Abnormality monitoring device and abnormality monitoring program

Description: TECHNICAL FIELD [Detailed description of the invention]

【0001】

[Technical field to which the invention belongs]

The present invention relates to an abnormality monitoring device that monitors the presence or absence of an abnormality in a diagnostic target that performs continuous operation over time, such as a rolling mill that continuously rolls, and an abnormality monitoring program that operates a computer as an abnormality monitoring device.

【0002】

[Conventional technology]

Conventionally, equipment diagnosis by various equipment diagnosis methods for determining the presence or absence of abnormality in equipment or equipment has been executed or proposed. This equipment diagnosis does not detect only catastrophic failures that require equipment to be destroyed or stopped immediately, but rather to damage bearings, such as in rotating machinery, prior to such catastrophic failure. It is possible to continue the operation sufficiently for now, such as entering or the wear of a certain moving part has progressed, but if it is left as it is, the abnormality that may lead to a serious failure in the future is targeted for detection. There is a need.

【0003】

As a typical example of such an equipment diagnosis method, for example, an acoustic vibration waveform when the equipment or equipment is in a normal state is obtained, the acoustic vibration waveform is spectrally analyzed to examine its characteristics, and the presence or absence of an abnormality is detected. At that time, the acoustic vibration waveform of the device or equipment is obtained and spectrum analysis is performed, and whether or not there is a peak of a specific frequency component that is not seen in the normal state in the spectrum, or when the combination of peaks is normal. It is known that anomalies are detected depending on whether or not they are the same.

【0004】

Further, in Japanese Patent Application Laid-Open No. 7-43259, an acoustic vibration waveform when the device or equipment is in a normal state is obtained, and an inverse filter is created based on the acoustic vibration waveform to detect the presence or absence of an abnormality. At that time, the acoustic vibration waveform of the equipment or equipment is obtained, the inverse filter obtained in advance is applied to the acoustic vibration waveform to obtain the residual signal, and the residual signal is analyzed to detect the abnormality of the equipment or equipment. It has been proposed to detect.

【0005】

Further, Japanese Patent Application Laid-Open No. 8-304124 obtains a plurality of acoustic vibration waveforms when the device or equipment is in a normal state, and reverses based on, for example, one of the plurality of acoustic vibration waveforms. A filter is created, and the inverse filter is applied to, for example, the remaining plurality of acoustic vibration waveforms to obtain a plurality of residual signals, and a plurality of statistical variables are obtained based on each of the plurality of residual signals. Even when detecting the presence or absence of an abnormality, a plurality of acoustic vibration waveforms of the device or equipment are obtained, and the above-mentioned inverse filter obtained in advance is applied to the plurality of acoustic vibration waveforms to obtain a plurality of residual signals. , Multiple statistical variables were obtained based on these multiple residual signals, and between the multiple statistical variables obtained when in the normal state and the multiple statistical variables obtained when detecting the presence or absence of anomalies. Therefore, it has been proposed to detect the presence or absence of an abnormality in the equipment or equipment by performing a test or estimation by a method such as an F test or a t test.

【0006】

The method of detecting abnormalities in equipment or equipment by performing the above spectrum analysis is also a fairly effective method depending on the nature of the equipment or equipment to be diagnosed, and the method of creating the above inverse filter or statistical test Etc. is a more effective method.

【0007】

[Problems to be Solved by the Invention]

However, regardless of which of the various equipment diagnosis methods described above is adopted, after the physical quantity reflecting the state of the diagnosis target is captured by the sensor, the calculation is performed until the presence or absence of the abnormality of the diagnosis target is determined. Is required, and if the diagnosis target performs continuous operation over time, it will continue to operate even while the calculation is being performed, but during that time there is a risk that it will be excluded from the abnormality detection target. is there.

【0008】

In view of the above circumstances, the present invention uses an abnormality monitoring device and a computer as such an abnormality monitoring device, which can continuously detect the presence or absence of an abnormality in a diagnosis target that performs continuous operation over time without a blank period. The purpose is to provide an abnormality monitoring program to operate.

【0009】

[Means for solving problems]

The abnormality monitoring device of the present invention that achieves the above object is an abnormality monitoring device that monitors the presence or absence of an abnormality of a diagnosis target that performs continuous operation over time, and captures a predetermined physical quantity that reflects the state of the diagnosis target. A sensor that obtains a time-continuous signal representing a physical quantity and each partial signal obtained by sequentially cutting out the signal obtained by the above sensor without interruption or partially overlapping are received and received alternately or cyclically. It is characterized by being provided with a plurality of abnormality monitoring units for determining the presence or absence of an abnormality to be diagnosed based on a partial signal.

【0010】

The abnormality monitoring device of the present invention includes a plurality of abnormality monitoring units, sequentially cuts out signals obtained from sensors without interruption or partially overlapping, and alternately cuts out each partial signal (when there are two abnormality monitoring units). Alternatively, the presence or absence of an abnormality is detected alternately or cyclically by sharing the signal (when there are three or more abnormality monitoring units), for example, when the occurrence of a single event is targeted for standby monitoring. Even so, the presence / absence of abnormality detection can be continued without causing a blank period for detecting the presence / absence of abnormality.

【0011】

Here, in the above-mentioned abnormality monitoring device of the present invention, a reference calculation unit that obtains reference data by performing an operation including an operation for obtaining an inverse filter based on a reference signal obtained by a sensor when the diagnosis target is in a normal state. All of the above-mentioned plurality of abnormality monitoring units perform an operation including an operation of obtaining a residual signal by applying an inverse filter to the diagnostic signal obtained at the time of abnormality monitoring by a sensor, and based on the result of the operation. , It is preferable that the presence or absence of an abnormality to be diagnosed is determined.

【0012】

Here, the above-mentioned "operation including the operation for obtaining the inverse filter" is a concept including the case where it is composed only of the operation for obtaining the inverse filter. In that case, the inverse filter can be used as the above reference data. it can. Further, the "operation including the operation for obtaining the inverse filter" may include the operation for obtaining the inverse filter, and as described above, the inverse filter is performed based on, for example, one reference signal among a plurality of reference signals. It may be an operation of creating and applying the inverse filter thereof to a plurality of other reference signals to obtain a plurality of statistical variables. In that case, the plurality of statistical variables thus obtained can be the reference data.

【0013】

Further, the above "calculation including the calculation for obtaining the residual signal by operating the inverse filter" is the same as the above, and may be composed only of the calculation for obtaining the residual signal, or the above-mentioned Japanese Patent Application Laid-Open No. 7 -43 As described in Japanese Patent Application Laid-Open No. 43259, calculations such as finding the moving average value of the power of the residual signal and finding data that is convenient for determining the presence or absence of an error by calculating the residual signal Alternatively, a plurality of residual signals are obtained by applying an inverse filter to a plurality of diagnostic signals as described in JP-A-8-304124 described above, and a plurality of statistics are obtained based on the plurality of residual signals. It may be an operation for obtaining a target variate.

【0014】

By using an inverse filter, it is possible to eliminate constant noise from the signal, and it is possible to determine the presence or absence of an abnormality to be diagnosed with higher accuracy.

【0015】

Further, the abnormality monitoring program of the present invention that achieves the above object operates in a computer, and the computer is operated as an abnormality monitoring device that monitors the presence or absence of an abnormality of a diagnosis target that continuously operates in time. In the program, this computer is connected to a sensor that captures a predetermined physical quantity that reflects the state of the diagnosis target and obtains a temporally continuous signal that represents the physical quantity, and is obtained by the above sensor. A plurality of abnormality monitoring units that alternately or cyclically share and receive each partial signal that is sequentially cut out without interruption or partially overlapping, and determine the presence or absence of an abnormality of the diagnosis target based on the received partial signal. Characterized by having Here, in the above-mentioned abnormality monitoring program of the present invention, a reference calculation unit that obtains reference data by performing an operation including an operation for obtaining an inverse filter based on a reference signal obtained by a sensor when the diagnosis target is in a normal state. All of the above-mentioned plurality of abnormality monitoring units perform an operation including an operation of obtaining a residual signal by applying an inverse filter to the diagnostic signal obtained at the time of abnormality monitoring by a sensor, and based on the result of the operation. Therefore, it is preferable to determine the presence or absence of an abnormality to be diagnosed.

【0016】

BEST MODE FOR CARRYING OUT THE INVENTION

Hereinafter, embodiments of the present invention will be described.

【0017】

FIG. 1 is a block diagram showing a basic embodiment of the abnormality monitoring device of the present invention.

【0018】

The abnormality monitoring device 10 is an abnormality monitoring device that monitors the presence or absence of an abnormality in the diagnosis target 20 that performs continuous operation over time, such as a rolling mill that continuously rolls, and is a sensor 11 and a reference calculation unit 12. , And a plurality of abnormality monitoring units 13A, 13B, ..., 13N.

【0019】

The sensor 11 of the abnormality monitoring device 10 captures a predetermined physical quantity (for example, sound, vibration, etc.) reflecting the state of the diagnosis target 20, and obtains a time-continuous signal representing the physical quantity.

【0020】

Further, the reference calculation unit 12 performs an operation including an operation of obtaining an inverse filter based on the reference signal obtained by the sensor 11 when the diagnosis target 20 is in a normal state, and the reference data is obtained by this. The reference data obtained by the reference calculation unit 12 may be the inverse filter itself or a statistical variable as described above. In this embodiment, the inverse filter itself is adopted as the reference data.

【0021】

When the diagnosis target 20 has a plurality of operation states, in the reference calculation unit 12, each inverse filter corresponding to each operation state is set based on each reference signal obtained when the diagnosis target 20 is in each operation state. Desired. However, in the case where one operating state or a state that can be regarded as one operating state is continuous, only one inverse filter may be obtained regardless of the operating state.

【0022】

For example, in the case of a rotating machine in which the diagnosis target 20 rotates slowly, each phase range when one rotation of the rotating machine is divided into a plurality of overlapping phase ranges without interruption or sequentially partially overlapped is regarded as each operating state. The inverse filter may be obtained for each partial reference signal corresponding to each of these phase ranges, and only one inverse filter may be obtained regardless of the phase range unless the operating state changes significantly depending on the phase of rotation. You may.

【0023】

Further, in the plurality of abnormality monitoring units 13A, 13B, ..., 13N, each partial diagnostic signal obtained by sequentially cutting out the diagnostic signal obtained at the time of abnormality monitoring by the sensor 11 without interruption or partially overlapping is alternately or cyclically. Based on the received partial diagnosis signal, the presence or absence of abnormality in the diagnosis target 20 is detected. Specifically, in these abnormality monitoring units 13A, 13B, ..., 13N, each partial diagnosis signal is the same as each operation state of the diagnosis target 20 when each partial diagnosis signal is obtained by the sensor 11. Each residual signal is obtained by operating an inverse filter corresponding to each operating state of. Alternatively, when the inverse filter is obtained regardless of the operating state, the same inverse filter is applied to each partial diagnosis signal to obtain each residual signal. Further, the power of each of these residual signals is obtained, and the presence or absence of an abnormality in the diagnosis target 20 is determined by comparing each of these powers with the threshold value.

【0024】

When the diagnosis target 20 is a rotating machine that rotates slowly, a plurality of abnormality monitoring units 13A, 13B, ..., 13N send a diagnostic signal obtained at the time of abnormality monitoring by the sensor 11 to one rotation of the rotating machine. Receives and receives each partial diagnostic signal alternately or cyclically when it is divided into multiple partial diagnostic signals corresponding to the same multiple phase ranges as the multiple phase ranges separated during the operation of the inverse filter. The residual signal is obtained by applying an inverse filter obtained based on the partial reference signal in the same phase range as the phase range corresponding to the partial diagnosis signal obtained by the reference calculation unit 12 on the partial diagnosis signal. By comparing the power of the residual signal with the threshold value, the presence or absence of abnormality of the rotating machine is determined.

【0025】

FIG. 2 is a system conceptual diagram showing an embodiment of the abnormality monitoring device of the present invention.

【0026】

The monitoring object 21 corresponds to an example of the diagnostic object referred to in the present invention. A vibration sensor 210A and an acoustic sensor 210B are attached to or installed in the vicinity of the monitoring object 21 as sensors for obtaining a signal for detecting the presence or absence of an abnormality. The vibration sensor 210A and the acoustic sensor 210B need not be provided with both of them, and may be provided with only one of them depending on the characteristics of the object to be monitored 21.

【0027】

The waveform signals obtained by these vibration sensors 210A and acoustic sensor 210B are input to the filter amplifier 211, and when an abnormality occurs in the monitored object 21, only the frequency band included in the sound or vibration caused by the abnormality passes through. It is filtered to do so and is amplified appropriately. The signal after passing through the filter amplifier 211 is input to the A / D converter 221 constituting the A / D conversion unit 220 and sampled, and the sampled data is sequentially stored in the ring structure memory 221. The ring structure memory 221 is a type of memory in which sampling data is sequentially stored for one round of the ring, and after one round, the memory is sequentially overwritten. Although a ring structure memory is shown here as an example, it does not have to be a memory having a ring structure as long as it can store sampling data over a sufficient length.

【0028】

The sampling data stored in the ring structure memory 221 is sequentially read out as waveform data for each memory area in which a part thereof is sequentially overlapped.

【0029】

In the abnormality monitoring computer 230, two comparative diagnosis tasks A and B are operating in the present embodiment, and the waveform data sequentially read from the ring structure memory 221 of the A / D conversion unit 220 is the two comparative diagnostic tasks. It is passed to tasks A and B alternately. When each of the comparative diagnosis tasks A and B receives the waveform data, it is determined whether or not there is an abnormality in the monitored object 21 based on the received waveform data. In detecting the presence or absence of abnormalities in these two comparative diagnostic tasks A and B, the data obtained by sampling in the same manner as above when it is known that the monitored object 21 is in a normal state in advance is used. An inverse filter is obtained based on the above, and in each comparative diagnosis task A and B, a residual signal is obtained by applying the inverse filter to the waveform data received at the time of abnormality monitoring, and the power of the residual signal is a threshold value. Is compared with, and the presence or absence of an abnormality is determined according to the magnitude. Here, if an abnormality is detected in either of the two comparative diagnostic tasks A and B, an alarm indicating that the abnormality has been detected is output.

【0030】

Only one inverse filter, which is required when the monitored object 21 is operating normally, needs to be obtained when the monitored object 21 is always operating in the same operating state, but it rotates slowly, for example. When a rotating machine is used as a monitoring object, one rotation of the rotating object is divided into a plurality of phase ranges, and each phase range is based on the waveform data obtained in each of the plurality of phase ranges. Each of the inverse filters was obtained, and even during abnormality monitoring, the waveform data obtained in the same phase range as the phase range when the reverse filter was obtained was read from the ring structure memory 221 and read out. An inverse filter obtained based on the data in the same phase range as the phase range obtained from the waveform data is applied to the waveform data. By doing so, it is possible to detect the presence or absence of an abnormality in the rotating machine with higher accuracy.

【0031】

However, even when a rotating machine that rotates slowly is the object to be monitored, each waveform data obtained in a plurality of phase ranges when the rotating machine is operating normally has the same statistically the same property. If it has (so to speak, the rotating machine rotates uniformly regardless of the phase of rotation), it seeks the only inverse filter regardless of the phase range, and its only inverse filter action for any phase range. You may let me.

【0032】

FIG. 3 is an external perspective view of an abnormality monitoring computer that operates as an embodiment of the abnormality monitoring device of the present invention. The abnormality monitoring device as an embodiment of the present invention is a combination of an abnormality monitoring device main body composed of the hardware of the diagnostic computer 100 and software executed therein, and a sensor or the like (not shown here). It has been realized.

【0033】

The abnormality monitoring computer 100 includes a main body 101 having a built-in CPU, RAM memory, magnetic disk, communication board, etc., a CRT display 102 that displays a screen on the display screen 102a according to an instruction from the main body, and the abnormality monitoring computer. It is equipped with a keyboard 103 for inputting operator instructions and character information, and a mouse 104 for inputting instructions according to an icon or the like displayed at an arbitrary position on the display screen.

【0034】

The main body 101 is loaded with a CD-ROM 105 (see FIG. 4) so that it can be taken out and loaded, and also has a built-in CD-ROM drive for driving the loaded CD-ROM 105.

【0035】

Here, the abnormality monitoring program is stored in the CD-ROM 105, the CD-ROM 105 is loaded in the main body 101, and the abnormality monitoring program stored in the CD-ROM 105 by the CD-ROM drive is the abnormality monitoring computer. Installed in 100 magnetic disks. When the abnormality monitoring program installed in the magnetic disk of the abnormality monitoring computer 100 is started, the abnormality monitoring computer 100 is an embodiment of the abnormality monitoring device main body excluding the sensor and the like among the abnormality monitoring devices of the present invention. Operate.

【0036】

FIG. 4 is a hardware configuration diagram of the abnormality monitoring computer 100 shown in FIG.

【0037】

In this hardware configuration diagram, the central processing unit (CPU) 111, RAM 112, magnetic disk controller 113, CD-ROM drive 115, mouse controller 116, keyboard controller 117, display controller 118, communication board 119, and A / D conversion boards 120 are shown, which are interconnected by bus 110.

【0038】

The CD-ROM drive 115 is loaded with the CD-ROM 105 and accesses the loaded CD-ROM 105, as described with reference to FIG.

【0039】

The communication board 119 is connected to a machine control device (not shown) that controls the diagnosis target, and the machine control device diagnoses the control state of the diagnosis target (whether it is in an operating state or a stationary state, or a rotating machine). Control state information indicating the current rotation angle (phase) of the rotating machine in the case of the target is input.

【0040】

Further, the A / D conversion board 120 is connected to a vibration sensor, an acoustic sensor 22, and the like shown in the sensor FIG. 1 for obtaining a signal for abnormality monitoring. This A / D conversion board 120 corresponds to the A / D conversion unit 220 shown in FIG. 2, and the signal picked up by the sensor is input, sampled, temporarily stored in the memory, and then partially overlapped one by one. It plays the role of reading each memory area and taking it inside.

【0041】

Further, FIG. 4 shows a magnetic disk 114 accessed by the magnetic disk controller 113, a mouse 104 controlled by the mouse controller 116, a keyboard 103 controlled by the keyboard controller 117, and a CRT display 102 controlled by the display controller 118. Is also shown.

【0042】

FIG. 5 is a diagram schematically showing an embodiment of the abnormality monitoring program of the present invention.

【0043】

Here, this abnormality monitoring program 130 is stored in the CD-ROM 105, and the CD-ROM 105 is loaded and driven in the abnormality monitoring computer 100 shown in FIGS. 3 and 4, and the abnormality stored in the CD-ROM 105 is stored. When the monitoring program 130 is installed and executed in the abnormality monitoring computer 100, the abnormality monitoring computer 100 operates as an embodiment of the abnormality monitoring device of the present invention together with a sensor for detecting the state to be diagnosed. ..

【0044】

The abnormality monitoring program 130 shown in FIG. 5 is composed of a reference calculation unit 131 and a plurality of abnormality monitoring units 132A, 132B, ......, 132N.

【0045】

The reference calculation unit 131 corresponds to the reference calculation unit 12 of the abnormality monitoring device 10 shown in FIG. 1, and the plurality of abnormality monitoring units 132A, 132B, ......, 132N refer to the abnormality monitoring device 10 of FIG. The configurations corresponding to the constituent abnormality monitoring units 13A, 13B, ..., 13N, respectively, except for the sensor 11 of the abnormality monitoring device 10 shown in FIG. 1, are the abnormality monitoring computer 100 shown in FIGS. When composed of the installed abnormality monitoring program shown in FIG. 5, the reference calculation unit 12 and the abnormality monitoring unit 132A, 132B, ......, 132N of the abnormality monitoring device 10 shown in FIG. 1 are all included. , Computer hardware, OS (operating system), and anomaly monitoring program as an application program. On the other hand, the anomaly monitoring program 130 shown in FIG. 5 is composed of only the application program among them. Has been done. The actions of the parts constituting the abnormality monitoring program 130 of FIG. 5 are the same as the actions of the corresponding parts of the abnormality monitoring device 10 of FIG.

【0046】

FIG. 6 is a flowchart of two comparative diagnostic tasks A and B executed in the abnormality monitoring computer shown in FIGS. 3 and 4. These two comparative diagnostic tasks A and B are one of the specific examples of the abnormality monitoring units 132A, 132B, ......, 132N of the abnormality monitoring program 130 shown in FIG.

【0047】

In the comparative diagnosis task A, waveform data consisting of data stored in a part of the memory area of the sampling data (A / D conversion data) stored in the ring structure memory in the A / D conversion board is generated. Input (step a1), comparison with the reference waveform data input when the diagnosis target is in the normal state is executed (in the present embodiment, as described above, an inverse filter based on the reference waveform data is obtained in advance. Then, the inverse filter is applied to the waveform data input this time) (step a2), and if there is an abnormality as a result of the comparison (step a3), a diagnostic alarm indicating that there was an abnormality is output. To.

【0048】

If there is no abnormality, wait for the data to be shared by this comparative diagnostic task A to be input to the ring structure memory (step a4), and the waveform data to be shared by this comparative diagnostic task A in the ring structure memory. Is stored (step a5), the waveform data is input (step a6), and the abnormality detection is repeated in the same manner thereafter. When the data to be shared by the comparative diagnosis task A is not stored in the ring structure memory for a predetermined time or longer, the system error handling is performed (step a7).

【0049】

The operation of the comparative diagnosis task B is the same as the operation of the comparative diagnosis task A, and the difference is that by receiving the waveform data alternately with the comparative diagnosis task A, the timing of receiving the waveform data and the subsequent abnormality determination The only difference is the timing.

【0050】

In this way, in the present embodiment, the abnormality to be diagnosed is monitored alternately by the two comparative diagnosis tasks A and B, as a whole, without interruption.

【0051】

FIG. 7 is a system structure diagram when a sheet rolling machine that transports sheets while rolling is targeted for diagnosis.

【0052】

Foreign matter may get on the sheet, and if the foreign matter is caught in the sheet rolling mill while it is on the sheet, the sheet may become defective. Therefore, here, a sound collecting microphone is provided to capture the abnormal sound generated when the foreign matter is bitten, and the acoustic waveform signal obtained by the sound collecting microphone is input to the abnormality monitoring device main body.

【0053】

In this sheet rolling mill, the roll feed speed is controlled by a machine control device to one of a plurality of speeds. The roll feed speed information is input to the abnormality monitoring device main body as control status information.

【0054】

In the abnormality monitoring device main body, when it is confirmed that no foreign matter is caught, an inverse filter is required for each feed rate controlled by the machine control device, and at the time of abnormality monitoring, two or three or more. The abnormal noise caused by the biting of foreign matter is detected without interruption by sharing the tasks alternately or cyclically.

【0055】

Although each of the above embodiments is an example of detecting an abnormality by capturing vibration or sound, an abnormality monitoring may be performed by capturing a different physical quantity other than vibration or sound.

【0056】

Next, an inverse filter and a method for detecting the presence or absence of an abnormality using the inverse filter will be described.

【0057】

Any time series signal can be regarded as the output when white noise is input to a suitable linear system. Determining the corresponding linear system from a given time series signal is called linear predictive analytics, and there is an established method. An autoregressive model (AR model) is usually obtained in this way. This is when the sampled and discretized time series signals are X (n), n = 1, 2, ..., the signal X (n) at the nth time point is the data at M points before that. It is decided as follows.

【0058】

[Number 1]

<img file="JP2003106946A_D0001.tif" />【0059】

Here, e (n) is a virtual input signal to the linear system and is white noise. Given a time series signal, a set of coefficients {A from that data_k} Determines an autoregressive model for that time-series signal.

【0060】

Now the set of coefficients {A_kWhen} is obtained, Y (n) is defined as follows using the time series signal data {X (n)}. At this time, Y (n) is said to be a linear predicted value of X (n).

【0061】

[Number 2]

<img file="JP2003106946A_D0002.tif" />【0062】

Therefore, when the following quantity is calculated, from equations (1) and (2), X (n)-Y (n) = e (n) ... (3) And the residual is white noise. That is, the input white noise is obtained by subtracting the predicted value Y (n) obtained from the previous M data from the time series signal data X (n) at the nth time point. Here, subtracting the predicted value Y (n) from X (n) to obtain the residual e (n) is referred to as the inverse filter acting. If a certain time-series signal can be represented by an appropriate autoregressive model in this way, white noise can be obtained by applying an inverse filter constructed using the time-series signal to the original time-series signal. That is, the input signal is whitened by the inverse filter. In this case, the input time series signal does not have to be the signal itself used when designing the inverse filter, and if the autoregressive model is the same, that is, a signal with the same characteristics, a whitened signal is obtained as an output. Can be done. However, if the characteristics of the time-series signal are different from those used in the design, whitening is not achieved even if the inverse filter is applied, and white noise cannot be obtained.

【0063】

Therefore, using the first time-series signal that carries the normal operating sound, vibration, etc. (operating sound, etc.), the inverse filter is configured in advance, and a new first that carries the operating sound, etc. at an arbitrary time point. By obtaining a time-series signal of 2 and monitoring the output by applying an inverse filter to the second time-series signal, it is possible to detect a time-series signal (residual signal) different from the normal time.

【0064】

Specifically, in the present embodiment, the following signal processing method can be adopted.

【0065】

First, in the reference calculation unit 12 of FIG. 1, FFT (Fast Fourier Transform) is performed using 1024 points of sound signal data or vibration signal data obtained when the diagnosis target is in a normal state, and then the power spectrum is obtained. Ask. Next, IFFT (Inverse Fast Fourier Transform) is performed to obtain the autocorrelation function, which is used to calculate by Levinson's algorithm (see, for example, Mikami's "Introduction to Digital Signal Processing" published by CQ Publishing), and the coefficient of the inverse filter. {A_k} To find.

【0066】

After that, the abnormality monitoring unit 13A, 13B, ..., 13N operates the inverse filter to obtain the residual signal, but in the present embodiment, the calculation for obtaining the residual signal is performed by the coefficient {a. It is performed by moving average calculation using [k]}.

【0067】

Here, in order to obtain the moving average of the power of the residual signal, 128 data are first extracted from the time series of the residual signal, the autocorrelation function is obtained through FFT, power spectrum calculation, and IFFT, and the autocorrelation function is obtained from the peak value of the origin. Seeking power. After that, the power is calculated sequentially while shifting the start point of the data by 50 points.

【0068】

As an inverse filter, as an example, the one in Table 1 is adopted as the order M = 27 and the coefficient {a [k]}.

【0069】

[table 1]

a [0] = 1.000000 a [1] =-2.887330 a [2] = 3.947344 a [3] =-3.535249 a [4] = 2.447053 a [5] =-1.620133 a [6] = 1.315352 a [7] =-1.268161 a [8] = 0.937471 a [9] =-0.380573 a [10] =-0.040919 a [11] = 0.284076 a [12] =-0.353665 a [13] = 0.397849 a [14] =-0.533185 a [15] = 0.501902 a [16] =-0.238178 a [17] =-0.003048 a [18] = 0.192420 a [19] =-0.166854 a [20] =-0.010498 a [21] = 0.061383 a [22] = 0.017323 a [23] =-0.014146 a [24] =-0.131247 a [25] = 0.239157 a [26] =-0.242444 a [27] = 0.115678 8 to 11 show an example of the waveform obtained from the equipment in the normal state, FIG. 8 shows the signal waveform of the sound signal collected from the equipment in the normal state, and FIG. 9 shows the inverse filter to this sound signal. The signal waveform of the residual signal after the action is applied, FIG. 10 shows the power spectrum of the residual signal, and FIG. 11 shows the moving average of the power of the residual signal.

【0070】

12 to 15 show an example of the waveform obtained when the equipment is in an abnormal state, and each figure shows a signal waveform having the same format as that of FIGS. 8 to 11. Even if FIG. 8 and FIG. 12, which show the waveforms of sound signals obtained from the equipment in the normal state and the abnormal state, respectively, are directly compared, it is difficult to immediately judge the normal or abnormal from these. However, it is possible to judge whether it is normal or abnormal by analyzing each other of FIGS. 9 and 13 showing the residual signals obtained by applying the inverse filter to these.

【0071】

As can be easily understood by comparing FIGS. 9 and 13, the amplitude of the residual signal is extremely small when the equipment is in a normal state, but the amplitude is extremely large when the equipment is in an abnormal state. Become. Therefore, it is possible to judge normality / abnormality based on the maximum value of the power in the residual signal. For example, if the equipment has a residual signal amplitude that is 10 dB or more larger than the maximum power of the signal obtained when the equipment is in a normal state, it is judged as abnormal, and if it has a residual signal amplitude less than this, it is judged as normal. , Can judge normal / abnormal.

【0072】

Further, as shown in FIGS. 10 and 14, in the spectrum obtained by Fourier transforming the residual signal, when an abnormality occurs, the power spectrum increases. For example, the power peak in FIG. 10 is 100 dB or less, but in FIG. 14, the power peak reaches almost 120 dB.

【0073】

Further, when the moving averages of the powers of the residual signals are compared with each other in FIGS. 11 and 15, it can be seen that the moving average increases due to the abnormality. For example, if the moving average data that is 20 dB or more larger than the maximum value of the moving average when the equipment is in the normal state is shown, it can be judged as abnormal, and if the data is less than this, it can be judged as normal. When this method is adopted, it is particularly easy to determine the presence or absence of an abnormality, and the abnormality can be detected in a short time. Therefore, it is possible to perform real-time detection at the site, which is particularly preferable. Depending on the type of abnormality, the presence of defects can be detected more accurately by the analysis of the power spectrum than by the analysis of the moving average of the power.

【0074】

In the embodiment described above, an inverse filter is created as the reference data referred to in the present invention, a residual signal is obtained, and the presence or absence of an abnormality is detected based on the residual signal. It is not always necessary to adopt this detection method, and for example, the above-mentioned method of spectrum analysis or a method of performing statistical estimation or testing may be adopted.

【0075】

[Effect of the invention]

As described above, according to the present invention, it is possible to continuously detect the presence or absence of an abnormality in the diagnosis target that performs continuous operation over time without setting a blank period.

[Simple explanation of drawings]

[Figure 1]

It is a block diagram which shows one basic embodiment of the abnormality monitoring apparatus of this invention.

[Figure 2]

It is a system conceptual diagram which shows one Embodiment of the abnormality monitoring apparatus of this invention.

[Fig. 3]

It is external perspective view of the abnormality monitoring computer which operates as one Embodiment of the abnormality monitoring apparatus of this invention.

[Fig. 4]

It is a hardware configuration diagram of the abnormality monitoring computer shown in FIG.

[Fig. 5]

It is a schematic diagram which shows one Embodiment of the abnormality monitoring program of this invention.

[Fig. 6]

It is a flowchart of two comparative diagnosis tasks A and B executed in the abnormality monitoring computer shown in Fig. 3 and Fig. 4.

[Fig. 7]

It is a system structure diagram when the sheet rolling machine which conveys while rolling a sheet is targeted for diagnosis.

[Fig. 8]

It is a waveform diagram of the sound signal obtained from the diagnosis target in a normal state.

[Fig. 9]

It is a signal waveform diagram obtained by applying an inverse filter to the signal of FIG.

[Fig. 10]

It is a power spectrum diagram obtained from the signal of FIG.

[Fig. 11]

It is a figure which shows the moving average of the electric power obtained from the signal of FIG.

[Fig. 12]

It is a waveform diagram of the sound signal obtained from the diagnosis target in an abnormal state.

[Fig. 13]

It is a signal waveform diagram obtained by applying an inverse filter to the signal of FIG.

[Fig. 14]

It is a power spectrum diagram obtained from the signal of FIG.

[Fig. 15]

It is a figure which shows the moving average of the electric power obtained from the signal of FIG.

[Explanation of symbols]

10 Abnormality monitoring device 11 Sensor 12 Reference calculation unit 13A, 13B, ..., 13N Abnormality monitoring unit 20 Diagnosis target 21 Observed object 100 Diagnostic computer 130 Abnormality monitoring program 131 Reference calculation unit 132A, 132B, ..., 132N Abnormality monitoring unit 211 filter amplifier 220 A / D conversion unit 221 A / D converter 222 Ring structure memory 230 Abnormality monitoring task