KR20220013256A - Method and system for diagnosing abnormality of a plurality of apparatus based on sequential learning - Google Patents

Abstract

Provided is a method for diagnosing abnormality of a plurality of devices to be diagnosed based on sequential learning, which automatically reconfigures an inspection schedule. According to the present invention, the method comprises the following steps: acquiring sensing data for ta plurality of devices to be diagnosed; predicting whether abnormality occurs in at least one of the plurality of devices to be diagnosed from an abnormal diagnosis prediction model, which predicts predicting whether abnormality occurs in the plurality of devices to be diagnosed, on the basis of the sensing data; accumulatively counting the number of occurrences of the abnormality by distinguishing the devices to be diagnosed for which the occurrence of the abnormality is predicted by each identification information; applying the cumulative counted result value as a weight to arrange an inspection order of the plurality of devices to be diagnosed; and accurately diagnosing whether the plurality of devices to be diagnosed are abnormal through a precise diagnosis model based on the sorted inspection order.

Description

Method and system for diagnosing abnormalities of a plurality of diagnosis target devices based on sequential learning techniques

The present invention relates to a method and system for diagnosing abnormalities in a plurality of diagnosis target devices based on a sequential learning technique.

When an abnormality occurs in machinery or equipment in industrial sites such as manufacturing and production industries or in sites equipped with power generation equipment, the economic loss is enormous.

In performing abnormal diagnosis, conventionally, abnormal diagnosis was performed for each individual device or individual facility. That is, when performing diagnosis of several devices and facilities sequentially, abnormal diagnosis is performed in a predetermined order by the manager, or a failure diagnosis is performed first by selecting a malfunctioning device.

This has a problem in that the efficiency of abnormal diagnosis is low because abnormality diagnosis is performed or precise diagnosis is performed without considering the greater or urgent degree of abnormality of the device or facility.

In recent years, anomaly diagnosis technology for machinery and equipment using artificial intelligence has been widely used. As an example, the abnormal diagnosis of the offshore wind power device is performed by selecting a blade on which the turbine is not driven, taking a large amount of images of the blades, and then performing an artificial intelligence-based abnormality diagnosis model on the captured visible light image. proceeds

In the case of wind power generation, since power generation is prioritized on the day the turbine is driven, the abnormal diagnosis test is postponed to a day when the turbine is not driven, that is, a day when the wind speed is low.

In this current method, the accuracy of diagnostic analysis can be increased through the visible light image-based abnormal diagnosis, but the inspection cycle is not clear, and the mechanical device in which the device abnormality is expected is detected and diagnosed in advance, or the diagnosis priority is raised compared to the general device. There is a problem that such a method is not presented.

An embodiment of the present invention predicts whether abnormal diagnosis occurs through an abnormal diagnosis prediction model before performing a precise diagnosis on the diagnosis target device, and adjusts the precision check weight of the diagnosis target device based on the prediction result, It is a method of diagnosing abnormalities of a plurality of diagnosis target devices based on a sequential learning technique that can accurately diagnose a diagnosis target device according to priority by sequentially applying an abnormal diagnosis predictive model and a precise diagnosis model, rather than a simple sequential abnormal diagnosis. Methods and systems are provided.

However, the technical task to be achieved by the present embodiment is not limited to the technical task as described above, and other technical tasks may exist.

As a technical means for achieving the above-described technical problem, the method for diagnosing whether a plurality of diagnosis target devices are abnormal based on the sequential learning technique according to the first aspect of the present invention includes sensing data for the plurality of diagnosis target devices. obtaining; predicting whether abnormal diagnosis has occurred in at least one of the plurality of diagnosis target devices from an abnormal diagnosis prediction model that predicts and diagnoses abnormalities of the plurality of diagnosis target devices based on the sensing data; accumulatively counting the number of occurrences of the abnormal diagnosis by distinguishing the diagnosis target device in which the occurrence of the abnormal diagnosis is predicted by respective identification information; arranging the inspection order of the plurality of diagnosis target devices by applying the cumulative counted result value as a weight; and precisely diagnosing whether the plurality of diagnosis target devices are abnormal through a precise diagnosis model based on the sorted inspection order.

In some embodiments of the present invention, the abnormal diagnosis prediction model may include at least one of a noise diagnosis model, a vibration diagnosis model, and a device sensing data diagnosis model.

In some embodiments of the present invention, the step of accumulatively counting the number of occurrences of the abnormal diagnosis by distinguishing the diagnosis target device for which the occurrence of the abnormal diagnosis is predicted by each identification information may include: The number of occurrences of abnormal diagnosis from at least one of the noise diagnosis model, the vibration diagnosis model, and the device sensing data diagnosis model may be counted cumulatively by classifying them with respective identification information.

Some embodiments of the present disclosure may further include learning an abnormality diagnosis predictive model for predicting and diagnosing abnormalities of the plurality of diagnosis target devices based on sensing data in a normal state among the sensing data.

Some embodiments of the present invention may include: setting a precise diagnosis result through the precise diagnosis model as a correct answer label; and updating the learned abnormal diagnosis predictive model based on a semi-supervised algorithm using the precise diagnosis result set as the correct answer label.

In some embodiments of the present invention, the step of updating the abnormal diagnosis predictive model includes using the precise diagnosis result set as the correct label between the normal feature vector region and the abnormal feature vector region of the abnormal diagnosis predictive model through a hyperplane. It can be updated to improve the classification performance.

In some embodiments of the present invention, the precise diagnostic model may include at least one of an IR-based thermal diagnostic model, a visible light sensor-based color diagnostic model, and a light reflection diagnostic model for lidar sensing.

Some embodiments of the present disclosure may further include learning a precision diagnosis model for accurately diagnosing abnormalities of the plurality of diagnosis target devices based on sensing data in a normal state among the sensing data.

In addition to this, another method for implementing the present invention, another system, and a computer-readable recording medium for recording a computer program for executing the method may be further provided.

According to the present invention as described above, before performing the detailed diagnosis of a plurality of mechanical devices and facilities, the prediction process for abnormal diagnosis is performed in advance, and the most precise diagnosis is urgently required by utilizing the results of the prediction process. The inspection schedule can be automatically reconfigured so that the target device can be precisely diagnosed first.

In addition, it is possible to continuously improve the performance of the abnormal diagnosis predictive model by performing a precise diagnosis and using the precise diagnosis result as a correct answer label.

Through this, when a device that operates normally and a device with a possibility of abnormal operation are mixed with each other, an optimal inspection sequence can be automatically established, and the possibility of a defect can be predicted more quickly, maintaining efficient operation of the device and life extension can be expected.

Effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the following description.

1 is a flowchart of a method of diagnosing whether a plurality of diagnosis target devices are abnormal based on a sequential learning technique according to an embodiment of the present invention.
FIG. 2 is a view for explaining the contents of arranging inspection procedures for a plurality of diagnosis target devices; FIG.
3 is a view for explaining the contents of updating the abnormal diagnosis predictive model.
4 is a block diagram of an abnormality diagnosis system according to an embodiment of the present invention.
5 is a diagram for explaining a function of an abnormality diagnosis system according to an embodiment of the present invention.

Advantages and features of the present invention and methods of achieving them will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only the present embodiments allow the disclosure of the present invention to be complete, and those of ordinary skill in the art to which the present invention pertains. It is provided to fully understand the scope of the present invention to those skilled in the art, and the present invention is only defined by the scope of the claims.

The terminology used herein is for the purpose of describing the embodiments and is not intended to limit the present invention. In this specification, the singular also includes the plural unless specifically stated otherwise in the phrase. As used herein, “comprises” and/or “comprising” does not exclude the presence or addition of one or more other components in addition to the stated components. Like reference numerals refer to like elements throughout, and "and/or" includes each and every combination of one or more of the recited elements. Although "first", "second", etc. are used to describe various elements, these elements are not limited by these terms, of course. These terms are only used to distinguish one component from another. Accordingly, it goes without saying that the first component mentioned below may be the second component within the spirit of the present invention.

Unless otherwise defined, all terms (including technical and scientific terms) used herein will have the meaning commonly understood by those of ordinary skill in the art to which this invention belongs. In addition, terms defined in a commonly used dictionary are not to be interpreted ideally or excessively unless specifically defined explicitly.

Hereinafter, a method of diagnosing abnormalities in a plurality of diagnosis target devices based on a sequential learning technique performed by the abnormality diagnosis system 100 will be described with reference to FIGS. 1 to 3 .

1 is a flowchart of a method for diagnosing whether a plurality of diagnosis target devices are abnormal based on a sequential learning technique according to an embodiment of the present invention.

In the method according to an embodiment of the present invention, first, sensing data for a plurality of devices to be diagnosed is acquired ( S110 ).

In this case, the sensed data means data sensed by noise, vibration, thermal image, color image, light reflection image, and other environmental data of the device to be diagnosed. Such sensing data may be stored and managed in a separate database.

Next, it is predicted whether abnormal diagnosis has occurred in at least one of a plurality of diagnosis target devices from the abnormal diagnosis prediction model (S120).

According to an embodiment of the present invention, the abnormal diagnosis prediction model primarily predicts and diagnoses abnormalities of a plurality of diagnosis target devices based on sensing data.

In this case, the abnormal diagnosis prediction model may include at least one of a noise diagnosis model, a vibration diagnosis model, and a device sensing data diagnosis model. Through this, an embodiment of the present invention outputs an abnormal diagnosis result in some or all of the noise diagnosis model, the vibration diagnosis model, and the device sensing data diagnosis model based on the value input from each sensor when an abnormal phenomenon occurs in the device to be diagnosed. By doing so, the occurrence of anomalies can be predicted primarily.

Meanwhile, according to an embodiment of the present invention, an abnormal diagnosis predictive model for predicting and diagnosing abnormalities of a plurality of diagnosis target devices based on sensing data in a normal state among the acquired sensing data may be trained in advance.

That is, at least one of the noise diagnosis model, the vibration diagnosis model, and the device sensing data diagnosis model of the abnormal diagnosis prediction model may be trained through the sensing data in a normal state among the sensing data stored in the database. Here, the device sensing data diagnosis model may be learned by setting various sensing data according to the diagnosis target device, such as temperature and ambient temperature and humidity measurement values of the diagnosis target device, as input values.

For example, the prediction of whether abnormal diagnosis occurs in the abnormal diagnosis predictive model, that is, the abnormal diagnosis result, is determined when sensing data in an abnormal state is input to an abnormal diagnosis predictive model that has learned only sensing data in a normal state. It can be made possible by configuring the feature vector or the loss value of the model output in the diagnostic prediction model to have a large difference between the case in the normal state and the case in the abnormal state.

Next, the number of occurrences of abnormal diagnosis is counted cumulatively by distinguishing the diagnosis target device in which the occurrence of abnormal diagnosis is predicted by identification number (S130), and the cumulative counted result value is applied as a weight to be prioritized in the inspection order Thus, the inspection order for the plurality of diagnosis target devices is arranged (S140). In this case, according to an embodiment of the present invention, the inspection order for a plurality of diagnosis target devices may be arranged in a preset period or in real time.

The precise diagnosis model is a model for precisely diagnosing the abnormality of a plurality of diagnosis target devices. At least one of an IR-based thermal diagnosis model, a visible light sensor-based color diagnosis model, and a light reflection diagnosis model for lidar sensing may include

Such a precise diagnosis model may be pre-trained based on sensing data in a normal state among sensing data.

In one embodiment, the present invention accumulates the number of occurrences of abnormal diagnosis from at least one of a noise diagnosis model, a vibration diagnosis model, and a device sensing data diagnosis model by classifying a diagnosis target device for which abnormal diagnosis occurrence is predicted by identification information. can be counted

FIG. 2 is a view for explaining the contents of arranging inspection procedures for a plurality of diagnosis target devices; FIG.

As an example, for the diagnosis target device ID "#4-solar-inverter" whose inspection order is the third, the total number of counting times for which noise anomalies are predicted through the abnormal diagnosis prediction model is 3 times, and vibration abnormalities are predicted. The number of counting times is 1, and the total number of occurrences of abnormal diagnosis is 4 times.

In this case, an embodiment of the present invention sorts the diagnosis target devices based on the total number of occurrences of abnormal diagnosis, and as a result, the diagnosis target device ID '#4-solar-inverter', which has a lower order of inspection, is the third in the order. The order of inspection may be adjusted.

Meanwhile, the remark item in FIG. 2 indicates the current state of the device to be diagnosed.

For example, 'in operation' refers to a diagnosis target device currently undergoing detailed diagnosis, and 'paused' refers to a diagnosis target device designated so that detailed diagnosis is not temporarily performed. In addition, 'waiting' means a device to be diagnosed on standby according to the order adjusted for a detailed diagnosis in the future, and 'delete' means a device to be diagnosed so that detailed diagnosis is not performed by an administrator.

In the case of such precise diagnosis, the majority of cases proceed slowly over a long period of time as in the case of inspection of wind power generation described in the background of the present invention, and one embodiment of the present invention first predicts whether abnormal diagnosis occurs for a device to be diagnosed. and, through the process of accumulating the results and adjusting the detailed diagnosis check order, it is possible to automatically designate the highest priority that requires detailed diagnosis.

Thereafter, based on the sorted inspection sequence, it is precisely diagnosed whether a plurality of diagnosis target devices are abnormal through the precise diagnosis model ( S150 ).

In the step of precisely diagnosing whether the device for diagnosis is abnormal, similar to the step of predicting whether or not an abnormal diagnosis has occurred, an abnormal diagnosis result is outputted from some or all of the thermal diagnostic model, the color diagnostic model, and the light reflection diagnostic model. Finally, the occurrence of anomalies is diagnosed.

According to this process, predictions and diagnostic results from the abnormal diagnosis prediction model and the precise diagnosis model are stored in a database, respectively, and the diagnosis results from the precise diagnosis model may be transmitted to the central management server.

On the other hand, an embodiment of the present invention may additionally perform the process of updating the learned abnormal diagnosis predictive model.

That is, according to an embodiment of the present invention, anomaly diagnosis learned based on a semi-supervised algorithm using the precision diagnosis result set as the correct answer label and setting the precise diagnosis result through the precise diagnosis model as the correct answer label The predictive model can be updated.

3 is a view for explaining the contents of updating the abnormal diagnosis predictive model.

In one embodiment of the present invention, in order to increase the accuracy of the prediction result of the abnormal diagnosis prediction model, the diagnosis result through the precise diagnosis model may be used as a correct answer label.

In addition, when a large number of unlabeled data and some correct labels exist, semi-supervised learning techniques can be applied to significantly increase the prediction probability to improve the primary anomaly diagnosis prediction performance of the anomaly diagnosis prediction model.

Referring to FIG. 3 , when there is no correct answer label, the normal feature vector region and the abnormal feature vector region in the abnormal diagnosis predictive model learned only from the sensing data in the normal state are separated from each other due to an error in the hyperplane that separates them. The problem is that the distinction is loose.

In contrast, an embodiment of the present invention can improve the performance of distinguishing between the normal feature vector region and the abnormal feature vector region of the anomaly diagnosis predictive model through the hyperplane by using the precise diagnosis result set as the correct answer label.

That is, by continuously providing precise diagnosis results through the precise diagnosis model as correct answer labels to update the abnormal diagnosis prediction model, the hyperplane error is gradually improved by Δp, and accordingly, the normal feature vector region and the abnormal feature vector region are more clearly defined can be distinguished, the possibility of prediction errors of the abnormal diagnosis predictive model can be significantly reduced.

In addition, by continuously applying the precise diagnosis results to the abnormal diagnosis predictive model, there is an advantage that the learning performance of the abnormal diagnosis predictive model can be automatically improved over time.

In the above description, steps S110 to S150 may be further divided into additional steps or combined into fewer steps, according to an embodiment of the present invention. In addition, some steps may be omitted if necessary, and the order between steps may be changed. In addition, the contents described in FIGS. 1 to 3 are also applied to FIGS. 4 to 5 even if other contents are omitted.

Hereinafter, an apparatus 100 (hereinafter referred to as an abnormality diagnosis system) for diagnosing abnormalities in a plurality of diagnosis target devices based on a sequential learning technique according to an embodiment of the present invention will be described with reference to FIGS. 4 to 5 . do.

4 is a block diagram of an abnormality diagnosis system 100 according to an embodiment of the present invention. 5 is a diagram for explaining the function of the abnormality diagnosis system 100 according to an embodiment of the present invention.

The abnormality diagnosis system 100 according to an embodiment of the present invention includes a communication module 110 , a memory 120 , and a processor 130 .

The communication module 110 transmits/receives data to and from a plurality of diagnosis target devices. Also, the communication module 110 acquires sensing data for a plurality of diagnosis target devices from a sensor module that senses the plurality of diagnosis target devices. In this case, the sensed data includes noise, vibration, thermal image, color image, light reflection image, and other environmental data of the device to be diagnosed.

Meanwhile, the communication module 110 may include both a wired communication module and a wireless communication module. The wired communication module may be implemented as a power line communication device, a telephone line communication device, a cable home (MoCA), Ethernet, IEEE1294, an integrated wired home network, and an RS-485 control device. In addition, the wireless communication module may be implemented with wireless LAN (WLAN), Bluetooth, HDR WPAN, UWB, ZigBee, Impulse Radio, 60GHz WPAN, Binary-CDMA, wireless USB technology, wireless HDMI technology, and the like.

The memory 120 predicts whether abnormal diagnosis occurs based on the sensed data, and precisely diagnoses abnormalities of a plurality of diagnosis target devices according to the detailed diagnosis check order. In this case, the memory 120 collectively refers to a non-volatile storage device and a volatile storage device that continuously maintain stored information even when power is not supplied.

As the processor 130 executes the program stored in the memory 120 , it performs a function according to the functional element shown in FIG. 5 .

Referring to FIG. 5 , the abnormality diagnosis system 100 according to an embodiment of the present invention includes an abnormality diagnosis prediction model 210 , a precise diagnosis scheduling unit 220 , and a precise diagnosis model 230 .

The abnormal diagnosis prediction model 210 includes at least one of a noise diagnosis model 211, a vibration diagnosis model 212, and a device sensing data diagnosis model 213, and whether a plurality of diagnosis target devices are abnormal based on the sensing data. Predictive diagnosis should be made.

The precise diagnosis scheduling unit 220 counts the number of occurrences of abnormal diagnosis by classifying the diagnosis target device for which the occurrence of abnormal diagnosis is predicted by respective identification information, and applies the accumulated counted result value as a weight to a plurality of diagnosis target devices Sort the check order for

The precision diagnostic model 230 includes at least one of an IR-based thermal diagnostic model 231, a visible light sensor-based color diagnostic model 232, and a light reflection diagnostic model 233 for lidar sensing, Based on the inspection sequence arranged through the diagnosis scheduling unit 220 , it is precisely diagnosed whether a plurality of diagnosis target devices are abnormal.

The precise diagnosis result is provided to the central management server, and each data from the sensing data, the abnormal diagnosis prediction model 210 , the precise diagnosis scheduling unit 220 , and the precise diagnosis model 230 is stored in the database 240 . and managed.

The method for diagnosing whether a plurality of diagnosis target devices are abnormal based on the sequential learning technique according to the embodiment of the present invention described above is implemented as a program (or application) to be executed in combination with a server, which is hardware, and is implemented as a medium can be stored in

The above-described program is C, C++, JAVA, machine language, etc. that a processor (CPU) of the computer can read through a device interface of the computer in order for the computer to read the program and execute the methods implemented as a program It may include code (Code) coded in the computer language of Such code may include functional code related to a function defining functions necessary for executing the methods, etc., and includes an execution procedure related control code necessary for the processor of the computer to execute the functions according to a predetermined procedure. can do. In addition, the code may further include additional information necessary for the processor of the computer to execute the functions or code related to memory reference for which location (address address) in the internal or external memory of the computer to be referenced. have. In addition, when the processor of the computer needs to communicate with any other computer or server located remotely in order to execute the above functions, the code uses the communication module of the computer to determine how to communicate with any other computer or server remotely. It may further include a communication-related code for whether to communicate and what information or media to transmit and receive during communication.

The storage medium is not a medium that stores data for a short moment, such as a register, a cache, a memory, etc., but a medium that stores data semi-permanently and can be read by a device. Specifically, examples of the storage medium include, but are not limited to, ROM, RAM, CD-ROM, magnetic tape, floppy disk, and an optical data storage device. That is, the program may be stored in various recording media on various servers accessible by the computer or in various recording media on the computer of the user. In addition, the medium may be distributed in a computer system connected to a network, and a computer-readable code may be stored in a distributed manner.

The steps of a method or algorithm described in relation to an embodiment of the present invention may be implemented directly in hardware, as a software module executed by hardware, or by a combination thereof. A software module may contain random access memory (RAM), read only memory (ROM), erasable programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, hard disk, removable disk, CD-ROM, or It may reside in any type of computer-readable recording medium well known in the art to which the present invention pertains.

As mentioned above, although embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art to which the present invention pertains know that the present invention may be embodied in other specific forms without changing the technical spirit or essential features thereof. you will be able to understand Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive.

100: abnormality diagnosis device
110: sensor module
120: communication module
130: memory
140: processor
210: abnormal diagnosis prediction model
220: precise diagnosis scheduling unit
230: precision diagnostic model
240: database

Claims (1)

A method performed by a computer comprising:
acquiring sensing data for the plurality of diagnosis target devices;
predicting whether abnormal diagnosis has occurred in at least one of the plurality of diagnosis target devices from an abnormal diagnosis prediction model for predicting and diagnosing abnormalities of the plurality of diagnosis target devices based on the sensing data;
accumulatively counting the number of occurrences of the abnormal diagnosis by distinguishing the diagnosis target device for which the occurrence of the abnormal diagnosis is predicted with respective identification information;
arranging the inspection order of the plurality of diagnosis target devices by applying the cumulative counted result value as a weight; and
and precisely diagnosing whether the plurality of diagnosis target devices are abnormal through a precise diagnosis model based on the sorted inspection order.
A method of diagnosing whether a plurality of diagnostic target devices are abnormal based on a sequential learning technique.

Images