Building operation knowledge graph construction method based on industry specifications and existing ontology

Abstract

The application discloses a building operation and maintenance knowledge graph construction method based on industry norms and existing ontology, relates to the field of knowledge graph construction, and comprises the following steps: reusing an existing building ontology to obtain an IfcOWL ontology; classifying norm provisions, performing word segmentation and entity identification on the norm provisions, constructing norm knowledge expressions, and obtaining a building operation and maintenance norm ontology by means of norm analysis; structuring the norm knowledge expressions, realizing norm translation based on an OWL language, and establishing the building operation and maintenance norm ontology; fusing the building operation and maintenance norm ontology with the IfcOWL ontology to construct a green building operation and maintenance knowledge ontology; and constructing a green building operation and maintenance knowledge graph based on a top-down and bottom-up combination method. The application constructs a green building operation and maintenance knowledge graph based on industry norms, which has great significance for giving full play to the role of norm knowledge and improving the green building operation and maintenance management level. The application improves the construction efficiency and scientific nature of the field knowledge graph.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph construction technology, specifically to a method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies. Background Technology

[0002] Building operation and maintenance is a necessary means to ensure the continuous and efficient operation of building facilities. Among them, the green building field has long suffered from the problem of "emphasizing design and neglecting operation and maintenance", and the level of intelligence of operation and maintenance management also needs to be improved. As a structured and relational semantic network graph, knowledge graph can integrate and manage data and information related to building operation and maintenance, and realize intelligent operation and maintenance. Zhu Qing et al. proposed a top-down construction method of the model layer and a bottom-up construction method of the data layer, which integrates the key element concepts and semantic relationships related to safety, quality and progress in the construction process of railway tunnels into the knowledge graph (Zhu Qing, Wang Suozhi, Ding Yulin, et al. Construction method of knowledge graph for safety, quality and progress of intelligent management of railway tunnel drilling and blasting construction [J]. Journal of Wuhan University (Information Science Edition), 2022, 47(8):1155-1164.). Yang Xiaoxia et al. proposed a knowledge graph construction and knowledge question answering method in the field of bridge inspection to address the problems of insufficient data extraction and integration in bridge inspection reports and insufficient knowledge question answering services in the maintenance decision-making process (Yang Xiaoxia, Yang Jianxi, Li Ren, et al. Knowledge graph construction and knowledge question answering method in the field of bridge inspection [J]. Computer Applications, 2022, 42(S1):28-36.). Liu Yue proposed a method for constructing a standard semantic model for fire protection review of BIM models to address the problems that have occurred in the current fire protection design review work (Liu Yue. Invention of a standard semantic model construction method for fire protection review of BIM models [D]. Beijing University of Civil Engineering and Architecture, 2022.). Chen Yuan et al. completed the compliance check process by conducting knowledge analysis on the clauses in the standard and summarizing the standard knowledge expression (Chen Yuan, Zhang Yu, Kang Hong. Invention of an automatic inspection system for compliance of BIM model building design based on knowledge management [J]. Journal of Graphics, 2020, 41(3):490-499.).

[0003] A review of existing knowledge graph construction methods in the construction field reveals that, although the application of knowledge graphs in the construction field is becoming increasingly diverse, there are still some shortcomings that need improvement:

[0004] (1) The application of knowledge graphs in the field of green building operation and maintenance still needs to be explored. Existing knowledge graphs and building ontologies lack knowledge in the field of green building operation and maintenance, while green buildings urgently need to use digital technology to achieve operation and maintenance goals. When using digital technology to solve green building operation and maintenance problems, most of them only focus on a single perspective such as water conservation, energy conservation and management, and lack technical support for the entire life cycle of operation and maintenance problems.

[0005] (2) Existing building ontologies were not fully reused when constructing the schema layer. Some existing technologies either did not reuse existing building ontologies or only referenced a small portion of them. Although there is no relatively unified reference standard for the hierarchical structure and logical relationships of the schema layer in many domain knowledge graph construction methods, the design of existing building ontologies usually involves the participation of multiple experts and the integration of domain knowledge, and can draw on the best knowledge and consensus from the perspectives and experiences of multiple experts. Therefore, the reuse or extension of existing building ontologies should be given priority when constructing the schema layer.

[0006] (3) Industry standard knowledge was neglected when constructing the pattern layer. Industry standards are formed based on years of practice and experience, representing the consensus of professionals in the industry, and establishing a unified set of methods and standards for the relevant fields. At present, the construction of the pattern layer is mainly based on certain specific methods, often extracting industry standard knowledge and knowledge from other sources on an equal footing, without making full use of industry standard knowledge.

[0007] Furthermore, a review of relevant literature on building operation and maintenance code knowledge ontology reveals that while some progress has been made in ontology-based compliance review, the following challenges and issues remain in code knowledge modeling and building ontology application:

[0008] (1) Most code provisions are difficult to express in a structured way. Existing code knowledge modeling only targets provisions that are easy to express in a structured way. These provisions generally have clear attribute values ​​or inclusion relationships and spatial relationships. However, the code also contains a large number of provisions that are difficult for computers to recognize. These provisions, as knowledge assets of the construction industry, carry a wealth of professional knowledge and experience. Their absence will greatly affect the completeness of the domain knowledge graph.

[0009] (2) Some information is missing during knowledge modeling. The main information in the normative clauses usually includes building components, component attributes, comparison terms, and attribute values, etc., and existing methods are all based on the main information for knowledge modeling. However, the normative clauses also include the normative number and modal words indicating the degree of strictness. This information, as part of the normative knowledge, is usually ignored in the knowledge modeling process. Storing this secondary information in a knowledge graph is beneficial to improving the accuracy of normative translation and further conducting knowledge reasoning in related fields.

[0010] (3) The translation of specifications and the extension of existing building ontology are quite difficult to develop. At present, the translation of specifications generally uses SWRL (Semantic Web Rule Language) rules to formally express the specification clauses, but the complex definition of SWRL rules greatly increases the development cost of ontology construction.

[0011] Furthermore, regarding the extension of IfcOWL (which defines a Building Information Modeling (IFC) ontology that can be represented using the OWL language), most scholars choose to extend the custom attribute set of the IFC standard based on the EXPRESS language, then establish a mapping framework between EXPRESS and OWL, and finally construct an IfcOWL ontology containing relevant domain knowledge. However, this extension path is rather cumbersome, and the extended content is strictly limited to the framework of the IFC standard itself. Summary of the Invention

[0012] To address the challenge of constructing a knowledge graph applicable to the field of building operation and maintenance based on existing building ontology and industry standards, this invention provides a method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontology. Constructing a green building operation and maintenance knowledge graph based on industry standards has practical significance for solving building operation and maintenance problems, and also provides a theoretical basis for the construction of knowledge graphs in the building field.

[0013] The technical solution adopted by this invention to solve the technical problem is as follows:

[0014] The method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontology, as described in this invention, includes the following steps:

[0015] Step 1: Reuse the existing building body and obtain the IfcOWL body;

[0016] Step 2: Through standard analysis, classify the standard provisions, perform word segmentation and entity recognition on the standard provisions, construct standard knowledge expressions, and obtain the building operation and maintenance standard ontology;

[0017] Step 3: Structure the standard knowledge expressions, realize standard translation based on OWL language, and establish a building operation and maintenance standard ontology; integrate the building operation and maintenance standard ontology with the IfcOWL ontology to construct a green building operation and maintenance knowledge ontology;

[0018] Step 4: Construct a green building operation and maintenance knowledge graph based on a combination of top-down and bottom-up approaches.

[0019] Furthermore, in step one, the existing building body reuse method is as follows:

[0020] (1) When a new node has the same meaning as an existing node, the existing node in IfcOWL is used directly;

[0021] (2) When the new node is a subclass of an existing node, the existing node needs to be extended.

[0022] Furthermore, in step one, ChatGPT is used to assist in the reuse of existing building structures.

[0023] Furthermore, in step two, the specific operational procedure for classifying the regulatory provisions is as follows:

[0024] (1) From the perspective of normative translation, normative clauses are divided into easy-to-structure and difficult-to-structure clauses; structuring is the process of converting normative clauses expressed in natural language into computer-readable statements; knowledge in knowledge graphs is represented by a structure of (entity)-[relationship]-(entity) or (entity)-{attribute:attribute value}. If a normative clause can be represented by a triple structure of a knowledge graph after being decomposed, it is called easy-to-structure clause; otherwise, it is called difficult-to-structure clause.

[0025] (2) From the perspective of the content of constraints, normative clauses are divided into six types: attribute constraints, relational constraints, normative constraints, action constraints, state constraints and non-constraints;

[0026] (3) From the perspective of constraint form, normative clauses with explicit values ​​are called formalizable normative clauses, and their meaning is expressed through data attributes in the OWL language; other normative clauses without explicit values ​​are called informal clauses or semi-formal clauses, and their meaning is expressed through object attributes.

[0027] Furthermore, in step two, the specific operational procedures for word segmentation and entity recognition of the specification text are as follows:

[0028] The standard clauses are processed by breaking down complex long sentences into short sentences, and then further breaking down these short sentences into individual words. After standard word segmentation, entity recognition is performed on various types of words, including: operation and maintenance objects, operation and maintenance attributes, operation and maintenance actions, operation and maintenance status, preconditions, modal words, quantity comparison words, relational words, attribute values, standard names, standard numbers, and clause numbers.

[0029] Furthermore, in step two, after completing the standardized word segmentation and entity recognition, each standardized clause is split into a set of multiple semantic elements.

[0030] Furthermore, in step three, the Neo4j graph database is used to store the green building operation and maintenance knowledge ontology.

[0031] Furthermore, the specific operational procedures for step three are as follows:

[0032] Step 3.1 involves directly extending IfcOWL on the Protégé platform using the OWL language;

[0033] Step 3.2 Utilize the Protégé platform and WebVOWL visualization tool to perform specification translation and achieve ontology modeling, establishing the building operation and maintenance specification ontology;

[0034] Step 3.3 Integrate the building operation and maintenance specification ontology with the IfcOWL ontology to construct a green building operation and maintenance knowledge ontology.

[0035] Furthermore, the specific operation procedure for step 3.2 is as follows:

[0036] ① The precondition is defined as a subclass of a certain object or property through semantic interpretation;

[0037] ② When there are clauses in the standard that are obviously missing sentence components, the missing components in the standard clauses shall be supplemented manually.

[0038] ③ When the set operations AND, OR, and NOT exist in the specification, set operations such as intersection, union, and complement are described in Protégé using AND, OR, and NOT.

[0039] ④ When there are implicit quantifier constraints and quantity constraints in some normative clauses, the existential quantifier "some" is used instead of the universal quantifier "only". The normative name and normative number are in one-to-one correspondence. This is a kind of quantity constraint, which is constrained by the form "Exactly 1".

[0040] Furthermore, the specific operational procedures for step four are as follows:

[0041] The top-down approach involves first performing domain feature analysis, reusing existing building ontology within its framework, constructing the schema layer of the knowledge graph using an ontology editor or other methods, and finally defining hierarchical, attribute, and semantic relationships to complete the construction of the green building operation and maintenance knowledge ontology. The bottom-up approach targets multi-source, multi-modal data in the green building operation and maintenance domain. It involves acquiring raw data, extracting entities and relationships using knowledge extraction algorithms, and supplementing and optimizing the schema layer with normative knowledge. Finally, it aligns, merges, and disambiguates normative knowledge from different sources, storing and applying the extracted normative knowledge according to the schema layer framework, forming a mapping from the schema layer to the data layer, and constructing a complete green building operation and maintenance knowledge graph.

[0042] The beneficial effects of this invention are:

[0043] Currently, inventions related to knowledge graph construction in the field of building operations and maintenance are still in their early stages, and many problems remain to be solved. Therefore, this invention proposes a complete and replicable construction method for building operations and maintenance knowledge graphs based on industry standards and existing ontologies.

[0044] Compared with the prior art, the present invention has the following advantages:

[0045] (1) This invention constructs a building operation and maintenance knowledge graph based on industry standards. Industry standards highly condense the knowledge of domain experts and can provide guidance on operation and maintenance issues throughout the entire cycle. Knowledge graph technology can extract knowledge from industry standards to achieve the integration and reasoning of knowledge in the field of building operation and maintenance.

[0046] (2) This invention improves existing building ontology based on normative knowledge. Existing building ontology lacks knowledge related to green building and building operation and maintenance, while the vast amount of knowledge contained in industry norms can effectively supplement the knowledge system of building ontology. At the same time, the former's knowledge system is conducive to sorting out the cumbersome and complex normative knowledge, and ontology technology can also convert normative knowledge into a computer-recognizable form.

[0047] (3) This invention proposes knowledge expressions and a method for translating a large number of difficult-to-structured normative clauses in building operation and maintenance specifications, supplements information that is often overlooked such as specification numbers and modal words, and constructs a building operation and maintenance specification ontology with high completeness.

[0048] (4) In the process of constructing the knowledge graph pattern layer, this invention makes full use of existing building ontology and building operation and maintenance specifications knowledge, thereby improving the construction efficiency and scientific nature of the domain knowledge graph from the perspective of construction method.

[0049] (5) The knowledge graph was constructed using the Protégé, WebVOWL, ChatGPT, and Neo4j platforms, which effectively built and presented a green building operation and maintenance knowledge graph, laying the foundation for intelligent operation and maintenance of green buildings. This invention constructs a green building operation and maintenance knowledge graph based on industry standards, which is of great significance for giving full play to the role of standard knowledge and improving the level of green building operation and maintenance management. Attached Figure Description

[0050] Figure 1 This is a flowchart of a method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontology, according to the present invention.

[0051] Figure 2 A diagram illustrating the classification of clauses to standardize their use.

[0052] Figure 3 The process of constructing a knowledge ontology for the operation and maintenance of green buildings.

[0053] Figure 4 A visual example with prerequisites.

[0054] Figure 5 This is a translation method for set operations.

[0055] Figure 6 A visual example of set operations and quantifier constraints.

[0056] Figure 7 The process of building a knowledge graph for green building operation and maintenance.

[0057] Figure 8 The query results are for action constraints.

[0058] Figure 9 This is the result of a related path query.

[0059] Figure 10 To standardize content search results. Detailed Implementation

[0060] The present invention will be further described in detail below with reference to the accompanying drawings.

[0061] See Figure 1 This invention provides a method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies. Before constructing the green building operation and maintenance knowledge graph, a green building operation and maintenance knowledge ontology must first be constructed. Before constructing the ontology, two fundamental tasks need to be completed: reusing existing building ontologies and designing the standard knowledge structure. Therefore, this invention mainly includes the following steps:

[0062] Step 1: Reuse the existing building body and obtain the IfcOWL body;

[0063] This invention integrates a building operations and maintenance specification ontology with the IfcOWL ontology. Therefore, when adding new nodes using specification knowledge, the reuse of existing nodes must be considered first. Node reuse refers to reusing existing nodes within the ontology project to accelerate the ontology development process, improve ontology quality, and promote knowledge sharing and interoperability. Node reuse includes two scenarios: first, the new node has the same meaning as an existing node, directly using existing nodes in IfcOWL, with some examples shown in Table 1; second, the new node is a subclass of an existing node, requiring extension of the existing node, with some examples shown in Table 2.

[0064] Table 1 Examples of Synonyms

[0065] Add a node Synonyms Add a node Synonyms sensor IfcSensor architecture IfcBuilding heater IfcSpaceHeater system IfcSystem humidifier IfcHumidifier area IfcSpatialZone transformer IfcTransformer Building load IfcStructuralLoad Actuator IfcActuator personnel IfcActor

[0066] Table 2 Examples of Extension Cases

[0067]

[0068] However, manually reusing IfcOWL nodes requires understanding the meaning and inheritance relationships of thousands of nodes. Therefore, this invention uses ChatGPT to assist in node reuse. In IfcOWL, classes related to solar thermal systems may include the following: 1. IfcSolarDevice: Represents solar energy equipment, used to describe components of a solar system such as solar collectors and solar panels. 2. IfcSolarCollector: Represents a solar collector, used to collect solar energy and convert it into heat energy. It can describe the type, size, material, and other attributes of the collector. 3. IfcSolarPanel: Represents a solar panel, used to convert solar energy into electrical energy. It can describe the type, power, efficiency, and other characteristics of the panel. 4. IfcSolarHeatingDevice: Represents a solar heating device, used to utilize solar energy for heating. It can describe the type, working principle, thermal efficiency, and other attributes of the heating device. These classes can be used to describe the components and characteristics of a solar thermal system. By defining object attributes and attribute values, the parameters, performance, and operation of the solar thermal system can be further described in detail. Please note that specific classes and attributes may vary depending on different data models and application requirements. Through simple training, ChatGPT can generally narrow down the reusability to within 4-5 nodes, and further consulting the definition of buildingSMART can determine the final number of nodes.

[0069] Step 2: Standardize the knowledge structure design and obtain the building operation and maintenance specification ontology;

[0070] Step 2.1 Normative analysis and classification of normative clauses;

[0071] Analysis of the characteristics of the "Technical Specification for Operation and Maintenance of Green Buildings" (JGJ / T391-2016) reveals that the provisions of the building operation and maintenance specification (hereinafter referred to as the specification provisions) are complex in content and structure, making it impossible to establish a unified specification knowledge expression applicable to the entire text. Therefore, this invention classifies the specification provisions from three dimensions: specification translation, constraint content, and constraint form.

[0072] like Figure 2 As shown, firstly, from the perspective of standard translation, standard provisions can be divided into easily structurable and difficult-to-structurable types. Structuring involves converting standard provisions expressed in natural language into computer-readable statements. Knowledge in knowledge graphs is usually represented using a (entity)-[relationship]-(entity) or (entity)-{attribute:attribute value} structure. Therefore, if a standard provision can be represented by a triple structure of a knowledge graph after decomposition, it is called easily structurable. Conversely, it is called difficult-to-structurable. For example, the terminology section of Chapter 2 of the standard provision defines terms in the field of green building operation and maintenance, but it lacks a complete and recognizable statement structure, making it difficult to represent using a triple structure.

[0073] Secondly, since most of the content of the normative provisions addresses constraints on activities in the construction field, they can be categorized into six types based on the content of the constraints: attribute constraints, relational constraints, normative constraints, action constraints, state constraints, and non-constraints. Table 3 provides examples of these six types of normative provisions.

[0074] Table 3 Examples of Six Categories of Regulatory Clauses and Constraints

[0075]

[0076] Among them, attribute constraints refer to the inclusion of "operation and maintenance attributes" in the code clauses with explicit attribute value restrictions. For example, code clause 6.1.5 explicitly stipulates that "the equipment integrity rate of green building equipment systems should not be less than 98%." Relational constraints describe certain relational constraints between "operation and maintenance objects," between "operation and maintenance attributes," or between "operation and maintenance objects" and "operation and maintenance attributes." These relationships may include hierarchical relationships, temporal relationships, spatial relationships, or semantic associations such as inclusion relationships. Code clause 3.0.1 expresses a constraint of inclusion relationship: "The operation and maintenance of green buildings should include comprehensive performance adjustment, delivery, operation and maintenance, and operation and maintenance management." Generally, attribute constraints and relational constraints can be directly represented in the form of triples. For example, code clause 6.1.5 can be decomposed into two triples:

[0077] (Green building equipment systems) - [Attributes] - (Equipment availability)

[0078] (Equipment availability) - {Limited value: 98%}

[0079] The descriptions of the constraints in other normative clauses are not clear or direct enough. Compared with the aforementioned attribute constraints and relational constraints, they lack clear implementation plans and require further summarization, organization, and analysis by domain experts so that computers can understand the knowledge contained in the normative clauses. Among them, normative constraints refer to mutual references between normative clauses. These normative clauses usually do not provide specific normative descriptions but are expressed in the form of "shall comply with XX norms XX regulations". For example, normative clause 3.0.3 clearly stipulates that "the actual measurement and evaluation of energy efficiency of green buildings shall comply with the relevant provisions of the current industry standard 'Technical Standard for Building Energy Efficiency Labeling' JGJ / T288-2012". State constraints are mainly based on the ontology-based knowledge modeling of building engineering quality hazard information proposed by Zhong Xueyan et al. (Zhong Xueyan, Shen Luoxin, Pan Xing, et al. Knowledge modeling of building engineering quality hazard information based on ontology [J]. Information Technology of Civil Engineering: 1-12.). The concept of "operation and maintenance status" is introduced, mainly to deal with semantically ambiguous adjectives in the code clauses. For example, code clause 5.3.5 clearly stipulates that "the water metering device should be in good working order and the data records should be complete." "In good working order" and "complete" are regarded as constraints on the state of "water metering device" and "water metering device". In addition, there is a type of code clause that is not a constraint on the relationship between "operation and maintenance object" and "operation and maintenance attribute", but a specific measure for a certain object or attribute. For example, code clause 6.2.6 clearly stipulates that "wind energy recovery system should be inspected and cleaned regularly". With the help of the definition of IfcActionRequest in IfcOWL, the concept of "operation and maintenance action" is introduced to summarize this type of code clause. In the "Technical Specification for Operation and Maintenance of Green Buildings", there are a few clauses that do not contain binding meanings. For example, clause 1.0.2 of the specification clearly stipulates that "this specification applies to the operation and maintenance of newly built, expanded and renovated green buildings". Such clauses are usually found in the general provisions and terminology chapters of the "Technical Specification for Operation and Maintenance of Green Buildings".

[0080] Furthermore, normative clauses with explicitly defined values ​​are generally referred to as formalized normative clauses. For example, attribute constraints are data attribute constraints, which determine whether the actual parameters of a construction project conform to the specified range. These types of normative clauses can be considered essentially equivalent to attribute constraint types, and this constraint form needs to be expressed through data attributes in the OWL language. Other normative clauses without explicitly defined values ​​are called informal or semi-formal clauses, namely project attribute constraints, such as normative constraints, state constraints, action constraints, and non-constraints. These require object attributes to express the computer-understandable meaning of the constraints.

[0081] Step 2.2 Standardize word segmentation and entity recognition;

[0082] Before semantically representing normative knowledge, the normative corpus needs to be segmented. In fact, every normative statement in natural language can be divided into combinations of multiple word sequences. First, the normative text is processed, breaking down complex long sentences into shorter ones, with the ultimate goal of representing knowledge in the form of triples. Then, these shorter sentences are further broken down, typically into individual words, to better analyze the characteristics of the sentences; this process is called normative word segmentation.

[0083] After standardized word segmentation, entity recognition is required for various types of words. The purpose is to extract abstract semantic elements to summarize and organize a certain type of words. Entity recognition can help understand important information and relationships in the text, providing a foundation for subsequent tasks such as information extraction and knowledge graph construction. As shown in Table 4, the following categories can be identified: maintenance object, maintenance attribute, maintenance action, maintenance status, precondition, modal words, quantity comparison words, relational words, attribute values, standard name, standard number, and clause number. Among them, modal words can be divided into four categories according to their strictness: strictly prohibited / must, should not / must not / should, inappropriate / appropriate, and impossible / permissible, corresponding to Must, Should, May, and Could in English, respectively. Taking standard clause 5.4.2 as an example, in the clause "The three-phase load imbalance of the power distribution system should not exceed 15%", "power distribution system" belongs to the maintenance object, "three-phase load imbalance" belongs to the maintenance attribute, "should not" is a modal word, "greater than" is a quantity comparison word, and "15%" is an attribute value. In addition, Table 4 shows what entity types or entity descriptions can be used to represent various semantic elements in the OWL language, which lays the foundation for subsequent specification translation.

[0084] Table 4 Entity Recognition of Semantic Elements

[0085] semantic elements Partial Examples Entity representation quantity Operation and maintenance objects Refrigeration equipment units and building re-commissioning plan kind 268 Operation and maintenance attributes Indoor particulate matter concentration, building load characteristics kind 140 Operations and maintenance actions Timely maintenance and regular inspection kind 49 Operation and maintenance status Complete and comprehensive kind 16 Prerequisites During renovation and before operation Subclass of the corresponding object or property 66 Standard Name Green building evaluation standards kind 15 Standard Number GB / T50378 kind 12 modal words Must, not advisable Object property / data property naming prefix 153 Relationship words Increase, adjust according to... object properties 126 Article Number 5.4.2 Object property / data property naming suffix 124 Quantity comparison words Greater than, not equal to Data attributes constrain classes 8 Attribute value 15％、0.93～0.98 Data attributes constrain classes 9 set operators And, or, except... Class description 10

[0086] Step 2.3 Standardize the construction of knowledge expressions;

[0087] After completing the standard word segmentation and entity recognition, each standard clause is broken down into a set of multiple semantic elements. In fact, if we consider semantic elements as abstract concepts of a certain type of semantic information, then the standard knowledge expression composed of multiple semantic elements is an abstract expression of a certain semantic information. By constructing standard knowledge expressions, the requirements and regulations in the "Technical Specification for Operation and Maintenance of Green Buildings" can be expressed in a structured form, making the knowledge in the standard clauses easier for computers to understand. Taking attribute constraint types as an example, the standard word segmentation and entity recognition of the clause "The power factor of the low-voltage side power system should be 0.93 to 0.98" are as follows:

[0088] Standardized word segmentation: low-voltage side + power system + power factor + suitable + for +0.93~0.98;

[0089] Entity recognition: Prerequisites + Operation and maintenance object + Operation and maintenance attributes + Modal words + Quantity comparison words + Attribute values;

[0090] The second row contains the canonical knowledge expression for the attribute constraint type. Since the canonical knowledge expression represents the richest combination of semantic elements for this constraint type, not all canonical provisions fully conform to it. However, canonical provisions of the same constraint type can use the same canonical knowledge expression, and the absence of a certain semantic element does not affect subsequent canonical translation. Table 5 lists the state constraint knowledge expressions and some examples.

[0091] Table 5. State constraint knowledge expressions and some examples.

[0092]

[0093]

[0094] Due to their small number and simple form, unconstrained type specification provisions will not be designed with separate specification knowledge expressions. Furthermore, set operators appear infrequently; if they do appear, ∩ (intersection), ∪ (union), and [other operators] can be used. (Supplement) Three symbols are used to represent the specifications. The "Technical Specification for Operation and Maintenance of Green Buildings" includes 7 attribute constraint provisions, 72 relationship constraint provisions, 12 normative constraint provisions, 9 state constraint provisions, and 30 action constraint provisions. It should be noted that a single provision may cover multiple constraint types. For example, provision 5.2.2 states that "In areas with centralized air conditioning and high population density, the fresh air volume during operation should be adjusted according to the actual indoor population demand and should comply with the relevant provisions of the current national standard 'Code for Design of Heating, Ventilation and Air Conditioning of Civil Buildings' GB 50736".

[0095] Existing building ontology reuse and normative knowledge structure design form the research foundation for constructing a green building operation and maintenance knowledge ontology. Based on ChatGPT, 58 nodes of the IfcOWL ontology were reused, including both synonym and extended cases. Through building operation and maintenance norm analysis, the normative clauses of the "Technical Specification for Operation and Maintenance of Green Buildings" were classified mainly according to their constraint content. Normative word segmentation and entity recognition were performed on 124 normative clauses, summarizing 13 semantic elements and 5 normative knowledge expressions, thus realizing the structured expression of normative semantic knowledge.

[0096] Step 3: Construction of a green building operation and maintenance knowledge ontology;

[0097] The "Technical Specifications for Operation and Maintenance of Green Buildings" is presented in natural language; however, this method only helps humans understand the specifications. To enable computers to understand the knowledge contained in the specifications, the text needs to be structurally transformed. For example... Figure 3 As shown, steps one and two are based on the "Technical Specification for Operation and Maintenance of Green Buildings" (…). Figure 3 The system performs standard analysis to classify the original standard (5.2.6 The frequency of frequency converters should not be lower than 30Hz). Furthermore, through standard word segmentation (5.2.6 / frequency converters / of / frequency / should not / lower / 30Hz) and entity recognition (number / object / attribute / modal word / comparison value / attribute value), various standard knowledge expressions (number + condition + object + attribute + modal word + comparison value + attribute value) are designed, laying the foundation for computers to understand the knowledge descriptions contained in the standard clauses.

[0098] The following steps will structure the specification knowledge expression based on the entity representation method in Table 6 of the Protégé platform, realize specification translation based on the OWL language, and further establish a building operation and maintenance specification ontology. Finally, the building operation and maintenance specification ontology will be integrated with the IfcOWL ontology to construct a green building operation and maintenance knowledge ontology. The specific operation process is as follows:

[0099] Step 3.1 IfcOWL ontology extension;

[0100] The IFC standard includes three extension mechanisms: extension of new entities, extension of IfcProxy entities, and extension of custom attribute sets. However, there is a contradiction between the extension cost and the significance of the extension, as shown in Table 6. Furthermore, regardless of the extension mechanism, it is necessary to extend the IFC standard based on the EXPRESS language and then establish an EXPRESS-OWL mapping framework to obtain the extended IfcOWL ontology. However, the IFC model is not easy for operators unfamiliar with the EXPRESS language to extend.

[0101] Table 6 Comparison of IFC Extension Mechanisms

[0102] Comparison Projects IfcProxy Entity Extension Add entity extension Attribute Set Extension Operability easy Disaster relatively easy compatibility excellent Difference good Operating efficiency Low high generally Extended meaning Small big Larger

[0103] The field of green building operation and maintenance encompasses all aspects of buildings, thus necessitating the addition of specific entities. Therefore, this invention extends IfcOWL directly using the OWL language on the Protégé platform. Extending IfcOWL with OWL can meet the needs of specific domains, allowing for the customization of concepts, attributes, and relationships based on specific domain requirements, enabling the ontology to better adapt to different application scenarios and needs. Furthermore, the ontology extended with OWL has good understandability, making subsequent maintenance and evolution easier. If new requirements or changes arise, the extension can be modified and adjusted relatively easily without requiring large-scale changes to the entire system. In addition, OWL has rich reasoning and query capabilities, helping developers quickly verify the correctness and consistency of the extension definitions. More importantly, OWL provides abundant tools and resources to accelerate the development process of extensions, reducing the workload and development cycle for developers.

[0104] Step 3.2 Utilize the Protégé platform and WebVOWL visualization tool to perform specification translation and achieve ontology modeling, establishing a building operation and maintenance specification ontology that can be understood by computers;

[0105] ① "Preconditions" are defined semantically as subclasses of a certain object or property. For example... Figure 4 As shown, for example, Article 6.1.4 of the standard clearly stipulates that "when repairing, renovating or transforming, (building products) should give priority to locally produced building materials." The "when repairing, renovating or transforming" is not directly defined as a class as a "prerequisite". Instead, it is a parent class of "building products" with three subclasses: "building products when repairing", "building products when renovating" and "building products when transforming".

[0106] ② The standard contains clauses with obvious missing sentence components. Examples include clause 6.3.1, which states "A greening management system should be formulated and publicized, and strictly implemented"; clause 6.1.4, which states "When repairing, renovating, or altering, locally produced building materials should be given priority"; and clause 6.2.11, which states "When replacing sanitary fixtures, those with lower water efficiency ratings should not be used." To address these clauses, this invention references the "Implementation Guidelines for Green Building Operation and Maintenance Technical Specifications" and the "Explanation of the Clauses of Green Building Operation and Maintenance Technical Specifications," and manually supplements the missing components of the standard clauses.

[0107] ③ The standard also includes set operators "AND", "OR", and "NOT". For example, "In areas with centralized air conditioning and high population density, the fresh air volume should meet the current national standards," meaning that such areas must simultaneously meet the conditions of being both areas with centralized air conditioning and areas with high population density. Figure 5 and Figure 6As shown, set operations such as intersection, union, and complement are generally described in Protégé using and, or, and not.

[0108] ④ Some normative provisions contain implicit quantifier constraints and quantity constraints. For example... Figure 6 As shown, for example, "fresh air volume" is an attribute of "region," but "region" has more than one attribute. Therefore, the existential quantifier "some" is used instead of the universal quantifier "only." Furthermore, the specification name and specification number have a one-to-one correspondence, which is a quantitative constraint that can be constrained using the form "Exactly 1 (cardinality)."

[0109] Step 3.3 Integrate the building operation and maintenance specification ontology with the IfcOWL ontology to construct a green building operation and maintenance knowledge ontology.

[0110] Ontologies play a crucial role in defining and organizing knowledge in knowledge graphs. This study utilizes the Protégé modeling tool and the WebVOWL visualization tool to model the ontology. Based on the specification translation of various modeling techniques, a computer-understandable building operation and maintenance specification ontology was constructed. After merging with the IfcOWL ontology, a green building operation and maintenance knowledge ontology was obtained, comprising 1937 classes, 2043 object attributes, 25 data attributes, and 24624 axioms. The Neo4j graph database was then used to store the green building operation and maintenance knowledge ontology.

[0111] Step 4: Construction of a green building operation and maintenance knowledge graph;

[0112] Because the application and invention of knowledge graphs in the field of green building operation and maintenance are relatively limited, the involvement of domain experts is needed to construct the basic data architecture of the knowledge graph from a top-down perspective. However, the "Technical Specifications for Operation and Maintenance of Green Buildings" is a high-quality knowledge resource within the industry, and the specification's knowledge data should be fully collected and organized to gradually construct and discover the structure and relationships of the knowledge graph. Therefore, this invention, based on a combination of top-down and bottom-up construction methods, further proposes a method for constructing a building operation and maintenance knowledge graph based on industry specifications and existing building ontology.

[0113] like Figure 7As shown, the top-down approach first performs domain feature analysis, reuses existing building ontology within its framework, constructs the schema layer of the knowledge graph using an ontology editor or other methods, and finally defines hierarchical, attribute, and semantic relationships to complete the construction of the green building operation and maintenance knowledge ontology. By defining hierarchical, categorical, and semantic relationships, a clear and explicit conceptual hierarchy can be formed. The bottom-up approach targets multi-source, multi-modal data in the green building operation and maintenance field, such as professional literature, expert experience, building operation and maintenance databases, and network resources. It acquires raw data and uses appropriate knowledge extraction algorithms to extract entities and relationships, with a focus on supplementing and optimizing the schema layer using normative knowledge. Finally, it aligns, merges, and disambiguates normative knowledge from different sources, stores and applies the extracted normative knowledge according to the schema layer framework, forming a mapping from the schema layer to the data layer, and constructing a complete green building operation and maintenance knowledge graph.

[0114] Based on the "Technical Specifications for Operation and Maintenance of Green Buildings" and existing building ontology, this invention improves the domain knowledge graph construction method that combines top-down and bottom-up approaches. Utilizing the Neo4j graph database to store the green building operation and maintenance knowledge ontology, a preliminary green building operation and maintenance knowledge graph has been constructed and applied, laying the foundation for large-scale storage of multi-source heterogeneous data.

[0115] A green building operation and maintenance knowledge graph was obtained using a method for constructing a building operation and maintenance knowledge graph based on industry standards and existing building ontology, as described in this invention. The Cypher query function of the Neo4j graph database was used to implement three types of queries based on the green building operation and maintenance knowledge graph: constraint content, association paths, and standard content. Details are as follows:

[0116] 1. Constraint Content Query. The knowledge provided by the "Technical Specification for Operation and Maintenance of Green Buildings" mainly concerns constraints on a certain object or attribute. Therefore, this function allows operation and maintenance personnel to quickly query all constraints on this node in the existing specification. In addition, this function can further compare the constraint content of different specifications on the same node, which is helpful in discovering conflicts and overlaps in specification clauses. Figure 8 The content of the action constraints and the query results of mutually exclusive classes are presented.

[0117] 2. Association Path Query. The building operation and maintenance field involves a large number of building entities, which are interconnected, forming a complex relationship network. Therefore, querying the association paths between nodes lays the foundation for assisting operation and maintenance decision-making, allowing maintenance personnel to consider management measures for individual nodes from multiple perspectives. Taking power distribution systems and three-phase load imbalance as an example, through... Figure 9 This indicates that the path between the two nodes represents a membership relationship.

[0118] 3. Standard Content Query. This function not only meets users' needs for consulting standards, but also allows them to further query which standard clause the constraint content originates from, since object attributes and data attributes are suffixed with the clause number. Figure 10 Taking the "Technical Specification for Operation and Maintenance of Green Buildings" as an example, this paper presents the chapters and attributes of the "Technical Specification for Operation and Maintenance of Green Buildings". Further exploration of the relevant chapters will allow you to view the specific provisions of the specification.

[0119] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies, characterized in that: Includes the following steps: Step 1: Reuse the existing building body and obtain the IfcOWL body; Step 2: Through standard analysis, classify the standard provisions, perform word segmentation and entity recognition on the standard provisions, construct standard knowledge expressions, and obtain the building operation and maintenance standard ontology; Step 3: Structure the standard knowledge expressions, realize standard translation based on OWL language, and establish a building operation and maintenance standard ontology; integrate the building operation and maintenance standard ontology with the IfcOWL ontology to construct a green building operation and maintenance knowledge ontology. Step 4: Construct a green building operation and maintenance knowledge graph based on a combination of top-down and bottom-up approaches; The top-down approach involves first performing domain feature analysis, reusing existing building ontology within its framework, constructing the schema layer of the knowledge graph using an ontology editor, and finally defining hierarchical, attribute, and semantic relationships to complete the construction of the green building operation and maintenance knowledge ontology. The bottom-up approach targets multi-source, multi-modal data in the green building operation and maintenance domain. It acquires raw data, extracts entities and relationships using knowledge extraction algorithms, and supplements and optimizes the schema layer with normative knowledge. Finally, it aligns, merges, and disambiguates normative knowledge from different sources, stores and applies the extracted normative knowledge according to the schema layer framework, forming a mapping from the schema layer to the data layer, and constructing a complete green building operation and maintenance knowledge graph.

2. The method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies according to claim 1, characterized in that, In step one, the method for reusing the existing building structure is as follows: (1) When a new node has the same meaning as an existing node, the existing node in IfcOWL is used directly; (2) When the new node is a subclass of an existing node, the existing node needs to be extended.

3. The method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies according to claim 1, characterized in that, In step one, ChatGPT is used to assist in the reuse of the existing building structure.

4. The method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies according to claim 1, characterized in that, In step two, the specific operational procedure for classifying the regulatory provisions is as follows: (1) From the perspective of normative translation, normative clauses are divided into easy-to-structure and difficult-to-structure clauses; structuring is the process of converting normative clauses expressed in natural language into computer-readable statements; knowledge in knowledge graphs is represented by a structure of (entity)-[relationship]-(entity) or (entity)-{attribute:attribute value}. If a normative clause can be represented by a triple structure of a knowledge graph after being decomposed, it is called easy-to-structure clause; otherwise, it is called difficult-to-structure clause. (2) From the perspective of the content of constraints, normative clauses are divided into six types: attribute constraints, relational constraints, normative constraints, action constraints, state constraints and non-constraints; (3) From the perspective of constraint form, normative clauses with explicit values ​​are called formalizable normative clauses, and their meaning is expressed through data attributes in the OWL language; other normative clauses without explicit values ​​are called informal clauses or semi-formal clauses, and their meaning is expressed through object attributes.

5. The method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies according to claim 1, characterized in that, In step two, the specific operational procedures for word segmentation and entity recognition of the specification text are as follows: The standard clauses are processed by breaking down complex long sentences into short sentences, and then further breaking down these short sentences into individual words. After standard word segmentation, entity recognition is performed on various types of words, including: operation and maintenance objects, operation and maintenance attributes, operation and maintenance actions, operation and maintenance status, preconditions, modal words, quantity comparison words, relational words, attribute values, standard names, standard numbers, and clause numbers.

6. The method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies according to claim 1, characterized in that, In step two, after completing the standardized word segmentation and entity recognition, each standardized clause is split into a set of multiple semantic elements.

7. The method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies according to claim 1, characterized in that, In step three, the Neo4j graph database is used to store the green building operation and maintenance knowledge ontology.

8. The method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies according to claim 1, characterized in that, The specific operation process for step three is as follows: Step 3.1 involves directly extending IfcOWL on the Protégé platform using the OWL language; Step 3.2 Utilize the Protégé platform and WebVOWL visualization tool to perform specification translation and achieve ontology modeling, establishing the building operation and maintenance specification ontology; Step 3.3 Integrate the building operation and maintenance specification ontology with the IfcOWL ontology to construct a green building operation and maintenance knowledge ontology.

9. The method for constructing a building operation and maintenance knowledge graph based on industry standards and existing ontologies according to claim 8, characterized in that, The specific operation procedure for step 3.2 is as follows: ① The precondition is defined as a subclass of a certain object or property through semantic interpretation; ② When there are clauses in the standard that are obviously missing sentence components, the missing components in the standard clauses shall be supplemented manually. ③ When the set operation words AND, OR, and NOT exist in the specification, set operations such as intersection, union, and complement are described in Protégé using AND, OR, and NOT. ④ When there are implicit quantifier constraints and quantity constraints in some normative clauses, the existential quantifier "some" is used instead of the universal quantifier "only". The normative name and normative number are in one-to-one correspondence. This is a kind of quantity constraint, which is constrained by the form "Exactly 1".

Images