JP2014105075A - Failure part estimation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately estimate a part where an abnormality occurs, even in the case where the magnitude of operation noise varies along with the position of a car as the car travels.SOLUTION: A failure part estimation device includes: a sound collector 1, provided in a moving body, for collecting operation noise generated from the moving body and devices disposed around the moving range of the moving body; reference sample series analysis means (a waveform acquisition unit 2, a time frequency analysis unit 3 and a sampling unit 4) for analyzing collected normal operation noise and obtaining reference sample series 104; target sample series analysis means (a waveform acquisition unit 2, a time frequency analysis unit 3 and a sampling unit 4) for analyzing collected diagnosis target operation noise and obtaining target sample series 105; a variation curve generation unit 6 for obtaining variation series between the reference sample series 104 and the target sample series 105 and generating a variation curve 106; a shape characteristics extraction unit 7 for extracting shape characteristics 107 of the variation curve 106; and a collation determination unit 9 for collating the shape characteristics 107 with a template and determining a part where a failure is occurring in the moving body and the devices disposed around the moving range of the moving body.

Description

  In the present invention, when an abnormality is detected in the operation sound generated by the sound collector provided in the moving body and generated from the moving body and devices located around the moving range of the moving body, the abnormality is detected. The present invention relates to a failure location estimation apparatus that estimates a location where a failure occurs.

  An apparatus as shown in Patent Document 1 is known for estimating a failure location when an abnormality occurs in sound. In the apparatus for detecting an abnormality in an elevator hoistway disclosed in Patent Document 1, when the peak value of a sound captured by a sound sensor installed in an elevator car is continuously increased or decreased gradually, It is determined that an error has occurred in the device. Further, the location where an abnormality has occurred is determined based on the monotonicity of the peak value of the captured sound or the maximum value when the monotonicity changes.

JP-A-9-208149

  However, in the conventional apparatus disclosed in Patent Document 1, when the car travels, the magnitude of the operation sound changes with the position of the car, and thus the location where the abnormality has occurred cannot always be accurately estimated. There was a problem.

  The present invention has been made to solve the above-described problems, and accurately estimates the location where an abnormality has occurred even when the magnitude of the operating sound changes with the position of the car when the car travels. It is an object of the present invention to provide a failure location estimation apparatus capable of performing the above-described problem.

  The failure location estimation device according to the present invention is provided in a moving body, and collects the sound generated from the moving body and devices located around the moving range of the moving body, and the sound collecting means. A normal sample sound that is analyzed to obtain a first sample series as a reference, and a diagnostic object operation sound collected by the sound collecting means is analyzed to obtain a second diagnostic target. A variation that generates a variation curve from the variation sequence by obtaining a variation sequence between the first and second sample sequences obtained by the target sample sequence analysis means for obtaining the sample series, the reference sample series analysis means, and the target sample series analysis means Curve generation means, shape feature extraction means for extracting shape characteristics of the mutation curve generated by the mutation curve generation means, and matching of the shape characteristics of the mutation curve extracted by the shape feature extraction means with a predetermined template In Rukoto, in which a determining match determination means a failure location that occurs in a device located around the moving range of the moving body and the moving body.

  According to this invention, since it comprised as mentioned above, even when the magnitude | size of an operation sound changes with the position of a car when a car drive | works, the effect that the location where abnormality generate | occur | produced can be estimated correctly. Play.

It is a block block diagram which shows the failure location estimation apparatus in Embodiment 1 of this invention. It is a flowchart which shows the process of the failure location estimation apparatus in Embodiment 1 of this invention. It is a flowchart which shows the learning process of the failure location estimation apparatus in Embodiment 1 of this invention. It is a flowchart which shows the diagnostic process of the failure location estimation apparatus in Embodiment 1 of this invention. It is a schematic diagram which shows the variation | mutation curve and shape characteristic in Embodiment 1 of this invention. It is a schematic diagram which shows the structure of the shape characteristic template memory | storage part in Embodiment 1 of this invention. It is a block block diagram which shows the failure location estimation apparatus in Embodiment 2 of this invention. It is a schematic diagram which shows the mutation curve produced | generated by the mutation curve production | generation part in Embodiment 2 of this invention. It is a flowchart which shows the failure location determination process of the failure location estimation apparatus in Embodiment 3 of this invention. In Embodiment 3 of this invention, it is a schematic diagram which shows the variation curve and area ratio when a failure location is a pit, and the variation curve and area ratio when a failure location is a cage | basket. It is a block block diagram which shows the failure location estimation apparatus in Embodiment 4 of this invention. It is a schematic diagram which shows creation of the template in Embodiment 4 of this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
Embodiment 1 FIG.
The present embodiment is implemented as software on a personal computer (hereinafter referred to as a PC) as a device for diagnosing abnormal sounds generated by a device to be inspected, and in a learning mode and a diagnosis (test) for capturing a normal waveform It has a diagnostic mode to capture the waveforms. The measurer installs a sound collector (sound collecting means) 1 such as a microphone, an acoustic sensor, an acceleration sensor, etc. in the inspection target device, and connects the sound collector 1 to an input terminal of a USB (Universal Serial Bus) interface of the PC. The operation in the learning mode and the diagnosis mode is performed. In the following, an elevator is assumed as the inspection target device, and the operation sound generated from the elevator car (moving body) and the device installed in the hoistway (device located around the moving range of the moving body) is collected by the sound collector 1. The case of collecting in will be explained.

FIG. 1 is a block configuration diagram showing a failure location estimation apparatus according to Embodiment 1 of the present invention.
In FIG. 1, reference numeral 1 denotes a sound collector such as a microphone, an acoustic sensor, or an acceleration sensor. Reference numeral 2 denotes a waveform acquisition unit that samples the measurement signal from the sound collector 1 and outputs the waveform data 101 converted into a digital signal. 3 is a spectral value indicating the time and frequency intensity of the waveform data 101 by applying a time window to the waveform data 101 and performing a fast frequency transform (hereinafter referred to as FFT) operation while shifting the time window in the time direction. It is a time frequency analysis part which outputs the time frequency distribution 102 which consists of. Reference numeral 4 denotes a sampling unit that outputs a sample sequence 103 of each frequency, which is a time series obtained by sampling a spectrum value indicating the intensity of each predetermined frequency from the time frequency distribution 102 at each time. The sample series 103 is a time series composed of sample values at each time for predetermined frequencies.
In the following description, a case where the time frequency distribution 102 output from the time frequency analysis unit 3 is set as sound data will be described. The sound data may be waveform data or other feature quantities obtained by analyzing the waveform data.

Reference numeral 5 denotes a data storage unit that stores a sample series 103 obtained from sound data acquired in the learning mode. Reference numeral 104 denotes a reference sample series (first sample series) that is stored in the storage unit 8 in the learning mode and includes the sample series 103 of each frequency that serves as a reference when calculating a later-described variation curve 106. Reference numeral 105 denotes a target sample series (second sample series) that is obtained from sound data acquired in the diagnosis mode and includes a sample series 103 of each frequency to be diagnosed when calculating a later-described mutation curve 106.
The waveform acquisition unit 2, the time frequency analysis unit 3, and the sampling unit 4 analyze the normal operating sound collected by the sound collection means of the present invention, and obtain the first sample series as a reference. This corresponds to “sample series analyzing means” and “target sample series analyzing means for analyzing the operation sound of the diagnosis target collected by the sound collecting means and obtaining the second sample series to be diagnosed”.

  Reference numeral 6 denotes a mutation curve generation unit (mutation curve generation means) that generates a mutation curve 106 with reference to the reference sample series 104 and the target sample series 105. Reference numeral 7 denotes a shape feature extraction unit (shape feature extraction means) that extracts the shape feature 107 from the mutation curve 106.

  Reference numeral 8 denotes a shape feature template storage unit that stores templates composed of predetermined shape features. Reference numeral 9 denotes a collation determination unit (collation determination unit) that collates the shape feature 107 with the template in the shape feature template storage unit 8, determines a failure location, and outputs the determination result 108.

Hereinafter, the operation of the failure location estimation apparatus will be described with reference to the flowcharts shown in FIGS. 1 and 2 to 4.
As shown in FIG. 2, first, in the learning mode or the diagnostic mode, the waveform acquisition unit 2 acquires the measurement signal output from the sound collector 1, amplifies it, performs sampling by AD conversion, and has a sampling frequency of 48 kHz. 16-bit linear PCM (pulse code modulation) digital signal waveform data 101 is converted (step ST201).

  Next, the time frequency analysis unit 3 cuts out the frame from the waveform data 101 output from the waveform acquisition unit 2 while shifting the time window of 1024 points in the time direction at intervals of 16 milliseconds. A frequency spectrum series y (t, f) is obtained by FFT operation and output as a time-frequency distribution 102 (step ST202). Here, t is a time index corresponding to the shift interval for shifting the time window, and f is an index indicating the frequency of the result of the FFT operation. The time t and the frequency f satisfy the relations 0 ≦ t ≦ T and 0 ≦ f ≦ F, respectively. Here, T is the number of frames in the time direction of the time frequency distribution 102, and F is a Nyquist frequency that is ½ of the sampling frequency fs of the waveform data 101 (F = fs / 2).

ここで、ｎは標本化された系列の時間のインデックスで、１〜Ｎの範囲の自然数（ただしＮは時間範囲の上限、Ｎ＝Ｔ／８で余りは切り捨て）、ｂは周波数のインデックスで１〜Ｂの範囲の自然数である。また、Ω（ｎ，ｂ）は、時間周波数分布ｙ（ｔ，ｆ）において、標本化のために総和をとる対象となる時間と周波数の組（ｔ，ｆ）の集合を表す。 Next, the sampling unit 4 uses the predetermined frequency bands (0.5 kHz, 1 kHz, 2 kHz, 4 kHz, and 8 kHz) as center frequencies in the time frequency distribution 102 output from the time frequency analysis unit 3, respectively. The sum of the spectrum values with 8 frames (256 ms) as one unit is obtained from the frequency components included in the width band, and sample values at predetermined times with the 8 frames as units are obtained as sample sequence 103. Output (step ST203). Now, assuming that the sample value in the sample sequence 103 at each frequency and each time is Y (n, b), Y (n, b) is calculated by Expression (1).

Here, n is a time index of a sampled sequence, a natural number ranging from 1 to N (where N is the upper limit of the time range, N = T / 8 and the remainder is rounded down), and b is a frequency index. It is a natural number in the range of ~ B. Ω (n, b) represents a set of time and frequency sets (t, f) to be summed for sampling in the time-frequency distribution y (t, f).

  When the sample series 103 is acquired by the sampling unit 4, the failure location estimating apparatus determines whether the failure mode is the learning mode or the diagnosis mode (step ST204). If it is determined in step ST204 that the mode is the learning mode, the process proceeds to the learning process shown in FIG. 3, and if it is determined that the mode is the diagnostic mode, the process proceeds to the diagnostic process shown in FIG.

Next, the operation of the learning process in the learning mode will be described.
When the failure location estimation device is in the learning mode, that is, when normal operating sounds are collected by the sound collector 1, as shown in FIG. 3, for each frequency (step ST301), the storage unit 8 The sample series 103 is stored as the reference sample series 104 (step ST302).
And after implementing the process of step ST302 with respect to all the frequencies (step ST303 'YES'), a learning process is complete | finished.

Next, the operation of the diagnosis process in the diagnosis mode will be described.
When the failure location estimation apparatus is in the diagnosis mode, that is, when the operation sound to be diagnosed is collected by the sound collector 1, as shown in FIG. Is the target sample series 105 (step ST402).

なお、式（３）中の“ｓｍｏｏｔｈ（ｘ（ｎ））”は系列ｘ（ｎ）のｎに関する移動平均を求める演算を表す。 Next, the variation curve generation unit 6 generates the variation curve 106 for the specific frequency from the reference sample sequence 104 stored in the storage unit 5 and the target sample sequence 105 in step ST402 (step ST403). Specifically, the amount of variation at each time n of a specific frequency b is calculated by the following procedures << A1-1 >> and << A1-2 >>.
<< A1-1 >> First, a sample sequence D (n, b) of a difference between the target sample sequence Y ₁ (n, b) at the frequency b and the reference sample sequence Y ₀ (n, b) at the frequency b is obtained. (Formula (2))
<< A1-2 >> Next, a moving average for the subscript n of the difference sample series D (n, b) is obtained and output as the variation H ₁ (n, b). (Formula (3))

Note that “smooth (x (n))” in equation (3) represents an operation for obtaining a moving average for n in the sequence x (n).

Next, the shape feature extraction unit 7 extracts the peak position and peak width of the variation curve 106 from the variation curve 106 output from the variation curve generation unit 6 as the shape feature 107 (step ST404).
FIG. 5 is a schematic diagram showing the variation curve 106 and the shape feature 107. 5A shows a reference sample sequence 104 and a target sample sequence 105 having a specific frequency. Moreover, (b) is a figure which shows the variation curve 106 calculated | required by Formula (2), (3), and the peak position and the peak width which are the shape characteristics 107. FIG. The peak position is the value on the x-axis of the peak of the mutation curve 106. The peak width is a peak spread in the x-axis direction (for example, a width that is 3 dB lower than the peak value is defined as a spread).

ここで、Ｓ（ｃ）はテンプレートのクラスｃに対する類似度である。
最後に、照合判定部９は、上記の類似度Ｓ（ｃ）が最大となるクラスｃ＊を求め、クラスｃ＊の「故障個所」属性値を判定結果１０８とする。詳細には式（５）の決定則によりクラスを決定する。

Next, the collation determination unit 9 collates the shape feature 107 output from the shape feature extraction unit 7 with the template stored in the shape feature template storage unit 8 to obtain the similarity for each failure location. The failure individuality value of the large template is determined as a failure location (step ST405).
FIG. 6 is a schematic diagram showing the configuration of the shape feature template storage unit 8. The template is stored as a table including, for example, feature amounts f ₁ “peak position” and f ₂ “peak spread” with respect to a class corresponding to a failure location (elevator pit, top, counter, car).
Now, the “peak position” feature quantity of the shape feature 107 is f ₁ , the “peak spread” feature quantity is f ₂ , the “peak position” feature quantity of the class c in the template is f ₁ (c), and “peak spread”. Assuming that the feature quantity is f ₂ (c), the similarity of the shape feature 107 to the class c of the template is calculated using, for example, an expression obtained by inverting the sign of the Euclidean distance calculation expression of Expression (4).

Here, S (c) is the similarity of the template to class c.
Finally, the collation determination unit 9 obtains the class c * having the maximum similarity S (c), and sets the “failure location” attribute value of the class c * as the determination result 108. Specifically, the class is determined according to the determination rule of Expression (5).

  As described above, according to the first embodiment, the normal operation sound collected by the sound collector 1 is analyzed to obtain the reference sample series 104, and the operation of the diagnosis target collected by the sound collector 1 is obtained. Since the target sample series 105 is obtained by analyzing the sound and the failure location is determined based on the reference sample series 104 and the variation curve 106 extracted from the target sample series 105, when the car runs, Even when the magnitude of the operating sound changes with the position, there is an effect that the location where the abnormality has occurred can be accurately estimated.

ここで、Ｓ（ｃ，ｂ）はある周波数ｂの変異曲線１０６から求めるクラスｃに対する類似度、ｃ＊は類似度の最も高い周波数の判定結果１０８である。 In the above description, the case where the variation curve 106 is obtained for a specific frequency and the failure location is determined is shown. However, the present invention is not limited to this. For each variation curve 106 obtained for each frequency, the failure location is determined, and the determination result of the frequency having the highest similarity is output as the final determination result 108. Also good. Specifically, the class is determined according to the determination rules of equations (6) and (7).

Here, S (c, b) is the similarity to the class c obtained from the variation curve 106 at a certain frequency b, and c * is the determination result 108 of the frequency with the highest similarity.

ここで、Ｈ_２（ｎ）は平均変異曲線、“ｍｅａｎ＿ｂ”は周波数ｂに関する平均演算を示す。 Further, as shown in the equation (8), after obtaining the variation curve 106 obtained by averaging the variation curves 106 of the respective frequencies, the failure location may be determined using the averaged variation curve 106.

Here, H ₂ (n) represents an average variation curve, and “mean_b” represents an average calculation regarding the frequency b.

  Further, instead of the average, the variation curve 106 of each frequency may be compared, and the variation curve 106 having a large variation may be selected and used for the determination of the failure location.

Embodiment 2. FIG.
In the second embodiment, a portion having a variation directly is extracted from the target sample series 105 and a variation curve 106 is obtained.
Only differences from the first embodiment will be described. The difference is the configuration of the mutation curve generation unit 6. The operation will be described below with reference to FIG. 7 and FIGS.

In FIG. 7, reference numeral 61 denotes a difference sequence generation unit that obtains a difference sequence between the reference sample sequence 104 and the target sample sequence 105. Reference numeral 62 denotes a load series generation unit that generates a load series according to the difference series. 63 is an extraction sequence generation unit that generates an extraction sequence (mutation sequence) from the target sample sequence 105 using the load sequence. Reference numeral 64 denotes a shaping unit that shapes the extracted sequence to generate the mutation curve 106. These constitute the mutation curve generation unit 6.
Below, the specific operation | movement of each structure part of the variation curve production | generation part 6 is demonstrated.

ここで、Ｄ（ｎ，ｂ）は差系列である。 First, the difference series generation unit 61 generates a difference series between the target sample series 105 and the reference sample series 104 using Expression (9).

Here, D (n, b) is a difference series.

ここで、ｗ（ｎ，ｂ）は荷重系列、ｓｇｍ（ｘ;α，β）はシグモイド関数を表す。また、αとβは、シグモイド関数の形状を特徴づけるパラメータであり、例えば、α＝１、β＝３と置くことができる。 The load series generation unit 62 generates a load series obtained by converting the difference series D (n, b) into a value between 0 and 1 using the nonlinear function shown in Expression (10).

Here, w (n, b) represents a load series, and sgm (x; α, β) represents a sigmoid function. Α and β are parameters that characterize the shape of the sigmoid function. For example, α = 1 and β = 3 can be set.

ここで、Ｅ（ｎ，ｂ）は抽出系列である。 The extraction sequence generation unit 63 generates an extraction sequence according to the load by multiplying the target sample sequence 105 by the load w (n, b) using Expression (11).

Here, E (n, b) is an extraction sequence.

ここで、Ｈ_３（ｎ，ｂ）は変異曲線１０６、ｓｍｏｏｔｈ＿ｎ（ｘ（ｎ））は系列ｘ（ｎ）のｎに関する平滑化演算を示し、例えば、移動平均演算により実現される。 Using the equation (12), the shaping unit 64 shapes the extracted series E (n, b) by smoothing with respect to n and generates the mutation curve 106.

Here, H ₃ (n, b) represents a variation curve 106, and smooth_n (x (n)) represents a smoothing operation relating to n of the sequence x (n), and is realized by, for example, a moving average operation.

  FIG. 8 is a schematic diagram comparing the variation curve 106 generated by the variation curve generation unit 6 of the second embodiment shown in (c) with the variation curve 106 of the first embodiment shown in (b). In this schematic diagram, as shown in (a), there is a case where there is a depression in the center of the reference sample series 104 and the abnormal sound component (dotted line) is large in the first half of the time axis. In this case, as shown in (b), it can be seen that the peak of the variation curve 106 according to the first embodiment moves to the center of the time axis and is greatly different from the shape of the abnormal sound component. On the other hand, as shown in (c), the variation curve 106 according to the second embodiment has a shape that is very similar to the abnormal sound component, and if the failure location is determined based on the variation curve 106, a more accurate failure is obtained. It is expected that a place will be obtained.

  As described above, according to the second embodiment, in order to prevent the shape of the variation curve 106 from being deformed by the shape of the reference sample sequence 104, the variation curve 106 is obtained directly from the target sample sequence 105. Therefore, even when the car travels, the magnitude of the operating sound changes with the position of the car, and even when the shape of the reference sample series 104 has a dent in the center of the time axis, for example, There is an effect that it can be estimated accurately.

Embodiment 3 FIG.
In the third embodiment, by using the peak position of the variation curve 106 and the concentration of mutation at the peak position as the shape feature 107 of the variation curve 106, a failure location (“top” from the peak position in the moving direction of the moving object is used. ”,“ Pit ”,“ counter ”), and whether or not the failure location is“ cage ”or“ cage ”is determined from the concentration of mutation at the peak position.
Only differences from the first embodiment will be described. The difference is the operation of the shape feature extraction unit 7 and the collation determination unit 9. FIG. 9 is a flowchart showing the failure location determination process.

  As shown in FIG. 9, the shape feature extraction unit 7 first extracts the peak position of the variation curve 106 from the variation curve 106 as the shape feature 107 (step ST901). Next, the collation determination unit 9 performs provisional determination based on the peak position (step ST902). In the tentative determination, it is determined by using the same method as in the first embodiment whether “pit”, “counter”, or “top” except “car” is similar, and output as tentative candidate x (step ST903). ). Here, x is any one of “pit”, “counter”, and “top”.

ここで、Ｒ＾ｘは面積比、Ｓ＾ｘは仮候補ｘに対して、テンプレートのピーク位置を中心とする所定幅に入る変異曲線１０６の面積、Ｓ＾ａｌｌは全体の面積である。
なお、面積比は、ピーク位置における変異の集中度を表す。 Next, the collation determination unit 9 uses the equation (13) to calculate the ratio (area ratio) of the area of the mutation curve 106 that falls within a predetermined width centered on the peak position of the template to the temporary candidate x and the entire area. Is obtained (step ST904).

Here, R ^ x is the area ratio, S ^ x is the area of the variation curve 106 that falls within a predetermined width around the peak position of the template with respect to the temporary candidate x, and S ^ all is the entire area.
The area ratio represents the degree of mutation concentration at the peak position.

  Next, the collation determination unit 9 compares the area ratio with a predetermined threshold to determine whether the area ratio is smaller than the predetermined threshold (step ST905). In step ST905, when the area ratio is smaller than the predetermined threshold value, it is determined that the failure location is “cage”.

  On the other hand, when the area ratio is greater than or equal to a predetermined threshold value in step ST905, it is determined whether or not the deviation between the peak position and the peak position of the template exceeds a predetermined deviation allowable value (step ST906). In step ST906, when the peak position deviation exceeds the deviation allowable value, it is determined that the failure location is “cage”.

  On the other hand, in step ST906, when the peak position deviation falls within the deviation allowable value, it is determined that the temporary candidate x is a failure location.

  FIG. 10 schematically shows (a) the variation curve 106 and the area ratio when the failure location is “pit”, and (b) the variation curve 106 and the area ratio when the failure location is “cage”. In the figure, S ^ pit is an area corresponding to the pit. In the case of (a), since the area ratio is large, pits are set as temporary candidates. Since the peak position deviation is small, it is determined that the pit is a failure location. On the other hand, in the case of (b), the peak position deviation is small, but the area ratio is small. Therefore, it is determined that there is a low possibility that the failure location is a pit, and it is determined that the car is a failure location.

  As described above, according to the third embodiment, using the peak position of the variation curve 106 and the concentration of mutation at the peak position, the failure location (“top”, “pit” from the peak position in the moving direction of the car). ”,“ Counter ”), and determination of whether the failure location is“ cage ”or other than“ cage ”from the degree of concentration of the mutation at the peak position, thereby reducing the erroneous determination of the failure location. There is an effect.

Embodiment 4 FIG.
In the fourth embodiment, a template stored in the shape feature template storage unit 8 is created from simulated abnormal sound data.
Only differences from the first embodiment will be described. The difference is that a simulated abnormal sound generating unit (simulated abnormal sound creating unit) 10 and a template creating unit (template creating unit) 11 are added to the failure location estimating apparatus. Hereinafter, the operation will be described with reference to FIGS.

  In FIG. 11, the simulated abnormal sound generation unit 10 generates simulated abnormal sound data by simulating the traveling operation of the moving body in which the sound collector 1 is installed. The template creation unit 11 uses the simulated abnormal sound data generated by the simulated abnormal sound generation unit 10 to obtain a variation curve with the reference sample series 104 stored in the storage unit 5 and extracts a shape feature from the variation curve. To create a template. The created template is stored in the shape feature template storage unit 8.

Conventionally, when creating a template of a predetermined shape feature in estimating the failure location of an elevator, it has been created from the approaching position between the failed component and the sound collector based on the component arrangement in the elevator hoistway. However, in many cases where the sound source intensity of the failed part changes with time according to the car traveling speed, the peak position of the actual variation curve and the approach position between the failed part and the sound collector do not necessarily match. Deviation occurs. This tendency is particularly remarkable at the pit and the top where the traveling speed of the car accelerates or decelerates.
Therefore, in this embodiment, a template is created using simulated abnormal sound data. As a result, the template to be created includes a positional shift due to a temporal change in the sound source intensity of the failed part, and therefore the peak position is expected to be close to the peak position of the actual mutation curve. FIG. 12 shows the difference between the conventional method and the template according to the present embodiment and its peak position.

  As described above, according to the fourth embodiment, since the template is created using the simulated abnormal sound data, the peak position of the template is brought closer to the peak position of the actual mutation curve compared to the conventional method. There is an effect that can be.

  In the present invention, within the scope of the invention, any combination of the embodiments, or any modification of any component in each embodiment, or omission of any component in each embodiment is possible. .

  1 sound collector, 2 waveform acquisition unit, 3 time frequency analysis unit, 4 sampling unit, 5 data storage unit, 6 mutation curve generation unit, 7 shape feature extraction unit, 8 shape feature template storage unit, 9 collation determination unit, DESCRIPTION OF SYMBOLS 10 Simulated abnormal sound production | generation part, 11 Template creation part, 61 Difference series production | generation part, 62 Load series production | generation part, 63 Extraction series production | generation part, 64 Shaping part, 101 Waveform data, 102 Time frequency distribution, 103 Sample series, 104 Reference sample Series, 105 target sample series, 106 mutation curve, 107 shape feature, 108 judgment result.

Claims (4)

A sound collecting means for collecting operating sound generated from the mobile body and a device located around the moving range of the mobile body, the mobile body;
A normal sample sound collected by the sound collecting means, and a reference sample series analyzing means for obtaining a first sample series as a reference;
Analyzing the operation sound of the diagnosis target collected by the sound collection means, and target sample series analysis means for obtaining a second sample series to be diagnosed;
A mutation curve generating means for obtaining a mutation sequence between the first and second sample series obtained by the reference sample series analysis means and the target sample series analysis means, and generating a mutation curve from the mutation series;
Shape feature extraction means for extracting shape characteristics of the mutation curve generated by the mutation curve generation means;
Collation determining means for determining a fault location occurring in the mobile body and a device located in the vicinity of the moving range of the mobile body by collating the shape feature of the variation curve extracted by the shape feature extracting means with a predetermined template And a failure location estimation device. The mutation curve generating means is:
Difference series generating means for generating a difference series between the first and second sample series;
Load sequence generation means for generating a load coefficient corresponding to the difference series generated by the difference series generation means;
A mutation sequence generation means for generating the mutation sequence by multiplying the second sample series by the load sequence generated by the load sequence generation means;
The failure location estimation apparatus according to claim 1, further comprising: shaping means for smoothing the mutation series generated by the mutation series generation means to generate the mutation curve. The shape characteristic of the mutation curve consists of the peak position of the mutation curve and the concentration of mutation at the peak position,
The collation determining means determines a failure location in the moving direction of the mobile body from the peak position, and determines whether or not the mobile body is faulty from a degree of mutation concentration at the peak position. The failure location estimation apparatus according to claim 1 or 2. Simulated abnormal sound generation means for generating simulated abnormal sound data by simulating the traveling operation of the mobile body,
A template that uses the simulated abnormal sound data generated by the simulated abnormal sound generation means to obtain a variation curve with the first sample sequence obtained by the reference sample sequence analysis means, and creates the template from the variation curve The fault location estimation apparatus according to any one of claims 1 to 3, further comprising: creation means.

Images