CN114153988A - Chemical abnormal event detection method and system based on matter map - Google Patents
Chemical abnormal event detection method and system based on matter map Download PDFInfo
- Publication number
- CN114153988A CN114153988A CN202111461707.9A CN202111461707A CN114153988A CN 114153988 A CN114153988 A CN 114153988A CN 202111461707 A CN202111461707 A CN 202111461707A CN 114153988 A CN114153988 A CN 114153988A
- Authority
- CN
- China
- Prior art keywords
- accident
- chemical
- detected
- graph
- report
- 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
- 239000000126 substance Substances 0.000 title claims abstract description 90
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 45
- 238000001514 detection method Methods 0.000 title claims abstract description 27
- 239000013598 vector Substances 0.000 claims abstract description 77
- 238000000034 method Methods 0.000 claims abstract description 53
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000003889 chemical engineering Methods 0.000 claims abstract description 10
- 230000008569 process Effects 0.000 claims description 16
- 238000000605 extraction Methods 0.000 claims description 12
- 230000015654 memory Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 10
- 230000011218 segmentation Effects 0.000 claims description 9
- 238000003860 storage Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 6
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 description 18
- 230000000875 corresponding effect Effects 0.000 description 12
- 238000004590 computer program Methods 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 230000001364 causal effect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000002372 labelling Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000012824 chemical production Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000006227 byproduct Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000009193 crawling Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 231100000572 poisoning Toxicity 0.000 description 1
- 230000000607 poisoning effect Effects 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001988 toxicity Effects 0.000 description 1
- 231100000419 toxicity Toxicity 0.000 description 1
- 230000009466 transformation Effects 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
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/194—Calculation of difference between files
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/237—Lexical tools
- G06F40/242—Dictionaries
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Animal Behavior & Ethology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a chemical engineering abnormal event detection method and a chemical engineering abnormal event detection system based on a matter graph, wherein the method comprises the following steps: acquiring a chemical accident report to be detected; preprocessing a chemical accident report to be detected to obtain a text vector of the chemical accident report to be detected; determining the accident type and case type corresponding to the chemical accident report to be detected according to the text vector of the chemical accident report to be detected and the trained graph attention network model; wherein, the graph attention network model is obtained based on the training of a physiological graph. So as to more accurately predict the type of accident and quickly provide an emergency treatment scheme for the accident.
Description
Technical Field
The invention relates to the technical field of chemical abnormal event detection, in particular to a chemical abnormal event detection method and system based on a physics map.
Background
The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.
The chemical production process has the characteristics of flammability and explosiveness, strong toxicity of byproducts, high process risk and the like, and a 'tiny' fault event in the production process can be evolved into a fire disaster, explosion, poisoning and other serious accidents. The heinrich law states that the occurrence of any accident is not an isolated event, but the result of a series of events occurring in succession. These "minor events" frequently occur in the production field but are often ignored, when these unsafe factors converge, they will cause larger events and even accidents, and when serious, they may cause catastrophic consequences, such as significant economic loss, casualties, environmental pollution, and even spread to adjacent plants, causing domino effect. Therefore, if the abnormal events which can be evolved into accidents can be accurately found, and the possible accident evolution situation can be correctly analyzed, the method has important significance in the aspects of ensuring the production safety, reducing the accident occurrence probability and the like.
The analysis and judgment of chemical engineering abnormal events are usually performed by means of expert experience. When an abnormal event occurs in the chemical production process, the expert analyzes and judges the evolution trend of the abnormal event according to the handling experience of the historical similar event, so that an abnormal event handling scheme is formed. Such methods, due to the complexity of the field environment, the diversity of event evolution and the differences in expert experience, pose a significant risk of uncertainty in the development of the project. How to form reusable expert knowledge from the handling experience of the historical accident is an important way for analyzing the current accident potential and finding out the possible accident evolution of the abnormal event by comprehensively analyzing and judging by means of an information technology and an artificial intelligence method.
Most of the existing abnormal event detection and analysis methods are based on accident analysis models, such as Fault Tree Analysis (FTA), Event Tree Analysis (ETA), bowknot analysis (Bow-Tie), hazard and operability analysis (HAZOP), and the like, and the analysis methods are based on cause analysis of a single accident or safety analysis results of a certain device and equipment, namely classification and summarization of a large number of historical cases are difficult to form. Therefore, no matter abnormal event detection, accident deduction and scenario analysis found on the basis of technologies such as a Bayesian network, a random network and a neural network, retrieval and analysis are only performed by taking an accident as a unit, convergence analysis of the accident is not formed, so that events of multi-source information cannot be fully expressed, implicit information of similar events is difficult to be fully mined, and a constructed scene cannot comprehensively capture changes and evolution of a historical case, so that the problems of insufficient accuracy and integrity exist in detection and evolution analysis of abnormal events.
Besides most accident analysis model-based methods, there are also atlas-based construction models. For example: the invention patent of China, publication No. CN110968699A, a logic map construction and early warning method and device based on affair recommendation, and the invention patent of China, publication No. CN108052576A, a affair knowledge map construction method and system.
The invention patent of China, publication No. CN 110968699A-a logic map construction and early warning method and device based on affair recommendation, the patent is in the aspect of constructing maps: and performing event extraction according to the type of the logic relationship among the specified events, and establishing an event map according to the event extraction result. The event map has the advantages of paying attention to the evolution logic relation between the exterior of the event and also paying attention to the rich attribute information of the event. However, the patent lacks a top-level structural design of an event abstraction layer, and only a map of the accident case.
The invention patent of China, publication No. CN 108052576A-a method and system for constructing a affair knowledge map, the invention constructs the affair knowledge map by a macroscopic event layer, a microscopic knowledge layer and a affair knowledge body layer, although the hierarchical network relation of the ontology, the microscopic knowledge layer and the affair knowledge body layer is adopted, the constructed affair map is separated from each other in the body layer, the microscopic knowledge layer and the event layer, and the information is not communicated and the deep commonality among the events is difficult to find.
Disclosure of Invention
In order to solve the defects of the prior art, the invention provides a chemical abnormal event detection method and system based on a physical map; the abnormal event detection based on the case map is provided, similar events are converged through a concept map abstract technology of the case map, and the specific cause and evolution expression of each accident case are formed through the case map, so that not only can common information of the abnormal events be found, but also individual change forms of different cases can be represented, and a basic knowledge organization form is provided for accurately detecting the historical common property and the different situation evolution of the abnormal events. Based on the knowledge organization form of the historical accident case, the invention provides the abnormal event detection method which integrates the process relation of the whole accident evolution, can more accurately predict the type of the accident and can rapidly provide an emergency disposal scheme for the accident.
In a first aspect, the invention provides a chemical engineering abnormal event detection method based on a physics map;
a chemical abnormal event detection method based on a physics map comprises the following steps:
acquiring a chemical accident report to be detected;
preprocessing a chemical accident report to be detected to obtain a text vector of the chemical accident report to be detected;
determining the accident type and case type corresponding to the chemical accident report to be detected according to the text vector of the chemical accident report to be detected and the trained graph attention network model;
wherein, the graph attention network model is obtained based on the training of a physiological graph.
In a second aspect, the invention provides a chemical engineering abnormal event detection system based on a physics map;
chemical engineering abnormal event detection system based on a matter map comprises:
an acquisition module configured to: acquiring a chemical accident report to be detected;
a pre-processing module configured to: preprocessing a chemical accident report to be detected to obtain a text vector of the chemical accident report to be detected;
a prediction module configured to: determining the accident type and case type corresponding to the chemical accident report to be detected according to the text vector of the chemical accident report to be detected and the trained graph attention network model;
wherein, the graph attention network model is obtained based on the training of a physiological graph.
In a third aspect, the present invention further provides an electronic device, including:
a memory for non-transitory storage of computer readable instructions; and
a processor for executing the computer readable instructions,
wherein the computer readable instructions, when executed by the processor, perform the method of the first aspect.
In a fourth aspect, the present invention also provides a storage medium storing non-transitory computer readable instructions, wherein the non-transitory computer readable instructions, when executed by a computer, perform the instructions of the method of the first aspect.
In a fifth aspect, the invention also provides a computer program product comprising a computer program for implementing the method of the first aspect when run on one or more processors.
Compared with the prior art, the invention has the beneficial effects that:
1. according to expert experience, the method designs a case map which describes evolution rules and modes among chemical events. The map provides a basis for subsequent chemical engineering event evolution and abnormal event detection;
2. based on a GAT fusion graph model, when chemical accident evolution reasoning is carried out, the sequential, causal, condition, upper and lower level and other affair logic relations among graph node events are considered, a high-quality data basis is provided for subsequent retrieval, evolution reasoning and classification decision, and the classification accuracy is improved;
3. the method fully excavates the hidden information in the data, efficiently expresses the multivariate information events and the sensitive information, and is beneficial to enhancing the expression of chemical abnormal events and the potential of fusing the information of surrounding nodes;
4. the model can efficiently match the event map nodes for the given abnormal events, quickly and accurately provide similar historical cases of the abnormal events, and provide a foundation for the subsequent evolution analysis and emergency treatment.
5. The method comprises the steps that on the basis of an event evolution structure, a hierarchical affair map network structure which contains both event evolution relations and event abstract relations is constructed; meanwhile, in the aspect of retrieval, the retrieval of the current abnormal event semantics can be completed based on the historical common events by adopting vectorization representation of the associated event structure relationship.
6. The method adopts a correlated hierarchical structure, namely a connecting line between a concept case affair layer and a case affair layer represents the abstract and generalized relation of an event, a network layer node representing the concept case affair is a more abstract semantic representation of the network layer node representing the case affair, and the network layer node representing the concept case affair can be generalized into any network layer node of the case affair. Semantic alignment between the concept event network layer and the case event network layer effectively discovers deep commonality among different events, and further can discover essential relation and knowledge of the events.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a chemical event map illustrating the rules and patterns of evolution between events as proposed by the present invention.
FIG. 2 is a method and system for detecting chemical abnormal events based on a case map.
Detailed Description
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all 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 is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise, and it should be understood that the terms "comprises" and "comprising", and any variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
In addition, in order to facilitate clear description of technical solutions of the embodiments of the present invention, in the embodiments of the present invention, terms such as "first" and "second" are used to distinguish the same items or similar items having substantially the same functions and actions. Those skilled in the art will appreciate that the words "first", "second", etc. do not necessarily define a quantity or order of execution and that the words "first", "second", etc. do not necessarily differ.
The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.
All data are obtained according to the embodiment and are legally applied on the data on the basis of compliance with laws and regulations and user consent.
Example one
The embodiment provides a chemical abnormal event detection method based on a matter map;
as shown in fig. 2, the chemical abnormal event detection method based on the event graph includes:
s101: acquiring a chemical accident report to be detected;
s102: preprocessing a chemical accident report to be detected to obtain a text vector of the chemical accident report to be detected;
s103: determining the accident type and case type corresponding to the chemical accident report to be detected according to the text vector of the chemical accident report to be detected and the trained graph attention network model;
wherein, the graph attention network model is obtained based on the training of a physiological graph.
Further, S101: acquiring a chemical accident report to be detected; wherein, the chemical accident report to be detected at least comprises: accident report title, accident report description information (time, place, cause of accident and casualties, property loss and the like caused by accident).
Further, S102: preprocessing a chemical accident report to be detected to obtain a text vector of the chemical accident report to be detected; the method specifically comprises the following steps:
performing word segmentation processing on a chemical accident report to be detected;
removing stop words from the word segmentation processing result;
performing stem extraction processing on the processing result of removing stop words to obtain a plurality of stems;
for each extracted stem, generating a stem vector;
and solving the average value of all the word stem vectors, and taking the average value as a text vector of the chemical accident report to be detected.
Illustratively, the stem extraction employs a HanLP segmentation tool.
HanLP is a Java toolkit consisting of a series of models and algorithms aimed at popularizing the application of natural language processing in a production environment. The HanLP has the characteristics of complete functions, high performance, clear architecture, novel linguistic data and self-definition.
Illustratively, for each stem extracted, generating each stem vector; implemented using word2 vec.
For example, after preprocessing such as data cleaning, word segmentation, word stop removal, word stem extraction and the like are performed on the matching events, the word2vec toolkit can be used for processing, so as to obtain a vector of given dimensionality for representing each word. A vector of text is then further computed that can be characterized by the average of the vectors of words in the text.
The method comprises the steps of considering the constitutional features of special nouns of dangerous chemicals, customizing domain lexicon information, adopting a maximum matching mode, using a jieba Chinese word segmentation tool, carrying out automatic word extraction on words based on a domain dictionary and general dictionary mixed mode, and designing and realizing a word extraction method of the opposite text.
When the GAT model is adopted to conduct surrounding node fusion, a constructed affair map is utilized to conduct word segmentation, word stop removal and word stem extraction, the segmented words are conducted on word2vec to obtain 128-dimensional vectors, and the average value of the word vectors is adopted to represent the vector of the event text.
Further, the graph attention network model includes: a first graph attention network and a second graph attention network connected in series.
Illustratively, the first graph attention network and the second graph attention network are both implemented using a graph attention network GAT. GAT is known in English as Graph Attention Network.
Further, as shown in fig. 1, the obtaining process of the case atlas includes:
constructing a training set and a test set; the training set and the test set respectively comprise M known accident types and a plurality of case types contained in each type of accident;
according to each accident type and a plurality of case types corresponding to each accident type, constructing a subgraph corresponding to each accident type to obtain a case affair network layer;
connecting all father nodes according to disaster factors, disaster-bearing bodies, accident types and accident link lines of accident consequences by combining the cause-and-effect relationship of accident occurrence and the accident evolution process to obtain a concept event network layer;
and combining the case affair network layer and the concept affair network layer to obtain the affair map.
Furthermore, in the process of constructing the subgraph, each accident type is used as a father node, all case types corresponding to the current accident type are used as child nodes, and connection is carried out according to whether the inclusion relationship exists between the father node and the child nodes.
Illustratively, the scheme of the present invention is illustrated in FIG. 2, a fact graph is constructed. Structurally, the event graph is a directed cyclic graph, wherein nodes represent events, and directed edges represent sequential, causal, conditional, upper and lower level and other event logical relations among the events.
The affair map adopts a layered network structure:
the first layer is a concept event network layer, wherein each node is an accident node event which is abstracted highly through event nodes reported by M accidents, and connecting lines among the nodes represent an evolution relation among the accidents.
The second layer is a case event network layer, wherein each node is a node of a case event abstracted from an accident report or a case fault tree event node, and connecting lines among the nodes represent an evolution relation among events in the accident.
The connection line between the first layer network node and the second layer network node represents the abstract and generalization relation of the event, the first layer network node represents the more abstract semantic representation of the second layer network node, and the first layer network node can be generalized into any second layer network node.
And (5) constructing a concept affair network layer. Based on the theory of correlation of the event atlas and the correlation research of the antecedent on the accident chain model, the invention establishes a general chemical field event atlas network model by combining the causal relationship and the accident evolution process of the accident according to the accident chain route of the disaster causing factor → disaster carrier → accident type → accident consequence, and further perfects the model according to the historical chemical accident data and the actual investigation.
And constructing a case affair network layer. Data crawling is carried out on accident report websites of related chemical fields through a web crawler, 1661 pieces of accident analysis reports are obtained in total, and 165 pieces of storage tank accident reports are obtained after the accident analysis reports are screened. And extracting events from the crawled accident cases according to the evolution relation of the accidents to describe a description chain of the events, and then constructing a case map of the accidents on the basis.
The concept case network layer is semantically aligned with the case network layer. The invention adopts a crowdsourcing mode to label the knowledge of the text accident report sub-event and the concept case graph nodes, the typical case logic tree nodes and the concept case graph nodes, and a label person can select a certain entity to label the entity, and the labeled knowledge is interconnected and intercommunicated in a plurality of data sources. If the knowledge to be labeled exists in the knowledge base, the completion is intelligently prompted, the labeling time is saved, the labeling efficiency is improved, and the labeling uniformity is ensured.
Further, the training process of the trained attention network model comprises the following steps:
acquiring M known accident types and a plurality of case types contained in each type of accident; wherein M is a positive integer;
constructing subgraphs for each accident type to obtain M subgraphs;
performing vector extraction on the accident report under each accident type to obtain a plurality of text vectors;
inputting M sub-graphs and a plurality of text vectors corresponding to each sub-graph into a first graph attention network together to obtain first vectors of M father nodes;
inputting the first vectors of the M father nodes, the M father nodes in the concept network layer and the connection relations among the M father nodes into a second graph attention network together to obtain second vectors of the M father nodes;
tensor decomposition is carried out on the report text vectors in the test set and second vectors of the M father nodes, vector similarity comparison is carried out, the first graph attention network and the second graph attention network are optimized, and the trained graph attention network model is obtained.
Further, S103: determining the accident type and case type corresponding to the chemical accident report to be detected according to the text vector of the chemical accident report to be detected and the trained graph attention network model; the method specifically comprises the following steps:
inputting the text vector of the chemical accident report to be detected and the subgraph of the chemical accident report to be detected into the trained graph attention network model to obtain a prediction vector of the chemical accident report to be detected;
similarity comparison is carried out on the prediction vector of the chemical accident report to be detected and the second vectors of the M father nodes, so that the accident report type corresponding to the node A with the highest similarity is obtained and is the accident type of the chemical accident report to be detected;
and comparing the similarity of the text vector of the chemical accident report to be detected with the text vectors of all case nodes under the highest similarity node A to obtain a case type corresponding to the highest similarity case node B, wherein the case type is the case type of the chemical accident report to be detected.
By designing a case map describing the evolution rule and mode among events, the case logic relations of sequence, cause and effect, conditions, upper and lower positions and the like among the events can be considered simultaneously in the case evolution deduction process, and the accuracy of abnormal event detection is ensured;
based on the GAT model, detection and event evolution reasoning of specific chemical abnormal events are realized, Embellding is carried out through metadata, and a new abnormal event detection model is constructed by fusing the physical and logical relations of surrounding events through the GAT deep network model.
Illustratively, the working principle of the first graph attention network is:
165 father nodes in the event graph and chemical accident cases under the father nodes are extracted to form 165 sub-graphs, and GAT operation is carried out on each sub-graph:
suppose that a parent node contains N nodes, the feature vector of each node is hi, and the dimension is F, as follows:
h＝{h1,h2,...,hN}，h1∈RF (1)
the node feature vector h is linearly transformed to obtain a new feature vector h 'i, the dimensionality is F', and W is a matrix of linear transformation as shown below:
hi'＝Whi，W∈RF'×F (2)
h'＝{h'1,h'2,...,h'N}，h'i∈RF (3)
if the node j is a neighbor of the node i, the importance of the node j to the node i can be calculated by using an Attention mechanism, namely the Attention Score:
eij＝Attention(Whi,Whj) (4)
GAT specifically orients by stitching together the feature vectors h ' i, h ' j of nodes i, j and then computing the inner product with a 2F ' dimensional vector a. The activation function uses LeakyReLU, and the formula is as follows:
the outputs of the K attentions are multiplied by the number of times on the line.
Beta is a value on the connecting line between the h1 node and the surrounding nodes
After GAT, a 128-dimensional vector representing 165 nodes on the standard graph is extracted.
The 128-dimensional vectors of 165 standard nodes generated by the first graph attention network GAT and fusing the information of the surrounding nodes are extracted as feature vectors of 165 nodes on the concept affairs network layer.
The generated event map is processed again through the attention network of the second graph, and a new 128-dimensional vector is generated in consideration of the event logical relationship such as the sequence, cause and effect, condition, and upper and lower levels between events.
For a new accident report, word segmentation, word stop removal and word stem extraction are adopted, 128 vectors are generated for each word through word2vec, and for a new chemical accident report, the vector of a new accident report text is represented by the average value of the vectors of the words in the text.
And carrying out tensor decomposition on the vectors generating the new accident report text and the generated 165 vectors, carrying out multi-dimensional vector similarity comparison, and returning the generated loss back to the two layers of GAT to optimize the network.
And predicting the node belonging to the standard graph of the accident report according to the trained model. And (3) calculating by adopting similarity: and calculating a vector of the new accident report text with all case nodes under the node. Obtaining the case with the highest matching degree and obtaining the evolution method thereof.
Example two
The embodiment provides a chemical abnormal event detection system based on a matter map;
chemical engineering abnormal event detection system based on a matter map comprises:
an acquisition module configured to: acquiring a chemical accident report to be detected;
a pre-processing module configured to: preprocessing a chemical accident report to be detected to obtain a text vector of the chemical accident report to be detected;
a prediction module configured to: determining the accident type and case type corresponding to the chemical accident report to be detected according to the text vector of the chemical accident report to be detected and the trained graph attention network model;
wherein, the graph attention network model is obtained based on the training of a physiological graph.
It should be noted here that the above-mentioned obtaining module, preprocessing module and prediction module correspond to steps S101 to S103 in the first embodiment, and the above-mentioned modules are the same as examples and application scenarios implemented by the corresponding steps, but are not limited to what is disclosed in the first embodiment. It should be noted that the modules described above as part of a system may be implemented in a computer system such as a set of computer-executable instructions.
In the foregoing embodiments, the descriptions of the embodiments have different emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
The proposed system can be implemented in other ways. For example, the above-described system embodiments are merely illustrative, and for example, the division of the above-described modules is merely a logical division, and in actual implementation, there may be other divisions, for example, multiple modules may be combined or integrated into another system, or some features may be omitted, or not executed.
EXAMPLE III
The present embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein, a processor is connected with the memory, the one or more computer programs are stored in the memory, and when the electronic device runs, the processor executes the one or more computer programs stored in the memory, so as to make the electronic device execute the method according to the first embodiment.
It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general purpose processors, digital signal processors DSP, application specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and so on. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory may include both read-only memory and random access memory, and may provide instructions and data to the processor, and a portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.
In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in a processor or instructions in the form of software.
The method in the first embodiment may be directly implemented by a hardware processor, or may be implemented by a combination of hardware and software modules in the processor. The software modules may be located in ram, flash, rom, prom, or eprom, registers, among other storage media as is well known in the art. The storage medium is located in a memory, and a processor reads information in the memory and completes the steps of the method in combination with hardware of the processor. To avoid repetition, it is not described in detail here.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
Example four
The present embodiments also provide a computer-readable storage medium for storing computer instructions, which when executed by a processor, perform the method of the first embodiment.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A chemical abnormal event detection method based on a matter map is characterized by comprising the following steps:
acquiring a chemical accident report to be detected;
preprocessing a chemical accident report to be detected to obtain a text vector of the chemical accident report to be detected;
determining the accident type and case type corresponding to the chemical accident report to be detected according to the text vector of the chemical accident report to be detected and the trained graph attention network model;
wherein, the graph attention network model is obtained based on the training of a physiological graph.
2. The method for detecting the chemical abnormal events based on the event graph according to claim 1, wherein the chemical accident report to be detected is preprocessed to obtain a text vector of the chemical accident report to be detected; the method specifically comprises the following steps:
performing word segmentation processing on a chemical accident report to be detected;
removing stop words from the word segmentation processing result;
performing stem extraction processing on the processing result of removing stop words to obtain a plurality of stems;
for each extracted stem, generating a stem vector;
and solving the average value of all the word stem vectors, and taking the average value as a text vector of the chemical accident report to be detected.
3. The method for chemical engineering abnormal event detection based on the event graph of claim 1, wherein the graph attention network model comprises: a first graph attention network and a second graph attention network connected in series.
4. The method for detecting the chemical abnormal events based on the event graph as claimed in claim 1, wherein the event graph obtaining process comprises:
constructing a training set and a test set; the training set and the test set respectively comprise M known accident types and a plurality of case types contained in each type of accident;
according to each accident type and a plurality of case types corresponding to each accident type, constructing a subgraph corresponding to each accident type to obtain a case affair network layer;
connecting all father nodes according to disaster factors, disaster-bearing bodies, accident types and accident link lines of accident consequences by combining the cause-and-effect relationship of accident occurrence and the accident evolution process to obtain a concept event network layer;
and combining the case affair network layer and the concept affair network layer to obtain the affair map.
5. The method for detecting the chemical abnormal events based on the event graph as claimed in claim 4, wherein in the process of constructing the sub-graph, each accident type is used as a father node, all case types corresponding to the current accident type are used as child nodes, and the connection is performed according to whether the inclusion relationship exists between the father node and the child nodes.
6. The method for detecting the chemical abnormal events based on the event graph as claimed in claim 1, wherein the training process of the graph attention network model comprises the following specific steps:
acquiring M known accident types and a plurality of case types contained in each type of accident; wherein M is a positive integer;
constructing subgraphs for each accident type to obtain M subgraphs;
performing vector extraction on the accident report under each accident type to obtain a plurality of text vectors;
inputting M sub-graphs and a plurality of text vectors corresponding to each sub-graph into a first graph attention network together to obtain first vectors of M father nodes;
inputting the first vectors of the M father nodes, the M father nodes in the concept network layer and the connection relations among the M father nodes into a second graph attention network together to obtain second vectors of the M father nodes;
tensor decomposition is carried out on the report text vectors in the test set and second vectors of the M father nodes, vector similarity comparison is carried out, the first graph attention network and the second graph attention network are optimized, and the trained graph attention network model is obtained.
7. The method for detecting chemical abnormal events based on a physics map as claimed in claim 1,
determining the accident type and case type corresponding to the chemical accident report to be detected according to the text vector of the chemical accident report to be detected and the trained graph attention network model; the method specifically comprises the following steps:
inputting the text vector of the chemical accident report to be detected and the subgraph of the chemical accident report to be detected into the trained graph attention network model to obtain a prediction vector of the chemical accident report to be detected;
comparing the similarity of the prediction vector of the chemical accident report to be detected with the M second vectors to obtain an accident report type corresponding to the node A with the highest similarity, wherein the accident report type is the accident type of the chemical accident report to be detected;
and comparing the similarity of the text vector of the chemical accident report to be detected with the text vectors of all case nodes under the highest similarity node A to obtain a case type corresponding to the highest similarity case node B, wherein the case type is the case type of the chemical accident report to be detected.
8. Chemical engineering abnormal event detection system based on matter map, characterized by includes:
an acquisition module configured to: acquiring a chemical accident report to be detected;
a pre-processing module configured to: preprocessing a chemical accident report to be detected to obtain a text vector of the chemical accident report to be detected;
a prediction module configured to: determining the accident type and case type corresponding to the chemical accident report to be detected according to the text vector of the chemical accident report to be detected and the trained graph attention network model;
wherein, the graph attention network model is obtained based on the training of a physiological graph.
9. An electronic device, comprising:
a memory for non-transitory storage of computer readable instructions; and
a processor for executing the computer readable instructions,
wherein the computer readable instructions, when executed by the processor, perform the method of any of claims 1-7.
10. A storage medium storing non-transitory computer-readable instructions, wherein the non-transitory computer-readable instructions, when executed by a computer, perform the instructions of the method of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111461707.9A CN114153988A (en) | 2021-12-02 | 2021-12-02 | Chemical abnormal event detection method and system based on matter map |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111461707.9A CN114153988A (en) | 2021-12-02 | 2021-12-02 | Chemical abnormal event detection method and system based on matter map |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114153988A true CN114153988A (en) | 2022-03-08 |
Family
ID=80456153
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111461707.9A Pending CN114153988A (en) | 2021-12-02 | 2021-12-02 | Chemical abnormal event detection method and system based on matter map |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114153988A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114707004A (en) * | 2022-05-24 | 2022-07-05 | 国网浙江省电力有限公司信息通信分公司 | Method and system for extracting and processing case-affair relation based on image model and language model |
-
2021
- 2021-12-02 CN CN202111461707.9A patent/CN114153988A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114707004A (en) * | 2022-05-24 | 2022-07-05 | 国网浙江省电力有限公司信息通信分公司 | Method and system for extracting and processing case-affair relation based on image model and language model |
| CN114707004B (en) * | 2022-05-24 | 2022-08-16 | 国网浙江省电力有限公司信息通信分公司 | Method and system for extracting and processing case-affair relation based on image model and language model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110968699B (en) | Logic map construction and early warning method and device based on fact recommendation | |
| CN112579477A (en) | Defect detection method, device and storage medium | |
| CN103544242A (en) | Microblog-oriented emotion entity searching system | |
| CN111386524A (en) | Facilitating domain and client specific application program interface recommendations | |
| Del Carpio et al. | Trends in software engineering processes using deep learning: a systematic literature review | |
| Gopalakrishnan et al. | Can latent topics in source code predict missing architectural tactics? | |
| CN111344695A (en) | Facilitating domain and client specific application program interface recommendations | |
| Cheema et al. | A natural language interface for automatic generation of data flow diagram using web extraction techniques | |
| Fazayeli et al. | Towards auto-labelling issue reports for pull-based software development using text mining approach | |
| Meusel et al. | Towards automatic topical classification of LOD datasets | |
| Bella et al. | ATLaS: A framework for traceability links recovery combining information retrieval and semi-supervised techniques | |
| Fiacco et al. | Deep neural model inspection and comparison via functional neuron pathways | |
| Gelman et al. | A language-agnostic model for semantic source code labeling | |
| Wang et al. | Exploring semantics of software artifacts to improve requirements traceability recovery: a hybrid approach | |
| Ke et al. | Interpretable test case recommendation based on knowledge graph | |
| CN114153988A (en) | Chemical abnormal event detection method and system based on matter map | |
| Zeman et al. | RDFRules: Making RDF rule mining easier and even more efficient | |
| Füßl et al. | Knowledge graph-based explainable artificial intelligence for business process analysis | |
| Daramola et al. | A conceptual framework for semantic case-based safety analysis | |
| Tovar et al. | Identification of Ontological Relations in Domain Corpus Using Formal Concept Analysis. | |
| Karthikeyan et al. | Ontology based concept hierarchy extraction of web data | |
| Abbas et al. | A module-based approach for structural matching of process models | |
| Gomes et al. | Bert-based feature extraction for long-lived bug prediction in floss: a comparative study | |
| Althar et al. | Software Source Code: Statistical Modeling | |
| Alhumam | Explainable Software Fault Localization Model: From Blackbox to Whitebox. |
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 |