CN111967486A - Complex equipment fault diagnosis method based on multi-sensor fusion - Google Patents

Complex equipment fault diagnosis method based on multi-sensor fusion Download PDF

Info

Publication number
CN111967486A
CN111967486A CN202010488395.XA CN202010488395A CN111967486A CN 111967486 A CN111967486 A CN 111967486A CN 202010488395 A CN202010488395 A CN 202010488395A CN 111967486 A CN111967486 A CN 111967486A
Authority
CN
China
Prior art keywords
complex equipment
data
fault
neural network
fault diagnosis
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.)
Pending
Application number
CN202010488395.XA
Other languages
Chinese (zh)
Inventor
李军
江水
徐启胜
后麒麟
梁天
都竞
张殷日
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Anhui Sanheyi Information Technology Co Ltd
Original Assignee
Anhui Sanheyi Information Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Anhui Sanheyi Information Technology Co Ltd filed Critical Anhui Sanheyi Information Technology Co Ltd
Priority to CN202010488395.XA priority Critical patent/CN111967486A/en
Publication of CN111967486A publication Critical patent/CN111967486A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D21/00—Measuring or testing not otherwise provided for
    • G01D21/02—Measuring two or more variables by means not covered by a single other subclass
    • 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
    • G06N3/08—Learning 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
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Abstract

The invention discloses a complex equipment fault diagnosis method based on multi-sensor fusion, which comprises the following steps: firstly, key data of a plurality of sensors during the operation of the complex equipment are collected, manual extraction of features and expert experience are not needed, original signal data are stacked in a hierarchical mode and are directly input into a convolutional neural network model, signal state feature indexes are extracted in a self-adaptive mode, time and space information of the original data from the plurality of sensors are fully considered, analyzed fault results are output, and the fault of the complex equipment which operates in real time is obtained. The method solves the problem that a single sensor cannot acquire information comprehensively, simultaneously solves the problem that expert experience is needed for manually extracting the characteristic indexes, directly extracts information from the original signals, effectively improves the accuracy of the fault diagnosis method of the complex equipment, realizes the output of end-to-end diagnosis results, and can visually reflect the faults occurring when the complex equipment operates in real time.

Description

Complex equipment fault diagnosis method based on multi-sensor fusion
Technical Field
The invention belongs to the technical field of state monitoring of mechanical equipment, and particularly relates to an intelligent fault diagnosis method for complex equipment, which is used for realizing state monitoring and fault diagnosis of the complex equipment.
Technical Field
With the trend of informatization, enlargement and complication of mechanical equipment in recent years, the requirements on the capacity and efficiency of complex equipment are continuously increased, and the complex equipment plays an irreplaceable role in modern industries and is more and more critical. However, the damage is inevitable due to factors such as long-term severe use working conditions, complex external environment, mechanical abrasion and the like, and once a fault occurs, serious economic loss is caused. Therefore, the safety of the complex equipment is more and more emphasized by people, so that the health management of the operation stability of the working state of the complex equipment is carried out, and the state detection and fault diagnosis of the complex equipment are of great significance.
With the progress of sensor technology and communication technology and the rise of cloud computing technology, the number, the size and the types of detection signals of machine equipment are continuously increased, researchers can monitor various signals of equipment such as vibration, acoustic emission, temperature, pressure and current, and the like, and a foundation is provided for intelligent state evaluation of the machine equipment, but at present, for a traditional state evaluation method of complex equipment, some problems still exist:
(1) most evaluation methods rely on manual feature extraction, require a lot of signal processing and diagnosis expertise, and involve much manual work. Although these methods, such as support vector machine, hidden markov algorithm, etc., have been proven to be successful in extracting the required features and performing fault diagnosis, the performance of these methods depends on the quality of feature design, the manual design of features is difficult, expert experience is required, and the existing methods can only be used for specific types of faults or machines, and have poor adaptability to new data fault prediction.
(2) Most evaluation methods adopt simple threshold value alarming, namely alarming when data exceeds a certain fixed value, so as to achieve the purpose of fault diagnosis. However, when the threshold value is set improperly, false alarm or missing occurs easily, and thus it is difficult to set the threshold value accurately. Meanwhile, the simple threshold alarm has poor adaptability, cannot detect the specific fault state of the complex equipment, and is difficult to maintain and repair equipment.
(3) Most of evaluation mode data are only acquired from a single sensor, and the single sensor has the problems of incomplete and inaccurate equipment information acquisition, no equipment space information and the like; meanwhile, a single sensor can only collect one data type, and can only reflect relevant characteristics from a certain angle, so that equipment information cannot be comprehensively known, and therefore fault diagnosis precision is low and stability is poor.
Aiming at the problems, the invention provides a complex equipment fault diagnosis method based on multi-sensor fusion. The method has the advantages that a plurality of sensor data are collected firstly, the data are fused from the data level without making relevant transformation on signals, equipment information is known most originally from multiple angles based on the sensors, the fault location and diagnosis are more accurate due to the fusion of the data level, and the method is suitable for real-time processing and diagnosis. Meanwhile, a plurality of sensors can be adopted to monitor a plurality of characteristic quantities, the characteristic quantities are fused, the complementarity of the sensors can be fully utilized, the information of complex equipment in various aspects is obtained, and the monitoring instability is reduced.
Meanwhile, compared with the traditional classical machine learning method, the deep learning method of the convolutional neural network is adopted for intelligent fault diagnosis, manual experience intervention in signal feature extraction is not needed, the deep learning method is used for processing input multi-sensor data layer by layer, high-level features closely related to complex equipment faults are gradually extracted, and adaptive extraction of complex equipment state features is achieved. The extracted high-level features are learned in a self-adaptive mode, and a function mapping relation between data and faults is established, so that intelligent fault diagnosis is achieved for complex equipment.
Disclosure of Invention
In order to effectively solve the problems and promote intelligent evaluation of fault diagnosis of complex equipment, the invention is realized by a complex equipment fault diagnosis method based on multi-sensor fusion, which specifically comprises the following steps:
step S1, signal acquisition: the method comprises the steps of arranging various sensors on complex equipment at certain intervals or key positions of the complex equipment, and collecting relevant detection signals of the running states of the sensors.
Step S2, data processing: and carrying out noise reduction and preprocessing on data acquired by the multiple sensors, stacking the one-dimensional time sequences line by line to form a two-dimensional input matrix, and realizing data-level multiple sensor fusion in a combined mode.
Step S3, training the model: a deep learning algorithm is adopted, a Convolutional Neural Network (CNN) is built, a corresponding label matrix is built on the basis of input matrix data to serve as a training set, and a Convolutional Neural Network model is trained.
Step S4, failure diagnosis: inputting the acquired real-time multi-sensor signals of the complex equipment, performing multi-sensor data fusion according to the step S2, and outputting the label of the complex equipment with the fault by using the model established in the step S3, so as to obtain the specific fault of the complex equipment at present according to the label.
Further, step S2 is to stack the multi-sensor data, where each sensor data is a row, and multiple sensors are stacked in multiple columns, to construct a two-dimensional input matrix, and to establish the time information and the spatial information from the sensors, and in this way, to construct in the input matrix, to directly combine the original signals at the data level, and to directly extract features from the combined data, so that the extraction is more accurate and the effect is more excellent.
Further, the step S3 trains the model, establishes a corresponding label matrix as a training set based on the input matrix data by using the data processed in the step S2, and based on a deep learning algorithm, directly inputs the original data as a convolutional neural network, calculates an output value of each neuron in a forward direction, calculates an error from a true value, and then propagates in a reverse direction, calculates a gradient of a connection weight of each neuron, updates a weight of each layer according to a gradient descent rule, and ends training when a set recognition accuracy is reached, thereby obtaining a model having a fault diagnosis capability for complex equipment.
Preferably, a convolutional neural network is used, the network comprising convolutional layers, pooling layers, fully-connected layers; inputting complex equipment multi-sensor data in a two-dimensional matrix form, and performing convolution operation on a convolution layer to extract features from the complex equipment multi-sensor data; in the pooling layer, pooling operation is carried out to reduce data dimensionality, and meanwhile, depth feature extraction is carried out to avoid overfitting; and converting shallow features into more abstract high-level features layer by layer through multilayer convolution and pooling, outputting a result to a full-connection layer, integrating the features extracted by the convolution layer and the pooling layer, and outputting high-level features of data.
Preferably, the neuron output of the last full-link layer is a complex equipment fault diagnosis result, and the Softmax activation function is used for multi-class classification, and assuming a training set with K-class samples, the formula is:
Figure RE-GDA0002712310950000031
wherein, WjIs a convolution kernel weight matrix; bKA convolution kernel bias matrix;
Figure RE-GDA0002712310950000032
is a characteristic index of the l-th layer; and classifying the fault diagnosis result through a Softmax activation function.
Further, convolutional neural network training belongs to supervised learning, the error is calculated by adopting a Mean Absolute Error (MAE) loss function and an L2 regularization method, and the formula is as follows:
Figure RE-GDA0002712310950000033
wherein h isW,b(x(i)) The convolution neural network outputs complex equipment faults; y is(i)Is a complex equipment actual failure; i sample number (i is 1,2, …, m);
Figure RE-GDA0002712310950000034
is a convolution kernel weight matrix; b(i)Is the grid deviation; λ is the regularization parameter.
Further, the error is propagated from output to input layer by layer to carry out back propagation, the gradient of the error to the layer of parameters is calculated through derivation, the weight of each characteristic index is updated according to a gradient descent rule, the training set is used for continuous training, the weight of the loss minimization function is found, error minimization is achieved, when the set working state identification accuracy is achieved, the training is finished, and the training model with better performance is obtained.
Further, in the step S4, in the fault diagnosis, the acquired real-time multi-sensor signal of the complex equipment is processed according to the step S2, and then is input to the model trained in the step S3, and the state of the complex equipment at this time, that is, the corresponding tag value, is output. If the fault exists, the specific fault type and the fault reason can be known according to the label, and the fault diagnosis of the complex equipment can be effectively carried out through the model.
The invention has the following beneficial effects:
(1) the invention relates to a complex equipment fault diagnosis method based on multi-sensor fusion, which is characterized by collecting data of various types of sensors of complex equipment, stacking the data of the multiple sensors to construct an input matrix, constructing time and space information of equipment data, and fully utilizing original data information to obtain higher fault diagnosis accuracy.
(2) The invention relates to a complex equipment fault diagnosis method based on multi-sensor fusion, which combines a convolution neural network model in deep learning, realizes the self-adaptive feature extraction of feature indexes through the mapping and conversion process of a plurality of hidden layers of the model, carries out intelligent fault diagnosis on real-time complex equipment based on a trained model, outputs the fault to the complex equipment, obtains more accurate fault of the complex equipment, and is beneficial to the development of the complex equipment fault diagnosis towards the direction of intellectualization.
(3) According to the complex equipment fault diagnosis method based on multi-sensor fusion, the output full-connection layer uses the Softmax function, and the output result can be effectively classified, so that the fault existing in the complex equipment is obtained.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a data structure diagram of a complex equipment fault diagnosis method based on multi-sensor fusion.
FIG. 2 is a schematic diagram of a convolutional neural network structure of a complex equipment fault diagnosis method based on multi-sensor fusion.
FIG. 3 is a flow chart of an implementation of a complex equipment fault diagnosis method based on multi-sensor fusion.
Detailed Description
The present invention will be described in further detail with reference to specific embodiments in order to make the technical field better understand the scheme of the present invention. It is to be understood that the embodiments described are only a few embodiments of the invention, and not all embodiments. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
The present invention is described in further detail below with reference to the attached drawing figures.
The invention provides a complex equipment fault diagnosis method based on multi-sensor fusion, which promotes the implementation of intelligent evaluation of complex equipment fault diagnosis, takes a public available roller bearing condition data set collected from a motor driving system by western reserve university (CWRU) as an example, firstly arranges a plurality of sensors at key positions of complex equipment, collects relevant monitoring signals of the running state of the complex equipment, then adaptively extracts signal state characteristic indexes based on a built deep convolutional neural network model, performs weighted fusion on the characteristic indexes, and finally outputs fault classification results to obtain the fault of the equipment, and the implementation steps are as follows:
step S1, signal acquisition: in this example, three different position information of the device are acquired by using 3 vibration acceleration sensors: drive end, fan end and base. The relevant state of the equipment can be fully obtained from a plurality of angles, and the vibration overall appearance of the complex equipment is basically reflected.
Step S2, data processing: according to the method in the invention, data collected by three vibration acceleration sensors are stacked in three columns in a row, so that the dimension of each obtained data sample is 3 × 1200, and the data structure is shown in fig. 1. 3600 samples are in total in the data set, 400 samples are in each bearing condition, an input matrix is constructed, and the training set and the testing set are further divided.
Preferably, the fault types in step S2 are testing the bearing under different conditions, including normal condition, rolling element fault, outer ring fault and inner ring fault. Each defect type has three levels of severity. Single point failures were introduced into each bearing using 0.18, 0.36 and 0.54mm machining diameter failure ratings for each bearing. In this case, the data set includes nine failed bearing states, as shown in table 1:
table 1 description of the failure conditions
Type of failure Failure size (mm) Label (R)
Failure of rolling body 0.18 1
Failure of rolling body 0.36 2
Failure of rolling body 0.54 3
Outer ring failure 0.18 4
Outer ring failure 0.36 5
Outer ring failure 0.54 6
Inner ring failure 0.18 7
Inner ring failure 0.36 8
Inner ring failure 0.54 9
Step S3, training the model: and (3) building a deep convolutional neural network based on a deep learning algorithm, inputting a training set into a model, directly learning the multi-sensor original data, extracting a deep characteristic index, obtaining more accurate equipment data information, and training the convolutional neural network model.
Preferably, the convolutional neural network model described in step S3 is trained to build a convolutional neural network, which includes convolutional layers, pooling layers, and fully-connected layers, and neurons of adjacent layers are connected in different ways to realize layer-by-layer transmission of input information; the network in this example has 5 layers, of which the first 4 layers are: the winding layer 1, the pooling layer 1, the winding layer 2 and the pooling layer 2, and the rear layer 1 is a full-connection layer. The first 4 layers are a feature extraction optimization part, the last 1 layer is responsible for classifying and outputting results, the neurons of the adjacent layers are connected in different modes, layer-by-layer transmission of input sample information is achieved, and the structure is shown in fig. 2.
Relevant parameters of each layer of the neural network are set firstly and are listed in table 2:
TABLE 2 neural network-related parameter settings
Network architecture Parameters (convolution kernel width stride channel number)
Convolutional layer 1 [3，17，1，64]
Pooling layer 1 8x8
Convolutional layer 2 [1，8，64，128]
Pooling layer 2 4x4
Full connection layer [128，35，1，1]
And further, inputting the data of the multiple sensors in the form of an input matrix, sending the data into a convolution layer, performing left-to-right weighted sum on convolution kernels in the convolution neural network, setting the size of the convolution kernels, during convolution operation, enabling a convolution kernel sliding window to pass through each region of the input matrix, multiplying corresponding elements of the region after overturning, accumulating, learning parameters in the convolution kernels, extracting bottom layer characteristics, and sending the characteristics into a pooling layer.
Further, pooling still needs to be performed after convolution, the method selects the largest pooling layer to perform pooling operation, namely, the largest value is found in each region, the dimensionality of data is reduced, and depth feature extraction is performed at the same time, and the largest pooling layer can effectively reduce estimated mean difference caused by parameter errors of the convolution layer relative to other types of pooling layers.
Further, after convolution and pooling of 4 layers of convolution layers, the bottom layer characteristics are converted into more abstract high layer characteristics layer by layer, and the results are output to a full connection layer.
Preferably, in this example, a ReLU activation function is used, and the activation function introduces a non-linear factor to the neuron, so that the neural network can arbitrarily approximate any non-linear function, and compared with other activation functions, the ReLU activation function has a higher calculation speed, and is beneficial to updating the weight in a backward direction.
Further, the full-connection layer is connected with the output of the pooling layer 2, the extracted features of the previous layers are integrated, the features learned by the network are mapped into the solving space of the sample, the final result is output, a Softmax activation function is adopted, the function output value is a corresponding fault sample label, and the fault of the label corresponding to the complex equipment can be known.
Further, outputting the obtained result, calculating errors among the sample labels, calculating the errors based on the MAE loss function in the invention content, and transmitting the errors from the output to the input layer by layer for back propagation. Firstly, each parameter Wij (l)And b(i)Initializing a small random value close to zero, updating the weight of each characteristic index according to a gradient descent rule, training the parameters by data for multiple times, continuously updating the weight and the deviation, finding W and b of the minimized loss function, further realizing error minimization, and finishing the training.
Preferably, the MAE loss function formula is selected in the example, the quality of the prediction capability of the model is measured by calculating the error between the sample label and the result obtained by outputting, and the method has a good effect on the classification problem and is very suitable for the example.
Preferably, in the embodiment, an algorithm for performing first-order gradient optimization on the random objective function by using an Adam optimizer has better low-order matrix estimation adaptability, is easy to implement, and has high calculation efficiency and lower memory requirement; meanwhile, the method has the invariance to the angle scaling, so that the method is very suitable for solving the problem with large-scale data or parameters and is very suitable for the example.
Further, Dropout technique and L2 regularization are employed in this example:
preferably, the Dropout technology randomly disables the weights of some nodes of the network at a certain probability in the training process, in this example, 50% of the node weights are randomly selected to suspend for one time to disable the operation according to uniform distribution for each data calculation, so that the generalization capability of the network is improved, and the over-fitting phenomenon is avoided.
Preferably, the L2 regularization adds the sum of squares of the weighting parameters to the original loss function in order to limit the parameters to be too large or too large, making the model more complex and avoiding overfitting.
Further, experimental result analysis shows that after training of the training set, the complex equipment fault diagnosis model with good performance and capable of being represented is finally obtained.
Step S4, failure diagnosis: in this example, the test set obtained after the data processing in step S2 is input to the model established in step S3, so as to output a corresponding label of the fault and obtain a corresponding fault of the current motor drive system.
According to the embodiment, the invention discloses a complex equipment fault diagnosis method based on multi-sensor fusion. The multi-sensor fusion is carried out from the data angle, the information contained in the original data can be fully utilized, and the characteristic extraction capability of the information is improved. The convolutional neural network in deep learning is further combined, the excellent automatic characteristic index extraction and nonlinear mapping functions of the convolutional neural network are utilized, an expert experience basis is not needed, the design of key characteristic indexes is not needed to be selected manually, the signal state characteristic indexes are directly extracted from original signals in a self-adaptive mode, characteristic fusion is carried out, an Adam optimizer, random gradient descent and L2 regularization are adopted in the training process, the training efficiency is improved, the over-fitting phenomenon is prevented, and the recognition rate of fault diagnosis of complex equipment is effectively improved. Therefore, the method provided by the invention can effectively diagnose the fault of the complex equipment and is beneficial to the development of the complex equipment towards the direction of intellectualization.

Claims (7)

1. A complex equipment fault diagnosis method based on multi-sensor fusion is characterized by comprising the following steps:
step S1: and acquiring related detection signals of the running state of the complex equipment based on various sensors.
Step S2: and denoising and preprocessing the data acquired by the multi-sensor, and stacking the one-dimensional time sequences line by line to form a two-dimensional input matrix to realize data-level multi-sensor fusion.
Step S3: and establishing a corresponding label matrix based on the input matrix data as a training set based on a deep learning algorithm, and training the convolutional neural network model.
Step S4: and inputting real-time complex equipment multi-sensor signal data based on building a trained convolutional neural network model, and performing fault diagnosis on complex equipment to output a result.
2. Step S1, according to claim 1, wherein the complex equipment is disposed at certain intervals, or the complex equipment is disposed at critical positions, and the critical positions will affect the normal operation of the complex equipment if a fault occurs. Meanwhile, a plurality of sensors can be adopted to monitor a plurality of characteristic quantities, the characteristic quantities are fused, the complementarity of the sensors can be fully utilized, the information of complex equipment in various aspects is obtained, and the monitoring instability is reduced.
3. Step S2 is characterized in that, a plurality of sensor data are stacked in rows and columns to construct a two-dimensional data matrix, time information and space information from the sensors are established, and the signals are directly combined into original signals at data level without making relevant transformation, so as to directly extract features from the data.
4. Step S3 of claim 1, wherein a training convolutional neural network is built, a corresponding label matrix is built based on input matrix data as a training set, raw data is directly input as the convolutional neural network based on a deep learning algorithm, an output value of each neuron is calculated forward, an error from a true value is calculated, the error is propagated backward, a gradient of a connection weight of each neuron is calculated, weights of each layer are updated according to a gradient descent rule, and when a set working state recognition accuracy is reached, training is ended, and a model having a fault diagnosis capability for complex equipment is obtained.
5. Step S4 is characterized in that, after the collected real-time multi-sensor signals of the complex equipment are processed according to step S2, the signals are input into the model trained in step S3, and the state of the complex equipment at that time, i.e. the corresponding label value, is output. If the fault exists, the specific fault type and the fault reason can be known according to the label, and the fault diagnosis of the complex equipment can be effectively carried out through the model.
6. The convolutional neural network as claimed in claim 4, wherein the signal state characteristic index is extracted in a self-adaptive manner by using the excellent automatic characteristic index extraction and nonlinear mapping functions of the convolutional neural network without an expert experience basis and without manually selecting and designing key characteristic indexes, so that the recognition rate of the state of the shore bridge operating mechanism is effectively improved.
7. The convolutional neural network model set up according to claim 4, wherein the convolutional neural network comprises convolutional layers, pooling layers and full-connection layers, and the last full-connection layer uses a Softmax activation function and can output a classification result to a neuron, so that a fault diagnosis result of complex equipment is obtained.
CN202010488395.XA 2020-06-02 2020-06-02 Complex equipment fault diagnosis method based on multi-sensor fusion Pending CN111967486A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010488395.XA CN111967486A (en) 2020-06-02 2020-06-02 Complex equipment fault diagnosis method based on multi-sensor fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010488395.XA CN111967486A (en) 2020-06-02 2020-06-02 Complex equipment fault diagnosis method based on multi-sensor fusion

Publications (1)

Publication Number Publication Date
CN111967486A true CN111967486A (en) 2020-11-20

Family

ID=73357964

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010488395.XA Pending CN111967486A (en) 2020-06-02 2020-06-02 Complex equipment fault diagnosis method based on multi-sensor fusion

Country Status (1)

Country Link
CN (1) CN111967486A (en)

Cited By (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112465052A (en) * 2020-12-07 2021-03-09 重庆忽米网络科技有限公司 Equipment fault diagnosis report generation method and system based on convolutional neural network
CN112528548A (en) * 2020-11-27 2021-03-19 东莞市汇林包装有限公司 Self-adaptive depth coupling convolution self-coding multi-mode data fusion method
CN112633358A (en) * 2020-12-19 2021-04-09 武汉第二船舶设计研究所(中国船舶重工集团公司第七一九研究所) Semantic-level-feature-learning-based steam generator water supply system diagnosis method and system
CN112684012A (en) * 2020-12-02 2021-04-20 青岛科技大学 Equipment key force-bearing structural part fault diagnosis method based on multi-parameter information fusion
CN112706901A (en) * 2020-12-31 2021-04-27 华南理工大学 Semi-supervised fault diagnosis method for main propulsion system of semi-submerged ship
CN112784490A (en) * 2021-01-25 2021-05-11 西安交通大学 Sensor position optimization method of helicopter health management system
CN113326896A (en) * 2021-06-25 2021-08-31 国网上海市电力公司 Fusion sensing method based on multiple types of sensors
CN113405646A (en) * 2021-06-17 2021-09-17 润智科技有限公司 Distributed vibration identification method based on dual-channel phi-OTDR (optical time Domain reflectometer) underground optical cable
CN113624466A (en) * 2021-07-08 2021-11-09 中南民族大学 Steam turbine rotor fault diagnosis method, device, equipment and storage medium
CN114002334A (en) * 2021-09-29 2022-02-01 西安交通大学 Structural damage acoustic emission signal identification method and device and storage medium
CN114137358A (en) * 2021-11-17 2022-03-04 国网天津市电力公司信息通信公司 Power transmission line fault diagnosis method based on graph convolution neural network
CN114186617A (en) * 2021-11-23 2022-03-15 浙江大学 Mechanical fault diagnosis method based on distributed deep learning
CN114611636A (en) * 2022-05-11 2022-06-10 深圳市三江电气有限公司 Method for realizing measured value analysis by fusing information of various sensors
CN114662371A (en) * 2022-05-24 2022-06-24 深圳市信润富联数字科技有限公司 Knowledge distillation-based light PHM system implementation method, device and system
CN114964841A (en) * 2022-05-07 2022-08-30 中国地质大学（武汉） Mobile robot fault detection system and method based on multi-sensor fusion
CN115683235A (en) * 2022-12-28 2023-02-03 常州检验检测标准认证研究院 Industrial robot servo motor vibration fault detection method and device
CN116304819A (en) * 2023-03-15 2023-06-23 四川大学 Nuclear reactor operation condition judging method based on LeNet-5 algorithm
WO2023139081A1 (en) 2022-01-21 2023-07-27 Siemens Aktiengesellschaft Method for producing a monitoring apparatus of an automation installation
US11789110B2 (en) 2020-09-03 2023-10-17 Honeywell International Inc. Fault detection, exclusion, isolation, and re-configuration of navigation sensors using an abstraction layer
CN117077040A (en) * 2023-09-04 2023-11-17 武汉蓝海科创技术有限公司 Large-scale complex equipment fault diagnosis and prediction system based on machine learning

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105372087A (en) * 2015-11-30 2016-03-02 南通大学 Fault diagnosis method based on multi-sensor signal analysis
CN109763944A (en) * 2019-01-28 2019-05-17 中国海洋大学 A kind of contactless monitoring system of offshore wind turbine blade fault and monitoring method
CN111022313A (en) * 2019-12-30 2020-04-17 哈尔滨工程大学 Ocean platform air compressor fault diagnosis method based on LSTM

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105372087A (en) * 2015-11-30 2016-03-02 南通大学 Fault diagnosis method based on multi-sensor signal analysis
CN109763944A (en) * 2019-01-28 2019-05-17 中国海洋大学 A kind of contactless monitoring system of offshore wind turbine blade fault and monitoring method
CN111022313A (en) * 2019-12-30 2020-04-17 哈尔滨工程大学 Ocean platform air compressor fault diagnosis method based on LSTM

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
任凤娟: "多传感器信息融合技术在液压系统故障诊断中的应用", 《液压气动与密封》, vol. 39, no. 07, pages 52 - 55 *

Cited By (27)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11789110B2 (en) 2020-09-03 2023-10-17 Honeywell International Inc. Fault detection, exclusion, isolation, and re-configuration of navigation sensors using an abstraction layer
CN112528548A (en) * 2020-11-27 2021-03-19 东莞市汇林包装有限公司 Self-adaptive depth coupling convolution self-coding multi-mode data fusion method
CN112684012A (en) * 2020-12-02 2021-04-20 青岛科技大学 Equipment key force-bearing structural part fault diagnosis method based on multi-parameter information fusion
CN112465052A (en) * 2020-12-07 2021-03-09 重庆忽米网络科技有限公司 Equipment fault diagnosis report generation method and system based on convolutional neural network
CN112465052B (en) * 2020-12-07 2023-04-07 重庆忽米网络科技有限公司 Equipment fault diagnosis report generation method and system based on convolutional neural network
CN112633358A (en) * 2020-12-19 2021-04-09 武汉第二船舶设计研究所(中国船舶重工集团公司第七一九研究所) Semantic-level-feature-learning-based steam generator water supply system diagnosis method and system
CN112633358B (en) * 2020-12-19 2022-11-25 武汉第二船舶设计研究所(中国船舶重工集团公司第七一九研究所) Semantic-level-feature-learning-based steam generator water supply system diagnosis method and system
CN112706901B (en) * 2020-12-31 2022-04-22 华南理工大学 Semi-supervised fault diagnosis method for main propulsion system of semi-submerged ship
CN112706901A (en) * 2020-12-31 2021-04-27 华南理工大学 Semi-supervised fault diagnosis method for main propulsion system of semi-submerged ship
CN112784490A (en) * 2021-01-25 2021-05-11 西安交通大学 Sensor position optimization method of helicopter health management system
CN113405646A (en) * 2021-06-17 2021-09-17 润智科技有限公司 Distributed vibration identification method based on dual-channel phi-OTDR (optical time Domain reflectometer) underground optical cable
CN113326896A (en) * 2021-06-25 2021-08-31 国网上海市电力公司 Fusion sensing method based on multiple types of sensors
CN113624466B (en) * 2021-07-08 2023-10-03 中南民族大学 Method, device, equipment and storage medium for diagnosing turbine rotor faults
CN113624466A (en) * 2021-07-08 2021-11-09 中南民族大学 Steam turbine rotor fault diagnosis method, device, equipment and storage medium
CN114002334A (en) * 2021-09-29 2022-02-01 西安交通大学 Structural damage acoustic emission signal identification method and device and storage medium
CN114137358A (en) * 2021-11-17 2022-03-04 国网天津市电力公司信息通信公司 Power transmission line fault diagnosis method based on graph convolution neural network
CN114137358B (en) * 2021-11-17 2023-04-28 国网天津市电力公司信息通信公司 Power transmission line fault diagnosis method based on graph convolution neural network
CN114186617A (en) * 2021-11-23 2022-03-15 浙江大学 Mechanical fault diagnosis method based on distributed deep learning
WO2023139081A1 (en) 2022-01-21 2023-07-27 Siemens Aktiengesellschaft Method for producing a monitoring apparatus of an automation installation
DE102022200694A1 (en) 2022-01-21 2023-08-10 Siemens Aktiengesellschaft Manufacturing method for a monitoring device for an automation system
CN114964841A (en) * 2022-05-07 2022-08-30 中国地质大学（武汉） Mobile robot fault detection system and method based on multi-sensor fusion
CN114611636A (en) * 2022-05-11 2022-06-10 深圳市三江电气有限公司 Method for realizing measured value analysis by fusing information of various sensors
CN114662371A (en) * 2022-05-24 2022-06-24 深圳市信润富联数字科技有限公司 Knowledge distillation-based light PHM system implementation method, device and system
CN115683235A (en) * 2022-12-28 2023-02-03 常州检验检测标准认证研究院 Industrial robot servo motor vibration fault detection method and device
CN116304819A (en) * 2023-03-15 2023-06-23 四川大学 Nuclear reactor operation condition judging method based on LeNet-5 algorithm
CN117077040A (en) * 2023-09-04 2023-11-17 武汉蓝海科创技术有限公司 Large-scale complex equipment fault diagnosis and prediction system based on machine learning
CN117077040B (en) * 2023-09-04 2024-02-23 武汉蓝海科创技术有限公司 Large-scale complex equipment fault diagnosis and prediction system based on machine learning

Similar Documents

Publication Publication Date Title
CN111967486A (en) Complex equipment fault diagnosis method based on multi-sensor fusion
Cheng et al. A convolutional neural network based degradation indicator construction and health prognosis using bidirectional long short-term memory network for rolling bearings
CN108334936B (en) Fault prediction method based on migration convolutional neural network
Wang et al. A method for rapidly evaluating reliability and predicting remaining useful life using two-dimensional convolutional neural network with signal conversion
Duan et al. Deep learning enabled intelligent fault diagnosis: Overview and applications
CN113255848B (en) Water turbine cavitation sound signal identification method based on big data learning
Zhang et al. A novel feature adaptive extraction method based on deep learning for bearing fault diagnosis
Yin et al. Wasserstein generative adversarial network and convolutional neural network (WG-CNN) for bearing fault diagnosis
CN109977869B (en) Shore bridge running mechanism state evaluation method based on deep learning
CN115018021B (en) Machine room abnormity detection method and device based on graph structure and abnormity attention mechanism
CN111562108A (en) Rolling bearing intelligent fault diagnosis method based on CNN and FCMC
CN113505655A (en) Bearing fault intelligent diagnosis method for digital twin system
CN111353373A (en) Correlation alignment domain adaptive fault diagnosis method
Zhang et al. Multi-model ensemble deep learning method for intelligent fault diagnosis with high-dimensional samples
CN112836604A (en) Rolling bearing fault diagnosis and classification method, system and equipment based on VMD-SSAE and storage medium thereof
CN116593157A (en) Complex working condition gear fault diagnosis method based on matching element learning under small sample
CN114091525A (en) Rolling bearing degradation trend prediction method
CN114266301A (en) Intelligent power equipment fault prediction method based on graph convolution neural network
CN116718377A (en) Bearing fault diagnosis method based on wavelet transformation and depth residual error attention mechanism
CN114819315A (en) Bearing degradation trend prediction method based on multi-parameter fusion health factor and time convolution neural network
CN114118138A (en) Bearing composite fault diagnosis method based on multi-label field self-adaptive model
CN113984389A (en) Rolling bearing fault diagnosis method based on multi-receptive-field and improved capsule map neural network
CN116738868A (en) Rolling bearing residual life prediction method
CN117272196A (en) Industrial time sequence data anomaly detection method based on time-space diagram attention network
CN108898157B (en) Classification method for radar chart representation of numerical data based on convolutional neural network

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