KR102418118B1 - Apparatus and method of deep learning-based facility diagnosis using frequency synthesis - Google Patents

Abstract

Disclosed are a method of diagnosing a facility based on deep learning and a system therefor. According to one embodiment of the present invention, an apparatus for diagnosing a facility based on deep learning may include: a signal generator generating a plurality of time series data signals by using a frequency signal acquired from the facility and performing signal synthesis for the plurality of time series data signals to convert them into time-domain signals by inverse Fourier transform (IFT); and a signal comparator receiving a pair of the time-domain signals to perform deep learning.

Description

DEEP LEARNING-BASED FACILITY DIAGNOSIS USING FREQUENCY SYNTHESIS

The present invention is a facility diagnosis technology using deep learning, which is a deep learning-based facility diagnosis device using frequency synthesis, which is used as a core technology that can automatically diagnose the status and failure of various manufacturing facilities using frequency synthesis. and methods.

In particular, the present invention is an essential technology for implementing a smart factory, which determines the state of equipment according to production operation conditions using frequency synthesis, predicts product quality according to equipment state, and optimizes production conditions. applied to technology.

The technology underlying the present invention is disclosed in the following documents.
1) Unexamined Patent Publication No. 10-2019-0114221 (2019.10.10.) “System and method for facility diagnosis based on deep learning”
2) Japanese Patent Publication No. 6801131 (2020.12.16.) "Diagnostic device, diagnostic method, and diagnostic program"
Artificial intelligence technology, which is represented by deep learning, is attracting attention from all over the society, including academia and industry, due to recent breakthroughs.

In the manufacturing industry, much effort is being made to commercialize artificial intelligence technology, which is one of the key technologies for smart factory implementation.

The smart factory is a more advanced form of factory automation, and includes technology that enables facility management using IoT, real-time diagnosis of the current state of equipment, and predicts failures to enable proactive measures.

In particular, equipment status and fault diagnosis are essential technologies for preventing mass defects, ensuring safety, stable operating conditions, and product quality. This technology is one of the large categories of PHM (Prognostics and Health management).

An automated system for diagnosing equipment failures may be configured as follows.

The sensor of the automation system consists of various sensors such as vibration, displacement, temperature, and ultrasonic waves, and collects signals that can indicate the state of the equipment. The transmission means of the automation system transmits the collected signals to the signal analysis PC in real time. The signal analysis PC processes the transmitted signals and uses deep learning technology to extract various conditions of equipment and diagnose failures. The signal analysis PC transmits the detected failure and status information to the data server, and the data server records the failure and status information in the data server together with equipment and product information produced by the equipment.

The existing fault diagnosis method determines whether the facility is normal or abnormal, based on the physical model of the facility.

However, the existing fault diagnosis method has a disadvantage in that the complexity of the facility increases and the operating state of the facility changes according to various environmental conditions, making it difficult to find a physical model of the facility.

Recently, many approaches to data analysis methods such as machine learning are being studied based on collected data rather than a physical model.

Accordingly, there is an urgent need to develop a more improved technology for performing fault diagnosis on equipment through learning data obtained by synthesizing and learning a plurality of frequency data collected from equipment.

An embodiment of the present invention aims to provide a deep learning-based equipment diagnosis apparatus and method using frequency synthesis, utilizing a high-precision frequency difference recognition technology using a signal generator.

In addition, an embodiment of the present invention aims to provide an innovative deep learning-based facility diagnosis apparatus and method that can be used without re-learning even when a new state is added.

In addition, an embodiment of the present invention aims to flexibly perform facility diagnosis without re-learning even if the number of state types of equipment to be diagnosed is changed.

In addition, an embodiment of the present invention aims to perform equipment diagnosis in a high-precision state capable of distinguishing various frequency differences due to a signal generator.

According to an embodiment of the present invention, a deep learning-based equipment diagnosis apparatus generates a plurality of time series data signals by using a frequency signal obtained from the equipment, synthesizes the plurality of time series data signals, and then performs inverse (IFT) a signal generator that transforms a time domain signal by Fourier Transform; and a signal comparator for deep learning by receiving the pair of time domain signals.

In addition, according to an embodiment of the present invention, a deep learning-based facility diagnosis method includes, in a signal generator, generating a plurality of time series data signals using a frequency signal obtained from the facility; synthesizing, in the signal generator, the plurality of time series data signals, and then converting them into time domain signals by IFT; and in the signal comparator, receiving the pair of time domain signals as input and performing deep learning.

According to an embodiment of the present invention, it is possible to provide a deep learning-based equipment diagnosis apparatus and method using frequency synthesis, utilizing a high-precision frequency difference recognition technology using a signal generator.

In addition, according to the present invention, even if a new state is added, it is possible to provide an innovative deep learning-based facility diagnosis apparatus and method that can be used without re-learning.

Also, according to the present invention, even if the number of state types of equipment to be diagnosed is changed, facility diagnosis can be flexibly performed without re-learning.

In addition, according to the present invention, it is possible to perform a high-precision equipment diagnosis capable of distinguishing various frequency differences due to the signal generator.

1 is a block diagram illustrating the configuration of a deep learning-based facility diagnosis apparatus according to an embodiment of the present invention.
2 is a diagram for explaining the structure and function of a signal generator according to the present invention.
3 is a diagram for explaining the structure and function of a signal comparator according to the present invention.
4 is a diagram illustrating a H/W configuration of a deep learning-based facility diagnosis apparatus according to the present invention.
5 is a flowchart illustrating a deep learning-based facility diagnosis method according to an embodiment of the present invention.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, since various changes may be made to the embodiments, the scope of the patent application is not limited or limited by these embodiments. It should be understood that all modifications, equivalents and substitutes for the embodiments are included in the scope of the rights.

The terms used in the examples are used for the purpose of description only, and should not be construed as limiting. The singular expression includes the plural expression unless the context clearly dictates otherwise. In this specification, terms such as "comprise" or "have" are intended to designate that a feature, number, step, operation, component, part, or a combination thereof described in the specification exists, but one or more other features It should be understood that this does not preclude the existence or addition of numbers, steps, operations, components, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiment belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present application. does not

In addition, in the description with reference to the accompanying drawings, the same components are assigned the same reference numerals regardless of the reference numerals, and the overlapping description thereof will be omitted. In describing the embodiment, if it is determined that a detailed description of a related known technology may unnecessarily obscure the gist of the embodiment, the detailed description thereof will be omitted.

1 is a block diagram illustrating the configuration of a deep learning-based facility diagnosis apparatus according to an embodiment of the present invention.

Referring to FIG. 1 , an apparatus 100 for diagnosing a deep learning-based facility according to an embodiment of the present invention may include a signal generator 110 and a signal comparator 120 . In addition, the deep learning-based facility diagnosis apparatus 100 may optionally include the diagnosis device 130 according to an embodiment.

The signal generator 110 generates a plurality of time series data signals by using the frequency signals obtained from the equipment. That is, the signal generator 110 receives a frequency signal from an arbitrary facility that becomes a standard, and selects and combines parameters constituting the input frequency signal in various ways to create a plurality of time series data signals. can

Here, the parameter may be the number of frequencies, a type of frequency, and a size for each frequency constituting the frequency signal.

In an embodiment, the signal generator 110 may generate the plurality of time series data signals by randomly selecting each of the number of frequencies, the type of frequency, and the size for each frequency with respect to the frequency signal.

That is, the signal generator 110 may generate a plurality of time-series data signals by time-series selecting the number of frequencies n, the frequency type f, and the size a for each frequency according to the number of cases.

For example, the signal generator 110 may generate a time series data signal a1f1 by selecting and combining the frequency type f1 and the size a1 for each frequency in the first frequency, and in the second frequency, the frequency type f2 and the size a2 by frequency can be selected and combined to generate the following time series data signal a2f2, and this can be repeated up to n times to generate a plurality of time series data signals.

In addition, the signal generator 110 synthesizes the plurality of time series data signals and then converts them into a time domain signal by Inverse Fourier Transform (IFT). That is, the signal generator 110 may play a role of calculating a specific signal using a frequency of a wavelength by sequentially synthesizing a time series data signal and then performing an inverse Fourier transform.

Inverse Fourier Transform (IFT) is a transform that is opposite to the Fourier transform, and may be a method of synthesizing a waveform in the time domain from the wave amplitude and phase of each frequency component.

The signal generator 110 may convert a frequency signal to generate a time domain signal.

The signal comparator 120 receives the pair of time domain signals and performs deep learning. That is, the signal comparator 120 may serve to compare two signal domain signals and determine whether the two signal domain signals are the same or different through learning.

The signal comparator 120 may receive a pair of time domain signals.

According to an embodiment, the signal comparator 120 may include a feature extractor and a classifier.

The feature extractor may receive the pair of time-domain signals and extract features for each pair of the time-domain signals. That is, the feature extractor may select features for each time domain signal.

The classifier may analyze the feature difference between the extracted features to perform deep learning whether the pair of time domain signals is the same. That is, the classifier can learn whether the pairs of time-domain signals are identical to each other by determining the degree of difference between features of the pairs of time-domain signals.

For example, if the feature difference is within a threshold range and the pair of time domain signals are the same, the classifier may converge the deep learning result to a predetermined value (eg, '1') and output it. On the other hand, the classifier can output the deep learning result as '0' by determining that the pair of time domain signals are different from each other when the feature difference is large outside the threshold range.

In one embodiment, the deep learning-based equipment diagnosis apparatus 100 enables the diagnosis of the state of the equipment to be diagnosed by mounting the signal comparator 120 on which deep learning has been completed to the equipment to be diagnosed.

To this end, the deep learning-based facility diagnosis apparatus 100 may further include a diagnosis device 130 .

The diagnostic device 130 counts the number N of states for the equipment to be diagnosed (where N is a natural number), and the signal comparator 120 on which the deep learning has been completed is mounted on the equipment to be diagnosed as many as the counted N pieces. , the diagnostic target facility can be diagnosed. That is, the diagnostic device 130 may install the deep-learning signal comparator 120 as much as the number of states expressed in the diagnostic target facility to estimate the current state of the diagnostic target facility.

In the diagnosis of the equipment to be diagnosed, the diagnostic device 130 is designed to be activated only for a state unique to each of the mounted signal comparators 120 , and the designed state of the activated signal comparator is converted to the current state of the equipment to be diagnosed. can be diagnosed

Specifically, the diagnostic device 130 may input n reference signals representing each state to each of the N signal comparators, and set each of the N signal comparators to a specific state. That is, the diagnostic device 130 may enable the corresponding signal comparator to be activated only for one specific state by inputting a unique reference signal representing a specific state to any one of a plurality of mounted signal comparators.

Also, the diagnostic device 130 may input the input signal collected from the diagnostic target facility to each of the N signal comparators. That is, the diagnostic device 130 may input the input signal acquired in real time from the diagnostic target facility to a plurality of mounted signal comparators.

Also, the diagnostic device 130 may identify a signal comparator that outputs the largest value among the output values of the N signal comparators. That is, the diagnostic device 130 may determine the signal comparator having the largest output value by comparing the output value according to the difference in characteristics between the input reference signal and the input signal for each N signal comparators.

In other words, since the reference signal and the input signal have similar state values, the diagnostic device 130 may identify a signal comparator that emits an output value closer to a predetermined value (eg, '1').

Also, the diagnostic device 130 may diagnose the state set in the identified signal comparator as the current state of the equipment to be diagnosed. That is, the diagnostic device 130 may check the state designed by the activated and identified signal comparator as the current state of the equipment to be diagnosed.

According to an embodiment of the present invention, it is possible to provide a deep learning-based equipment diagnosis apparatus and method using frequency synthesis, utilizing a high-precision frequency difference recognition technology using a signal generator.

In addition, according to the present invention, even if a new state is added, it is possible to provide an innovative deep learning-based facility diagnosis apparatus and method that can be used without re-learning.

Also, according to the present invention, even if the number of state types of equipment to be diagnosed is changed, facility diagnosis can be flexibly performed without re-learning.

In addition, according to the present invention, it is possible to perform a high-precision equipment diagnosis capable of distinguishing various frequency differences due to the signal generator.

In the present invention, by analyzing time series data such as vibration signals using deep learning, a new smart facility diagnosis system for diagnosing the condition of equipment is proposed.

In general, in facility diagnosis using deep learning, the size of the deep learning model output value is determined according to the type to be diagnosed.

For example, if a specific facility has 0~9 states, the output of the deep learning model should be 10 to determine the states 0~9.

If the state of equipment is newly added to the deep learning model trained with 10 outputs, the deep learning model should change the structure of the model to 11 outputs and relearn so that 11 outputs come out again.

Due to these technical limitations, deep learning-based facility diagnostic devices have great limitations in scalability and performance.

The present invention proposes an innovative deep learning-based facility diagnosis device that can be used without re-learning even when the state of the facility is newly added.

The deep learning-based facility diagnosis apparatus 100 of the present invention may be configured to include a signal generator and a signal comparator in a learning process.

When a specific frequency is input, the signal generator functions to generate a time series data signal having the input specific frequency.

The signal generator receives a plurality of specific frequencies according to a frequency type f, the number of frequencies n, and a size a for each type, synthesizes the signals, and then generates a signal in the time domain through Inverse Fourier Transform (IFT).

2 is a diagram for explaining the structure and function of a signal generator according to the present invention.

As shown in FIG. 2 , the signal generator may randomly receive a plurality of time series data signals having a frequency type f, the number n of frequencies, and a size a for each frequency type. In addition, the signal generator may generate a time domain signal by synthesizing a plurality of inputted time series data signals and then performing IFT transformation.

The time domain signal becomes the input of the signal comparator configured in deep learning.

3 is a diagram for explaining the structure and function of a signal comparator according to the present invention.

A signal comparator has two inputs, and an output indicates whether the two signals are equal or different. The output of the signal comparator may converge to a predetermined value (eg, '1') if the two signals are the same.

The deep learning structure of the signal comparator is a structure including a feature extractor and a classifier.

The feature extractor is applied equally to the two inputs, so it is possible to extract the feature difference between the two signals.

The classifier is located at the rear end of the feature extractor, and can classify whether the signal is the same or a different signal by receiving the feature difference between the two extracted signals.

3 , the signal comparator may extract features 1 and 2 by the feature extractor from the input signals 1 and 2, respectively.

The signal comparator may input the feature difference between feature 1 and feature 2 into the classifier.

The classifier may analyze the input feature difference, check the similarity between the signal 1 and the signal 2, converge to a predetermined value if the signal is the same, and output the same.

Various time domain signals from the signal generator are input to the signal comparator and learned.

After learning, the signal comparator is disconnected from the signal generator and configured one by one for each type of equipment condition to be diagnosed.

For example, if the state of the specific equipment is 10, the signal comparators may be composed of 10, and if the state of the specific equipment is 5, the signal comparators may be composed of 5 units.

In the diagnosis process, if the equipment state is N, the deep learning-based equipment diagnosis apparatus 100 inputs one reference signal for each N state to each of the N signal comparators. In addition, the deep learning-based facility diagnosis apparatus 100 inputs a frequency signal related to an actual state obtained from the facility to each of the N signal comparators.

In each signal comparator, a reference signal and a frequency signal are used for comparison, an output value is output, and the state of the equipment is determined using the largest value among the output values.

4 is a diagram illustrating a H/W configuration of a deep learning-based facility diagnosis apparatus according to the present invention.

As shown in FIG. 4 , for a facility having N states, the deep learning-based facility diagnosis apparatus 100 may be equipped with N signal comparators for which learning has been completed, and input reference signals for each of the N signal comparators. can do.

Here, the reference signal may be a signal that sets the corresponding signal comparator to diagnose only a specific state.

That is, the deep learning-based facility diagnosis apparatus 100 may be equipped with N signal comparators to which reference signals for each state are input so that the N states can be independently diagnosed.

Thereafter, the deep learning-based facility diagnosis apparatus 100 inputs an input signal to N signal comparators, and outputs the largest value (a meaningful value, for example, 1) among the output values S1 to Sn from the N signal comparators. The state assigned to the signal comparator can be diagnosed as the current state of the equipment.

If the proposed structure is used, the diagnosis system can be configured flexibly without re-learning even if the number of state types of equipment to be diagnosed changes, and a high-precision state diagnosis system that can distinguish various frequency differences due to the signal generator is possible. .

Hereinafter, a work flow of the deep learning-based facility diagnosis apparatus 100 according to embodiments of the present invention will be described in detail with reference to FIG. 5 .

5 is a flowchart illustrating a deep learning-based facility diagnosis method according to an embodiment of the present invention.

The deep learning-based equipment diagnosis method according to the present embodiment may be performed by the deep learning-based equipment diagnosis apparatus 100 .

In the signal generator 110 of the deep learning-based equipment diagnosis apparatus 100, a plurality of time series data signals are generated using the frequency signal obtained from the equipment (510). Step 510 may be a process of generating a plurality of time series data signals by receiving a frequency signal from an arbitrary standard equipment and selecting and combining parameters constituting the received frequency signal in various ways.

Here, the parameter may be the number of frequencies, a type of frequency, and a size for each frequency constituting the frequency signal.

In an embodiment, the signal generator may generate the plurality of time series data signals by randomly selecting each of the number of frequencies, types of frequencies, and magnitudes for each frequency with respect to the frequency signal.

That is, the signal generator may generate a plurality of time-series data signals by time-series selecting the number of frequencies n, the frequency type f, and the size a for each frequency according to the number of cases.

For example, the signal generator can generate a time series data signal a1f1 by selecting and combining the frequency type f1 and the size a1 for each frequency in the first frequency, and in the second frequency, select and combine the frequency type f2 and the size a2 for each frequency Thus, the following time series data signal a2f2 may be generated, and this may be repeated up to n times to generate a plurality of time series data signals.

In addition, the signal generator of the deep learning-based facility diagnosis apparatus 100 synthesizes the plurality of time series data signals and then converts them into time domain signals by Inverse Fourier Transform (IFT) ( 520 ). Step 520 may be a process of calculating a specific signal using a frequency of a wavelength by sequentially synthesizing a time series data signal and performing inverse Fourier transform.

Inverse Fourier Transform (IFT) is a transform that is opposite to the Fourier transform, and may be a method of synthesizing a waveform in the time domain from the wave amplitude and phase of each frequency component.

A signal generator may convert a frequency signal to generate a time domain signal.

Subsequently, the signal comparator of the deep learning-based facility diagnosis apparatus 100 receives the pair of time domain signals and performs deep learning ( 530 ). Step 530 may be a process of comparing two signal domain signals and determining whether the two signal domain signals are the same or different through learning.

The signal comparator may receive a pair of time domain signals.

The signal comparator may receive the pair of time domain signals through a feature extractor, and extract features of each pair of time domain signals. That is, the feature extractor may select features for each time domain signal.

In addition, the signal comparator may analyze the feature difference between the extracted features through the classifier to perform deep learning whether the pair of the time domain signals is the same. That is, the classifier can learn whether the pairs of time-domain signals are identical to each other by determining the degree of difference between features of the pairs of time-domain signals.

For example, if the feature difference is within a threshold range and the pair of time domain signals are the same, the classifier may converge the deep learning result to a predetermined value (eg, '1') and output it. On the other hand, the classifier can output the deep learning result as '0' by determining that the pair of time domain signals are different from each other when the feature difference is large outside the threshold range.

In one embodiment, the deep learning-based equipment diagnosis apparatus 100 enables the diagnosis of the state of the equipment to be diagnosed by mounting the signal comparator 120 on which deep learning has been completed to the equipment to be diagnosed.

To this end, the diagnostic device of the deep learning-based facility diagnosis apparatus 100 counts the number N of states for the equipment to be diagnosed (where N is a natural number), and uses the deep learning-completed signal comparator as many as the counted N; It may be mounted on the diagnostic target facility to diagnose the diagnostic target facility. That is, the diagnostic device may be equipped with a deep-learning signal comparator as much as the number of states expressed in the diagnostic target facility to estimate the current state of the diagnostic target facility.

In the diagnosis of the equipment to be diagnosed, the diagnostic device may be designed to be activated only for a state unique to each of the mounted signal comparators, and the designed state of the activated signal comparator may be diagnosed as the current state of the equipment to be diagnosed.

Specifically, the diagnostic device may input n reference signals representing each state to each of the N signal comparators to set each of the N signal comparators to a specific state. That is, by inputting a unique reference signal representing a specific state to any one of a plurality of mounted signal comparators, the diagnostic device can activate the corresponding signal comparator only for one specific state.

Also, the diagnostic device may input the input signal collected from the diagnostic target facility to each of the N signal comparators. That is, the diagnostic device may input the input signals acquired in real time from the diagnostic target facility to a plurality of mounted signal comparators.

Also, the diagnostic device may identify a signal comparator that outputs the largest value among the output values of the N signal comparators. That is, the diagnostic device may determine the signal comparator having the largest output value by comparing the output value according to the difference in characteristics between the input reference signal and the input signal for each N signal comparators.

In other words, since the reference signal and the input signal have similar state values, the diagnostic device may identify a signal comparator that emits an output value closer to a predetermined value (eg, '1').

In addition, the diagnostic device may diagnose the state set in the identified signal comparator as the current state of the equipment to be diagnosed. That is, the diagnostic device may check the activated state designed by the identified signal comparator as the current state of the equipment to be diagnosed.

According to an embodiment of the present invention, it is possible to provide a deep learning-based equipment diagnosis apparatus and method using frequency synthesis, utilizing a high-precision frequency difference recognition technology using a signal generator.

In addition, according to the present invention, even if a new state is added, it is possible to provide an innovative deep learning-based facility diagnosis apparatus and method that can be used without re-learning.

Also, according to the present invention, even if the number of state types of equipment to be diagnosed is changed, facility diagnosis can be flexibly performed without re-learning.

In addition, according to the present invention, it is possible to perform a high-precision equipment diagnosis capable of distinguishing various frequency differences due to the signal generator.

The deep learning-based facility diagnosis method according to the embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Software may comprise a computer program, code, instructions, or a combination of one or more thereof, which configures a processing device to operate as desired or is independently or collectively processed You can command the device. The software and/or data may be any kind of machine, component, physical device, virtual equipment, computer storage medium or apparatus, to be interpreted by or to provide instructions or data to the processing device. , or may be permanently or temporarily embody in a transmitted signal wave. The software may be distributed on a networked computer system and stored or executed as a distributed deep learning-based facility diagnostic method. Software and data may be stored in one or more computer-readable recording media.

As described above, although the embodiments have been described with reference to the limited drawings, those skilled in the art may apply various technical modifications and variations based on the above. For example, the described techniques are performed in an order different from the described deep learning-based equipment diagnostic method, and/or the described components of a system, structure, device, circuit, etc. are different from the described deep learning-based equipment diagnostic method Appropriate results may be achieved even if they are combined or combined in a form, or replaced or substituted by other elements or equivalents.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

100: Deep learning-based facility diagnosis device
110: signal generator 120: signal comparator
130: diagnostic device

Claims (10)

a signal generator that generates a plurality of time series data signals using a frequency signal obtained from a facility, synthesizes the plurality of time series data signals, and then converts them into time domain signals by Inverse Fourier Transform (IFT);
a signal comparator for deep learning by receiving the pair of time domain signals; and
Counting the number N of states for the equipment to be diagnosed (where N is a natural number), and mounting the signal comparator on which the deep learning has been completed, as many as the counted N pieces, to the equipment to diagnose the equipment to be diagnosed diagnostic device
including,
The diagnostic device,
inputting n reference signals representing each state to each of the N signal comparators to set each of the N signal comparators to a specific state;
inputting the input signal collected from the diagnostic target facility to each of the N signal comparators;
Identifies a signal comparator that outputs the largest value among the output values of the N signal comparators,
Diagnosing the state set in the identified signal comparator as the current state of the equipment to be diagnosed
Deep learning-based facility diagnosis device. According to claim 1,
The signal generator is
generating the plurality of time series data signals by randomly selecting each of the number of frequencies, the type of frequency, and the size for each frequency with respect to the frequency signal
Deep learning-based facility diagnosis device. According to claim 1,
The signal comparator is
a feature extractor that receives the pair of time domain signals and extracts features for each pair of time domain signals; and
A classifier that analyzes the feature difference between the extracted features to perform deep learning whether the pair of time domain signals are the same, and if the pairs of time domain signals are the same, converges the deep learning result to a predetermined value and outputs it
A deep learning-based facility diagnosis device comprising a. delete delete generating, in a signal generator, a plurality of time series data signals using a frequency signal obtained from a facility;
synthesizing, in the signal generator, the plurality of time series data signals, and then converting them into time domain signals by IFT;
deep learning by receiving the pair of time domain signals in a signal comparator;
counting, in a diagnostic device, the number of states N (where N is a natural number) for the equipment to be diagnosed; and
Diagnosing the equipment to be diagnosed by mounting, in the diagnostic device, the number of N counted signal comparators on the equipment to be diagnosed
including,
The step of diagnosing the equipment to be diagnosed includes:
setting each of the N signal comparators to a specific state by inputting n reference signals representing each state to each of the N signal comparators;
inputting the input signal collected from the diagnostic target facility to each of the N signal comparators;
identifying a signal comparator that outputs the largest value among the output values of the N signal comparators; and
Diagnosing, in the diagnostic device, the state set in the identified signal comparator as the current state of the equipment to be diagnosed
A deep learning-based facility diagnosis method comprising a. 7. The method of claim 6,
The generating of the plurality of time series data signals comprises:
generating the plurality of time series data signals by randomly selecting each of the number of frequencies, the type of frequency, and the size for each frequency with respect to the frequency signal
A deep learning-based facility diagnosis method comprising a. 7. The method of claim 6,
The deep learning step is
receiving the pair of time-domain signals and extracting features of each pair of the time-domain signals; and
A step of deep learning whether the pair of time domain signals is the same by analyzing the feature difference between the extracted features, and if the pair of time domain signals are the same, converging the deep learning result to a predetermined value and outputting it
A deep learning-based facility diagnosis method comprising a. delete delete

Images