JP7188463B2 - ANALYSIS DEVICE, ANALYSIS METHOD, AND PROGRAM - Google Patents

Description

TECHNICAL FIELD The present invention relates to an analysis device, an analysis method, and a program, and more particularly to an analysis device, an analysis method, and a program for analyzing data from sensors that monitor the state of equipment.

There is a method of condition monitoring using vibration and acoustic sensors for manufacturing quality control of production materials using mechanical equipment. For example, by acquiring vibration data generated by a processing machine during processing of production materials and stopping processing of production materials when abnormal vibrations are detected, production loss can be avoided, and the operating status of production equipment can be monitored based on vibrations. It is attracting attention as a technology that improves the production efficiency of the manufacturing industry by monitoring various parts of the body, making maintenance more efficient, and finding the optimal operating conditions for extending the life of equipment.

Patent Literature 1 discloses a method of attaching a sensor to equipment to be monitored and monitoring the equipment based on time-series data measured by the sensor.

JP 2009-270843 A

If an attempt is made to detect equipment anomalies based on data collected by sensors, as in the system described in Patent Literature 1, the amount of data to be analyzed may increase. It is also conceivable to build a normal/abnormal state model by machine learning such as deep learning, but this method only extracts states that are inferred to be abnormal, and only takes actions such as stopping operations. Not applicable.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a technique for efficiently and highly accurately monitoring the state of production equipment.

Each aspect of the present invention employs the following configurations in order to solve the above-described problems.

A first aspect relates to an analysis device.
A first analysis device according to the first aspect includes:
image processing means for imaging the detection result of the vibration sensor provided in the production equipment;
generating means for generating a discriminator by subjecting the imaged data to machine learning processing;
and analysis means for performing state analysis processing of the production equipment using the discriminator.
A second analysis device according to the first aspect includes:
determination means for performing abnormality determination processing of the production equipment using a first discriminator based on the detection result of the vibration sensor provided in the production equipment;
An image processing means for imaging the detection result that could not be determined as normal or abnormal by the first discriminator;
generating means for generating a second discriminator using the imaged data as a target for machine learning processing;
and analysis means for performing state analysis processing of the production equipment using the second discriminator.

A second aspect relates to at least one computer-implemented analysis method.
A first analysis method according to the second aspect is
the analysis device
Image the detection result of the vibration sensor installed in the production equipment,
generating a discriminator by subjecting the imaged data to machine learning processing;
and performing state analysis processing of the production equipment using the discriminator.
A second analysis method according to the second aspect is
the analysis device
performing abnormality determination processing of the production facility using a first discriminator based on the detection result of the vibration sensor provided in the production facility;
Imaging the detection result that could not be determined as normal or abnormal by the first discriminator,
generating a second discriminator using the imaged data as a target for machine learning processing;
and performing state analysis processing of the production facility using the second discriminator.

As another aspect of the present invention, it may be a program that causes at least one computer to execute the method of the second aspect, or a computer-readable recording medium recording such a program. may This recording medium includes a non-transitory tangible medium.
The computer program includes computer program code which, when executed by a computer, causes the computer to perform the analysis method on the analysis device.

Any combination of the above constituent elements, and any conversion of expressions of the present invention into methods, devices, systems, recording media, computer programs, etc. are also effective as embodiments of the present invention.

In addition, the various constituent elements of the present invention do not necessarily have to exist independently of each other. A component may be part of another component, a part of a component may overlap a part of another component, and the like.

In addition, although the method and computer program of the present invention describe multiple procedures in order, the order of description does not limit the order in which the multiple procedures are performed. Therefore, when implementing the method and computer program of the present invention, the order of the plurality of procedures can be changed within a range that does not interfere with the content.

Furthermore, the multiple steps of the method and computer program of the present invention are not limited to being performed at different times. Therefore, the occurrence of another procedure during the execution of a certain procedure, or the overlap of some or all of the execution timing of one procedure with the execution timing of another procedure, and the like are acceptable.

According to each aspect described above, it is possible to provide a technique for efficiently and highly accurately monitoring the state of production equipment.

The above objectives, as well as other objectives, features and advantages, will become further apparent from the preferred embodiments described below and the accompanying drawings below.

1 is a diagram conceptually showing a system configuration of an equipment monitoring system using an analysis device according to an embodiment of the present invention; FIG. It is a figure which shows an example of the data structure of the vibration data which the memory|storage device of this embodiment memorize|stores, and equipment information. It is a block diagram which illustrates the hardware constitutions of each apparatus of this embodiment. It is a functional block diagram showing the logical configuration of the analysis device of the present embodiment. is a diagram showing time-series data of measurement data before imaging. It is a figure which shows the image data imaged by the image processing part. It is a flow chart which shows an example of operation of an analysis device of this embodiment. FIG. 8 is a flowchart showing an example of a detailed flow of imaging processing in step S101 of FIG. 7; FIG. It is a functional block diagram showing the logical configuration of the analysis device of the present embodiment. It is a flow chart which shows an example of operation of an analysis device of this embodiment. 5 is a flow chart showing an example of the procedure of abnormality determination processing in the analysis device of the present embodiment; It is a functional block diagram which shows the logical structure of the image processing part of the analysis apparatus of this embodiment. It is a flow chart which shows an example of operation of an image processing part of an analysis device of this embodiment. It is a functional block diagram showing the logical configuration of the analysis device of the present embodiment. 3 is a diagram showing an example of a data structure of correction information 30; FIG. 9 is a flowchart showing an example of a detailed flow of discrimination processing by an analysis unit; FIG. 17 is a flow chart showing an example of a discriminator update processing procedure using data extracted in step S405 of FIG. 16; FIG. FIG. 17 is a flow chart showing an example of a processing procedure when it is determined to be normal in step S401 of the state analysis processing of FIG. 16; 4 is a flow chart showing an example of the procedure of fault model building processing of the analysis device; FIG. 2 is a flowchart for explaining the analysis device of Example 1; FIG. 10 is a flowchart for explaining the analysis device of Example 2;

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings. In addition, in all the drawings, the same constituent elements are denoted by the same reference numerals, and the description thereof will be omitted as appropriate.

(First embodiment)
FIG. 1 is a diagram conceptually showing the system configuration of a facility monitoring system 1 using an analysis device according to an embodiment of the present invention.
Equipment to be monitored by the equipment monitoring system 1 is the production equipment 10, and in this embodiment, a belt conveyor will be described as an example. In the illustrated example, a plurality of sensors 12 for monitoring the belt conveyor are installed at a plurality of locations along the moving direction of the belt conveyor. Each sensor 12 is, for example, a vibration sensor. Moreover, in the case of a vibration sensor that detects vibration in one direction, a plurality of vibration sensors may be installed at one location in order to detect vibration in multiple directions.

The measurement data output from the vibration sensor is time-series data representing a vibration waveform.
The vibration sensor measures vibrations occurring in the monitored production equipment 10 . The vibration sensor may be a uniaxial acceleration sensor that measures acceleration in a uniaxial direction, a triaxial acceleration sensor that measures acceleration in 3 axial directions, or other sensors. The plurality of vibration sensors may be vibration sensors of the same type, or multiple types of vibration sensors may be mixed.

The analysis device 100 is connected to a gateway (GW) 5 via a network 3 and receives detection results from a plurality of sensors 12 provided in the production facility 10 . Analysis device 100 is connected to storage device 20 . The storage device 20 stores vibration data analyzed by the analysis device 100 . The storage device 20 may be a device separate from the analysis device 100, a device included inside the analysis device 100, or a combination thereof.

In each drawing of the embodiment, the configuration of parts that are not related to the essence of the present invention are omitted and not shown.

FIG. 2 is a diagram showing an example of the data structure of the vibration data 22 and the facility information 24 stored in the storage device 20 of this embodiment.
The vibration data 22 is associated with time information and measurement data for each sensor ID that identifies the vibration sensor. In the facility information 24, the sensor ID of at least one vibration sensor is associated with each facility ID that identifies the facility.

FIG. 3 is a block diagram illustrating the hardware configuration of each device of this embodiment. Each device has a processor 50 , a memory 52 , an input/output interface (I/F) 54 , peripheral circuits 56 and a bus 58 . The peripheral circuit 56 includes various modules. The processing device need not have peripheral circuitry 56 .

A bus 58 is a data transmission path through which the processor 50, memory 52, peripheral circuit 56 and input/output interface 54 mutually transmit data. The processor 50 is an arithmetic processing device such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The memory 52 is , for example, a RAM (Random Access Memory) or a ROM (Read Only Memory). The input/output interface 54 includes an interface for acquiring information from an input device, an external device, an external server, a sensor, or the like, an interface for outputting information to an output device, an external device, an external server, or the like. The input device is, for example, a keyboard, mouse, microphone, or the like. The output device is, for example, a display, a speaker, a printer, a mailer, or the like. The processor 50 can issue commands to each module and perform calculations based on their calculation results.

Each component of the analysis device 100 of this embodiment shown in FIG. 4, which will be described later, is realized by an arbitrary combination of computer hardware and software shown in FIG. It should be understood by those skilled in the art that there are various modifications to the implementation method and apparatus. A functional block diagram showing an analysis apparatus of each embodiment described below shows blocks in units of logical functions, not in units of hardware.

Each function of each unit of the analysis device 100 of FIG. 4 can be realized by the processor 50 reading the program into the memory 52 and executing it.

The computer program of the present embodiment provides a computer (processor 50 in FIG. 3) for realizing the analysis device 100 with a procedure for imaging the detection result of the vibration sensor 12 provided in the production facility 10 and the imaged data. It is described so as to execute a procedure for generating a discriminator as a target of machine learning processing and a procedure for performing abnormality determination processing of the production facility 10 using the discriminator.

The computer program of this embodiment may be recorded on a computer-readable recording medium. The recording medium is not particularly limited, and various forms are conceivable. Also, the program may be loaded from a recording medium into the computer's memory 52 (FIG. 3), or may be downloaded to the computer through a network and loaded into the memory 52 .

A recording medium for recording a computer program includes a non-transitory tangible computer-usable medium in which a computer-readable program code is embedded. When the computer program is executed on the computer, it causes the computer to execute the analysis method of the present embodiment that implements the analysis apparatus 100 .

FIG. 4 is a functional block diagram showing the logical configuration of the analysis device 100 of this embodiment. The analysis device 100 includes an image processing unit 102 , a generation unit 104 and an analysis unit 106 .
The image processing unit 102 converts the detection result of the vibration sensor provided in the production equipment 10 into an image. The generating unit 104 generates the discriminator 110 by subjecting the imaged data to machine learning processing. The analysis unit 106 performs state analysis processing of the production equipment 10 using the discriminator 110 .

Vibration consists of three components, amplitude, frequency, and phase, and the measured data can be expressed in terms of three parameters, displacement, velocity, and acceleration. Since the measurement data of the vibration sensor is time-series data containing multiple vibration waveforms indicated by multiple parameters, even if it is directly applied to machine learning, the characteristics cannot be captured, and an appropriate learning model cannot be generated. Therefore, in this embodiment, the image processing unit 102 converts the measurement data into an image. FIG. 5 is a diagram showing time-series data of measurement data before imaging. FIG. 6 is a diagram showing image data imaged by the image processing unit 102. As shown in FIG.

The image processing unit 102 acquires the detection result (hereinafter also referred to as vibration data) of the sensor 12 of the production equipment 10 . Then, the image processing unit 102 subjects the obtained vibration data to FFT processing and frequency-divides the data, and converts the obtained vibration spectrum data into an image to obtain image data. By these processes, the amount of data can be compressed, and the machine learning process and the discrimination process can be speeded up.

In the embodiment, "acquisition" means that the own device goes to get data or information stored in another device or storage medium (active acquisition), and that the device is output from another device Including at least one of entering data or information (passive acquisition). Examples of active acquisition include requesting or interrogating other devices and receiving their replies, and accessing and reading other devices or storage media. Also, examples of passive acquisition include receiving information that is distributed (or sent, pushed, etc.). Furthermore, "acquisition" may be selecting and acquiring received data or information, or selecting and receiving distributed data or information.

The discriminator 110 is generated by the generation unit 104 by subjecting the imaged data to machine learning processing. The analysis unit 106 uses this discriminator 110 to perform state analysis processing of the production equipment 10 .

Machine learning can use, for example, deep learning, but is not limited to this. The state of the production facility 10 determined by the discriminator 110 is, for example, normal or otherwise. Hereinafter, states other than normal are also referred to as "Unknown." In this embodiment, the discriminator 110 does not discriminate an abnormal state. The machine learning processing of the discriminator 110 will be described in detail in an embodiment described later.

The determination result of the discriminator 110 may be output so that the operator can refer to it. The output method may be, for example, display on the display of the analysis device 100, printout from the analysis device 100 to a printer, or transmission to another device (eg, operation terminal, etc.) via a communication line. may be

The operation of the analysis device 100 of this embodiment configured in this way will be described. FIG. 7 is a flow chart showing an example of the operation of the analysis device 100 of this embodiment.
First, the image processing unit 102 images the detection result of the vibration sensor provided in the production equipment 10 (step S101). Then, the generation unit 104 generates the discriminator 110 by subjecting the imaged data to machine learning processing (step S103). Then, the analysis unit 106 uses the discriminator 110 to perform state analysis processing of the production equipment 10 (step S105).

FIG. 8 is a flow chart showing an example of the detailed flow of the imaging process in step S101 of FIG. First, the image processing unit 102 performs FFT processing on the measurement data of the sensor 12 and then frequency-divides (step S113), converts the obtained vibration spectrum data into an image, and outputs the image data (step S115). The image data obtained in step S115 undergoes state analysis processing of the production facility 10 using the discriminator 110 by the analysis unit 106 in step S105 of FIG.

As described above, in the present embodiment, after the measurement data of the sensor 12 is imaged by the image processing unit 102, the image data imaged by the generation unit 104 is subjected to machine learning processing, and the discriminator 110 Then, using this discriminator 110, the analysis unit 106 performs state analysis processing of the production equipment 10. FIG. As described above, according to the present embodiment, when machine learning is performed on the detection result of the vibration sensor, the amount of data can be compressed by imaging the vibration waveform data to be machine-learned, thereby reducing the processing load. The processing speed can be increased while reducing.

(Second embodiment)
FIG. 9 is a functional block diagram showing the logical configuration of the analysis device 100 of this embodiment. The analysis device 100 includes an image processing unit 102, a generation unit 104, and an analysis unit 106 similar to the analysis device 100 of FIG.

The abnormality determination unit 120 uses the first discriminator 124 to perform abnormality determination processing of the production equipment 10 based on the detection result of the vibration sensor provided in the production equipment 10 . The abnormality determination unit 120 outputs the result of abnormality determination processing using the first discriminator 124 . The output result may be used for monitoring processing of the production equipment 10 in the equipment monitoring system 1 .

In addition, the image processing unit 102 of the present embodiment performs the image processing of the above embodiment in that the detection result to be imaged is a detection result for which the first discriminator 124 cannot determine whether it is normal or abnormal. It is different from the part 102 . The generation unit 104 generates the second discriminator 126 for performing machine learning processing on the measured data for which it was not possible to discriminate whether it is normal or abnormal by the threshold analysis.

The vibration measured by the vibration sensor includes multiple vibration waveforms due to multiple factors. In general, measurement data detected by a vibration sensor undergoes frequency analysis by performing Fast Fourier Transform (FFT) processing. Characteristic frequencies (peaks) are detected by subjecting the measured data to FFT processing, and abnormality diagnosis can be made by determining whether the detected peak level is normal or abnormal using a threshold value.

However, it may not be possible to distinguish between normality and abnormality by this threshold analysis. The generation unit 104 generates the second discriminator 126 by subjecting the measured data, which could not be determined as normal or abnormal by the threshold analysis, to machine learning processing. This second classifier 126 corresponds to the classifier 110 in FIG.

The case where "cannot be discriminated by the first discriminator 124" is, for example, when the vibration waveform pattern registered in the first discriminant model 128 and the detection result do not match, the likelihood of the matching result is equal to or greater than a predetermined value. This is the case when the reliability is not obtained and the reliability of the standard or higher (for example, the detection rate of 90% or higher) required for equipment diagnosis is not obtained.

The analysis unit 106 uses the second discriminator 126 to perform state analysis processing of the production equipment 10 . The analysis unit 106 uses the second discriminator 126 to perform state analysis processing of the production equipment 10 on the image data of the detection result for which the first discriminator 124 could not discriminate between normal and abnormal.

The first classifier 124 uses the first discriminant model 128 to discriminate between normal and abnormal. The first discrimination model 128 is, for example, a model using vibration spectrum pattern matching or threshold determination.

As an example, the first discriminator 124 determines the peak level of a specific frequency, the ratio of the maximum and average peak levels, the S/N ratio (Signal-to -Noise ratio), and at least one of the peak level integrated values in a specific frequency range, set a threshold value for the obtained value, and perform abnormality determination processing based on whether it is within the threshold range. conduct. If the value is within the threshold range, the first discriminator 124 determines that it is normal, and if it is outside the threshold range, it determines that it is abnormal.

The first discriminator 124 may perform the determination process described above using a plurality of specific frequencies. In that case, the first discriminator 124 may output a determination result for each specific frequency. In addition, the first discriminator 124 may determine that there is an abnormality when even one of the plurality of specific frequencies has a value outside the threshold value, or may determine that a predetermined number or more of the plurality of specific frequencies exceed the threshold value. It may be determined that there is an abnormality when the frequency has a value outside the threshold, it may be determined as an abnormality when all of the plurality of specific frequencies are outside the threshold, or all of the plurality of specific frequencies is within the threshold range, it may be determined to be normal. Further, a combination of a plurality of specific frequency values for each failure event is modeled and registered in the first discriminant model 128, and the first discriminator 124 performs pattern matching with the first discriminant model 128 to determine the failure Events may be determined. The abnormality determination unit 120 may notify the operator of the corresponding malfunction event item for the determined malfunction event.

The threshold may be set automatically or manually by an operator. In the case of automatic setting, the generation unit 104 generates the peak level of a specific frequency, the ratio of the maximum and average peak levels, the S/N ratio ( (Signal-to-Noise ratio), and at least one of the peak level integrated values in the specific frequency range, and the threshold may be set by detecting the boundary range between the normal state and the abnormal state.

In the case of manual setting, the generation unit 104 calculates the peak level of a specific frequency, the ratio of the maximum and average peak levels, the S/N ratio (Signal -to-Noise ratio), and at least one of the peak level integrated values in the specific frequency range is presented to the operator, and the setting of the threshold value of the first discriminator 124 is operated by the operator. may be accepted by Various methods of presenting the results to the operator are conceivable. For example, the results may be displayed on the display of the analysis device 100, printed out from the analysis device 100 on a printer, or sent to another operator via a communication line. It may be transmitted to a device (for example, an operation terminal, etc.).

The second discriminator 126 is generated by the generator 104 and corresponds to the discriminator 110 generated by the generator 104 in the above embodiment. The generator 104 performs machine learning only on the data determined to be normal by the second discriminator 126 , normalizes the normal state, and updates the second discriminant model 130 . Here, normalization refers to modeling the pattern of imaging data of measurement data of the vibration sensor during normal operation. The second discriminator 126 extracts data out of the range of the second discriminant model 130 as, for example, "Unknown". That is, the analysis unit 106 uses the second discriminator 126 to determine the state of the production equipment 10 based on the second discriminant model 130 for the detection result that could not be discriminated by the first discriminator 124. , is determined to be normal or “Unknown”. The second discriminator 126 does not discriminate an abnormal state of the production facility 10, and discriminates a state other than normal as "Unknown".

The data determined as "Unknown" by the second discriminator 126 may be referred to and analyzed by the operator. The result of analysis by the operator may be reflected in the threshold of the abnormality determination process of the detection result. A method of using data determined as "Unknown" will be described in detail in an embodiment described later.

The operation of the analysis device 100 of this embodiment configured in this way will be described. FIG. 10 is a flow chart showing an example of the operation of the analysis device 100 of this embodiment.
First, the abnormality determination unit 120 uses the first discriminator 124 to perform abnormality determination processing for the production equipment 10 based on the detection result of the vibration sensor provided in the production equipment 10 (step S121). Then, the image processing unit 102 forms an image of the detection result (NO in step S123) for which the first discriminator 124 could not discriminate between normal and abnormal (step S125). If it can be determined (YES in step S123), this process ends.

Then, the generation unit 104 generates the second discriminator 126 by subjecting the imaged data to machine learning processing (step S127). Then, the analysis unit 106 uses the second discriminator 126 to perform state analysis processing of the production equipment 10 (step S129).

Further, the computer program of the present embodiment causes the computer (processor 50 in FIG. 3) for realizing the analysis device 100 to detect an abnormality in the production equipment 10 based on the detection result of the vibration sensor 12 provided in the production equipment 10 . A procedure for performing determination processing using the first discriminator 124, a procedure for imaging the detection result that could not be discriminated as normal or abnormal by the first discriminator 124, and the imaged data as a target for machine learning processing. and the procedure for performing the state analysis processing of the production equipment 10 using the second discriminator 126 are described to be executed.

As described above, in the present embodiment, when the abnormality determination unit 120 uses the first discriminator 124 and the abnormality determination process based on the detection result of the vibration sensor fails to obtain a determination result, the vibration at that time is determined. The image processing unit 102 converts the detection result of the sensor into an image, and the second discriminator 126 discriminates whether it is normal or "Unknown". Thus, according to the present embodiment, the second discriminator 126 performs machine learning only on the detection results of the vibration sensors corresponding to the normal state, and normalizes the normal state. In production, etc., even if manufacturing conditions are always in flux and it is difficult to accumulate information on defect events, the accuracy of discrimination can be improved.

Further, as in the above-described embodiment, when machine learning is performed on the detection results of the vibration sensor, the amount of data can be compressed by imaging the vibration waveform data that is the target of machine learning processing, thereby reducing the processing load. At the same time, the processing speed can be increased.

(Third Embodiment)
FIG. 11 is a flow chart showing an example of the procedure of abnormality determination processing in the analysis device 100 of this embodiment. This embodiment is the same as the analysis device 100 of FIG. 9 except that it has a configuration for outputting the result of the abnormality determination processing of the production equipment 10 using the first discriminator 124 .

First, the abnormality determination unit 120 acquires vibration data from a vibration sensor provided in the production equipment 10 (step S301). Then, the abnormality determination unit 120 executes abnormality determination processing for the acquired vibration data using the first discriminator 124 (step S303). Here, after the vibration data is subjected to FFT processing, normality or abnormality is determined by the first determination model 128 of the abnormality determination unit 120 (step S305).

From the frequency distribution obtained by FFT processing, the abnormality determination unit 120 determines the peak level of a specific frequency, the ratio of the maximum and average peak levels, the signal-to-noise ratio (S/N ratio), and the range of the specific frequency. are obtained for at least one of the integrated values of the peak levels of , and a threshold value is used to determine whether these values are normal or abnormal (step S305). If it is within the threshold range (NO in step S305), the first discriminator 124 determines that the production equipment 10 is normal (step S307), and the abnormality determination unit 120 determines that the production equipment 10 is normal. is output (step S311).

If it is outside the threshold range (YES in step S305), the first discriminator 124 determines that the production facility 10 is in an abnormal state (step S309), and the determination result indicates that the production facility 10 is in an abnormal state. is output (step S311).

Further, when the first discriminator 124 cannot discriminate between normality and abnormality (discrimination not possible in step S305), the abnormality determination unit 120 outputs vibration data to the image processing unit 102 (step S313). 10 to step S125.

In this embodiment, the first discriminator 124 is used for the abnormality determination processing of the production equipment 10, and the second discriminator 126 is not used for the abnormality determination processing. The discrimination result of the second discriminator 126 is used to update the first discriminant model 128 of the first discriminator 124, and this configuration will be described later in the embodiment.

As described above, in the present embodiment, the abnormality determination unit 120 uses the first discriminator 124 to perform abnormality determination processing for the production facility 10 , and the abnormality determination unit 120 uses the second discriminator 126 as the production facility 10 not used for the abnormality determination process. The first discriminator 124 can be updated using the discrimination result of the second discriminator 126 updated by the generation unit 104 . According to this configuration, the abnormality determination process can be performed using the first discriminator 124 reflecting the determination result of the second discriminator 126, so that the accuracy of the determination result can be improved.

(Fourth embodiment)
FIG. 12 is a functional block diagram showing the logical configuration of the image processing unit 102 of the analysis device 100 of this embodiment. The analysis apparatus 100 of the present embodiment is the same as the above embodiment except that the image processing unit 102 performs processing for removing noise from the measurement data and then performs imaging processing. The configuration of this embodiment may be combined with the configuration of any other embodiment.

The image processing unit 102 includes a noise removal unit 112 and an imaging processing unit 114 . The noise removal unit 112 performs noise removal processing on the detection result. The imaging processing unit 114 images the detection result after the noise removal processing by the noise removal unit 112 has been performed. Noise removal makes the vibration waveform of the measurement data clearer.

The noise removal process includes, for example, pre-measurement and storage of environmental noise from a plurality of arranged vibration sensors, and differential processing based on the noise data. However, noise removal processing may be performed by other methods.

FIG. 13 is a flow chart showing an example of the operation of the image processing unit 102 of the analysis device 100 of this embodiment. The flowchart of FIG. 13 further includes step S111 in addition to steps S113 and S115 of the flowchart of FIG. 8 described in the above embodiment.

In step S111, the noise removal unit 112 performs noise removal processing on the measurement data. Then, the image processing unit 114 performs FFT processing on the data subjected to the noise removal processing in step S111, and then performs frequency division processing (step S113). ). The image data obtained in step S115 undergoes state analysis processing of the production equipment 10 using the discriminator 110 by the analysis unit 106 in step S103 of FIG.

As described above, in the present embodiment, the measurement data of the sensor 12 subjected to noise removal processing by the noise removal unit 112 is imaged by the image processing unit 114 . With this configuration, according to the present embodiment, it is possible to speed up the machine learning process or the discrimination process as in the above embodiment, and furthermore, the vibration waveform of the measurement data becomes clearer by removing noise, and the FFT process and The accuracy of imaging processing is improved, and the accuracy and reliability of analysis results of measurement data are improved.

(Fifth embodiment)
FIG. 14 is a functional block diagram showing the logical configuration of the analysis device 100 of this embodiment. The analysis device 100 of the present embodiment extracts data determined to be abnormal by the state analysis processing of the production equipment 10, and updates the first discrimination model 128 of the first discriminator 124 based on the data. It is the same as the above embodiment except that it has

The analysis device 100 includes an image processing unit 102, a generation unit 104, an analysis unit 106, and an abnormality determination unit 120 similar to the analysis device 100 of FIG.

The extraction unit 140 extracts data determined to be abnormal (“Unknown”) in the state analysis process using the second discriminator 126 . The generation unit 104 receives correction information based on the extracted data and updates the first discrimination model 128 of the first discriminator 124 . Here, the correction information includes information that associates the malfunction event with the vibration characteristic.

The data to be extracted includes the raw vibration waveform data received from the vibration sensor before image processing by the image processing unit 102, corresponding to the imaging data determined as "Unknown", and the time information of the data. including. The vibration indicated by the vibration data determined as "Unknown" may be caused by an unspecified malfunction of the production equipment 10 .

Therefore, an operator manually analyzes the extracted vibration data together with the time information, using the operation information, state information, information on the workpiece, etc. of the production equipment 10 possessed by the equipment monitoring system 1. Identify the failure event that caused the

Therefore, the analysis device 100 outputs the data extracted by the extraction unit 140 and presents it to the operator. Various methods of presenting the data to the operator are conceivable. For example, the data may be displayed on the display of the analysis device 100, printed out from the analysis device 100 on a printer, or may be presented to the operator via a communication line. It may be transmitted to a device (for example, an operation terminal, etc.).

The operator inputs the correction information 30 shown in FIG. 15, in which the vibration characteristic information of the corresponding vibration is linked to the defect event specified by the operator, into the analysis device 100 using the operation screen or the like. The generation unit 104 receives the input correction information 30 and updates the first discrimination model 128 of the first discriminator 124 .

FIG. 16 is a flowchart showing an example of a detailed flow of discrimination processing by the analysis unit 106. As shown in FIG. The analysis unit 106 applies the image data output in step S115 of FIG. 13 to the second discriminator 126, and performs state analysis processing of the production facility 10 (step S401). Data determined to be normal by the second discriminator 126 ("normal" in step S401) is transferred to the generation unit 104 and machine-learned as normal data (step S403). On the other hand, the analysis unit 106 extracts data ("Unknown" in step S401) outside the normalization range by the second discriminator 126 (step S405).

Note that the image data output in step S115 of FIG. 8 may be similarly processed using the flow of FIG.

FIG. 17 is a flow chart showing an example of a classifier update processing procedure using the data extracted in step S405 of FIG.
First, the extraction unit 140 outputs the data (vibration data and time information) extracted in step S405 (step S411). Here, for example, it is displayed on the display of the analysis device 100 . Then, the operator analyzes the vibration data displayed in step S411 together with the time information, using the operation information, state information, workpiece information, etc. of the production facility 10 possessed by the facility monitoring system 1, and Identify the fault event that caused the vibration. The operator creates information that associates the specified failure event with the corresponding vibration characteristic information, and inputs it as correction information for the threshold value of the first discriminator 124 according to the operation screen of the analysis device 100 .

The generation unit 104 receives the input correction information (step S413), and updates the first discriminator 124 using the received correction information (step S415).

Specifically, the vibration characteristic information and the failure event included in the received correction information are registered in the first discrimination model 128, and the peak level of the specific frequency, the ratio of the maximum and average peak levels, the S/N ratio ( (Signal-to-Noise ratio), and at least one of the peak level integrated values in the specific frequency range, and the boundary range between normal and abnormal states is detected and a threshold is set. Alternatively, the operator may set each threshold value based on each value obtained from the vibration characteristic information corresponding to the specified malfunction event and input it using the operation screen.

As described above, in the present embodiment, the extraction unit 140 extracts the "Unknown" data of the second discriminator 126, presents it to the operator, and the generation unit 104 analyzes the malfunction event and vibration characteristic information by the operator. is received, and the first discriminator 124 is updated based on the correction information. As described above, according to the present embodiment, the measurement data that the first discriminator 124 could not discriminate is further discriminated by the second discriminator 126, and the data marked as "Unknown" is extracted. Since the operator analyzes and the result is reflected in the first classifier 124, the accuracy of the abnormality determination process can be improved.

In addition, since the second discriminator 126 is machine-learned only on the information of the normal state, it is difficult to accumulate information on defective events because the manufacturing conditions are always in flux , such as in small-lot, high-mix, variable-variety production. However, since the first discrimination model 128 can be updated, the accuracy of discrimination of the abnormal state of the production facility 10 can be improved.

(Modification of Fifth Embodiment)
The analysis device 100 in FIG. 14 has a configuration in which the extraction unit 140 is added to the configuration of the analysis device 100 in FIG. 9 . As a modification thereof, the analysis device 100 of FIG. 4 may be configured to include the extraction unit 140 .

The analysis device 100 further includes an abnormality determination section 120 and an extraction section 140 . The abnormality determination unit 120 performs characteristic analysis processing of the vibration of the vibration sensor indicated by the detection result, and performs abnormality determination processing of the production facility 10 using a threshold value. The extraction unit 140 extracts data determined by the state analysis processing that the production facility 10 is not in a normal state. The generation unit 104 receives correction information based on the extracted data and updates the threshold. The correction information includes information that associates the malfunction event with the vibration characteristic.

Although the embodiments of the present invention have been described above with reference to the drawings, these are examples of the present invention, and various configurations other than those described above can also be adopted.
For example, FIG. 18 is a flow chart showing an example of a processing procedure when it is determined to be normal in step S401 of the state analysis processing of FIG. This embodiment is the same as the above embodiment except that the detection result determined as normal by the second discriminator 126 is used as training data for the second discriminator 126 .

The generation unit 104 uses the detection result determined to be normal by the state analysis processing using the second discriminator 126 (normal in step S401 in FIG. The second discriminant model 130 of the discriminator 126 is updated (step S501).

According to this configuration, the generation unit 104 updates the second discrimination model 130 by using the detection result determined to be normal by the state analysis processing using the second discriminator 126 as teaching data of the normal state of the production equipment 10. be done. In this way, it is possible to generate normal training data from measured data that could not be discriminated in the abnormality determination process by the first discriminator 124, to update the second discriminant model 130, and to update the second discriminant model 126. can improve the accuracy of discrimination.

Furthermore, in addition to the first discriminator 124 and the second discriminator 126 of the above embodiments, the analysis device 100 may include a third discriminator (not shown). FIG. 19 is a flow chart showing an example of a processing procedure in which the analysis device 100 constructs a defect model using the third discriminator and identifies a defect event. The third discriminator acquires measurement data determined to be normal by the first discriminator 124 (step S601). Furthermore, the third discriminator acquires the measurement data discriminated as Unknown by the second discriminator 126 (step S603). Then, the third discriminator machine-learns these data and builds a model that classifies the failure event (step S605). The noise removal processing and imaging processing described in the above embodiment may also be performed on each measurement data machine-learned by the third discriminator.

Furthermore, using a third discriminator, it is determined whether the measurement data of the sensor 12 of the production equipment 10 is normal or abnormal, and if abnormal, a malfunction event is identified (step S607).

In addition, the position information of each of the plurality of sensors 12 is stored in the facility information 24, and the relationship between the vibration data and the position information is further machine-learned, and the classification model 202, the first discrimination model 128, and the second At least one of the discriminant models 130 may be reflected.

Furthermore, information such as measurement conditions (equipment type, environment (temperature, humidity), etc.) for measurement data of a plurality of sensors 12 may be stored in the facility information 24 . By grouping measurement data of conditions close to the measurement conditions, operating states, operating conditions, etc. included in the operation information of the production equipment 10, machine learning is performed on the measurement data for each group, and a classification model 202 and a first discrimination model 128 are generated. , and the second discriminant model 130 .

(Example 1)
FIG. 20 is a flowchart for explaining the analysis device of Example 1. FIG.
First, when vibration data is input from the sensor 12 of the production equipment 10, the abnormality determination unit 120 performs FFT processing in the first discriminator 124 (step S11). At this time, the first discriminant model 128 is used to identify the vibration characteristics by pattern matching processing. Then, the first discriminator 124 determines that the vibration characteristic is normal when the vibration characteristic is within the range of the threshold, and determines that it is abnormal when it is outside the range of the threshold . The abnormality determination unit 120 outputs this result to the equipment monitoring system 1 as the abnormality determination result of the production equipment 10 (not shown).

Further, the abnormality determination unit 120 extracts the data for which the first discriminator 124 could not determine whether or not it was normal in step S11 (step S13), and uses it as a target for machine learning processing by the second discriminator 126. It is passed to the analysis unit 106 .

The analysis unit 106 performs noise removal processing on the measurement data of the sensor 12 that could not be discriminated by the first discriminator 124 (step S15), performs frequency analysis, and forms an image (step S17). The analysis unit 106 discriminates the imaged data using the second discriminator 126 (step S19), and discriminates whether or not the state of the production equipment 10 is normal (step S21). If determined to be normal (YES in step S21), the generator 104 machine-learns the measured data determined to be normal, and updates the second discrimination model 130 of the second discriminator 126 (step S23). .

On the other hand, if it is not determined to be normal ("Unknown") (NO in step S21), the extraction unit 140 extracts and outputs the measurement data that is Unknown (step S31). The operator refers to this extracted unknown data, analyzes it together with the operation information of the production equipment 10, etc., and identifies the failure event. Then, correction information linking the malfunction event and the vibration characteristic is input to the analysis device 100 (step S33).

The generation unit 104 receives the input correction information, and updates the first discrimination model 128 and the threshold based on the received correction information (step S35).

In this way, the second discriminator 126 performs machine learning on the measurement data for which the first discriminator 124 could not determine an abnormality by the analysis device 100, thereby extracting abnormal unknown data and 10 operation information by the operator to identify the failure event, input to the analysis device 100 as correction information that links the vibration characteristics and the failure event, and based on the correction information, the first discriminator 124 The first discriminant model 128 and thresholds can be updated.

(Example 2)
FIG. 21 is a flowchart for explaining the analysis device of Example 2. FIG.
The analysis apparatus 100 of this embodiment further includes a third discriminator 200 in addition to the first discriminator 124 and the second discriminator 126 .
The third discriminator 200 performs machine learning using the measured data determined as normal by the first discriminator 124 and the measured data determined as Unknown by the second discriminator 126 (step S41). The second discriminator 126 builds a normal/abnormal classification model 202 by machine learning. Classification model 202 classifies failure events.

Using this classification model 202, the third discriminator 200 can discriminate whether the measured data is normal or abnormal, and can also identify and specify a failure event.

Although the present invention has been described with reference to the embodiments and examples, the present invention is not limited to the above embodiments and examples. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.
In the present invention, acquisition and use of information relating to users shall be done legally.

Some or all of the above embodiments can also be described as the following additional remarks, but are not limited to the following.
1. image processing means for imaging the detection result of the vibration sensor provided in the production equipment;
generating means for generating a discriminator by subjecting the imaged data to machine learning processing;
an analysis device that uses the discriminator to analyze the state of the production equipment.
2. 1. In the analysis device according to
determination means for performing characteristic analysis processing of the vibration of the vibration sensor indicated by the detection result, and performing abnormality determination processing of the production equipment using a threshold value;
an extraction means for extracting data determined by the state analysis processing that the production equipment is not in a normal state,
The generating means receives correction information based on the extracted data, updates the threshold,
The analysis device, wherein the correction information includes information that associates a malfunction event with a vibration characteristic.
3. determination means for performing abnormality determination processing of the production equipment using a first discriminator based on the detection result of the vibration sensor provided in the production equipment;
An image processing means for imaging the detection result that could not be determined as normal or abnormal by the first discriminator;
generating means for generating a second discriminator using the imaged data as a target for machine learning processing;
analysis means for performing state analysis processing of the production equipment using the second discriminator;
comprising
analysis equipment.
4. 3. In the analysis device according to
further comprising extracting means for extracting data determined by the state analysis processing that the production facility is not in a normal state;
The generating means receives correction information based on the extracted data and updates the first discriminator,
The analysis device, wherein the correction information includes information that associates a malfunction event with a vibration characteristic.
5. 1. to 4. In the analysis device according to any one of
The generating means is an analysis device that uses the data determined to be normal by the state analysis process as teacher data for the machine learning process.
6. 1. to 5. In the analysis device according to any one of
Further comprising processing means for performing noise removal processing on the detection result,
The analysis device, wherein the image processing means forms an image of the detection result after the noise removal processing is performed by the processing means.
7. 1. to 6. In the analysis device according to any one of
The production equipment is a belt conveyor,
The analysis device, wherein the vibration sensor is a plurality of vibration sensors provided on the belt conveyor.

8. the analysis device
Image the detection result of the vibration sensor installed in the production equipment,
generating a discriminator by subjecting the imaged data to machine learning processing;
performing state analysis processing of the production equipment using the classifier;
analysis method.
9. 8. In the analysis method described in
The analysis device further
Performing characteristic analysis processing of the vibration of the vibration sensor indicated by the detection result, performing abnormality determination processing of the production equipment using a threshold value,
extracting data determined by the state analysis processing that the production facility is not in a normal state;
Receiving correction information based on the extracted data, updating the threshold,
The analysis method, wherein the correction information includes information that associates a malfunction event with a vibration characteristic.
10. the analysis device
performing abnormality determination processing of the production facility using a first discriminator based on the detection result of the vibration sensor provided in the production facility;
Imaging the detection result that could not be determined as normal or abnormal by the first discriminator,
generating a second discriminator using the imaged data as a target for machine learning processing;
Performing state analysis processing of the production equipment using the second discriminator,
analysis method.
11. 10. In the analysis method described in
The analysis device further
extracting data determined by the state analysis processing that the production facility is not in a normal state;
Receiving correction information based on the extracted data and updating the first discriminator;
The analysis method, wherein the correction information includes information that associates a malfunction event with a vibration characteristic.
12. 8. to 11. In the analysis method according to any one of
The analysis device further
An analysis method, wherein the data determined to be normal by the state analysis processing is used as teaching data for the machine learning processing.
13. 8. to 12. In the analysis method according to any one of
The analysis device further
Performing noise removal processing on the detection result,
An analysis method, wherein the detection result after the noise removal processing is performed is imaged.
14. 8. to 13. In the analysis method according to any one of
The production equipment is a belt conveyor,
The analysis method, wherein the vibration sensor is a plurality of vibration sensors provided on the belt conveyor.

15. to the computer,
A procedure for imaging the detection result of the vibration sensor provided in the production equipment,
A procedure for generating a discriminator by subjecting the imaged data to machine learning processing,
A program for executing a procedure for performing state analysis processing of the production equipment using the discriminator.
16. 15. In the program described in
A procedure of performing characteristic analysis processing of the vibration of the vibration sensor indicated by the detection result and performing abnormality determination processing of the production equipment using a threshold value;
A procedure for extracting data determined by the state analysis processing that the production facility is not in a normal state;
causing the computer to further execute a step of receiving correction information based on the extracted data and updating the threshold value in the generating step;
The program, wherein the correction information includes information that associates a malfunction event with a vibration characteristic.
17. to the computer,
A procedure for performing abnormality determination processing of the production equipment using a first discriminator based on the detection result of the vibration sensor provided in the production equipment,
A procedure for imaging the detection result that could not be determined as normal or abnormal by the first discriminator,
A procedure for generating a second discriminator using the imaged data as a target for machine learning processing,
A program for executing a procedure for performing state analysis processing of the production equipment using the second discriminator.
18. 17. In the program described in
A procedure for extracting data determined by the state analysis processing that the production facility is not in a normal state;
causing the computer to further execute a procedure of accepting correction information based on the extracted data and updating the first discriminator in the generating procedure;
The program, wherein the correction information includes information that associates a malfunction event with a vibration characteristic.
19. 15. to 18. In the program according to any one of
A program for causing a computer to further execute a procedure, in the generating procedure, of using the data determined to be normal by the state analysis processing as training data for the machine learning processing.
20. 15. to 19. In the program according to any one of
a procedure for performing noise removal processing on the detection result;
A program for causing a computer to further execute a procedure of imaging the detection result after the noise removal process in the imaging procedure.
21. 15. to 20. In the program according to any one of
The production equipment is a belt conveyor,
The program, wherein the vibration sensor is a plurality of vibration sensors provided on the belt conveyor.

This application claims priority based on Japanese Patent Application No. 2019-019068 filed on February 5, 2019, and the entire disclosure thereof is incorporated herein.

Claims (7)

determination means for performing abnormality determination processing of the production equipment using a first discriminator based on the detection result of the vibration sensor provided in the production equipment;
An image processing means for imaging the detection result that could not be determined as normal or abnormal by the first discriminator;
generating means for generating a second discriminator using the imaged data as a target for machine learning processing;
analysis means for performing state analysis processing of the production equipment using the second discriminator;
comprising
analysis equipment. In the analysis device according to claim 1 ,
further comprising extracting means for extracting data determined by the state analysis processing that the production facility is not in a normal state;
The generating means receives correction information based on the extracted data and updates the first discriminator,
The analysis device, wherein the correction information includes information that associates a malfunction event with a vibration characteristic. In the analysis device according to claim 1 or 2 ,
The generating means is an analysis device that uses the data determined to be normal by the state analysis process as teacher data for the machine learning process. In the analysis device according to any one of claims 1 to 3 ,
Further comprising processing means for performing noise removal processing on the detection result,
The analysis device, wherein the image processing means forms an image of the detection result after the noise removal processing is performed by the processing means. In the analysis device according to any one of claims 1 to 4 ,
The production equipment is a belt conveyor,
The analysis device, wherein the vibration sensor is a plurality of vibration sensors provided on the belt conveyor. the analysis device
performing abnormality determination processing of the production facility using a first discriminator based on the detection result of the vibration sensor provided in the production facility;
Imaging the detection result that could not be determined as normal or abnormal by the first discriminator,
generating a second discriminator using the imaged data as a target for machine learning processing;
Performing state analysis processing of the production equipment using the second discriminator,
analysis method. A program that causes at least one computer to execute the analysis method according to claim 6 .

Images