JP2015161506A - Signal analyzer, signal analysis method, degradation diagnosis apparatus, and degradation diagnosis method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To overcome the problem that, at a time of extracting the feature of a signal waveform exhibiting periodicity in equipment operation so as to grasp the state of the equipment mounted in a vehicle, a waveform is unable to be appropriately extracted if a reference waveform differs in shape from a target waveform and a high correlation coefficient is not obtained.SOLUTION: A signal analyzer 1 comprises: an intersection extraction unit 3 extracting intersections for a signal waveform; a target-waveform generation unit 6 generating a comparison target waveform from the waveform and the extracted intersections; a candidate-waveform extraction unit 4 extracting a candidate waveform from the waveform and the extracted intersections; a comparison-waveform generation unit 5 generating a comparison waveform equal in length to the comparison target waveform by repeating the candidate waveform; a correlation-coefficient calculation unit (evaluation-value calculation unit) 7 calculating a correlation coefficient (evaluation value) between the comparison target waveform and the comparison waveform generated by repeating the candidate waveform; and a waveform determination unit 9 determining a waveform corresponding to one period on the basis of the calculated correlation coefficient (evaluation value).

Description

  The present invention relates to a signal analysis apparatus and a signal analysis method for extracting and analyzing a waveform of a signal having periodicity in the operation of an apparatus.

Mainly in the railway field, it is necessary to grasp the state of the equipment in order to check the equipment mounted on the vehicle at an appropriate timing.
For the purpose of grasping the state of a device mounted on a vehicle, a deterioration diagnosis device and a deterioration diagnosis method capable of extracting the characteristics from the waveform of a signal having periodicity in the operation of the device are described ( Patent Document 1). The deterioration diagnosis apparatus described in Patent Document 1 determines deterioration by comparing a normal waveform held in advance with a newly measured waveform.
Further, Fourier analysis and wavelet analysis are generally used as methods for analyzing periodic signals (see Patent Document 2). In Patent Literature 2, a waveform obtained by approximating a certain section on the time axis with a wavelet function as a reference waveform for a waveform having no noise of the same type as the target waveform, and the target waveform on which the reference waveform and noise are superimposed Describes a technique for extracting a waveform based on the correlation coefficient between

JP2011-257154A JP 2000-276463 A

  Since the techniques described in Patent Documents 1 and 2 both require the preparation of a reference waveform, when the shape of the reference waveform is different from the target waveform and a high correlation coefficient cannot be obtained, waveform extraction can be performed appropriately. There was no problem.

  The present invention has been made to solve the above-described problem, and an object of the present invention is to obtain a signal analysis apparatus that takes into account the constraints caused by the time relationship.

  The signal analysis device according to the present invention is a candidate from an intersection extraction unit that extracts an intersection with a signal waveform, a target waveform generation unit that generates a comparison target waveform from the waveform and the extracted intersection, and a waveform and the extracted intersection A candidate waveform extraction unit for extracting a waveform, a comparison waveform having the same length as the comparison target waveform generated by repeating the candidate waveform, and an evaluation value calculation unit for calculating an evaluation value of the comparison target waveform. A waveform determining unit for determining a waveform for one period based on the evaluation value.

  According to the present invention, a candidate waveform is obtained from an intersection extraction unit that extracts an intersection with respect to a waveform of a signal, a target waveform generation unit that generates a comparison target waveform from the waveform and the extracted intersection, and a waveform and the extracted intersection. A candidate waveform extraction unit to extract, a comparison waveform having the same length as the comparison target waveform generated by repeating the candidate waveform, an evaluation value calculation unit that calculates an evaluation value of the comparison target waveform, and a calculated evaluation Since it has a waveform determination unit that determines the waveform for one cycle based on the value, there is no need to prepare a reference waveform different from the target waveform in advance, and waveform extraction for one cycle is possible without depending on the specific waveform shape can do.

1 is a configuration diagram of a signal analysis device according to Embodiment 1. FIG. 3 is a flowchart showing an operation in the first embodiment. An example of extracting intersection information from the target waveform in the first embodiment. FIG. 3 shows an example of generating a comparison target waveform from the target waveform and intersection information in the first embodiment. The example in which the candidate waveform extraction part 4 in Embodiment 1 extracts a candidate waveform. The figure which shows the comparison waveform group produced | generated based on the candidate waveform group of FIG.5 (b). FIG. 3 is a configuration diagram of a signal analysis device according to a second embodiment. The flowchart which shows the operation | movement in Embodiment 2 of this invention. Explanatory drawing of the coordinate distance of the coordinate space which uses the 1st feature-value and 2nd feature-value as coordinate axes in Embodiment 2, and a comparison object waveform and each comparison waveform.

Embodiment 1 FIG.
In the railway field, state maintenance is to determine the inspection timing and inspection contents of railway vehicles according to the state of the equipment. If the equipment status cannot be accurately grasped, maintenance measures are performed at inappropriate timing or maintenance procedures with inappropriate contents are performed. is necessary.
In the operation of equipment mounted on railway vehicles, signals indicating voltage, current, temperature, vibration, etc. obtained from equipment are waveforms having periodicity, and waveforms for comparison with waveforms in a normal state. A signal analysis apparatus that performs extraction will be described below.

FIG. 1 is a configuration diagram of a signal analysis apparatus according to Embodiment 1 of the present invention.
The signal analysis apparatus 1 includes a target waveform acquisition unit 2, an intersection extraction unit 3, a candidate waveform extraction unit 4, a comparative waveform generation unit 5, a target waveform generation unit 6, a correlation coefficient calculation unit 7, and a correlation A number storage unit 8 and a waveform determination unit 9 for determining a waveform for one cycle are provided.

  The target waveform acquisition unit 2 acquires state information that is signal information such as voltage, current, temperature, and vibration from a device or a sensor attached to the device. A waveform with the vertical axis representing the value of the acquired device status information and the horizontal axis representing time is taken as the target waveform. The target waveform acquisition unit 2 outputs the acquired target waveform to the intersection extraction unit 3, candidate waveform extraction unit 4, and target waveform generation unit 6.

  The intersection extraction unit 3 takes the intersection of a specific value and the target waveform among the values of the target waveform that is the state information input from the target waveform acquisition unit 2, and extracts the value in the horizontal axis direction of the intersection as intersection information. . In addition, the intersection extraction unit 3 outputs the extracted intersection information to the candidate waveform extraction unit 4 and the target waveform generation unit 6.

  The candidate waveform extraction unit 4 sets a specific intersection as a starting point for waveform extraction based on the target waveform transmitted from the target waveform acquisition unit 2 and the intersection information transmitted from the intersection extraction unit 3, and after the start point A specific intersection is set as the end point of waveform extraction, and a waveform in the range from the start point to the end point is extracted as a candidate waveform. Further, the candidate waveform extraction unit 4 transmits the extracted candidate waveform group to the comparison waveform generation unit 5.

  Based on the target waveform transmitted from the target waveform acquisition unit 2 and the intersection information transmitted from the intersection extraction unit 3, the target waveform generation unit 6 performs a horizontal axis that is a time axis in the intersection group from the start point of the target waveform. A waveform obtained by removing up to the intersection where the value in the axial direction is minimum is extracted as a comparison target waveform. Further, the target waveform generation unit 6 transmits the comparison target waveform to the correlation coefficient calculation unit 7 and transmits the length of the comparison target waveform in the horizontal axis direction to the comparison waveform generation unit 5.

  The comparison waveform generation unit 5 generates each candidate waveform based on the candidate waveform group transmitted from the candidate waveform extraction unit 4 and the length in the horizontal axis direction of the comparison target waveform transmitted from the comparison target waveform generation unit 5. , And a comparison waveform group listed in the horizontal axis direction until a length equal to the length in the horizontal axis direction of the comparison target waveform is generated. Further, the comparison waveform generation unit 5 sets the comparison waveform group and a candidate waveform that is a generator of each comparison waveform as a set, and transmits the set to the correlation coefficient calculation unit 7.

  The correlation coefficient calculation unit 7 calculates a correlation coefficient between the comparison target waveform transmitted from the target waveform generation unit 6 and each comparison waveform of the comparison waveform group transmitted from the comparison waveform generation unit 5. The calculated correlation coefficient is output to the correlation coefficient storage unit 8 together with a candidate waveform that is a generator of each comparison waveform.

  The correlation coefficient storage unit 8 sets a correlation coefficient between the comparison target waveform and each comparison waveform transmitted from the correlation coefficient calculation unit 7 and a candidate waveform that is a generator of each comparison waveform. And output to the waveform determination unit 9.

  The waveform determination unit 9 extracts a correlation coefficient held in the correlation coefficient storage unit 8 that exceeds a set threshold as a waveform for one cycle. When there is no correlation coefficient exceeding the threshold value, the candidate waveform extraction unit 4 is supported to continue the operation with the starting point in extracting the candidate waveform as the next intersection of the current starting point.

Next, the operation of the present invention will be described.
The outline of the operation is to acquire state information that is signal information such as voltage, current, temperature, vibration, etc. from sensors attached to the device, and determine which cycle is correct as state information from the acquired state information. To do. The acquired state information is a waveform having a plurality of periodicities, and by obtaining an evaluation value (correlation coefficient in this embodiment) for a candidate waveform of a period for a waveform in a certain period, a plurality of periodicities can be obtained. It is possible to determine which period is correct as the state information among the waveforms having. It is also possible to use the determined period and the state information to measure the degree of deterioration of the state information of the state information device.
The operation procedure will be described below.

FIG. 2 is a flowchart showing the operation in the first embodiment of the present invention.
In step S <b> 101, the target waveform acquisition unit 2 acquires state information that is signal information such as voltage, current, temperature, and vibration from a device or a sensor attached to the device. A waveform with the vertical axis representing the value of the acquired device status information and the horizontal axis representing time is taken as the target waveform. The target waveform acquisition unit 2 outputs the acquired target waveform to the intersection extraction unit 3, candidate waveform extraction unit 4, and target waveform generation unit 6.

In step S <b> 102, the intersection extraction unit 3 takes an intersection between a specific value and the target waveform among the values of the target waveform that is the state information input from the target waveform acquisition unit 2, and calculates a value in the horizontal axis direction of the intersection. Extracted as intersection information. In addition, the intersection extraction unit 3 outputs the extracted intersection information to the candidate waveform extraction unit 4 and the target waveform generation unit 6.
FIG. 3 shows an example of extracting intersection information from the target waveform in the first embodiment.
FIG. 3 shows a case where the specific value that intersects the target waveform is the median value calculated from the maximum value and the minimum value of the target waveform. The vertical axis represents the value of the target waveform, and the horizontal axis represents time. For a specific value, the value may be set freely, and the value may be set to 0. For example, among the positions between the maximum and minimum values, the shortest interval between the intersections can be taken by the target waveform. By setting the value to be the longest among the specific values and to avoid the proximity of the intersections most, it is possible to avoid the influence on unnecessary noise and reduce the number of candidate waveforms to be described later.

In step S103, the target waveform generation unit 6 removes a portion from the start point of the target waveform to the intersection having the smallest value in the horizontal axis direction from the start point of the target waveform based on the target waveform and the intersection information. Are generated as a comparison target waveform, and the generated comparison target waveform is output to the correlation coefficient calculation unit 7.
FIG. 4 shows an example of generating a comparison target waveform from the target waveform and the intersection information in the first embodiment. The upper graph in FIG. 4 shows the target waveform, and the lower graph shows the comparison target waveform generated from the target waveform and the intersection information.

In step S104 and step S105, the candidate waveform extraction unit 4 starts the waveform extraction from a specific intersection based on the target waveform transmitted from the target waveform acquisition unit 2 and the intersection information transmitted from the intersection extraction unit 3. The specific intersection after the start point is set as the end point of waveform extraction, and the waveform in the range from the start point to the end point is extracted as a candidate waveform. Further, the candidate waveform extraction unit 4 outputs the extracted candidate waveform group to the comparison waveform generation unit 5.
FIG. 5 shows an example in which the candidate waveform extraction unit 4 in Embodiment 1 extracts candidate waveforms. Upper (a) shows a target waveform, and (b) shows a candidate waveform extracted from the target waveform and intersection information.

In step S104, the candidate waveform extraction unit 4 first sets the starting point for extracting the candidate waveform based on the intersection information as an intersection (at (a) in FIG. 5A) having the smallest value in the horizontal axis direction in the intersection group. 1) (The circle is indicated by () and is defined as (1)).
Next, in step S105, the candidate waveform extraction unit 4 sets the end points from which the candidate waveform is extracted at the intersections ((2) to (11 in FIG. 5A) in the descending order of the value in the horizontal axis direction after the current start point. )) Respectively. The candidate waveform extraction unit 4 is a group of candidate waveforms ((1) to (2), (1) to (3),..., The last intersection point) from the start point and the intersection point as shown in FIG. (1) to (11)) are extracted, and the candidate waveform group is output to the comparison waveform generation unit 5.

2, the comparison waveform generation unit 6 generates a comparison waveform group based on the candidate waveform group and the length of the comparison target waveform in the horizontal axis direction, and the candidate waveform group and the comparison waveform group are also generated. The set is output to the correlation coefficient calculation unit 7.
FIG. 6 shows a comparison waveform group generated based on the candidate waveform group of FIG.
Each comparison waveform is a waveform that is duplicated in the horizontal axis direction and enumerated in the horizontal axis direction until it reaches the length in the horizontal axis direction of the comparison target waveform.

  In step S107 of FIG. 2, the correlation coefficient calculation unit 7 calculates the correlation coefficient between the comparison target waveform and each comparison waveform. Also, the candidate waveform that is a generator of each comparison waveform and a correlation coefficient are paired and output to the correlation coefficient storage unit 8, and the correlation coefficient storage unit 8 is a candidate that is a generator of each comparison waveform A waveform and a correlation coefficient are stored as a pair.

In step S <b> 108, the waveform determination unit 9 evaluates whether there is a correlation coefficient stored in the correlation coefficient storage unit 8 that exceeds the set threshold value.
In step S <b> 109, when there is a correlation coefficient between the comparison target waveform and each comparison waveform that exceeds the threshold value, the waveform determination unit 9 selects a candidate waveform paired with the correlation coefficient exceeding the threshold value for one cycle. Extracted as a waveform. Further, the extracted waveform portion in the comparison target waveform is removed from the comparison target waveform. If there are a plurality of samples exceeding the threshold, the candidate waveform paired with the highest correlation coefficient is extracted as a waveform for one cycle.

In step S110, the waveform determination unit 9 evaluates whether another intersection exists after the intersection corresponding to the end point of the waveform extracted from the comparison target waveform in step S109. If there is no other intersection, it ends. If it demonstrates with FIG.5 (b), it will correspond when it is a comparison waveform produced | generated using candidate waveform (1)-(11).
In step S111, when another intersection exists after the intersection corresponding to the end point of the extracted candidate waveform in the comparison target waveform, the waveform determination unit 9 sets the current end point as a new start point. Then, it returns to step S105 and repeats waveform extraction for one period.
In FIG. 5B, for example, when the comparison waveform generated using the candidate waveforms (1) to (6) exceeds the threshold in the correlation coefficient, the target waveform (6) is set as the starting point.

In step S112, the waveform determination unit 9 evaluates whether there is another intersection after the intersection that is the current start point in the target waveform when there is no threshold exceeding the threshold in the evaluation in step S108.
Explaining with FIG. 5B, for example, when the comparison waveform generated using the candidate waveforms (1) to (6) does not exceed the threshold in the correlation coefficient, there is an intersection after (1) of the target waveform. Evaluate what to do.

When another intersection exists after the intersection that is the current start point, the waveform determination unit 9 sets the start point when extracting the candidate waveform as the next intersection ((2) in FIG. 5B) of the current start point. Thus, returning to S105, the waveform extraction is repeated (S113).
If there is no other intersection after the intersection corresponding to the end point of the extracted waveform in the evaluation of S110, or if there is no other intersection after the intersection that is the current start point in the evaluation of S112, the waveform extraction ends. .

  As described above, the signal analysis apparatus 1 according to the first embodiment includes the intersection extraction unit 3 that extracts intersections with respect to the waveform of the signal, and the target waveform generation unit 6 that generates a comparison target waveform from the waveforms and the extracted intersections. A candidate waveform extraction unit 4 that extracts a candidate waveform from the waveform and the extracted intersection, a comparison waveform generation unit 5 that generates a comparison waveform having the same length as the comparison target waveform by repeating the candidate waveform, and a comparison target Based on a correlation coefficient calculation unit (evaluation value calculation unit) 7 that calculates a correlation coefficient (evaluation value) between a waveform and a comparison waveform generated by repeating the candidate waveform, and the calculated correlation coefficient (evaluation value) Since the waveform determining unit 9 for determining the waveform for one period is provided, it is not necessary to prepare a reference waveform different from the target waveform in advance, and the waveform for one period is extracted without depending on a specific waveform shape. Can do.

In addition, in order to diagnose the deterioration state of equipment such as trains, the comparison target waveform needs to be measured when the equipment is unused, new, or normal. It is necessary to measure at the time of use including the time of deterioration.
For example, if a reference waveform is used as a comparison target waveform and a comparison waveform is generated using a candidate waveform that is a waveform at the stage of use, the advantage of evaluating similarity is that it can be evaluated on an actual machine basis. It is in. In other words, the conventional example extracts a waveform for one cycle using a virtual waveform as a reference waveform. However, since the reference waveform cannot completely simulate the normal waveform of the actual machine, the waveform can be extracted appropriately. As a result, there is a possibility of diagnosing an abnormality even though the device is normal.

In that respect, the signal analysis apparatus according to the present embodiment can solve the above-mentioned problems, and uses the waveform at the reference time of the device as a comparison target waveform and compares it using a candidate waveform that is a waveform at the use stage. When a waveform is generated, a correlation coefficient when a waveform for one period is determined by the waveform determination unit is held, so that it is possible to record a change over time of each actual machine. By recording the secular change based on the standard time of the actual machine, there is also an advantage that the amount of change from the standard time can be known, and it is easy to predict where the failure will occur from the change in the value.
By using the evaluation value (correlation coefficient) calculated by the signal analysis apparatus for deterioration diagnosis in the deterioration diagnosis apparatus, it is possible to perform deterioration diagnosis based on the reference time of the actual machine.

Embodiment 2. FIG.
In the first embodiment, a comparison waveform having a waveform for one period as a generator is determined based on a correlation coefficient between the comparison waveform and the comparison target waveform. Next, a comparison target waveform and a comparison waveform group are used. Singular value decomposition is performed, and the distance (coordinate distance) in the coordinate space of the comparison target waveform and each comparison waveform in the coordinate space with the obtained first feature value and second feature value as coordinate axes is obtained. By calculating and determining the comparison waveform that uses the waveform for one cycle as a generator, the comparison waveform has a waveform shape that is closer to the intersection start target waveform, and the waveform for one cycle is higher than the correlation coefficient. The case of extracting with accuracy is shown. Hereinafter, the difference will be described based on the contents described in the first embodiment.

FIG. 7 is a configuration diagram of the control device support apparatus according to the second embodiment.
7 is different from the first embodiment in that a coordinate distance calculation unit 10 is provided instead of the correlation coefficient calculation unit 7 and a coordinate distance storage unit 11 is provided instead of the correlation coefficient calculation unit 8. Further, the waveform determining unit 9 is different in that the determination is based on the coordinate distance, not based on the correlation coefficient.

The operation will be described below.
The outline of the operation is the same as in the first embodiment, from the sensor attached to the device, from the state information that is signal information such as voltage, current, temperature, vibration, etc. In this form, it is used to measure the degree of deterioration of the device by obtaining the coordinate distance). Therefore, the time to acquire the comparison target waveform from the target waveform described below is acquired from the time when the equipment is new or normal, and the candidate waveform is measured at the time when the equipment is used and periodically inspected. Can be used to determine deterioration of train equipment.
FIG. 8 is a flowchart showing the operation in the second embodiment of the present invention. Note that description of the same operation as in the first embodiment will be omitted.
Steps S201 and S202 in FIG. 8 are the same as steps S101 and S102 in FIG.
In step S203 in FIG. 8, instead of the target waveform generation unit 6 in step S103 in FIG. 2 outputting the comparison target waveform to the correlation coefficient calculation unit 7, the target waveform generation unit 6 converts the comparison target waveform into a coordinate distance calculation unit. 10 is output.

Steps S204 and S205 in FIG. 8 are the same as steps S104 and S105 in FIG.
In step S206 in FIG. 8, instead of the comparison waveform generation unit 5 in step S106 of FIG. 2 outputting each candidate waveform and the comparison waveform to the correlation coefficient calculation unit 7, the comparison waveform generation unit 5 The waveform and the comparison waveform are output to the coordinate distance calculation unit 10.

  In step S207 in FIG. 8, the coordinate distance calculation unit 10 performs singular value decomposition using the comparison target waveform and the comparison waveform group as inputs. Then, a coordinate distance between the comparison target waveform and each comparison waveform in a coordinate space having the obtained first feature value and second feature value as coordinate axes is calculated. The calculated coordinate distance is output to the coordinate distance storage unit 11, and the coordinate distance storage unit 11 holds the coordinate distance. At this time, the coordinate distance of the comparison waveform is held in combination with a candidate waveform that is a generator of each comparison waveform.

The significance of the coordinate distance between the comparison target waveform and each comparison waveform in the coordinate space having the first feature value and the second feature value calculated by performing the singular value decomposition performed by the coordinate distance calculation unit 10 as coordinate axes will be described.
FIG. 9 is an explanatory diagram of the coordinate distance between the first feature value and the second feature value as coordinate axes, and the comparison target waveform and each comparison waveform in the second embodiment.

Singular value decomposition is a method of matrix decomposition used in signal processing and statistics. By using singular value decomposition, it is possible to calculate a feature space with a smaller number of dimensions that well reflects the characteristics by combining the axes of multidimensional numerical data. In general, when the matrix SP to be input, the data SP _{i in} the i-th row of SP, the singular value σ, the singular vector coefficient u, and the singular vector v are set, the following relationship is established.

Since the feature amounts are generally expressed as a singular vector v, the singular vector v is used as a coordinate in the feature space, and the first feature amount and the second feature amount are evaluated in two dimensions.
Singular value decomposition is an algorithm that quantifies (features) components with different characteristics between a plurality of data patterns. For this reason, a larger difference in feature amount is shown for irregular changes than for regular changes. For example, a waveform data group in which the amplitude and period change constantly has a small difference in feature amount, and when a waveform data having a different change pattern is evaluated, a significantly different feature amount is derived.

  The first feature amount is, for example, an evaluation value regarding the offset of the amplitude component of the waveform between the comparison target waveform and the comparison waveform. That is, the value indicates how much the comparison waveform is shifted in the amplitude direction with respect to the comparison target waveform. A waveform having a large value of the first feature value obtained by the singular value decomposition is entirely shifted in the amplitude direction with respect to the waveform to be compared.

  The second feature amount is, for example, an evaluation value related to a difference in waveform shape. That is, in the state where the difference in amplitude component between the waveform to be compared and the comparison waveform is canceled, the difference considering the amplitude level of each waveform is added (for example, for a plurality of evaluation points). . A waveform having a large second feature value obtained by singular value decomposition generally has a different waveform shape from other waveforms.

Therefore, a sample of a certain waveform in a coordinate space having the first feature value (evaluation value of deviation in amplitude direction) and the second feature value (evaluation value of waveform shape difference) obtained by singular value decomposition as coordinate axes, The distance from another waveform sample represents the similarity of the two waveforms, and can be determined with higher accuracy than the determination based on the correlation coefficient (overlapping waveform), and extracted as a waveform for one cycle. The comparison waveform generated based on the candidate waveform to be determined can be more accurately determined.
Specifically, when the correlation coefficient in the first embodiment is used, when comparing periodic signals, the amplitudes are different, but the phases are matched, the correlation is made. Since there is no change, as long as the phase matches, the anomaly due to the magnitude of the amplitude cannot be detected and it can be detected as normal, but it was obtained by singular value decomposition In the case of the feature amount, since it appears as a value at least in the second feature amount, it is possible to evaluate the waveform for one period in consideration of the magnitude of the amplitude.

In step S <b> 208 in FIG. 8, the waveform determination unit 9 evaluates whether or not there is a value below the set threshold within the coordinate distance held in the correlation coefficient storage unit 8.
In step S <b> 109, if there is a coordinate distance within the threshold among the comparison distances between the comparison target waveform and each comparison waveform, the waveform determination unit 9 selects a candidate waveform paired with the coordinate distance within the threshold for one cycle. Extracted as a waveform. Further, the extracted waveform portion in the comparison target waveform is removed from the comparison target waveform. When there is a coordinate distance within the threshold, a candidate waveform paired with the coordinate distance that is the smallest value is extracted as a waveform for one cycle.

In step S210, the waveform determination unit 9 evaluates whether another intersection exists after the intersection corresponding to the end point of the waveform extracted from the comparison target waveform in step S209. If there is no other intersection, it ends. Similar to the first embodiment, the case illustrated in FIG. 5B corresponds to the case where the comparison waveforms are generated using the candidate waveforms (1) to (11).
In step S211, when another intersection exists after the intersection corresponding to the end point of the extracted candidate waveform in the comparison target waveform, the waveform determination unit 9 sets the current end point as a new start point. Then, it returns to step S205 and repeats waveform extraction for one period.
Similar to the first embodiment, FIG. 5B illustrates, for example, when the comparison waveform generated using the candidate waveforms (1) to (6) is within the threshold value in the correlation coefficient, 6) is the starting point.

In step S212, the waveform determination unit 9 evaluates whether there is another intersection after the intersection that is the current start point in the target waveform when there is no threshold value within the evaluation in step S108.
5B, as in the first embodiment, for example, when the comparison waveform generated using the candidate waveforms (1) to (6) does not fall within the threshold in the correlation coefficient, (1 ) Evaluate if there is an intersection after that.

When another intersection exists after the intersection that is the current start point, the waveform determination unit 9 sets the start point when extracting the candidate waveform as the next intersection ((2) in FIG. 5B) of the current start point. Thus, returning to S105, the waveform extraction is repeated (S213).
If there is no other intersection after the intersection corresponding to the end point of the extracted waveform in the evaluation of S210, or if there is no other intersection after the intersection that is the current start point in the evaluation of S112, the waveform extraction ends. .

  As described above, the signal analyzing apparatus 1 according to the second embodiment includes the intersection extraction unit 3 that extracts intersections with respect to the waveform of the signal, and the target waveform generation unit 6 that generates a comparison target waveform from the waveforms and the extracted intersections. The candidate waveform extraction unit 4 that extracts a candidate waveform from the target waveform during operation and the extracted intersection point, and the comparison waveform generation unit 5 that generates a comparison waveform having the same length as the comparison target waveform by repeating the candidate waveform And a coordinate distance calculation unit (evaluation value calculation unit) 10 that calculates a coordinate distance (evaluation value) between the comparison target waveform and the comparison waveform generated by repeating the candidate waveform, and the calculated coordinate distance (evaluation value). Since the waveform determination unit 9 for determining the waveform for one period is provided, it is not necessary to prepare a reference waveform different from the target waveform in advance, and the waveform for one period is extracted without depending on a specific waveform shape. be able to.

In addition, by using the coordinate distance obtained from the first feature value (evaluation value of deviation in the amplitude direction) and the second feature value (evaluation value of the difference in waveform shape), the evaluation is performed in consideration of the magnitude of the amplitude. Is possible.
Further, depending on the content of the waveform to be detected, the coordinate distance may be obtained using only the first feature value or the second feature value, and the coefficient for the first feature value or the second feature value is used. The coordinate distance may be obtained.

In addition, in order to diagnose the deterioration state of equipment such as trains, the comparison target waveform needs to be measured when the equipment is unused, new, or normal. It is necessary to measure at the time of use including the time of deterioration.
For example, if a reference waveform is used as a comparison target waveform and a comparison waveform is generated using a candidate waveform that is a waveform at the stage of use, the advantage of evaluating similarity is that it can be evaluated on an actual machine basis. It is in. In other words, the conventional example uses some virtual waveform as the reference waveform, but there is a problem that the reference waveform can completely simulate the normal waveform of the actual machine, and it can be abnormal even if it operates normally There is sex.

In that respect, the signal analysis apparatus according to the present embodiment can solve the above-described problems and can record the aging change of each actual machine. By recording the secular change based on the standard time of the actual machine, there is also an advantage that the amount of change from the standard time can be known, and it is easy to predict where the failure will occur from the change in the value.
In particular, by using the evaluation values (first feature value (evaluation value of deviation in amplitude direction) and second feature value (evaluation value of difference in waveform shape)) calculated by the signal analysis device for deterioration diagnosis in the deterioration diagnosis device. In addition to the diagnosis of deterioration based on the reference time of the actual machine, it is also possible to identify the cause of deterioration of the sensor or the like.

DESCRIPTION OF SYMBOLS 1 Signal analyzer 2 Target waveform acquisition part 3 Intersection extraction part 4 Candidate waveform extraction part 5 Comparison waveform generation part 6 Target waveform generation part 7 Correlation coefficient calculation part 8 Correlation coefficient storage part 9 Waveform determination part 10 Coordinate distance calculation part 11 Coordinate distance storage

Claims (6)

An intersection extraction unit for extracting an intersection with respect to the waveform of the signal;
A target waveform generation unit that generates a comparison target waveform from the waveform and its intersection;
A candidate waveform extraction unit for extracting a candidate waveform from the waveform and its intersection;
A comparison waveform having the same length as the comparison target waveform generated by repeating the candidate waveform, and an evaluation value calculation unit that calculates an evaluation value of the comparison target waveform;
A signal analysis apparatus comprising a waveform determination unit that determines a waveform for one period based on a calculated evaluation value.   The signal analysis apparatus according to claim 1, wherein the evaluation value calculation unit calculates an evaluation value using a correlation coefficient between the comparison waveform and the comparison target waveform.   The evaluation value calculation unit calculates an evaluation value using a coordinate distance using either or both of an evaluation value for a deviation in amplitude direction between the comparison waveform and the comparison target waveform or an evaluation value for a waveform shape. The signal analysis device according to claim 1, wherein   A deterioration diagnosis apparatus that performs deterioration diagnosis using the evaluation value calculated by the signal analysis apparatus according to any one of claims 1 to 3. An intersection extraction step for extracting an intersection with respect to the waveform of the signal;
A target waveform generation step for generating a comparison target waveform from the waveform and its intersection;
A candidate waveform extraction step of extracting a candidate waveform from the waveform and its intersection;
A comparison waveform having the same length as the comparison target waveform generated by repeating the candidate waveform, and an evaluation value calculating step of calculating an evaluation value of the comparison target waveform;
A signal analysis method comprising a waveform determining step for determining a waveform for one period based on a calculated evaluation value.   6. A deterioration diagnosis method, wherein deterioration diagnosis is performed using the evaluation value calculated by the signal analysis method according to claim 5.

Images