CN113743024A - Artificial intelligence twin type training method based on building information model - Google Patents
Artificial intelligence twin type training method based on building information model Download PDFInfo
- Publication number
- CN113743024A CN113743024A CN202111293215.3A CN202111293215A CN113743024A CN 113743024 A CN113743024 A CN 113743024A CN 202111293215 A CN202111293215 A CN 202111293215A CN 113743024 A CN113743024 A CN 113743024A
- Authority
- CN
- China
- Prior art keywords
- training
- model
- clustering function
- artificial intelligence
- initial
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention discloses an artificial intelligence twin type training method based on a building information model, which comprises the following steps: acquiring a plurality of building design training sets, and acquiring initial feature vectors of members corresponding to the building design training sets from a building information model for each building design training set; inputting the initial characteristic vector into an initial model for each stage of the building design, and training the initial model to obtain an artificial intelligence stage model; the initial model is obtained by cascading a clustering function and a neural network; and fusing the plurality of artificial intelligence stage models, and training the plurality of fused artificial intelligence stage models to obtain an artificial intelligence twin model. According to the invention, each stage of the building design is trained and corrected, and the clustering function and the neural network are cascaded, so that the simulation precision of the artificial intelligent twin model based on the building information model is higher.
Description
Technical Field
The invention relates to the technical field of building design, in particular to an artificial intelligence twin type training method based on a building information model.
Background
With the development of the technology, various people have gradually adopted the Artificial Intelligence (AI) technology to solve the problem of part of the work with high repeatability and long periodicity. However, in the current building design process, a lot of work with high repeatability and long periodicity exists, manual design is adopted, so that a lot of time cost and working cost are wasted, and the prior art also has AI simulation.
Thus, there is still a need for improvement and development of the prior art.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide an artificial intelligence twin training method based on a building information model, aiming at solving the problem that the simulation method in the prior art corrects the calculation according to the final result of the design, does not pay attention to the correction of each stage process, and thus the AI simulation precision is not high.
The technical scheme adopted by the invention for solving the problems is as follows:
in a first aspect, an embodiment of the present invention provides an artificial intelligence twin training method based on a building information model, where the method includes:
acquiring a plurality of building design training sets, and acquiring initial feature vectors of members corresponding to the building design training sets from a building information model for each building design training set;
inputting the initial characteristic vector into an initial model for each stage of the building design, and training the initial model to obtain an artificial intelligence stage model; the initial model is obtained by cascading a clustering function and a neural network;
and fusing the plurality of artificial intelligence stage models, and training the plurality of fused artificial intelligence stage models to obtain an artificial intelligence twin model.
In one implementation, the building design training set includes, among other things, a project name, a project number, and a project identification number.
In one implementation, the initial feature vector includes a component identification number, a component parameter value, a distance of the component from an adjacent component, a mapping vector of the component from the adjacent component, and a membership of the component to the adjacent component.
In one implementation, the obtaining a plurality of building design training sets includes, for each of the building design training sets, after obtaining initial feature vectors of members corresponding to the building design training set from a building information model, the method includes:
for each of the building design training sets, tracking a state of variation of components in the building design training set and adding the state of variation to the initial feature vector.
In one implementation, the inputting the initial feature vector into an initial model for each stage of the building design, and training the initial model to obtain an artificial intelligence stage model includes:
inputting the initial characteristic vector into a clustering function, and training the clustering function to obtain a trained clustering function;
freezing the trained clustering function, inputting the output of the trained clustering function into a neural network model for training, and finishing the training of the neural network model when the training of the neural network model meets the preset condition;
and cascading the trained clustering function and the trained neural network model to obtain an artificial intelligence stage model.
In one implementation, the inputting the initial feature vector into a clustering function and training the clustering function to obtain a trained clustering function includes:
acquiring user drawing data;
inputting the initial characteristic vector into a clustering function to obtain output data;
and training the clustering function according to the user drawing data and the output data to obtain the trained clustering function.
In an implementation manner, the training the clustering function according to the user drawing data and the output data to obtain a trained clustering function includes:
when the user drawing data and the output data are different, sending prompt information;
acquiring feedback information sent by a user according to the prompt information; wherein the feedback information comprises an error and a correct;
and training the clustering function according to the feedback information to obtain the trained clustering function.
In an implementation manner, the training the clustering function according to the feedback information to obtain a trained clustering function includes:
when the feedback information is wrong, storing the initial characteristic vector corresponding to the feedback information in a wrong set, and inputting the initial characteristic vector corresponding to the feedback information into a clustering function to obtain intermediate data;
calculating a mean square error value of the intermediate data and the user rendered data;
repeatedly executing the steps of storing the initial characteristic vector corresponding to the feedback information in an error set and inputting the initial characteristic vector corresponding to the feedback information into a clustering function to obtain intermediate data when the feedback information is wrong;
and when the mean square error value is smaller than a preset value or the times of training the clustering function reach the preset times, stopping training to obtain the trained clustering function.
In a second aspect, an embodiment of the present invention further provides an artificial intelligence twin training apparatus based on a building information model, where the apparatus includes:
the initial feature vector acquisition unit is used for acquiring a plurality of building design training sets, and for each building design training set, acquiring an initial feature vector of a component corresponding to the building design training set from a building information model;
the artificial intelligence stage model obtaining unit is used for inputting the initial characteristic vector into an initial model for each stage of the building design and training the initial model to obtain an artificial intelligence stage model; the initial model is obtained by cascading a clustering function and a neural network;
and the artificial intelligence twin model acquisition unit is used for fusing the artificial intelligence stage models and training the fused artificial intelligence stage models to obtain an artificial intelligence twin model.
In a third aspect, an embodiment of the present invention further provides an intelligent terminal, including a memory, and one or more programs, where the one or more programs are stored in the memory, and configured to be executed by one or more processors includes a method for performing the artificial intelligence twin training method based on the building information model as described in any one of the above.
In a fourth aspect, embodiments of the present invention also provide a non-transitory computer-readable storage medium, wherein instructions of the storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the artificial intelligence twin training method based on a building information model as described in any one of the above.
The invention has the beneficial effects that: firstly, acquiring a plurality of building design training sets, and acquiring an initial feature vector of a component corresponding to each building design training set from a building information model for each building design training set; then inputting the initial characteristic vector into an initial model for each stage of the building design, and training the initial model to obtain an artificial intelligence stage model; the initial model is obtained by cascading a clustering function and a neural network; finally, fusing the plurality of artificial intelligence stage models, and training the plurality of fused artificial intelligence stage models to obtain an artificial intelligence twin model; therefore, the embodiment of the invention trains and corrects each stage of the building design and cascades the clustering function and the neural network, so that the simulation precision of the artificial intelligent twin model based on the building information model is higher.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of an artificial intelligence twin training method based on a building information model according to an embodiment of the present invention
Fig. 2 is a schematic block diagram of an artificial intelligence twin training device based on a building information model according to an embodiment of the present invention.
Fig. 3 is a schematic block diagram of an internal structure of an intelligent terminal according to an embodiment of the present invention.
Detailed Description
The invention discloses an artificial intelligence twin type training method based on a building information model, and in order to make the purpose, technical scheme and effect of the invention clearer and clearer, the invention is further described in detail below by referring to the attached drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. As used herein, the term "and/or" includes all or any element and all combinations of one or more of the associated listed items.
It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
In the prior art, the simulation method corrects the calculation according to the final result of the design, and correction of each stage process is not concerned, so that the AI simulation precision is not high.
In order to solve the problems in the prior art, the embodiment provides an artificial intelligence twin training method based on a building information model, and each stage of a building design is trained and corrected, and a clustering function and a neural network are cascaded, so that the simulation precision of the artificial intelligence twin model based on the building information model is higher. When the method is specifically implemented, a plurality of building design training sets are obtained, and for each building design training set, an initial feature vector of a component corresponding to the building design training set is obtained from a building information model; then inputting the initial characteristic vector into an initial model for each stage of the building design, and training the initial model to obtain an artificial intelligence stage model; the initial model is obtained by cascading a clustering function and a neural network; and finally, fusing the plurality of artificial intelligence stage models, and training the plurality of fused artificial intelligence stage models to obtain an artificial intelligence twin model.
Exemplary method
The embodiment provides an artificial intelligence twin type training method based on a building information model, and the method can be applied to an intelligent terminal for building design. As shown in fig. 1 in detail, the method includes:
s100, obtaining a plurality of building design training sets, and for each building design training set, obtaining an initial feature vector of a component corresponding to the building design training set from a building information model;
in this embodiment, because part of the work of the architectural designer has repeatability and a long period, in order to reduce the time and cost waste caused by the part of the work, the artificial intelligence AI is trained to assist the design, and in order to realize the training of the artificial intelligence twin model, training data needs to be acquired, in the present invention, the artificial intelligence twin model includes a clustering function and a neural network, and 3 databases are configured at the same time: coefficient library, training set library, error set library. Wherein the coefficient bank is used to store the coefficients of the clustering function, such as f (x) = Ax + Bx + C, and A, B, C here are the coefficients stored in the coefficient bank. The training set is used to store various data in the project design process. The error set is used for storing data judged as errors by designers in the project design process. In order to realize more accurate simulation of the trained model, the more abundant the building design training sets are, the better the acquired building design training sets are, the obtained building design training sets are stored in a training set library, and practically, all the building design training sets can be trained. Each building design training set comprises a project name, a project number and a project identity identification number (ID), and the project name, the project number and the project identity identification number (ID) are obtained through an API provided by BIM software and are used for distinguishing the identification of each training set. When training is carried out each time, a building design training set is taken, an initial feature vector of a member corresponding to the building design training set is obtained through an API provided by BIM software, and the initial feature vector comprises a member identification number, a member parameter value, a distance between the member and an adjacent member, a drawing vector between the member and the adjacent member and a membership between the member and the adjacent member, wherein the member identification number, the member parameter value, the distance between the member and the adjacent member, the drawing vector between the member and the adjacent member and the membership between the member and the adjacent member are all member information increased in an item process, and the member parameter value comprises: the names, sizes, serial numbers, materials, belonged specialties, positioning lines, positioning points, bottom elevations, top elevations, bottom offsets, top offsets, volumes, projected areas, belonged floors, component purposes, reinforcement information, structural purposes, creation stages, demolition stages, reference elevations, starting point offsets, end point offsets, cross section rotation angles, circumferences, thicknesses, slopes, room boundaries, contour lines, main body IDs and geometric center coordinates of the components. The distance between a member and an adjacent member is the distance between the geometric centers of the two members, and the distance value is used for collecting rich member information and inputting the rich member information into the clustering function. The drawing vector of the member and the adjacent member is a unit vector from the geometric center of the current member to the geometric center of the adjacent member. The dependency relationship between a component and an adjacent component refers to the ID value of the main body of the component, which is recorded in the BIM software, for example, if the main body of a door or a window is a wall, the ID value of the main body of the wall is recorded in the parameter column of the BIM software.
In one implementation, the obtaining a plurality of building design training sets includes, for each of the building design training sets, the following steps after obtaining initial feature vectors of members corresponding to the building design training set from a building information model: for each of the building design training sets, tracking a state of variation of components in the building design training set and adding the state of variation to the initial feature vector.
Specifically, since the mistaken deletion is inevitable during the building design process, for each building design training set, the change state of the components in the building design training set is tracked, that is, the API provided by the BIM software is used to track the modification of each ID in the current training set during the design process, for example, the parameters of the components are modified or the positions of the components are moved, if the components are deleted during the project design process, the last positions of the deleted components are marked as mark points, and if the newly added components are the same type of components, the deleted components are associated, and the tracking period is consistent with the original period of the project. That is, in the process of replacing the component in the item, the identification ID of the component is not changed, the component deleted by mistake by the user is prevented from being regarded as a newly added component, the last position of the deleted component is recorded, if the same type of component is newly added at the same position later, the previously deleted component is regarded as deleted by mistake, the change state is deleted by mistake, otherwise, the change state is added. The variation state is added to the initial feature vector.
After the initial feature vector is obtained, the following steps can be performed as shown in fig. 1: s200, inputting the initial characteristic vector into an initial model for each stage of building design, and training the initial model to obtain an artificial intelligence stage model; the initial model is obtained by cascading a clustering function and a neural network;
specifically, the conventional artificial intelligence method for building design corrects the calculation correctness of the AI according to the final result of the whole project process, and generally does not pay attention to the result of the process, which may result in low precision of the AI-aided simulation design. Generally, a project is divided into a plurality of stages, such as the steps of firstly establishing a shaft network, then making an elevation, then designing a column, a beam and a wall, and finally designing a door and a window on the wall. According to the method, the initial model of each stage in the project is trained and corrected independently to obtain the artificial intelligence stage model, each stage is trained and corrected, and finally the artificial intelligence twin model obtained after the artificial intelligence stage models are fused is more accurate in auxiliary design simulation, wherein in order to further improve the precision, the initial model comprises a clustering function and a neural network and is formed by cascading the clustering function and the neural network. Correspondingly, for each stage of the building design, inputting the initial feature vector into an initial model, and training the initial model to obtain an artificial intelligence stage model, including the following steps:
s201, inputting the initial characteristic vector into a clustering function, and training the clustering function to obtain a trained clustering function;
s202, freezing the trained clustering function, inputting the output of the trained clustering function into a neural network model for training, and finishing the training of the neural network model when the training of the neural network model meets the preset condition;
and S203, cascading the trained clustering function and the trained neural network model to obtain an artificial intelligence stage model.
Specifically, the clustering function is cascaded with the neural network, and the input data may be input to the neural network first and then the output of the neural network is input to the clustering function, or the input data may be input to the clustering function first and then the output of the clustering function is input to the neural network. In this embodiment, the initial feature vector is first input to the clustering function, and then the output of the clustering function is input to the neural network. In one implementation, the initial feature vector is input into a clustering function, and the clustering function is trained to converge preferentially. Correspondingly, the step of inputting the initial feature vector into a clustering function and training the clustering function to obtain a trained clustering function comprises the following steps: acquiring user drawing data; inputting the initial characteristic vector into a clustering function to obtain output data; and training the clustering function according to the user drawing data and the output data to obtain the trained clustering function.
Specifically, in order to converge the clustering function, a correct reference value is required, so that user drawing data is required to be obtained first, and the user drawing data includes the position of the primitive, the parameter value, and the relationship with the adjacent primitive. Meanwhile, the initial feature vector is input into a clustering function to obtain output data, then the difference value between the user drawing data and the output data can be minimized, and the output data can be controlled to approach the user drawing data to train the clustering function so as to obtain a trained clustering function. In an implementation manner, the training the clustering function according to the user drawing data and the output data to obtain a trained clustering function includes the following steps: when the user drawing data and the output data are different, sending prompt information; acquiring feedback information sent by a user according to the prompt information; wherein the feedback information comprises an error and a correct; and training the clustering function according to the feedback information to obtain the trained clustering function.
In practice, when the user drawing data and the output data are different, prompt information is sent, such as prompting the user: the result of the artificial intelligent twin model simulation is inconsistent with the current drawn component, whether the result of the artificial intelligent twin model simulation is correct or not is judged, and the user can judge whether the result is correct or wrong according to the result of the artificial intelligent twin model simulation. Then the system acquires feedback information sent by the user according to the prompt information; wherein the feedback information comprises an error and a correct; and finally, training the clustering function according to the feedback information to obtain the trained clustering function. Correspondingly, the training the clustering function according to the feedback information to obtain a trained clustering function includes the following steps: when the feedback information is wrong, storing the initial characteristic vector corresponding to the feedback information in a wrong set, and inputting the initial characteristic vector corresponding to the feedback information into a clustering function to obtain intermediate data; calculating a mean square error value of the intermediate data and the user rendered data; repeatedly executing the steps of storing the initial characteristic vector corresponding to the feedback information in an error set and inputting the initial characteristic vector corresponding to the feedback information into a clustering function to obtain intermediate data when the feedback information is wrong; and when the mean square error value is smaller than a preset value or the times of training the clustering function reach the preset times, stopping training to obtain the trained clustering function.
