JP6455852B2 - Signal analysis system, method and program - Google Patents

Description

  The present invention relates to a signal analysis system, and more particularly to a system for analyzing fluctuations in a periodic signal.

  Conventionally, a method for diagnosing a damaged state of a rolling bearing using an acoustic emission method (Acoustic Emission, AE) is known. For example, Japanese Patent Application Laid-Open No. 8-159151 (Patent Document 1) discloses a bearing by measuring ultrasonic waves generated from a bearing with a microphone and comparing the component intensity of the target frequency of the acoustic signal with a master curve prepared in advance. Disclosed is a bearing diagnosis method for predicting the damage state and estimating the remaining life.

JP-A-8-159151

  The present invention has been made in view of the above-described prior art, and an object of the present invention is to provide a novel signal analysis system that can detect a sign of destruction or collapse of an object.

  In the process of studying a new signal analysis system that can detect the sign of destruction and collapse of an object, the present inventor, when a vibrating object is in the process of breaking or collapsing, is caused by strain generated inside the object. Then, it was discovered that fluctuation occurred in the peak interval of the vibration waveform emitted from the object. Based on this discovery, the present inventor can grasp the signs of destruction and collapse of objects from temporal fluctuations of the scaling index by creating a mechanism that continuously analyzes fluctuations of vibration waveforms emitted from the objects using DFA (Detrended Fluctuation Analysis). I got the idea that I could do it. And as a result of earnest examination about the realization of this idea, the following configuration has been conceived and the present invention has been achieved.

  That is, according to the present invention, a peak time acquisition unit that acquires a peak time of an analysis target signal, a DFA target data generation unit that generates DFA target data based on the peak time acquired by the peak time acquisition unit, A DFA execution unit that executes a DFA based on the DFA target data to acquire a scaling index; and a scaling index integration unit that integrates the acquired scaling index. The DFA target data generation unit includes the peak time The peak times acquired by the acquisition unit are loaded in order of time series (A + 1), A peak intervals are calculated based on the loaded (A + 1) peak times, and the A peak intervals are The process of generating pseudo time-series data including M elements as the DFA target data is repeated based on the time-series data (A and M are both natural numbers, M is greater than A. The same applies hereinafter), and the DFA execution unit loads the DFA target data generated by the DFA target data generation unit in the order of generation and loads them. The process of acquiring the scaling index based on the DFA target data is repeatedly executed, and the scaling index integration unit loads the scaling indexes acquired by the DFA execution unit in the acquisition order, and the loaded scaling index is a predetermined value. When the scaling index is not below the threshold value, the scaling index is added to the latest integrated value, and when the loaded scaling index is below the predetermined threshold value, the process of resetting the latest integrated value is repeatedly executed. A signal analysis system is provided.

  As described above, according to the present invention, a novel signal analysis system capable of capturing a sign of destruction or collapse of an object is provided.

1 is a schematic configuration diagram of a signal analysis system of an embodiment. The signal analysis system function block diagram of this embodiment. The figure which shows the waveform of the analysis object signal in a 1st measuring method. The figure which shows the waveform of the analysis object signal in a 2nd measuring method. The figure which shows the waveform of the analysis object signal in a 3rd measuring method. The flowchart showing the peak time acquisition process of this embodiment. The conceptual diagram for demonstrating the peak time acquisition process of this embodiment. The flowchart showing the DFA object data generation process of this embodiment. The conceptual diagram for demonstrating the DFA object data generation process of this embodiment. The conceptual diagram for demonstrating the DFA object data generation process of this embodiment. The conceptual diagram for demonstrating the DFA object data generation process of this embodiment. The flowchart showing DFA of this embodiment. The flowchart showing DFA of this embodiment. The figure which shows the box size data of this embodiment. The conceptual diagram for demonstrating DFA of embodiment. The flowchart showing the integration process of the scaling index of this embodiment. The figure which shows the time series graph of the integrated value of the scaling index | exponent of this embodiment.

  Hereinafter, the present invention will be described with reference to embodiments shown in the drawings, but the present invention is not limited to the embodiments shown in the drawings. In the drawings referred to below, the same reference numerals are used for common elements, and the description thereof is omitted as appropriate.

  FIG. 1 shows a schematic configuration of a signal analysis system 100 according to an embodiment of the present invention. As shown in FIG. 1, the signal analysis system 100 of this embodiment includes a vibration sensor 10, a vibrator 12, and a computer 20. The vibration sensor 10 is a vibration detection unit that converts vibration generated by an object to be measured into a voltage signal. The vibration sensor 10 may be a contact type using a piezo element or the like, or may be a non-contact type using a laser or the like. The vibrator 12 is a means for inputting periodic vibration to an object to be measured. The vibrator 12 can be configured by using an appropriate vibrator according to the scale of the measurement target, and may be either a contact type or a non-contact type.

  On the other hand, the computer 20 is an information processing device for analyzing fluctuations in the sensor signal input from the vibration sensor 10, and preferably outputs an excitation signal (sine wave signal) to the vibrator 12. In addition, the configuration is provided. In FIG. 1, a notebook PC is exemplarily shown as the computer 20, but the computer 20 may be a tablet PC or a desktop PC, and is used to catch a sign of destruction or collapse of a stationary object. It may be a workstation or server specialized in the above, or an embedded computer mounted on a terminal device specialized for the application. Further, transmission / reception of signals between the computer 20 and the vibration sensor 10 and the vibrator 12 may be performed by either wired communication or wireless communication.

  The outline configuration of the signal analysis system 100 of the present embodiment has been described above. Next, the functional configuration of the computer 20 will be described based on the functional block diagram shown in FIG.

  As shown in FIG. 2, the computer 20 includes a peak time acquisition unit 21, a measurement mode setting unit 22, a DFA target data generation unit 23, a storage unit 24, a DFA execution unit 25, and a scaling index integration unit 26. And an output information generation unit 27. Each functional unit described above is realized by installing a dedicated program in the computer 20.

  The peak time acquisition unit 21 is a functional unit for acquiring the peak time of the input periodic signal, and the measurement mode setting unit 22 allows the user to select a measurement mode (described later) that the signal analysis system 100 has. This is a functional unit for presenting and setting various parameters according to the measurement mode selected by the user.

  The DFA target data generation unit 23 is a functional unit for generating DFA (Detrended Fluctuation Analysis) target data based on the peak time acquired by the peak time acquisition unit 21, and the DFA execution unit 25 generates the DFA target data. This is a functional unit for executing DFA based on the target data generated by the unit 23.

  The scaling index integration unit 26 is a functional unit for integrating the scaling index acquired by DFA, and the output information generation unit 27 is a functional unit for generating output information based on the integrated value of the scaling index.

  The computer 20 further includes a sine wave oscillation circuit 34 and a signal superimposing circuit 36. The sine wave oscillating circuit 34 is a circuit for generating and oscillating a sine wave signal having a specified frequency, and the signal superimposing circuit 36 is a circuit for superimposing and outputting two input signals.

  Here, a measurement method using the signal analysis system 100 will be described with reference to FIG. In the present embodiment, one of the following three types of measurement methods can be applied depending on the measurement target.

(First measurement method)
When measuring a moving object that periodically vibrates in the vicinity of a predetermined frequency when normal, the first measurement method can be applied. In the first measurement method, the vibration sensor 10 detects periodic vibrations that accompany the movement of the moving object, and the computer 20 analyzes the sensor signal transmitted from the vibration sensor 10. Examples of the moving object to which the first measurement method is applied include a rotary drive device such as a motor and an internal combustion engine, and a rotary body such as a rolling bearing connected to the rotary drive device and rotating.

  In the first measurement method, the computer 20 turns on the switch SW shown in FIG. 1 so that the sensor signal of the vibration sensor 10 is directly input to the peak time acquisition unit 21.

  FIG. 3 illustrates the vibration waveform of the sensor signal input from the vibration sensor 10 that detects the periodic vibration of the motor to the peak time acquisition unit 21. In this case, as shown in FIGS. 3A and 3B, the vibration waveform input to the peak time acquisition unit 21 sequentially changes according to the change in the state of the motor.

(Second measurement method)
When measuring a stationary object that does not vibrate itself but is capable of inputting vibration from the outside, the second measurement method can be applied. In the second measurement method, the periodic vibration of the excited stationary object is detected by the vibration sensor 10 in a state where the periodic vibration is input from the vibrator 12 to the stationary object, and the sensor signal transmitted from the vibration sensor 10 is detected. Is analyzed by the computer 20. Examples of stationary objects to which the second measurement method is applied include skeletons (building columns, beams, walls, etc.).

  In the second measurement method, the computer 20 controls the vibration signal generated by the vibration signal generation unit 32 to be transmitted to the vibration exciter 12, and the sensor signal of the vibration sensor 10 is the peak time acquisition unit 21. The switch SW shown in FIG. 1 is turned on.

  FIG. 4 illustrates a vibration waveform of a sensor signal input from the vibration sensor 10 to the peak time acquisition unit 21 when periodic vibration is input to the aluminum steel material. In this case, as shown in FIGS. 4A and 4B, the vibration waveform input to the peak time acquisition unit 21 sequentially changes according to the state change of the aluminum steel material.

(Third measurement method)
The third measurement method can be applied when measuring a stationary object that generates weak vibration and cannot receive vibration from the outside. In the third measurement method, a weak vibration of a stationary object is detected by the vibration sensor 10, and a superimposed signal obtained by superimposing a sensor signal transmitted from the vibration sensor 10 and a predetermined sine wave signal is analyzed by the computer 20. As the stationary object to which the third measurement method is applied, the ground can be exemplified.

  In the third method, the computer 20 performs control so that a sine wave signal having a predetermined frequency generated by the sine wave oscillation circuit 34 is input to the signal superimposing circuit 36, and the sensor signal of the vibration sensor 10 is the signal superimposing circuit. As shown in FIG. 36, the switch SW shown in FIG. As a result, the sine wave signal input from the sine wave oscillation circuit 34 and the sensor signal input from the vibration sensor 10 are superimposed in the signal superimposing circuit 36, and the superimposed signal is input to the peak time acquisition unit 21. .

  FIG. 5A illustrates a vibration waveform of a sensor signal input to the signal superimposing circuit 36 from a seismometer (corresponding to the vibration sensor 10) installed on the ground, and FIG. 5B illustrates a sine wave oscillation circuit. The sine wave signal input into the signal superimposition circuit 36 from 34 is illustrated. In this case, in the signal superimposing circuit 36, the two signals shown in FIGS. 5A and 5B are superposed, and as a result, the superposed signal shown in FIG.

  As described above, in the present embodiment, either the sensor signal of the vibration sensor 10 or the superimposed signal obtained by superimposing the sensor signal of the vibration sensor 10 and the sine wave signal is input to the peak time acquisition unit 21 according to the measurement method to be applied. Will be. The signals input to the peak time acquisition unit 21 are all periodic signals, and the computer 20 analyzes fluctuations of the periodic signals. In the following, the sensor signal and the superimposed signal input to the peak time acquisition unit 21 are collectively referred to as “analysis target signal”.

  The functional configuration of the computer 20 has been outlined above. Next, specific contents of the processing executed by each functional unit described above will be described. In the following description, FIG. 2 will be referred to as appropriate.

  First, the processing executed by the peak time acquisition unit 21 will be described based on the flowchart shown in FIG. 6 and the schematic waveform diagram of the analysis target signal shown in FIG.

In step 101, voltage monitoring of the input analysis target signal is started. The voltage monitoring is continued until the signal voltage of the analysis target signal exceeds a predetermined threshold voltage _VTh (S101, No), and when the signal voltage of the analysis target signal exceeds the threshold voltage _VTh (S101, Yes). The processing proceeds to step 102.

In step 102, the voltage value of the signal to be analyzed is sampled at a predetermined sampling rate over the detection period T1, and sampling is stopped when the detection period T1 has passed (S103). In the subsequent step 104, the time at which the maximum value V _max among the sampled voltage values is sampled is acquired as the peak time t, and the time series data {t _i } at the peak time t is stored in the storage unit 24.

  Here, in order to analyze the fluctuation of the analysis target signal with high accuracy, it is important to accurately detect the peak of the analysis target signal. Therefore, it is necessary to set the sampling rate in step 102 to an appropriate value according to the time scale of fluctuation occurring in the analysis target signal. When the peak time t of the analysis target signal fluctuates in units of β milliseconds, the sampling rate is In general, it is preferably set to (1 / β) kHz or more. For example, when a precise artificial rotating body such as a rolling bearing is to be measured, it is preferable to set the sampling rate to 200 kHz or higher, and a complex formed by nonlinearly coupling a plurality of parts including the rotating body When the measurement target is (for example, an automobile), the sampling rate is preferably set to 20 kHz or higher. Further, when the biological heart rate signal is to be measured, the sampling rate is preferably set to 1 kHz or higher. .

  In the subsequent step 105, it is determined whether or not the standby period T2 has elapsed after the sampling is stopped. Thereafter, this determination process is looped until the waiting period T2 elapses (S105, No).

  When the standby period T2 elapses (S105, Yes), the process returns to step 101 again, and the voltage monitoring of the analysis target signal is resumed. Thereafter, the above-described series of processing (S101 to S105) is repeatedly executed.

  As described above, in the present embodiment, by interrupting the voltage monitoring of the analysis target signal in the standby period T2 that is continuous to the detection period T1, unexpected noise that occurs in the standby period T2 is counted as a peak. To prevent. In the present embodiment, as the waiting period T2, a time length corresponding to the difference between the time length of approximately half the peak interval of the analysis target signal obtained from the normal measurement target and the time length of the detection period T1 is set. It is preferable.

  The processing executed by the peak time acquisition unit 21 has been described above. Subsequently, the processing executed by the DFA target data generation unit 23 will be described based on the flowchart shown in FIG.

In step 201, (A + 1) peak times t are loaded in time series order from the time series data {t _i } stored in the storage unit 24 by the peak time acquisition unit 21. Here, “A” is a parameter that determines the number of elements of time-series data {p _i } described later. In the present embodiment, the value of “A” is preferably set to a natural number in the range of 200 to 300, and is preferably set to “250”.

In the subsequent step 202, A peak intervals x are calculated from (A + 1) peak times t loaded from the storage unit 24, and time series data {p _i } having the calculated A peak intervals x as elements is obtained. Generate. Specifically, (A + 1) peak times t loaded from the storage unit 24 are arranged in time series, and the time difference between the (i + 1) th time and the (i) th time is calculated as the peak interval x. As a result, {p _i } is generated as time-series data including A peak intervals x.

In subsequent step 203, the time-series data { _pi } generated in step 202 is divided into B segments. At this time, a serial number is assigned to each segment. Here, “B” is a parameter that determines the number of divisions of the time-series data {p _i }. In the present embodiment, the value of “B” is preferably set to a natural number in the range of 4 to 6, and is preferably set to “5”. For example, when A = 250 and B = 5 are set, as shown in FIG. 9A, the time-series data { _pi } is divided into five types of segments each consisting of 50 elements. Are numbered 1-5.

In the following step 204, one segment is selected at random from the B types of segments, and the selected segment is combined behind the time-series data { _pi }. For example, in the case of the time series data {x _i } shown in FIG. 9A, a random number generator (for example, Mersenne Twister pseudo random number generator) outputs random numbers 1 to 5 according to the numbers. The selected segment is selected, and the selected segment is coupled to the tail of the time series data {p _i }. Hereinafter, the pseudo time series data extended by the combination of segments is referred to as time series data {x _i }.

In the following step 205, it is determined whether or not the number of elements of the time series data {x _i } has reached the target number M. As a result, when the number of elements of the time series data {x _i } has not reached the target number M (No at Step 205), the process returns to Step 204, and the randomly selected segment is stored in the time series data {x _i }. Join at the end. Thereafter, until the number of elements of {x _i } reaches the target number M, as shown in FIG. 9B, the segment combination is repeated (step 205, No → step 204), and the number of elements of {x _i } When the target number M is reached (step 205, Yes), the process proceeds to step 206. Here, “M” is a parameter that determines the number of elements of continuous data to be subjected to DFA. In the present embodiment, the value of “M” is preferably set to a natural number in the range of 25000 to 35000, and is preferably set to “30000”.

In step 206, when after calculating the average value _{x ave} of the M peak interval x constituting the series data _{{x i},} subtracting the mean value _{x ave} from the elements constituting the series data _{{x i}} When Thus, time-series data {(x _i -x _ave ) _i } is generated. FIGS. 10A and 10B exemplarily show a graph of time series data {x _i } and a graph of time series data {(x _i −x _ave ) _i }, respectively.

In the subsequent step 207, the time series data {(x _i -x _ave ) _i } is integrated to generate time series data {y _i }. The following formula (1) shows a calculation formula of the time series data {y _i }.

Specifically, the time series data {y _i } is generated by adding the elements of the time series data {(x _i -x _ave ) _i } in time series order. FIG. 10C exemplarily shows a graph of the time series data {y _i } generated in step 207.

In the subsequent step 208, the pseudo time-series data {y _i } generated by the above-described series of procedures is numbered as DFA target data in the order of generation and stored in the storage unit 24. Thereafter, the process returns to step 201 again, and thereafter, steps 201 to 208 are repeatedly executed. At this time, in step 201 after the second round, (A + 1) at (A + 1) peak times t loaded in the immediately preceding step 201 from the time-series data {t _i } stored in the storage unit 24. It is preferable to load (A + 1) elements so as to overlap the peak time t.

  In step 201 after the second round, (A + 1) elements may be loaded in such a manner that two or more elements in (A + 1) elements loaded in the immediately preceding step 201 overlap. . For example, as shown in FIG. 11, when A = 250 is set, 251 elements (first data) are loaded in the first step 201, and then the second half of the first data in step 201 of the second round. 251 elements (second data) obtained by adding 125 unloaded elements to 126 elements of the second data are loaded, and in step 201 of the third round, the latter 126 elements of the second data are not loaded. The procedure of loading 251 elements (third data) obtained by adding the 125 elements may be repeated.

  The process executed by the DFA target data generation unit 23 has been described above. Next, the process executed by the DFA execution unit 25 will be described based on the flowcharts shown in FIGS. 12 and 13.

First, at the start of processing, a counter [c] indicating the generation order of the time series data {y _i } is set to an initial value “1” (step 301).

In subsequent step 302, the c-th (first at the start) time-series data {y _i } is loaded from the storage unit 24.

  In the subsequent step 303, the box size data in the box size range set from the box size data stored in the storage unit 24 is loaded. Here, the box size data means a set of a plurality of box sizes (integers) used in the DFA, the box size means the number of data elements, and the box size range is used in the DFA. It means the range that box size (integer) takes.

  FIG. 14 exemplarily shows box size data stored in the storage unit 24. The box size data shown in FIG. 14 has 136 integers in the range of 10 to 1000 as the box size, and the 136 integers are incremented by 1 between 10 and 100, During 500, it is incremented by 10 and between 500 and 1000 is incremented by 100.

  In step 303, an integer group in the set box size range is loaded from the integer groups shown in FIG. Here, as the box size range, an appropriate range according to the measurement target is set in advance. For example, when a precise artificial rotating body such as a rolling bearing is used as a measurement object, it is preferable to set the box size range to a range centering on 10 to 30. If the frequency of the analysis target signal related to the rotating body is 100 Hz, the box size range can be grasped by replacing it with a time length range centering on 0.1 second to 0.3 second. Moreover, when measuring the composite_body | complex (for example, motor vehicle) which a several component containing a rotary body couple | bonds nonlinearly, it is preferable to set a box size range to the range centering on 30-60. If the frequency of the signal to be analyzed related to the complex is 100 Hz, the box size range can be grasped by replacing it with a time length range centering on 0.3 seconds to 0.6 seconds. Furthermore, when a periodic biological signal typified by a heartbeat signal is to be measured, it is preferable to set the box size range to a range centering on 30 to 270. If the frequency of the biological signal is 1 Hz, the box size range can be grasped by replacing it with a time length range centering on 30 seconds to 270 seconds.

  In the subsequent step 304, the first box size [N] (for example, [10]) is set from the loaded box size data.

In the following step 305, the time series data {y _i } loaded in step 302 is divided by the box size [N] set at that time. For example, when the box size set at that time is [10], the time-series data {y _i } is divided into small sections (hereinafter referred to as boxes) including 10 elements. In this case, since the time series data {y _i } is composed of M elements, in step 305, the time series data {y _i } is divided into M / N boxes. FIG. 15A shows time-series data {y _i } divided by the box size [10]. In this case, each box (BOX (1), BOX (2), BOX (3)...) Includes 10 elements.

In the subsequent step 306, for each of the divided boxes (BOX (1), BOX (2), BOX (3)...), An approximate curve is fitted to N data existing in the box, The value on the approximate curve is determined as the local trend y _v for each box. Here, fitting of the approximate curve can be performed by a least square method using a linear function to a quartic function. The approximate curve here is a concept including a straight line. FIG. 15B shows approximate curves y _v (y _v (1), y _v (2), y _v (3) for each box (BOX (1), BOX (2), BOX (3)...). ) ...) is shown in a fitted state. In FIG. 15B, the approximate curve is shown as a linear function for convenience of explanation.

In the subsequent step 308, for each of the boxes (BOX (1), BOX (2), BOX (3)...), The local trend y _v determined for the box is subtracted from each element constituting the box to obtain a time series. Data {z _i } is generated. The following equation (2) shows an equation of time series data {z _i }, and FIG. 15C shows a graph of the time series data {z _i } generated in step 306.

In the subsequent step 308, each of the boxes (BOX (1), BOX (2), BOX (3)...) Of the time series data {z _i } is shown by enclosing the box with the first element (dotted circle). ) And the value of the last element (indicated by a dotted square).

Note that the following method may be employed in place of the processing in steps 307 and 308 described above to reduce the amount of calculation. That is, each box (BOX (1), BOX ( 2), BOX (3) ...) by subtracting the corresponding local trend _{y v} from corresponding local trend _{y v} values are subtracted and trailing elements from the first element in the This is a method for obtaining a difference between values.

  In the subsequent step 309, the root mean square [S] of the difference (difference between the beginning and the end) obtained for all boxes is calculated.

  In the subsequent step 310, a set of numerical values (N, S) consisting of the root mean square [S] calculated in step 309 and the box size [N] set at that time is recorded.

  In the subsequent step 311, it is determined whether or not a set (N, S) has been recorded for all box sizes [N] included in the box size data loaded in step 303. As a result, when the recording of the group (N, S) is not completed for all box sizes [N] (No at Step 311), the process proceeds to Step 312.

  In step 312, the next box size [N] (for example, [11]) is newly set from the values included in the box size data loaded in step 303. Thereafter, the process returns to step 305 again, and thereafter, until it is determined in step 311 that the set (N, S) has been recorded for all box sizes [N], the above-described series of processes (S305 to S305). S312) is repeated.

  In other words, in the present embodiment, by executing the above-described series of processing (S304 to S312) for all the box sizes included in the box size data loaded in step 303, a set of numerical values corresponding to the number of box sizes ( N, S) is recorded.

Here, the relationship between the box size [N] and the root mean square [S] is defined as a function S (N) shown in the following equation (3). In the following formula (3), “M” indicates the number of elements of the time series data {y _i }, “N” indicates the box size, and “z _{jN + N} −z _{jN + 1} ” indicates the top of the j-th box. The difference (displacement) between the element (z _{jN + 1} ) and the last element (z _{jN + N} ) is shown.

  On the other hand, if it is determined in step 311 that a set (N, S) has been recorded for all box sizes [N] (step 311, Yes), the process proceeds to step 313 shown in FIG. 13.

  In step 313, the recorded set (N, S) is plotted on a log-log scale, and a linear function is fitted thereto.

  In the subsequent step 314, the slope of the fitted linear function is acquired as the scaling index α.

  In the subsequent step 315, the scaling index α is numbered in the order of acquisition and stored in the storage unit 24.

  In the following step 316, [c] is incremented by 1, and then the process returns to step 302 shown in FIG. Thereafter, the above-described series of processing (S302 to S316) is repeatedly executed.

As described above, the process in which the DFA execution unit 25 obtains the scaling index α based on the time series data {y _i } has been described. The DFA method described above is a preferable method because it is effective not only when the distribution function of fluctuation of the signal to be analyzed is a Gaussian distribution but also when it is a Levy distribution. However, when the distribution function of fluctuation of the signal to be analyzed is a Gaussian distribution, the scaling index may be obtained by another known DFA method. For example, the average F (N) of variance in each box of the time series data {y _i } is calculated based on the following formulas (4) and (5), the box size N and the variance F are plotted on a logarithmic scale, The slope of the linear function fitted thereto may be obtained as the scaling index α. In the following formulas (4) and (5), “N” indicates the box size, “M” indicates the number of elements of the time series data {y _i }, and “y _v (jN + k)” indicates the location trend. The fitting function of is shown.

