KR102419740B1 - Apparatus and method for diagnosis of system condition - Google Patents

Abstract

An apparatus and method for diagnosing a system state are disclosed. The system state diagnosis device includes an acoustic signal collector that collects acoustic signals generated by the system, a spectrogram image generator that processes the collected acoustic signals to generate a spectrogram image, and a pre-defined system that stores classification values for each symptom of the system. A diagnostic unit that obtains a classification value corresponding to the spectrogram image using a storage and machine learning-based diagnostic model, and determines the current state of the system based on a comparison result between the obtained classification value and the stored classification values for each symptom can

Description

Apparatus and method for diagnosis of system condition

The following embodiments relate to system state diagnosis technology.

In systems such as vacuum pumps used in manufacturing and research, maintenance and repair are required to efficiently use the system for a long period of time. As a method of improving the lifespan of equipment by performing maintenance before it is no longer usable due to system failure, there are concepts such as post-maintenance, preventive maintenance, condition-based maintenance, and predictive maintenance. The occurrence of system failure is difficult to predict, and if there is a standard that can detect failure in advance, time and economic loss can be reduced and equipment operation rate can be improved.

US 2017-0058797 B1 US 2018-0029543 A

According to an exemplary embodiment, an apparatus for diagnosing a system state includes an acoustic signal collector configured to collect an acoustic signal generated in a system; a spectrogram image generator for generating a spectrogram image by processing the collected sound signal; a storage for storing predefined classification values for each symptom of the system; and a diagnosis of obtaining a classification value corresponding to the spectrogram image using a machine learning-based diagnostic model, and determining the current state of the system based on a comparison result between the obtained classification value and the stored classification values for each symptom may include wealth.

The spectrogram image generator may obtain an MFCC transformed image by applying a mel frequency cepstral coefficient (MFCC) transform to the spectrogram image.

The MFCC converted image may be input to the diagnostic model.

The diagnosis unit may calculate a similarity between the obtained classification value and the stored classification value for each symptom, and determine a current state of the system from among predefined states based on the calculated similarity.

The diagnostic model may be a machine-learned model based on acoustic signals acquired when the system is in a normal state and acoustic signals acquired when the system is in a faulty state.

The spectrogram image generator may reconstruct the spectrogram image in log scale or mel scale.

The acoustic signal collection, the image generation, the storage, and the determination may be performed non-periodically.

The machine learning-based diagnostic model may be obtained by learning a spectrogram image generated by processing a normal acoustic signal of the system.

The machine learning-based diagnostic model may be generated using an image recognition model as a machine learning model.

The storage may store the degree of similarity, and the diagnosis unit may further determine the degree of obsolescence of the system based on data at an operation start time of the system and a trend of similarity change according to operation time.

A method for diagnosing a system state according to an embodiment may include collecting an acoustic signal generated in a system; generating a spectrogram image by processing the collected sound signal; storing classification values for each symptom of the system defined in advance; and obtaining a classification value corresponding to the spectrogram image using a machine learning-based diagnostic model, and determining the current state of the system based on a comparison result between the obtained classification value and the stored classification values for each symptom. may include.

The generating of the spectrogram image may include obtaining an MFCC transformed image by applying a mel frequency cepstral coefficient (MFCC) transform to the spectrogram image.

The MFCC converted image may be input to the diagnostic model.

The determining of the current state of the system may include calculating a similarity between the obtained classification value and the stored classification value for each symptom, and determining the current state of the system from among predefined states based on the calculated similarity. have.

The diagnostic model may be a machine-learned model based on acoustic signals acquired when the system is in a normal state and acoustic signals acquired when the system is in a faulty state.

The generating of the spectrogram image may include reconstructing the spectrogram image in a log scale or a mel scale.

The collecting of the sound signal, the generating of the image, the storing, and the determining may be performed aperiodically.

The machine learning-based diagnostic model may be obtained by learning a spectrogram image generated by processing a normal acoustic signal of the system.

The machine learning-based diagnostic model may be generated using an image recognition method as a machine learning model.

The storing may further include storing the degree of similarity, and the determining may further determine the degree of obsolescence of the system based on data at an operation start time of the system and a similarity change trend according to operation time.

1 is a diagram for describing an overall system of an apparatus for diagnosing a system state according to an exemplary embodiment.
2 is a block diagram illustrating a configuration of an apparatus for diagnosing a system state according to an exemplary embodiment.
3A is a flowchart illustrating a machine learning process of a machine learning-based diagnostic model of an apparatus for diagnosing a system state according to an exemplary embodiment.
3B is a flowchart illustrating a system state diagnosis process using a machine learning-based diagnosis model of the system state diagnosis apparatus according to an exemplary embodiment.
4A to 4E are diagrams for explaining an acoustic signal processing process of an apparatus for diagnosing a system state according to an exemplary embodiment.
5 is a flowchart illustrating a method for diagnosing a system state according to an exemplary embodiment.

Specific structural or functional descriptions of the embodiments are disclosed for purposes of illustration only, and may be changed and implemented in various forms. Accordingly, the actual implementation form is not limited to the specific embodiments disclosed, and the scope of the present specification includes changes, equivalents, or substitutes included in the technical spirit described in the embodiments.

Although terms such as first or second may be used to describe various elements, these terms should be interpreted only for the purpose of distinguishing one element from another. For example, a first component may be termed a second component, and similarly, a second component may also be termed a first component.

When a component is referred to as being “connected” to another component, it may be directly connected or connected to the other component, but it should be understood that another component may exist in between.

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 the described feature, number, step, operation, component, part, or combination thereof exists, and includes one or more other features or numbers, It should be understood that the possibility of the presence or addition of steps, operations, components, parts or combinations thereof is not precluded in advance.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in a commonly used dictionary 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 specification. does not

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. 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.

1 is a diagram for describing an overall system of an apparatus for diagnosing a system state according to an exemplary embodiment.

Referring to FIG. 1 , a system 110 to be diagnosed, a sensor 105 that receives an acoustic signal from the system 110 in a non-contact manner, and a system state diagnosis apparatus 100 that receives information received by the sensor 105 ) is shown.

In an embodiment, the system state diagnosis apparatus 100 may determine the state by collecting non-contact data using an acoustic signal generated by the system 110 and analyzing the machine learning method. If the system state diagnosis apparatus 100 is used, a failure of the system 110 essential for research and manufacturing may be detected in advance.

In one embodiment, system 110 may be a vacuum system, such as a vacuum pump. However, the present invention is not limited thereto, and the system state diagnosis apparatus 100 may be applied to other systems generating complex mechanical noise.

In an embodiment, the sensor 105 sensing an acoustic signal may collect an acoustic signal generated when the system 110 operates in a non-contact manner with the system 110 . The system state diagnosis apparatus 100 may receive an acoustic signal sensed from the sensor 105 . The received sound signal may be stored in the system state diagnosis apparatus 100 in the form of a sound file. The stored sound file can be used for signal processing by increasing the number of data by dividing it into several files of a unit time. For example, the unit time may be 1 second. The system state diagnosis apparatus 100 may remove noise from an acoustic signal and improve integrity through a preprocessing process.

In an embodiment, the system state diagnosis apparatus 100 may analyze an acoustic signal to extract a feature from the acoustic signal and generate a spectrogram image from the acoustic signal.

In an embodiment, the system state diagnosis apparatus 100 may convert a sound file into a spectrogram in the form of an image in order to diagnose the state of the system 110 using image recognition. In general, an acoustic signal for determining whether the system 110 has failed has a meaningful signal in a low frequency band. However, the image displayed in the form of a spectrogram may not represent the signal characteristics well. In an embodiment, the system state diagnosis apparatus 100 converts the spectrogram image to log-scale or mel-scale in order to better represent signal characteristics by different expressions of the low-frequency band and the high-frequency band. can be reconstructed. In an embodiment, the system state diagnosis apparatus 100 extracts a feature from the reconstructed image and expresses the feature value as an image by using a Mel Frequency Cepstral Coefficient (MFCC) transformation on the reconstructed image for use in machine learning and diagnosis. can

In an embodiment, the system state diagnosis apparatus 100 obtains a classification value corresponding to a spectrogram image using a machine learning-based diagnosis model, and compares the obtained classification value with predefined classification values for each symptom. Based on the , the current state of the system 110 may be determined.

In an embodiment, the machine learning-based diagnostic model may use a CNN algorithm for image recognition. In one embodiment, the machine learning-based diagnostic model may be a VGG16 model, and in another embodiment, the machine learning-based diagnostic model may be a Resnet50 model. However, the CNN used in the machine learning-based diagnostic model is not limited thereto, and various CNN algorithms may be used.

In an embodiment, the machine learning-based diagnostic model may be a machine-learning model based on acoustic signals acquired when the system 110 is in a normal state and acoustic signals acquired when the system is in a faulty state.

In one embodiment, the system state diagnosis apparatus 100 collects and processes a normal acoustic signal of the system 110 in a normal state, and obtains an MFCC transformed image in which the characteristic value of the acoustic signal is expressed as an image by using MFCC transformation. can The MFCC transformed image may be input to a machine learning-based diagnostic model. The machine learning-based diagnostic model may learn a feature for a steady state using MFCC transformed images obtained from a normal acoustic signal and generate a feature value. In an embodiment, the machine learning-based diagnostic model learns features for each symptom from MFCC-converted images corresponding to acoustic signals that appear as symptoms for each cause of failure of the system 110, and generates a classification value for each symptom. . The system state diagnosis apparatus 100 may include a storage, and may store and store the generated characteristic value of a normal state and a classification value for each symptom.

In an embodiment, the system state diagnosis apparatus 100 collects an acoustic signal from the system 110 in a current state to generate a spectrogram image and an MFCC converted image, and machine learning-based diagnosis in which features for a normal state are learned It can be determined whether the system 110 is normal by using the MFCC-converted image corresponding to the model and the current state. In an embodiment, the system state diagnosis apparatus 100 obtains a feature value corresponding to the spectrogram image of the current state of the system 110 , and based on a comparison result between the acquired feature value and the stored normal state feature value It may be determined whether the system 110 is normal. In an embodiment, the system state diagnosis apparatus 100 may calculate a similarity between the acquired feature value and the stored normal state feature value, and determine whether the system 110 is normal based on the calculated similarity.

In an embodiment, the system state diagnosis apparatus 100 collects acoustic signals from the system 110 in a current state, generates a spectrogram image and an MFCC converted image, and generates a machine learning-based diagnostic model in which characteristics for each symptom are learned. It can be used to determine the current state of the system 110 . The system state diagnosis apparatus 100 obtains a classification value corresponding to the spectrogram image of the current state of the system 110 , and based on a comparison result between the obtained classification value and the stored classification values for each symptom, the current state of the system 110 . status can be judged. In an embodiment, the system state diagnosis apparatus 100 calculates a similarity between the obtained classification value and the stored classification value for each symptom, and determines the current state of the system 110 from among predefined states based on the calculated similarity. can

In an embodiment, the system state diagnosis apparatus 100 may aperiodically collect, generate, store, and determine an acoustic signal. In an embodiment, the system state diagnosis apparatus 100 may monitor the state of the system 110 by continuously collecting, generating, storing, and determining an acoustic signal. By the system state diagnosis apparatus 100 monitoring the system 110 aperiodically or continuously, it is possible to identify the failure signs before the failure of the system 110, and to estimate the remaining life, and the system 110 no longer functions. It can be maintained and repaired before it becomes impossible.

In an embodiment, the system state diagnosis apparatus 100 collects data over time through monitoring, and the degree of similarity between the obtained classification value and the stored classification value for each symptom and the similarity between the obtained feature value and the stored normal state feature value can be saved. The system state diagnosis apparatus 100 may determine the degree of aging of the system based on the data at the start time of the system operation and the similarity change trend according to the operation time.

2 is a block diagram illustrating a configuration of an apparatus for diagnosing a system state according to an exemplary embodiment.

Referring to FIG. 2 , the system state diagnosis apparatus 100 according to an embodiment may include an acoustic signal collector 205 , a spectrogram image generator, a storage 215 , and a diagnosis unit 220 .

In one embodiment, the acoustic signal collector 205 may collect acoustic signals generated in the system. A sensor for sensing the acoustic signal may be included in the acoustic signal collector 205 or may be separately provided outside the acoustic signal collection.

In an embodiment, the spectrogram image generator 210 may generate a spectrogram image by processing the collected sound signal. The spectrogram image generator 210 may reconstruct the spectrogram image in log scale or mel scale so that the expression of the low-frequency region and the high-frequency region is different in order to express signal characteristics well. In an embodiment, the spectrogram image generator 210 may obtain an MFCC transformed image expressing characteristic values of an acoustic signal by applying a mel frequency cepstral coefficient (MFCC) transform to the spectrogram image. The obtained MFCC transformed image may be input to a machine learning-based diagnosis model of the system state diagnosis apparatus 100 .

In an embodiment, the storage 215 may store predefined classification values for each symptom of the system. In an embodiment, the storage 215 stores the characteristic values of the normal state generated by the machine learning-based diagnostic model learning the characteristics of the acoustic signal of the system in the normal state and the symptom-specific features of the acoustic signal of the system in the faulty state. can be learned and the generated classification value for each symptom can be stored.

In an embodiment, the diagnosis unit 220 obtains a feature value corresponding to the spectrogram image using a machine learning-based diagnostic model, and based on a comparison result between the acquired feature value and the stored normal-state feature value It is possible to determine whether the system is normal or not.

The diagnosis unit 220 may calculate a similarity between the acquired feature value and the stored feature value, and determine whether the system is normal based on the calculated similarity.

In an embodiment, the diagnosis unit 220 obtains a classification value corresponding to the spectrogram image by using a machine learning-based diagnosis model, and based on a comparison result between the obtained classification value and the stored classification values for each symptom, the system The current state can be determined.

The diagnosis unit 220 may calculate a similarity between the obtained classification value and the stored classification value for each symptom, and determine a current state of the system from among predefined states based on the calculated similarity.

3A is a flowchart illustrating a machine learning process of a machine learning-based diagnostic model of the system state diagnosis apparatus according to an embodiment, and FIG. 3B is a machine learning-based diagnostic model of the system state diagnosis apparatus according to an embodiment. It is a flowchart for explaining the system state diagnosis process used.

Referring to FIG. 3A , a process for learning the characteristics of an acoustic signal of a system in a normal state by a machine learning-based diagnostic model of the system state diagnosis apparatus is illustrated.

In operation 305 , the acoustic signal of the system in a normal state may be collected through the acoustic signal collector of the system state diagnosis apparatus. The received sound signal may be stored in the storage of the system state diagnosis apparatus in the form of a sound file. The stored sound file can be used for signal processing by increasing the number of data by dividing it into several files of a unit time.

In operation 310, the spectrogram image generator of the system state diagnosis apparatus may generate a spectrogram image by processing the received sound signal. The spectrogram image generator may reconstruct the spectrogram image in log scale or mel scale, and apply a mel frequency cepstral coefficient (MFCC) transform to the spectrogram image to obtain an MFCC transformed image. The obtained MFCC converted image may be stored in the storage of the system state diagnosis apparatus in step 315 .

In step 320, the obtained MFCC transformed image may be input to the diagnostic model for training. The diagnostic model may learn the characteristics of the normal state using MFCC transformed images obtained from the normal acoustic signal, and may generate a training file in step 325 . The training file may include feature values. The generated learning file may be stored in storage.

Although not shown in FIG. 3A , the diagnostic model may learn the characteristics of the acoustic signal for each failure symptom in the same manner as in FIG. 3A . In an embodiment, an acoustic signal of a system in a state in which an abnormal symptom appears may be collected through an acoustic signal collector of the system status diagnosis apparatus. The spectrogram image generator may generate a spectrogram image by processing the received sound signal. The spectrogram image generator may reconstruct the spectrogram image in log scale or mel scale, and apply a mel frequency cepstral coefficient (MFCC) transform to the spectrogram image to obtain an MFCC transformed image. The obtained MFCC transformed image may be input to a diagnostic model for training. The diagnostic model may learn characteristics for each symptom from MFCC-converted images corresponding to acoustic signals that appear as symptoms for each cause of failure of the system, and may generate a learning file including a classification value for each symptom. The generated learning file may be stored in storage.

In an embodiment, data for model training may be provided externally without going through steps 305 to 315 .

Referring to FIG. 3B , a process for diagnosing the state of a system using a learned machine learning-based diagnostic model is illustrated.

In operation 330 , an acoustic signal may be collected from the current system through the acoustic signal collector of the system state diagnosis apparatus. The received sound signal may be stored in the storage of the system state diagnosis apparatus in the form of a sound file. The stored sound file can be used for signal processing by increasing the number of data by dividing it into several files of a unit time.

In operation 335, the spectrogram image generator of the system state diagnosis apparatus may generate a spectrogram image by processing the received sound signal. The spectrogram image generator may reconstruct the spectrogram image in log scale or mel scale, and apply a mel frequency cepstral coefficient (MFCC) transform to the spectrogram image to obtain an MFCC transformed image. The obtained MFCC converted image may be stored in the storage of the system state diagnosis apparatus in step 340 .

In step 345 , the obtained MFCC transformed image may be input to a diagnostic model for diagnosing the state of the system. In step 345, the diagnosis unit of the system state diagnosis apparatus obtains a classification value from the MFCC converted image using the diagnosis model, and performs a comparison between the obtained classification value and the classification values for each symptom stored in the storage of the system state diagnosis apparatus. can

In operation 350, the diagnostic unit may calculate a similarity between the obtained classification value and the stored classification value for each symptom, and determine the current state of the system from among predefined states based on the calculated similarity.

When the diagnostic unit determines only whether the system is normal, obtains the feature value of the current state from the MFCC converted image using the diagnostic model, and compares the acquired feature value with the feature value stored in the storage of the system state diagnosis device, Whether the system is normal may be determined based on the similarity between the acquired feature value and the stored feature value.

4A to 4E are diagrams for explaining a sound signal processing process of an apparatus for diagnosing a system state according to an exemplary embodiment.

Referring to FIG. 4A , an acoustic signal collected through the acoustic signal collector of the system state diagnosis apparatus is shown in the form of an acoustic file, including strength and frequency information of the acoustic signal. In an embodiment, the sound file may be used for signal processing by increasing the number of data by dividing the sound file into several files of a unit time, and the unit time is 1 second in FIG. 4A .

In an embodiment, the spectrogram image generator of the system diagnosis apparatus may display the sound file in the form of a spectrogram to be used for image recognition. Referring to FIG. 4B , a spectrogram image generated from a sound file is shown. Since the low frequency band and the high frequency band appear linearly in the spectrogram, the signal characteristics of the low frequency band that occur in the event of a system failure are not well represented.

In an embodiment, the spectrogram image generator may reconstruct the spectrogram image in log-scale or mel-scale to have different representations of the low-frequency band and the high-frequency band. Referring to (A) of FIG. 4C, a log-spectrogram obtained by reconstructing a spectrogram image in a log scale is shown, and referring to FIG. 4C (B), the spectrogram image is reconstructed in a Mel scale. A mel-power spectrogram is shown.

In one embodiment, the spectrogram image generator facilitates feature extraction of an acoustic signal from the spectrogram image and applies a Mel Frequency Cepstral Coefficient (MFCC) transform to the spectrogram of FIG. 4C to input it into a machine learning-based diagnostic model. An MFCC-converted image can be obtained. Referring to FIG. 4D , an MFCC-converted image in which the spectrogram of FIG. 4C is MFCC-converted is shown.

In an embodiment, the MFCC transformed image may be input to a diagnostic model for training of a diagnostic model based on machine learning and diagnosing a system. Referring to FIG. 4E , a diagnostic model according to an exemplary embodiment is illustrated. A diagnostic model can be constructed using a CNN algorithm for image recognition. In the embodiment of Figure 4e, the diagnostic model is constructed using the VGG16 model. The VGG16 model can be applied to the diagnostic model by modifying only the classification layer through the transfer learning method for the classification layer among the feature extraction layer 405 and the classification layer 410 of the VGG16 model. In an embodiment, for cross-validation, training and testing may be performed by dividing the training data and the test data in a ratio of 8 to 2. The CNN algorithm that can be used for the diagnostic model is not limited thereto, and various image recognition CNN algorithms may be used.

5 is a flowchart illustrating a method for diagnosing a system state according to an exemplary embodiment.

Referring to FIG. 5 , the system state diagnosis apparatus may collect sound signals generated in the system in operation 505 . The sound signal may be stored in the system state diagnosis apparatus in the form of a sound file. The stored sound file can be used for signal processing by increasing the number of data by dividing it into several files of a unit time.

In operation 510, the system state diagnosis apparatus may generate a spectrogram image by processing the collected acoustic signal. The system state diagnosis apparatus may reconstruct the spectrogram image in log-scale or mel-scale in order to better represent signal characteristics by different expressions of the low-frequency band and the high-frequency band. The system state diagnosis apparatus extracts a feature from the reconstructed image and may express the feature value as an image by using a Mel Frequency Cepstral Coefficient (MFCC) transformation on the reconstructed image for use in machine learning and diagnosis.

In operation 515 , the system state diagnosis apparatus may store predefined classification values for each symptom of the system in the storage. In step 520, the system state diagnosis apparatus obtains a classification value corresponding to the spectrogram image using a machine learning-based diagnosis model, and based on the comparison result between the obtained classification value and the stored classification values for each symptom, the system The current state can be determined.

In an embodiment, the system state diagnosis apparatus may calculate a similarity between the obtained classification value and the stored classification value for each symptom, and determine the current state of the system from among predefined states based on the calculated similarity.

The embodiments described above may be implemented by a hardware component, a software component, and/or a combination of a hardware component and a software component. For example, the apparatus, methods and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate (FPGA). array), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions, may be implemented using a general purpose computer or special purpose computer. The processing device may execute an operating system (OS) and a software application running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For convenience of understanding, although one processing device is sometimes described as being used, one of ordinary skill in the art will recognize that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that may include For example, the processing device may include a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as parallel processors.

The 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 device, 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 over networked computer systems and stored or executed in a distributed manner. Software and data may be stored in a computer-readable recording medium.

The method according to the embodiment may be implemented in the form of program instructions 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, and the program instructions recorded on the medium are specially designed and configured for the embodiment or may be known and available to those skilled in the art of computer software. may be 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 a plurality of software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with reference to the limited drawings, a person skilled in the art may apply various technical modifications and variations based thereon. For example, the described techniques are performed in an order different from the described method, and/or the described components of the system, structure, apparatus, circuit, etc. are combined or combined in a different form than the described method, or other components Or substituted or substituted by equivalents may achieve an appropriate result.

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

Claims (12)

an acoustic signal collector for collecting acoustic signals generated by the system;
a spectrogram image generator that processes the collected sound signal to generate a spectrogram image;
a storage for storing predefined classification values for each symptom of the system; and
A diagnosis unit that obtains a classification value corresponding to the spectrogram image using a machine learning-based diagnostic model, and determines the current state of the system based on a comparison result between the obtained classification value and the stored classification values for each symptom
including,
The diagnostic model includes a classification layer trained to obtain a classification value corresponding to the spectrogram image using transfer learning,
The diagnostic unit,
calculating the similarity between the obtained classification value and the stored classification value for each symptom,
determining a current state of the system from among predefined states based on the calculated similarity,
A system state diagnosis apparatus for determining the degree of obsolescence of the system based on the data at the start time of the operation of the system and the similarity change trend according to the operation time. The method of claim 1,
The spectrogram image generating unit,
Obtaining an MFCC transformed image by applying a mel frequency cepstral coefficient (MFCC) transform to the spectrogram image,
and the MFCC converted image is input to the diagnostic model. According to claim 1,
The diagnostic unit,
and determining a current state of the system from among predefined states based on the calculated similarity. According to claim 1,
The diagnostic model is
An apparatus for diagnosing a system state, which is a machine-learned model based on an acoustic signal acquired when the system is in a normal state and acoustic signals acquired when the system is in a fault state. According to claim 1,
The spectrogram image generating unit,
A system state diagnosis apparatus for reconstructing the spectrogram image in log scale or mel scale. According to claim 1,
The acoustic signal collection, the image generation, the storage and the determination are performed aperiodically,
The machine learning-based diagnostic model is a machine learning model, which is generated by using an image recognition model. delete In the system state diagnosis method,
collecting acoustic signals generated by the system;
generating a spectrogram image by processing the collected sound signal;
storing classification values for each symptom of the system defined in advance; and
obtaining a classification value corresponding to the spectrogram image using a machine learning-based diagnostic model, and determining the current state of the system based on a comparison result between the obtained classification value and the stored classification values for each symptom
including,
The diagnostic model includes a classification layer trained to obtain a classification value corresponding to the spectrogram image using transfer learning,
The step of determining the current state of the system is
calculating a similarity between the obtained classification value and the stored classification value for each symptom; and
determining a current state of the system from among predefined states based on the calculated similarity
including,
The system state diagnosis method comprises:
Determining the degree of obsolescence of the system based on the data at the start time of the operation of the system and the similarity change trend according to the operating time
Further comprising, a system state diagnosis method. 9. The method of claim 8,
The generating of the spectrogram image comprises:
Obtaining an MFCC transformed image by applying a mel frequency cepstral coefficient (MFCC) transform to the spectrogram image,
The MFCC converted image is input to the diagnostic model, the system state diagnosis method. delete delete A computer program stored in a computer-readable recording medium in combination with hardware to execute the method of any one of claims 8 and 9.

Images