Specifically, when the feedback information is correct, no operation is performed, and the system continues to track the variation state of the components in the building design training set and add the variation state to the initial feature vector. When the feedback information is wrong, storing the initial characteristic vector corresponding to the feedback information in an error set according to a preset period, correcting the coefficient of the clustering function under the action of the error set, and verifying the accuracy of artificial intelligence twin model (AI) aided design. In the present invention, there is a period from the beginning to the end of the project, and since each project has uniqueness and uniqueness, the preset period is uncertain, and whether the project is ended is generally determined according to the project status set by the BIM project management platform. When the number of the initial feature vectors of the newly added component in the error set is reduced again, it means that the judgment of the artificial intelligence twin model AI tends to be correct. Meanwhile, inputting the initial characteristic vector corresponding to the feedback information into a clustering function to obtain intermediate data; calculating a mean square error value of the intermediate data and the user rendered data; that is, the user drawing data is a correct value, the intermediate data is an error value, the mean square error of the correct value and the error value is calculated, the correct value is used as a reference, the coefficient of the clustering function is corrected, the intermediate data approaches to the correct value, the clustering function coefficient corresponding to the correct value is recorded in the coefficient library, and the correct interval of the clustering function coefficient is obtained. And repeating the steps of storing the initial characteristic vector corresponding to the feedback information in an error set when the feedback information is wrong, inputting the initial characteristic vector corresponding to the feedback information into a clustering function to obtain intermediate data, further reducing a correct interval of a clustering function coefficient, stopping training when the mean square error value is smaller than a preset value or the number of times of training the clustering function reaches a preset number (for example, the number of times of repetition is 10000), obtaining a trained clustering function and a correct interval value of the clustering function coefficient, wherein after multiple times of correction, the interval range is smaller and smaller, and any value in the interval can be infinitely close to correct as a coefficient. The clustering function can provide basis for the design of a follow-up artificial intelligence twin model and the training of a neural network.
After the clustering function is trained, the trained clustering function can be frozen, the output of the trained clustering function is input into a neural network model for training, and when the training of the neural network model meets the preset condition, the training of the neural network model is completed; since the training method for the neural network is already available in the prior art, it is not described herein again. And finally, cascading the trained clustering function and the trained neural network model to obtain an artificial intelligence stage model.
After obtaining the artificial intelligence phase model, the following steps can be performed as shown in fig. 1: s300, fusing the plurality of artificial intelligence stage models, and training the plurality of fused artificial intelligence stage models to obtain an artificial intelligence twin model.
Specifically, the artificial intelligence stage model is a sub-model generated by simulating different stages of a project, all artificial intelligence stage models are fused in order to complete a complete project, namely, the artificial intelligence stage models are cascaded according to the sequence of the project, and a plurality of the artificial intelligence stage models after fusion are trained to finally form an artificial intelligence twin model.
Exemplary device
As shown in fig. 2, an embodiment of the present invention provides an artificial intelligence twin training apparatus based on a building information model, the apparatus includes an initial feature vector obtaining unit 401, an artificial intelligence stage model obtaining unit 402, and an artificial intelligence twin model obtaining unit 403, wherein: an initial feature vector obtaining unit 401, configured to obtain a plurality of building design training sets, and for each building design training set, obtain an initial feature vector of a component corresponding to the building design training set from a building information model;
an artificial intelligence stage model obtaining unit 402, configured to input the initial feature vector into an initial model for each stage of the building design, and train the initial model to obtain an artificial intelligence stage model; the initial model is obtained by cascading a clustering function and a neural network;
an artificial intelligence twin model obtaining unit 403, configured to fuse the plurality of artificial intelligence stage models, and train the plurality of artificial intelligence stage models after being fused, so as to obtain an artificial intelligence twin model.
Based on the above embodiment, the present invention further provides an intelligent terminal, and a schematic block diagram thereof may be as shown in fig. 3. The intelligent terminal comprises a processor, a memory, a network interface, a display screen and a temperature sensor which are connected through a system bus. Wherein, the processor of the intelligent terminal is used for providing calculation and control capability. The memory of the intelligent terminal comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the intelligent terminal is used for being connected and communicated with an external terminal through a network. The computer program is executed by a processor to implement a method of artificial intelligence twin training based on a building information model. The display screen of the intelligent terminal can be a liquid crystal display screen or an electronic ink display screen, and the temperature sensor of the intelligent terminal is arranged inside the intelligent terminal in advance and used for detecting the operating temperature of internal equipment.
It will be understood by those skilled in the art that the schematic diagram in fig. 3 is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation to the intelligent terminal to which the solution of the present invention is applied, and a specific intelligent terminal may include more or less components than those shown in the figure, or combine some components, or have different arrangements of components.
In one embodiment, an intelligent terminal is provided that includes a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for:
acquiring a plurality of building design training sets, and acquiring initial feature vectors of members corresponding to the building design training sets from a building information model for each building design training set;
inputting the initial characteristic vector into an initial model for each stage of the building design, and training the initial model to obtain an artificial intelligence stage model; the initial model is obtained by cascading a clustering function and a neural network;
and fusing the plurality of artificial intelligence stage models, and training the plurality of fused artificial intelligence stage models to obtain an artificial intelligence twin model.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, databases, or other media used in embodiments provided herein may include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
In summary, the invention discloses an artificial intelligence twin type training method based on a building information model, which comprises the following steps: acquiring a plurality of building design training sets, and acquiring initial feature vectors of members corresponding to the building design training sets from a building information model for each building design training set; inputting the initial characteristic vector into an initial model for each stage of the building design, and training the initial model to obtain an artificial intelligence stage model; the initial model is obtained by cascading a clustering function and a neural network; and fusing the plurality of artificial intelligence stage models, and training the plurality of fused artificial intelligence stage models to obtain an artificial intelligence twin model. According to the invention, each stage of the building design is trained and corrected, and the clustering function and the neural network are cascaded, so that the simulation precision of the artificial intelligent twin model based on the building information model is higher.
Based on the above embodiments, the present invention discloses an artificial intelligence twin training method based on a building information model, it should be understood that the application of the present invention is not limited to the above examples, and it will be obvious to those skilled in the art that modifications and variations can be made in the light of the above description, and all such modifications and variations are within the scope of the appended claims.
Claims (10)
1. An artificial intelligence twin type training method based on a building information model is characterized by comprising the following steps:
acquiring a plurality of building design training sets, and acquiring initial feature vectors of members corresponding to the building design training sets from a building information model for each building design training set;
inputting the initial characteristic vector into an initial model for each stage of the building design, and training the initial model to obtain an artificial intelligence stage model; the initial model is obtained by cascading a clustering function and a neural network;
and fusing the plurality of artificial intelligence stage models, and training the plurality of fused artificial intelligence stage models to obtain an artificial intelligence twin model.
2. The artificial intelligence twin training method based on building information model according to claim 1, wherein the building design training set includes an item name, an item number, and an item identification number.
3. The artificial intelligence twin training method based on building information model according to claim 1, wherein the initial feature vector includes a component identification number, a component parameter value, a distance of a component from an adjacent component, a mapping vector of a component from an adjacent component, and a membership of a component to an adjacent component.
4. The method of claim 1, wherein the obtaining a plurality of training sets of building design comprises, for each training set of building design, obtaining initial feature vectors of components corresponding to the training set of building design from the building information model, and then:
for each of the building design training sets, tracking a state of variation of components in the building design training set and adding the state of variation to the initial feature vector.
5. The artificial intelligence twin training method based on building information model according to claim 4, wherein for each stage of building design, inputting the initial feature vector into an initial model and training the initial model, and obtaining an artificial intelligence stage model comprises:
inputting the initial characteristic vector into a clustering function, and training the clustering function to obtain a trained clustering function;
freezing the trained clustering function, inputting the output of the trained clustering function into a neural network model for training, and finishing the training of the neural network model when the training of the neural network model meets the preset condition;
and cascading the trained clustering function and the trained neural network model to obtain an artificial intelligence stage model.
6. The artificial intelligence twin training method based on building information model according to claim 5, wherein the inputting the initial feature vector into a clustering function and training the clustering function to obtain a trained clustering function comprises:
acquiring user drawing data;
inputting the initial characteristic vector into a clustering function to obtain output data;
and training the clustering function according to the user drawing data and the output data to obtain the trained clustering function.
7. The artificial intelligence twin training method based on building information model according to claim 6, wherein the training the clustering function according to the user drawing data and the output data to obtain the trained clustering function comprises:
when the user drawing data and the output data are different, sending prompt information;
acquiring feedback information sent by a user according to the prompt information; wherein the feedback information comprises an error and a correct;
and training the clustering function according to the feedback information to obtain the trained clustering function.
8. The artificial intelligence twin training method based on building information model according to claim 7, wherein the training the clustering function according to the feedback information to obtain a trained clustering function comprises:
when the feedback information is wrong, storing the initial characteristic vector corresponding to the feedback information in a wrong set, and inputting the initial characteristic vector corresponding to the feedback information into a clustering function to obtain intermediate data;
calculating a mean square error value of the intermediate data and the user rendered data;
repeatedly executing the steps of storing the initial characteristic vector corresponding to the feedback information in an error set and inputting the initial characteristic vector corresponding to the feedback information into a clustering function to obtain intermediate data when the feedback information is wrong;
and when the mean square error value is smaller than a preset value or the times of training the clustering function reach the preset times, stopping training to obtain the trained clustering function.
9. An intelligent terminal comprising a memory, and one or more programs, wherein the one or more programs are stored in the memory, and wherein the one or more programs being configured to be executed by the one or more processors comprises instructions for performing the method of any of claims 1-8.
10. A non-transitory computer-readable storage medium, wherein instructions in the storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the method of any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111293215.3A CN113743024B (en) | 2021-11-03 | 2021-11-03 | Artificial intelligence twin type training method based on building information model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111293215.3A CN113743024B (en) | 2021-11-03 | 2021-11-03 | Artificial intelligence twin type training method based on building information model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113743024A true CN113743024A (en) | 2021-12-03 |
| CN113743024B CN113743024B (en) | 2022-02-18 |
Family
ID=78727230
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111293215.3A Active CN113743024B (en) | 2021-11-03 | 2021-11-03 | Artificial intelligence twin type training method based on building information model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113743024B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190163215A1 (en) * | 2017-11-27 | 2019-05-30 | General Electric Company | Building energy modeling tool systems and methods |
| CN110245625A (en) * | 2019-06-19 | 2019-09-17 | 山东浪潮人工智能研究院有限公司 | A kind of field giant panda recognition methods and system based on twin neural network |
| US20200042903A1 (en) * | 2018-08-03 | 2020-02-06 | Samsung Electronics Co., Ltd. | Multi-layered machine learning system to support ensemble learning |
| CN111562769A (en) * | 2019-02-14 | 2020-08-21 | 罗克韦尔自动化技术公司 | AI extension and intelligent model validation for industrial digital twinning |
| WO2021062113A1 (en) * | 2019-09-27 | 2021-04-01 | Dirtt Environmental Solutions, Ltd. | Architectural design software incorporating intelligent data-driven user guidance |
| CN112784130A (en) * | 2021-01-27 | 2021-05-11 | 杭州网易云音乐科技有限公司 | Twin network model training and measuring method, device, medium and equipment |
| CN113065276A (en) * | 2021-03-09 | 2021-07-02 | 北京工业大学 | Intelligent construction method based on digital twins |
| CN113592078A (en) * | 2021-08-09 | 2021-11-02 | 郑州大学 | Deep learning network training method based on artificial intelligence |
-
2021
- 2021-11-03 CN CN202111293215.3A patent/CN113743024B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190163215A1 (en) * | 2017-11-27 | 2019-05-30 | General Electric Company | Building energy modeling tool systems and methods |
| US20200042903A1 (en) * | 2018-08-03 | 2020-02-06 | Samsung Electronics Co., Ltd. | Multi-layered machine learning system to support ensemble learning |
| CN111562769A (en) * | 2019-02-14 | 2020-08-21 | 罗克韦尔自动化技术公司 | AI extension and intelligent model validation for industrial digital twinning |
| CN110245625A (en) * | 2019-06-19 | 2019-09-17 | 山东浪潮人工智能研究院有限公司 | A kind of field giant panda recognition methods and system based on twin neural network |
| WO2021062113A1 (en) * | 2019-09-27 | 2021-04-01 | Dirtt Environmental Solutions, Ltd. | Architectural design software incorporating intelligent data-driven user guidance |
| CN112784130A (en) * | 2021-01-27 | 2021-05-11 | 杭州网易云音乐科技有限公司 | Twin network model training and measuring method, device, medium and equipment |
| CN113065276A (en) * | 2021-03-09 | 2021-07-02 | 北京工业大学 | Intelligent construction method based on digital twins |
| CN113592078A (en) * | 2021-08-09 | 2021-11-02 | 郑州大学 | Deep learning network training method based on artificial intelligence |
Non-Patent Citations (2)
| Title |
|---|
| 李通等: ""基于聚类的无监督神经网络模型"", 《2018中国自动化大会（CAC2018）》 * |
| 栾宇等: ""智慧园区关键技术与解决方案研究"", 《中国移动5G天线产业技术研讨会》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113743024B (en) | 2022-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105786715A (en) | Program static automatic analysis method | |
| TWI708140B (en) | Automatic generation of test layouts for verifying a drc deck | |
| US20220414066A1 (en) | Data management system, management method, and storage medium | |
| JPH1092938A (en) | Layout method, layout apparatus and database | |
| CN110222407A (en) | A kind of fusion method and device of BIM data | |
| CN114882211A (en) | Time sequence data automatic labeling method and device, electronic equipment, medium and product | |
| CN113743024B (en) | Artificial intelligence twin type training method based on building information model | |
| CN114168694A (en) | Road network fusion method and device, computer equipment and storage medium | |
| CN116630990B (en) | RPA flow element path intelligent restoration method and system | |
| CN117573093A (en) | Chip code processing method based on functional characteristics, electronic equipment and medium | |
| CN110704898A (en) | Method for placing connecting node between tiger window model and roof truss girder model and product | |
| CN113987803B (en) | Frequency control method, device, equipment and storage medium based on model correction | |
| CN116342745A (en) | Editing method and device for lane line data, electronic equipment and storage medium | |
| CN113283213B (en) | Circuit design system and method based on machine learning | |
| CN115455899A (en) | Analytic layout method based on graph neural network | |
| CN110704897A (en) | Method for placing connecting node between wall keel model and bottom guide beam model and product | |
| CN110765509B (en) | Method and device for generating main roof hole-opening reinforcing node | |
| CN113033128A (en) | Method for selecting initial value of Newton iteration in circuit simulation | |
| CN110704902A (en) | Method for placing connecting node of combined lintel model and combined column model and product | |
| CN110795811A (en) | MEP pipeline adjusting method and device, computer equipment and storage medium | |
| CN110765513B (en) | Method for placing connecting node of wall keel model and L-shaped top guide beam model and product | |
| CN110765508B (en) | Method and device for generating tiger window gable truss node and storage medium | |
| CN110750817B (en) | Connecting piece placement method, adjacent relation determination method, device, equipment and medium | |
| CN112464324B (en) | Adjacent relation acquisition method, adjacent state set generation method and related products | |
| CN110704900B (en) | Method for placing connection node between keel column model and wall keel model and product |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |