A real-time comparison method for construction engineering quality data

Abstract

The present invention provides a real-time comparison method for construction engineering quality data, which belongs to the field of building construction technology. The method forms a domain ontology with ontology as the expression form from standards, specifications, architectural design requirements and some typical cases, and then expands the domain ontology to actual application scenarios to form a knowledge application framework, establishes a multi-level data network based on ontology, and forms an entity node network oriented to application scenarios. On this basis, an activation diffusion method is used to form an activation area, and a set of entity nodes closely associated with stimulus input data is identified from a large data set, and then a type and similarity-based matching is performed from the set to find the target entity, thereby realizing real-time comparison of input stimulus data and target entity, and outputting the comparison result. The present invention avoids the probability of finding irrelevant but conceptually similar data comparison items by performing direct text query in a large number of construction engineering data sets, and improves the accuracy and efficiency of data comparison.

Description

Technical Field

[0001] The invention belongs to the technical field of building construction, and in particular relates to a real-time comparison method for building engineering quality data. Background Art

[0002] The construction project involves data from all aspects during the design, construction, acceptance, operation and maintenance process. These data come from the construction site, including material supply, construction process, completion acceptance, data archiving, and multiple dimensions of "personnel-machinery-materials-methods-environment" (abbreviated as "personnel-machine-materials-methods-environment"). The collection of these data not only reflects the current status of the construction project itself, but also reflects the information on the personnel, equipment, materials, methods and environment currently operating the construction project. These data often need to be compared with standards, architectural design requirements or some related cases related to the construction project to clarify the differences with related measurement requirements.

[0003] The existing comparison is still mainly done manually. After obtaining a piece of data, experts or technicians find the corresponding clauses, requirements or refer to relevant cases based on their understanding of the standard for comparison. With the development of artificial intelligence and big data, there are also some methods that use computer methods for comparison, but they often require experts to set the comparison rules in advance, such as entering a numerical value, then finding a specific field from a specified database, and then comparing the numerical value with the content obtained from the field, or just performing text-based matching, which has low search efficiency and low accuracy. At the same time, such systems often have poor flexibility in the actual application stage. Once data elements or standards and specifications are updated, a lot of content needs to be adjusted, which is prone to errors and cannot automatically compare related cases. Summary of the invention

[0004] To this end, the present invention provides a real-time comparison method for construction project quality data in response to the existing technical problems. This method forms a domain ontology in the form of ontology based on standards, specifications, architectural design requirements and some typical cases, and establishes a multi-level network structure based on ontology. Using the activation diffusion method, after a node in the entity network is activated by inputting stimulus data, the activation state will spread along the network to form an activation area. This area is a set of entity nodes that are closely associated with the stimulus input data identified from a large data set. In this way, while inputting the stimulus data, a subset associated with the input stimulus data is formed from the huge construction project quality data set, and then a type and similarity-based matching is performed from the subset to find the target entity, thereby realizing the comparison between the input stimulus data and the target entity, and outputting the comparison result in real time. The specific scheme is as follows:

[0005] A real-time comparison method for construction project quality data collects standard or specification data of construction projects, structures and processes the data in a knowledge-based manner to construct an ontology layer; the ontology layer defines the meaning and attributes of concepts based on the hierarchical structure and logical relationship of the concepts, wherein each concept forms a named node in the ontology layer network, and the graphical expression is generally represented by an ellipse, and its attributes include object attributes and data attributes, and the attributes are generally expressed as directed arrows in the graph, and the node and its attributes constitute a named entity in the ontology layer, and each named entity has a unique number that is distinguished from other named entities in the entire data set.

[0006] On the basis of the ontology layer, the concepts and attributes of named entities are extended to construct an application layer; the application layer supplements the parameter acquisition steps and calculation process in the standard or specification data, and establishes constraints on the conclusions or numerical descriptions in the standard or specification data; the application layer can also structure some typical cases into instance nodes to provide reference for subsequent similar applications.

[0007] Based on the application layer, with the node and attribute network as the basic framework, the correlation of the concepts represented by each node is calculated to indicate the connection strength between each entity, and the connection layer is constructed with the connection strength as the edge.

[0008] On the basis of the connection layer, concepts or data related to the quality of construction projects are input as incentive data to activate the corresponding nodes in the connection layer network and diffuse to the affected nodes. These nodes and edges become the diffusion layer. After an activation diffusion event is completed, the instance nodes formed by this activation can be merged into the application layer to become instance nodes for subsequent reference.

[0009] Based on the multi-layer data organization structure constructed above, real-time comparison of construction project quality data is performed. Within the activation range of the diffusion layer, the matching target node is searched for the node formed by the core vocabulary in the input stimulus data; filtering is performed based on the type of the core vocabulary to form a filter node set; the similarity between the concept corresponding to the core vocabulary and each concept in the filter node set is calculated; if the core vocabulary is text, the result with the highest semantic similarity is selected as the comparison result output; if the core vocabulary is a numerical or time type, the closest result and size relationship are output.

[0010] The multi-layer data organization structure can be used as a single layer or as a composite layer with multiple layers stacked together.

[0011] Specifically, the knowledge processing includes knowledge modeling, named entity recognition, and relationship recognition and extraction.

[0012] Specifically, the attributes include object attributes and data attributes. The object attributes are defined as the association between two nodes; the data attributes define the descriptive or numerical characteristics of the concept; the domain and value range of the attribute form a constraint relationship on the concepts, numerical values ​​and texts involved; the node has data attributes to indicate the specific content and meaning of the concept represented by the node, and has object attributes pointing to other nodes related to it.

[0013] Specifically, the connection strength of the connection layer can be represented by the correlation between node concepts sim_use, which consists of two parts: the semantic similarity relationship sim_base between two concepts in a general context and the correlation sim_field between concepts in the field of construction engineering.

[0014] The semantic similarity between two concepts in a general context is sim_base. According to the semantic similarity formula, the semantic similarity formula between concepts A and B is as follows:

[0015] (1)

[0016] Among them, f(A) is the number of web pages in the search engine library that contain concept A, f(B) is the number of web pages in the search engine library that contain concept B, f(A,B) is the number of web pages in the search engine library that contain both A and B, and N is the total number of web pages on the Internet referenced by the search engine library.

[0017] The correlation between concepts in the field of construction engineering is sim_field, which collects and organizes technical documents, test reports, and operation and maintenance instructions for each link of construction engineering design, construction, acceptance, and operation and maintenance to form a data set and semantic vector for construction engineering applications. In this set, the correlation sim_field between concepts of each node is calculated through the semantic vector of the concept. Sim_field is calculated using the cosine similarity method. The specific formula is as follows:

[0018] (2)

[0019] Among them, A and B represent the semantic vectors of two concept texts. and They represent the magnitude of the vectors respectively.

[0020] The correlation between the concepts represented by the two nodes, sim_use, is formed by combining sim_base and sim_field according to the weights:

[0021] (3)

[0022] Among them, a and b are hyperparameters and must satisfy 0≤a+b≤1, a≥0, b≥0.

[0023] Specifically, during the construction of the diffusion layer, the degree of activation of any node in the connection layer is the activation value act_total. When the node activation value is lower than the set value act_min, the node is not activated. All activated nodes with activation values ​​greater than or equal to act_min constitute the diffusion layer by reference or generated instances.

[0024] The activation value act_total_i of a node i consists of two parts: ① the basic activation value, which represents the impact of the initial input of the incentive concept on node i through multi-level diffusion; ② the impact of the activated nodes around node i on node i. The calculation formula of act_total_i is as follows:

[0025] The calculation formula of the node base activation value act_base_i is:

[0026] (4)

[0027] Among them, act_start is the previous stage excitation intensity, is a hyperparameter, which represents the loss in the transmission process.

[0028] On this basis, several nodes around entity node i are represented by set j, among which there are m activated nodes, k is any activated node in j, and the calculation formula of act_total_i is:

[0029] (5)

[0030] in, The initial value is set to 1 / m, where m is the number of active nodes in the j set. is the correlation between the concept represented by node k and the concept represented by node i.

[0031] When inputting stimulus data, the node where the stimulus data is located is taken as the center, and the scope of influence of the stimulus from near to far is calculated, and a data diffusion zone is formed along the direction of the attribute. When the diffusion to the node activation value is lower than the set value act_min, the diffusion stops. At this time, the diffusion zone contains all the nodes affected by the stimulus, and the target nodes that can match the stimulus data are searched within the node range of the diffusion zone for comparison.

[0032] Specifically, in the real-time comparison of construction project quality data, construction project quality data or text and its related context are input to form a context data set; the related text in the context data is segmented to form an input vocabulary set, and the vocabulary set includes a core vocabulary and several related vocabulary. The core vocabulary can be manually specified, or the subject words can be extracted from the input vocabulary set as the core vocabulary by artificial intelligence; the vocabulary in the vocabulary set is divided into concepts and instances, and the concept mainly reflects an abstract expression of a class of physical objects and behavior descriptions, and the node containing the concept can be directly referenced from the connection layer to the diffusion layer. The instance is the specific data for this application, and is instantiated in the diffusion layer according to a node in the connection layer to form an instance node.

[0033] Furthermore, the hyperparameters need to be set before starting the activation process; if the target node for comparison is not within the region of the diffusion layer, the hyperparameters are adjusted.

[0034] Furthermore, if during comparison, the node concept formed by the core word is consistent with the concept represented by the target node, but one of the nodes does not have a corresponding attribute value, the ontology layer knowledge structure is adjusted to supplement the node attributes.

[0035] Furthermore, in the process of constructing a multi-layer data organization, while inputting the construction project quality data, a comparison result is formed manually to establish a data training set for training the connection layer network. .

[0036] Beneficial effects:

[0037] The present invention proposes a real-time comparison method for construction engineering quality data. The method forms a domain ontology with ontology as the expression form by standards, specifications, architectural design requirements and some typical cases, and then expands the domain ontology to actual application scenarios to form a knowledge application framework. A multi-level data network based on ontology is established to form a named entity node network oriented to application scenarios. On this basis, an activation diffusion method is adopted. When a node in the entity network is activated by inputting stimulus data, the activation state will be diffused along the network to form an activation area. This area is a named entity node set closely associated with the stimulus input data identified from a large data set, so that the problem of directly performing text queries in a large number of construction engineering data sets to find conceptually similar but unrelated data comparison items is avoided to a certain extent. In this way, a subset associated with the input stimulus data is formed from a huge construction engineering quality data set while inputting the stimulus data, and then a target entity is found from the subset by matching based on type and similarity, thereby realizing real-time comparison of the input stimulus data and the target entity, and outputting the comparison result. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings:

[0039] Figure 1 The structure of the main body layer in the embodiment of the present invention;

[0040] Figure 2 is a structure in the application layer of an embodiment of the present invention;

[0041] Figure 3 The structure of the connection layer in the embodiment of the present invention;

[0042] Figure 4 The structure of the diffusion layer in the embodiment of the present invention;

[0043] Figure 5 This is the data hierarchical organization structure of the embodiment of the present invention. DETAILED DESCRIPTION

[0044] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. The principles and features of the present invention are described below in combination with the accompanying drawings. It should be noted that the embodiments in this application and the features in the embodiments can be combined with each other without conflict. The embodiments are only used to explain the present invention and are not used to limit the scope of the present invention.

[0045] A real-time comparison method for construction engineering quality data collects standards or specifications involved in each link of construction engineering design, construction, completion and operation and maintenance, and structures and processes the standards or specifications one by one to construct an ontology layer; the ontology layer defines the meaning of concepts and the attributes they should have by establishing a hierarchical structure and logical relationship of concepts, wherein each basic concept forms a uniquely named node in the ontology layer network, and its attributes include object attributes and data attributes, and the node and its attributes constitute a named entity in the ontology layer. In practical applications, the relevant national, industry, local, and group standards or specifications in the process of construction engineering design, construction, completion acceptance, and operation and maintenance are first collected and sorted. The above standards and specifications are structured one by one, and then processed in a knowledgeable way, so that each part forms an independent knowledge system. Among these standards or specifications, there will be some identical or similar meanings, or upper and lower conceptual relationships, or some meaningful associations. By aligning the same or similar entities in these standards and associating related entities through object attributes, these domain ontologies that were originally formed by different standards and are formally independent of each other are essentially associated in connotation to form an ontology knowledge network. Each named entity has a unique identifier in the knowledge network.

[0046] On the basis of the ontology layer, the concepts and attributes of the entity are further extended and constructed for the application process to obtain the application layer; the application layer supplements the parameter acquisition steps and calculation process in the standards or specifications, and establishes constraints on the conclusions or numerical descriptions in the standards or specifications, so that the entity can establish a direct association with the actual data entered on site. In the ontology layer, the basic concept hierarchy and association relationship of the entity have been clarified, but some nodes involved in the comparison will not be directly obtained in actual applications. The formation of these nodes requires some basic data or original data, and also requires some specific methods or steps to process and form. Basic data or original data often come directly from the construction site, and specific methods or steps are generally described in the specifications. The application layer realizes the transition from basic data or original data to nodes and some specific calculation methods. At the same time, some typical cases are structured and processed into instance nodes to provide reference for subsequent similar applications.

[0047] Based on the application layer, with the node and attribute network as the basic framework, the correlation of the concepts represented by each node is calculated to indicate the connection strength between entities, and the connection layer is constructed with the connection strength as the edge.

[0048] On the basis of the connection layer, concepts or data related to the quality of construction projects are input as incentive data to activate the corresponding nodes in the connection layer network and diffuse to the affected nodes. These nodes and edges form a diffusion layer. After an activation event is completed, the instance node formed by this activation can be merged into the application layer to become an instance node for subsequent reference.

[0049] Based on the multi-layer data organization structure constructed above, real-time comparison of construction project quality data is performed. Within the activation range of the diffusion layer, the matching target node is searched for the node formed by the core vocabulary in the input stimulus data; filtering is performed based on the type of the core vocabulary, the main types include numerical values, texts, data sequences and time, to form a set of filtered nodes; the similarity between the concept corresponding to the core vocabulary and each concept in the set of filtered nodes is calculated; if the core vocabulary to be compared is text, the num results with the highest similarity are selected as the comparison result output; if it is a numerical value or time type, the num closest results and their size relationship are output.

[0050] In the real-time comparison of construction engineering quality data, it is often not just a data or a text that is input. In practical applications, it is often necessary to enter data or text and its related context to form a context data set; the relevant text in the context data is segmented to form an input vocabulary set, which includes a core vocabulary and several related vocabulary. The core vocabulary can be manually specified, obtained through some steps, or extracted from the input vocabulary set as the core vocabulary by artificial intelligence. The vocabulary in the vocabulary set is divided into concepts and instances. For concepts, the node containing the concept is directly referenced from the connection layer to the diffusion layer. For instances, it is first identified which concept is concretized, and based on a node in the connection layer, it is instantiated in the diffusion layer to form an instance node. Through the above four layers of data organization, the comparison scope is narrowed from the massive big data area to the limited diffusion layer range, and the diffusion layer not only has the excitation data and the target node and the nodes in the diffusion path process, but also establishes an association network between the excitation data and the target node.

[0051] Specifically, the knowledge processing includes knowledge modeling, named entity recognition, and relationship recognition and extraction. Specifically, the attributes include object attributes and data attributes. The object attributes establish associations between different concepts, and the data attributes define the descriptive or numerical characteristics of the concepts. The domain and range of the attributes form constraints on the concepts, numerical values, and texts involved. The nodes have data attributes that indicate the specific content and meaning of the concept represented by the node, and the nodes have object attributes that point to other nodes related to them.

[0052] Specifically, the connection strength of the connection layer can be represented by the correlation of node concepts sim_use, which consists of two parts: the semantic similarity relationship sim_base between two concepts in a general context and the correlation sim_field between concepts in the field of construction engineering.

[0053] The semantic similarity relationship between two node concepts in a general context is sim_base. According to the semantic similarity formula, the semantic similarity formula between concepts A and B is as follows:

[0054] (1)

[0055] Among them, f(A) is the number of web pages in the search engine library that contain concept A, f(B) is the number of web pages in the search engine library that contain concept B, f(A,B) is the number of web pages in the search engine library that contain both A and B, and N is the total number of web pages on the Internet referenced by the search engine library.

[0056] The similarity between concepts in the field of construction engineering is sim_field. The technical documents, test reports, and operation and maintenance instructions of each link of the construction engineering are collected and sorted to form a data set and semantic vector for construction engineering applications. In this set, the correlation sim_field between each concept is calculated through the semantic vector of the concept. Sim_field is calculated using the cosine similarity method. The specific formula is as follows:

[0057] (2)

[0058] Among them, A and B represent the semantic vectors of two concept texts. and They represent the magnitude of the vectors respectively.

[0059] The correlation between two concepts, sim_use, is formed by combining sim_base and sim_field according to the weights:

[0060] (3)

[0061] Among them, a and b are hyperparameters and must satisfy 0≤a+b≤1, a≥0, b≥0.

[0062] Specifically, during the construction of the diffusion layer, the degree of activation of any node in the connection layer is the activation value act_total. When the node activation value is lower than the set value act_min, the node is not activated. All activated nodes with activation values ​​greater than or equal to act_min constitute the diffusion layer by referencing or generating instances.

[0063] The activation value act_total_i of a node consists of two parts: ① the basic activation value, which represents the impact of the initial input of the incentive concept on node i through multi-level diffusion; ② the impact of the activated nodes around node i on node i. The calculation formula of act_total_i is as follows:

[0064] The calculation formula of the basic activation value act_base_i of node i is:

[0065] (4)

[0066] Among them, act_start is the previous stage excitation intensity, is a hyperparameter, which represents the loss in the transmission process.

[0067] On this basis, several nodes around entity node i are represented by set j, among which there are m activated nodes, k is any activated node in j, and the calculation formula of act_total_i is:

[0068] (5)

[0069] in, The initial value is set to 1 / m, where m is the number of active nodes in the j set. is the correlation between the concept represented by node k and the concept represented by node i.

[0070] When inputting stimulus data, the node where the stimulus data is located is taken as the center, and the scope of influence of the stimulus from near to far is calculated, and a data diffusion zone is formed along the direction of the attribute. When the diffusion to the node activation value is lower than the set value act_min, the diffusion stops. At this time, the diffusion zone contains all the nodes affected by the stimulus, and the target nodes that can match the stimulus data are searched in the node range of the diffusion zone for comparison.

[0071] The diffusion steps are as follows:

[0072] 1) Information is manually entered or transmitted by on-site devices. The information may be the name of the building structure, the number of the building structure component, the test data, the on-site personnel, the test equipment, etc.;

[0073] 2) Input data is associated or matched by default through the software interface, triggering a node node_center in the connection layer, referencing the node in the diffusion layer to form a node representing a concept, and forming an instance node of the node. If necessary, the on-site user is prompted to complete other attribute information in the instance node. This attribute information can also be regarded as incentive data and trigger the incentive diffusion process;

[0074] 3) Set the node node_center instance to be activated, and set act_base to the initial value, indicating full activation;

[0075] 4) Take the node node_center as the previous node, and search for other concepts associated with all object attributes of node_center to form the next node set node_L1[node_L1_1,node_L1_2,…,node_L1_n];

[0076] 5) Calculate the act_base of each node in the set node_L1 using formula (4), where act_start is the act_base of the previous node;

[0077] 6) With node_L1_1 as the previous node, repeat steps 4) to 5) to calculate the basic activation value of the subsequent node of node_L1_1;

[0078] 7) Adopt the breadth - first method, and calculate the act_base of each subsequent node along the direction of the attributes of each activated node object until the act_base of a first - level node is less than act_min or the pointed - to node has been activated, then stop the diffusion;

[0079] 8) The nodes with activation values greater than or equal to act_min form the activated node set node_activated_set1;

[0080] 9) Calculate the act_total of each node in node_activated_set1 one by one using formula (5);

[0081] 10) For nodes with act_total ≥ act_min, act_base < act_min and there are unactivated subsequent nodes, they can continue to diffuse to the subsequent level; otherwise, stop the diffusion;

[0082] 11) The newly diffused nodes calculate act_total according to formula (4) and formula (5), and repeat step 10) to form the set of all activated nodes node_activated_set2;

[0083] 12) Refer the nodes in node_activated_set2 to the diffusion layer to form the corresponding node network. The connection strength between the activated nodes can be obtained from the connection layer. At this time, the construction of the diffusion layer is completed.

[0084] Furthermore, the hyperparameters need to be set before starting the activation process; if the target node for comparison is not within the area of the diffusion layer, then adjust the hyperparameters.

[0085] Furthermore, if during the comparison, the node concepts formed by the core words are the same as the concepts represented by the target node, but one of them lacks the corresponding attribute value, then adjust the knowledge structure of the ontology layer.

[0086] Furthermore, during the process of constructing the multi - layer data organization, while inputting the building engineering quality data, form the comparison results manually to establish a data training set for training in the .

[0087] Take the example of using the rebound method to detect the compressive strength of concrete in a building and conduct data comparison to illustrate the process of data organization and comparison at each level:

[0088] 1. Ontology layer

[0089] In the process of construction project acceptance and inspection, many acceptance specifications and inspection specifications are involved, such as "Technical Standard for On-site Inspection of Concrete Structures" (GB / T 50784-2013), "Technical Specification for Testing Concrete Compressive Strength by Rebound Method" (JGJ / T23-2011), etc. Now, taking these two standards as examples, we will conduct structured processing respectively, extract the concepts and relationships between them through named entity recognition, relationship recognition and extraction, form corresponding nodes and attributes of connections between nodes, and form two independent domain ontology networks. Figure 1 The ontology layer nodes of the standard knowledge graph are represented by ellipses, using the namespace (ns:) custom namespace. The descriptive text of the concept is represented by a rectangle, and directed arrows represent attributes and their directions.

[0090] Using entity alignment technology, two ontology networks can be merged through nodes with the same or similar concepts. Figure 1 In the ontology layer of the standard knowledge graph, the independent ontology concept network formed by the two specifications of "Technical Standard for Field Testing of Concrete Structures" (GB / T 50784-2013) and "Technical Code for Testing Concrete Compressive Strength by Rebound Method" (JGJ / T23-2011) is integrated through the node of AreaReboundValue to form a mutually integrated professional domain ontology. These two domains of knowledge are related to the node formed by the concept of area rebound value. Appendix A of "Technical Standard for Field Testing of Concrete Structures" (GB / T 50784-2013) indicates that the rebound average value should be corrected according to the relevant provisions of the current industry standard "Technical Code for Testing Concrete Compressive Strength by Rebound Method" (JGJ / T 23-2011). Therefore, this concept is also a key concept in "Technical Code for Testing Concrete Compressive Strength by Rebound Method" (JGJ / T 23-2011), so a connection point of the two standard knowledge graphs is formed here.

[0091] 2. Application layer

[0092] Figure 2It is the structure in the application layer, which is the situation of the application layer formed by expanding on the basis of the domain ontology formed by the original standard. The original ontology layer only involves the calculation method of the regional rebound tester and the regional rebound value. In the application layer, according to the description of data calculation in the standard, 3 steps and 16 raw data are expanded. The 16 raw data correspond to the 16 field collection data (ns:ReboundRawValue1~ns:ReboundRawValue16) that need to be entered on site, and the corresponding test operator (ns:Tester1), data collection time (ns:TestDateTime), and the instrument used (ns:TestDevice) and other detailed information need to be supplemented. The three steps include: 1) data sorting (orderSet); 2) obtaining the middle 10 data (getMiddeleTen); 3) calculating the average value of the 10 data (getAverageTen). In this embodiment, a=0.4 and b=0.6 are set.

[0093] 3. Connection layer

[0094] The node network framework established in the application layer is used to calculate the correlation between each node. The correlation calculation is mainly based on the semantic similarity between the concepts of two nodes. The established correlation network, such as Figure 3 As shown in the structure of the connection layer, the numbers on the connection lines between nodes represent the similarity values.

[0095] 4. Diffusion layer

[0096] The diffusion layer is formed according to the on-site data entry situation. In this embodiment, 16 raw data input text boxes are set in the web page or App interface, which are associated with the nodes in ns:ReboundRawValue1~ns:ReboundRawValue16 in the connection layer. When data is input, a node is referenced from the connection layer, and then a corresponding node is generated in the diffusion layer to represent the input value, such as Figure 4 The structure in the diffusion layer is shown.

[0097] Finally, the 4-layer structure diagram of data organization is as follows Figure 5 shown.

[0098] The diffusion steps of an application case are as follows:

[0099] 1) The first rebound value data entered manually or transmitted by the on-site rebound hammer is labeled as "Rebound Value 1" and its value is 44;

[0100] 2) A diffusion layer data instance is generated based on the data, with a unique naming number of ns:RRValue5A2Z18E001. The instance is pointed to by the attribute ns:is_a by the ns:ReboundRawValue1 node referenced by the connection layer. The ns:hasValue attribute indicates that the current value is 44. The user is required to complete key information such as the inspection personnel, the equipment used, and the inspection area. The inspection time can be automatically entered by the system or manually supplemented by the user.

[0101] 3) At this time, the node ns:RRValue5A2Z18E001 instance is in the activated state, and act_base is set to 1.0, indicating full activation. In this embodiment, the diffusion strength defaults to ;

[0102] 4) Take the class node ns:ReboundRawValue1 of the node ns:RRValue5A2Z18E001 as the previous node, and search for the subsequent node set node_L1 connected to the object attribute of ns:ReboundRawValue1. In this example, there is only one node ns:areaReboundValueSet in this level of diffusion.

[0103] 5) Calculate the act_base of areaReboundValueSet using formula (4), where act_start is the act_base of the ns:ReboundRawValue1 node;

[0104] 6) With areaReboundValueSet as the front node, repeat steps 4) to 5) to implement the basic activation value calculation of the back nodes such as ns:ReboundRawValue2, ns:computingMethodFor, areaReboundValue, etc.

[0105] 7) Using the breadth-first method, calculate the act_base of each subsequent level along the direction of the object attributes of each activated node until the first-level act_base is less than act_min or the node it points to is activated, then stop spreading;

[0106] 8) The nodes whose activation values ​​are greater than or equal to act_min form the activated node set node_activated_set1;

[0107] 9) Use formula (5) to calculate the act_total of each node in node_activated_set1 one by one;

[0108] 10) For nodes where act_total ≥ act_min, act_base < act_min and there are unactivated nodes in the subsequent stage, they can continue to spread to the subsequent stage; otherwise, stop spreading.

[0109] 11) The newly spread nodes calculate act_total according to formula (4) and formula (5), and repeat step 10) to form the set of all activated nodes node_activated_set2.

[0110] 12) The nodes in node_activated_set2 are referenced to the diffusion layer to form the corresponding node network. The connection strength between the activated nodes can be obtained from the connection layer. At this time, the diffusion layer composed of several instances formed by inputting the original data of a rebound hammer is completed.

[0111] 13) Complete the activation and diffusion process of the input data such as the tester, testing equipment, and measuring area according to steps 3) to 12).

[0112] The real-time comparison steps are as follows:

[0113] 1) In this embodiment, 16 rebound hammer measurement values are input in batches, and the corresponding information such as the tester ns:Tester, the measuring area ns:MeasureAera, and the testing instrument ns:measuringDevice used is supplemented. The "presumed value of compressive strength" is manually set as the core vocabulary.

[0114] 2) The 16 rebound hammer measurement values, personnel, measuring area, and testing instrument are all data that need to be input in real time for this measurement. In the interface, "tester" is a description of a node concept, "Zhang San" is a text description of the data attribute of a tester, "measuring area" is a description of a node concept, and "5A2Z18-E" is the number of a measuring area instance.

[0115] 3) For the concept of a node, the node representing this concept is referenced from the connection layer and continues to be used in the diffusion layer.

[0116] 4) The input data such as the rebound hammer measurement values and the measuring area are formed into instance nodes. <ns:rrvalue5a2z18e001> ~ <ns:rrvalue5a2z18e016> 、 <ns:meaaera5a2z18e>;

[0117] 5) The application will calculate the data generated by the input data according to the requirements of the associated standards in the interface. In this example, the actual instance nodes that need to be compared are calculated according to the specification "Technical Specification for Testing Concrete Compressive Strength by Rebound Method" (JGJ / T 23-2011) <ns:fcue5a2z18e>;

[0118] 6) The vocabulary in the entire context data set forms a diffusion layer composed of multiple concept nodes and instance nodes through the activation and diffusion process;

[0119] Among them, input measurement area <ns:meaaera5a2z18e>The number is used as the activation <ns:measureaera>After that, the system will activate the component where the measurement area is located. <ns:bc2d32001>.member <ns:bc2d32001>Has attributes <ns:designrequirement>Point to the design requirements of this component <ns:bc2d32001designrequire>The design requirements include the design strength grade <ns:bc2d32001sg>, which contains a string "MPa" representing the physical unit and a numeric attribute 30, indicating the design strength grade C30.

[0120] 7) Within the activation scope, for instance nodes <ns:fcue5a2z18e>The node formed as the core incentive data looks for the target data node for comparison;

[0121] 8) Node <ns:ns:fcue5a2z18e>The node attributes of contain three attributes: a description text, a value, and a physical unit text. In the diffusion layer, the target node set that also contains three similar attributes is found through the sparql (SPARQL Protocol and RDF QueryLanguage) query statement;

[0122] 9) Calculate the similarity between the description text of the data attribute in the core vocabulary instance and the data attribute in each node in the target node set, which can be calculated using formula (1) or formula (2);

[0123] 10) If the core vocabulary to be compared is text, the three results with the highest similarity are selected as the real-time comparison results and output, and the similarity compared in step 8) is output at the same time, and the results of comparing these three results with the stimulus data values ​​are also given. For example, the output result in this example is:

[0124] Similar node identifiers, similarity, and comparison relationships

[0125] <ns:bc2d32001sg> ,0.9, <ns:ns:fcue5a2z18e>larger <ns:bc2d32001sg>.

[0126] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0127] It should be noted that, in this article, the terms "comprise", "include" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0128] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0129] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of them according to actual needs to achieve the purpose of the present embodiment, and no limitation is made here.

[0130] It is understandable that the system provided by the embodiment of the present invention corresponds to the method provided by the embodiment of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the above method.

[0131] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention. < / ns:bc2d32001sg> < / ns:designrequirement> < / ns:measureaera> < / ns:rrvalue5a2z18e016> < / ns:rrvalue5a2z18e001>

Claims

1. A real-time comparison method for construction engineering quality data, characterized in that: Collect standard or specification data of construction projects, and construct an ontology layer by structuring and knowledge processing the data; the ontology layer defines the meaning and attributes of concepts according to the hierarchical structure and logical relationship of concepts, wherein each concept forms a named node in the ontology layer network, and its attributes include object attributes and data attributes. The node and its attributes constitute a named entity in the ontology layer, and each named entity has a unique number that distinguishes it from other named entities in the entire data set; On the basis of the ontology layer, the concepts and attributes of the named entities are extended to construct an application layer; the application layer supplements the parameter acquisition steps and calculation process in the standard or specification data, and establishes constraints on the conclusions or numerical descriptions in the standard or specification data; The application layer can structure some typical cases into instance nodes to provide reference for subsequent similar applications; Based on the application layer, the node and attribute network is used as the basic framework, the correlation of the concepts represented by each node is calculated to indicate the connection strength between each entity, and the connection layer is constructed with the connection strength as the edge; On the basis of the connection layer, the quality concept or data of the construction project is input as the incentive data, the corresponding nodes in the connection layer network are activated and diffused to the affected nodes, and these node edges become the diffusion layer; after an activation event is completed, the instance node formed by this activation can be merged into the application layer to become the instance node for subsequent reference; Based on the multi-layer data organization structure constructed above, real-time comparison of construction project quality data is performed; within the activation range of the diffusion layer, matching target nodes are searched for nodes formed by core words in the input stimulus data; filtering is performed based on the type of core words to form a filter node set; the similarity between the concept corresponding to the core word and each concept in the filter node set is calculated; If the core vocabulary is a text, the result with the highest semantic similarity is selected as the comparison result output; If the core word is of numerical or time type, output the closest result and its size relationship; The multi-layer data organization structure can be used as a single layer or as a composite layer with multiple layers stacked together.

2. A real-time comparison method for construction engineering quality data according to claim 1, characterized in that: The knowledge processing includes knowledge modeling, named entity recognition, and relationship recognition and extraction; The attributes include object attributes and data attributes, wherein the object attributes are defined as the association between two nodes; and the data attributes are defined as the descriptive or numerical characteristics of a concept; The domain and range of the attribute form a constraint relationship for the concepts, values ​​and texts involved; the node has data attributes to indicate the specific content and meaning of the concept represented by the node, and object attributes to point to other nodes related to it.

3. A real-time comparison method for construction engineering quality data according to claim 1, characterized in that: The connection strength of the connection layer can be represented by the correlation between node concepts sim_use, which consists of two parts: the semantic similarity relationship sim_base between two concepts in a general context and the correlation sim_field between concepts in the field of construction engineering; The semantic similarity between two concepts in a general context is sim_base. According to the semantic similarity formula, the semantic similarity formula between concepts A and B is as follows: (1) Where f(A) is the number of web pages in the search engine library that contain concept A, f(B) is the number of web pages in the search engine library that contain concept B, f(A,B) is the number of web pages in the search engine library that contain both A and B, and N is the total number of web pages on the Internet referenced by the search engine library; The correlation between concepts in the field of construction engineering is sim_field. Technical documents, test reports, and operation and maintenance instructions for each link of construction engineering design, construction, acceptance, and operation and maintenance are collected and sorted to form a data set for construction engineering applications. In this set, the correlation sim_field between concepts of each node is formed by semantic vectors. Sim_field is calculated using the cosine similarity method. The specific formula is as follows: (2) Among them, A and B represent the semantic vectors of two concept texts. and They represent the modulus of the vector respectively; The correlation between the concepts represented by the two nodes, sim_use, is formed by combining sim_base and sim_field according to the weights: (3) Among them, a and b are hyperparameters and must satisfy 0≤a+b≤1, a≥0, b≥0.

4. A real-time comparison method for construction engineering quality data according to claim 1, characterized in that: In the construction of the diffusion layer, the activation degree of any node in the connection layer is the activation value act_total. When the node activation value is lower than the set value act_min, the node is not activated. All activated nodes with activation values ​​greater than or equal to act_min constitute the diffusion layer by reference or instance generation. The activation value act_total_i of a node i consists of two parts: ① the basic activation value, which represents the impact of the initial input of the incentive concept on node i through multi-level diffusion; ② the impact of the activated nodes around node i on node i. The calculation formula of act_total_i is as follows: The calculation formula of the basic activation value act_base_i of node i is: (4) Among them, act_start is the previous stage excitation intensity, is a hyperparameter, indicating the loss during the transmission process; On this basis, several nodes around node i are represented by set j, among which there are m activated nodes, k is any activated node in j, and the calculation formula of act_total_i is: (5) in, The initial value is set to 1 / m, where m is the number of active nodes in the j set. is the correlation between the concept represented by node k and the concept represented by node i; When inputting stimulus data, the node where the stimulus data is located is taken as the center, and the scope of influence of the stimulus from near to far is calculated, and a data diffusion zone is formed along the direction of the attribute. When the diffusion to the node activation value is lower than the set value act_min, the diffusion stops. At this time, the diffusion zone contains all the nodes affected by the stimulus, and the target nodes that can match the stimulus data are searched within the node range of the diffusion zone for comparison.

5. A real-time comparison method for construction engineering quality data according to claim 1, characterized in that: In the real-time comparison of construction project quality data, the construction project quality data or text and its related context are input to form a context data set; the related text in the context data is segmented to form an input vocabulary set, and the vocabulary set includes a core vocabulary and several related vocabulary. The core vocabulary can be manually specified, or the subject words can be extracted from the input vocabulary set as the core vocabulary by artificial intelligence; the vocabulary in the vocabulary set is divided into concepts and instances, and the concept mainly reflects an abstract expression of a class of physical objects and behavior descriptions. The node containing the concept can be directly referenced from the connection layer to the diffusion layer. The instance is the specific data for this application, and is instantiated in the diffusion layer according to a node in the connection layer to form an instance node.

6. A real-time comparison method for construction engineering quality data according to any one of claims 3 or 4, characterized in that: The hyperparameters need to be set before starting the activation process; if the target node for comparison is not in the region of the diffusion layer, the hyperparameters are adjusted.

7. A real-time comparison method for construction engineering quality data according to claim 1, characterized in that: If, during comparison, the node concept formed by the core vocabulary is consistent with the concept represented by the target node, but one of the nodes does not have a corresponding attribute value, the ontology layer knowledge structure is adjusted to supplement the node attributes.

8. A real-time comparison method for construction engineering quality data according to claim 4, characterized in that: In the process of building a multi-layer data organization, while inputting the construction project quality data, artificially forming the comparison results, and establishing the data training set for training the connection layer network. .

Images