Method for generating semantic network of building equipment

Abstract

The invention discloses a building equipment semantic network generation method, which is a building metadata model generation method based on a deep learning algorithm, and can extract and analyze building metadata from a two-dimensional or three-dimensional building design drawing, and automatically construct a corresponding building metadata model. Brick Schema is used as a structure and a rule of a building metadata model; transform is used as the core of a deep learning algorithm. According to the method, Transform is used as a core of a deep learning algorithm, Brick Schema is used as a structure and a rule of the building metadata model, and by means of the algorithm, the structure and the rule, the high-quality building metadata model is automatically generated from a two-dimensional or three-dimensional building design drawing so as to meet the requirement of building intelligent application.

Description

Technical Field

[0001] The invention relates to a method for generating a semantic network of building equipment. Background Art

[0002] Building Information Modeling is a digital technology-based building design, construction and operation method that can improve the efficiency and quality of building projects. Building Information Modeling contains information about the physical, logical and virtual assets of a building and their relationships, which is called Building Metadata. Building Metadata is the basis of building intelligent applications and can support functions such as energy management, security monitoring, and equipment control of buildings.

[0003] However, the existing methods for generating architectural metadata models have the following shortcomings:

[0004] Manual generation method: It requires manual extraction and analysis of architectural metadata from two-dimensional or three-dimensional architectural design drawings, and the construction of architectural metadata models based on certain rules and standards. This method is not only time-consuming and labor-intensive, but also prone to errors and omissions.

[0005] Automated generation method: Use computer programs to extract and analyze architectural metadata from two-dimensional or three-dimensional architectural design drawings, and build architectural metadata models based on certain logical algorithms and templates. Although this method can improve efficiency, these fixed logical processing methods and fixed templates are difficult to handle the diversity, complexity and dynamics of architectural design drawings, as well as the consistency, scalability and interoperability of architectural metadata models.

[0006] Therefore, the technical problem to be solved by the present invention is: using artificial intelligence technology to automatically generate high-quality architectural metadata models from two-dimensional or three-dimensional architectural design drawings to meet the needs of intelligent building applications.

[0007] In order to solve the above-mentioned problems, the present invention is proposed. Summary of the invention

[0008] The present invention aims to provide a method for generating a semantic network of building equipment to overcome the deficiencies in the prior art solutions.

[0009] To achieve the above objectives, the following technical solutions are adopted:

[0010] A method for generating a semantic network of building equipment is a method for generating a building metadata model based on a deep learning algorithm. The method can extract and analyze building metadata from two-dimensional or three-dimensional building design drawings and automatically construct a corresponding building metadata model; Brick Schema is used as the structure and rules of the building metadata model; Transformer is used as the core of the deep learning algorithm.

[0011] Specifically, the generation method is divided into the following steps:

[0012] The first step is to convert the two-dimensional or three-dimensional architectural design drawings into serialized data in Protocol Buffer format;

[0013] The second step is to input the serialized data in Protocol Buffer format into the Transformer model. The Transformer identifies different types of building metadata and converts them into concepts and relationships defined in the Brick Schema.

[0014] The third step is to construct the corresponding building metadata model based on the Brick Schema concepts and relationships output by the Transformer model and store it in an RDF format file;

[0015] The fourth step is to verify and optimize the generated building metadata model.

[0016] Furthermore, in the first step, the Protocol Buffer codec is used to implement the conversion between two-dimensional or three-dimensional architectural design drawings and Protocol Buffer format serialized data.

[0017] Furthermore, in the second step, the encoder-decoder structure of Transformer is used to realize the conversion between ProtocolBuffer format serialized data and Brick Schema concepts and relationships.

[0018] Furthermore, in the third step, an RDF serializer is used to implement the conversion between Brick Schema concepts and relationships and RDF format files.

[0019] Furthermore, in the fourth step, the Brick Schema validator and optimizer are used to implement the validation and optimization of the building metadata model.

[0020] Beneficial technical effects:

[0021] The present invention uses Transformer as the core of the deep learning algorithm and Brick Schema as the structure and rules of the building metadata model. With the help of the above algorithms, structures and rules, high-quality building metadata models are automatically generated from two-dimensional or three-dimensional architectural design drawings to meet the needs of building intelligent applications. DETAILED DESCRIPTION

[0022] A method for generating a semantic network of building equipment is a method for generating a building metadata model based on a deep learning algorithm. The method can extract and analyze building metadata from two-dimensional or three-dimensional building design drawings and automatically construct a corresponding building metadata model. Brick Schema is used as the structure and rules of the building metadata model. Brick Schema is a unified building metadata specification that can describe the physical, logical and virtual assets of a building and their relationships.

[0023] The present invention uses Transformer as the core of the deep learning algorithm. Transformer is a neural network architecture based on the self-attention mechanism, which can effectively process serialized data such as natural language and meaningful serialized data.

[0024] The method for generating a building equipment semantic network of the present invention is divided into the following steps:

[0025] The first step is to convert the two-dimensional or three-dimensional architectural design drawings into serialized data in Protocol Buffer format. Protocol Buffer is a lightweight data exchange format that can accurately represent structured objects; use Protocol Buffer's codec to realize the conversion between two-dimensional or three-dimensional architectural design drawings and Protocol Buffer format serialized data.

[0026] The second step is to input the serialized data in Protocol Buffer format into the Transformer model. The trained Transformer model can identify the features such as graphics, text, and symbols in the serialized data, and convert them into the concepts and relationships defined in the Brick Schema according to the corresponding relationships of different types of building metadata. The encoder-decoder structure of Transformer is used to realize the conversion between the serialized data in Protocol Buffer format and the concepts and relationships of Brick Schema.

[0027] The third step is to build the corresponding building metadata model based on the Brick Schema concepts and relationships output by the Transformer model and store it in RDF format, which is a semantic web standard for describing resources and relationships. The RDF serializer is used to implement the conversion between Brick Schema concepts and relationships and RDF format files.

[0028] The fourth step is to verify and optimize the generated building metadata model, using different evaluation indicators, methods and tools to detect and improve the accuracy, completeness, consistency and usability of the building metadata model. Use BrickSchema's validator and optimizer to implement verification and optimization of the building metadata model.

[0029] Example 1

[0030] The present invention provides a method for generating a building metadata model based on an artificial intelligence algorithm, which can automatically generate a high-quality building metadata model from a two-dimensional or three-dimensional building design drawing to meet the needs of building intelligent applications.

[0031] The generation method of the present invention is described in detail below with the help of some codes and taking an existing data as an example.

[0032] The existing data is a two-dimensional architectural design drawing, which represents a simple office space, including some equipment entities, such as doors, windows, lights, air conditioners, temperature sensors, humidity sensors, etc. The process described in this patent can generate a building metadata model that conforms to the Brick Schema from the drawing to facilitate intelligent management and control of these equipment entities.

[0033] The specific steps are:

[0034] The first step is to convert the two-dimensional architectural design drawings into serialized data in Protocol Buffer format. Use the Protocol Buffer codec to implement this step. The codec includes an encoder and a decoder. The encoder converts the graphics, text, symbols and other information in the drawings into the message type of Protocol Buffer and serializes it into a byte stream. The decoder deserializes the byte stream into the message type and restores it to the information in the drawing. For example, the following message type can be defined to represent the device entity:

[0035]

[0036] The above message type {device} includes the following devices: doors, windows, lights, air conditioners, temperature sensors and humidity sensors. At the same time, a label is defined for each device, for example, the door is numbered 0, the window is numbered 1, the light is numbered 2, the air conditioner is numbered 3, the temperature sensor is numbered 4, and the humidity sensor is numbered 5.

[0037] At the same time, the following message contents are defined for each device: device type, device number, device name, device location, and device function.

[0038] Then use the encoder to convert each device entity in the drawing into the corresponding message type and serialize it into a byte stream. For example, convert the door in the upper left corner of the drawing into the following message type:

[0039]

[0040] Device type: DOOR, device number: D1, device name: Door_1, device location: North_Wall, device function: Entrance. The message type is then serialized into a byte stream: 0a020800120244311a06446f6f725f12284e6f7274685f57616c6c1a07456e7472616e6365.

[0041] Similarly, other device entities in the drawing can also be converted into corresponding byte streams, and all byte streams can be spliced ​​together to form the final Protocol Buffer format serialized data.

[0042] In the second step, the serialized data in Protocol Buffer format is input into the Transformer model. The Transformer model identifies different types of building metadata based on the features in the data and converts them into concepts and relationships defined in BrickSchema. The encoder-decoder structure of Transformer is used to implement this step. Among them, the encoder can map the message type in the serialized data into a vector representation, and extract its semantic features through the self-attention mechanism and feedforward neural network. The decoder can map the vector representation to the concepts and relationships in Brick Schema, and generate its semantic representation through the self-attention mechanism, the encoder-decoder attention mechanism and the feedforward neural network. For example, the vector representation output by the encoder can be used as the input of the decoder, and then the decoder can generate the following BrickSchema concepts and relationships:

[0043]

[0044]

[0045] These concepts and relationships represent the types, locations, functions, and connections between equipment entities.

[0046] The third step is to construct the corresponding building metadata model based on the Brick Schema concepts and relationships output by the Transformer model, and store it as a file in RDF format. This patent uses an RDF serializer to implement this step. The serializer includes a mapper and a serializer. The mapper can map the concepts and relationships in the Brick Schema to resources, attributes, and values ​​in RDF, and generate the corresponding URI based on the namespace and vocabulary of the Brick Schema. The serializer serializes the resources, attributes, and values ​​in RDF into formats such as RDF / XML, N-Triples, Turtle, etc., and stores them as files.

[0047] For example, a mapper can be used to map the concepts and relationships in the Brick Schema to the following RDF resources, properties, and values:

[0048]

[0049]

[0050] These resources, properties, and values ​​are then serialized into RDF / XML format using a serializer and stored in a file called metadata.xml.

[0051] The fourth step is to use Brick Schema's validator and optimizer to validate and optimize the generated architectural metadata model. The validator can detect whether there are syntax errors, logic errors, conflict errors, etc. in the architectural metadata model, and give corresponding prompts and correction suggestions. The optimizer can analyze whether there is redundant information, missing information, inconsistent information, etc. in the architectural metadata model, and give corresponding deletion, supplementation, and modification suggestions.

[0052] The above are only implementation modes of the present invention, and are not intended to limit the patent scope of the present invention. On the other hand, any equivalent structure made using the contents of the present invention specification, directly or indirectly used in other related technical fields, is also within the patent protection scope of the present invention.

Claims

1. A method for generating a semantic network of building equipment, characterized in that: It is a method for generating architectural metadata models based on deep learning algorithms. This method can extract and analyze architectural metadata from two-dimensional or three-dimensional architectural design drawings and automatically construct the corresponding architectural metadata model. Brick Schema is used as the structure and rules of the building metadata model; Transformer is used as the core of the deep learning algorithm.

2. A method for generating a building equipment semantic network according to claim 1, characterized in that: The generation method is divided into the following steps: The first step is to convert the two-dimensional or three-dimensional architectural design drawings into serialized data in Protocol Buffer format; The second step is to input the serialized data in Protocol Buffer format into the Transformer model. The Transformer identifies different types of building metadata and converts them into concepts and relationships defined in the Brick Schema. The third step is to construct the corresponding building metadata model based on the Brick Schema concepts and relationships output by the Transformer model and store it in an RDF format file; The fourth step is to verify and optimize the generated building metadata model.

3. The method for generating a building equipment semantic network according to claim 1, characterized in that: In the first step, the Protocol Buffer codec is used to convert two-dimensional or three-dimensional architectural design drawings into serialized data in Protocol Buffer format.

4. The method for generating a building equipment semantic network according to claim 1, characterized in that: In the second step, the encoder-decoder structure of Transformer is used to realize the conversion between Protocol Buffer format serialized data and BrickSchema concepts and relationships.

5. A method for generating a building equipment semantic network according to any one of claims 1 to 4, characterized in that: In the third step, the RDF serializer is used to convert Brick Schema concepts and relationships into RDF format files.

6. The method for generating a building equipment semantic network according to claim 1, characterized in that: In the fourth step, the Brick Schema validator and optimizer are used to validate and optimize the building metadata model.

Images