KR20210067108A - Fault diagnosis apparatus and method using frequency analysis based on machine-learning - Google Patents

Abstract

The present invention relates to a failure diagnosis technology and, in particular, to a failure diagnosis device for analyzing features and frequency features of a sensor signal according to the time change based on machine learning and efficiently diagnosing a state of complex equipment that features change according to the time, and a method thereof. To this end, the failure diagnosis device using machine learning-based frequency analysis according to the present invention includes: an input vector generation unit which receives a sensing signal for equipment and generates a signal input vector; a signal conversion unit which frequency-converts the sensing signal to generate a frequency-converted vector; and an analysis unit which receives the signal input vector and the frequency conversion vector, analyzes the vector through a deep learning-based model, and outputs a state diagnosis result for the equipment.

Description

Fault diagnosis apparatus and method using frequency analysis based on machine-learning

The present invention relates to a failure diagnosis technology, and in particular, a machine learning-based analysis that analyzes the characteristics and frequency characteristics of a sensor signal according to time change based on machine learning to effectively diagnose the state of complex equipment whose characteristics change over time. It relates to a failure diagnosis apparatus and method using frequency analysis.

Recently, interest in artificial intelligence has increased significantly due to the recent confrontation between AlphaGo and Lee Sedol. In particular, academia and industry research on deep learning, known as AlphaGo's core technology, has exploded.

Deep learning improves the performance of many known problems of artificial neural networks (ie, vanishing problem, overfitting, etc.) by developing an activation function (ReLU) and improving algorithms such as drop-out. Also, thanks to the development of hardware such as GPU (Graphic Processing Units) and the power of big data that can learn complex structures, it is showing excellent performance in various fields recently.

These deep learning technologies are being developed rapidly by many overseas companies (Google, Facebook, Apple, Microsoft, Alibaba, Baidu) and are being applied to fields such as face recognition, voice recognition, natural language processing, search service, and medical care. It is urgent to secure the latest technology of deep learning, which is developing rapidly, and further preempt the application fields and rapidly commercialize it.

A defect classification method used in a conventional defect inspection equipment is shown in FIG. 1 .

An algorithm developer extracts features that are likely to be classified well from an image with an image processing algorithm, and then learns these features with a classifier (SVM, Decision Tree).

An optical device is constructed using lighting and a camera, and the S/N ratio of the defective area is increased by imaging the change in the amount of light entering the camera as the light path changes in the defective area. In such an image, a defect detection algorithm detects a defect candidate, and the defect is detected and classified using a feature extraction algorithm and a classification algorithm for the defect candidate image.

The performance limit of this method is how well a person can design a feature extraction algorithm to extract features. In addition, there is a problem in that classification performance is limited according to selected features, and classification performance varies according to rotation of an image, change in brightness, change in size, and the like. Since each product has different characteristics of the image, it is analyzed and developed, but there is a disadvantage that it takes a lot of time.

Among the latest technologies, deep learning, a classification algorithm applying the CNN method is a method in which artificial intelligence extracts and learns features from an image by itself. The above problem can be solved by constructing a defect classification algorithm by applying deep learning technology. Among the various deep learning structures, deep learning in the image field uses a structure called Convolutional Neural Network (CNN).

Image classification using CNN has the feature that CNN itself extracts and learns features that can improve classification performance. If this feature is applied to the image-based defect detection field, a dramatic performance improvement can be expected.

The configuration of a general classifier using deep learning is shown in FIG. 2 .

After applying a convolution layer and an activation function to the input image, the feature map is reduced in size through a pooling layer and transferred to the next convolution layer. do. By repeatedly and deeply stacking these basic structures, it is possible to effectively extract features for fault classification or fault diagnosis from images.

On the other hand, the smart factory includes technologies that are an evolved form of factory automation, that is, facility management using IoT, real-time diagnosis of the current state of the facility, and technology that can predict failures to take preemptive measures.

In particular, equipment status and fault diagnosis are essential technologies to prevent mass defects, ensure safety, stable operating conditions, and product quality. This technology is one of the large areas of failure prediction and health management (PHM).

In general, the configuration of an automated system for equipment failure diagnosis is as follows.

It collects signals that can indicate the status of equipment from various sensors such as vibration, displacement, temperature, and ultrasonic waves. These signals are transmitted in real time to the signal analysis PC. Signal analysis PC uses signal processing and deep learning technology to extract various statuses of equipment and diagnose failures. The detected fault and status information is sent to the data server, and this information is recorded in the data server together with product information produced by the equipment.

The conventional fault diagnosis method determines whether it is normal or abnormal based on the physical model of the facility. However, as the complexity of the facility increases and the operating state of the facility changes according to various environmental conditions, it is difficult to find a physical model accordingly. Recently, many approaches according to data analysis methods such as machine learning have been studied based on collected data rather than a physical model.

Specifically, it senses signals such as vibration, displacement, temperature, and ultrasound generated from an object, and based on these sensed values, time series analysis through linear prediction coefficients, frequency analysis through fast Fourier transform, and discrete analysis for each frequency band are performed. After analyzing the values and variances, a method in which the failure characteristics are extracted was adopted by classifying the results of testing and validation of these data through a multi-layer perceptron network.

However, the conventional method has a problem in that there is a limitation in diagnosing equipment conditions and failures for complex equipment whose characteristics are variously changed according to time because the conventional method extracts the failure characteristics using only the sensing signal according to time change.

Korea Patent Publication No. 2017-0173034

The present invention has been devised to solve the above problems, and an object of the present invention is to improve the state diagnosis performance of complex equipment whose characteristics are variously changed over time.

To this end, the apparatus for diagnosing a failure using machine learning-based frequency analysis according to the present invention includes an input vector generator that receives a sensing signal for a facility and generates a signal input vector, and frequency-converts the sensing signal to generate a frequency conversion vector and an analysis unit that receives the signal input vector and frequency conversion vector, analyzes it through a deep learning-based model, and outputs a state diagnosis result for the facility.

In addition, the failure diagnosis apparatus using machine learning-based frequency analysis according to the present invention includes an input vector generator that receives a sensing signal for a facility and generates a one-dimensional signal input vector, and frequency-converts the sensing signal to generate a frequency conversion vector. A signal converter for generating, a feature extraction module for extracting a signal feature vector from the one-dimensional signal input vector, a combiner that combines the signal feature vector and the frequency transform vector, and analyzes the combined vector through the combiner It includes a classification module that outputs a failure diagnosis result of the equipment.

In addition, the failure diagnosis apparatus using machine learning-based frequency analysis according to the present invention includes an input vector generator that receives a sensing signal for a facility and generates a one-dimensional signal input vector, and frequency-converts the sensing signal to generate a frequency conversion vector. a signal converter generating a signal; a first feature extraction module for extracting a signal feature vector from the one-dimensional signal input vector; a second feature extraction module for extracting a frequency feature vector from the frequency transform vector; A combiner for combining the frequency feature vector, and a classification module for outputting a failure diagnosis result of the equipment by analyzing the vector combined through the combiner.

In addition, the fault diagnosis method using machine learning-based frequency analysis according to the present invention is a method for performing fault diagnosis based on machine learning in a fault diagnosis device that outputs a state diagnosis result of a facility. Generating an input vector, frequency-converting the sensing signal to generate a frequency conversion vector, receiving the signal input vector and frequency conversion vector and analyzing it through a deep learning-based model to obtain a state diagnosis result for the facility output step.

As described above, the AI-based fault diagnosis technology according to the present invention can increase the failure diagnosis performance of equipment because it simultaneously uses various signal characteristics according to the time change of the sensing signal and the frequency characteristics of the sensing signal in a deep learning-based model. there is an effect

In addition, since the present invention uses a vector combining signal characteristics and frequency characteristics as an input vector, the amount of computation can be dramatically reduced compared to a method using a spectrogram that visualizes a signal in terms of frequency, amplitude, and time, so that the calculation speed is fast. It has the effect of performing fault diagnosis.

The failure diagnosis technology using deep learning according to the present invention can be used not only as a core technology for diagnosing failures and conditions of various industrial facilities, but also in diagnostic fields such as non-destructive testing.

In addition, it will be applicable to technology that optimizes production conditions by judging the state of equipment according to production operation conditions and predicting product quality according to equipment conditions as an essential technology for implementing a smart factory.

1 is a view for explaining a conventional defect classification method;
2 is a view for explaining a general deep learning-based classification method.
3 is a diagram showing a schematic configuration of a failure diagnosis apparatus using machine learning-based frequency analysis according to the present invention.
4 is a detailed configuration diagram of a failure diagnosis apparatus using machine learning-based frequency analysis according to the first embodiment of the present invention.
5 is a detailed configuration diagram of a failure diagnosis apparatus using machine learning-based frequency analysis according to the first embodiment of the present invention.
6 is a flowchart of a failure diagnosis method using machine learning-based frequency analysis according to the first embodiment of the present invention.
7 is a flowchart of a failure diagnosis method using machine learning-based frequency analysis according to a second embodiment of the present invention.
8 is a diagram illustrating cropping a sensing signal according to the present invention.
9 is a view showing a plurality of sensing signals for measuring the equipment state according to the present invention.

Hereinafter, with reference to the accompanying drawings, embodiments of the present invention will be described in detail so that those of ordinary skill in the art to which the present invention pertains can easily implement them.

However, the present invention may be embodied in many different forms and is not limited to the embodiments described herein.

And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout the specification, when a part "includes" a certain component, it means that other components may be further included, rather than excluding other components, unless otherwise stated.

In addition, the terms "... unit" and "... module" described in the specification mean a unit that processes at least one function or operation, which may be implemented by hardware or software or a combination of hardware and software.

Hereinafter, a failure diagnosis apparatus and method using machine learning-based frequency analysis according to an embodiment of the present invention will be described in detail with reference to the drawings.

3 shows the configuration of a failure diagnosis apparatus using machine learning-based frequency analysis according to the present invention.

Referring to FIG. 1 , a failure diagnosis apparatus using machine learning-based frequency analysis (hereinafter referred to as a failure diagnosis apparatus) includes an input vector generation unit 10, a signal conversion unit 20, an analysis unit 30, and the like as main components. do.

The input vector generator 10 receives a sensing signal for the facility and generates a signal input vector. Various sensors are installed in equipment (factory equipment, equipment) or objects (composite structures, products) to output signals indicating the state of equipment or objects, that is, sensing signals. Sensors include vibration, displacement, temperature, and ultrasonic sensors.

The input vector generator generates a one-dimensional signal input vector by cropping the sensing signal to a predetermined window size at regular intervals and extracting the signal size at each sampling period.

The signal converter 20 frequency-converts the sensed signal to generate a frequency-converted vector. The signal converter 20 converts a sensing signal that changes with time into a frequency component by using a Fourier transform.

The analysis unit 30 receives a signal input vector and a frequency conversion vector, analyzes it through a deep learning-based model, and outputs a state diagnosis result for the facility.

The analysis unit 30 extracts a signal feature vector from a one-dimensional signal input vector through a convolutional neural network (CNN), combines the extracted signal feature vector and a frequency conversion vector, and analyzes the combined vector through a deep neural network to determine the state of the facility The diagnostic result is output.

As another embodiment, the analysis unit 30 extracts a signal feature vector from a one-dimensional input vector through a first convolutional neural network (CNN), extracts a frequency feature vector from a frequency transform vector through a second convolutional neural network, and the extracted signal By combining the feature vector and the frequency feature vector and analyzing the combined vector through a deep neural network, it is possible to output the state diagnosis result for the facility.

The input vector generating unit 10 , the signal converting unit 20 , the analyzing unit 30 , etc. constituting the failure diagnosis apparatus may be implemented in software or hardware. The analysis unit 30 applies a model for performing fault diagnosis of equipment, and this model is a model learned based on machine learning.

The failure diagnosis apparatus according to the present invention may be used in the field of product defect classification as well as equipment failure diagnosis. In addition, as an essential technology for implementing a smart factory, it can be used to determine the quality of a product according to production operating conditions.

4 is a detailed view of the configuration of a failure diagnosis apparatus according to the first embodiment of the present invention.

Referring to FIG. 4 , the analysis unit 30 of the failure diagnosis apparatus according to the first embodiment includes a first feature extraction module 32 , a combiner 34 , a classification module 36 , and the like.

The first feature extraction module 32 extracts a signal feature vector from the one-dimensional signal input vector output from the input vector generator 10 using a one-dimensional convolutional neural network (CNN).

The first feature extraction module 32 is composed of a plurality of layers performing convolution, batch normalization, activation function (ReLU), and maximization pooling to extract a signal feature vector from a signal input vector. will do

The combiner 33 combines the signal feature vector extracted from the first feature extraction module 32 and the frequency conversion vector output from the signal converter 20 .

The classification module 36 analyzes the vector combined through the combiner 33 through a deep neural network (DNN) and outputs a failure diagnosis result of the facility.

The joint vector output from the combiner 33 is connected to a fully connected layer having the same size as the number of types of diagnostic states to be finally classified. That is, the element value of the coupling vector passes through the neural network of the fully connected layer to classify the diagnostic state of the equipment.

The classification module 36 may classify the failure or state of the equipment into a plurality of classifications into normal, failure 1, failure 2, ... failure N.

5 is a detailed view of the configuration of a failure diagnosis apparatus according to a second embodiment of the present invention.

Referring to FIG. 4 , the analysis unit 30 of the failure diagnosis apparatus according to the second embodiment includes a first feature extraction module 32 , a second feature extraction module 33 , a combiner 34 , and a classification module 36 . etc.

Unlike the first embodiment, the second embodiment further includes a second feature extraction module 33 for extracting features of the frequency signal.

The first feature extraction module 32 extracts a signal feature vector from the one-dimensional signal input vector output from the input vector generator 10 using a one-dimensional convolutional neural network (CNN).

The second feature extraction module 33 uses a one-dimensional convolutional neural network (CNN) applied to the first feature extraction module 32 and a one-dimensional convolutional neural network different from the frequency feature vector from the frequency transformation vector output from the signal converter 20 . to extract

The combiner 33 combines the signal feature vector extracted by the first feature extraction module 32 and the frequency feature vector extracted by the second feature extraction module 33 .

The classification module 36 analyzes the vector combined through the combiner 33 through a deep neural network (DNN) and outputs a failure diagnosis result of the facility.

When there are few sensing signals, there is no big problem even if the frequency transformation vector is directly input into the deep neural network, but when there are many sensing signals, the size of the frequency transformation vector increases and it is difficult to secure diagnostic performance when input directly into the deep neural network. In the embodiment, a separate feature extraction module capable of extracting the features of the frequency transformation vector is provided.

6 is a flowchart of a failure diagnosis method using machine learning-based frequency analysis according to the first embodiment of the present invention.

Referring to FIG. 6 , first, a sensing signal is received from a sensor installed in the facility, and the sensing signal is cropped ( S10 ).

Cropping processing for the sensing signal will be described with reference to FIGS. 8 and 9 .

Referring to FIG. 8 , the sensing signal is a time-varying signal, and the sensing signal is cut into a predetermined window size at regular time intervals (stride).

The cropped signal is converted into a signal input vector to be input to the analysis unit 30 (S12). For example, when the time interval is 50 ms, the window size is 100 ms, and the sampling period is 1 ms, a signal input vector including 100 signal values may be generated every 50 ms.

If, as shown in FIG. 9 , when three signals are output from one sensor in the x-axis, y-axis, and z-axis directions like a vibration sensor, the three signals are cropped in the same way and converted into a signal input vector. can be

At the same time, the cropped sensing signal is generated as a frequency transform vector through Fourier transform (S14).

Next, the signal input vector is extracted as a signal feature vector through the convolutional neural network (S16), and the signal feature vector and the frequency conversion vector thus extracted are combined (S18).

For example, when the magnitude of the signal feature vector is 100 and the magnitude of the frequency transformation vector is 100, concatenating two vectors results in a vector of magnitude 200.

After the vector combined in this way is analyzed through the deep neural network (S20), the failure or state of the equipment is classified and output in the output layer of the deep neural network (S22).

7 is a flowchart of a failure diagnosis method using machine learning-based frequency analysis according to a second embodiment of the present invention.

Referring to FIG. 7 , in the second embodiment, unlike the first embodiment, a cropped sensing signal is generated into a frequency transform vector through Fourier transform, and then is not directly combined with a signal feature vector, but is obtained from the frequency transform vector. The step of extracting (S17) is further included so that the extracted frequency feature vector is combined with the signal feature vector.

That is, a signal feature vector is extracted from a signal input vector through a first convolutional neural network (CNN) in step S16, and a frequency conversion vector is extracted from a second convolutional neural network (CNN) different from the first convolutional neural network (CNN) in step S17. A frequency feature vector is extracted.

Thereafter, the extracted signal feature vector and frequency feature vector are combined (S18), and the combined vector is analyzed through a deep neural network as in the first embodiment (S20), and a state diagnosis result for the facility is output (S22).

The above description is merely illustrative of the present invention, and various modifications may be made by those of ordinary skill in the art to which the present invention pertains without departing from the technical spirit of the present invention.

Accordingly, the embodiments disclosed in the specification of the present invention are not intended to limit the present invention. The scope of the present invention should be construed by the following claims, and all technologies within the scope equivalent thereto should be construed as being included in the scope of the present invention.

10: input vector generating unit 20: signal converting unit
30: analysis unit 32: first feature extraction module
33: second feature extraction module 34: combiner
36: classification module

Claims (12)

An input vector generating unit that receives a sensing signal for a facility and generates a signal input vector;
a signal converter for frequency-converting the sensing signal to generate a frequency-converted vector;
A failure diagnosis apparatus using machine learning-based frequency analysis, comprising an analysis unit that receives the signal input vector and frequency conversion vector, analyzes it through a deep learning-based model, and outputs a state diagnosis result for the facility. According to claim 1,
The input vector generator crops the sensing signal to a predetermined window size at regular intervals and then extracts the signal size at every sampling period to generate a one-dimensional signal input vector. fault diagnosis device used. 3. The method of claim 2,
The analysis unit extracts a signal feature vector from the one-dimensional signal input vector through a convolutional neural network (CNN), combines the extracted signal feature vector with the frequency conversion vector, and analyzes the combined vector through a deep neural network to determine the state of the facility A failure diagnosis device using machine learning-based frequency analysis, characterized in that it outputs a diagnosis result. 3. The method of claim 2,
The analysis unit extracts a signal feature vector from the one-dimensional input vector through a first convolutional neural network (CNN), extracts a frequency feature vector from the frequency transform vector through a second convolutional neural network, and the extracted signal feature vector and the A failure diagnosis device using machine learning-based frequency analysis, characterized in that it outputs a state diagnosis result for a facility by combining frequency feature vectors and analyzing the combined vector through a deep neural network. An input vector generator that receives a sensing signal for a facility and generates a one-dimensional signal input vector;
a signal converter for frequency-converting the sensing signal to generate a frequency-converted vector;
a feature extraction module for extracting a signal feature vector from the one-dimensional signal input vector;
a combiner for combining the signal feature vector and the frequency conversion vector;
A failure diagnosis apparatus using machine learning-based frequency analysis including a classification module for outputting a failure diagnosis result of a facility by analyzing the vector combined through the combiner. An input vector generator that receives a sensing signal for a facility and generates a one-dimensional signal input vector;
a signal converter for frequency-converting the sensing signal to generate a frequency-converted vector;
a first feature extraction module for extracting a signal feature vector from the one-dimensional signal input vector;
a second feature extraction module for extracting a frequency feature vector from the frequency transformation vector;
a combiner for combining the signal feature vector and the frequency feature vector;
A failure diagnosis apparatus using a machine learning-based frequency analysis including a classification module for outputting a failure diagnosis result of a facility by analyzing the vector combined through the combiner. 7. The method according to claim 5 or 6,
The input vector generator crops the sensing signal to a predetermined window size at regular intervals and then extracts the signal size at every sampling period to generate a one-dimensional signal input vector. fault diagnosis device used. 7. The method according to claim 5 or 6,
The first feature extraction module or the second feature extraction module is a failure diagnosis apparatus using machine learning-based frequency analysis, characterized in that it is composed of a one-dimensional convolutional neural network. A method for performing fault diagnosis based on machine learning in a fault diagnosis device that outputs a state diagnosis result of a facility, comprising:
receiving a sensing signal for the facility and generating a signal input vector;
frequency-converting the sensing signal to generate a frequency-converted vector;
A failure diagnosis method using a machine learning-based frequency analysis comprising the step of receiving the signal input vector and the frequency conversion vector as input, analyzing it through a deep learning-based model, and outputting a state diagnosis result for the facility. 10. The method of claim 9,
The generating of the signal input vector comprises cropping the sensing signal to a predetermined window size at regular intervals and then extracting the signal size at every sampling period to generate a one-dimensional signal input vector. Fault diagnosis method using based frequency analysis. 10. The method of claim 9,
The step of outputting the state diagnosis result for the facility includes the process of extracting a signal feature vector from the signal input vector through a convolutional neural network (CNN);
combining the extracted signal feature vector and the frequency conversion vector;
Failure diagnosis method using machine learning-based frequency analysis, characterized in that it comprises the step of analyzing the combined vector through a deep neural network and outputting a state diagnosis result for the facility. 10. The method of claim 9,
The step of outputting the state diagnosis result for the facility includes the process of extracting a signal feature vector from the signal input vector through a first convolutional neural network (CNN);
The process of extracting a frequency feature vector from the frequency transformation vector through a second convolutional neural network (CNN);
combining the extracted signal feature vector and the frequency feature vector;
Failure diagnosis method using machine learning-based frequency analysis, characterized in that it comprises the step of analyzing the combined vector through a deep neural network and outputting a state diagnosis result for the facility.

Images