  The processing executed by the DFA execution unit 25 has been described above. Subsequently, the processing executed by the scaling index integration unit 26 will be described based on the flowchart shown in FIG.

First, at the start of processing, a threshold _αth is set (step 401). Here, the threshold α _th is set with reference to the scaling index acquired by DFA of the analysis target signal when the object to be measured is normal. Specifically, a value obtained by adding an appropriate margin to the normal scaling index is set as the threshold α _th .

For example, the normal scaling index of a precision artificial rotating body represented by a precision bearing is a value close to “0” in general. An appropriate value in the range of 0.5 can be set as the threshold value _αth . In addition, since the scaling index of composites (motors, internal combustion engines, automobiles equipped with these, etc.) that are formed by nonlinearly coupling multiple parts is a value centered on "0.5", such composites When the body is a measurement target, an appropriate value in the range of 0.6 to 1.0 can be set as the threshold α _th .

  In the subsequent step 402, the counter [d] indicating the acquisition order of the scaling index α is set to the initial value “1”, and the integration counter [I] for integrating the values of the scaling index α is set to the initial value “0”. Set (step 402).

  In subsequent step 403, the d-th scaling index α is loaded from the storage unit 24.

In the next step 404, it is determined whether or not the loaded scaling index α is below the threshold value _αth . As a result, when the threshold value _αth is not below (No in Step 404), the value of the scaling index α loaded is added to the latest value of [I] (Step 405), and then the process proceeds to Step 407. On the other hand, if the threshold value _αth is below (step 404, Yes), the latest [I] is reset to the initial value [0] (step 406), and then the process proceeds to step 407.

  In the subsequent step 407, [d] is incremented by 1, and then the process returns to step 403 again. Thereafter, the series of processes (S403 to S407) described above are repeatedly executed.

Here, the reason why the latest [I] is reset to the initial value [0] when the loaded scaling index α is below the threshold α _th will be described.

FIG. 17A shows a time series graph 200 of the scaling index α acquired by the DFA execution unit 25. As shown in FIG. 17A, the scaling index α acquired by the DFA execution unit 25 greatly increases / decreases over time. Therefore, in the method of paying attention to the value of the scaling index α itself, the scaling index α is set inside the object to be measured. It is difficult to evaluate the resulting strain. With respect to this point, the present invention is characterized in that the state in which the value of the scaling index α _stays high over the threshold value αth as time passes is regarded as a state in which strain is accumulated inside the object to be measured. That is, in the present embodiment, by adopting the above-described step of resetting the integrated value (step 404 → step 406), the length of the period during which the scaling index α remains high is represented by the size of the integrated value. ing.

In order to understand the above-described points, FIG. 17B shows a time series graph 300 of integrated values of the scaling index α derived by the scaling index integration unit 26. The time series graph 300 is a time series graph of integrated values when the threshold α _th is set to [1.0] for the scaling index α shown in FIG. Here, if the time-series graph 200 and the time-series graph 300 are compared, in the integrated graph 200, the continuation of the period during which time the scaling index α remains high (enclosed by a broken line in the time-series graph 200). It will be understood that the integrated value has increased.

  The processing executed by the scaling index integration unit 26 has been described above. Subsequently, the processing executed by the output information generation unit 27 will be described.

  The output information generation unit 27 generates output information based on the latest integration value of the scaling index α derived by the scaling index integration unit 26. The output information here is a concept including image data to be output to the display 40, audio data to be output to the speaker 50, and control data for controlling another device 60.

  In the preferred embodiment, the output information generation unit 27 generates image data for displaying a time series graph of the integrated value of the scaling index α as shown in FIG. 17B based on the value of the counter [I]. To do. The output information generation unit 27 generates image data for displaying a time series graph reflecting the latest integrated value every time the integrated value of the scaling index α is updated, and outputs the image data to the display 40. In this case, in the time series graph displayed on the display 40, the user intuitively grasps that the strain is accumulated inside the object to be measured from the state that the triangular waveform of the line graph grows greatly. Can be seen as a sign of the destruction or collapse of the object. In another method, the latest integrated value of the scaling index α can be expressed by a bar graph. In this case, the user can grasp that the bar graph displayed on the display 40 is extended as a sign of destruction or collapse of the object.

  Further, the output information generation unit 27 generates audio data for generating an alarm sound according to the latest integrated value of the scaling index α based on the value of the counter [I]. Specifically, an appropriate threshold value of 1 or more is set for the integrated value, and when the integrated value reaches a predetermined threshold value, a sound for generating an alarm sound having a timbre or volume corresponding to the magnitude of the threshold value. Data is generated and output to the speaker 50. In this case, the user can recognize the alarm sound generated by the speaker 50 as a sign of destruction or collapse of the object to be measured.

  Further, the output information generation unit 27 generates control data for controlling the other device 60 based on the value of the counter [I]. Here, the other device 60 is, for example, the control device 60 for the moving object to be measured. In this case, when the integrated value reaches a predetermined threshold value, the output information generation unit 27 generates control data of the moving object to be measured according to the magnitude of the threshold value, and outputs the control data to the control device 60. The control data here is, for example, speed control data for controlling the speed of the moving object, and is preferably an emergency stop signal. In this case, the moving object is prevented from being destroyed or collapsed by receiving an emergency stop upon receiving the signal.

  The process executed by the output information generation unit 27 has been described above. Next, the process executed by the measurement mode setting unit 22 will be described.

In the present embodiment, a plurality of measurement modes corresponding to the type of measurement object to be planned are prepared, and preset data of various parameters described above is stored in the storage unit 24 for each measurement mode. As preset data in this embodiment, parameters used by the peak time acquisition unit 21 (signal sampling rate, detection period T1, standby period T2, threshold voltage V _Th ), parameters used by the DFA target data generation unit 23 (A, B, M), parameters used by the DFA execution unit 25 (box size range), parameters used by the scaling index integration unit 26 (threshold α _th ), the frequency of the excitation signal generated by the excitation signal generation unit 32, and sine The frequency of the sine wave signal to be oscillated by the wave oscillation circuit 34 can be exemplified. In the present embodiment, the measurement mode setting unit 22 loads preset data corresponding to the measurement mode selected by the user from the storage unit 24, and sets each parameter included in the loaded preset data in each corresponding function unit.

  As described above, according to the signal analysis system of this embodiment, by continuously analyzing the fluctuation of the vibration waveform generated by the object using the DFA, the accumulation state of the strain generated inside the object is integrated with the scaling index. The value can be quantitatively evaluated. For example, when the above-described parameter “A” is set to 300 for an analysis target signal having a frequency of 300 Hz, the signal analysis system according to the present embodiment displays the accumulation state of strain generated inside the object with a time resolution of about 1 second. Can be evaluated.

  In addition, according to the signal analysis system of this embodiment, by constantly installing a vibration sensor on an object to be measured, it is possible to constantly monitor the state of the object. In this case, the sensor signal of the vibration sensor may be acquired continuously all the time, or the acquisition of the sensor signal over a predetermined period may be repeated intermittently.

  As described above, the present invention has been described with the embodiment. However, the present invention is not limited to the above-described embodiment, and as long as the operations and effects of the present invention are exhibited within the scope of embodiments that can be considered by those skilled in the art. It is included in the scope of the present invention.

  Each function of the computer 20 described above can be realized by a device-executable program written in C, C ++, C #, Java (registered trademark), etc. The program of this embodiment includes a hard disk device, It can be stored and distributed on a device-readable recording medium such as a CD-ROM, MO, DVD, flexible disk, EEPROM, EPROM, etc., and can be transmitted via a network in a format that other devices can.

  Hereinafter, the signal analysis system of the present invention will be described more specifically with reference to examples. However, the present invention is not limited to the examples described later.

(Experiment 1: Measurement of a moving object that oscillates periodically)
After setting the motor on the fixed base, the motor was driven at about 300 Hz. In this experiment, the motor was covered with glass wool to encourage the destruction of the motor so that heat was trapped inside. In this state, the sensor signal of the vibration sensor (piezo element) attached to the fixed base was analyzed using the signal analysis system of the present invention. Table 1 below collectively shows the parameter values set in the signal analysis system in Experiment 1.

(Result of Experiment 1)
A short time after starting the motor drive, the triangular waveform began to grow greatly in the line graph displayed by the signal analysis system, but at this point in time, no particular change was felt. After a while, an abnormal noise was generated from the motor, and finally the motor stopped.

(Experiment 2: Measurement of a stationary object)
After fixing a bar made of aluminum on a fixed base, vibration was input by fixing a speaker that transmits a sound of frequency 300 Hz to the bar. On top of that, stress was generated in the bar by gradually injecting water into a water tank hung on one end of the bar. In this state, the sensor signal of the vibration sensor (piezo element) attached to the bar was analyzed using the signal analysis system of the present invention. Table 2 below collectively shows the parameter values set in the signal analysis system in Experiment 2.

(Result of Experiment 2)
A short time after starting the injection of water into the water tank, a triangular waveform began to grow greatly in the line graph displayed by the signal analysis system of the present invention. After a while, the bar was bent by the weight of Minamata.

  From the above experimental results, it was shown that according to the signal analysis system of the present invention, a sign of destruction or collapse of an object can be captured based on the integral value of the scaling index.

DESCRIPTION OF SYMBOLS 10 ... Vibration sensor 12 ... Exciter 20 ... Computer 21 ... Peak time acquisition part 22 ... Measurement mode setting part 23 ... DFA object data generation part 24 ... Storage part 25 ... DFA execution part 26 ... Scaling index integration part 27 ... Output information Generation unit 32 ... Excitation signal generation unit 34 ... Sine wave oscillation circuit 36 ... Signal superposition circuit 40 ... Display 50 ... Speaker 60 ... Other device 100 ... Signal analysis system 200 ... Time series data of scaling index 300 ... Integration of scaling index Value time series data

Claims (22)

A peak time acquisition unit for acquiring the peak time of the analysis target signal;
A DFA target data generation unit that generates DFA target data based on the peak time acquired by the peak time acquisition unit;
A DFA execution unit that acquires a scaling index by executing DFA based on the DFA target data;
A scaling index integration unit for integrating the acquired scaling index;
Including
The DFA target data generation unit
The peak times acquired by the peak time acquisition unit are loaded in order of time series (A + 1), A peak intervals are calculated based on the loaded (A + 1) peak times, and the A peaks A process of repeatedly generating pseudo time-series data including M elements as the DFA target data based on time-series data having intervals as elements (A and M are both natural numbers, and M is more than A Big.
The DFA execution unit
Loading the DFA target data generated by the DFA target data generation unit in the order of generation, repeatedly executing the process of obtaining the scaling index based on the loaded DFA target data,
The scaling index integration unit is
The scaling exponents acquired by the DFA execution unit are loaded in the order of acquisition, and if the loaded scaling exponent is not below a predetermined threshold, the scaling exponent is added to the latest integrated value, and the loaded scaling exponent is If it is below the predetermined threshold, repeatedly execute the process of resetting the latest integrated value,
Signal analysis system. The DFA target data generation unit
The time series data having the A peak interval as an element is divided into a plurality of segments, and one randomly selected segment is added to the rear of the time series data to repeat the M peak intervals as elements. Generating pseudo time-series data, and generating pseudo time-series data obtained by integrating values obtained by subtracting an average value from each element of the pseudo-time-series data as the DFA target data.
The signal analysis system according to claim 1.   The signal analysis system according to claim 1, further comprising an output information generation unit that generates output information based on the integrated value.   The signal analysis system according to claim 3, wherein the output information is image data for displaying a graph reflecting the latest integrated value.   The signal analysis system according to claim 4, wherein the graph is a time series graph of the integrated values.   The signal output system according to any one of claims 3 to 5, wherein the output information is audio data for generating an alarm sound.   The signal analysis system according to any one of claims 3 to 6, wherein the output information is control data for controlling another device.   The signal analysis system according to any one of claims 1 to 6, wherein A is a natural number in a range of 200 to 300, and M is a natural number in a range of 25000 to 35000.   The signal analysis system according to claim 1, wherein the analysis target signal is a sensor signal that detects vibration of a moving object. The moving object is a bearing;
The signal analysis system according to claim 9, wherein a range of a box size for acquiring a scaling index by the DFA execution unit is a range centering on 10 to 30.   The signal analysis system according to claim 1, wherein the analysis target signal is a sensor signal for detecting vibration of an object to which periodic vibration is input from the outside.   The signal analysis system according to claim 1, wherein the analysis target signal is a superimposed signal of a sensor signal for detecting vibration of an object and a predetermined sine wave signal.   The signal analysis system according to any one of claims 9 to 12, wherein the predetermined threshold is set with reference to a scaling index acquired by a DFA of the analysis target signal related to the object in a normal state.   According to the selected measurement mode, one or more parameters are automatically selected from among the parameters related to the A, M, and the DFA execution unit for obtaining the scaling index and the box size range and the predetermined threshold value, respectively. The signal analysis system according to claim 1, further comprising a measurement mode setting unit that is set to A method for analyzing a signal comprising:
Obtaining a peak time of the signal to be analyzed;
Generating DFA target data based on the acquired peak time;
Performing a DFA based on the DFA target data to obtain a scaling index;
Accumulating the obtained scaling index;
Including
The step of generating the DFA target data includes:
The peak times acquired by the peak time acquisition unit are loaded in order of time series (A + 1), A peak intervals are calculated based on the loaded (A + 1) peak times, and the A peaks It is a step of repeatedly executing a process of generating pseudo time-series data including M elements as the DFA target data based on time-series data having intervals as elements (A and M are natural numbers, and M is Greater than A. The same shall apply hereinafter.)
Obtaining the scaling index comprises:
Loading the generated DFA target data in the order of generation and repeatedly executing the process of obtaining the scaling index based on the loaded DFA target data;
The step of accumulating the scaling index includes:
The acquired scaling exponents are loaded in the order of acquisition, and when the loaded scaling exponent is not below a predetermined threshold, the scaling exponent is added to the latest integrated value, and the loaded scaling exponent is reduced to the predetermined threshold. It is a step that repeatedly executes the process of resetting the latest integrated value if it is below,
Method.   The method according to claim 15, wherein the analysis target signal is a sensor signal for detecting vibration of a moving object. The moving object is a bearing;
The method of claim 16, wherein a range of box sizes for obtaining a scaling index in the DFA is a range centered around 10-30. The moving object is a complex formed by nonlinearly coupling a plurality of parts,
The method of claim 16, wherein a range of box sizes for obtaining a scaling index in the DFA is a range centered around 30-60.   The method according to claim 15, wherein the analysis target signal is a sensor signal for detecting vibration of an object to which periodic vibration is input from the outside.   The method according to claim 15, wherein the analysis target signal is a superimposed signal of a sensor signal for detecting vibration of an object and a predetermined sine wave signal. The analysis target signal is a periodic biological signal,
The method of claim 15, wherein a range of box sizes for obtaining a scaling index in the DFA is a range centered around 30-270. A computer-executable program for causing a computer to execute the steps of the method according to any one of claims 15 to 21.