CN116595444A - Fault category detection method and system for medical instrument based on deep learning - Google Patents
Fault category detection method and system for medical instrument based on deep learning Download PDFInfo
- Publication number
- CN116595444A CN116595444A CN202310883905.7A CN202310883905A CN116595444A CN 116595444 A CN116595444 A CN 116595444A CN 202310883905 A CN202310883905 A CN 202310883905A CN 116595444 A CN116595444 A CN 116595444A
- Authority
- CN
- China
- Prior art keywords
- signal
- fault
- vibration
- interpolation
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 32
- 238000013135 deep learning Methods 0.000 title claims abstract description 28
- 238000013145 classification model Methods 0.000 claims abstract description 49
- 238000000034 method Methods 0.000 claims abstract description 30
- 238000007781 pre-processing Methods 0.000 claims abstract description 28
- 230000000153 supplemental effect Effects 0.000 claims abstract description 28
- 239000013598 vector Substances 0.000 claims description 106
- 238000012549 training Methods 0.000 claims description 31
- 239000011159 matrix material Substances 0.000 claims description 27
- 230000006870 function Effects 0.000 claims description 26
- 238000012360 testing method Methods 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 13
- 230000002457 bidirectional effect Effects 0.000 claims description 12
- 230000007246 mechanism Effects 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 10
- 230000002708 enhancing effect Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 230000000295 complement effect Effects 0.000 abstract description 7
- 230000008859 change Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000006073 displacement reaction Methods 0.000 description 4
- 230000001965 increasing effect Effects 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 230000001133 acceleration Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 3
- 238000004590 computer program Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
- G16H40/40—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the management of medical equipment or devices, e.g. scheduling maintenance or upgrades
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Business, Economics & Management (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- General Business, Economics & Management (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- Multimedia (AREA)
- Databases & Information Systems (AREA)
- Human Resources & Organizations (AREA)
- Public Health (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Primary Health Care (AREA)
- Epidemiology (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The application belongs to the field of medical instruments, and particularly relates to a fault class detection method and system of a medical instrument based on deep learning, wherein the method comprises the following steps: acquiring a vibration signal of the medical instrument, and preprocessing the vibration signal; performing interpolation operation on the preprocessed signals; adding random disturbance into the interpolation signal to obtain a first supplementary signal; randomly oversampling the vibration signal to generate a second supplemental signal; combining the second supplementary signal with the vibration signal to obtain an expanded signal; converting the expansion signal into a fault image, inputting the fault image into a multi-layer classification model for fault classification, and obtaining the fault category of the medical instrument; the method can simulate noise and variation in an actual vibration signal by adding random disturbance to the interpolation signal, and enhance the anti-interference capability of data. The first complementary signals are randomly oversampled, so that the number of signals of different types of medical instrument fault categories can be balanced, and the accuracy of fault classification of the medical instrument is improved.
Description
Technical Field
The application relates to the field of medical instruments, in particular to a fault type detection method and system for a medical instrument based on deep learning.
Background
With the development of medical technology, at present, doctors provide important medical evidence for the detection result of medical instruments in the disease diagnosis process, and the occurrence of medical instrument faults is one of the main factors for increasing medical accidents, but the faults cannot be found in time by simple manual detection. For this reason, related research and development personnel have developed a GIS vibration signal detection apparatus for detecting a failure of a medical instrument by performing vibration detection.
Although the detection of the faults of the medical instrument based on the vibration signals is more convenient, the vibration signals in the running state of the medical instrument are often complex, and contain a certain noise interference to influence the analysis and processing effects of the vibration signals. Therefore, the fault type detection method based on deep learning is adopted, and the acquired signals are subjected to improvements such as preprocessing, so that the problem of uneven samples in training data is solved, and the accurate detection of faults is realized.
Disclosure of Invention
The application aims at: the utility model provides a fault class detection method and system of medical instrument based on deep learning, which can solve the problem that the existing medical instrument fault detection technology has unbalanced types of vibration signals, thereby reducing the accuracy of judging the fault class of the medical instrument.
In order to achieve the above object, the present application provides a fault class detection method for a medical device based on deep learning, including:
acquiring a vibration signal of a medical instrument, and preprocessing the vibration signal to obtain a preprocessed signal;
performing interpolation operation on the preprocessing signal to obtain an interpolation signal;
adding random disturbance to the interpolation signal to obtain a first supplementary signal;
randomly oversampling the vibration signal to generate a second supplemental signal;
splicing the first supplementary signal, the second supplementary signal and the vibration signal to obtain an expansion signal;
converting the expansion signal into a fault image, inputting the fault image into a multi-layer classification model for fault classification, and obtaining the fault category of the medical instrument; the multi-layer classification model is obtained by training a model to be trained by using the expanding signals.
Preferably, the inputting the fault image into a multi-layer classification model for fault classification to obtain a medical instrument fault class includes:
inputting the fault image into a convolution layer of the multi-layer classification model for feature extraction to obtain image features;
inputting the image features into a first bidirectional LSTM layer of the multi-layer classification model for time series processing to obtain a first intermediate result;
inputting the first intermediate result into a second bidirectional LSTM layer of the multi-layer classification model for time series processing to obtain a second intermediate result;
inputting the second intermediate result into an attention mechanism layer of the multi-layer classification model to perform attention enhancement, so as to obtain an attention enhancement result;
and inputting the attention enhancement result into a classifier of the multi-layer classification model to classify, so as to obtain the medical instrument fault class.
Preferably, the inputting the second intermediate result into the attention mechanism layer of the multi-layer classification model performs attention enhancement to obtain an attention enhancement result, including:
the second intermediate result is attentive by the following formula:
；
；
；
wherein ,for the j-th unnormalized attention score,for the j-th attention weight,in order to achieve the above-mentioned attention-enhancing result,as a first matrix of parameters,as a matrix of the second parameters,for the third parameter matrix, T represents a matrix transpose operation,in order to query the matrix,for the j-th key matrix,tanh represents the hyperbolic tangent function, softmax represents the normalized exponential function,is the j-th value matrix.
Preferably, the preprocessing the vibration signal to obtain a preprocessed signal includes:
normalizing the vibration signal to obtain a normalized signal;
filtering the normalized signal to obtain a filtered signal;
and filling the missing value of the filtered signal to obtain the preprocessing signal.
Preferably, before the fault image is input into the multi-layer classification model to perform fault classification, training the model to be trained is further included, and the training process includes:
setting a minimum vector and a maximum vector according to the first supplemental signal;
randomly generating a target vector according to the minimum vector and the maximum vector;
obtaining differential weights, and generating disturbance vectors according to the differential weights and the target vectors;
setting the target vector or the disturbance vector as a test vector according to the cross discriminant;
detecting whether the fitness of the test vector is higher than that of the target vector, if so, replacing the target vector with the test vector;
adding Gaussian disturbance to the test vector to obtain a disturbance adding vector;
and detecting whether the current training times is greater than or equal to the maximum training times, and if not, randomly generating the target vector according to the minimum vector and the maximum vector.
Preferably, the randomly oversampling the vibration signal to generate a second supplemental signal includes:
randomly selecting a first supplementary data point from the vibration signal, and calculating a first threshold value according to the oversampling ratio;
randomly copying the vibration data points of the first threshold value from the vibration signals to obtain intermediate signals;
and carrying out interpolation operation on the intermediate signal, adding random disturbance, and generating the second supplementary signal.
Preferably, the performing interpolation operation on the preprocessed signal to obtain an interpolation signal includes:
selecting different preprocessed data points in the preprocessed signal;
interpolation is performed on the different preprocessed data points using bilinear interpolation according to the following formula:
；
wherein ,for the interpolation coefficient to be used,for the ith pre-processed data point,for the jth pre-processed data point,for the kth interpolated data point, i < j;
and combining all the preprocessed data points and all the interpolated data points into the interpolated signal.
Preferably, the adding random disturbance in the interpolation signal to obtain a first complementary signal includes:
randomly generating a plurality of random disturbances through normal distribution; wherein the number of random perturbations is the same as the number of interpolated data points;
and adding each interpolation data point to the corresponding random disturbance to obtain the first supplementary signal.
Preferably, the converting the expansion signal into a fault image includes:
and performing short-time Fourier transform on the expansion signal to obtain the fault image.
The application provides a fault class detection system of a medical instrument based on deep learning, which comprises the following components:
the pretreatment module is used for obtaining vibration signals of the medical instrument, and carrying out pretreatment on the vibration signals to obtain pretreatment signals;
the interpolation operation module is used for carrying out interpolation operation on the preprocessing signals to obtain interpolation signals;
the random disturbance adding module is used for adding random disturbance into the interpolation signal to obtain a first supplementary signal;
the random oversampling module is used for carrying out random oversampling on the vibration signal and generating a second supplementary signal;
the signal combination module is used for splicing the first supplemental signal, the second supplemental signal and the vibration signal to obtain an expansion signal;
the fault classification module is used for converting the expansion signals into fault images, inputting the fault images into a multi-layer classification model for fault classification, and obtaining the fault types of the medical instruments; the multi-layer classification model is obtained by training a model to be trained by using the expanding signals.
The fault type detection method of the medical instrument based on deep learning comprises the steps of obtaining a vibration signal of the medical instrument, and preprocessing the vibration signal to obtain a preprocessed signal; performing interpolation operation on the preprocessing signal to obtain an interpolation signal; adding random disturbance into the interpolation signal to obtain a first supplementary signal; randomly oversampling the vibration signal to generate a second supplemental signal; combining the second supplementary signal with the vibration signal to obtain an expanded signal; converting the expansion signal into a fault image, inputting the fault image into a multi-layer classification model for fault classification, and obtaining the fault category of the medical instrument; the multi-layer classification model is obtained by training a model to be trained by using the expanding signals. According to the method, the number of data points in the preprocessed signals can be increased through interpolation operation, random disturbance is added to the interpolation signals, noise and change in actual vibration signals can be simulated, and the anti-interference capability of the data is improved. The first complementary signals are randomly oversampled, so that the number of signals of different types of medical instrument fault categories can be balanced, and the accuracy of fault classification of the medical instrument is improved.
Drawings
FIG. 1 is a flow chart of a fault class detection method of a deep learning-based medical device according to an embodiment;
FIG. 2 is a flow chart of preprocessing a vibration signal according to an embodiment;
FIG. 3 is a flow chart of random oversampling of vibration signals according to an embodiment;
FIG. 4 is a schematic flow chart of inputting a fault image into a multi-layer classification model for fault classification according to an embodiment;
FIG. 5 is a schematic diagram of a process for training a model to be trained according to an embodiment;
FIG. 6 is a block diagram schematically illustrating the structure of a fault class detection system for a deep learning-based medical device according to an embodiment;
fig. 7 is a block diagram schematically illustrating a structure of a computer device according to an embodiment.
The achievement of the objects, functional features and advantages of the present application will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless expressly stated otherwise, as understood by those skilled in the art. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, modules, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, modules, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. The term "and/or" as used herein includes all or any module and all combination of one or more of the associated listed items.
It will be understood by those skilled in the art that all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs unless defined otherwise. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
In one embodiment, referring to fig. 1, a flow chart of a fault class detection method of a medical device based on deep learning disclosed in the present application includes:
s1: and acquiring a vibration signal of the medical instrument, and preprocessing the vibration signal to obtain a preprocessed signal.
Vibration signals of the medical instrument in the running process are acquired through a vibration sensor, and the vibration sensor is arranged at a key part of the medical instrument, such as a motor and/or a bearing, to measure the vibration signals. The vibration sensor converts the vibration signal into an electric signal, and then the electric signal is connected to a data acquisition system or a data recorder for data acquisition and recording.
The vibration signal includes a vibration amplitude, a vibration frequency, a vibration acceleration, a vibration velocity, and a vibration displacement. The vibration amplitude represents the magnitude of the vibration or the maximum deflection of the vibration. Typically expressed in millimeters (mm) or micrometers (μm), the vibration frequency represents the frequency of the vibration signal, i.e., the periodic variation of the vibration. A common unit is hertz (Hz), which represents the number of vibration cycles that occur per second. The vibration acceleration represents the acceleration of the vibration signal, i.e., the rate of change of the vibration speed per unit time. The usual unit is meters per square second (m/s). The vibration speed represents the speed of the vibration signal, i.e., the rate of change of vibration. The usual unit is meter per second (m/s), and the vibration displacement represents the displacement amount of the vibration signal, i.e., the displacement amount of the vibration. Commonly used units are millimeters (mm) or micrometers (μm).
In one embodiment, medical device vibration fault data is collected, including normal data collection and fault data collection. And a GIS vibration information detection device is adopted to collect vibration signals of the medical instrument in the operation process.
Referring to fig. 2, the vibration signal is preprocessed to obtain a preprocessed signal, which includes the following steps:
s12: normalizing the vibration signal to obtain a normalized signal.
Step S11 is further included before step S12: a vibration signal of the medical instrument is acquired.
Preferably, the vibration signal is labeled for a medical instrument failure category including bearing wear, imbalance, misalignment, and looseness.
Normalization is an important step in the preprocessing process to eliminate the magnitude difference of the data, so that the data are on the same scale. The normalization method adopted in this embodiment is a min-max normalization method. Specifically, the vibration signal is denoted as X, the vibration signal includes N vibration data points, and the ith vibration data point is denoted as X i Wherein i=1, 2,..n. The minimum value in the vibration signal is a, the maximum value is b, and the vibration signal is normalized by the following formula:
；
wherein ,the normalized signal is represented by X, a, and b, respectively.
S13: and filtering the normalized signal to obtain a filtered signal.
The filtering is a process of eliminating noise and extracting useful signals, specifically, the filtering operation is performed by using a low-pass filter in this embodiment, and the formula of filtering the normalized signal is as follows:
Y =H（ sd(X norm )）
wherein Y is a filtered signal,to normalize the signal, H () is a filter function, sd () is a complex frequency conversion function, and X norm Converted to complex frequency and input into the filter function H ()For filtering the signal, K is the gain coefficient of the low-pass filter, T is the time constant of the low-pass filter, and s is the complex frequency.
S14: and filling the missing value of the filtered signal to obtain the preprocessing signal.
In this embodiment, the missing value filling is performed by means of mean filling, specifically, the mean value of all the filtered data points in the filtered signal isIf the ith filtered data point Y in the filtered signal i Is a missing value, then。
S2: and carrying out interpolation operation on the preprocessing signal to obtain an interpolation signal.
Selecting different preprocessed data points in the preprocessed signal;
interpolation is performed on the different preprocessed data points using bilinear interpolation according to the following formula:
；
wherein ,for the interpolation coefficient to be used,for the ith pre-processed data point,for the jth pre-processed data point,for the kth interpolated data point, i < j;
and combining all the preprocessed data points and all the interpolated data points into the interpolated signal.
The interpolation coefficient represents the distance of position k relative to i and j, with the interpolation coefficient being greater than 0 and less than 1. By using bilinear interpolation for the preprocessed signals, the gap between different preprocessed data points can be filled, and a continuous interpolation signal can be obtained.
S3: and adding random disturbance to the interpolation signal to obtain a first supplementary signal.
Randomly generating a plurality of random disturbances through normal distribution; wherein the number of random perturbations is the same as the number of interpolated data points;
and adding each interpolation data point to the corresponding random disturbance to obtain the first supplementary signal.
A random disturbance is added to each interpolated data point to simulate the noise and variation of the vibration signal in actual use. Each random disturbance is represented by a mean value of 0 and a variance of 0Is randomly generated. Alternatively, the variance is calculatedSet to 0.1.
Adding random perturbations to the interpolated signal by the following formula:
；
wherein ,for the kth first supplemental data point,for the kth interpolated data point,is the kth random perturbation.
S4: and randomly oversampling the vibration signal to generate a second complementary signal.
Referring to fig. 3, step S4 includes the following steps S41 to S43:
s41: a first supplemental data point is randomly selected from the vibration signal, and a first threshold is calculated based on the oversampling ratio.
Randomly selecting a first supplemental data pointThe formula for calculating the first threshold is as follows:
；
wherein ,as a result of the first threshold value being set,in order to achieve an oversampling ratio,for the total number of vibration data points of the vibration signal []Representing a down-rounding operation or a rounding operation,representing oversampling ratio and vibration data pointIs the product of the total number of (c).
S42: randomly copying the vibration data points of the first threshold value from the vibration signals to obtain intermediate signals.
The intermediate signal contains a first threshold vibration data point.
S43: and carrying out interpolation operation on the intermediate signal, adding random disturbance, and generating the second supplementary signal.
The interpolation operation on the intermediate signal is the same as step S2, and will not be described here again. The addition of random disturbance is the same as step S3 and will not be described here again.
Steps S41-S43 are repeated until the number of generated second supplemental signals is greater than or equal to the second threshold. The second threshold may be set to 2 or 5, which is not limited herein.
S5: and splicing the first supplementary signal, the second supplementary signal and the vibration signal to obtain an expansion signal.
The first supplemental signal, the second supplemental signal, and the vibration signal may be spliced in a random order or a fixed order, as an example, the first supplemental signal being spliced after the vibration signal and the second supplemental signal being spliced after the first supplemental signal.
S6: converting the expansion signal into a fault image, inputting the fault image into a multi-layer classification model for fault classification, and obtaining the fault category of the medical instrument; the multi-layer classification model is obtained by training a model to be trained by using the expanding signals.
And performing short-time Fourier transform on the expansion signal to obtain the fault image.
Performing short-time Fourier transform on the expansion signal according to the following formula:
wherein I is a fault image,is the data filled with the missing values,is a function of the window and,the time is represented by the time period of the day,representing frequency, in the formulaRepresenting imaginary units.Is a virtual variable for performing integration over time,representing the integral variable, all of which are traversed during the integration processI.e. all possible points in time, to calculate the time-frequency conversion result.
The multi-layer classification model includes a convolution layer, a first bi-directional LSTM layer, a second bi-directional LSTM layer, an attention mechanism layer, and a classifier. The fault image sequentially passes through the convolution layer, the first bidirectional LSTM layer, the second bidirectional LSTM layer, the attention mechanism layer and the classifier to obtain the fault category of the medical instrument.
The fault type detection method of the medical instrument based on deep learning comprises the steps of obtaining a vibration signal of the medical instrument, and preprocessing the vibration signal to obtain a preprocessed signal; performing interpolation operation on the preprocessing signal to obtain an interpolation signal; adding random disturbance into the interpolation signal to obtain a first supplementary signal; randomly oversampling the vibration signal to generate a second supplemental signal; combining the second supplementary signal with the vibration signal to obtain an expanded signal; converting the expansion signal into a fault image, inputting the fault image into a multi-layer classification model for fault classification, and obtaining the fault category of the medical instrument; the multi-layer classification model is obtained by training a model to be trained by using the expanding signals. According to the method, the number of data points in the preprocessed signals can be increased through interpolation operation, random disturbance is added to the interpolation signals, noise and change in actual vibration signals can be simulated, and the anti-interference capability of the data is improved. The first complementary signals are randomly oversampled, so that the number of signals of different types of medical instrument fault categories can be balanced, and the accuracy of fault classification of the medical instrument is improved.
Referring to fig. 4, in one embodiment, the inputting the fault image into the multi-layer classification model to perform fault classification, to obtain a medical device fault class includes:
s62: and inputting the fault image into a convolution layer of the multi-layer classification model to perform feature extraction, so as to obtain image features.
Before step S62, step S61 is further included: and converting the expansion signal into a fault image.
The convolution layer is used to extract image features, and the convolution operation can be expressed as:
；
wherein I is a fault image,in the form of a convolution kernel,for the image feature, u denotes the width of the failure image, v denotes the height of the failure image, m denotes the number of rows of the convolution kernel, and n denotes the number of columns of the convolution kernel.
And stretching the image features obtained by convolution operation into one-dimensional vectors.
S63: and inputting the image features into a first bidirectional LSTM layer of the multi-layer classification model for time series processing to obtain a first intermediate result.
The bi-directional LSTM layer has two sets of weight coefficients and bias coefficients, one set for forward sequences and the other set for backward sequences.
S64: and inputting the first intermediate result into a second bidirectional LSTM layer of the multi-layer classification model to perform time sequence processing to obtain a second intermediate result.
The image features are time-sequentially processed by the first bi-directional LSTM layer and the second bi-directional LSTM layer such that the second intermediate result has temporal features.
S65: and inputting the second intermediate result into an attention mechanism layer of the multi-layer classification model to perform attention enhancement, so as to obtain an attention enhancement result.
The second intermediate result is an image and the attention mechanism layer is used for assigning different attention weights to different areas of the second intermediate result. Specifically, the attention mechanism layer assigns a greater attention weight to the portion of the second intermediate result that is associated with the medical instrument failure category and a lesser attention weight to the portion that is not associated with the medical instrument failure category, thereby highlighting the portion of the second intermediate result that is associated with the medical instrument failure category.
The second intermediate result is attentive by the following formula:
；
；
；
wherein ,for the j-th unnormalized attention score,for the j-th attention weight,in order to achieve the above-mentioned attention-enhancing result,as a first matrix of parameters,as a matrix of the second parameters,for the third parameter matrix, T represents a matrix transpose operation,in order to query the matrix,for the j-th key matrix, tanh represents the hyperbolic tangent function, softmax represents the normalized exponential function,is the j-th value matrix.
S66: and inputting the attention enhancement result into a classifier of the multi-layer classification model to classify, so as to obtain the medical instrument fault class.
Preferably, the classifier uses a normalized exponential function to translate the attention-enhancing results into a probability distribution for multi-classification tasks. The normalized exponential function is expressed as:
；
wherein ,is the number of categories of medical instrument failure,is the firstThe attention of the individual categories enhances the outcome,is the firstThe attention of the individual categories enhances the outcome,is the firstPrediction probabilities for individual categories. The medical instrument fault class output by the model is the class corresponding to the maximum value of the prediction probability.
As described above, the fault images are input into the multi-layer classification model, the fault images sequentially undergo feature extraction by the convolution layer, the first bidirectional LSTM layer performs time-series processing on the first intermediate result, and the second bidirectional LSTM layer performs time-series processing on the second intermediate result. And inputting the second intermediate result into an attention mechanism layer, and distributing a larger attention weight to the part of the second intermediate result related to the medical instrument fault class so as to highlight the part related to the medical instrument fault class. And converting the attention enhancement result into probability distribution through a classifier, and taking the category corresponding to the maximum predicted probability as the medical instrument fault category.
In one embodiment, referring to fig. 5, before the fault image is input into the multi-layer classification model for fault classification, the training process further includes training the model to be trained, where the training process includes the following steps S51'-S57':
s51': and setting a minimum vector and a maximum vector according to the first supplementary signal.
The minimum vector is expressed asThe maximum vector is expressed as。
S52': and randomly generating a target vector according to the minimum vector and the maximum vector.
The target vector is generated by the following formula:
；
wherein ,is the ith target vector at the G generation,to randomly generate a random function that is greater than 0 and less than 1, NP is the total number of target vectors.
Optionally, NP is set to 10.
S53': and obtaining differential weights, and generating disturbance vectors according to the differential weights and the target vectors.
The disturbance vector is generated by the following formula:
；
wherein ,as a vector of the disturbance,is the firstThe object vector corresponding to the index is selected,is the firstThe object vector corresponding to the index is selected,is the firstThe object vector corresponding to the index is selected,is three different indices randomly chosen from {1,2,..np } and F is the differential weight.
S54': the target vector or the disturbance vector is set as a test vector according to a cross discriminant.
The diversity of the data can be increased by performing the cross operation by the cross discriminant, which is as follows:
；
wherein CR is the cross probability,is an index randomly selected from {1,2,.,. D., },an ith test vector corresponding to the jth dimension of the (G+1) -th generation,an ith parameter vector corresponding to the jth dimension of the G generation,the ith disturbance vector corresponding to the jth dimension of the (G+1) -th generation.
The differential weights and the crossover probabilities are adaptively updated. The differential weights F and the crossover probabilities CR are dynamically adjusted during operation according to the previous search performance. Specifically, an initial candidate list is set for F and CR, such as:
；
；
in each generation, for each individual, the values of F and CR are randomly selected:
；
；
wherein ,for the differential weight corresponding to the i-th individual,for the crossover probability corresponding to the ith individual,as a function of the random value of the code,a function is calculated for the length.
Further, after the selection operation, the candidate list of F and CR is updated according to the quality of the new solution. If the new solution is better than the current solution, the corresponding F and CR values are added to the candidate list.
Further, the number of successes for each F and CR value in the past search is tracked by a variable S. Then in each generation, the values of F and CR are updated according to the following formula:
；
；
wherein the argmax function represents the set of arguments that find when a given function takes a maximum,representing differential weights defining when the fitness function takes a maximum value in the search tracking variable in the set corresponding to F,the crossover probability when the fitness function takes the maximum value in the search tracking variable in the set corresponding to CR is defined.
S55': and detecting whether the fitness of the test vector is higher than that of the target vector, and if so, replacing the target vector with the test vector.
And if the fitness of the test vector is lower than or equal to the fitness of the target vector, keeping the target vector unchanged.
Updating the target vector or leaving the target vector unchanged by the following formula:
；
wherein ,is the ith test vector of the G+1th generation,is the ith parameter vector at the G generation,as a fitness function. The fitness function is expressed as:
；
wherein ,as a function of the loss of the model,is a small positive number to prevent the denominator from being zero. In particular, the loss function of the model is a cross entropy loss function.
S56': and adding Gaussian disturbance to the test vector to obtain a disturbance adding vector.
When a new better solution is found, a local search process is started to refine the new better solution, so that better balance between the quality of the solution and the search efficiency is achieved.
Specifically, when a new more optimal solution is found, a small gaussian perturbation is added to the new more optimal solution, as follows:
；
wherein ,for a new and better solution to be obtained,in order to add the post-disturbance solution,is an adjustable parameter for controlling the magnitude of the disturbance,representing a standard normal distribution.
Further, a new solution is calculatedIs used for the adaptation degree of the device. If a new solution ratioMore preferably, then replace with a new solution. Otherwise, keepIs unchanged. Can be expressed as:
；
s57': and detecting whether the current training times is greater than or equal to the maximum training times, and if not, randomly generating the target vector according to the minimum vector and the maximum vector.
If the current training times are smaller than the maximum training times, repeating the steps S53'-S56'; and if the current training times are greater than or equal to the maximum training times, stopping training the model to be trained to obtain the multi-layer classification model.
As described above, by comparing the fitness of the test vector and the fitness of the target vector, the target vector can be replaced constantly, and the parameters of each layer of the model to be trained can be updated.
Referring to fig. 6, there is a schematic block diagram of a fault class detection system for a deep learning-based medical device according to the present disclosure, the system comprising:
the preprocessing module 10 is used for acquiring a vibration signal of the medical instrument, preprocessing the vibration signal and obtaining a preprocessed signal;
the interpolation operation module 20 is configured to perform interpolation operation on the preprocessed signal to obtain an interpolation signal;
a random disturbance adding module 30, configured to add a random disturbance to the interpolation signal, so as to obtain a first supplemental signal;
a random oversampling module 40, configured to randomly oversample the vibration signal to generate a second supplemental signal;
the signal combining module 50 is configured to splice the first supplemental signal, the second supplemental signal, and the vibration signal to obtain an extended signal;
the fault classification module 60 is configured to convert the expansion signal into a fault image, and input the fault image into a multi-layer classification model for fault classification, so as to obtain a medical instrument fault class; the multi-layer classification model is obtained by training a model to be trained by using the expanding signals.
The fault class detection system of the medical instrument based on the deep learning is used for realizing a fault class detection method of the medical instrument based on the deep learning.
In one embodiment, the fault classification module 60 includes:
the convolution unit is used for inputting the fault image into a convolution layer of the multi-layer classification model to perform feature extraction so as to obtain image features;
the first time sequence processing unit is used for inputting the image characteristics into a first bidirectional LSTM layer of the multi-layer classification model to perform time sequence processing to obtain a first intermediate result;
the second time sequence processing unit is used for inputting the first intermediate result into a second bidirectional LSTM layer of the multi-layer classification model to perform time sequence processing to obtain a second intermediate result;
the attention enhancing unit is used for inputting the second intermediate result into an attention mechanism layer of the multi-layer classification model to enhance the attention and obtain an attention enhancing result;
and the classification unit is used for inputting the attention enhancement result into a classifier of the multi-layer classification model to classify, so as to obtain the medical instrument fault class.
In one embodiment, the attention enhancing unit includes:
an attention enhancing subunit for enhancing the attention of the second intermediate result by the following formula:
；
；
；
wherein ,for the j-th unnormalized attention score,for the j-th attention weight,in order to achieve the above-mentioned attention-enhancing result,as a first matrix of parameters,as a matrix of the second parameters,for the third parameter matrix, T represents a matrix transpose operation,in order to query the matrix,for the j-th key matrix, tanh represents the hyperbolic tangent function, softmax represents the normalized exponential function,is the j-th value matrix.
In one embodiment, the preprocessing module 10 includes:
the normalization unit is used for normalizing the vibration signal to obtain a normalized signal;
the filtering unit is used for filtering the normalized signal to obtain a filtered signal;
and the missing value filling unit is used for filling the missing value of the filtering signal to obtain the preprocessing signal.
In one embodiment, the fault class detection system of a deep learning-based medical instrument further comprises a model training module comprising:
a vector setting unit for setting a minimum vector and a maximum vector according to the first supplemental signal;
the first target vector generation unit is used for randomly generating a target vector according to the minimum vector and the maximum vector;
the disturbance vector generation unit is used for acquiring the differential weight and generating a disturbance vector according to the differential weight and the target vector;
the test vector setting unit is used for setting the target vector or the disturbance vector as a test vector according to the cross discrimination;
the target vector replacement unit is used for detecting whether the fitness of the test vector is higher than that of the target vector, and if so, replacing the target vector with the test vector;
the Gaussian disturbance adding unit is used for adding Gaussian disturbance to the test vector to obtain a disturbance adding vector;
and the second target vector generation unit is used for detecting whether the current training times are greater than or equal to the maximum training times, and if not, randomly generating the target vector according to the minimum vector and the maximum vector.
In one embodiment, the random oversampling module 40 includes:
the first threshold calculating unit is used for randomly selecting a first supplementary data point from the vibration signal and calculating a first threshold according to the oversampling ratio;
a vibration data point copying unit, configured to randomly copy the vibration data points of the first threshold value from the vibration signal, to obtain an intermediate signal;
and the second complementary signal generating unit is used for carrying out interpolation operation on the intermediate signal and adding random disturbance to generate the second complementary signal.
In one embodiment, the interpolation module 20 includes:
a preprocessing data point selecting unit, configured to select different preprocessing data points from the preprocessing signal;
an interpolation operation unit, configured to perform interpolation operation on different preprocessed data points by using bilinear interpolation according to the following formula:
；
wherein ,for the interpolation coefficient to be used,for the ith pre-processed data point,for the jth pre-processed data point,for the kth interpolated data point, i < j;
an interpolation signal composing unit for composing all the preprocessed data points and all the interpolation data points into the interpolation signal.
In one embodiment, the random disturbance adding module 30 includes:
a random disturbance generation unit for randomly generating a plurality of random disturbances through normal distribution; wherein the number of random perturbations is the same as the number of interpolated data points;
and the random disturbance adding unit is used for adding each interpolation data point with the corresponding random disturbance to obtain the first supplementary signal.
In one embodiment, the fault classification module 60 includes:
and the fault classification unit is used for carrying out short-time Fourier transform on the expansion signals to obtain the fault image.
Corresponding to the above method embodiment, a computer device is further provided in the embodiment of the present application, and the internal structure of the computer device may be as shown in fig. 7. The computer device includes a computer processor, a storage medium, a memory, a network interface, and an input device, the computer processor exchanging information with the storage medium, the storage medium including an operating system, a computer program, and a database. The computer processor is connected with the memory, the network interface and the input device through a system bus, and the network interface is used for carrying out network communication with external equipment.
Corresponding to the above method embodiment, the embodiment of the present application further provides a computer readable storage medium, where a computer program is stored, and the computer program when executed by a computer processor implements the steps of the fault class detection method for a medical apparatus based on deep learning.
It should also be noted that in this document relational terms such as first and second are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. The medical instrument fault type detection method based on deep learning is characterized by comprising the following steps of:
acquiring a vibration signal of a medical instrument, and preprocessing the vibration signal to obtain a preprocessed signal;
performing interpolation operation on the preprocessing signal to obtain an interpolation signal;
adding random disturbance to the interpolation signal to obtain a first supplementary signal;
randomly oversampling the vibration signal to generate a second supplemental signal;
splicing the first supplementary signal, the second supplementary signal and the vibration signal to obtain an expansion signal;
converting the expansion signal into a fault image, inputting the fault image into a multi-layer classification model for fault classification, and obtaining the fault category of the medical instrument; the multi-layer classification model is obtained by training a model to be trained by using the expanding signals.
2. The fault class detection method for deep learning-based medical devices according to claim 1, wherein the step of inputting the fault image into a multi-layer classification model for fault classification to obtain a medical device fault class comprises the steps of:
inputting the fault image into a convolution layer of the multi-layer classification model for feature extraction to obtain image features;
inputting the image features into a first bidirectional LSTM layer of the multi-layer classification model for time series processing to obtain a first intermediate result;
inputting the first intermediate result into a second bidirectional LSTM layer of the multi-layer classification model for time series processing to obtain a second intermediate result;
inputting the second intermediate result into an attention mechanism layer of the multi-layer classification model to perform attention enhancement, so as to obtain an attention enhancement result;
and inputting the attention enhancement result into a classifier of the multi-layer classification model to classify, so as to obtain the medical instrument fault class.
3. The method for detecting a fault class of a deep learning-based medical device according to claim 2, wherein the inputting the second intermediate result into the attention mechanism layer of the multi-layer classification model for attention enhancement, to obtain an attention enhancement result, comprises:
the second intermediate result is attentive by the following formula:
；
；
；
wherein ,for the j-th unnormalized attention score, a. About.>For the j-th attention weight, +.>For the attention enhancing result, +.>For the first parameter matrix, < >>For the second parameter matrix->For the third parameter matrix, T represents the matrix transpose operation, ">For inquiring the matrix +.>For the j-th key matrix, tanh represents the hyperbolic tangent function, softmax represents the normalized exponential function, ++>Is the j-th value matrix.
4. The fault class detection method of a deep learning-based medical device according to claim 1, wherein the preprocessing the vibration signal to obtain a preprocessed signal includes:
normalizing the vibration signal to obtain a normalized signal;
filtering the normalized signal to obtain a filtered signal;
and filling the missing value of the filtered signal to obtain the preprocessing signal.
5. The method for detecting a fault class of a deep learning-based medical device according to claim 1, wherein before the fault image is input into a multi-layer classification model for fault classification, the method further comprises training the model to be trained, and the training process comprises:
setting a minimum vector and a maximum vector according to the first supplemental signal;
randomly generating a target vector according to the minimum vector and the maximum vector;
obtaining differential weights, and generating disturbance vectors according to the differential weights and the target vectors;
setting the target vector or the disturbance vector as a test vector according to the cross discriminant;
detecting whether the fitness of the test vector is higher than that of the target vector, if so, replacing the target vector with the test vector;
adding Gaussian disturbance to the test vector to obtain a disturbance adding vector;
and detecting whether the current training times is greater than or equal to the maximum training times, and if not, randomly generating the target vector according to the minimum vector and the maximum vector.
6. The method of fault class detection for deep learning based medical devices of claim 1, wherein said randomly oversampling the vibration signal to generate a second supplemental signal comprises:
randomly selecting a first supplementary data point from the vibration signal, and calculating a first threshold value according to the oversampling ratio;
randomly copying the vibration data points of the first threshold value from the vibration signals to obtain intermediate signals;
and carrying out interpolation operation on the intermediate signal, adding random disturbance, and generating the second supplementary signal.
7. The fault class detection method of a deep learning-based medical device according to claim 1, wherein the interpolating the preprocessed signal to obtain an interpolated signal includes:
selecting different preprocessed data points in the preprocessed signal;
interpolation is performed on the different preprocessed data points using bilinear interpolation according to the following formula:
；
wherein ,for interpolation coefficients +.>For the ith pre-processed data point, +.>For the j-th pre-processed data point, +.>For the kth interpolated data point, i < j;
and combining all the preprocessed data points and all the interpolated data points into the interpolated signal.
8. The method for detecting a fault class of a deep learning-based medical device according to claim 7, wherein adding random disturbance to the interpolation signal to obtain a first supplemental signal comprises:
randomly generating a plurality of random disturbances through normal distribution; wherein the number of random perturbations is the same as the number of interpolated data points;
and adding each interpolation data point to the corresponding random disturbance to obtain the first supplementary signal.
9. The method for detecting a fault class of a deep learning-based medical device according to claim 1, wherein the converting the expansion signal into a fault image includes:
and performing short-time Fourier transform on the expansion signal to obtain the fault image.
10. A deep learning-based fault class detection system for a medical instrument, comprising:
the pretreatment module is used for obtaining vibration signals of the medical instrument, and carrying out pretreatment on the vibration signals to obtain pretreatment signals;
the interpolation operation module is used for carrying out interpolation operation on the preprocessing signals to obtain interpolation signals;
the random disturbance adding module is used for adding random disturbance into the interpolation signal to obtain a first supplementary signal;
the random oversampling module is used for carrying out random oversampling on the vibration signal and generating a second supplementary signal;
the signal combination module is used for splicing the first supplemental signal, the second supplemental signal and the vibration signal to obtain an expansion signal;
the fault classification module is used for converting the expansion signals into fault images, inputting the fault images into a multi-layer classification model for fault classification, and obtaining the fault types of the medical instruments; the multi-layer classification model is obtained by training a model to be trained by using the expanding signals.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310883905.7A CN116595444B (en) | 2023-07-19 | 2023-07-19 | Fault category detection method and system for medical instrument based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310883905.7A CN116595444B (en) | 2023-07-19 | 2023-07-19 | Fault category detection method and system for medical instrument based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116595444A true CN116595444A (en) | 2023-08-15 |
| CN116595444B CN116595444B (en) | 2023-09-26 |
Family
ID=87594168
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310883905.7A Active CN116595444B (en) | 2023-07-19 | 2023-07-19 | Fault category detection method and system for medical instrument based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116595444B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110657091A (en) * | 2019-09-20 | 2020-01-07 | 中国农业大学 | Circulating pump bearing fault diagnosis method and system |
| CN112633132A (en) * | 2020-12-18 | 2021-04-09 | 合肥工业大学 | Bearing fault diagnosis method and system based on two-dimensional vibration image enhancement |
| CN115510965A (en) * | 2022-09-22 | 2022-12-23 | 同济大学 | Bearing imbalance fault diagnosis method based on generated data fusion |
| CN115859165A (en) * | 2022-11-25 | 2023-03-28 | 中国矿业大学(北京) | Rolling bearing intelligent fault diagnosis method based on deep learning |
| WO2023096185A1 (en) * | 2021-11-29 | 2023-06-01 | 주식회사 써로마인드 | Method for diagnosing machine failure on basis of deep-learning by using sounds and vibrations, and diagnosis device using same |
| CN116226704A (en) * | 2022-12-05 | 2023-06-06 | 重庆邮电大学 | Ship variable rotation speed bearing fault diagnosis method based on multi-feature fusion and improved SheffeNetV 2 |
-
2023
- 2023-07-19 CN CN202310883905.7A patent/CN116595444B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110657091A (en) * | 2019-09-20 | 2020-01-07 | 中国农业大学 | Circulating pump bearing fault diagnosis method and system |
| CN112633132A (en) * | 2020-12-18 | 2021-04-09 | 合肥工业大学 | Bearing fault diagnosis method and system based on two-dimensional vibration image enhancement |
| WO2023096185A1 (en) * | 2021-11-29 | 2023-06-01 | 주식회사 써로마인드 | Method for diagnosing machine failure on basis of deep-learning by using sounds and vibrations, and diagnosis device using same |
| CN115510965A (en) * | 2022-09-22 | 2022-12-23 | 同济大学 | Bearing imbalance fault diagnosis method based on generated data fusion |
| CN115859165A (en) * | 2022-11-25 | 2023-03-28 | 中国矿业大学(北京) | Rolling bearing intelligent fault diagnosis method based on deep learning |
| CN116226704A (en) * | 2022-12-05 | 2023-06-06 | 重庆邮电大学 | Ship variable rotation speed bearing fault diagnosis method based on multi-feature fusion and improved SheffeNetV 2 |
Non-Patent Citations (1)
| Title |
|---|
| 王荟儒: "基于集成深度学习的滚动轴承故障诊断方法研究", 《中国优秀硕士学位论文全文数据库 工程科技II辑》, no. 1, pages 18 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116595444B (en) | 2023-09-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hakim et al. | A systematic review of rolling bearing fault diagnoses based on deep learning and transfer learning: Taxonomy, overview, application, open challenges, weaknesses and recommendations | |
| CN109974782B (en) | Equipment fault early warning method and system based on big data sensitive characteristic optimization selection | |
| CN111914883B (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
| CN110823574A (en) | Fault diagnosis method based on semi-supervised learning deep countermeasure network | |
| CN113405825B (en) | Belt conveyor fault diagnosis method based on sound signals | |
| CN113920375B (en) | Fusion characteristic typical load identification method and system based on combination of Faster R-CNN and SVM | |
| KR20090078075A (en) | Fault diagnosis of inductirn motors by dft and wavelet | |
| CN111189638B (en) | HMM and QPSO optimization algorithm-based bearing fault degree identification method | |
| Jiang et al. | Rolling bearing fault identification using multilayer deep learning convolutional neural network | |
| CN109993232A (en) | Turbine Vibrationfault Diagnosis method based on deep learning artificial neural network | |
| CN117104377B (en) | Intelligent management system and method for electric bicycle | |
| Islam et al. | Time–frequency envelope analysis-based sub-band selection and probabilistic support vector machines for multi-fault diagnosis of low-speed bearings | |
| CN116380445B (en) | Equipment state diagnosis method and related device based on vibration waveform | |
| CN114429152A (en) | Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption | |
| CN113611293A (en) | Mongolian data set expansion method | |
| Wei et al. | Intelligent fault diagnosis of rotating machinery using ICD and generalized composite multi-scale fuzzy entropy | |
| CN116595444B (en) | Fault category detection method and system for medical instrument based on deep learning | |
| CN113408578A (en) | Harmonic reducer fault diagnosis method and system based on generative countermeasure network | |
| CN116879910A (en) | Laser scanning distance measuring device and method thereof | |
| JP4827285B2 (en) | Pattern recognition method, pattern recognition apparatus, and recording medium | |
| CN115171730A (en) | Voice data processing method and system based on maximum LPP and global variance | |
| CN115753102A (en) | Bearing fault diagnosis method based on multi-scale residual error sub-domain adaptation | |
| Wu et al. | Time Series Classification Based on Adaptive Feature Adjustment and Multi-scale AGRes2Net | |
| Brandao et al. | Fault Diagnosis of Rotary Machines Using Machine Learning | |
| Hong et al. | Adaptive fusion based on physics-constrained dictionary learning for fault diagnosis of rotating machinery |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |