CN110263185B - Construction method of heat consumption knowledge map for novel dry-process cement clinker production - Google Patents
Construction method of heat consumption knowledge map for novel dry-process cement clinker production Download PDFInfo
- Publication number
- CN110263185B CN110263185B CN201910575949.7A CN201910575949A CN110263185B CN 110263185 B CN110263185 B CN 110263185B CN 201910575949 A CN201910575949 A CN 201910575949A CN 110263185 B CN110263185 B CN 110263185B
- Authority
- CN
- China
- Prior art keywords
- knowledge
- agent
- heat consumption
- cement clinker
- basic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention belongs to the technical field of cement clinker heat consumption, and discloses a method for constructing a heat consumption knowledge map for novel dry cement clinker production, which is characterized in that the heat consumption knowledge map for the novel structured dry cement clinker production is constructed by taking the heat consumption knowledge for the cement clinker production as a center and dividing the relationship among basic knowledge, depth knowledge and comprehensive application knowledge; the distributed agent mode is adopted, the agent and the corresponding sub-agent are designed based on three-level characteristics of basic knowledge, depth knowledge and comprehensive application knowledge, and the user can acquire related knowledge by distributing instructions to the agent for execution. The invention analyzes the heat consumption from a basic level, a deep level and an application level, finds the potential relation between the entity and the entity, further finds out the key factors influencing the heat consumption and an effective and efficient method for reducing the heat consumption, thereby achieving the direct purpose of reducing the energy consumption and finally realizing the optimization of the cement industry.
Description
Technical Field
The invention belongs to the technical field of cement clinker heat consumption, and particularly relates to a construction method of a heat consumption knowledge map for novel dry-process cement clinker production.
Background
Currently, the closest prior art: the cement industry is an important basic material industry in China, is also a main high-energy-consumption and high-emission industry in China, and is one of the key points of energy conservation and emission reduction in China. The increase of the energy conservation and emission reduction strength becomes a difficult and urgent task in the cement industry. Therefore, the energy conservation and emission reduction are very slow for the cement industry of China.
The energy consumption of the cement industry is mainly derived from the calcination production of cement clinker. The calcination of cement clinker is a process link with high energy consumption and high heat loss in the cement production process, and has important influence on the heat consumption of cement production. The cement production method generally adopted in China at present is a novel dry-method cement production technology, and the technology gains wide acceptance due to the remarkable superiority and becomes a new development trend of cement production technology, but the defects of high energy consumption, low capacity utilization rate and the like still exist. In order to achieve the aim of energy conservation and emission reduction, the heat consumption analysis of the production process of the novel dry-process cement clinker is necessary. However, the heat consumption analysis is not enough, so that a heat consumption knowledge map for the production of the novel dry-method cement clinker needs to be constructed.
However, in the research at home and abroad, most of the heat consumption in the production process of the novel dry-method cement clinker is analyzed, and no scholars construct a knowledge map of the heat consumption in the production of the novel dry-method cement clinker. Therefore, the construction of the heat consumption knowledge map for the production of the novel dry-process cement clinker has very important significance and value
In summary, the problems of the prior art are as follows: the field of cement clinker production does not form a knowledge base in the field, and most of the production process or policy making process is based on the experience and actual experience of experts or engineers, so that the conclusion and refinement of the knowledge in the field are not achieved; moreover, the knowledge is scattered, distributed in different places, different in appearance form and different in the knowledge required by different people, so that higher requirements are put on knowledge acquisition in the field.
In the research at home and abroad, most of the heat consumption in the production process of the novel dry-method cement clinker is analyzed, and no scholars construct a knowledge map of the heat consumption in the production of the novel dry-method cement clinker.
The difficulty of solving the technical problems is as follows:
how to build a knowledge framework according to the reality of cement clinker production; how to determine relevant knowledge according to different requirements; how to make the knowledge present a dynamic state, that is, different knowledge can be responded automatically according to the requirement, and then the knowledge is recombined to serve different users so as to solve the practical problem.
The significance of solving the technical problems is as follows:
the construction of the novel dry-method cement clinker production heat consumption knowledge map can represent the complex and unintelligible cement clinker heat consumption knowledge field and knowledge system through data mining, information processing, information metering and other modes, not only can provide the reference of all aspects, scientificity, integrity and relation chains for the research of the novel dry-method cement clinker production heat consumption analysis, can help enterprise personnel to obtain relevant information in the industry through searching questions and answers, but also can provide scientific and reliable help for the aspects of decision making, business management and the like of the actual production of the enterprise.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a construction method of a heat consumption knowledge map for producing novel dry-process cement clinker.
The invention is realized in this way, a method for constructing a heat consumption knowledge map for the production of novel dry-process cement clinker, which comprises the following steps:
step one, taking the heat consumption knowledge of cement clinker production as a center, and constructing a structured novel dry-method cement clinker production heat consumption knowledge map by dividing the relationship among basic knowledge, depth knowledge and comprehensive application knowledge.
And step two, designing the agent and the corresponding sub-agent based on three-level characteristics of basic knowledge, depth knowledge and comprehensive application knowledge by adopting a distributed agent mode, and distributing instructions to the agent for execution so that a user can acquire related knowledge.
Further, in the step one, the relationship among the basic knowledge, the depth knowledge and the comprehensive application knowledge is as follows:
the depth knowledge includes a combination of relevant basic knowledge and depth knowledge.
The integrated application knowledge comprises a combination of relevant integrated application knowledge, depth knowledge and basic knowledge.
The concrete description is as follows:
depth knowledge | ∪ basic knowledge | (depth knowledge ∪ basic knowledge).
Integrated application knowledge | ∪ depth knowledge | (integrated application knowledge ∪ depth knowledge).
Wherein, each knowledge includes knowledge in cement clinker production, and can be divided into the following 9 types:
| knowledge categories | Content of related knowledge |
| Knowledge of basic information | Basic information of cement plant (including production line information, cement plant information, product information, production task, etc.) |
| Standard knowledge | The national standard, provincial standard and relevant important document stipulate the standard |
| Knowledge of raw materials | Production of raw materials and contents of the reactions thereof |
| Knowledge of fuel | Production of fuels and contents of reactions thereof |
| Knowledge of the operating procedure | Production process and operation |
| Operational management knowledge | Operation management (function of staff at each level; monitoring of current operation state; abnormality warning, alarming and fault diagnosis) |
| Quality monitoring knowledge | Quality monitoring |
| Knowledge of environmental monitoring | Environmental monitoring |
| Knowledge of equipment maintenance | Maintenance and management of production equipment (electronic instruments, mechanical equipment) |
Further, after the structured heat consumption knowledge graph of the novel dry-method cement clinker production is constructed in the first step, the following steps are required:
the named entity identification through BiLSTM-CRF combined information entropy specifically comprises the following steps:
and in the named entity recognition process, word segmentation and labeling are carried out on the national standard document text, model building and training are completed through word vectorization and feature vectors, named entity recognition is realized, and a recognition result is output.
And further, in the second step, the Agent and corresponding sub-agents are designed based on three levels of characteristics of basic knowledge, depth knowledge and comprehensive application knowledge, the field knowledge of the cement clinker is constructed by utilizing the abstract components, and nine different sub-agents are divided to form the abstract component Agent. The abstract component Agent specifically comprises a basic information knowledge subagent, a standard knowledge subagent, a raw material knowledge subagent, a fuel knowledge subagent, an operation process knowledge subagent, an operation management knowledge subagent, a quality monitoring knowledge subagent, an environment monitoring knowledge subagent and an equipment maintenance management knowledge subagent, wherein each seed Agent manages a knowledge component of a corresponding type; the abstract component Agent is an abstract component of 3 different levels of knowledge.
Further, the second step is to assign the instruction to the agent for execution, so that the method for the user to acquire the relevant knowledge comprises the following steps:
firstly, when a user initiates an access request to the system, Agenthub preliminarily analyzes the input content of the user, scores according to the matching degree of the vocabulary library and the number of entities or concepts contained in the input content based on three vocabulary libraries including basic knowledge, deep analysis and comprehensive application.
And secondly, when the difference of the three scores is obvious, transmitting the analysis result to the field with the highest score and the best matching degree. And when the three scores are smaller in difference, reminding the user to select the fields of basic knowledge, depth knowledge and comprehensive application knowledge.
And after the analysis result of the user is transmitted to the corresponding field, the field performs deep analysis on the analysis result to determine the affiliated category, and allocates the search task to the affiliated category sub-agent, and the affiliated category agent performs related knowledge search.
When a user result is fed back, if only knowledge in a certain single level is fed back, the user purpose is often not achieved; ordering the three levels of complexity from high to low into comprehensive application knowledge > depth knowledge > basic knowledge; in order to realize multi-dimension of output knowledge, when a user searches for high-level knowledge, the subagent also calls the same-name subagent in a low-level field to inquire.
And thirdly, transmitting the search results of each layer to a knowledge gathering agent for integration, returning the results to the Agenthub by the knowledge gathering agent after integration, and feeding back the results to the user by the Agenthub.
The invention also aims to provide the novel heat consumption knowledge map for the dry-process cement clinker production, which is constructed by the construction method for the heat consumption knowledge map for the dry-process cement clinker production.
The invention has the advantages and positive effects that:
the invention takes the heat consumption knowledge of cement clinker production as the center, and constructs a structured novel dry-method cement clinker production heat consumption knowledge map from top to bottom from three layers of foundation, depth and application through Agenthub so as to realize the completeness and hierarchy of the knowledge map. By adopting a distributed agent mode, the sub-agents are designed based on three-level characteristics, and instructions are distributed to a plurality of sub-agents to be executed, so that a user can more efficiently, accurately and quickly acquire related knowledge, and the high efficiency and the accuracy of knowledge acquisition are realized.
The invention takes the heat consumption knowledge of cement clinker production as the center, and constructs a structured novel heat consumption knowledge map of dry cement clinker production from three levels of foundation, depth and application so as to realize the completeness and hierarchy of the knowledge map.
By adopting a distributed agent mode, the agent and the corresponding sub-agents are designed based on three-level characteristics, and the instructions are distributed to the agent for execution, so that a user can more efficiently, accurately and rapidly acquire related knowledge, and the high efficiency and the accuracy of knowledge acquisition are realized.
The invention utilizes the technologies of data mining, intelligent analysis and the like to establish a comprehensive and complete novel dry-method cement heat consumption knowledge map from three layers of foundation, depth and application, through named entity recognition of BilSTM-CRF combined with information entropy, through entity relation extraction of a BilSTM model based on attention mechanism, and carries out conflict resolution on the established knowledge. The established knowledge platform can provide corresponding knowledge for the production of the cement clinker and guidance for solving related problems, can be better referred by workers or engineers on site, and provides a guidance scheme, so that the aims of reducing the heat consumption in the whole production process of the cement clinker, saving fuel and reducing the production cost are fulfilled, and the energy conservation and optimization of the production are realized.
According to the invention, by constructing a heat consumption knowledge map for the production of the novel dry-method cement clinker, the heat consumption in the production process of the novel dry-method cement clinker is more scientifically and efficiently analyzed from a basic level, a depth level and an application level, the potential relation between an entity and the entity is found, and then key factors influencing the heat consumption are found out and an effective and efficient method for reducing the heat consumption is found, so that the direct purpose of reducing the energy consumption is achieved, and the optimization of the cement industry is finally realized.
Drawings
FIG. 1 is a flow chart of a construction method of a heat consumption knowledge map for production of novel dry cement clinker provided by the embodiment of the invention.
FIG. 2 is a diagram of the knowledge architecture of the basic, deep and application three-level build provided by an embodiment of the present invention.
Fig. 3 is an abstraction component Agent diagram provided by an embodiment of the present invention.
FIG. 4 is an architecture diagram of a knowledge graph provided by an embodiment of the present invention.
FIG. 5 is a diagram of an improved BilSTM-CRF model according to an embodiment of the present invention.
FIG. 6 is an exemplary diagram of input corpora of relationship extraction according to an embodiment of the present invention.
Fig. 7 is a diagram of basic knowledge evaluation analysis provided by an embodiment of the present invention.
In the figure: (a) and basic knowledge entity distribution. (b) The underlying knowledge constitutes an analysis.
FIG. 8 is a diagram of an example of a basic knowledge portion atlas visualization provided by an embodiment of the invention.
Fig. 9 is a basic knowledge query information presentation 1 provided by the embodiment of the present invention.
Fig. 10 is a basic knowledge query information example presentation 2 provided by the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In the research at home and abroad, most of the heat consumption in the production process of the novel dry-method cement clinker is analyzed, and no scholars construct a knowledge map of the heat consumption in the production of the novel dry-method cement clinker.
Aiming at the problems in the prior art, the invention provides a method for constructing a heat consumption knowledge map in the production of novel dry-method cement clinker, and the invention is described in detail below by combining the attached drawings.
As shown in fig. 1, the method for constructing a heat consumption knowledge map for the production of the novel dry cement clinker provided by the embodiment of the invention comprises the following steps:
s101, taking the heat consumption knowledge of cement clinker production as a center, and constructing a structured novel dry-process cement clinker production heat consumption knowledge map by dividing the relationship among basic knowledge, depth knowledge and comprehensive application knowledge.
S102, designing the agent and the corresponding sub-agents based on three-level characteristics of basic knowledge, depth knowledge and comprehensive application knowledge by adopting a distributed agent mode, and enabling a user to acquire related knowledge by distributing instructions to the agents to execute.
Step S101, the relationship among the basic knowledge, the depth knowledge and the comprehensive application knowledge is as follows:
the depth knowledge includes a combination of relevant basic knowledge and depth knowledge.
The integrated application knowledge comprises a combination of relevant integrated application knowledge, depth knowledge and basic knowledge.
The concrete description is as follows:
depth knowledge | ∪ basic knowledge | (depth knowledge ∪ basic knowledge).
Integrated application knowledge | ∪ depth knowledge | (integrated application knowledge ∪ depth knowledge).
Step S101, after a structured heat consumption knowledge map for the production of the novel dry-process cement clinker is constructed, the following steps are carried out:
the named entity identification through BiLSTM-CRF combined information entropy specifically comprises the following steps:
and in the named entity recognition process, word segmentation and labeling are carried out on the national standard document text, model building and training are completed through word vectorization and feature vectors, named entity recognition is realized, and a recognition result is output.
Step S102, designing an Agent and corresponding sub-agents based on three levels of characteristics of basic knowledge, depth knowledge and comprehensive application knowledge, and constructing the field knowledge of the cement clinker by using the abstract components and dividing nine different sub-agents to form the abstract component Agent. The abstract component Agent concretely comprises a basic information knowledge sub-Agent, a standard knowledge sub-Agent, a raw material knowledge sub-Agent, a fuel knowledge sub-Agent, an operation process knowledge sub-Agent, an operation management knowledge sub-Agent, a quality monitoring knowledge sub-Agent, an environment monitoring knowledge sub-Agent and an equipment maintenance management knowledge sub-Agent, wherein each sub-Agent manages knowledge components of corresponding types.
Step S102, distributing instructions to agents for execution, and the method for enabling users to acquire relevant knowledge comprises the following steps:
firstly, when a user initiates an access request to the system, Agenthub preliminarily analyzes the input content of the user, scores according to the matching degree of the vocabulary library and the number of entities or concepts contained in the input content based on three vocabulary libraries including basic knowledge, deep analysis and comprehensive application.
And secondly, when the difference of the three scores is obvious, transmitting the analysis result to the field with the highest score and the best matching degree. And when the three scores are smaller in difference, reminding the user to select the fields of basic knowledge, depth knowledge and comprehensive application knowledge.
And after the analysis result of the user is transmitted to the corresponding field, the field performs deep analysis on the analysis result to determine the affiliated category, and allocates the search task to the affiliated category sub-agent, and the affiliated category agent performs related knowledge search.
When a user result is fed back, if only knowledge in a certain single level is fed back, the subagent calls a homonymous subagent in a low-level field to inquire and complete knowledge supplementation in the certain single level by utilizing a principle of sequencing comprehensive application knowledge, depth knowledge and basic knowledge from top to bottom.
And thirdly, transmitting the search results of each layer to a knowledge gathering agent for integration, returning the results to the Agenthub by the knowledge gathering agent after integration, and feeding back the results to the user by the Agenthub.
The invention is further described below with reference to specific examples and analyses.
1 knowledge architecture.
Throughout the cement production process, a great deal of knowledge is involved. In order to ensure the integrity and accuracy of the constructed knowledge graph as much as possible, the knowledge constructed from three levels of foundation, depth and application is determined, as shown in a knowledge framework of FIG. 2.
In FIG. 2, each level of knowledge is an independent knowledge component. The depth knowledge includes a combination of relevant basic knowledge and depth knowledge. The integrated application knowledge comprises a combination of relevant integrated application knowledge, depth knowledge and basic knowledge.
Depth knowledge | ∪ basic knowledge | (depth knowledge ∪ basic knowledge).
Integrated application knowledge | ∪ depth knowledge | (integrated application knowledge ∪ depth knowledge).
In addition, according to factors such as the process and the process of cement clinker production, in order to improve the accuracy of acquiring related heat consumption knowledge, the cement clinker production knowledge is roughly divided into nine types, see table 1, and the types of knowledge basically cover main application scenes such as the process, the management, the raw materials and the environmental protection of the cement clinker production.
TABLE 1 Cement Clinker production-related knowledge partitioning and Contents
2 building domain knowledge of cement clinker using abstract components.
Aiming at the commonalities of the knowledge of the fields related to the basic, deep and comprehensive applications, the method designs nine different sub-agents to form an abstract component Agent according to the knowledge division of the table 1, as shown in fig. 3.
In fig. 3, the abstract component Agent includes a basic information knowledge sub-Agent (Agent _ BasicInfo), a standard knowledge sub-Agent (Agent _ Norm), a raw material knowledge sub-Agent (Agent _ raw material), a Fuel knowledge sub-Agent (Agent _ Fuel), an operation process knowledge sub-Agent (Agent _ process operation), an operation management knowledge sub-Agent (Agent _ running management), a quality monitoring knowledge sub-Agent (Agent _ QualityMonitor), an environment monitoring knowledge sub-Agent (Agent _ environment manager), and a device dimension management knowledge sub-Agent (Agent _ equipment configuration), each of which manages a corresponding type of knowledge component.
The three levels of knowledge can complete the realization of different functions by inheriting the abstract component Agent.
3. Architecture and mechanism of knowledge graph.
According to the knowledge construction and the three-level knowledge construction method, a knowledge graph framework as shown in FIG. 4 is designed.
FIG. 4 illustrates well the advantages of the architecture of the present knowledge-graph, which is generally qualitative (since the architecture diagram is most important):
the Agenthub can be used for collecting the requirements of users and assigning different task downlinks according to specific requirements, so that bottleneck is not caused when the requirements are acquired, and congestion which may occur when the requirements are acquired is solved;
different agents can process different knowledge, so that the search space is reduced, and the processing efficiency is improved;
the division of the three levels of knowledge is also from the perspective of applying knowledge, so that different levels of knowledge services can be provided for users.
In order to more accurately acquire and feed back relevant knowledge required by a user, the AgentHub is provided for analyzing the user problem and summarizing a final result.
When a user initiates an access request to the system, firstly Agenthub preliminarily analyzes the input content of the user, and scores the input content according to the matching degree of the input content and the vocabulary library and the number of entities or concepts contained in the input content based on three vocabulary libraries including basic knowledge, deep analysis and comprehensive application. And when the difference of the three scores is obvious, transmitting the analysis result to the field with the highest score and the best matching degree. When the three scores are smaller in difference, the user is reminded to select the three fields, and the accuracy of the output result is improved.
And after the analysis result of the user is transmitted to the corresponding field, the field performs deep analysis on the analysis result to determine the affiliated category, and allocates the search task to the affiliated category sub-agent, and the affiliated category agent performs related knowledge search.
When the user result is fed back, if only knowledge in a certain single level is fed back, the user purpose is often not achieved. The three levels of complexity are considered to be ordered from high to low as "comprehensive application knowledge > deep knowledge > basic knowledge". In order to realize multi-dimension of output knowledge, when a user searches for high-level knowledge, the subagent also calls the same-name subagent in a low-level field to inquire.
For example, when a user searches basic information knowledge of a cement plant subordinate to the integrated application level, the subagent Agent _ BasicInfo of the integrated application level calls the subagent Agent _ BasicInfo of the deep knowledge level and the basic knowledge level respectively to perform information query, so as to supplement and improve the integrated application knowledge.
Finally, the search results of each level are transmitted to a knowledge gathering agent for integration. And the integrated knowledge summarizing agent returns the result to the Agenthub and feeds the result back to the user by the Agenthub.
4. The invention is further described below in connection with the construction of a knowledge graph (examples).
Reading knowledge from national standards is taken as an example to construct basic knowledge. The depth knowledge is substantially similar to the method of applying knowledge synthetically.
Knowledge is represented by a triple < entity 1, the relationship between entity 1 and entity 2, entity 2 >. The experiment of entity relationship extraction and named entity identification is evaluated by the accuracy P, the recall ratio R and the F1, and the corresponding calculation is respectively shown as formula (1), formula (2) and formula (3).
P＝n/N (1)。
R＝n/M (2)。
F1＝2×P×R/(P+R) (3)。
Wherein: n represents the number of correct relationships in the experimental results, N represents the total number of entity relationship labels in the experimental results, and M represents the total number of entity relationships in the test set.
4.1 named entity recognition procedure.
In consideration of using an algorithm with high recognition accuracy of the named entity as much as possible, the invention combines a BilSTM-CRF-MI system with information entropy to perform word segmentation and labeling on the national standard document text, completes model construction and training from two aspects of word vectorization and feature vector construction, further realizes the recognition of the named entity, and outputs a recognition result, wherein the model structure is shown in figure 5. The results are shown in Table 2 in comparison with CRF-MI.
TABLE 2 experimental results for CRF-MI
As can be seen from the experimental results in table 2, a screening algorithm of mutual information entropy (MI) is also added to the CRF layer, and the recognition accuracy of the BiLSTM-CRF for 5 types of named entities, namely terms, equipment names, material names, entities and dimension units, is relatively high, and is significantly improved compared with the CRF model. It can be seen that the learning model based on the BilSTM-CRF achieves better effect on the named entity recognition task aiming at the related field. Especially for some combined named entities, such as words containing a large number of combinations in terms, the ability to learn long-term dependencies in the LSTM model achieves better performance. From the recognition rate, BilSTM has a larger improvement on the 5 types of named entities compared with the traditional CRF model, which should be because CRF is a probability-based model, and the named entities in the cement production energy consumption field are usually spliced by some basic named entities, thus causing a certain interference to the classification of CRF. From the point of view of the overall recall rate, the recall rate of the BilSTM and the CRF is similar only on the dimension unit, which is also probably because the dimension is not distributed in the national standard text of the cement production, so that the related training corpora are not enough to make the feature learning unobvious.
In summary, from experimental comparison results, in the named entity identification task in the subdivided field of the national standard field of energy consumption for cement production, the model learning identification method based on the BilSTM-CRF-MI algorithm has greater advantages than the traditional statistical learning-based method.
4.2 entity relationship extraction.
After identifying the named entities, and after inputting the expectation of FIG. 6, the BiLSTM-Attention model of Attention mechanism is introduced to extract the entity relationship of national standard documents of energy consumption for cement production, and the experimental results of the named entities, CRF and BiLSTM are shown in Table 3.
TABLE 3 results of the experiment
From the experimental results, it can be seen that, in the extraction experiment based on the context domain relationship, the effect of the model based on the BiLSTM-Attention on relationship identification is greatly improved compared with the BiLSTM model without the Attention mechanism. This should be because when the sequence of the attention mechanism is input, as the sequence grows, the performance of the recognition method of the original BilSTM will decrease, and the corresponding entity words before and after the sentence cannot be taken care of. This is due to the structural problem of the BiLSTM original model design, i.e., all context information is limited to a fixed length, so that the learning capability of the BiLSTM model is limited.
4.3 basic domain knowledge component analysis.
The field knowledge base established by the national standard related to the cement production energy consumption has complete entity types including terms, parameter names, equipment names, material names, dimension units and the like, and is shown in figure 7((a) basic knowledge entity distribution and (b) basic knowledge composition analysis). Since the knowledge information is generally derived from national standard documents for energy consumption of cement production issued by the national standards institute, the authority of the knowledge is not in doubt. In general, the national standard knowledge base in the field of cement production energy consumption constructed by the method has higher potential application value, and related construction methods and experiences can be extended and reused in other fields.
In an embodiment of the invention, fig. 8 is a diagram of an example of visualization of a basic knowledge portion atlas. Fig. 9 shows basic knowledge query information 1. Fig. 10 shows fig. 2 for an example of basic knowledge query information.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (5)
1. A construction method of a heat consumption knowledge map for the production of novel dry-process cement clinker is characterized by comprising the following steps:
dividing the relation among basic knowledge, depth knowledge and comprehensive application knowledge in the cement clinker production heat consumption knowledge to construct a structured novel dry-method cement clinker production heat consumption knowledge map;
designing an agent and a corresponding sub-agent based on three-level characteristics of basic knowledge, depth knowledge and comprehensive application knowledge by adopting a distributed agent mode, and enabling a user to acquire related knowledge by distributing instructions to the agent for execution;
the method for distributing the instructions to the agents to be executed to enable the users to acquire the relevant knowledge comprises the following steps:
firstly, when a user initiates an access request to a system, Agenthub preliminarily analyzes input contents of the user, scores according to matching degree with a vocabulary library and the number of entities or concepts contained in the input contents based on three vocabulary libraries including basic knowledge, deep analysis and comprehensive application;
secondly, when the difference of the three scores is obvious, transmitting the analysis result to the field with the highest score and the best matching degree; when the three scores are smaller in difference, reminding a user to select the fields of basic knowledge, depth knowledge and comprehensive application knowledge;
after the analysis result of the user is transmitted to the corresponding field, the field carries out deep analysis on the analysis result to determine the category of the analysis result, the search task is distributed to the sub-agents of the category of the analysis result, and the agents of the category of the analysis result carry out searching of related knowledge;
when a user result is fed back, if only knowledge in a certain single level is fed back, the user purpose is often not achieved; ordering the three levels of complexity from high to low into comprehensive application knowledge > depth knowledge > basic knowledge; in order to realize multi-dimensionality of output knowledge, when a user searches for high-level knowledge, the subagents also call the same-name subagents in the low-level field to inquire;
and thirdly, transmitting the search results of each layer to a knowledge gathering agent for integration, returning the results to the Agenthub by the knowledge gathering agent after integration, and feeding back the results to the user by the Agenthub.
2. The method for constructing the heat consumption knowledge-graph in the production of the novel dry-process cement clinker as claimed in claim 1, wherein in the step one, the relationship among the basic knowledge, the depth knowledge and the comprehensive application knowledge is as follows:
the depth knowledge comprises a combination of related basic knowledge and depth knowledge;
the comprehensive application knowledge comprises the combination of related comprehensive application knowledge, depth knowledge and basic knowledge;
the concrete description is as follows:
depth knowledge = depth knowledge | ∪ basic knowledge | (depth knowledge ∪ basic knowledge);
integrated application knowledge = integrated application knowledge | ∪ depth knowledge | (integrated application knowledge ∪ depth knowledge).
3. The method for constructing the heat consumption knowledge map for the production of novel dry cement clinker as claimed in claim 1, wherein the step one is carried out after constructing the heat consumption knowledge map for the production of the structured novel dry cement clinker:
the named entity identification through BiLSTM-CRF combined information entropy specifically comprises the following steps:
and in the named entity recognition process, word segmentation and labeling are carried out on the national standard document text, model building and training are completed through word vectorization and feature vectors, named entity recognition is realized, and a recognition result is output.
4. The method for constructing the heat consumption knowledge map for the production of the novel dry-process cement clinker as claimed in claim 1, wherein in the second step, the Agent and the corresponding sub-agency are designed based on three levels of characteristics of basic knowledge, depth knowledge and comprehensive application knowledge, the field knowledge of the cement clinker is constructed by using the abstract component, and nine different sub-agents are divided to form the abstract component Agent; the abstract component Agent concretely comprises a basic information knowledge sub-Agent, a standard knowledge sub-Agent, a raw material knowledge sub-Agent, a fuel knowledge sub-Agent, an operation process knowledge sub-Agent, an operation management knowledge sub-Agent, a quality monitoring knowledge sub-Agent, an environment monitoring knowledge sub-Agent and an equipment maintenance management knowledge sub-Agent, wherein each sub-Agent manages knowledge components of corresponding types.
5. A novel heat consumption knowledge map for dry cement clinker production constructed by the construction method of the heat consumption knowledge map for dry cement clinker production according to claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910575949.7A CN110263185B (en) | 2019-06-28 | 2019-06-28 | Construction method of heat consumption knowledge map for novel dry-process cement clinker production |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910575949.7A CN110263185B (en) | 2019-06-28 | 2019-06-28 | Construction method of heat consumption knowledge map for novel dry-process cement clinker production |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110263185A CN110263185A (en) | 2019-09-20 |
| CN110263185B true CN110263185B (en) | 2020-06-30 |
Family
ID=67922866
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910575949.7A Active CN110263185B (en) | 2019-06-28 | 2019-06-28 | Construction method of heat consumption knowledge map for novel dry-process cement clinker production |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110263185B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111191047A (en) * | 2019-12-31 | 2020-05-22 | 武汉理工大学 | Knowledge graph construction method for human-computer cooperation disassembly task |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107341215A (en) * | 2017-06-07 | 2017-11-10 | 北京航空航天大学 | A kind of vertical knowledge mapping classification ensemble querying method of multi-source based on Distributed Computing Platform |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9542648B2 (en) * | 2014-04-10 | 2017-01-10 | Palo Alto Research Center Incorporated | Intelligent contextually aware digital assistants |
-
2019
- 2019-06-28 CN CN201910575949.7A patent/CN110263185B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107341215A (en) * | 2017-06-07 | 2017-11-10 | 北京航空航天大学 | A kind of vertical knowledge mapping classification ensemble querying method of multi-source based on Distributed Computing Platform |
Non-Patent Citations (1)
| Title |
|---|
| 自动化构建的中文知识图谱系统;鄂世嘉 等;《计算机应用》;20160410;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110263185A (en) | 2019-09-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112685385B (en) | Big data platform for smart city construction | |
| Wielinga et al. | Methods and techniques for knowledge management: what has knowledge engineering to offer? | |
| US7383234B2 (en) | Extensible data mining framework | |
| CN106202207A (en) | A kind of index based on HBase ORM and searching system | |
| CN109344223A (en) | A kind of Building Information Model management system and method based on cloud computing technology | |
| CN104111998A (en) | Method and device for sorting coding and integrated exchange and management of heterogeneous data of enterprise | |
| CN112241402A (en) | Empty pipe data supply chain system and data management method | |
| Utepbergenov et al. | About the concept of information support system for innovative economy in the Republic of Kazakhstan | |
| CN103164476A (en) | Execution method and execution device of applying metadata to describe files in business intelligence (BI) | |
| CN100530086C (en) | Embedded software drive management method | |
| CN110263185B (en) | Construction method of heat consumption knowledge map for novel dry-process cement clinker production | |
| CN106599216A (en) | Computer based training courseware publishing system | |
| Radermacher et al. | Trends in environmental information processing | |
| Wang et al. | A Web-based CBR knowledge management system for PC troubleshooting | |
| CN111784192A (en) | Industrial park emergency plan executable system based on dynamic evolution | |
| Abrahiem | A new generation of middleware solutions for a near-real-time data warehousing architecture | |
| WO2021066231A1 (en) | Sql query recommendation method and system | |
| Shayakhmetova et al. | Descriptive big data analytics in the field of education | |
| CN114418585A (en) | Intelligent contract generation method and device and related equipment | |
| Clements | Comparing the SEI's views and beyond approach for documenting software architectures with ansi-ieee 1471-2000 | |
| Zhou | Knowledge dichotomy and semantic knowledge management | |
| Gringinger et al. | Ontology-based CNS software development | |
| KR101979242B1 (en) | Knowledge providing method in construction knowledge hub-center | |
| Osonuga et al. | Lifecycle of a sub-metered tertiary multi-use (GreEn-ER) building’s open energy data: from resource mobilisation to data re-usability | |
| Zhang et al. | Research on knowledge management for group enterprise in cloud manufacturing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |