Automatic generation method for building drawing version updating content based on large modelTechnical Field
The invention relates to the technical field of constructional engineering, in particular to a method for automatically generating updated contents of a version of a construction drawing based on a large model.
Background
In process management of construction of a building structure, when one building drawing is subjected to multiple modification drawing and multiple versions appear, it is necessary to compare differences between versions before and after updating, that is, in the latter version, update contents compared with the former version are described by referring to the side of the building drawing in a text description manner. The text description content of the update content compared with the version before and after the update is manually compared, manually described and manually annotated by a plotter according to the building drawings before and after the modification, the possibility of omission is extremely high, and the time and the labor are wasted when the modified content is identified by naked eye comparison.
Disclosure of Invention
The invention aims to provide a large model-based automatic generation method for updating content of a building drawing version, which aims to solve the problems that after the updating of the building drawing version, the updating content is manually described by manually comparing the version before and after the updating of the building drawing version, and the manual annotating efficiency and the accuracy are low.
In order to solve the technical problems, the invention provides a building drawing version updating content automatic generation method based on a large model, which comprises the following steps:
Providing primitive list information of a certain number of groups of building drawings before and after version updating and respective text description information thereof, and distinguishing content text information of the building drawings after version updating of each group relative to the building drawings before version updating as a fine tuning data set of a large language model;
Step two, fine tuning is carried out on the large language model through the fine tuning data set;
inputting primitive list information and respective text description information of the target building drawing before and after the version update into the trimmed large language model, automatically outputting text information of the updated target building drawing relative to the updated content of the target building drawing before the version update through the large language model, and automatically annotating the text information on the target building drawing after the version update.
Further, the method for automatically generating the updated content of the building drawing version based on the large model provided by the invention comprises the following steps of:
Inputting the primitive list and the text description information of the building drawing before and after the version updating of each group in the fine adjustment data set into a large language model, and automatically comparing the primitive list and the text description information of the building drawing before and after the version updating of each group by the large language model and then automatically outputting the difference text information of the primitive list and the text description information of the building drawing before and after the version updating of each group;
And comparing the difference text information of the building drawing before and after each group of version updating with the difference content text information in the corresponding fine-tuning data set through the large language model, and when the difference text information and the difference content text information are inconsistent, fine-tuning the parameter information of the large language model based on the difference content text information in the fine-tuning data set until the difference text information of the building drawing before and after each group of version updating is output by the large language model is consistent with the difference content text information in the corresponding fine-tuning data set, so as to obtain the fine-tuned large language model.
Further, according to the building drawing version updating content automatic generation method based on the large model, the text information of the updating content attached to the target building drawing after version updating is endowed with a striking color.
Further, according to the building drawing version updating content automatic generation method based on the large model, text information of updating content attached to the target building drawing after version updating is assigned to an independent layer.
Further, according to the building drawing version updating content automatic generation method based on the large model, text information of updating content automatically output by the large language model is annotated beside a primitive or on one side of a frame of a target building drawing after version updating.
Compared with the prior art, the invention has the following beneficial effects:
According to the large model-based automatic generation method for the construction drawing version updating content, an artificial intelligent large language model technology is utilized, primitive list information of the construction drawing before and after version updating and respective text description information of the construction drawing before and after version updating are utilized, and distinguishing content text information of the construction drawing after version updating relative to the construction drawing before version updating in each group is used as a fine adjustment data set of the large language model to carry out fine adjustment on the large language model, primitive list information of the target construction drawing before and after version updating and respective text description information of the target construction drawing after version updating are automatically compared through the fine adjustment large language model, text information of updating content of the target construction drawing after version updating relative to the target construction drawing before version updating is automatically output, and text information of the updating content is automatically annotated on the target construction drawing after version updating, so that automatic generation and automatic annotating of the text information of the updating content are realized on the target construction drawing after version updating. Compared with manual comparison, manual description and manual notes, the method has the advantages of being high in efficiency, high in speed, high in accuracy and intelligent.
Drawings
FIG. 1 is a flow chart of a method for automatically generating updated content of a version of a construction drawing based on a large model.
Detailed Description
The advantages and features of the invention will become apparent from the following detailed description of the invention when taken in conjunction with the accompanying drawings. It should be noted that the drawings are in a very simplified form and are all to a non-precise scale, merely for convenience and clarity in aiding in the description of embodiments of the invention.
Referring to fig. 1, an embodiment of the present invention provides a method for automatically generating a building drawing version update content based on a large model, which may include the following steps:
Providing primitive list information of a certain number of groups of building drawings before and after version updating and respective text description information thereof, and distinguishing content text information of the building drawings after version updating of each group relative to the building drawings before version updating as a fine tuning data set of a large language model. Wherein the large language model may select a GPT or other transform-based large language model. The text description information here includes, but is not limited to, information of relative position relation, connection relation, filling, block reference, text label, axis net, color, layer of the building drawing and the like of each building component in the primitive information of the building drawing. Including but not limited to walls, beams, panels, columns, etc. Some example text descriptions may be "structural plates at the axle net 36 to the axle net 37, the elevation of the structural plates being 12.80m", "a circular filling at the (100,123) position in the drawing, etc. The primitive list information, the text description information, and the differentiated content text information here serve as input contents of the large language model.
And step two, fine tuning is carried out on the large language model through the fine tuning data set. The method comprises the steps of inputting a primitive list and text description information of a building drawing before and after version updating of each group in a fine-tuning data set into a large language model, automatically comparing the primitive list and the text description information of the building drawing before and after version updating of each group by the large language model, and automatically outputting difference text information of the primitive list and the text description information of the building drawing before and after version updating of each group, comparing the difference text information of the building drawing before and after version updating of each group with corresponding difference text information in the fine-tuning data set by the large language model, and when the difference text information is inconsistent with the difference text information in the fine-tuning data set, fine-tuning parameter information of the large language model by taking the difference text information in the fine-tuning data set as a basis until the difference text information of the building drawing before and after version updating of each group is output by the large language model is consistent with the corresponding difference text information in the fine-tuning data set, and obtaining the fine-tuned large language model. An example of the automatic comparison output text may be "structural beams exist from the axis 15 to the axis 16 in the text 1, and no structural beams exist in the text 2", and an example of the two comparison result data may be "structural slab elevation from the axis 36 to the axis 37 in the text 2 is reduced by 300mm from the axis 37 in the text 1", where the comparison result is updated content. In order to avoid disordered version sequence of the building drawings updated by each group of versions, the building drawings before and after version update can be automatically numbered or prompting guide words such as 'drawing 1 is the previous version' and 'drawing 2 is the latter version' are added.
Inputting primitive list information and respective text description information of the target building drawing before and after the version update into the trimmed large language model, automatically outputting text information of the updated target building drawing relative to the updated content of the target building drawing before the version update through the large language model, and automatically annotating the text information on the target building drawing after the version update. And automatically obtaining an update log of the target building drawing after version update relative to the target building drawing before version update through the trimmed large language model. An example version update log output text may be "structural beam cancellation at shaft web 15 through shaft web 16".
According to the large model-based automatic generation method for the construction drawing version updating content, which is provided by the embodiment of the invention, by utilizing an artificial intelligence large language model technology, the large language model is finely adjusted by using a plurality of groups of primitive list information of construction drawings before and after version updating and respective text description information thereof and distinguishing content text information of the construction drawings after version updating relative to the construction drawings before version updating as a fine adjustment data set of the large language model, primitive list information of target construction drawings before and after version updating and respective text description information thereof are automatically compared through the fine adjustment large language model, text information of updating content of the target construction drawings after version updating relative to the target construction drawings before version updating is automatically output, and text information of updating content is automatically annotated on the target construction drawings after version updating, so that automatic generation and automatic annotating of the text information of the updating content on the target construction drawings after version updating are realized. Compared with manual comparison, manual description and manual notes, the method has the advantages of being high in efficiency, high in speed, high in accuracy and intelligent.
In order to facilitate the identification and viewing of updated contents, the method for automatically generating the updated contents of the version of the building drawing based on the large model provided by the embodiment of the invention gives the text information of the updated contents attached to the updated version of the target building drawing a striking color.
In order to facilitate later editing and management, the method for automatically generating the version update content of the building drawing based on the large model, provided by the embodiment of the invention, assigns text information of the update content, which is annotated on the target building drawing after version update, on a single layer.
In order to avoid influencing the readability of the primitive information of the target building drawing after version updating, the method for automatically generating the building drawing version updating content based on the large model provided by the embodiment of the invention is used for automatically outputting the text information of the updating content of the target building drawing after version updating through the large language model to be attached to the side of the primitive of the target building drawing after version updating or one side of a drawing frame.
The method for automatically generating the updated content of the building drawing version based on the large model can avoid the problem of omission caused by negligence of manual comparison, avoid visual fatigue caused by manual comparison, and has the advantages of time saving and labor saving.
The present invention is not limited to the above-described embodiments, but rather, the above-described embodiments are merely examples of some, but not all embodiments of the present invention. All other embodiments, which are obtained by a person skilled in the art based on the described embodiments of the invention, fall within the scope of protection of the invention. Other levels of modification and variation to the present invention may occur to those skilled in the art. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims.