Method, device, equipment and medium for identifying defects in power transmission and transformation engineering design schemes

Abstract

The present application provides a method, device, equipment and medium for identifying defects in design schemes for power transmission and transformation projects. The method includes obtaining a design drawing data set and a scheme text data set with defect marks set; inputting the design drawing data set and the corresponding defect description information into a graphic encoder and training the graphic encoder to obtain a design drawing defect recognition model, and generating a first feature vector corresponding to the design drawing feature and a second feature vector of the defect description information, and associating the first feature vector with the second feature vector; inputting the scheme text data set and the corresponding defect description information into a scheme defect recognition model and training the scheme defect recognition model to obtain the defect description information; inputting the design drawing to be reviewed into the design drawing defect recognition model to obtain the defect description information of the corresponding design drawing; inputting the scheme text to be reviewed into the scheme defect recognition model to determine the scheme defect description information. The present application automates the identification and intelligent diagnosis of common problems in design schemes for power transmission and transformation projects, greatly improving the efficiency of the review work.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and in particular to a method, device, equipment and medium for identifying defects in a design scheme of a power transmission and transformation project. Background Art

[0002] The current design review process for power transmission and transformation projects is that the review unit forms a review team after receiving the project review task and engineering design related materials. The review team conducts the review work in accordance with the relevant regulations and requirements of the country, industry and company. The scope and depth of the review meet the requirements of the regulations and relevant regulations. Based on the summary of the review issues, the design unit revise the design materials, and then the review unit issues the review opinion.

[0003] However, the review results are often limited by the experts' personal knowledge accumulation and experience judgment, resulting in certain differences in review standards and conclusions among experts, and even possible disagreements. More importantly, in the review process, some common, repeated or frequently occurring problems are often encountered. These problems are typical and representative. They often involve technical solutions for engineering design, unreasonable use of specifications, insufficient design depth and insufficient professional cooperation. However, the review and diagnosis of these problems currently still rely on the manual judgment and item-by-item review of experts, which not only increases the cost of the review and is inefficient, but may also lead to deviations or omissions in the review results due to human factors.

[0004] In view of this, there is an urgent need to provide an intelligent identification and diagnosis method for evaluating whether there are defects in the design scheme of power transmission and transformation projects. Summary of the invention

[0005] In order to overcome the problems existing in the related art, the present disclosure provides a method, device and medium for identifying and diagnosing common problems in power transmission and transformation engineering design schemes, so as to solve the technical problems in the related art that common problems in power transmission and transformation engineering design schemes rely on manual review, which is inefficient and has deviations in the review results.

[0006] One or more embodiments of this specification provide a method for identifying defects in a power transmission and transformation engineering design scheme, including the steps of:

[0007] Obtain a design drawing dataset and a solution text dataset with defect marks set, and establish a defect corpus to store defect corpora and defect description information corresponding to each corpus, so as to describe the defects of each design drawing or solution text;

[0008] Input the design drawing data set and the defect description information corresponding to each design drawing and train the image-text encoder to obtain a design drawing defect recognition model, generate a first feature vector corresponding to the design drawing feature and a second feature vector of the defect description information, and associate the first feature vector with the second feature vector;

[0009] The solution text data set and the defect description information corresponding to each solution text are input into the solution defect recognition model, and the defect description information of the solution text is identified and determined by the solution defect recognition model;

[0010] Inputting the design drawing to be reviewed into the design drawing defect recognition model to obtain an image feature vector, performing similarity calculation between the image feature vector and each first feature vector to obtain the first feature vector with the highest similarity, and determining the defect description information of the corresponding design drawing according to the associated second feature vector; and

[0011] The solution text to be reviewed is input into the solution defect identification model to determine the defects in the solution text and the defect description information.

[0012] One or more embodiments of this specification provide a device for identifying defects in a power transmission and transformation engineering design scheme, including:

[0013] The defect corpus construction module is used to obtain the design drawing data set and the solution text data set with defect marks, and establish a defect corpus to store the defect corpus and the defect description information corresponding to each corpus, so as to describe the defects of each design drawing or solution text;

[0014] The first model training module is used to input the design drawing data set and the defect description information corresponding to each design drawing and train the image-text encoder to obtain a design drawing defect recognition model, generate a first feature vector corresponding to the design drawing feature and a second feature vector of the defect description information, and associate the first feature vector with the second feature vector;

[0015] The second model training module inputs the solution text data set and the defect description information corresponding to each solution text and trains the solution defect recognition model, and identifies and determines the defect description information of the solution text through the solution defect recognition model;

[0016] a first review module, configured to input the design drawing to be reviewed into the design drawing defect recognition model to obtain an image feature vector, perform similarity calculation between the image feature vector and each first feature vector to obtain the first feature vector with the highest similarity, and determine defect description information of the corresponding design drawing according to the associated second feature vector; and

[0017] The second review module is used to input the solution text to be reviewed into the solution defect identification model to determine the defects of the solution text and the defect description information.

[0018] One or more embodiments of the present specification provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for identifying defects in a design scheme for a power transmission and transformation project when executing the computer program.

[0019] One or more embodiments of the present specification provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method for identifying defects in a design scheme for a power transmission and transformation project as described above.

[0020] The present disclosure provides a method, device, equipment and medium for identifying defects in a power transmission and transformation engineering design scheme. The advantages are that a defect corpus corresponding to the defects of design drawings and design schemes commonly seen in the power transmission and transformation engineering design review process is constructed, and the defect corpus is preset with description information of the defects of the design drawings and design schemes, and models for identifying defects in design drawings and design schemes are obtained through training with different corpora respectively, and the association between defect image features and scheme defect features and defect information is learned and established. The model can learn the connection between image and text concepts or between texts, and train a large language model for deep learning, which enables the model to automatically learn effective feature representations from data, and realizes the comprehensive construction of a knowledge base for the power transmission and transformation engineering design review link. The method of this embodiment relies on the setting of a drawing defect recognition model and a scheme defect recognition model, and can realize the review capability of design drawings and / or design schemes to meet more review usage scenarios, realize the automatic recognition and intelligent diagnosis of common diseases in the power transmission and transformation engineering design scheme, greatly improve the efficiency of the review work, and significantly shorten the review cycle, saving a lot of manpower, material resources and time costs for the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 A flow chart of a method for identifying defects in a power transmission and transformation engineering design scheme provided for one or more embodiments of this specification;

[0023] Figure 2 A framework diagram for classifying common defects in power transmission and transformation engineering design by professional category for one or more embodiments of this specification;

[0024] Figure 3A framework diagram for classifying common defects in power transmission and transformation engineering design according to defect categories provided for one or more embodiments of this specification;

[0025] Figure 4 A schematic diagram of a design drawing defect identification diagnosis and question-and-answer interface provided for one or more embodiments of this specification;

[0026] Figure 5 A large language model network architecture diagram provided for one or more embodiments of this specification;

[0027] Figure 6 A block diagram of a device for identifying defects in a power transmission and transformation engineering design scheme provided for one or more embodiments of this specification; and

[0028] Figure 7 A schematic diagram of the structure of a computer device provided for one or more embodiments of this specification. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the protection scope of the present invention.

[0030] The present invention is described in detail below in conjunction with specific implementation methods and the accompanying drawings.

[0031] Method Embodiment

[0032] According to an embodiment of the present invention, a method for identifying defects in a power transmission and transformation engineering design scheme is provided. Figure 1 As shown, it is a flow chart of the method for identifying defects in the design scheme of a power transmission and transformation project provided in this embodiment. The method for identifying defects in the design scheme of a power transmission and transformation project according to the embodiment of the present invention includes:

[0033] Step S1, obtain a design drawing dataset and a solution text dataset with defect marks, and establish a defect corpus (Corpus) to store defect corpuses and defect description information corresponding to each corpus, so as to describe the defects of each design drawing or solution text.

[0034] The defect description information includes defect ID, professional sub-item, defect object, defect problem description, defect type, image / text number, cause of the defect and solution, and other field information.

[0035] Step S2, input the design drawing data set and the defect description information corresponding to each design drawing and train the image-text encoder to obtain a design drawing defect recognition model, and generate a first feature vector corresponding to the design drawing feature and a second feature vector of the defect description information, associate the first feature vector with the second feature vector and store them in a sub-dataset.

[0036] Step S3, inputting the solution text data set and the defect description information corresponding to each solution text into a solution defect recognition model, and identifying and determining the defect description information of the solution text through the solution defect recognition model.

[0037] Step S4, input the design drawing to be reviewed into the design drawing defect recognition model to obtain an image feature vector, calculate the similarity between the image feature vector and each first feature vector to obtain the first feature vector with the highest similarity, and determine the defect description information of the corresponding design drawing based on the associated second feature vector.

[0038] Step S5, inputting the solution text to be reviewed into the solution defect identification model to determine the defects of the solution text and the defect description information.

[0039] The method provided in this embodiment constructs a corresponding defect corpus for the defects of design drawings and design schemes commonly seen in the design review process of power transmission and transformation projects, and the defect corpus presets description information of the defects of design drawings and design schemes, and obtains models for identifying defects of design drawings and design schemes respectively through training with different corpora, learns and establishes the association between defect image features and scheme defect features and defect information, and the model can learn the connection between image and text concepts or between texts, and trains deep learning of a large language model, which enables the model to automatically learn effective feature representations from data, and realizes the comprehensive construction of a knowledge base for the design review link of power transmission and transformation projects. The method in this embodiment relies on the setting of a drawing defect recognition model and a scheme defect recognition model, and can realize the review capability of design drawings and / or design schemes to meet more review usage scenarios, and provide corresponding defect causes and solutions, and the automatic recognition and intelligent diagnosis of common diseases in the design schemes of power transmission and transformation projects greatly improves the efficiency of the review work, and also significantly shortens the review cycle, saving a lot of manpower, material resources and time costs for the enterprise.

[0040] In this embodiment, in order to improve the generalization and adaptability of the model, the graphic encoder will be learned by obtaining new design drawings for setting defect marks and defect description information, or the solution defect recognition model will be learned by obtaining new solution texts for setting defect marks and defect description information, so that it can adapt to and meet the review requirements.

[0041] In this embodiment, the image-text encoder may use a Clip (Contrastive Language-Image Pre-Training) model or a Transformer model, and the initial model of the solution defect recognition model may use a Transformer model.

[0042] In this embodiment, reference Figure 2-3 As shown, Figure 2 The common defects in the design of power transmission and transformation projects provided in this embodiment are divided into a framework diagram according to professional categories. Figure 3 A framework diagram for dividing common defects in power transmission and transformation engineering design according to defect categories, which include insufficient design depth, unreasonable use of specifications, unreasonable design solutions, insufficient professional cooperation, and others. Meanwhile, referring to Table 1 below, a defect corpus list established for this embodiment is provided, in which defect description information is composed of fields such as defect ID, professional sub-item, defect object, defect problem description, defect type, number corresponding to the image / text, cause of the defect, and solution.

[0043] Table 1. List of defective corpora

[0044]

[0045]

[0046] In order to achieve more intelligent solution review and provide reasonable solutions, and further improve the practicality of the method of this embodiment, this embodiment also includes step A1, according to the defect description information of the design drawing obtained in step S4, or the solution defect description information obtained in step S5, and then according to the input question information, determine the intention of the question, and obtain corresponding reply information by matching the information with the defect description information.

[0047] The specific implementation process of the above step A1 includes the following steps:

[0048] Step A11, based on the defect description information identified by the design drawing defect recognition model in step S4, determine the question intent based on the question information, match the information with the field names of the defect description information, and obtain the corresponding field information from the defect description information as the reply information.

[0049] Step A21, based on the defect description information identified by the defect identification model in step S5, determine the question intent based on the question information, match the information with the field names of the defect description information, and obtain the corresponding field information from the defect description information as the reply information.

[0050] For ease of understanding, in a specific embodiment, reference is made to Figure 4As shown, it is a schematic diagram of the design drawing defect identification diagnosis and question-and-answer interface provided in this embodiment. For example, the question information is what problem exists in the solution or the solution to the problem. According to the existing problem or solution in the question information, the "defect problem description" and "cause and solution" in the corresponding extracted defect description information can be determined.

[0051] In step S3 of this embodiment, the large language model used uses a neural network to identify defects in the solution text. The network model reference Figure 5 , the input text sequence is passed and processed through a multi-layer intermediate network to map the input sequence to the output text sequence. The Transformer network uses associations between words at different positions to obtain long-distance dependencies, and uses input text encoded based on specific rotation positions of absolute positions to enhance the context perception of the model. The Transformer network consists of two parts: an encoder and a decoder. Among them, the encoder is responsible for converting the input text sequence into a high-dimensional representation for subsequent processing. The encoder is composed of multiple identical layers stacked together, each of which contains two main sublayers: multi-head attention and feedforward neural network. Multi-head attention establishes an attention relationship between the word at each position and the words at all other positions to capture global context information; the feedforward neural network maps the vector at each position to a higher-dimensional space, which helps the model capture more complex semantic information. Each sublayer is followed by accumulation and layer normalization to stabilize training and accelerate convergence. The encoder fuses the features of the input sequence and adopts a multi-layer stacking structure to facilitate layer-by-layer extraction of the model. The difference between the decoder and the encoder is that masked multi-head attention is also introduced in the decoder to associate the input sequence with the output of the current decoding position in order to generate the desired output sequence.

[0052] In this embodiment, the solution defect recognition model is trained using a pre-training plus fine-tuning learning method. First, pre-training is performed on a large amount of unsupervised corpus (solution text data set and defect description information corresponding to each solution text) to learn the general knowledge of the language model, and then instruction fine-tuning and alignment fine-tuning are performed on a supervised data set of a specific task to adapt to the training data of the specific task.

[0053] Specifically, obtain each solution text and the defect description information corresponding to each solution text, and construct a text sequence S = {S 1 ,S 2 ,...S n}, where the text sequence includes the preceding text S 1 ,S 2 ,...S i-1 and predict the target text S i, the task goal of the Transformer network is to predict the target text, that is, the decoder input, based on the previous text in the sequence, that is, the encoder input, through autoregression. The objective function L(s) of the Transformer network is the maximum likelihood function shown in the following formula:

[0054]

[0055] Where n is the number of texts and P(·) is the likelihood function.

[0056] In this embodiment, the text sequence is S = {design plan, professional sub-item, defect object, number, defect name, problem type, defect problem description, cause and solution}, where {design plan, professional sub-item, defect object, number, defect name, problem type} is the preceding text, and {defect problem description, cause and solution} is the target text to optimize the learning rate, optimizer and training stability of the Transformer network.

[0057] In the fine-tuning stage, the instruction fine-tuning of the solution defect recognition model is implemented through the following steps:

[0058] Collect and construct instructional instances, adopt the instance ratio mixing strategy to fine-tune the initial solution defect recognition model, and increase the sampling ratio of the text sequence data set during the instruction fine-tuning process to obtain the initial solution defect recognition model after instruction fine-tuning.

[0059] In this embodiment, instruction fine-tuning further optimizes the performance of the model, so that the parameters of the model are optimized according to the loss function of the task, so that it can better adapt to the requirements of the specific task. Instruction fine-tuning first collects and constructs instruction instances, and then fine-tunes the parameters of the large language model in a supervised manner. In this embodiment, a formatted instruction instance includes a task description instruction, a pair of input (preface text) and output (target text) and a small number of examples, in order to realize the conversion of natural language processing tasks into natural language instructions, and then put them into the model for training, so that the model can better understand and generate text specific to the field, thereby improving the applicability and accuracy of the model in this field.

[0060] In this embodiment, the alignment and fine-tuning of the solution defect recognition model is specifically achieved through the following steps:

[0061] Construct a supervised dataset containing pre-order text and target text;

[0062] Input each preceding text in the supervision data set into the initial solution defect recognition model after instruction fine-tuning, and obtain multiple output ranking results corresponding to each input;

[0063] Obtaining human feedback data corresponding to a plurality of output texts that meet a preset alignment standard, and training a reward model using a reinforcement learning algorithm and the human feedback data, so that the trained reward model can determine an output ranking result corresponding to the plurality of output texts that meets the preset alignment standard when obtaining a plurality of output texts corresponding to each input;

[0064] The trained reward model is connected to the initial solution defect recognition model after instruction fine-tuning, so that the output content generated by the initial solution defect recognition model after instruction fine-tuning according to the input is input into the trained reward model, and according to the output sorting result corresponding to the output content generated by the trained reward model that meets the preset alignment standard, the parameters of the initial pre-trained model after instruction fine-tuning are optimized to obtain the fine-tuned solution defect recognition model.

[0065] Device Embodiment

[0066] According to an embodiment of the present invention, a device for identifying defects in a power transmission and transformation engineering design scheme is provided. Figure 6 As shown, it is a block diagram of a power transmission and transformation engineering design scheme defect identification device provided in this embodiment. The power transmission and transformation engineering design scheme defect identification device according to the embodiment of the present invention comprises:

[0067] The defect corpus construction module 10 is used to obtain a design drawing dataset and a solution text dataset with defect marks set, and establish a defect corpus (Corpus) to store defect corpuses and defect description information corresponding to each corpus, so as to describe the defects of each design drawing or solution text, wherein the defect description information includes field information such as defect ID, professional sub-item, defect object, defect problem description, defect type, drawing / text number, cause of the defect and solution.

[0068] The first model training module 20 is used to input the design drawing data set and the defect description information corresponding to each design drawing and train the image-text encoder to obtain a design drawing defect recognition model to generate a first feature vector corresponding to the design drawing feature and a second feature vector of the defect description information, associate the first feature vector with the second feature vector and store them in a sub-dataset.

[0069] The second model training module 30 inputs the solution text data set and the defect description information corresponding to each solution text and trains the solution defect recognition model, and determines the defect description information of the solution text through the solution defect recognition model.

[0070] The first review module 40 is used to input the design drawing to be reviewed into the design drawing defect recognition model to obtain an image feature vector, calculate the similarity between the image feature vector and each first feature vector to obtain the first feature vector with the highest similarity, and determine the defect description information of the corresponding design drawing based on the associated second feature vector.

[0071] The second review module 50 is used to input the solution text to be reviewed into the solution defect identification model to determine the defects of the solution text and the defect description information.

[0072] The device provided in this embodiment constructs a defect corpus corresponding to the defects of design drawings and design schemes commonly seen in the design review process of power transmission and transformation projects, and the defect corpus presets descriptive information of the defects of design drawings and design schemes, and obtains models for identifying the defects of design drawings and design schemes respectively through training with different corpora, learns and establishes the association between defect image features and scheme defect features and defect information, and the model can learn the connection between image and text concepts or between texts, and trains deep learning of large language models, which enables the model to automatically learn effective feature representations from data, realizes the comprehensive construction of the knowledge base of the power transmission and transformation project design review link, and can also provide corresponding defect causes and solutions, realizes the automatic identification and intelligent diagnosis of common diseases of power transmission and transformation project design schemes, greatly improves the efficiency of the review work, and significantly shortens the review cycle, saving a lot of manpower, material and time costs for the enterprise.

[0073] In order to improve the generalization and adaptability of the model, this embodiment further provides a model optimization module 60, which is used to learn the graphic encoder through the obtained new design drawings for setting defect marks and defect description information, or to learn the solution defect recognition model through the obtained new solution texts for setting defect marks and defect description information, so that it can adapt to and meet the review requirements.

[0074] In this embodiment, the image-text encoder may be a Clip (Contrastive Language-Image Pre-Training) model, or the initial model of the image-text encoder and the solution defect recognition model may use a Transformer model.

[0075] In this embodiment, in order to achieve more intelligent review and provide reasonable solutions, and further improve the practicality of the method of this embodiment, a question and answer module 70 is also set up, which is used to determine the intention of the question based on the defect description information of the design drawing obtained by the first review module 40, or the defect description information of the solution obtained by the second review module 50, and then based on the input question information, and obtain the corresponding reply information by matching the information with the defect description information.

[0076] In this embodiment, the question-answering module 70 is configured to perform the following steps to implement question-answering:

[0077] Step B11, based on the defect description information identified by the design drawing defect identification model in the first review module 40, determine the question intent based on the question information, match the information with the field names of the defect description information, and obtain the corresponding field information from the defect description information as the reply information.

[0078] Step B21, based on the defect description information identified by the solution defect identification model in the second audit module 50, determine the question intent based on the question information, and match the information with the field names of the defect description information, and obtain the corresponding field information from the defect description information as the reply information.

[0079] In order to improve the recognition accuracy and parameter optimization of the solution defect recognition model, the solution defect recognition model is trained by a pre-training plus fine-tuning learning method. First, pre-training is performed on a large amount of unsupervised corpus (solution text data set and defect description information corresponding to each solution text) to learn the general knowledge of the language model, and then instruction fine-tuning and alignment fine-tuning are performed on the supervised data set of the specific task to adapt to the training data of the specific task. Therefore, an instruction fine-tuning module 80 and an alignment fine-tuning module 90 are also set to perform instruction fine-tuning and alignment fine-tuning on the solution defect recognition model.

[0080] The instruction fine-tuning module 80 is configured to implement instruction fine-tuning of the solution defect identification model through the following steps, specifically including the steps of:

[0081] Collect and construct instructional instances, adopt the instance ratio mixing strategy to fine-tune the initial solution defect recognition model, and increase the sampling ratio of the text sequence data set during the instruction fine-tuning process to obtain the initial solution defect recognition model after instruction fine-tuning.

[0082] The alignment fine-tuning module 90 is configured to implement the alignment fine-tuning of the solution defect recognition model through the following steps, specifically including the steps of:

[0083] Construct a supervised dataset containing pre-order text and target text;

[0084] Input each preceding text in the supervision data set into the initial solution defect recognition model after instruction fine-tuning, and obtain multiple output ranking results corresponding to each input;

[0085] Obtaining human feedback data corresponding to a plurality of output texts that meet a preset alignment standard, and training a reward model using a reinforcement learning algorithm and the human feedback data, so that the trained reward model can determine an output ranking result corresponding to the plurality of output texts that meets the preset alignment standard when obtaining a plurality of output texts corresponding to each input;

[0086] The trained reward model is connected to the initial solution defect recognition model after instruction fine-tuning, so that the output content generated by the initial solution defect recognition model after instruction fine-tuning according to the input is input into the trained reward model, and according to the output sorting result corresponding to the output content generated by the trained reward model that meets the preset alignment standard, the parameters of the initial pre-trained model after instruction fine-tuning are optimized to obtain the fine-tuned solution defect recognition model.

[0087] The embodiment of the present invention is an apparatus embodiment corresponding to the above-mentioned method embodiment. The specific operations of the processing steps of each module can be understood by referring to the description of the method embodiment, which will not be repeated here.

[0088] like Figure 7 As shown, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for identifying defects in the design scheme of a power transmission and transformation project in the above-mentioned embodiment, or, when executed by a processor, implements the method for identifying defects in the design scheme of a power transmission and transformation project in the above-mentioned embodiment.

[0089] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can 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. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0090] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and the contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A method for identifying defects in a power transmission and transformation engineering design scheme, characterized in that Includes steps: Obtain a design drawing dataset and a solution text dataset with defect marks set, and establish a defect corpus to store defect corpora and defect description information corresponding to each corpus, so as to describe the defects of each design drawing or solution text; Input the design drawing data set and the defect description information corresponding to each design drawing and train the image-text encoder to obtain a design drawing defect recognition model, generate a first feature vector corresponding to the design drawing feature and a second feature vector of the defect description information, and associate the first feature vector with the second feature vector; The solution text data set and the defect description information corresponding to each solution text are input into the solution defect recognition model, and the defect description information of the solution text is identified and determined by the solution defect recognition model; Input the design drawing to be reviewed into the design drawing defect recognition model to obtain an image feature vector, perform similarity calculation between the image feature vector and each first feature vector to obtain the first feature vector with the highest similarity, and determine the defect description information of the corresponding design drawing according to the associated second feature vector; as well as The solution text to be reviewed is input into the solution defect identification model to determine the defects in the solution text and the defect description information.

2. The method for identifying defects in a power transmission and transformation engineering design scheme according to claim 1, characterized in that: The defect description information includes field information of defect ID, professional sub-item, defect object, defect problem description, defect type, image / text number, cause of the defect and solution.

3. The method for identifying defects in a power transmission and transformation engineering design scheme according to claim 1, characterized in that: It also includes the question-answering processing steps: Based on the acquired defect description information of the design drawing or the acquired defect description information of the solution, and then based on the input question information, the question intent is determined, and the corresponding reply information is obtained by matching the information with the defect description information.

4. The method for identifying defects in a power transmission and transformation engineering design scheme according to claim 3, characterized in that: The question-answering process specifically includes the following steps: According to the defect description information identified by the design drawing defect recognition model, determine the question intent according to the question information, match the information with the field names of the defect description information, and obtain the corresponding field information from the defect description information as the reply information; or According to the defect description information identified by the solution defect identification model, the question intention is determined according to the question information, and the information is matched with the field names of the defect description information, and the corresponding field information is obtained from the defect description information as the reply information.

5. The method for identifying defects in a power transmission and transformation engineering design scheme according to claim 1, characterized in that: The initial image-text encoder is a Clip model or a Transformer model, and the initial defect recognition model is a Transformer model.

6. The method for identifying defects in a power transmission and transformation engineering design scheme according to claim 5, characterized in that: It also includes instruction fine-tuning and alignment fine-tuning training for the initial scheme defect recognition model.

7. A device for identifying defects in a power transmission and transformation engineering design scheme, characterized in that include: The defect corpus construction module is used to obtain the design drawing data set and the solution text data set with defect marks, and establish a defect corpus to store the defect corpus and the defect description information corresponding to each corpus, so as to describe the defects of each design drawing or solution text; The first model training module is used to input the design drawing data set and the defect description information corresponding to each design drawing and train the image-text encoder to obtain a design drawing defect recognition model, generate a first feature vector corresponding to the design drawing feature and a second feature vector of the defect description information, and associate the first feature vector with the second feature vector; The second model training module inputs the solution text data set and the defect description information corresponding to each solution text and trains the solution defect recognition model, and identifies and determines the defect description information of the solution text through the solution defect recognition model; A first review module is used to input the design drawing to be reviewed into the design drawing defect recognition model to obtain an image feature vector, perform similarity calculation between the image feature vector and each first feature vector to obtain the first feature vector with the highest similarity, and determine the defect description information of the corresponding design drawing according to the associated second feature vector; as well as The second review module is used to input the solution text to be reviewed into the solution defect identification model to determine the defects of the solution text and the defect description information.

8. The power transmission and transformation engineering design scheme defect identification device according to claim 7, characterized in that: A question-and-answer module is also provided for determining the intent of the question based on the acquired defect description information of the design drawing or the acquired defect description information of the solution and the input question information, and performing information matching with the defect description information to obtain corresponding reply information.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for identifying defects in the design scheme of a power transmission and transformation project as described in any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a method for identifying defects in a power transmission and transformation engineering design scheme as claimed in any one of claims 1 to 6 is implemented.

Images