Dam defect processing method based on large language model and related equipment

Abstract

The invention relates to the field of intelligent water and electricity management, in particular to a dam defect processing method based on a large language model and related equipment. Constructing a professional knowledge base based on dam full life cycle data, industry specifications, enterprise systems and scientific and technical literatures; constructing a dam defect analysis model constructed based on Transform on the basis of a professional knowledge base, and training the dam defect analysis model to obtain a trained dam defect analysis model; inputting the obtained dam defect problem into a dam defect analysis model for analysis to obtain an analysis result; and retraining the dam defect analysis model according to an analysis result to realize iterative upgrading. A dam defect analysis model is obtained by constructing a professional knowledge base fusing full-life-cycle data, industry specifications and scientific and technical literatures and training a Transform architecture, and the model can analyze input defect problems in real time; and the traditional manual data consulting time is obviously shortened.

Description

Technical Field

[0001] This invention relates to the field of smart hydropower management, specifically to a method and related equipment for handling dam defects based on a large language model. Background Technology

[0002] With the emergence of various large speech models, such as GPT, significant changes in production methods have been triggered in many fields; the ability to quickly and accurately obtain the required information using large language models has become an unstoppable trend of the times.

[0003] In the field of hydropower, the construction and operation of dams is a complex process, during which various undetected hidden dangers may exist, often leading to a variety of thorny problems. Current methods for dealing with these problems require professionals to spend a significant amount of time searching through massive amounts of data before implementing solutions. This process has two major drawbacks: First, it is too time-consuming, making it difficult to quickly address complex problems. In emergencies, delays in intervention may worsen the problem, causing incalculable losses. Second, the results are often unsatisfactory. Because complex problems often involve multiple factors such as geology, meteorology, and engineering structure, manual data searching is rarely comprehensive. This can easily lead to information gaps for professionals, resulting in biased analysis and judgment, and ultimately, inaccurate and ineffective solutions.

[0004] In order to effectively improve defect management and problem analysis in the dam construction and maintenance process, help practitioners break free from the constraints of traditional models, quickly and accurately obtain the necessary information, and thus improve the safety status of the dam, it is essential to introduce large-scale models. Therefore, it is particularly important to combine language models with dam problems for processing. Summary of the Invention

[0005] To address the problems mentioned in the prior art, this invention proposes a method and related equipment for handling dam defects based on a large language model. To achieve the above objectives, the present invention adopts the following technical solution: This invention discloses a method for handling dam defects based on a large language model, comprising the following steps: A professional knowledge base will be built based on dam life-cycle data, industry standards and corporate regulations, and scientific and technological literature. Based on a professional knowledge base, a dam defect analysis model based on Transformer is constructed, and the dam defect analysis model is trained to obtain a trained dam defect analysis model. The identified defects in the dam are input into the dam defect analysis model for analysis, and the analysis results are obtained. The dam defect analysis model was retrained based on the analysis results to achieve iterative upgrades.

[0006] As a further improvement of the present invention, the dam's full life cycle data includes the feasibility study report, design drawings, and construction technical parameters before the dam's construction. The construction technical parameters include concrete grade, performance parameters, and mixing ratio.

[0007] As a further improvement of the present invention, the industry standards and corporate regulations include comparing dam design indicators with actual monitoring data to determine whether the stress, temperature, and strength exceed the compliance threshold range.

[0008] As a further improvement to this invention, a well-trained dam defect analysis model is obtained, including: The professional knowledge base data is cleaned, segmented, and encoded to obtain the training set. Configure the number of Transformer architecture layers, the number of attention heads, and the dimensions of hidden layers; The training data is input into the dam defect analysis model for training, resulting in a trained dam defect analysis model. The parameters of the dam defect analysis model are then adjusted using hydropower professional data.

[0009] As a further improvement to the present invention, the process of obtaining the analytical results includes: Extract keywords related to the defect issue and search for relevant information in the professional knowledge base; Compare the abnormal data with the threshold values ​​of the dam's design parameters; The reason for sorting by relevance from high to low.

[0010] As a further improvement and iterative upgrade of the present invention, it includes: The feedback on the processing based on the analysis results is categorized as follows: completely resolved, partially resolved, and the measures are ineffective. The feedback results are input into the professional knowledge base, allowing the dam defect analysis model to learn incrementally.

[0011] As a further improvement of the present invention, the sources of the defects are at least one of the following: sudden anomalies identified by the dam monitoring system, existing problems discovered by manual inspections, defects in registration and scheduled inspections, and abnormal technical indicators during production.

[0012] This invention proposes a dam defect processing system based on a large language model, comprising: The module is used to build a professional knowledge base based on dam life-cycle data, industry standards and corporate regulations, and scientific and technological literature. The training module is used to construct a dam defect analysis model based on Transformer based on a professional knowledge base, and to train the dam defect analysis model to obtain a trained dam defect analysis model. The analysis module is used to input the identified dam defects into the dam defect analysis model for analysis and to obtain the analysis results. The upgrade module is used to retrain the dam defect analysis model based on the analysis results to achieve iterative upgrades.

[0013] This invention proposes a dam defect processing device based on a large language model, comprising a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the dam defect processing method based on the large language model as described above.

[0014] This invention proposes a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the dam defect processing method based on a large language model as described above.

[0015] Compared with the prior art, the present invention achieves the following technical effects: This invention constructs a professional knowledge base that integrates full lifecycle data, industry standards, and scientific literature, and trains a Transformer architecture to obtain a dam defect analysis model. This model can analyze the input defect problem in real time. Through keyword retrieval, threshold comparison, and relevance ranking mechanisms, it outputs possible causes and solutions within seconds, significantly shortening the time required for traditional manual data retrieval. Especially for sudden defects, rapid response can prevent the accident from escalating.

[0016] Compared to traditional methods that rely on human experience and easily overlook multi-factor coupling issues, this invention utilizes the powerful contextual association capabilities of Transformer to output a comprehensive diagnostic report by comparing abnormal data with design thresholds and sorting the causes by relevance, thereby reducing misjudgments caused by missing human information. This invention categorizes the processing results into three categories: completely resolved, partially resolved, and ineffective measures, and then feeds them back to the knowledge base, triggering dynamic adjustments to the model parameters. This closed-loop mechanism enables the model to continuously adapt to various defects, thereby avoiding the lag of traditional static knowledge bases. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall process of the present invention; Figure 2 This is a schematic diagram of the dam defect processing system based on a large language model according to the present invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0019] See Figure 1 This invention proposes a method for handling dam defects based on a large language model, comprising the following steps: A professional knowledge base will be built based on dam life-cycle data, industry standards and corporate regulations, and scientific and technological literature. Based on a professional knowledge base, a dam defect analysis model based on Transformer is constructed, and the dam defect analysis model is trained to obtain a trained dam defect analysis model. The identified defects in the dam are input into the dam defect analysis model for analysis, and the analysis results are obtained. The dam defect analysis model was retrained based on the analysis results to achieve iterative upgrades.

[0020] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments: Step 1: This embodiment requires three types of data to build a professional knowledge base: dam life cycle data, industry standards and corporate regulations, and scientific and technological literature. Based on these data, the model can have a comprehensive understanding of the dam. Compared with manual searching, the model can more accurately call up these data. The following is a detailed explanation of these three types of data.

[0021] The full life cycle data of a dam includes all the key information such as design indicators and structural parameters of a specific dam. The main data consists of the feasibility study report before the dam construction, which includes key information such as the dam overview, water temperature, geology, scale, and electromechanical systems; the design drawings, which need to be marked with detailed information such as the distribution of structural stress; and finally, the technical information during construction, such as the concrete grade, performance parameters, mixing ratio, and other material characteristics.

[0022] Industry standards and corporate regulations, such as the provisions in the "Technical Specification for Safety Monitoring of Concrete Dams" and the internal inspection procedures of a certain company, are essential standards that must be followed in the construction of a dam. They serve as the basis for the large language model to measure the actual state of the dam. Through training with this type of information, the large language model can learn to judge whether various parts of the dam are compliant and whether the dam body has returned to normal after the hidden dangers have been dealt with. The principle is to compare the actual state of the dam with the design indicators. Specifically, during the operation of the dam, the values ​​of its stress, temperature, strength, etc. will change, and it is necessary to monitor whether they are within the normal range.

[0023] Scientific and technological literature refers to various high-quality scientific and technological documents, such as defect cases and the latest research results that can be obtained from the market. This type of data can provide descriptions of various hidden dangers of dams, their causes, mechanisms and solutions, and are specific problem-solving methods.

[0024] Once these three types of information are fully compiled, the professional knowledge base of the model is constructed, which is the foundation for the model's operation. In practical applications, by using common language model dialogue organization algorithms and keyword association analysis techniques, logically sound, accurately descriptive, and easy-to-read dialogue text can be organized to achieve information interaction with people.

[0025] Step 2: This embodiment is a dam defect analysis model built based on Transformer. The construction process includes: First, the data in the professional knowledge base is preprocessed. First, invalid or outdated data is cleaned. Then, the cleaned data is standardized and segmented into words using a hydropower professional dictionary. Finally, the construction parameters are converted into high-dimensional vector representations.

[0026] Then, the number of layers is configured according to the model specifications. In this embodiment, an encoder with more than 12 layers is set to enhance semantic understanding capabilities, 16 attention heads to improve the effect of multi-parameter coupling analysis, and 1024-dimensional hidden layer space to accommodate complex engineering features.

[0027] The entire training process is divided into two stages. The first stage is pre-training, which involves establishing basic language comprehension capabilities on general water conservancy engineering texts. The second stage involves optimization and fine-tuning based on specific hydropower data and tasks, such as fine-tuning dam defect records. The training process is iterated and repeated to obtain a well-trained dam defect analysis model.

[0028] The model in this embodiment is continuously trained based on the input questions to improve accuracy. Specifically, based on the actual operation and construction of the dam, defects will be continuously discovered, such as sudden anomalies identified by the dam monitoring system, existing problems found during the dam inspection process, defects found during the dam registration and maintenance phase, and problems reflected when the main technical indicators are monitored abnormally during dam production. Through the concept of this invention, these are integrated into the safety defect input of the large language model, i.e., examples, and handed over to the large model for training and analysis.

[0029] Step 3: When a defect problem is input into the dam defect analysis model, for example, when the problem is an abnormal increase in seepage flow, the model first extracts seepage flow as the core keyword; then it searches for related content in the knowledge base, specifically retrieving the design parameter thresholds for that dam section from the knowledge base, then compares the abnormal value with the design parameter thresholds, and through learning and searching the knowledge base literature, it conducts a preliminary analysis of the possible causes of the problem and provides feasible further checks and solutions; finally, the above methods are transcribed into human language dialogue text and pushed to users through terminals such as computers and mobile phones.

[0030] Therefore, the possible causes of this problem are abnormal seepage flow, which could be due to excessive uplift pressure at a certain point in the dam, the appearance of cracks, instrument malfunctions, etc. All of these causes are possible, listed in descending order of relevance. This way, users will clearly understand the potential location of the problem upon seeing the text, and can thoroughly and accurately confirm it using the given specific methods, then handle it according to the corresponding solutions. In this way, a complete defect handling procedure is formed. Users respond according to the measures, conduct on-site verification, and resolve the problem. The entire process, relying on the parallel computing advantages of Transformer, is completed in seconds compared to traditional manual analysis.

[0031] Step 4: The dam defect analysis model in this embodiment can self-iterate and upgrade. After personnel complete the verification and closed-loop processing of the problem, the result will inevitably be one of the following three: 1. The problem is completely resolved; 2. The problem is partially resolved, and after handling according to the measures given by the large language model, the hidden dangers are repaired and the problem improves, but some anomalies still exist; 3. The handling measures are ineffective, and the problem is actually solved by other means, and the cause of the problem is different from that described in the measures. Regardless of which of the above three results is correct, it is a valid feedback. It has a good guiding role in the analysis of similar anomalies in the future.

[0032] Therefore, in this embodiment, the feedback results are edited into defect handling cases, re-input into the professional knowledge base, and the model is retrained. Cases that are fully resolved are added to the professional knowledge base to strengthen the experience; cases that are partially resolved are labeled with unresolved factors to generate new training data; and cases where the measures are ineffective are corrected by using the professional knowledge base to correct errors in the knowledge base.

[0033] Therefore, when a similar problem arises again, if practitioners turn to the model, the resulting solutions will include the causes and solutions for that specific case. Applying this to real-world situations allows for more accurate judgment and handling. This iterative process makes the large language model increasingly accurate in analyzing similar problems, thus continuously approaching intelligent processing methods. To achieve this, the model in this example uses incremental learning to achieve dynamic evolution. Each update mixes 10% historical data with new cases for training, absorbing new knowledge while avoiding forgetting historical experience, ensuring the model maintains the stability of its core capabilities when adapting to new scenarios.

[0034] This invention, based on the Transformer architecture, allows the model to directly locate relevant knowledge nodes when processing queries, avoiding the massive data sifting required by traditional manual retrieval and resulting in faster response times. Furthermore, model iteration and upgrades are achieved through closed-loop feedback, with each processing result being converted into training data, enabling the model to continuously accumulate engineering practice experience and avoiding the lag in professional knowledge bases. This system employs automated processing, reducing reliance on engineers. Practitioners can obtain professional-level analysis simply by inputting a description of the defects in natural language. Simultaneously, the deployment on devices and storage media supports localized operation, ensuring data security and reducing cloud service costs.

[0035] Based on the same inventive concept, this invention also provides a dam defect processing system based on a large language model. Since the principle of this dam defect processing system based on a large language model is similar to that of the aforementioned dam defect processing method based on a large language model, the implementation of this dam defect processing system based on a large language model can refer to the implementation of the dam defect processing method based on a large language model, and the repeated parts will not be described again.

[0036] In specific implementation, the dam defect processing system based on a large language model provided in this embodiment of the invention specifically includes: The module is used to build a professional knowledge base based on dam life-cycle data, industry standards and corporate regulations, and scientific and technological literature. The training module is used to construct a dam defect analysis model based on Transformer based on a professional knowledge base, and to train the dam defect analysis model to obtain a trained dam defect analysis model. The analysis module is used to input the identified dam defects into the dam defect analysis model for analysis and to obtain the analysis results. The upgrade module is used to retrain the dam defect analysis model based on the analysis results to achieve iterative upgrades. Accordingly, embodiments of the present invention also provide a dam defect processing device based on a large language model, including a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the dam defect processing method based on a large language model as provided in the embodiments of the present invention.

[0037] For more detailed information on the above methods, please refer to the relevant content disclosed in the foregoing embodiments, which will not be repeated here.

[0038] Accordingly, embodiments of the present invention also provide a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the dam defect processing method based on a large language model as described above in embodiments of the present invention.

[0039] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems, devices, and storage media disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0040] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0041] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0042] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0043] The above provides a detailed description of the dam defect processing method, system, equipment, and storage medium based on a large language model provided by this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for handling dam defects based on a large language model, characterized in that, Includes the following steps: A professional knowledge base will be built based on dam life-cycle data, industry standards and corporate regulations, and scientific and technological literature. Based on a professional knowledge base, a dam defect analysis model based on Transformer is constructed, and the dam defect analysis model is trained to obtain a trained dam defect analysis model. The identified defects in the dam are input into the dam defect analysis model for analysis, and the analysis results are obtained. The dam defect analysis model was retrained based on the analysis results to achieve iterative upgrades.

2. The dam defect processing method based on a large language model according to claim 1, characterized in that, The dam's full life-cycle data includes feasibility study reports, design drawings, and construction technical parameters before dam construction; The construction technical parameters include concrete grade, performance parameters, and mixing ratio.

3. The dam defect processing method based on a large language model according to claim 1, characterized in that, The industry standards and corporate regulations include comparing dam design parameters with actual monitoring data to determine whether stress, temperature, and strength exceed compliance thresholds.

4. The dam defect processing method based on a large language model according to claim 1, characterized in that, The trained dam defect analysis model is obtained, including: The professional knowledge base data is cleaned, segmented, and encoded to obtain the training set. Configure the number of Transformer architecture layers, the number of attention heads, and the dimensions of hidden layers; The training data is input into the dam defect analysis model for training, resulting in a trained dam defect analysis model. The parameters of the dam defect analysis model are then adjusted using hydropower professional data.

5. The dam defect processing method based on a large language model according to claim 1, characterized in that, The process of obtaining the analysis results includes: Extract keywords related to the defect issue and search for relevant information in the professional knowledge base; Compare the abnormal data with the threshold values ​​of the dam's design parameters; The reason for sorting by relevance from high to low.

6. The dam defect processing method based on a large language model according to claim 1, characterized in that, Iterative upgrades, including: The feedback on the processing based on the analysis results is categorized as follows: completely resolved, partially resolved, and the measures are ineffective. The feedback results are input into the professional knowledge base, allowing the dam defect analysis model to learn incrementally.

7. The dam defect processing method based on a large language model according to claim 1, characterized in that, The sources of the defects are at least one of the following: sudden anomalies identified by the dam monitoring system, existing problems discovered by manual inspections, defects in registered and scheduled inspections, and abnormal technical indicators during production.

8. A dam defect processing system based on a large language model, characterized in that, include: The module is used to build a professional knowledge base based on dam life-cycle data, industry standards and corporate regulations, and scientific and technological literature. The training module is used to construct a dam defect analysis model based on Transformer based on a professional knowledge base, and to train the dam defect analysis model to obtain a trained dam defect analysis model. The analysis module is used to input the identified dam defects into the dam defect analysis model for analysis and to obtain the analysis results. The upgrade module is used to retrain the dam defect analysis model based on the analysis results to achieve iterative upgrades.

9. A dam defect processing device based on a large language model, characterized in that, It includes a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the dam defect processing method based on a large language model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the dam defect handling method based on a large language model as described in any one of claims 1 to 7.

Images