JPWO2015068446A1 - Abnormal sound diagnosis device - Google Patents

Abstract

The time frequency analysis unit 3 performs spectrum analysis on the operating sound acquired by the sound collector 1 to acquire a time frequency distribution. The specific component detection unit 4 detects a predetermined specific component from the time-frequency distribution. The specific component counting unit 5 counts the number of detection times of the specific component detected by the specific component detection unit 4 in the time frequency distribution. The increment calculation unit 7 compares the specific component count value at the normal time and the specific component count value at the time of diagnosis to calculate whether there is an increase over the set value. If there is an increase, the bearing abnormality determination unit 8 Is determined as abnormal.

Description

  The present invention relates to an abnormal sound diagnosis apparatus for diagnosing abnormal sound generated from a diagnosis target such as a machine having a rotating body such as an elevator or a vehicle.

2. Description of the Related Art Conventionally, for diagnosis of abnormal noise generated from a rotating body, particularly abnormal noise generated from a rolling bearing, devices as shown in Patent Document 1 and Patent Document 2, for example, are known.
In the abnormality inspection apparatus of Patent Document 1, a plurality of feature amounts (standard deviation of sound pressure per unit time, number of peaks, fluctuation range, etc.) calculated from sound pressure or vibration are learned by a neural network, and a rotating machine Is disclosed.
Further, in the bearing diagnostic device of Patent Document 2, the vibration of the bearing is converted into a time series signal for each frequency band, the maximum value and the average effective value of the converted time series signal are extracted, and set by a normal bearing. It is disclosed to compare with the determined value. Further, it is disclosed that a damaged position of a bearing is specified based on a pulse-like peak period.

JP-A-11-241945 JP 2002-022617 A

  However, since the conventional abnormal sound diagnosis device is intended for abnormal sound generated from a rotating body, particularly a rolling bearing, it is applied to a machine including a rotating body and other parts and having the rotating body. In this case, there is a problem that the accuracy is lowered due to a sound similar to an abnormal sound generated from another part and generated from the rotating body. That is, in the diagnosis of abnormal sound generated from the rotating body, there has been a problem that components generated from other parts similar to the abnormal sound generated from the rotating body are erroneously detected as abnormal sounds generated from the rotating body. Also, in the diagnosis of abnormal sounds generated from other parts, the detection accuracy of abnormal sounds generated from other parts is reduced by superimposing abnormal sounds generated from the rotating body on abnormal sounds generated from other parts. There was a problem.

  The present invention has been made to solve the above-described problems, and provides an abnormal sound diagnostic apparatus that can prevent abnormal sounds generated due to other factors from being erroneously detected as abnormal sounds to be diagnosed. The purpose is to obtain.

  The abnormal sound diagnosis apparatus according to the present invention compares the operation sound at the time of diagnosis with the operation sound at the time of diagnosis, and diagnoses the abnormality of the operation sound. The time frequency analysis unit for acquiring the frequency distribution, the specific component detection unit for detecting a predetermined specific component from the time frequency distribution, and the number of detections of the specific component detected by the specific component detection unit in the time frequency distribution. A specific component counting unit for counting, a specific component count value at the time of normality, and a determination unit for comparing with the specific component count value at the time of diagnosis and determining an abnormality when there is an increase beyond a set value is there.

  The abnormal sound diagnosis apparatus according to the present invention counts the number of times a specific component is detected in a time-frequency distribution, and determines that an abnormality occurs when the specific component count value at the time of diagnosis is greater than a set value by a specific component count value at normal time. Therefore, it is possible to prevent the abnormal sound generated due to other factors from being erroneously detected as the abnormal sound to be diagnosed.

It is a block diagram which shows the abnormal sound diagnostic apparatus by Embodiment 1 of this invention. It is a flowchart which shows operation | movement of the abnormal sound diagnostic apparatus by Embodiment 1 of this invention. It is explanatory drawing which shows the bearing abnormal sound data of the abnormal sound diagnostic apparatus by Embodiment 1 of this invention in time series. It is explanatory drawing which shows the data preserve | saved at the count log | history storage part of the abnormal sound diagnostic apparatus by Embodiment 1 of this invention. It is a block diagram which shows the abnormal sound diagnostic apparatus by Embodiment 2 of this invention. It is explanatory drawing which shows the bearing abnormal sound data of the abnormal sound diagnostic apparatus by Embodiment 2 of this invention in time series. It is a block diagram which shows the abnormal sound diagnostic apparatus by Embodiment 3 of this invention. It is a flowchart which shows operation | movement of the specific component removal part and abnormality detection part of the abnormal sound diagnostic apparatus by Embodiment 3 of this invention. It is explanatory drawing which shows the relationship between the abnormal mode and determination output of the abnormal sound diagnostic apparatus by Embodiment 3 of this invention.

  The present invention is implemented as software on a personal computer (hereinafter referred to as a PC) as a device for diagnosing abnormal sounds generated by a machine to be diagnosed. Has a diagnostic mode to capture sound. Sound collectors such as microphones, acoustic sensors, and acceleration sensors are installed so as to be able to collect the operating sound of the machine to be inspected, and signals from the sound collectors are connected via a USB (Universal Serial Bus) interface. Is taken into the PC.

  Moreover, this invention is an elevator installed in the hoistway as a machine made into a diagnostic object, for example. In this elevator, a counter is suspended from one end of a rope extending from a pulley provided at the upper part of the hoistway toward the lower part of the hoistway, and a car is suspended from the other end. In the lower part of the car (hereinafter referred to as “car lower”), a pulley (a hanging car) having a rolling bearing is disposed in order to reverse the direction of the rope for hanging the car. The sound collectors are installed under the car and above the car (hereinafter referred to as “above the car”). During learning and diagnosis, the car is moved up and down in the elevator to collect the operating sound generated from each device in the hoistway and the car equipment, and the collected operating sound is taken into the PC. The PC software diagnoses the presence or absence of abnormal sounds generated from each device. Hereinafter, an embodiment in which such an elevator is a diagnosis target will be described.

Hereinafter, in order to explain the present invention in more detail, modes for carrying out the present invention will be described with reference to the accompanying drawings.
Embodiment 1 FIG.
FIG. 1 is a block diagram showing an abnormal sound diagnosis apparatus according to Embodiment 1 of the present invention.
The abnormal sound diagnosis apparatus shown in FIG. 1 includes a sound collector 1, a waveform acquisition unit 2, a time frequency analysis unit 3, a specific component detection unit 4, a specific component counting unit 5, a counting history storage unit 6, an increment calculation unit 7, a bearing. An abnormality determination unit 8 is provided. The sound collector 1 is a sound collector installed under the elevator car. The waveform acquisition unit 2 is a processing unit that samples a signal from the sound collector 1, converts it into a digital signal, and outputs waveform data 21. The time frequency analysis unit 3 applies a time window to the waveform data 21 and performs time frequency analysis on the waveform data 21 by fast Fourier transform (hereinafter referred to as FFT) while shifting the time window in the time direction, and the time and frequency intensity. Is a processing unit that obtains a time-frequency distribution composed of spectrum values indicating, and outputs a time-frequency distribution 311 at learning 31 and a time-frequency distribution 321 at diagnosis 32. The specific component detection unit 4 is a processing unit that refers to the time frequency distribution 311 and detects an impact component as a specific component included in the time frequency distribution 321. The specific component counting unit 5 is a processing unit that counts the number of detections of the impact component detected by the specific component detection unit 4, and the count history storage unit 6 counts the number of detections counted by the specific component counting unit 5 together with the diagnosis time. It is a memory | storage part to preserve | save. The increment calculation unit 7 is a processing unit that calculates the increment of the count value at the time of diagnosis with respect to the count value at the time of learning stored in the count history storage unit 6. The bearing abnormality determination unit 8 is a determination unit that determines a bearing abnormality based on the increment calculated by the increment calculation unit 7 and outputs a bearing abnormality determination result 81. In FIG. 1, a solid line block indicates a processing unit or a storage unit, a broken line block indicates data output from the processing unit, and an alternate long and short dash line block indicates a type of processing.

  Moreover, said waveform acquisition part 2-specific component counting part 5, the increment calculation part 7, and the bearing abnormality determination part 8 are implement | achieved by respectively executing the corresponding software using hardware, such as CPU and memory. . Alternatively, at least one of the configurations may be configured with dedicated hardware.

Next, the operation of the abnormal sound diagnosis apparatus of the first embodiment will be described.
FIG. 2 is a flowchart showing the operation of the abnormal sound diagnosis apparatus of the first embodiment.
At the time of learning or diagnosis, the waveform acquisition unit 2 acquires a measurement signal output from the sound collector 1, performs amplification by performing AD conversion, performs sampling, and performs 16-bit linear PCM (pulse code modulation) with a sampling frequency of 24 kHz. ) Is converted into waveform data 21 of the digital signal (step ST1).
The time frequency analysis unit 3 cuts out a frame from the waveform data 21 output by the waveform acquisition unit 2 while shifting the time window of 256 points in the time direction at intervals of 5 milliseconds, and performs FFT calculation on each frame. A frequency spectrum series y (t, f) is obtained and output as a time frequency distribution 311 at the time of learning 31 or a time frequency distribution 321 at the time of diagnosis 32 (steps ST2 to ST5). That is, in step ST3, when the learning mode (normal time) is set, the time frequency distribution 311 at the time of learning 31 is output, and when not in the learning mode (in the diagnosis mode), the time frequency distribution 321 at the time of diagnosis 32 is output. Output.

  Here, t is a time index corresponding to the shift interval for shifting the time window, and f is a frequency index indicating the frequency of the result of the FFT operation (hereinafter, “time index t” is “ (Time t) and “frequency index” as “frequency f”). Further, y (t, f) is a decibel value indicating the magnitude of power at time t and frequency f. The time t and the frequency f satisfy the relations 1 ≦ t ≦ T and 0 ≦ f ≦ F, respectively. Here, T is the number of frames in the time direction of the time frequency distributions 311 and 321, and F is a Nyquist frequency (F = fs / 2) that is ½ of the sampling frequency fs of the waveform data 21.

ここで、ｆＬ及びｆＵは、それぞれ、所定の周波数範囲を示す下限の周波数と上限の周波数である。例えば、ｆＬ＝１０００Ｈｚ，ｆＵ＝９０００Ｈｚと設定する。The specific component detection unit 4 detects the specific component from the time frequency distribution 321 at the time of diagnosis 32 as follows (step ST6).
First, from the time frequency distribution 321 at the time of diagnosis 32, the sum of power values included in a predetermined frequency band at each time is obtained to obtain a time series of band power (decibel values). Specifically, the band power time series B (t) with time t is calculated as shown in Equation (1).

Here, fL and fU are a lower limit frequency and an upper limit frequency indicating a predetermined frequency range, respectively. For example, fL = 1000 Hz and fU = 9000 Hz are set.

ここで、「Ｐｅａｋ（ｔ）」は１のときピークがあることを示し、０のときピークがないことを示す。なお、窓幅Ｗは例えば、Ｗ＝４１フレーム（時間幅換算で２０５ミリ秒）とする。Next, the band power time series B (t) is searched in the time direction to detect its peak. The peak is provided with a time window having a predetermined window width (set width) W centered at time t, and the value of the band power at time t matches the maximum value of the band power time series in this time window. , Detecting that there is a peak at time t. That is, the peak is detected as shown in Equation (2).

Here, “Peak (t)” indicates that there is a peak when it is 1, and indicates that there is no peak when it is 0. The window width W is, for example, W = 41 frames (205 milliseconds in terms of time width).

ここで、Ｃｏｕｎｔは、検出された特定成分の計数値（ピークの数）である。The specific component counting unit 5 counts the number of specific components detected from the detection result of the specific component obtained by the specific component detection unit 4 (step ST7). Specifically, the number of detected specific components is calculated as shown in Equation (3), where the number of peaks obtained by the specific component detection unit 4 is the sum of Peak (t) time ranges 1 ≦ t ≦ T. To do.

Here, Count is a count value (number of peaks) of the detected specific component.

  FIG. 3 is a diagram showing bearing abnormal sound data in time series, where (a) shows the time frequency distribution (spectrogram), (b) shows the band power time series (b), and (C) in the figure shows the detection result of the specific component by a circle.

The count history storage unit 6 stores the count value acquired by the specific component counting unit 5 together with the diagnosis time (date and time when the diagnosis is performed) (step ST8).
FIG. 4 shows the configuration of data stored in the count history storage unit 6 and an example of stored data. In the figure, the column “Operation” records the operation mode when the count value is obtained, the column “Date” records the date of learning or diagnosis, and the column “Time” The time when the sound is started is recorded, and the column indicated by “Count Value” records the count value (Count in Expression (3)).

ここで、「Ｃｏｕｎｔ（学習時）」は学習時の計数値、「Ｃｏｕｎｔ（診断時）」は診断時の計数値である。If the learning mode is determined in step ST9, the process ends. On the other hand, in the diagnosis mode, the increment calculation unit 7 calculates the increment of the count value based on the count value recorded in the count history storage unit 6 (step ST10). That is, the increment of the count value is calculated as shown in Equation (4).

Here, “Count (at the time of learning)” is a count value at the time of learning, and “Count (at the time of diagnosis)” is a count value at the time of diagnosis.

ここで、「ＪｕｄｇＢｅａｒｉｎｇ」は軸受異常の判定結果で、１は異常があることを示し、０は異常がないことを示す。また、「ＴｈｒｅｓｈｏｌｄＢｅａｒｉｎｇ」は軸受異常判定の閾値である。ＴｈｒｅｓｈｏｌｄＢｅａｒｉｎｇは、例えば、特定成分が毎秒２回ずつ検出される場合異常であるとして、４２と設定する。When the increment of the count value exceeds a predetermined threshold based on the increment of the count value calculated by the increment calculation unit 7 (when there is an increase beyond a preset value), the bearing abnormality determination unit 8 Is determined and output as a bearing abnormality determination result 81 (step ST11). That is, a bearing abnormality is determined as shown in Equation (5).

Here, “JudgBearing” is a bearing abnormality determination result, where 1 indicates that there is an abnormality and 0 indicates that there is no abnormality. Further, “Threshold Bearing” is a bearing abnormality determination threshold value. Threshold Bearing is set to 42, for example, because it is abnormal when a specific component is detected twice per second.

  In the example of FIG. 4, since the count value exceeds 87 and the threshold value 42 at the time of diagnosis started on October 29, 2012, 3:02:35, it is determined that there is a bearing abnormality.

  As described above, the abnormal sound diagnosis apparatus according to the present embodiment has an abnormality generated from the rotating body based on the increment of the count value of the specific component detected at the time of diagnosis with respect to the count value of the specific component detected at the time of learning. Since the sound is detected, the effect of the specific component existing from the learning time is eliminated, and the detection of the specific component is facilitated.

  In the above example, the count value over the entire time interval of the time frequency distribution of the specific component is used as the count value of the detection number of the specific component, but as the count value of the detection number per unit time of the specific component, Needless to say, it has a similar effect.

  As described above, according to the abnormal sound diagnosis apparatus of the first embodiment, in the abnormal sound diagnosis apparatus that compares the operation sound at the time of diagnosis and the operation sound at the time of normality to diagnose abnormality of the operation sound, Spectral analysis of the target operating sound to obtain a time frequency distribution, a time frequency analysis unit that detects a predetermined specific component from the time frequency distribution, and a specific component detection unit in the time frequency distribution The specific component counting unit that counts the number of times the specific component detected is detected, and the normal component count value is compared with the specific component count value at the time of diagnosis. Therefore, it is possible to prevent abnormal sounds generated due to other factors from being erroneously detected as abnormal sounds to be diagnosed.

  Further, according to the abnormal sound diagnosis apparatus of the first embodiment, the specific component detection unit obtains the band power time series from the time frequency distribution, and detects the peak included in the time window of the set width from the band power time series. By doing so, the specific component is detected, so that the abnormal sound can be detected with high detection accuracy.

Embodiment 2. FIG.
FIG. 5 is a configuration diagram illustrating the abnormal sound diagnosis apparatus of the second embodiment.
In the second embodiment, the specific component detection unit 40 removes the normal time (learning 31 o'clock) time frequency distribution 311 from the time frequency distribution 321 at the time of diagnosis 32 and removes the normal time frequency distribution 311 from the time frequency distribution 311. A specific component is detected from the time-frequency distribution. Other configurations of the sound collector 1 to the bearing abnormality determination unit 8 are the same as those of the first embodiment shown in FIG.

ここで、ｙ′（ｔ，ｆ）は診断時の時間周波数分布３２１から学習時の時間周波数分布３１１を除去した後の時間周波数分布、ｙ（ｔ，ｆ）は診断時の時間周波数分布３２１、ｙ０（ｔ，ｆ）は学習時の時間周波数分布３１１である。Next, the operation of the abnormal sound diagnosis apparatus of the second embodiment will be described.
The waveform acquisition unit 2 converts the measurement signal output from the sound collector 1 into the waveform data 21, and the time frequency analysis unit 3 outputs the time frequency distribution 311 at the time of learning 31 based on the waveform data 21, and at the time of diagnosis 32 The time frequency distribution 321 is output as in the first embodiment.
In the diagnosis mode, the specific component detection unit 40 removes the time frequency distribution 311 at the time of learning 31 from the time frequency distribution 321 at the time of diagnosis 32 and detects the specific component from the time frequency distribution after the removal. Specifically, first, the time frequency distribution after removal is obtained as shown in Equation (6).

Here, y ′ (t, f) is a time frequency distribution after removing the learning time frequency distribution 311 from the time frequency distribution 321 at the time of diagnosis, and y (t, f) is a time frequency distribution 321 at the time of diagnosis. y0 (t, f) is the time frequency distribution 311 during learning.

ここで、ｆＬ及びｆＵは、それぞれ、所定の周波数範囲を示す下限の周波数と上限の周波数である。Next, the specific component detection unit 40 detects the specific component from the time frequency distribution y ′ (t, f) after removal as follows. First, a band power time series is obtained by obtaining the sum of power values included in a predetermined frequency band at each time with respect to the time frequency distribution y ′ (t, f) after removal. That is, the band power time series B ′ (t) at time t is calculated as in Expression (7).

Here, fL and fU are a lower limit frequency and an upper limit frequency indicating a predetermined frequency range, respectively.

  Next, the specific component detection unit 40 detects the peak of the band power time series B ′ (t). As shown in the above equation (2), the peak is provided with a time window having a predetermined window width W around the time t, and the band power value at the time t is the band power time series in the time window. Is detected as a peak at time t.

  The specific component counting unit 5 counts the number of detections of the specific component from the detection result of the specific component obtained by the specific component detection unit 40 as shown in the above-described equation (3). This counting process is the same as step ST7 in FIG. 2 of the first embodiment, and the processes after step ST7 are also the same as those of the first embodiment.

  FIG. 6 is a diagram showing time-sequential data of bearing abnormality in the same manner as FIG. 3 of the first embodiment. FIG. 6A shows a time frequency distribution (spectrogram) and FIG. The power time series and the detection result of the specific component are shown in FIG. Comparing FIG. 3C and FIG. 6C, the detected peak values are aligned in FIG. 6C. Therefore, the detection of a peak with a small value is suppressed by a simple but robust process of comparing a peak value with a certain threshold value (for example, threshold value 5). Peaks can be suppressed.

  As described above, the abnormal sound detection device according to the present embodiment removes the band power at the time of learning from the band power at the time of diagnosis when detecting a specific component, so even when the band power changes with time, The detected peak values are aligned. For this reason, it is possible to suppress a small peak by comparing with a predetermined threshold value newly provided, and to reduce the possibility of erroneously detecting a small peak caused by a component different from the bearing abnormal noise.

  In addition, the band power time series at the time of learning is obtained from the time frequency distribution at the time of learning, the band power time series at the time of diagnosis is obtained from the time frequency distribution at the time of diagnosis, and the band at the time of learning is obtained from the band power time series at the time of diagnosis. Even if a peak is detected for a band power time series from which the power time series is removed, the same effect can be obtained.

  As described above, according to the abnormal sound diagnosis apparatus of the second embodiment, the specific component detection unit detects the band power from the time frequency distribution obtained by removing the normal time frequency distribution from the time frequency distribution at the time of diagnosis. Since the specific component is detected by obtaining the time series and detecting the peak included in the time window of the set width from the band power time series, the abnormal sound can be detected with high detection accuracy.

  In addition, according to the abnormal sound diagnosis apparatus of the second embodiment, the specific component detection unit obtains the normal band power time series from the normal time frequency distribution, and determines the normal time band from the time frequency distribution at the time of diagnosis. Determined by detecting the peak included in the time window of the set width using the band power time series obtained by obtaining the power time series and removing the normal band power time series from the band power time series at the time of diagnosis Since the component is detected, the abnormal sound can be detected with high detection accuracy.

Embodiment 3 FIG.
The third embodiment is provided with multi-channel sound collecting means composed of a plurality of sound collectors capable of collecting sounds in synchronization with each other, and eliminates the influence of abnormal sound generated from a rotating body under the cage. Thus, an abnormal sound generated from other parts is detected.

FIG. 7 is a configuration diagram illustrating the abnormal sound diagnosis apparatus of the third embodiment.
The abnormal sound diagnosis apparatus of Embodiment 3 includes sound collectors 1a and 1b, waveform acquisition units 2a and 2b, time frequency analysis units 3a and 3b, a specific component detection unit 4, a specific component counting unit 5, and a counting history storage unit 6. , An increment calculation unit 7, a bearing abnormality determination unit 8, specific component removal units 9a and 9b, and abnormality detection units 10a and 10b.

  The sound collector 1a is a sound collector installed under the car of a predetermined specific channel, and the waveform acquisition unit 2a samples the signal from the sound collector 1a and converts it into a digital signal to convert the waveform data 21a. The output processing unit, the time frequency analysis unit 3a, is a processing unit that obtains the time frequency distribution from the waveform data 21a and outputs the time frequency distribution 311a at the time of learning 31a and the time frequency distribution 321a at the time of diagnosis 32a.

  Similarly, the sound collector 1b is a sound collector installed on the car, the waveform acquisition unit 2b is a processing unit that samples the signal from the sound collector 1b, converts it into a digital signal, and outputs the waveform data 21b. The analysis unit 3b is a processing unit that obtains the time frequency distribution from the waveform data 21b and outputs the time frequency distribution 311b at the time of learning 31b and the time frequency distribution 321b at the time of diagnosis 32b. The waveform acquisition unit 2a and the waveform acquisition unit 2b operate in synchronization with each other.

  The specific component detection unit 4 refers to the time frequency distribution 311a and detects a shock component as a specific component included in the time frequency distribution 321a. The specific component count unit 5 is detected by the specific component detection unit 4. A processing unit for counting the number of detections of impact components, a count history storage unit 6 is a storage unit for storing the number of detections counted by the specific component counting unit 5 together with a diagnosis time, and an increment calculation unit 7 is stored in the count history storage unit 6. The processing unit that calculates the increment from the count value at the time of learning, the bearing abnormality determination unit 8 is a processing unit that determines a bearing abnormality based on the increase calculated by the increment calculation unit 7 and outputs a bearing abnormality determination result 81. is there. That is, the sound collector 1a, the waveform acquisition unit 2a, the time frequency analysis unit 3a, the specific component detection unit 4 to the bearing abnormality determination unit 8 have the same configuration as the sound collector 1 to the bearing abnormality determination unit 8 in the first embodiment. It is.

  The specific component removal unit 9a refers to the output of the specific component detection unit 4, removes the specific component from the time frequency distribution 321a at the time of diagnosis, and outputs the time frequency distribution 91a after removal of the specific component, an abnormality detection unit Reference numeral 10a denotes a processing unit that refers to the time frequency distribution 311a at the time of learning, detects an abnormal sound component included in the time frequency distribution 91a after removal of the specific component, and outputs an undercarriage abnormality determination result 101a. Similarly, the specific component removal unit 9b refers to the output of the specific component detection unit 4, removes the specific component from the time frequency distribution 321b at the time of diagnosis, and outputs the time frequency distribution 91b after removal of the specific component, abnormality detection The unit 10b is a processing unit that detects an abnormal sound component included in the time frequency distribution 91b after removal of the specific component with reference to the time frequency distribution 311b at the time of learning and outputs an abnormality determination result 101b on the car.

Next, the operation of the abnormal sound diagnosis apparatus according to the third embodiment will be described. FIG. 8 is a flowchart showing the operations of the specific component removal units 9a and 9b and the abnormality detection units 10a and 10b. The operation from the sound collector 1a to the bearing abnormality determination unit 8 is the same as that in the first embodiment, and will be described with reference to the flowchart of FIG.
The specific component detection unit 4 obtains a band power time series from the time frequency distribution 321a at the time of diagnosis, and outputs a peak detection result of the band power time series (step ST6 in FIG. 2). The position of the detected peak is represented as a set of t where Peak (t) = 1 in the same equation (8) as the equation (2).

  The specific component removal unit 9a removes the specific component detected by the specific component detection unit 4 from the time frequency distribution 321a at the time of diagnosis as follows (step ST12 in FIG. 8). First, the position of the detected peak is obtained from the peak detection result output by the specific component detection unit 4. Next, for the time frequency distribution 321a at the time of diagnosis, a mask having a predetermined duration and a frequency band determined so as to be suitable for covering a specific component at each detected peak position is generated. The value of the time frequency component to be covered is replaced with a special numerical value (denoted as NA) indicating that there is a mask, and output as a time frequency distribution 91a after removal of the specific component. The mask duration and bandwidth are, for example, 10 frames (50 milliseconds) and 100 Hz to 12000 Hz, respectively.

  Similarly, the specific component removal unit 9b obtains the position of the detected peak from the peak detection result output by the specific component detection unit 4, and determines the position of the detected peak with respect to the time frequency distribution 321b at the time of diagnosis. A mask having a predetermined duration and frequency band determined to be suitable for covering a specific component is generated, and the value of the time frequency component covered by the mask is replaced with a special numerical value (NA) having a mask, and the specific component Output as a time frequency distribution 91b after removal.

ここで、ｙ′（ｔ，ｆ）は特定成分除去後の時間周波数分布、ｙ０（ｔ，ｆ）は学習時の時間周波数分布、Ｋは所定の時間周波数領域の数であり、Ω（ｋ）はｋ番目（１≦ｋ≦Ｋ）の所定の時間周波数領域で、総和記号の下付きの条件式は、Ω（ｋ）に含まれ、かつ、マスク有という特別の数値でないような時間ｔ及び周波数ｆの組について総和をとることを表す。所定の時間周波数領域としては、例えば、時間周波数分布の全体を覆う領域と、時間軸をＮ分割、周波数軸をＭ分割した各部分領域とすることができる。The abnormality detection unit 10a compares the time frequency distribution 91a after removal of the specific component with the time frequency distribution 311a at the time of learning, and determines the presence or absence of abnormality as follows (step ST13 in FIG. 8). First, based on empirical values of time frequency distribution of abnormal sounds that can occur in an elevator to be diagnosed, defined on the time frequency distribution and previously determined as prior knowledge for K predetermined time frequency regions Thus, an average intensity calculated from a numerical value other than NA from the time frequency distribution 91a after removal of the specific component and an average intensity calculated from the value of the time frequency distribution 311a during learning corresponding to the numerical value other than NA. Calculate the difference. That is, the difference is calculated as in equation (9).

Here, y ′ (t, f) is the time frequency distribution after removal of the specific component, y 0 (t, f) is the time frequency distribution during learning, K is the number of predetermined time frequency regions, and Ω (k) Is the k-th (1 ≦ k ≦ K) predetermined time-frequency region, and the subscript conditional expression of the sum symbol is included in Ω (k) and is not a special numerical value with a mask t and It represents taking the sum for the set of frequencies f. As the predetermined time frequency region, for example, a region covering the entire time frequency distribution and a partial region obtained by dividing the time axis into N and dividing the frequency axis into M can be used.

ここで、「ＪｕｄｇＯｔｈｅｒｓ」は、判定結果で、１は異常、０は正常であることを示す。また、「ＴｈｒｅｓｈｏｌｄＯｔｈｅｒｓ」は閾値である。この閾値は、例えば、６デシベルとする。Next, when this difference exceeds a predetermined threshold value, it is determined that there is an abnormality under the car in a predetermined time frequency region, and an under car abnormality determination result 101a is output. That is, the determination is made as shown in Equation (10).

Here, “JudgOthers” is a determination result, which indicates that 1 is abnormal and 0 is normal. “ThresholdOthers” is a threshold value. This threshold is, for example, 6 dB.

  Similarly, the abnormality detection unit 10b is calculated from numerical values other than a special numerical value having a mask from the time frequency distribution 91b after removal of the specific component in K predetermined time frequency regions determined in advance as prior knowledge. The average intensity and the average intensity calculated from the corresponding values of the time frequency distribution 311b at the time of learning are calculated. When the difference between the average intensity exceeds a predetermined threshold value, It is determined that there is an abnormality in the car, and an abnormality determination result 101b on the car is output.

In the present embodiment, it is known that the bearing abnormal noise generated under the car reaches the sound collector 1b on the car through the space in the hoistway around the car.
In addition, the abnormal sound diagnosis apparatus of the present embodiment has three types of abnormality determination results (generally, the number of sound collectors + 1), that is, a bearing abnormality determination result 81, a car bottom abnormality determination result 101a, and a car top. The abnormality determination result 101b is output.

FIG. 9 is a diagram illustrating a relationship between an abnormal mode that may occur in an elevator and a determination output according to the present embodiment.
When no bearing abnormal noise is generated under the car, the bearing abnormality determination result 81 is always normal. Further, the under-car abnormality determination result 101a is an abnormality under the car, and if there is an abnormality in parts other than the bearing, the abnormality is determined and output. Similarly, if there is an abnormality on the car, the abnormality determination result 101b on the car determines and outputs the abnormality.
On the other hand, when a bearing abnormality occurs under the cage, the bearing abnormality determination result 81 is always abnormal. Further, the under-car abnormality determination result 101a is an abnormality under the car, and if there is an abnormality in parts other than the bearing, the abnormality is determined and output. Similarly, if there is an abnormality on the car, the abnormality determination result 101b on the car determines and outputs the abnormality.

  As described above, the abnormal sound diagnosis apparatus according to the third embodiment can detect the abnormality of other parts even when the bearing abnormality does not occur under the cage, and even when the bearing abnormality occurs under the cage. The effect of the abnormal bearing noise generated under the car is removed, and the abnormal noise of other parts can be detected.

  As described above, according to the abnormal sound diagnosis apparatus of the third embodiment, the operation sound to be diagnosed is one channel, and the operation sound at the time of diagnosis and the operation sound at the time of normal of each channel are obtained. In comparison, an abnormal sound diagnosis device for diagnosing abnormalities in operating sound of a plurality of channels, wherein a plurality of time frequency analyzing units that respectively perform spectrum analysis of the operating sound of a plurality of channels and obtain a time frequency distribution; Among the time-frequency analysis units, one time-frequency analysis unit is set as a specific time-frequency analysis unit, and a specific component that detects a predetermined specific component from the time-frequency distribution of a specific channel acquired by the specific time-frequency analysis unit A detection unit, a specific component counting unit that counts the number of times of detection of the specific component detected by the specific component detection unit in the time-frequency distribution, a specific component count value in a normal state, and a diagnosis time The specific component detection unit respectively determines from the determination unit that compares with the constant component count value and determines that there is an abnormality when there is an increase over the set value, and the time frequency distribution of the plurality of channels acquired by the plurality of time frequency analysis units. A plurality of specific component removal units for removing a detected specific component and obtaining a time frequency distribution from which the specific component has been removed; a plurality of time frequency distributions obtained by a plurality of specific component removal units; and a plurality of time frequency analysis units Since it includes a plurality of abnormality detection units that perform abnormality detection in each channel based on the normal time frequency distribution, the specific component detection unit of the first or second embodiment is used as the specific component detection unit. It is possible to prevent abnormal sounds generated by other factors from being erroneously detected as abnormal sounds to be diagnosed and to prevent abnormal sounds generated by other factors from occurring. It would reduce the precision output can be prevented.

  In the first to third embodiments, the abnormality of the elevator bearing is determined as the abnormality detection. However, the present invention is not limited to this as an object to be diagnosed, and is an apparatus having a mechanism that performs a periodic operation. Applicable if available.

  Further, within the scope of the present invention, the invention of the present application can be freely combined with each embodiment, modified with any component in each embodiment, or omitted with any component in each embodiment. .

  The abnormal sound diagnosis apparatus according to the present invention counts the number of times a specific component is detected in a time-frequency distribution, and determines that an abnormality occurs when the specific component count value at the time of diagnosis is greater than a set value by a specific component count value at normal time. Therefore, it is possible to prevent the abnormal sound generated due to other factors from being erroneously detected as the abnormal sound to be diagnosed. For this reason, it is suitable for diagnosing abnormal noise generated from a machine having a rotating body such as an elevator or a vehicle.

  1, 1a, 1b Sound collector, 2, 2a, 2b Waveform acquisition unit, 3, 3a, 3b Time frequency analysis unit, 4, 40 Specific component detection unit, 5 Specific component counting unit, 6 Count history storage unit, 7 Increment Calculation unit, 8 bearing abnormality determination unit, 9a, 9b specific component removal unit, 10a, 10b abnormality detection unit, 21, 21a, 21b waveform data, 31, 31a, 31b learning, 32, 32a, 32b diagnosis, 81 bearing abnormality determination Results: 91a, 91b Temporal frequency distribution after removal of specific components, 101a Car lower abnormality determination result, 101b Car upper abnormality determination result, 311, 311a, 311b Time frequency distribution, 321, 321a, 321b Time frequency distribution.

The abnormal sound diagnosis apparatus according to the present invention operates a plurality of channels by comparing the operation sound at the time of diagnosis of each channel with the operation sound at the time of normal using each operation sound to be diagnosed as one channel. a abnormal sound diagnostic apparatus for diagnosing an abnormality in the sound, and a plurality of time-frequency analysis unit for acquiring respective spectral analysis and time-frequency distribution the operation sound of the plurality of channels, among the plurality of time-frequency analysis unit, one the time-frequency analyzer and a specific time-frequency analysis unit, a time-frequency distribution or these particular channels to which the specific time frequency analysis unit obtains a specific component detector for detecting a predefined specific component, time-frequency Compares the specific component count unit that counts the number of detections of the specific component detected by the specific component detector in the distribution with the specific component count value at normal time and the specific component count value at diagnosis And abnormality determination unit when there is an increase of more than set value each, from the time frequency distribution of a plurality of channels in which a plurality of time-frequency analysis unit obtains, respectively, to remove certain components detected the specific component detection unit, A plurality of specific component removal units for obtaining a time frequency distribution from which the specific component has been removed, a plurality of time frequency distributions obtained by a plurality of specific component removal units, and a time frequency distribution at a normal time of a plurality of time frequency analysis units Are provided with a plurality of abnormality detection units for detecting abnormality in each channel .

Claims (5)

In the abnormal sound diagnosis device that compares the operation sound at the time of diagnosis and the operation sound at the normal time to diagnose the abnormality of the operation sound,
A time frequency analysis unit that performs spectrum analysis of the operating sound to be diagnosed to obtain a time frequency distribution;
A specific component detection unit for detecting a predetermined specific component from the time-frequency distribution;
A specific component counting unit that counts the number of detections of the specific component detected by the specific component detection unit in the time-frequency distribution;
An abnormal sound diagnosis comprising: a determination unit that compares the specific component count value at the normal time and the specific component count value at the time of diagnosis and determines an abnormality when there is an increase over a set value apparatus.   The specific component detection unit obtains a band power time series from the time frequency distribution, and detects the specific component by detecting a peak included in a time window of a set width from the band power time series. The abnormal sound diagnosis apparatus according to claim 1.   The specific component detection unit obtains a band power time series from a time frequency distribution obtained by removing the normal time frequency distribution from the time frequency distribution at the time of diagnosis, and calculates a time of a set width from the band power time series. The abnormal sound diagnosis apparatus according to claim 1, wherein the specific component is detected by detecting a peak included in the window.   The specific component detection unit obtains the normal band power time series from the normal time frequency distribution and obtains the normal band power time series from the time frequency distribution at the diagnosis. Using the band power time series obtained by removing the normal band power time series from the band power time series, the specific component is detected by detecting a peak included in a time window of a set width. The abnormal sound diagnosis apparatus according to claim 1. Abnormal sound diagnosis device for diagnosing abnormality of operation sound of multiple channels by comparing operation sound at diagnosis of each channel and operation sound at normal time with operation sound to be diagnosed as one channel Because
A plurality of time-frequency analysis units that respectively perform spectrum analysis of the operation sounds of the plurality of channels to obtain a time-frequency distribution;
Among the plurality of time frequency analysis units, one time frequency analysis unit is set as a specific time frequency analysis unit, and a specific component defined in advance is obtained from the time frequency distribution of the specific channel acquired by the specific time frequency analysis unit. A specific component detection unit to detect;
A specific component counting unit that counts the number of detections of the specific component detected by the specific component detection unit in the time-frequency distribution;
A determination unit that compares the specific component count value at the time of normality with the specific component count value at the time of diagnosis and determines an abnormality when there is an increase over a set value,
A plurality of specific components for removing the specific components detected by the specific component detection unit from the time frequency distributions of the plurality of channels acquired by the plurality of time frequency analysis units and obtaining the time frequency distribution from which the specific components are removed A removal section;
A plurality of anomaly detection units that detect anomalies in each channel by using the plurality of time frequency distributions obtained by the plurality of specific component removal units and the time frequency distributions at the normal time of the plurality of time frequency analysis units; An abnormal sound diagnosis apparatus characterized by that.

Images