (S5);
Performing a high-frequency cutoff process on the thinned signal train (S6),
A signal processing method in which the signal sequence subjected to the high-frequency cutoff processing is used as an expected convergence value of the output signal of the sensor. A / D conversion means (21) for oversampling the output signal of the sensor (1) and converting it to a digital original signal sequence;
AC component detecting means (23) for detecting an AC component of the original signal sequence;
Correction means (24, 25) for matching the phase and amplitude of the signal sequence of the AC component output from the AC component detection means and the original signal sequence;
An arithmetic means (26) for performing addition or subtraction between the signal sequence of the AC component corrected by the correction means and the original signal sequence, and detecting the signal sequence of the DC component of the original signal sequence;
Thinning means (30) for performing thinning processing on the signal sequence of the DC component output from the computing means;
A signal processing apparatus comprising: a low-pass filter (31) that performs high-frequency cutoff processing on the signal sequence of the DC component that has been subjected to the thinning-out processing and outputs the processing result as a predicted convergence value of the output signal of the sensor.

Images