CN117142009B - Scraper conveyor health state assessment method based on graph rolling network - Google Patents
Scraper conveyor health state assessment method based on graph rolling network Download PDFInfo
- Publication number
- CN117142009B CN117142009B CN202311412381.XA CN202311412381A CN117142009B CN 117142009 B CN117142009 B CN 117142009B CN 202311412381 A CN202311412381 A CN 202311412381A CN 117142009 B CN117142009 B CN 117142009B
- Authority
- CN
- China
- Prior art keywords
- scraper conveyor
- data
- graph
- health
- node
- 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.)
- Active
Links
- 230000036541 health Effects 0.000 title claims abstract description 144
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000005096 rolling process Methods 0.000 title claims abstract description 27
- 238000012544 monitoring process Methods 0.000 claims abstract description 67
- 230000003993 interaction Effects 0.000 claims abstract description 32
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 48
- 238000009826 distribution Methods 0.000 claims description 30
- 230000006870 function Effects 0.000 claims description 20
- 239000003638 chemical reducing agent Substances 0.000 claims description 16
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 230000003862 health status Effects 0.000 claims description 12
- 238000010276 construction Methods 0.000 claims description 10
- 238000013507 mapping Methods 0.000 claims description 8
- 238000012952 Resampling Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 5
- 230000006866 deterioration Effects 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 abstract description 23
- 238000010586 diagram Methods 0.000 abstract description 4
- 230000008878 coupling Effects 0.000 abstract description 2
- 238000010168 coupling process Methods 0.000 abstract description 2
- 238000005859 coupling reaction Methods 0.000 abstract description 2
- 230000002708 enhancing effect Effects 0.000 abstract 1
- 238000012854 evaluation process Methods 0.000 abstract 1
- 239000003245 coal Substances 0.000 description 7
- 238000005065 mining Methods 0.000 description 7
- 238000012545 processing Methods 0.000 description 7
- 239000000306 component Substances 0.000 description 6
- 238000013527 convolutional neural network Methods 0.000 description 6
- 230000005540 biological transmission Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000003595 spectral effect Effects 0.000 description 5
- 238000003745 diagnosis Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- NAXKFVIRJICPAO-LHNWDKRHSA-N [(1R,3S,4R,6R,7R,9S,10S,12R,13S,15S,16R,18S,19S,21S,22S,24S,25S,27S,28R,30R,31R,33S,34S,36R,37R,39R,40S,42R,44R,46S,48S,50R,52S,54S,56S)-46,48,50,52,54,56-hexakis(hydroxymethyl)-2,8,14,20,26,32,38,43,45,47,49,51,53,55-tetradecaoxa-5,11,17,23,29,35,41-heptathiapentadecacyclo[37.3.2.23,7.29,13.215,19.221,25.227,31.233,37.04,6.010,12.016,18.022,24.028,30.034,36.040,42]hexapentacontan-44-yl]methanol Chemical compound OC[C@H]1O[C@H]2O[C@H]3[C@H](CO)O[C@H](O[C@H]4[C@H](CO)O[C@H](O[C@@H]5[C@@H](CO)O[C@H](O[C@H]6[C@H](CO)O[C@H](O[C@H]7[C@H](CO)O[C@@H](O[C@H]8[C@H](CO)O[C@@H](O[C@@H]1[C@@H]1S[C@@H]21)[C@@H]1S[C@H]81)[C@H]1S[C@@H]71)[C@H]1S[C@H]61)[C@H]1S[C@@H]51)[C@H]1S[C@@H]41)[C@H]1S[C@H]31 NAXKFVIRJICPAO-LHNWDKRHSA-N 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000013210 evaluation model Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000010006 flight Effects 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 230000033001 locomotion Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000007790 scraping Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000012733 comparative method Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000036961 partial effect Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 230000002829 reductive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000000547 structure data Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G19/00—Conveyors comprising an impeller or a series of impellers carried by an endless traction element and arranged to move articles or materials over a supporting surface or underlying material, e.g. endless scraper conveyors
- B65G19/18—Details
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G43/00—Control devices, e.g. for safety, warning or fault-correcting
-
- 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
- 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
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G06N3/042—Knowledge-based neural networks; Logical representations of neural 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
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
- G06N3/048—Activation functions
-
- 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
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G2203/00—Indexing code relating to control or detection of the articles or the load carriers during conveying
- B65G2203/02—Control or detection
- B65G2203/0266—Control or detection relating to the load carrier(s)
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G2203/00—Indexing code relating to control or detection of the articles or the load carriers during conveying
- B65G2203/04—Detection means
- B65G2203/042—Sensors
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Mechanical Engineering (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention provides a scraper conveyor health state assessment method based on a graph rolling network. The method comprises the following steps: determining monitoring indexes of the scraper conveyor according to the structure and the working flow of the scraper conveyor, and acquiring index data by using a sensor; constructing a sensor interaction diagram structure based on a KNN algorithm; the acquired index data is processed by a variation self-encoder to obtain a scraper conveyor health state index curve, and the acquired index data is classified into different grades according to the scraper conveyor health state index curve; the sensor interaction graph structure is input into the graph rolling network GCN for training. According to the invention, the coupling between the sensor state data is considered in the evaluation process, the space information between the sensors is extracted, and the characteristic information of multiple sensors is fused better. Meanwhile, in the aspect of the division of the health states of the scraper conveyor, the method relieves the problems of excessive human intervention and data uncertainty, thereby enhancing the rationality and accuracy of the evaluation result.
Description
Technical Field
The invention relates to the technical field of fully mechanized mining, in particular to a scraper conveyor health state assessment method based on a graph rolling network.
Background
The scraper conveyor is used as one of important equipment for realizing mechanized coal mining on a working face, bears the key task of conveying raw coal outwards from the working face, is suitable for comprehensive intelligent development, and continuously develops towards long-distance, large-capacity, high-reliability and intelligent directions, and meanwhile, the scraper groove is used as a track for supporting normal operation of the coal mining machine. Thus, the scraper conveyor occupies a very important position in the equipment composition of the fully mechanized mining face. In high efficiency, high yield fully mechanized coal mining face, the scraper conveyor is faced with complex load changes and severe conditions. The normal operation and the safe use of the mine shaft are significant for ensuring the production efficiency of the mine shaft and improving the benefit of enterprises. Therefore, it is important to identify the current health status of the scraper conveyor in time. Through the accurate identification of the health state of the scraper conveyor, the fault prediction and preventive maintenance management of the scraper conveyor can be realized. This may increase reliability and availability of the device, reduce the probability of failure, increase the life of the device, and reduce production downtime and maintenance costs due to the failure.
At present, information reflecting the state of the scraper conveyor is collected through various sensors, and the health management of the scraper conveyor is gradually trended by using signal processing, deep learning, artificial intelligence and other emerging technologies, and Yang Junshe proposes a method for predicting the abrasion of a middle groove of the scraper conveyor based on PSO-CNN. The convolutional neural network CNN structure suitable for wear prediction is constructed, the weight of CNN is evaluated and optimized by using a particle swarm algorithm PSO, and the network is prevented from being trapped into local optimization. And establishing a scraper conveyor fault diagnosis model based on a fuzzy neural network in Guozhong, and researching the basis of fuzzy clustering and the learning flow of the RBF neural network. The original mind proposes a fault diagnosis method for the bearing change working condition of the scraper conveyor speed reducer based on subspace learning SL. According to the method, firstly, an original signal is subjected to fast Fourier transform, then, the spectrum energy of the original signal is mapped to a high-dimensional space by utilizing a principal component analysis method, a training data and test data nuclear energy subspace is obtained, and finally, the fault type of the rolling bearing is classified by utilizing a support vector machine SVM classifier. Wang Jinhui the principle, structure and learning algorithm of the depth sparse coding are researched, and the depth sparse self-coding network is applied to fault diagnosis of the rolling bearing of the scraper conveyor speed reducer. Ma Hailong a fuzzy expert system based on fuzzy theory was constructed. The structure of the expert system is provided, and the fuzzy relation matrix and the membership function are provided through empirical data and expert experience, so that the fuzzy relation expression of the fault symptoms and the fault reasons of the scraper conveyor speed reducer is realized. However, the conventional convolutional network, LSTM, fully connected network, etc. generally perform data processing based on a regular grid structure, and cannot fully capture the complex topological structure of the scraper conveyor and the nonlinear and multi-modal relationship between the sensor data. A scraper conveyor can generally be seen as a graph or topology in which nodes represent different components or sensors and edges represent the relationship between them. Traditional convolutional networks, LSTM and fully-connected networks do not take into account the connection relationships between nodes, and cannot accurately model and capture the influence and dependency relationships between nodes. Scraper conveyors are typically made up of a large number of nodes and edges, and conventional convolutional networks, LSTM, and fully-connected networks are very computationally and memory intensive in handling large-scale graphs, resulting in inefficiency. Meanwhile, when a new sample or an unseen abnormal situation is processed, there may be a problem of insufficient generalization capability. They are typically trained on small sample data and, for complex scraper conveyor complete machine health assessment tasks, are generalized to perform poorly in large-scale, complex situations.
The disadvantages of the traditional convolutional network, LSTM and fully-connected network can be avoided by using the graph convolution neural network, but when the health state of the scraper conveyor is identified by using the graph convolution neural network, the identification accuracy of the graph neural network on the health state of the scraper conveyor is not high because of the fact that too much human participation occurs when the label is added to the data of the scraper conveyor and irrelevant node association is introduced if a connection strategy is not proper when the graph structure is constructed.
Disclosure of Invention
The invention provides a scraper conveyor health state assessment method based on a graph rolling network, which can solve the problem in the prior art that the accuracy of identifying the health state of the scraper conveyor by a graph neural network is not high due to the fact that too much human participation occurs when labels are added to data of the scraper conveyor and irrelevant nodes are not properly introduced if a connection strategy is not proper when a graph structure is constructed.
The invention provides a scraper conveyor health state assessment method based on a graph rolling network, which comprises the following steps:
determining a scraper conveyor monitoring index, and acquiring scraper conveyor monitoring index data by using a sensor corresponding to the scraper conveyor monitoring index;
constructing a sensor interaction graph structure, wherein the sensor interaction graph comprises: the connection relation among the nodes formed by the sensors and the monitoring index data of the scraper conveyor contained in each node;
inputting the collected monitoring index data of the scraper conveyor into a trained variable self-encoder to obtain a health state index curve of the scraper conveyor; dividing monitoring index data of the scraper conveyor into different grades according to a health state index curve of the scraper conveyor, and distributing grade labels for the monitoring index data;
inputting the sensor interaction graph structure into a graph rolling network GCN, and training the graph rolling network GCN by minimizing the difference between the health state recognition result of the sensor interaction graph structure and the grade label distributed by the scraper conveyor monitoring index data in the sensor interaction graph structure;
and inputting the sensor interaction graph structure of the scraper conveyor to be detected into a trained graph rolling network GCN to identify the health state of the scraper conveyor.
Further, the scraper conveyor monitoring index includes:
blade temperature C 11 Vibration C of scraper 12 Blade crack C 13 The method comprises the steps of carrying out a first treatment on the surface of the Chain vibration C 21 Degree of tightness C of chain 22 Degree of chain deformation C 23 The method comprises the steps of carrying out a first treatment on the surface of the Sprocket bearing vibration C 31 Sprocket bearing temperature C 32 Oil C of chain wheel 33 Sprocket bearing torque C 34 The method comprises the steps of carrying out a first treatment on the surface of the Rail temperature C 41 Vibration C of guide rail 42 The method comprises the steps of carrying out a first treatment on the surface of the Vibration C of speed reducer bearing 51 Oil C of speed reducer 52 Bearing torque C of speed reducer 53 Speed reducer sound C 54 Motor current C 55 Temperature C of motor 56 。
Further, the constructing a sensor interaction graph structure includes:
obtaining a node set consisting of N sensors:
（1）
where V denotes a node set made up of N sensors, where,representing the node formed by the ith sensor, wherein N is the monitoring index number of the scraper conveyor;
obtaining the connection relation between nodes formed by the sensors through a K neighbor algorithm:
（2）
wherein E represents a compound represented byAn edge set formed by edges, each edge representing an interaction relationship between the sensors;
constructing a sensor interaction graph structure based on the node set and the connection relation between the nodes:
（3）
wherein,
representing a feature matrix formed of N nodes, wherein,characteristic vector representing the i-th node, < +.>Is the feature vector of node i at time T, T being the number of time steps;
is an adjacency matrix if node->Node->With influence, node->And node->There is a directed edge between, adjoining elements in matrix A +.>Otherwise, let(s)>。
Further, the obtaining the connection relation between the nodes formed by the sensors through the K-nearest neighbor algorithm includes:
using K-nearest neighbor method, based on the eigenvector of each sensor nodeK adjacent nodes of each sensor node are found;
further, the step of inputting the collected monitoring index data of the scraper conveyor into the trained variation self-encoder to obtain a health state index curve of the scraper conveyor comprises the following steps:
obtaining a trained variation self-encoder;
mapping the collected scraper conveyor monitoring index data into one-dimensional health indexes by using the trained variable self-encoder, and outputting the mean value and variance of the probability distribution of the one-dimensional health indexes:
（4）
（5）
（6）
（7）
wherein,
x is the feature matrix of the input and,a nonlinear activation function;
is the variation from the first of the encoders in the encoder>Output result of layer hidden layer, < >>And->Are respectively->The weight matrix and bias of the layer hiding layer, L is the total layer number of the encoder in the variable self-encoder;
w and b are respectively a weight matrix and bias of probability distribution fitting;
and->The mean value and the variance of the probability distribution of the one-dimensional health index are respectively;
sampling in a one-dimensional health index probability distribution by a resampling method:
（8）
wherein,
is distributed from normal>Data obtained by sampling;
z is resampled data, and the expression of the formula (8) means that the resampled data is compressed by an encoder from original data;
fitting the resampled data to obtain a scraper conveyor health state index curve.
Further, the obtaining trained variational self-encoder includes:
constructing a variation self-encoder model, including an encoder and a decoder;
the encoder comprises a plurality of neural network layers and is used for mapping input scraper conveyor monitoring index data into one-dimensional health indexes;
the decoder is used for mapping the data obtained by resampling back to the original data space to generate reconstruction data;
calculating a gap between the reconstruction data and the input scraper conveyor monitoring index data by using a loss function, wherein the loss function is as follows:
（9）
wherein,is one-dimensional health index probability distribution->Is a standard normal distribution,/->The KL divergence between the one-dimensional health index probability distribution and the standard normal distribution; />The error between the reconstruction data and the input monitoring index data of the scraper conveyor;
minimizing the loss function by an optimization algorithm, and obtaining a trained variational self-encoder when the loss function converges.
Further, the classifying the monitoring index data of the scraper conveyor into different grades according to the health state index curve of the scraper conveyor comprises the following steps:
curve section of health state index of scraper conveyor，/>) The corresponding scraper conveyor monitoring index data are divided into health data, and the grade label is 0;
curve section of health state index of scraper conveyor，/>) The corresponding scraper conveyor monitoring index data are divided into good data, and the grade label is 1;
curve section of health state index of scraper conveyor，/>) The corresponding scraper conveyor monitoring index data are divided into deterioration data, and the grade label is 2;
curve section of health state index of scraper conveyor，/>) The corresponding scraper conveyor monitoring index data are divided into fault data, and the grade label is 3;
wherein, ~/>、/>~/>the grading threshold value is obtained by combining the healthy state index curve of the scraper conveyor, the real running condition of the scraper conveyor and expert experience.
Further, the judging result of the graph rolling network GCN on the input sensor interaction graph structure includes:
the sensor interaction graph structure is input into a graph rolling network GCN to conduct graph convolution:
（10）
wherein,
，/>is an N-order identity matrix,>is a diagonal matrix with diagonal elements +.>；/>Is an activation function; />Is a parameter matrix of the network to be learned, r represents the dimension of the output,/->Is a node characteristic matrix, which is a node characteristic matrix constructed by taking out each node characteristic at the current moment in the characteristic matrix X, and is +.>Is the output matrix of the primary graph convolution;
carrying out multiple graph convolutions, and cascading or summing each node characteristic in an output matrix Z obtained by the multiple graph convolutions to obtain the characteristic of each graph structure;
inputting the characteristics of each graph structure into a softmax classifier to perform health state recognition, and outputting a health state recognition result, wherein the health state recognition result comprises the following steps: healthy, good, worsening, malfunctioning.
The embodiment of the invention provides a scraper conveyor health state assessment method based on a graph rolling network, which has the following beneficial effects compared with the prior art:
the invention obtains the index curve from the encoder based on the variation, and automatically marks the original data instead of relying on a tedious and subjective manual marking process. The variable self-encoder can automatically grasp potential characteristics between normal and abnormal states by learning a large amount of scraper conveyor data, thereby realizing automatic learning of the characteristics and modes of the scraper conveyor. Thus, by using the variance from the health indicator generated by the encoder, the present invention is able to more accurately tag the raw data. This not only improves the accuracy of the label, but also eliminates the subjective complexity of the manual marking process. The method solves the problem that labels are inaccurate due to excessive manual intervention in the construction of health indexes, and further improves the identification accuracy of the graph neural network.
Besides, the sensor interaction graph structure is constructed through the K neighbor algorithm, and the K neighbor method is to find K neighbor nodes of each sensor node according to the feature vector of each sensor node, so that the adjacent K sensor nodes are connected to capture the relation and similarity between the sensor nodes, and the spatial information between the sensors is extracted, so that the sensor interaction graph structure constructed based on the K neighbor algorithm is input into the graph roll-up neural network GNN, the graph roll-up neural network GNN is facilitated to better understand the relevance between the nodes, the multi-sensor information fusion is completed, and the recognition accuracy of the graph roll-up neural network GNN is higher.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention. In the drawings:
FIG. 1 is a flow chart of a VAE-based health index construction for a scraper conveyor provided herein;
FIG. 2 is a schematic diagram of a multi-layer index system for health assessment of a scraper conveyor provided herein;
FIG. 3 is a flow chart of the scraper conveyor health identification based on a graph rolling network provided herein;
FIG. 4 is a schematic view of a health index profile of the scraper conveyor provided herein;
FIG. 5 is a schematic diagram of the sensor interaction structure at different k values provided in the present specification.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, but it should be understood that the protection scope of the present invention is not limited by the specific embodiments.
Examples
A scraper conveyor health state assessment method based on a graph rolling network specifically comprises the following steps:
step 1, constructing health indexes of scraper conveyor based on VAE
Step 1.1, analysis of the action of the Key Components of the scraper conveyor
The key parts of the scraper conveyor are divided into a scraper, a chain wheel, a guide rail, a sliding block and a transmission system.
A scraper blade: the flights are the core components of the flight conveyor for scraping material from beneath the conveyor. The health of the blade can be identified by monitoring its temperature, vibration, wear, and crack indicators. For example, a sensor may be used to monitor the vibration frequency and amplitude of the screed.
Chain and sprocket: the chain and sprocket are used to drive the movement of the flights. It can monitor the tightness degree of the chain, the abrasion degree of the chain wheel and the vibration condition of the chain. For example, vibration sensors or accelerometers may be used to monitor the vibration of the chain.
Guide rail and slider: the guide rail and the slide block are used for supporting and guiding the movement of the scraping plate. Their health status can be achieved by monitoring the wear, deformation and surface temperature parameters of the guide rail and the slide block. The infrared thermal imager, the camera or the temperature sensor and other devices can be used for real-time monitoring.
A transmission system: the transmission system of the scraper conveyor comprises a motor, a speed reducer, a bearing and the like. The health status of these critical components can be determined by monitoring current, temperature, vibration, noise, etc. For example, sensors may be used to monitor the current and temperature of the motor, and their health status may be identified and predicted by a fault diagnosis algorithm or model.
Step 1.2, selecting principle of monitoring index of scraper conveyor
The invention takes the scraper conveyor as a research object, the scraper conveyor is used for guaranteeing the coal mining efficiency, the dimension of the parameters of the state monitoring of each device is extremely large, and if all the monitoring parameters are used as the input indexes of the evaluation model, the evaluation efficiency is affected and the evaluation workload is increased. Therefore, in the selection of the evaluation index of the health state of the scraper conveyor, each selected index must be reasonable and enough to represent the health state of the scraper conveyor, so that the rationality and completeness of the evaluation system can be ensured. Therefore, the evaluation of the health state of the scraper conveyor is carried out efficiently, the evaluation system of the health state of the coal mine equipment to be built is scientific and reasonable, and key indexes which have important influence on the health state of the scraper conveyor are selected through expert experience and various state parameter correlation analysis methods at the present stage, so that the evaluation working efficiency is improved. In the selection of state index data, the following principles should be followed:
(1) Scientific principle
The rationality of the index parameters of the scraper conveyor is that the selected index is scientific, namely that the index should express the working principle and objective practical condition of the scraper conveyor. Meanwhile, scientificity and uniformity of a calculation method, a conversion relation and the like of the parameter indexes are guaranteed, so that the selected indexes are standard, objective and scientific.
(2) Principle of applicability
The selected health state index of the scraper conveyor needs to be applicable, and a practical acquisition channel is needed, so that workers can measure the health state index through a normal approach method. Meanwhile, the sensor has practical operability, and the sensor is convenient and reasonable to install, so that data acquisition and data transmission are facilitated.
(3) Principle of testability
The state index parameters of the scraper conveyor are determined for realizing the evaluation of the health state of the scraper conveyor, and the state characteristic attribute of the scraper conveyor is required to be quantitatively represented, so that the index parameters can be obtained by a measuring instrument, a statistical method or a mathematical expression and the like.
(4) Consistency principle
The index object is ensured to be consistent with the evaluation target when the index parameter of the scraper conveyor is selected, namely, all the selected indexes can express the current health state of the scraper conveyor to a certain extent.
(5) Representative principle
The selected index parameters should not be too much, so that the selection of representative index parameters is required, thereby ensuring that the selected index data can all participate.
In the process of selecting the evaluation indexes, all index parameters meeting the standards are selected by combining the five principles and expert experience, so that preparation is made for the health status recognition of the scraper conveyor below.
Step 1.3, selecting monitoring indexes of the scraper conveyor
The scraper conveyor is a machine integrating electric, mechanical and hydraulic systems and the key component functions of the scraper conveyor are analyzed in the steps. Therefore, on the basis of referring to domestic and foreign documents, collecting data and consulting specialists, the invention combines the five principles and the self structure of the scraper conveyor, screens out most indexes which have great influence on the health state of the scraper conveyor, and determines all the health state evaluation indexes of the scraper conveyor. The monitoring index of the scraper conveyor constructed in the specification is shown in figure 1, a scraper conveyor health state assessment index system is obtained according to the principle, and then the variation self-encoder obtains the scraper conveyor health state index. Fig. 2 shows all evaluation indexes, and the scraper conveyor health status evaluation index system C specifically includes:
scraper C 1 : blade temperature C 11 Vibration C of scraper 12 Blade crack C 13 ；
Chain C 2 : chain vibration C 21 Degree of tightness C of chain 22 Degree of chain deformation C 23 ；
Sprocket C 3 : sprocket bearing vibration C 31 Sprocket wheelBearing temperature C 32 Oil C of chain wheel 33 Sprocket bearing torque C 34 ；
Guide rail C 4 : rail temperature C 41 Vibration C of guide rail 42 ；
Transmission system C 5 : vibration C of speed reducer bearing 51 Oil C of speed reducer 52 Bearing torque C of speed reducer 53 Speed reducer sound C 54 Motor current C 55 Temperature C of motor 56 。
Step 1.4, constructing a health state index curve of the scraper conveyor based on the VAE
The variant self-encoder VAE is a depth generation model that describes the observation of potential space by way of probability, unlike conventional self-encoders. The VAE-based flight conveyor health index construction has a number of advantages. First, it can automatically learn the potential representation between the normal and abnormal conditions of the scraper conveyor. By learning a large amount of flight conveyor data, the VAE is able to automatically learn the characteristics and patterns of the flight conveyor from the data, describing its potential characteristics in a probabilistic distribution manner, without requiring manual definition of the characteristics or rules. Secondly, in the evaluation of the health state of the scraper conveyor, the monitoring index of the scraper conveyor based on the VAE can better describe the evolution process of the equipment from health to degradation to failure through the grading of the health state of the scraper conveyor, and provide real-time and visual qualitative description for the state of the equipment. The existing dividing method depends on priori knowledge and expert experience, and lacks objective scraper conveyor state sample grade information. To solve this problem, the present invention employs the VAE to obtain the health index of the scraper conveyor, and the potential space of the VAE has continuity, which means that similar health states are also close in the potential space, and the health states are further divided by controlling and analyzing the potential space. The construction method can improve the objectivity and accuracy of the health state assessment of the scraper conveyor and provide powerful support for equipment operation and maintenance.
Step 1.4.1, obtaining a trained variable self-encoder
Constructing a variation self-encoder model, including an encoder and a decoder;
the encoder comprises a plurality of neural network layers and is used for mapping input scraper conveyor monitoring index data into one-dimensional health indexes;
the decoder is used for mapping the data obtained by resampling back to the original data space to generate reconstruction data;
calculating the difference between the reconstruction data and the input monitoring index data of the scraper conveyor by using a loss function, wherein in the training process, the optimization target of the part is to reduce the loss of a variation self-encoder through a gradient descent algorithm, the loss is realized through maximizing a variation lower bound, namely, minimizing the sum of a KL divergence term and a reconstruction error term, the KL divergence term is the KL divergence between the probability distribution of a health index z and the standard normal distribution of prior distribution, the reconstruction error term is the error between the monitoring parameter and the original data, and finally the optimized loss function of the part is as follows:
wherein,is one-dimensional health index probability distribution->Is a standard normal distribution,/->The KL divergence between the one-dimensional health index probability distribution and the standard normal distribution; />The error between the reconstruction data and the input monitoring index data of the scraper conveyor;
minimizing the loss function by an optimization algorithm, and obtaining a trained variational self-encoder when the loss function converges.
Step 1.4.2, obtaining a health state index curve based on the VAE model
The input of this part is obtained in step 1.3The variation self-encoder fuses and compresses the 18 paths of monitoring signals into one-dimensional health indexes through a full-connection layer to finally generate the average value of probability distribution of the health indexesVariance->Its dimensions are allIn order to ensure that the model is guided, the health index with robustness is obtained by sampling in the probability distribution of the health index through a resampling method. The specific calculation formula is as follows:
wherein,
x is the feature matrix of the input and,representing a feature matrix formed by N nodes, wherein, < +.>Characteristic vector representing the i-th node, < +.>Is the feature vector of node i at time T, T being the number of time steps;
a nonlinear activation function;
is the variation from the first of the encoders in the encoder>Output result of layer hidden layer, < >>And->Are respectively->The weight matrix and bias of the layer hiding layer, L is the total layer number of the encoder in the variable self-encoder;
w and b are respectively a weight matrix and bias of probability distribution fitting;
and->The mean value and the variance of the probability distribution of the one-dimensional health index are respectively;
is distributed from normal>Data obtained by sampling;
z is resampled data, meaning that the resampled data is compressed by the encoder from the original data; finally, sampling to obtain a robust scraper conveyor health state index curve.
Step 1.4.3, label is allocated for the original data
In combination with expert experience, the health of the scraper conveyor can be classified into different classes according to these index curves, and each class is assigned a respective label:
curve section of health state index of scraper conveyor，/>) The corresponding scraper conveyor monitoring index data are divided into health data, and the grade label is 0;
curve section of health state index of scraper conveyor，/>) The corresponding scraper conveyor monitoring index data are divided into good data, and the grade label is 1;
curve section of health state index of scraper conveyor，/>) The corresponding scraper conveyor monitoring index data are divided into deterioration data, and the grade label is 2;
curve section of health state index of scraper conveyor，/>) The corresponding scraper conveyor monitoring index data are divided into fault data, and the grade label is 3.
Step 2, identifying health state of scraper conveyor based on graph rolling network
Step 2.1, model frame
The health state evaluation framework of the scraper conveyor based on the graph convolution network, which is provided by the invention, is shown in fig. 3, and a graph sample is constructed by means of a graph data representation method. By calculating the degree of similarity between the sensor state data (using euclidean distance), the sensors are mapped to nodes in the graph, and a graph structure between the sensors is constructed using KNN algorithm to show the spatial relationship between the sensors. The graph structure is regarded as the structure information of the graph sample, and the state data acquired by each sensor is regarded as the attribute characteristics of the nodes, thereby completing the construction of the graph sample. Finally, the pattern book at each moment is input into the graph rolling network GCN to complete the health state evaluation of the pattern samples at each moment.
Step 2.2, pattern book structure of scraper conveyor
The invention reflects the interaction structure between the sensors through the graph, and the graph provides an intuitive visual mode to display the connection and interaction relationship between the sensors. The sensors are mapped into nodes in the graph, and the relation modes among the sensors can be clearly presented through the modes of the nodes and the edges. The invention constructs the picture sampleWhereinRepresenting a node set consisting of N sensor nodes, < >>The representation is composed ofEach side represents whether interaction exists between sensors, and each node in the network is attached with descriptive characteristic information, which is expressed as a characteristic matrix X, and node +.>Is expressed as +.>The feature set of all N nodes is denoted +.>. The trajectory of node i is denoted +.>Where T is the number of time steps. Finally, all track data are recorded as。/>Is an adjacency matrix of the network if the node +.>Node->With influence, a directed edge exists between the adjacent matrix elements +.>Otherwise, let(s)>The adjacency matrix reflects the direct acting relationship between the sensor nodes.
The adjacency matrix of the graph sample is determined by a K-nearest neighbor algorithm, and the distance between two points is measured in a feature space for measuring the distance using euclidean distance. First, full life cycle characteristic information of the sensor node is extracted. K neighbor nodes for a given node are then found from these features using the K neighbor (KNN) method. And constructing a connecting edge among K sensor nodes closest to the current sensor node to obtain a graph structure. Finally, the graph structure is regarded as the structure information of the graph sample, and the state data acquired by each sensor is regarded as the attribute characteristics of the nodes, so that the construction of the graph sample is completed.
Step 2.3, health State identification
According to the invention, the health state assessment of the scraper conveyor is completed by using a GCN (Graph Convolutional Network) algorithm, the GCN is a deep learning model for processing the structural data of the graph, each convolution layer consists of a plurality of convolution units, and the parameters of each convolution unit layer can be continuously and iteratively updated through a back propagation algorithm. When the graph data structure is subjected to similar convolution operation, the graph data structure can be divided into a spectrum domain graph convolution and a space domain graph convolution. Spectral domain convolution, as the name implies, convolves the graph data in spectral space, and spatial domain convolution, as the name implies, convolves the graph data directly in time domain space. Spectral domain map convolution while data is first de-processed from the frequency domain perspective, spectral domain map convolution is not suitable for use with directed maps, and the present invention uses spatial domain map convolution because the application of spectral domain map convolution only exists in undirected maps.
In terms of the flight conveyor health assessment, the GCN utilizes the ability of the graph structure data processing to model and analyze the relationships between the various components or parts of the flight conveyor, providing a more comprehensive system level assessment. By representing individual components or parts of the scraper conveyor as nodes in the graph and considering the connection relationships between them, the GCN is able to learn and extract useful representations of the health of the scraper conveyor from the characteristics and topology information of these nodes, and thus learn complex relationships between the topology and characteristics of the scraper conveyor from the sensor data, thereby improving the accuracy and reliability of the health.
The input of the graph rolling network of the invention is node characteristic matrix in non-European spaceAnd adjacency matrix between nodes->. n represents the number of nodes and h represents the feature dimension of the node. The primary convolution is calculated as follows:
wherein,
，/>is an N-order identity matrix,>is a diagonal matrix with diagonal elements +.>；/>Is an activation function; />Is a parameter matrix of the network to be learned, r represents the dimension of the output,/->Is a node characteristic matrix, which is a node characteristic matrix constructed by taking out each node characteristic at the current moment in the characteristic matrix X, and is +.>Is the output matrix of the primary graph convolution;
carrying out multiple graph convolutions, and cascading or summing each node characteristic in an output matrix Z obtained by the multiple graph convolutions to obtain the characteristic of each graph structure;
inputting the characteristics of each graph structure into a softmax classifier to perform health state recognition, and outputting a health state recognition result, wherein the health state recognition result comprises the following steps: healthy, good, worsening, malfunctioning.
The specification also provides experimental verification of a scraper conveyor health state assessment method based on a graph rolling network, which comprises the following steps:
1. data sources and experimental environment
The invention is based on the Weibull distribution rule, combines Pu Bai existing partial scraper conveyor state data, determines Weibull distribution parameters, simulates scraper conveyor operation state data by using Python software, and completes experimental verification according to real data and simulation data. And finally, 2500 pieces of scraper conveyor operation data of real and simulated 18 paths of monitoring signals are added.
2. Data preprocessing
Since the dimensions of 18 signals are not uniform, normalization or normalization processing is generally required for the case of non-uniform signal dimensions for use in the model. In this way, the consistency of the value ranges of different signals can be ensured, and the excessive or insufficient influence of certain signals on the training and prediction of the model is avoided. The invention adopts normalization processing to complete signal dimension unification, and the specific processing formula is as follows:
3. constructing distribution labels based on health metrics
And inputting the processed 18-dimensional monitoring parameter data into a VAE network for health index construction. The number 1 of output nodes of the final coding network corresponds to the final health state index of the scraper conveyor. The final obtained health index curve of the scraper conveyor is shown in fig. 4.
According to the constructed health state indexes, the health state of the scraper conveyor is divided into 4 grades by combining the real running condition of the scraper conveyor and expert experience, and each grade and the corresponding health state index interval are shown in the table 1:
TABLE 1 health status correspondence description of scraper conveyor
And based on the health index interval of the health state level, carrying out sample division on the original monitoring data. The result is 609 "healthy" samples, 1717 "good" samples, 162 "bad" samples, and 12 "faulty" samples.
4. Contrast method
According to the invention, 4 main stream health state identification methods are selected for comparison experiments. The methods selected can be divided into two categories. The first two methods are machine learning methods, and the second two methods are traditional deep learning methods. The following is a detailed description of the comparative method.
XGBoost: the method comprises the steps of screening out a coal mining machine health state evaluation index through calculating a correlation coefficient between state parameters; the whole sample is trained by adopting an XGBoost integrated learning algorithm, and proper parameters are selected by a cross-validation method to establish an optimal model.
PCA-SVM: the method adopts PCA algorithm to reduce the dimension of the obtained multidimensional state parameters, and finally uses SVM to complete the health state identification.
ICNN: according to the method, the self-encoder is used for realizing data dimension reduction and feature extraction through unsupervised training and noise reduction, so that the health state index is constructed; and then training and improving a convolutional neural network model according to the noise-reduced monitoring data and the health state indexes to realize automatic identification of the health state.
1DCNN: the method is based on a one-dimensional convolutional neural network to construct a health state evaluation model, and the obtained health evaluation index vector is input into the model to perform training and health state evaluation.
5. Health status recognition results
According to the invention, after all samples are disordered, 90% of the samples are taken as training set samples, and the rest 10% are taken as test set samples. The obtained accuracy, micro-F1 and Macro-F1 results are shown in Table 2. From the results, the VAE-GCN method provided by the invention has the best effect, because the method effectively extracts the space information among the monitoring parameters, and better fusion among the multiple sensor parameters is achieved by using the GCN.
TABLE 2 health status recognition results
6. Effects of different K values in graph sample construction
In order to verify the influence of different k values on an evaluation result when a graph structure is constructed by using a k nearest neighbor algorithm, 5 k values are taken, namely 2, 4, 6, 8 and 10 respectively, and finally the evaluation result corresponding to the different k values is shown in a table 3, the result obtained by using 5 k values is shown in a table 5, and the invention can see that the evaluation effect is best when the k value is taken for 8, and the accuracy is lower when the k value is smaller than 8, because a plurality of sensor nodes interacted with the current sensor node are ignored, so that information is lost; when the k value is greater than 8, the result is not slightly degraded because the information of the sensor node which has not interacted with the current sensor node is aggregated to the current sensor node, thus causing the addition of error information.
TABLE 3 health status recognition results at different k values
In order to intuitively observe the interaction condition among the sensor nodes, the invention draws a graph structure diagram under different k values, as shown in fig. 5, when k is equal to 2, two clusters appear in the graph, so that a plurality of sensors do not interact with each other; when k=4 and 6, there is less interaction between the sensors, and part of the information is missing; when k=8, combining the above health status recognition results, the interactive structure is optimal; when k=10, more edges are obviously seen in the graph, resulting in information redundancy.
7. Conclusion(s)
1) The health state index of the scraper conveyor is constructed based on the VAE, so that noise interference caused by data inclusion of the scraper conveyor is considered, and unsupervised training is adopted, and the problem that the recognition accuracy is low due to too much artificial participation in the construction of the health state index can be effectively solved.
2) The health state identification method based on the GCN model and with the multi-sensor monitoring data effectively fused is provided, the spatial information among the sensors is effectively extracted by using the graph structure information obtained by the K nearest neighbor algorithm, the multi-sensor information fusion is effectively completed by using the graph rolling network GCN, and finally, the interactive structure among the sensors is visually displayed, so that the problems that the health state identification is difficult due to the fact that the monitoring parameters of the scraper conveyor are multiple and the monitoring parameters have coupling are successfully solved.
The above examples illustrate only a few embodiments of the invention, which are described in detail and are not to be construed as limiting the scope of the invention. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention. Accordingly, the scope of protection of the present invention is to be determined by the appended claims.
Claims (5)
1. A method for assessing the health status of a scraper conveyor based on a graph roll-up network, comprising:
determining a scraper conveyor monitoring index, and acquiring scraper conveyor monitoring index data by using a sensor corresponding to the scraper conveyor monitoring index;
the scraper conveyor monitoring index comprises:
blade temperature C 11 Vibration C of scraper 12 Blade crack C 13 The method comprises the steps of carrying out a first treatment on the surface of the Chain vibration C 21 Degree of tightness C of chain 22 Degree of chain deformation C 23 The method comprises the steps of carrying out a first treatment on the surface of the Sprocket bearing vibration C 31 Sprocket bearing temperature C 32 Oil C of chain wheel 33 Sprocket bearing torque C 34 The method comprises the steps of carrying out a first treatment on the surface of the Rail temperature C 41 Vibration C of guide rail 42 The method comprises the steps of carrying out a first treatment on the surface of the Vibration C of speed reducer bearing 51 Oil C of speed reducer 52 Bearing torque C of speed reducer 53 Speed reducer sound C 54 Motor current C 55 Temperature C of motor 56 ；
Constructing a sensor interaction graph structure, wherein the sensor interaction graph comprises: the connection relation among the nodes formed by the sensors and the monitoring index data of the scraper conveyor contained in each node;
the construction sensor interaction graph structure comprises:
obtaining a node set consisting of N sensors:
V＝{v 1 ,v 2 ,…,v i ,…,v N } (1)
wherein V represents a node set composed of N sensors, wherein V i Representing the node formed by the ith sensor, wherein N is the monitoring index number of the scraper conveyor;
obtaining the connection relation between nodes formed by the sensors through a K neighbor algorithm:
wherein E represents a set of edges consisting of |E| edges, each edge representing an interaction relationship between sensors;
constructing a sensor interaction graph structure based on the node set and the connection relation between the nodes:
G＝(V,E,X,T,A) (3)
wherein,
X＝{X 1 ,X 2 ,…,X i ,…,X N -representing a feature matrix formed by N nodes, wherein,feature vector representing the i-th node, +.>Is the feature vector of node i at time T, T being the number of time steps;
A∈R N×N is an adjacency matrix if node v i Opposite node v j With influence, node v i And node v j There is a directed edge between adjacent elements a in matrix a ij =1, otherwise, a ij ＝0；
Inputting the collected monitoring index data of the scraper conveyor into a trained variable self-encoder to obtain a health state index curve of the scraper conveyor; comprising the following steps:
obtaining a trained variation self-encoder;
mapping the collected scraper conveyor monitoring index data into one-dimensional health indexes by using the trained variable self-encoder, and outputting the mean value and variance of the probability distribution of the one-dimensional health indexes:
Y 1 ＝σ(Xw 1 +b 1 ) (4)
Y l ＝σ(Y l-1 w l +b l ),l＝2,…,L (5)
u HI ＝Y l w+b (6)
σ HI ＝Y l w+b (7)
wherein,
x is an input feature matrix, and sigma (°) is a nonlinear activation function;
Y l is the output result of the hidden layer of the first layer in the encoder in the variable self-encoder, w l And b l The weight matrix and the bias of the first hidden layer are respectively, and L is the total layer number of the encoders in the variable self-encoder;
w and b are respectively a weight matrix and bias of probability distribution fitting;
u HI sum sigma HI The mean value and the variance of the probability distribution of the one-dimensional health index are respectively;
sampling in a one-dimensional health index probability distribution by a resampling method:
z＝u HI +σ HI ⊙∈ (8)
wherein,
e is data sampled from a standard normal distribution N (0,I);
z is resampled data, and the expression of the formula (8) means that the resampled data is compressed by an encoder from original data;
fitting the resampled data to obtain a scraper conveyor health state index curve;
dividing monitoring index data of the scraper conveyor into different grades according to a health state index curve of the scraper conveyor, and distributing grade labels for the monitoring index data;
inputting the sensor interaction graph structure into a graph rolling network GCN, and training the graph rolling network GCN by minimizing the difference between the health state recognition result of the sensor interaction graph structure and the grade label distributed by the scraper conveyor monitoring index data in the sensor interaction graph structure;
and inputting the sensor interaction graph structure of the scraper conveyor to be detected into a trained graph rolling network GCN to identify the health state of the scraper conveyor.
2. The method for estimating health of a scraper conveyor based on a graph rolling network according to claim 1, wherein the obtaining the connection relation between the nodes constituted by the sensors by the K-nearest neighbor algorithm comprises:
using K-nearest neighbor method, according to the characteristic vector X of each sensor node i K adjacent nodes of each sensor node are found;
k sensor nodes closest to each sensor node are connected.
3. The method of claim 1, wherein the obtaining trained variance self-encoders comprises:
constructing a variation self-encoder model, including an encoder and a decoder;
the encoder comprises a plurality of neural network layers and is used for mapping input scraper conveyor monitoring index data into one-dimensional health indexes;
the decoder is used for mapping the data obtained by resampling back to the original data space to generate reconstruction data;
calculating a gap between the reconstruction data and the input scraper conveyor monitoring index data by using a loss function, wherein the loss function is as follows:
L vae ＝-KL[q(z)||N(0,I)]+E q(z|x) (log(p(x|z))) (9)
wherein q (z) isOne-dimensional health index probability distribution, N (0,I) is a standard normal distribution, KL [ q (z) ||N (0,I)]The KL divergence between the one-dimensional health index probability distribution and the standard normal distribution; e (E) q(z|x) (log (p (x|z))) is the error of the reconstruction data from the inputted flight conveyor monitoring index data;
minimizing the loss function by an optimization algorithm, and obtaining a trained variational self-encoder when the loss function converges.
4. The method for estimating health of a scraper conveyor based on a graph roll-up network according to claim 1, wherein the classifying the scraper conveyor monitoring index data into different levels according to the scraper conveyor health index curve comprises:
curve section alpha of health state index of scraper conveyor 0 ，β 0 ) The corresponding scraper conveyor monitoring index data are divided into health data, and the grade label is 0;
curve section alpha of health state index of scraper conveyor 1 ，β 1 ) The corresponding scraper conveyor monitoring index data are divided into good data, and the grade label is 1;
curve section alpha of health state index of scraper conveyor 2 ，β 2 ) The corresponding scraper conveyor monitoring index data are divided into deterioration data, and the grade label is 2;
curve section alpha of health state index of scraper conveyor 3 ，β 3 ) The corresponding scraper conveyor monitoring index data are divided into fault data, and the grade label is 3;
wherein alpha is 0 ～α 3 、β 0 ～β 3 The grading threshold value is obtained by combining the healthy state index curve of the scraper conveyor, the real running condition of the scraper conveyor and expert experience.
5. The method for estimating a health state of a scraper conveyor based on a graph roll-up network according to claim 1, wherein the determining result of the graph roll-up network GCN on the input sensor interaction graph structure includes:
the sensor interaction graph structure is input into a graph rolling network GCN to conduct graph convolution:
wherein,
I N is an N-order identity matrix,>is a diagonal matrix with diagonal elements +.>Sigma is the activation function; />Is a parameter matrix of the network to be learned, r represents the dimension of the output,/->Is a node characteristic matrix, which is a node characteristic matrix constructed by taking out each node characteristic at the current moment in the characteristic matrix X, and is +.>Is the output matrix of the primary graph convolution;
carrying out multiple graph convolutions, and cascading or summing each node characteristic in an output matrix Z obtained by the multiple graph convolutions to obtain the characteristic of each graph structure;
inputting the characteristics of each graph structure into a softmax classifier to perform health state recognition, and outputting a health state recognition result, wherein the health state recognition result comprises the following steps: healthy, good, worsening, malfunctioning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311412381.XA CN117142009B (en) | 2023-10-30 | 2023-10-30 | Scraper conveyor health state assessment method based on graph rolling network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311412381.XA CN117142009B (en) | 2023-10-30 | 2023-10-30 | Scraper conveyor health state assessment method based on graph rolling network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117142009A CN117142009A (en) | 2023-12-01 |
| CN117142009B true CN117142009B (en) | 2024-03-15 |
Family
ID=88908434
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311412381.XA Active CN117142009B (en) | 2023-10-30 | 2023-10-30 | Scraper conveyor health state assessment method based on graph rolling network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117142009B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117932322B (en) * | 2024-03-22 | 2024-06-04 | 山东冠县鑫恒祥面业有限公司 | Flour equipment fault diagnosis method and system |
| CN118062489B (en) * | 2024-04-19 | 2024-09-10 | 山东科技大学 | Chain transmission tension pulsation real-time monitoring and controlling system of scraper conveyor |
Citations (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR1602068A (en) * | 1968-01-02 | 1970-10-05 | ||
| AU4217593A (en) * | 1992-08-10 | 1994-02-17 | Pit Top Mining Pty Ltd | Conveyor scraping system |
| DE29721296U1 (en) * | 1997-12-02 | 1998-02-05 | SMD-Fördertechnik GmbH, 42477 Radevormwald | Articulated chain and continuous conveyor |
| US6142291A (en) * | 1998-12-31 | 2000-11-07 | Sidney Manufacturing Company | Self-cleaning inclined section for drag conveyor |
| US6170642B1 (en) * | 1997-05-30 | 2001-01-09 | Investronica Sistemas S.A. | Aerial system for transportation, storage, classification and control of light products |
| JP2012147126A (en) * | 2011-01-07 | 2012-08-02 | Fujifilm Corp | Density unevenness correction value calculation method and image processing method and image processing apparatus |
| CN103213820A (en) * | 2013-04-11 | 2013-07-24 | 湖北博尔德科技股份有限公司 | Intelligent detection method and equipment for bulk embedded scraper conveying equipment |
| CN103662636A (en) * | 2013-11-29 | 2014-03-26 | 西安重装蒲白煤矿机械有限公司 | Chute of self-lubricating wear-resistant light scraper conveyor and manufacturing method for chute |
| CN104003141A (en) * | 2014-04-21 | 2014-08-27 | 中国矿业大学 | Device and method for wirelessly monitoring scraper conveyer |
| CN104444227A (en) * | 2014-09-30 | 2015-03-25 | 宁夏天地奔牛实业集团有限公司 | Broken chain monitor device and method for coal face scraper conveyer |
| CN106115178A (en) * | 2016-07-27 | 2016-11-16 | 西安重装蒲白煤矿机械有限公司 | A kind of combination type connection end and preparation method thereof |
| AU2018101509A4 (en) * | 2018-02-21 | 2018-12-13 | Wear Systems Solutions Pty Ltd | Conveyor belt wear assessment and conveyor belt coating |
| CN109911550A (en) * | 2019-04-17 | 2019-06-21 | 华夏天信（北京）智能低碳技术研究院有限公司 | Scratch board conveyor protective device based on infrared thermal imaging and visible light video analysis |
| CN110344882A (en) * | 2019-07-24 | 2019-10-18 | 精英数智科技股份有限公司 | Monitor method, system and the storage medium of drag conveyor staff operation |
| CN110422588A (en) * | 2019-07-29 | 2019-11-08 | 精英数智科技股份有限公司 | Monitoring method, device, equipment and the storage medium of drag conveyor |
| CN111735670A (en) * | 2020-07-15 | 2020-10-02 | 辽宁机电职业技术学院 | Belt middle sampler and control method thereof |
| CN111924427A (en) * | 2020-07-16 | 2020-11-13 | 精英数智科技股份有限公司 | Scraper conveyor floating chain fault identification method and device and electronic equipment |
| US10882700B1 (en) * | 2019-06-26 | 2021-01-05 | China University Of Mining And Technology | Embedded scraper rotation angle detection device for scraper conveyor and detection method |
| CN112193727A (en) * | 2020-09-11 | 2021-01-08 | 徐州工程学院 | Scraper conveyor health management system based on multi-sensing fusion and operation method |
| CN112215352A (en) * | 2020-09-28 | 2021-01-12 | 武汉科技大学 | BP neural network algorithm based system and method for predicting and warning deviation of conveying belt |
| CN113184483A (en) * | 2021-04-22 | 2021-07-30 | 武汉菲舍控制技术有限公司 | Anti-tearing system and early warning method for conveying belt |
| CN113375933A (en) * | 2021-05-31 | 2021-09-10 | 中国矿业大学 | Fault diagnosis system and method for scraper conveyor |
| CN113567159A (en) * | 2021-06-17 | 2021-10-29 | 太原理工大学 | Scraper conveyor state monitoring and fault diagnosis method based on edge cloud cooperation |
| US11353889B1 (en) * | 2021-06-16 | 2022-06-07 | Beta Air, Llc | System for electric aircraft navigation |
| CN114940343A (en) * | 2022-05-24 | 2022-08-26 | 无锡雪浪数制科技有限公司 | Double-chain scraper conveyor anomaly detection method based on deep learning |
| CN115272980A (en) * | 2022-09-22 | 2022-11-01 | 常州海图信息科技股份有限公司 | Conveying belt surface detection method and system based on machine vision |
| CN115577309A (en) * | 2022-09-27 | 2023-01-06 | 上海交通大学 | Belt conveyor sudden fault early warning method and system |
| CN116142726A (en) * | 2023-03-22 | 2023-05-23 | 山西潞安矿业(集团)有限责任公司 | Fault monitoring and early warning system of scraper conveyor chain transmission system |
| CN116576363A (en) * | 2023-04-17 | 2023-08-11 | 中国矿业大学 | Device and method for recognizing coal and rock through audio frequency, hyperspectrum and camera fusion |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9067728B2 (en) * | 2012-03-01 | 2015-06-30 | Clasen Quality Coatings, Inc. | Container including a magnetically operated scraper |
| PL3241646T3 (en) * | 2016-05-03 | 2019-05-31 | Comau Spa | Production line |
| EP3530360A1 (en) * | 2018-02-21 | 2019-08-28 | Siemens Aktiengesellschaft | Detection of non-conveyable mail items |
| CA3039328A1 (en) * | 2018-04-06 | 2019-10-06 | Sst Systems, Inc. | Conveyor system with automated carriers |
| CA3041600A1 (en) * | 2019-04-25 | 2020-10-25 | The Dun & Bradstreet Corporation | Machine learning classifier for identifying internet service providers from website tracking |
| US11363909B2 (en) * | 2020-04-15 | 2022-06-21 | Air Products And Chemicals, Inc. | Sensor device for providing control for a food processing system |
-
2023
- 2023-10-30 CN CN202311412381.XA patent/CN117142009B/en active Active
Patent Citations (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR1602068A (en) * | 1968-01-02 | 1970-10-05 | ||
| AU4217593A (en) * | 1992-08-10 | 1994-02-17 | Pit Top Mining Pty Ltd | Conveyor scraping system |
| US6170642B1 (en) * | 1997-05-30 | 2001-01-09 | Investronica Sistemas S.A. | Aerial system for transportation, storage, classification and control of light products |
| DE29721296U1 (en) * | 1997-12-02 | 1998-02-05 | SMD-Fördertechnik GmbH, 42477 Radevormwald | Articulated chain and continuous conveyor |
| US6142291A (en) * | 1998-12-31 | 2000-11-07 | Sidney Manufacturing Company | Self-cleaning inclined section for drag conveyor |
| JP2012147126A (en) * | 2011-01-07 | 2012-08-02 | Fujifilm Corp | Density unevenness correction value calculation method and image processing method and image processing apparatus |
| CN103213820A (en) * | 2013-04-11 | 2013-07-24 | 湖北博尔德科技股份有限公司 | Intelligent detection method and equipment for bulk embedded scraper conveying equipment |
| CN103662636A (en) * | 2013-11-29 | 2014-03-26 | 西安重装蒲白煤矿机械有限公司 | Chute of self-lubricating wear-resistant light scraper conveyor and manufacturing method for chute |
| CN104003141A (en) * | 2014-04-21 | 2014-08-27 | 中国矿业大学 | Device and method for wirelessly monitoring scraper conveyer |
| CN104444227A (en) * | 2014-09-30 | 2015-03-25 | 宁夏天地奔牛实业集团有限公司 | Broken chain monitor device and method for coal face scraper conveyer |
| CN106115178A (en) * | 2016-07-27 | 2016-11-16 | 西安重装蒲白煤矿机械有限公司 | A kind of combination type connection end and preparation method thereof |
| AU2018101509A4 (en) * | 2018-02-21 | 2018-12-13 | Wear Systems Solutions Pty Ltd | Conveyor belt wear assessment and conveyor belt coating |
| CN109911550A (en) * | 2019-04-17 | 2019-06-21 | 华夏天信（北京）智能低碳技术研究院有限公司 | Scratch board conveyor protective device based on infrared thermal imaging and visible light video analysis |
| US10882700B1 (en) * | 2019-06-26 | 2021-01-05 | China University Of Mining And Technology | Embedded scraper rotation angle detection device for scraper conveyor and detection method |
| CN110344882A (en) * | 2019-07-24 | 2019-10-18 | 精英数智科技股份有限公司 | Monitor method, system and the storage medium of drag conveyor staff operation |
| CN110422588A (en) * | 2019-07-29 | 2019-11-08 | 精英数智科技股份有限公司 | Monitoring method, device, equipment and the storage medium of drag conveyor |
| CN111735670A (en) * | 2020-07-15 | 2020-10-02 | 辽宁机电职业技术学院 | Belt middle sampler and control method thereof |
| CN111924427A (en) * | 2020-07-16 | 2020-11-13 | 精英数智科技股份有限公司 | Scraper conveyor floating chain fault identification method and device and electronic equipment |
| CN112193727A (en) * | 2020-09-11 | 2021-01-08 | 徐州工程学院 | Scraper conveyor health management system based on multi-sensing fusion and operation method |
| CN112215352A (en) * | 2020-09-28 | 2021-01-12 | 武汉科技大学 | BP neural network algorithm based system and method for predicting and warning deviation of conveying belt |
| CN113184483A (en) * | 2021-04-22 | 2021-07-30 | 武汉菲舍控制技术有限公司 | Anti-tearing system and early warning method for conveying belt |
| CN113375933A (en) * | 2021-05-31 | 2021-09-10 | 中国矿业大学 | Fault diagnosis system and method for scraper conveyor |
| US11353889B1 (en) * | 2021-06-16 | 2022-06-07 | Beta Air, Llc | System for electric aircraft navigation |
| CN113567159A (en) * | 2021-06-17 | 2021-10-29 | 太原理工大学 | Scraper conveyor state monitoring and fault diagnosis method based on edge cloud cooperation |
| CN114940343A (en) * | 2022-05-24 | 2022-08-26 | 无锡雪浪数制科技有限公司 | Double-chain scraper conveyor anomaly detection method based on deep learning |
| CN115272980A (en) * | 2022-09-22 | 2022-11-01 | 常州海图信息科技股份有限公司 | Conveying belt surface detection method and system based on machine vision |
| CN115577309A (en) * | 2022-09-27 | 2023-01-06 | 上海交通大学 | Belt conveyor sudden fault early warning method and system |
| CN116142726A (en) * | 2023-03-22 | 2023-05-23 | 山西潞安矿业(集团)有限责任公司 | Fault monitoring and early warning system of scraper conveyor chain transmission system |
| CN116576363A (en) * | 2023-04-17 | 2023-08-11 | 中国矿业大学 | Device and method for recognizing coal and rock through audio frequency, hyperspectrum and camera fusion |
Non-Patent Citations (3)
| Title |
|---|
| 周昭茂等.电力需求侧管理技术支持系统.中国电力出版社,2007,(第1版),218. * |
| 基于小波分析的刮板输送机断链故障检测系统;温晓荣等;电气与自动化;216-219 * |
| 赵文才等.煤矿智能化技术.煤炭工业出版社,2020,(第1版),214. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117142009A (en) | 2023-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117142009B (en) | Scraper conveyor health state assessment method based on graph rolling network | |
| Udo et al. | Data-driven predictive maintenance of wind turbine based on SCADA data | |
| CN117234785B (en) | Centralized control platform error analysis system based on artificial intelligence self-query | |
| Xu et al. | A novel health indicator for intelligent prediction of rolling bearing remaining useful life based on unsupervised learning model | |
| CN114282443B (en) | Residual service life prediction method based on MLP-LSTM supervised joint model | |
| CN117421582A (en) | Equipment health analysis method based on multi-source data driving | |
| CN114519923B (en) | Intelligent diagnosis and early warning method and system for power plant | |
| CN109492790A (en) | Wind turbines health control method based on neural network and data mining | |
| Son et al. | Deep learning-based anomaly detection to classify inaccurate data and damaged condition of a cable-stayed bridge | |
| CN110757510A (en) | Method and system for predicting remaining life of robot | |
| CN118133203A (en) | Fault diagnosis method for electric energy metering detection information | |
| CN113962308A (en) | Aviation equipment fault prediction method | |
| Zhu et al. | RGCNU: recurrent graph convolutional network with uncertainty estimation for remaining useful life prediction | |
| Ding et al. | Artificial intelligence based abnormal detection system and method for wind power equipment | |
| Dang et al. | seq2graph: Discovering dynamic non-linear dependencies from multivariate time series | |
| Bond et al. | A hybrid learning approach to prognostics and health management applied to military ground vehicles using time-series and maintenance event data | |
| CN117872219A (en) | Open-circuit fault diagnosis method and system for train traction converter | |
| Jastrzebska et al. | Measuring wind turbine health using fuzzy-concept-based drifting models | |
| CN114638039B (en) | Structural health monitoring characteristic data interpretation method based on low-rank matrix recovery | |
| CN114548701B (en) | Full-measurement-point-oriented coupling structure analysis and estimation process early warning method and system | |
| Yuan et al. | Deep learning approach to multiple features sequence analysis in predictive maintenance | |
| CN113469228A (en) | Power load abnormal value identification method based on data flow space-time characteristics | |
| CN118568471B (en) | Intelligent power distribution station operation fault prediction method and system | |
| CN117930815B (en) | Wind turbine generator remote fault diagnosis method and system based on cloud platform | |
| CN118657404A (en) | Multi-mode-based power grid health assessment and analysis method |
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 | ||
| CB03 | Change of inventor or designer information |
Inventor after: Su Le Inventor after: Zhai Lina Inventor after: Zhao Chenglong Inventor after: Yang Huan Inventor after: Zhang Tianliang Inventor after: Wang Huanle Inventor after: Cheng Yongjun Inventor after: Wang Meng Inventor after: Wang Bo Inventor after: Yang Xiongwei Inventor after: Feng Jiong Inventor before: Su Le Inventor before: Zhai Lina Inventor before: Zhao Chenglong Inventor before: Yang Huan Inventor before: Zhang Tianliang Inventor before: Wang Huanle Inventor before: Cheng Yongjun Inventor before: Wang Meng Inventor before: Wang Bo Inventor before: Yang Xiongwei Inventor before: Feng Jiong |
|
| CB03 | Change of inventor or designer information | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20240204 Address after: 715517, approximately 200 meters east of Dongfang Road residential community in Pucheng County, Weinan City, Shaanxi Province, Xi'an Zhongzhuang Pubai Coal Mine Machinery Co.,Ltd. Applicant after: XI'AN HEAVY EQUIPMENT PUBAI COAL MINING MACHINERY Ltd. Country or region after: China Address before: Small and Micro Park, Xiaotunzhuang Village, Huangzhai Town, Yangqu County, Taiyuan City, Shanxi Province, 030100 Applicant before: Shanxi Haicheng Intelligent Manufacturing Co.,Ltd. Country or region before: China Applicant before: XI'AN HEAVY EQUIPMENT PUBAI COAL MINING MACHINERY Ltd. |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |