CN113449072A - Construction method of excavator fault knowledge map based on deep learning - Google Patents
Construction method of excavator fault knowledge map based on deep learning Download PDFInfo
- Publication number
- CN113449072A CN113449072A CN202110660314.4A CN202110660314A CN113449072A CN 113449072 A CN113449072 A CN 113449072A CN 202110660314 A CN202110660314 A CN 202110660314A CN 113449072 A CN113449072 A CN 113449072A
- Authority
- CN
- China
- Prior art keywords
- fault
- text
- entity
- network model
- excavator
- 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
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 15
- 238000010276 construction Methods 0.000 title abstract description 19
- 238000003745 diagnosis Methods 0.000 claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000000034 method Methods 0.000 claims description 48
- 238000013527 convolutional neural network Methods 0.000 claims description 27
- 230000006870 function Effects 0.000 claims description 21
- 238000012423 maintenance Methods 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 13
- 238000003062 neural network model Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 7
- 238000012795 verification Methods 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 5
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 description 12
- 238000000605 extraction Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 7
- 238000001228 spectrum Methods 0.000 description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000003672 processing method Methods 0.000 description 3
- 230000001502 supplementing effect Effects 0.000 description 3
- 238000012800 visualization Methods 0.000 description 3
- 230000003213 activating effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000009435 building construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000012407 engineering method Methods 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
-
- 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
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Economics (AREA)
- Evolutionary Biology (AREA)
- Marketing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Quality & Reliability (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Animal Behavior & Ethology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses an excavator fault knowledge map construction method based on deep learning, relates to the field of excavator fault diagnosis, is good in instantaneity and is suitable for online fault diagnosis. The invention comprises the following steps: receiving a fault work order reported by a client, extracting an entity text from the fault work order and marking the entity text; performing model training by using the marked entity text; constructing RDF triples and importing the RDF triples into a knowledge map database, wherein the constructed RDF triples comprise: head, relationship and tail entities; and aiming at the current excavator fault to be processed, carrying out knowledge search by using the knowledge map database needle to obtain a fault diagnosis result. The invention is suitable for the on-line diagnosis of the excavator faults.
Description
Technical Field
The invention relates to the field of excavator fault diagnosis, in particular to an excavator fault knowledge map construction method based on deep learning.
Background
The crawler excavator is an essential tool in civil building construction, hydraulic engineering, road and bridge engineering and other capital construction projects. Compared with a wheeled machine, the crawler excavator is worse in environment and vibration conditions during operation. The crawler excavator can easily break down after long-time operation in a severe environment, construction stagnation is caused, construction risks are increased, construction progress is seriously influenced, and economic benefits of related enterprises and life safety of constructors are further influenced. Therefore, the method for diagnosing the faults of the crawler excavator is researched, the fault diagnosis and maintenance time is shortened, the diagnosis reliability is improved, and the method has a vital significance for guaranteeing the safe operation of the excavator, promoting the safety management work and increasing the economic benefits of related enterprises.
At present, fault diagnosis and maintenance of the crawler excavator mostly depend on the technical level and experience of maintenance personnel, and the interval time from the occurrence of a fault to the arrival of the maintenance personnel at an operation site for diagnosing and removing the fault can lead to construction stagnation, so that the problems of uneven diagnosis accuracy, low reliability, low efficiency and the like can be caused under the current situation. In addition, in the field of large-scale mechanical fault diagnosis, the real-time performance of most of the methods used in the current research is poor, and the methods are not suitable for online fault diagnosis.
Disclosure of Invention
The embodiment of the invention provides an excavator fault knowledge map construction method based on deep learning, which is good in instantaneity and suitable for online fault diagnosis.
In order to achieve the above purpose, the embodiment of the invention adopts the following technical scheme:
receiving a fault work order reported by a client, extracting an entity text from the fault work order and marking the entity text; performing model training by using the marked entity text; constructing RDF triples and importing the RDF triples into a knowledge map database, wherein the constructed RDF triples comprise: head, relationship and tail entities; and aiming at the current excavator fault to be processed, carrying out knowledge search by using the knowledge map database needle to obtain a fault diagnosis result.
The extracting and marking of the entity text from the fault work order comprises the following steps: extracting text data representing fault description, processing measures and processing results from the fault work order entity through character recognition; preprocessing the extracted text data to obtain an entity text; and marking the entity text according to the category. The categories include at least: parts, fault words, fault phenomena, fault reasons, excavator types, excavator models and maintenance methods.
Classifying the marked entity texts to obtain a training set, a verification set and a test set, wherein the training set is used for training the convolutional neural network model, the verification set is used for training the cyclic neural network model, and the test set verifies the accuracy of the model.
The model training by using the labeled entity text comprises the following steps: constructing a text CNN network model and a text RNN network model, and setting a loss function according to an optimized target; performing iterative operation on the constructed CNN network model and the text RNN network model until the loss function is stable, and obtaining a parameter result of the network model; and verifying the accuracy of the constructed CNN network model and the text RNN network model under the obtained parameter result through the test set.
The excavator fault knowledge map construction method based on deep learning provided by the embodiment of the invention extracts entities and relations among various entities from unstructured and semi-structured excavator fault work orders; classifying the entities by using a neural network model based on deep learning, constructing triples of the entities and the relations based on rules and completing construction of a known spectrum; and new knowledge provided by a user is automatically absorbed, and the failure knowledge map is perfected and updated. The fault of the excavator mainly occurs in the operation period, and an operator is required to have rich maintenance experience, so that the difficulty is increased for fault diagnosis of the excavator, and the difficulty in diagnosis can be reduced by constructing a fault knowledge map. According to the technical scheme, the fault knowledge map of the excavator can be constructed to assist diagnosis decision, and the accuracy and the rapidity of fault diagnosis are improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic diagram of a logic flow in a possible implementation manner provided by an embodiment of the present invention;
fig. 2 is a schematic diagram of a text CNN network structure provided in an embodiment of the present invention;
fig. 3 is a schematic diagram of a text RNN network structure according to an embodiment of the present invention;
FIG. 4 is a partial schematic diagram of a fault knowledge map provided by an embodiment of the present invention;
fig. 5 is a flowchart illustrating a method according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the present invention will be described in further detail with reference to the accompanying drawings and specific embodiments. Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention. As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or coupled. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items. It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
At present, knowledge is extracted from text files such as fault work orders, product specifications and the like, a fault diagnosis knowledge map which is easy to operate, can be searched in real time and is complete is constructed for field operators, and the fault diagnosis knowledge map has great significance for supplementing learning capacity. In general, a practical engineering method is still lacked for how to efficiently construct the excavator fault diagnosis knowledge graph in the current application scheme. Therefore, further research and development are still needed for the construction and completion method of the excavator fault diagnosis knowledge graph.
The design idea of this embodiment lies in: an excavator fault knowledge map construction and completion method based on deep learning. The diagnosis decision of field operation personnel is assisted by introducing the knowledge map into the field of fault diagnosis; different entity classification network models are built by utilizing a deep learning algorithm, and finally, the model with the best effect is selected to complete a prediction task through competition among the models so as to improve the accuracy; the built failure knowledge graph is stored by using a Neo4j graph database and knowledge search is carried out based on a Cypher query language. By the method, the efficient and reliable excavator fault knowledge map can be constructed, stored, supplemented and searched.
The embodiment of the invention provides an excavator fault knowledge graph construction method based on deep learning, which comprises the following steps of:
and S1, receiving the fault work order reported by the client, extracting the entity text from the fault work order and marking the entity text.
Wherein, the information in the fault work order at least should include: fault phenomena, corresponding processing methods and processing results.
And S2, performing model training by using the labeled entity text.
Specifically, the method is used for performing text description, entity extraction and classification on faults occurring in the operation process of the excavator. When a new fault occurs, the fault is subjected to text description and entity extraction. And carrying out entity recognition and automatic classification through the model with the best training effect.
S3, constructing RDF triples and importing the RDF triples into a knowledge graph database, wherein the constructed RDF triples comprise: head entities, relationships, and tail entities.
Specifically, constructing RDF triples may be understood as: and forming the RDF triples of the head entity, the relation and the tail entity by the entities and the relation according to a ternary reasoning rule. Then importing Neo4j database: and storing the processed RDF triples in a csv file format, building an interface through a py2Neo library, importing data into a Neo4j knowledge spectrum database, realizing knowledge visualization, and providing bottom support for knowledge search.
And S4, aiming at the current excavator fault to be processed, carrying out knowledge search by using the knowledge map database needle to obtain a fault diagnosis result.
Wherein, dynamically perfecting and supplementing the fault knowledge map: and newly generated fault knowledge is supplemented to the original knowledge map, so that the knowledge map is continuously perfected and enriched, and the reliability and integrity of the knowledge map are improved. In practical application, the current reported excavator fault which needs to be diagnosed can also be reported in a fault work order form. The excavator fault knowledge map construction method based on deep learning provided by the embodiment of the invention extracts entities and relations among various entities from unstructured and semi-structured excavator fault work orders; classifying the entities by using a neural network model based on deep learning, constructing triples of the entities and the relations based on rules and completing construction of a known spectrum; and new knowledge provided by a user is automatically absorbed, and the failure knowledge map is perfected and updated. The fault of the excavator mainly occurs in the operation period, and an operator is required to have rich maintenance experience, so that the difficulty is increased for fault diagnosis of the excavator, and the difficulty in diagnosis can be reduced by constructing a fault knowledge map. According to the technical scheme, the fault knowledge map of the excavator can be constructed to assist diagnosis decision, and the accuracy and the rapidity of fault diagnosis are improved.
In this embodiment, the extracting and labeling the entity text from the fault work order includes:
and extracting text data representing fault description, processing measures and processing results from the fault work order entity through character recognition. And preprocessing the extracted text data to obtain an entity text. And marking the entity text according to the category. Wherein the categories include at least: parts, fault words, fault phenomena, fault reasons, excavator types, excavator models and maintenance methods. Specifically, in the process of extracting and marking the fault work order entity, the entity text is extracted from the fault work order by adopting a named entity identification method based on rules, the method depends on manually designed rules and dictionaries, and the method is simple and effective to use. And (4) combining a named entity library, carrying out weight assignment on each rule, and then carrying out type judgment according to the condition that the entity conforms to the rule. And marking the extracted entities as various types such as parts, fault words, fault phenomena, fault reasons, excavator types, excavator models, maintenance methods, processing methods and the like.
In this embodiment, the method further includes: classifying the marked entity texts to obtain a training set, a verification set and a test set, wherein the training set is used for training the convolutional neural network model, the verification set is used for training the cyclic neural network model, and the test set verifies the accuracy of the model.
The model training by using the labeled entity text comprises the following steps: a text CNN (Convolutional Neural Network) Network model and a text RNN (recurrent Neural Network) Network model are constructed, and a loss function is set according to an optimized target. And carrying out iterative operation on the constructed CNN network model and the text RNN network model until the loss function is stable, and obtaining a parameter result of the network model. And verifying the accuracy of the constructed CNN network model and the text RNN network model under the obtained parameter result through the test set.
The method comprises the following steps that a text CNN and a text RNN network model can be constructed based on a Tensorflow architecture; training a network model by using entity text data with labels; and (4) checking the accuracy of the trained network model through the test set, and selecting the network model with better performance to perform subsequent prediction tasks. A cross entropy loss function can be employed to evaluate the effectiveness of the neural network model and define the goal of optimization. And (4) after the network model is subjected to repeated iterative operation, and finally the loss function is stable and does not fall any more, stopping training to obtain parameters of different network models. And (4) checking evaluation indexes such as accuracy and recall rate of different models through the test set, and selecting the model with the best effect to perform subsequent prediction tasks.
Specifically, the CNN network model is composed of a knowledge representation layer and a convolutional neural network, and includes a word embedding layer, a convolutional layer, a maximum pooling layer, and a full connection layer. The word vector dimension, the number of convolution kernels, the size of the convolution kernels, the learning rate, the total iteration number and the like of the network can be specified by adjusting various hyper-parameters of the network. The main calculation work of the text CNN network model is completed by a word embedding layer and a convolution layer. The word embedding layer converts the text object into a matrix form, the dimension of a word vector is defined as dim, the length of a sentence is len, then the dimension of the text matrix can be expressed as len multiplied by dim, and the text matrix is recorded asUsing A [ i: j is a function of]And a sub-matrix composed of the ith row to the jth row elements of the matrix A is represented. The convolutional layer extracts features from the input matrix for subsequent classification. Output sequence of convolutional layersThe calculation formula of (2) is as follows: outi=w·A[i:i+s-1]Where w is the parameter to be trained, s is the height of the convolution kernel, i is 1, …, len-h +1, for each outiAdding bias termsAnd activating the function Act to obtain the feature mappingWhereinAnd constructing a text RNN network model. The model is different from the text CNN in that the feature extraction is carried out by using a circulation layer, so that the context information can be better expressed.
The effect of the neural network model is evaluated using the cross-entropy loss function commonly used in classification problems and the goal of optimization is defined:
where y represents the true value of the tag,the label prediction values representing the neural network. In the back propagation process, an Adam optimizer is used to direct the loss function values to continuously approach the global minimum. The concrete algorithm of the Adam optimizer is as follows:
wt+1=wt-ηt
wherein etatFor the descending gradient at the present moment, α is the learning rate, mtIs a first order momentum, VtIs a second order momentum, beta1And beta2Are two hyper-parameters, wtAs a parameter of the current time, wt+1For the next moment parameter, the specific expressions of the first-order momentum and the second-order momentum are as follows:
mt=β1·mt-1+(1-β1)·gt
wherein g istIs the gradient of the loss function with respect to the current parameter.
And when the network model is subjected to repeated iterative operation and finally the loss function is stable and does not decrease any more, stopping training to obtain parameters of different network models. And (4) checking evaluation indexes such as accuracy and recall rate of different models through the test set, and selecting the model with the best effect to perform subsequent prediction tasks.
In this embodiment, the process of constructing the RDF triple includes: extracting an entity text from the character description of the fault during the field operation of the excavator, labeling, inputting the trained CNN network model and the trained text RNN network model, and obtaining a prediction result of the type of the entity text:
wherein h isw(xi) Is a sample xiProbability of belonging to each label, wjIs the jth component value of the trained network parameter w, i is a positive integer, y is a real label (total 10 categories), and P is xiThe category is the probability of y ═ j (j ═ 1, 2, …, 10). Predictive tagThe prediction tag may also be referred to as a prediction result.
Specifically, character descriptions of faults occurring during operation of the excavator are collected, entity texts are generated through preprocessing after the character descriptions are sorted, and the entity texts are input into a trained network model with the best performance to perform prediction classification. And a new RDF triple formed by the classification result is added into the original fault knowledge map as supplementary knowledge.
In this embodiment, the constructing RDF triples and importing the RDF triples into a knowledge graph database includes: and constructing RDF triples by using sixteen different entity texts according to a preset rule and storing the RDF triples as csv files. And (4) building an interface through a py2Neo library, and importing the obtained csv file into a Neo4j knowledge map database.
The preset rule may be a relationship between entities of a certain category, and the preset rule may be artificially preset, for example, the relationship between a fault phenomenon and a fault cause is preset as "fault cause", and a fault cause completely combined as a certain fault phenomenon is a certain fault cause. Ten different entities are embodied as characters: "degree adverb, auxiliary entity, method, failure word, failure phenomenon, failure cause, precondition, phenomenon word, component, and maintenance method". The sixteen relationships are embodied as characters: "the failure cause and the phenomenon are, the processing method is, the failure predicate is, the main cause and the method is, the leading level of … is, the possible method is, the cause is, the leading level of the preceding condition is, the main phenomenon is, the auxiliary maintenance method is, the associated failure phenomenon is, the auxiliary failure cause is".
Specifically, ten different entities with labels and sixteen relations are combined into an RDF triple of < head entity, relation and tail entity > according to a ternary reasoning rule and stored as a standard csv file. Then importing Neo4j database: and storing the processed RDF triples in a csv file format, building an interface through a py2Neo library, importing data into a Neo4j knowledge spectrum database, realizing knowledge visualization, and providing bottom support for knowledge search.
Specifically, for example, in practical application, the method realizes the construction and completion of the excavator fault knowledge graph based on deep learning, and mainly comprises two stages, and the specific flow is shown in fig. 1. The first stage is the construction of the excavator fault knowledge map, and the second stage is the supplement of the excavator fault knowledge map. The method mainly comprises the following steps:
the first stage, 1, collecting fault work orders and carrying out entity extraction and marking on the fault work orders; 2, building and training a text CNN and a text RNN for entity classification; 3 constructing RDF triples; 4, importing a Neo4j database to construct a fault knowledge map;
specifically, the step 1 of collecting the fault work order and performing entity extraction and labeling on the fault work order includes:
the method is based on rules, named entity recognition method is adopted to extract entity texts from fault work orders, and the method relies on manually designed rules and dictionaries, and is simple and effective to use. And (4) combining a named entity library, carrying out weight assignment on each rule, and then carrying out type judgment according to the condition that the entity conforms to the rule. The extracted entities are marked as ten different categories of fault phenomena, fault reasons, parts, phenomenon words, maintenance methods and the like.
Specifically, the 2 building and training a text CNN and a text RNN for entity classification includes:
(1) and constructing a text CNN network model. The model is composed of a knowledge representation layer and a convolutional neural network, and comprises a word embedding layer, a convolutional layer, a maximum pooling layer and a full connection layer, as shown in FIG. 2. The word vector dimension, the number of convolution kernels, the size of the convolution kernels, the learning rate, the total iteration number and the like of the network can be specified by adjusting various hyper-parameters of the network.
The main calculation work of the text CNN network model is completed by a word embedding layer and a convolution layer.
The word embedding layer converts the text object into a matrix form, the dimension of a word vector is defined as dim, the length of a sentence is len, then the dimension of the text matrix can be expressed as len multiplied by dim, and the text matrix is recorded asUsing A [ i: j ]]And a sub-matrix composed of the ith row to the jth row elements of the matrix A is represented.
The convolutional layer extracts features from the input matrix for subsequent classification. Output sequence of convolutional layersThe calculation formula of (2) is as follows:
outi=w·A[i:i+s-1]
where w is the parameter to be trained, s is the height of the convolution kernel, i ═ 1, …, len-h +1, for each outiAdding bias termsAnd activating the function Act to obtain the feature mapping ml:
mi=Act(outi+bias)
A textual RNN network model was constructed as shown in fig. 3. The model is different from the text CNN in that the feature extraction is carried out by using a circulation layer, so that the context information can be better expressed.
(2) Using a cross entropy loss function commonly used in classification problems to evaluate the effect of the neural network model and define the goal of optimization:
where y represents the true value of the data,representing the predicted value of the neural network.
In the back propagation process, an Adam optimizer is used to direct the loss function values to continuously approach the global minimum. The specific algorithm of the Adam optimizer is as follows:
wt+1=wt-ηt
wherein etatFor the descending gradient at the present moment, α is the learning rate, mtIs a first order momentum, VtIs a second order momentum, beta1And beta2Are two hyper-parameters, wtAs a parameter of the current time, wt+1For the next moment parameter, the specific expressions of the first-order momentum and the second-order momentum are as follows:
mt=β1·mt-1+(1-β1)·gt
wherein g istIs the gradient of the loss function with respect to the current parameter.
(3) And (4) after the network model is subjected to repeated iterative operation, and finally the loss function is stable and does not fall any more, stopping training to obtain parameters of different network models. And (4) checking evaluation indexes such as accuracy and recall rate of different models through the test set, and selecting the model with the best effect to perform subsequent prediction tasks.
Specifically, 3, constructing an RDF triple includes:
and forming RDF triples of < head entities, relations and tail entities > by the ten different entities with the labels and the sixteen relations according to a ternary reasoning rule, and storing the RDF triples as standard csv files.
Specifically, 4, importing a Neo4j database to construct a fault knowledge graph, which comprises the following steps:
and storing the processed RDF triples in a csv file format, building an interface through a py2Neo library, importing data into a Neo4j knowledge spectrum database, realizing knowledge visualization, and providing bottom support for knowledge search.
And in the second stage, the fault knowledge graph is supplemented, and 5, the faults occurring in the operation process of the excavator are subjected to character description, entity extraction and classification and are added into the original fault knowledge graph as supplementary knowledge.
Specifically, 5 carry out description of characters, entity extraction and classification to the trouble that the excavator operation in-process appears, add into original trouble knowledge map as supplementary knowledge, include:
acquiring character description of faults when the excavator works on site, generating an entity text through the preprocessing in the step 1), inputting the entity text into the best-performing network model trained in the step 2), and predicting the type of the entity:
wherein h isw(xi) Is a sample xiProbability of belonging to each label, wjIs the jth component value of the trained network parameter w. Finally, the label is predictedThis can be derived from the following equation:
and a new RDF triple formed by the classification result is added into the original fault knowledge map as supplementary knowledge.
In the embodiment, the method can be used for building the excavator fault knowledge base with high reliability, and the integrity of the map is ensured by continuously supplementing new knowledge. The method has the advantages that the existing fault work order is used for training the deep learning network model, so that the model has the capability of predicting the entity type, and the knowledge graph can be supplemented and perfected efficiently. By using the Neo4j database to store and display the knowledge graph, reliable functions of constructing, storing, displaying, complementing and searching the excavator fault knowledge graph can be realized.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the apparatus embodiment, since it is substantially similar to the method embodiment, it is relatively simple to describe, and reference may be made to some descriptions of the method embodiment for relevant points. The above description is only for the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (7)
1. A method for constructing a fault knowledge graph of an excavator based on deep learning is characterized by comprising the following steps:
receiving a fault work order reported by a client, extracting an entity text from the fault work order and marking the entity text;
performing model training by using the marked entity text;
constructing RDF triples and importing the RDF triples into a knowledge map database, wherein the constructed RDF triples comprise: head, relationship and tail entities;
and aiming at the current excavator fault to be processed, carrying out knowledge search by using the knowledge map database needle to obtain a fault diagnosis result.
2. The method of claim 1, wherein extracting and labeling entity text from a trouble order comprises:
extracting text data representing fault description, processing measures and processing results from the fault work order entity through character recognition;
preprocessing the extracted text data to obtain an entity text;
and marking the entity text according to the category.
3. The method according to claim 2, characterized in that said categories comprise at least: parts, fault words, fault phenomena, fault reasons, excavator types, excavator models and maintenance methods.
4. The method of claim 1, further comprising:
classifying the marked entity texts to obtain a training set, a verification set and a test set, wherein the training set is used for training the convolutional neural network model, the verification set is used for training the cyclic neural network model, and the test set verifies the accuracy of the model.
5. The method of claim 4, wherein the model training using the labeled entity text comprises:
constructing a text CNN network model and a text RNN network model, and setting a loss function according to an optimized target;
performing iterative operation on the constructed CNN network model and the text RNN network model until the loss function is stable, and obtaining a parameter result of the network model;
and verifying the accuracy of the constructed CNN network model and the text RNN network model under the obtained parameter result through the test set.
6. The method of claim 1, wherein the process of constructing the RDF triples comprises:
extracting an entity text from the character description of the fault during the field operation of the excavator, labeling, inputting the trained CNN network model and the trained text RNN network model, and obtaining a prediction result of the type of the entity text:
7. The method of claim 1 or 6, wherein the constructing of RDF triples and importing of RDF triples into a knowledge graph database comprises:
constructing RDF triples by using sixteen different entity texts according to a preset rule and storing the RDF triples as csv files;
and (4) building an interface through a py2Neo library, and importing the obtained csv file into a Neo4j knowledge map database.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110660314.4A CN113449072A (en) | 2021-06-15 | 2021-06-15 | Construction method of excavator fault knowledge map based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110660314.4A CN113449072A (en) | 2021-06-15 | 2021-06-15 | Construction method of excavator fault knowledge map based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113449072A true CN113449072A (en) | 2021-09-28 |
Family
ID=77811308
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110660314.4A Pending CN113449072A (en) | 2021-06-15 | 2021-06-15 | Construction method of excavator fault knowledge map based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113449072A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114417015A (en) * | 2022-01-26 | 2022-04-29 | 西南交通大学 | Method for constructing maintainability knowledge graph of high-speed train |
CN115544265A (en) * | 2022-09-13 | 2022-12-30 | 南京航空航天大学 | Bearing fault diagnosis method based on bearing fault knowledge graph |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111737496A (en) * | 2020-06-29 | 2020-10-02 | 东北电力大学 | Power equipment fault knowledge map construction method |
US20210103256A1 (en) * | 2019-09-06 | 2021-04-08 | Intelligent Fusion Technology, Inc. | Decision support method and apparatus for machinery control |
-
2021
- 2021-06-15 CN CN202110660314.4A patent/CN113449072A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20210103256A1 (en) * | 2019-09-06 | 2021-04-08 | Intelligent Fusion Technology, Inc. | Decision support method and apparatus for machinery control |
CN111737496A (en) * | 2020-06-29 | 2020-10-02 | 东北电力大学 | Power equipment fault knowledge map construction method |
Non-Patent Citations (2)
Title |
---|
丁頔 等: "CNN-RNN融合法在旋转机械故障诊断中的应用", 轻工学报, vol. 35, no. 1, pages 102 - 108 * |
吴定海 等: "基于卷积神经网络的机械故障诊断方法综述", 机械强度, vol. 42, no. 5, pages 1025 - 1032 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114417015A (en) * | 2022-01-26 | 2022-04-29 | 西南交通大学 | Method for constructing maintainability knowledge graph of high-speed train |
CN115544265A (en) * | 2022-09-13 | 2022-12-30 | 南京航空航天大学 | Bearing fault diagnosis method based on bearing fault knowledge graph |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110134757B (en) | Event argument role extraction method based on multi-head attention mechanism | |
CN108984775B (en) | Public opinion monitoring method and system based on commodity comments | |
CN111966076A (en) | Fault positioning method based on finite-state machine and graph neural network | |
CN111079430A (en) | Power failure event extraction method combining deep learning and concept map | |
CN110162478B (en) | Defect code path positioning method based on defect report | |
CN111427775B (en) | Method level defect positioning method based on Bert model | |
CN113255321B (en) | Financial field chapter-level event extraction method based on article entity word dependency relationship | |
CN113449072A (en) | Construction method of excavator fault knowledge map based on deep learning | |
CN109471793A (en) | A kind of webpage automatic test defect positioning method based on deep learning | |
CN113742733B (en) | Method and device for extracting trigger words of reading and understanding vulnerability event and identifying vulnerability type | |
CN109947898B (en) | Equipment fault testing method based on intellectualization | |
CN113688169B (en) | Mine potential safety hazard identification and early warning system based on big data analysis | |
CN109345272A (en) | One kind is based on the markovian shop credit risk forecast method of improvement | |
CN112036185B (en) | Method and device for constructing named entity recognition model based on industrial enterprise | |
CN112328475B (en) | Defect positioning method for multiple suspicious code files | |
CN108920446A (en) | A kind of processing method of Engineering document | |
CN112434390A (en) | PCA-LSTM bearing residual life prediction method based on multi-layer grid search | |
CN114757557A (en) | On-site operation risk assessment prediction method and device based on electric work ticket | |
CN114662405A (en) | Rock burst prediction method based on few-sample measurement and ensemble learning | |
CN112559741B (en) | Nuclear power equipment defect record text classification method, system, medium and electronic equipment | |
CN112699018A (en) | Software defect positioning method based on software defect correlation analysis | |
Xie et al. | The twenty-first century of structural engineering research: A topic modeling approach | |
CN116975161A (en) | Entity relation joint extraction method, equipment and medium of power equipment partial discharge text | |
CN116974799A (en) | Fault diagnosis method and system of lighting system based on knowledge graph | |
CN114881172A (en) | Software vulnerability automatic classification method based on weighted word vector and 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 |