Quality inspection management risk analysis and feedback method and device

Abstract

The application discloses a kind of quality inspection management risk analysis and feedback method and device, the method includes: constructing multi-source heterogeneous railway engineering quality inspection business file database, and the multiple-format files in it are preprocessed, generate standardization structured data set;Based on standardization structured data set and railway engineering specification and historical case construction training knowledge base, based on training knowledge base and the target AI big model of pre-set multi-modal big model;The business file after pre-processing is input into target AI big model and carries out multidimensional logic check, obtains multidimensional logic check result;Based on multidimensional logic check result, risk information is identified and risk level is evaluated, and the development trend of risk is predicted in linkage target AI big model, and targeted disposal suggestion is generated, to realize quality inspection audit process intelligentization, risk front prewarning and closed-loop management and control, improve railway engineering quality inspection management audit precision and management efficiency.

Description

Technical Field

[0001] This application relates to the field of engineering management technology, and more specifically, to a method and apparatus for risk analysis and feedback in quality inspection management. Background Technology

[0002] In railway engineering construction, quality inspection management is a key link to ensure the quality and durability of the project. The current quality inspection management of railway engineering has the following problems: (1) The current quality inspection management has quality management problems such as using raw materials before inspection and unqualified inspection. The review cycle of test and inspection reports is long, the quality problems are not handled in a timely manner, and the transmission and handling of unqualified results lack closed-loop management, which may lead to the accumulation of hidden dangers in the project quality; (2) The business document review dimension is single, and it can only check the surface information of the document. It is difficult to achieve a comprehensive verification of the integrity of the form, the compliance of the data, and the correlation of the document. It is impossible to accurately identify deep-seated problems such as the false test data and the logical contradictions in the document; (3) The ability to predict quality risks is insufficient. It relies on manual experience for risk assessment and it is difficult to discover potential quality defects in advance based on massive business data, resulting in poor accuracy and timeliness of risk warning. Summary of the Invention

[0003] This invention provides a method and apparatus for risk analysis and feedback in quality inspection management, to solve the technical problems of the prior art as described in the background section. The method includes: A multi-source heterogeneous railway engineering quality inspection business document database is constructed, and the multi-format files in the multi-source heterogeneous railway engineering quality inspection business document database are preprocessed to generate a standardized structured dataset. A training knowledge base is constructed based on the standardized structured dataset, railway engineering specifications, and historical cases. The target AI large model is obtained based on the training knowledge base and the preset multimodal large model. The preprocessed business file is input into the target AI model for multi-dimensional logical verification to obtain the multi-dimensional logical verification results. Based on the multi-dimensional logical verification results, risk information is identified and the risk level is assessed. The target AI big model is then linked to predict the risk development trend and generate targeted handling suggestions.

[0004] In some specific embodiments, a multi-source heterogeneous railway engineering quality inspection business document database is constructed, including: Collect testing and inspection business documents and process management documents, and construct a multi-source heterogeneous railway engineering quality inspection business document database based on the testing and inspection business documents and process management documents.

[0005] In some specific embodiments, the multi-format files in the multi-source heterogeneous railway engineering quality inspection business document database are preprocessed to generate a standardized structured dataset, specifically as follows: Data cleaning is performed on files of various formats. The data cleaning process includes deduplication of redundant data, filtering of noisy data, and labeling of missing data. Using deep learning-based OCR technology, PDF scans from various file formats are converted into editable text; The program parses and encapsulates the format of Word documents, extracts text from image files and standardizes their format, resulting in standardized files. Image detection algorithms are used to locate image regions in image files, extract the images and generate URLs, parse the visual information in the images, and retain the contextual annotations of the images. Based on the railway engineering quality inspection specifications, the specifications are classified and sorted according to the railway engineering professional categories. The inspection technical indicators, judgment standards and related parameters of each professional category are extracted. Combined with the key node information and data format requirements of different inspection links, the specification clauses are transformed into extraction rules to build a key field extraction rule library. A standardized structured dataset is constructed based on the editable text, normalized files, and extracted images by combining named entity recognition and rule matching.

[0006] In some specific embodiments, a training knowledge base is constructed based on the standardized structured dataset and railway engineering specifications and historical cases, specifically as follows: The standardized structured dataset is used as the training sample input to a pre-set large AI model based on the Transformer architecture, and a training knowledge base is constructed by combining relevant standards, historical problem cases, and risk level classification rules related to railway engineering quality inspection.

[0007] In some specific embodiments, the target AI large model is obtained based on the training knowledge base and the preset multimodal large model, specifically as follows: Based on a multimodal large model, structured numerical data is transformed into text descriptions and input into a text encoder along with the canonical text in the knowledge base. Image features are extracted using a visual encoder and mapped to a text embedding space to achieve alignment between visual and text features. Adversarial training techniques are employed to obtain diverse training data, and the multimodal large model is optimized through fine-tuning training methods. This enables the multimodal large model to learn the relationship between perturbation data and quality inspection standards and specifications, thereby gradually improving its adaptability and judgment accuracy for different quality inspection tasks. By simulating changes in the field environment, enhanced image data is generated, and the defect recognition ability of the multimodal large model under complex visual conditions is trained, so that the multimodal large model can enhance its adaptability to data fluctuations and uncertainties in the field environment during the learning process. The target AI large model is obtained.

[0008] In some specific embodiments, the preprocessed business file is input into the target AI model for multi-dimensional logical verification to obtain the multi-dimensional logical verification result, specifically: The form integrity of the preprocessed business file is verified based on the target AI large model, and a first verification result is obtained; The data compliance of the preprocessed business file is verified based on the target AI large model, and a second verification result is obtained; Based on the target AI large model, the logical correlation and data consistency between different preprocessed business files are verified to obtain a third verification result; The first verification result, the second verification result, and the third verification result are used as the multi-dimensional logical verification result.

[0009] In some specific embodiments, risk information is identified and the risk level is assessed based on the multi-dimensional logical verification results. The target AI model is then used to predict risk development trends and generate targeted response suggestions, specifically: Based on the multi-dimensional logical verification results and combined with historical problem cases and risk level classification rules in the training knowledge base, risk information is identified, the risk information is classified and the risk level of the risk information is evaluated. The classification results include form-type problems, data-type problems and correlation-type problems, and the risk levels include high risk, medium risk and low risk. Based on the target AI big model, the development trend of the risk information is predicted, and the prediction result is obtained; Based on the prediction results, classification results, and risk levels, risk information will be pushed to the corresponding responsible parties and corresponding handling opinions will be generated.

[0010] Accordingly, the present invention also proposes a quality inspection management risk analysis and feedback device, the device comprising: The preprocessing module is used to construct a multi-source heterogeneous railway engineering quality inspection business document database, and to preprocess the multi-format files in the multi-source heterogeneous railway engineering quality inspection business document database to generate a standardized structured dataset. The construction module is used to build a training knowledge base based on the standardized structured dataset, railway engineering specifications and historical cases, and to obtain the target AI large model based on the training knowledge base and the preset multimodal large model. The verification module is used to input the preprocessed business file into the target AI model for multi-dimensional logical verification and obtain the multi-dimensional logical verification results. The identification module is used to identify risk information and assess risk level based on the multi-dimensional logical verification results, link the target AI big model to predict the risk development trend, and generate targeted handling suggestions.

[0011] One embodiment of the present invention also provides a computing device, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, wherein when the computer-executable instructions are executed by the processor, the steps of the quality inspection management risk analysis and feedback method as described above are implemented.

[0012] One embodiment of the present invention also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the quality inspection management risk analysis and feedback method as described in any of the above claims.

[0013] By applying the above technical solutions, a risk analysis and feedback method for quality inspection management is proposed. The method includes: constructing a multi-source heterogeneous railway engineering quality inspection business document database; preprocessing the multi-format files in the database to generate a standardized structured dataset; constructing a training knowledge base based on the standardized structured dataset, railway engineering specifications, and historical cases; obtaining a target AI model based on the training knowledge base and a pre-set multimodal large model; inputting the preprocessed business documents into the target AI model for multi-dimensional logical verification to obtain multi-dimensional logical verification results; identifying risk information and assessing risk levels based on the multi-dimensional logical verification results; linking the target AI model to predict risk development trends and generate targeted handling suggestions; and using the AI ​​model to perform comprehensive verification of business documents to ensure their quality and improve the efficiency of business review and risk assessment. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of a quality inspection management risk analysis and feedback method provided in an embodiment of this application; Figure 2This is a schematic diagram of the structure of a quality inspection management risk analysis and feedback device provided in an embodiment of this application; Figure 3 This is a structural block diagram of a computing device provided in an embodiment of this application. Detailed Implementation

[0016] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0017] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0018] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0019] In response to the problems existing in the quality inspection management of railway engineering construction, such as using raw materials before inspection, non-closed-loop handling of unqualified tests, insufficient intelligent review of business documents, and low accuracy of risk warning, this invention provides a risk analysis and feedback method for quality inspection management. This method enables automated and comprehensive logical verification of quality inspection business documents, accurately identifies problems, predicts risks, and constructs a risk closed-loop handling mechanism to improve the level of railway engineering quality inspection management. Specifically, it includes: (1) a business document data collection and preprocessing module, which collects multi-source engineering quality inspection management business documents, performs standardization conversion and cleaning preprocessing, and generates a structured dataset; (2) an AI large model training and optimization module, which constructs a Transformer architecture AI model, combines standard and case training and optimization, and obtains a railway quality inspection-specific model; (3) a business document logical verification module, which inputs the preprocessed documents into the model and performs multi-dimensional verification of form completeness, data compliance, and document association; (4) a quality risk identification, prediction and closed-loop handling module, which accurately identifies quality problems such as using raw materials before inspection and unqualified quality inspection, pushes risk warnings in a targeted manner, generates handling plans, tracks progress and verifies results, and realizes closed-loop management.

[0020] Specifically, such as Figure 1 As shown, this application proposes a risk analysis and feedback method for quality inspection management, the method comprising the following steps: Step S101: Construct a multi-source heterogeneous railway engineering quality inspection business document database, and preprocess the multi-format files in the multi-source heterogeneous railway engineering quality inspection business document database to generate a standardized structured dataset.

[0021] In one possible implementation, a multi-source heterogeneous railway engineering quality inspection business document database is constructed, including: Collect testing and inspection business documents and process management documents, and construct a multi-source heterogeneous railway engineering quality inspection business document database based on the testing and inspection business documents and process management documents.

[0022] In one possible implementation, the multi-format files in the multi-source heterogeneous railway engineering quality inspection business document database are preprocessed to generate a standardized structured dataset, specifically as follows: Data cleaning is performed on files of various formats. The data cleaning process includes deduplication of redundant data, filtering of noisy data, and labeling of missing data. Using deep learning-based OCR technology, PDF scans from various file formats are converted into editable text; The program parses and encapsulates the format of Word documents, extracts text from image files and standardizes their format, resulting in standardized files. Image detection algorithms are used to locate image regions in image files, extract the images and generate URLs, parse the visual information in the images, and retain the contextual annotations of the images. Based on the railway engineering quality inspection specifications, the specifications are classified and sorted according to the railway engineering professional categories. The inspection technical indicators, judgment standards and related parameters of each professional category are extracted. Combined with the key node information and data format requirements of different inspection links, the specification clauses are transformed into extraction rules to build a key field extraction rule library. A standardized structured dataset is constructed based on the editable text, normalized files, and extracted images by combining named entity recognition and rule matching.

[0023] Specifically, a multi-source heterogeneous database of railway engineering quality inspection business documents will be constructed. The scope of data collection will be clearly defined, including inspection and testing documents such as raw material arrival inspection, semi-finished product testing, and appearance quality documents, as well as process management documents such as construction record sheets, construction logs, inspection batches, and images of concealed works. Preprocessing will be performed on the collected multi-format documents (inspection and testing business documents and process management business documents, covering tables, Word documents, standard specifications PDFs, images, etc.). For PDF scans (mainly quality inspection management-related standard specifications documents – usually in PDF format), deep learning-based OCR technology will be used to convert them into editable text. Word documents such as inspection reports, inspection batch tables, and construction record sheets will undergo format parsing and unified encapsulation. Image files will undergo text extraction and format standardization. Image detection algorithms are used to locate image regions in image files, extract the images and generate URLs, and parse the visual information in the images while retaining contextual annotations. Data cleaning is performed through methods such as redundancy removal, noise filtering, and missing data annotation. Based on railway engineering quality inspection standards, the data is categorized and sorted according to typical railway engineering professional categories such as bridges, tunnels, tracks, and roadbeds. Inspection technical indicators, judgment standards, and related parameters for each specialty are extracted. Combined with the inspection workflow, key node information and data format requirements for different inspection stages such as on-site acceptance, process inspection, and final acceptance are sorted out, and the standard clauses are transformed into extraction rules to build a key field extraction rule library. A standardized structured dataset is constructed based on the editable text, standardized documents, and extracted images using a combination of named entity recognition and rule matching.

[0024] Step S102: Construct a training knowledge base based on the standardized structured dataset, railway engineering specifications, and historical cases; and obtain the target AI large model based on the training knowledge base and the preset multimodal large model.

[0025] In one possible implementation, a training knowledge base is constructed based on the standardized structured dataset and railway engineering specifications and historical cases, specifically as follows: The standardized structured dataset is used as the training sample input to a pre-set large AI model based on the Transformer architecture, and a training knowledge base is constructed by combining relevant standards, historical problem cases, and risk level classification rules related to railway engineering quality inspection.

[0026] In one possible implementation, the target AI large model is obtained based on the training knowledge base and a preset multimodal large model, specifically as follows: Based on a multimodal large model, structured numerical data is transformed into text descriptions and input into a text encoder along with the canonical text in the knowledge base. Image features are extracted using a visual encoder and mapped to a text embedding space to achieve alignment between visual and text features. Adversarial training techniques are employed to obtain diverse training data, and the multimodal large model is optimized through fine-tuning training methods. This enables the multimodal large model to learn the relationship between perturbation data and quality inspection standards and specifications, thereby gradually improving its adaptability and judgment accuracy for different quality inspection tasks. By simulating changes in the field environment, enhanced image data is generated, and the defect recognition ability of the multimodal large model under complex visual conditions is trained, so that the multimodal large model can enhance its adaptability to data fluctuations and uncertainties in the field environment during the learning process. The target AI large model is obtained.

[0027] In this embodiment, a large-scale AI model based on the Transformer architecture is constructed. A preprocessed, standardized structured dataset is used as training samples, and a training knowledge base is built by combining relevant standards, historical case studies, and risk level classification rules related to railway engineering quality inspection. The Qwen3 multimodal model is used to convert structured numerical data (such as detection indicators) into text descriptions, which are then input into a text encoder along with the standard text from the knowledge base. Image features are extracted through a visual encoder and mapped to the text embedding space to align visual and text features. Adversarial training techniques are employed, actively introducing perturbations into the data to obtain diverse training data. The Qwen3 model is further optimized through fine-tuning of the training method. By learning the relationship between perturbation data and quality inspection standards, the model gradually improves its adaptability and accuracy for different quality inspection tasks. Simultaneously, by simulating changes in the on-site environment, enhanced image data is generated, enabling the model to accurately identify defects even under complex visual conditions. This enhances the model's adaptability to data fluctuations and uncertainties in the on-site environment during the learning process, ensuring the accuracy and reliability of quality inspection results. This results in a dedicated large-scale AI model suitable for risk analysis in railway engineering quality inspection.

[0028] Step S103: Input the preprocessed business file into the target AI large model for multi-dimensional logical verification to obtain the multi-dimensional logical verification result.

[0029] In one possible implementation, the preprocessed business file is input into the target AI large model for multi-dimensional logical verification to obtain the multi-dimensional logical verification result, specifically: The form integrity of the preprocessed business file is verified based on the target AI large model, and a first verification result is obtained; The data compliance of the preprocessed business file is verified based on the target AI large model, and a second verification result is obtained; Based on the target AI large model, the logical correlation and data consistency between different preprocessed business files are verified to obtain a third verification result; The first verification result, the second verification result, and the third verification result are used as the multi-dimensional logical verification result.

[0030] In this embodiment, the preprocessed business documents are input into the trained AI model for multi-dimensional logical verification. Verification includes checking the completeness of the forms, verifying whether the documents contain all required key fields, and checking for missing fields or non-standard filling. Data compliance is also verified, ensuring that the data submitted in the documents meets relevant standards and specifications, including compliance with data range, accuracy, and format, while identifying potential issues such as abnormal data fluctuations and data falsification. Finally, the logical relationships and data consistency between different business documents are verified, including automatic correlation verification between inspection batches (indicators such as 28-day strength values ​​of entities, earthwork backfill compaction, and rebar processing data) and test reports (indicators such as representative strength values, on-site measured dry density, and rebar bending test records), and correlation between inspection batches (indicators such as construction time and acceptance location) and construction record sheets (indicators such as construction record time and construction location).

[0031] Step S104: Based on the multi-dimensional logical verification results, identify risk information and assess the risk level, link the target AI big model to predict the risk development trend, and generate targeted handling suggestions.

[0032] In one possible implementation, risk information is identified and the risk level is assessed based on the multi-dimensional logical verification results. The target AI model is then used to predict risk development trends and generate targeted response suggestions, specifically: Based on the multi-dimensional logical verification results and combined with historical problem cases and risk level classification rules in the training knowledge base, risk information is identified, the risk information is classified and the risk level of the risk information is evaluated. The classification results include form-type problems, data-type problems and correlation-type problems, and the risk levels include high risk, medium risk and low risk. Based on the target AI big model, the development trend of the risk information is predicted, and the prediction result is obtained; Based on the prediction results, classification results, and risk levels, risk information will be pushed to the corresponding responsible parties and corresponding handling opinions will be generated.

[0033] In this embodiment, quality risk identification and prediction are carried out based on multi-dimensional logical verification results, combined with historical problem cases and risk level classification rules in the training knowledge base. Problems discovered during verification are categorized and labeled, including form-related problems (missing, non-standard), data-related problems (non-compliance, anomalies, data falsification), and correlation-related problems (logical contradictions, unreasonable timing, untested raw materials used before inspection, untreated unqualified tests, etc.). Based on the severity, scope of impact, and probability of occurrence, quality risks are divided into high, medium, and low levels. Leveraging the time-series analysis capabilities of the AI ​​large-scale model, the temporal correlation characteristics of multi-dimensional business data within continuous time periods are deeply mined, performing feature extraction, correlation pattern mining, and trend fitting. Intelligent algorithms are used to deduce the evolution patterns and development trends of various quality problems, predicting the development trends of low, medium, and high-level risks. Based on continuous time-series business data, risk development trends are predicted, and intelligent early warning technology based on the model achieves graded and accurate advance warnings, providing early warning of potential hazards. Identified risk information is automatically pushed to the corresponding responsible parties, and targeted handling suggestions are generated for high and medium-risk problems based on problem type, risk level, and historical handling cases. By tracking the progress of the handling in real time and recording the rectification measures and results, the AI ​​big data model will conduct a second verification of the results after the responsible party provides the handling materials. Once the problem is confirmed to be resolved, a complete risk closed-loop management will be achieved.

[0034] Compared with the prior art, the present invention has the following significant advantages: (1) Accurately identify key quality risks: For typical problems such as using raw materials without inspection, unqualified tests, and inaccurate test data, accurate identification can be achieved through document correlation verification and data compliance analysis, so as to discover potential quality defects in advance and improve the accuracy and timeliness of risk warning.

[0035] (2) Construct a closed-loop risk management system: form a closed-loop management system of "problem identification - risk warning - handling tracking - result verification - archiving optimization" to ensure that quality problems are resolved in a timely and effective manner, avoid risk accumulation, and protect the quality of railway engineering.

[0036] (3) Model adaptive optimization: Through continuous iterative training, the AI ​​large model can adapt to the dynamic changes in railway engineering quality inspection and management, and continuously improve the level of intelligent control.

[0037] In summary, this invention proposes a method for risk analysis and feedback in quality inspection management. The method includes: constructing a multi-source heterogeneous railway engineering quality inspection business document database; preprocessing the multi-format files in the database to generate a standardized structured dataset; constructing a training knowledge base based on the standardized structured dataset, railway engineering specifications, and historical cases; obtaining a target AI model based on the training knowledge base and a pre-set multimodal large model; inputting the preprocessed business documents into the target AI model for multi-dimensional logical verification to obtain multi-dimensional logical verification results; identifying risk information and assessing risk levels based on the multi-dimensional logical verification results; linking the target AI model to predict risk development trends and generate targeted handling suggestions; and using the AI ​​model to perform comprehensive verification of business documents to ensure their quality and improve the efficiency of business review and risk assessment.

[0038] This application also proposes a quality inspection management risk analysis and feedback device, such as... Figure 2 As shown, the device includes: The preprocessing module 10 is used to construct a multi-source heterogeneous railway engineering quality inspection business document database, and to preprocess the multi-format files in the multi-source heterogeneous railway engineering quality inspection business document database to generate a standardized structured dataset. Module 20 is used to build a training knowledge base based on the standardized structured dataset, railway engineering specifications and historical cases, and to obtain the target AI large model based on the training knowledge base and the preset multimodal large model. The verification module 30 is used to input the preprocessed business file into the target AI large model for multi-dimensional logical verification and obtain multi-dimensional logical verification results. The identification module 40 is used to identify risk information and assess risk level based on the multi-dimensional logical verification results, link the target AI big model to predict the risk development trend, and generate targeted handling suggestions.

[0039] Figure 3 A structural block diagram of a computing device 400 according to one embodiment of this specification is shown. The components of the computing device 400 include, but are not limited to, a memory 410 and a processor 420. The processor 420 is connected to the memory 410 via a bus 430, and a database 450 is used to store data.

[0040] The computing device 400 also includes an access device 440, which enables the computing device 400 to communicate via one or more networks 460. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 440 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0041] In one embodiment of this specification, the aforementioned components of the computing device 400 and Figure 3 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 3 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0042] The computing device 400 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 400 can also be a mobile or stationary server.

[0043] The processor 420 is configured to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the aforementioned quality inspection management risk analysis and feedback method. The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the aforementioned quality inspection management risk analysis and feedback method belong to the same concept. Details not described in detail in the technical solution of the computing device can be found in the description of the technical solution of the aforementioned quality inspection management risk analysis and feedback method.

[0044] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described quality inspection management risk analysis and feedback method.

[0045] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the technical solution of the aforementioned quality inspection management risk analysis and feedback method. Details not described in detail in the technical solution of the storage medium can be found in the description of the technical solution of the aforementioned quality inspection management risk analysis and feedback method.

[0046] An embodiment of this specification also provides a computer program, wherein when the computer program is executed in a computer, it causes the computer to perform the steps of the above-described quality inspection management risk analysis and feedback method.

[0047] The above is an illustrative example of a computer program according to this embodiment. It should be noted that the technical solution of this computer program and the technical solution of the aforementioned quality inspection management risk analysis and feedback method belong to the same concept. Details not described in detail in the computer program's technical solution can be found in the description of the technical solution of the aforementioned quality inspection management risk analysis and feedback method.

[0048] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0049] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content contained in the computer-readable medium may be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0050] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0051] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0052] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A method for risk analysis and feedback in quality inspection management, characterized in that, The method includes: A multi-source heterogeneous railway engineering quality inspection business document database is constructed, and the multi-format files in the multi-source heterogeneous railway engineering quality inspection business document database are preprocessed to generate a standardized structured dataset. A training knowledge base is constructed based on the standardized structured dataset, railway engineering specifications, and historical cases. The target AI large model is obtained based on the training knowledge base and the preset multimodal large model. The preprocessed business file is input into the target AI model for multi-dimensional logical verification to obtain the multi-dimensional logical verification results. Based on the multi-dimensional logical verification results, risk information is identified and the risk level is assessed. The target AI big model is then linked to predict the risk development trend and generate targeted handling suggestions.

2. The method according to claim 1, characterized in that, Construct a multi-source heterogeneous railway engineering quality inspection business document database, including: Collect testing and inspection documents and process acceptance documents, and construct a multi-source heterogeneous railway engineering quality inspection document database based on the testing and inspection documents and process acceptance documents.

3. The method according to claim 2, characterized in that, The multi-format files in the multi-source heterogeneous railway engineering quality inspection business document database are preprocessed to generate a standardized structured dataset, specifically as follows: Data cleaning is performed on files of various formats. The data cleaning process includes deduplication of redundant data, filtering of noisy data, and labeling of missing data. Using deep learning-based OCR technology, PDF scans from various file formats are converted into editable text; The program parses and encapsulates the format of Word documents, extracts text from image files and standardizes their format, resulting in standardized files. Image detection algorithms are used to locate image regions in image files, extract the images and generate URLs, parse the visual information in the images, and retain the contextual annotations of the images. Based on the railway engineering quality inspection specifications, the specifications are classified and sorted according to the railway engineering professional categories. The inspection technical indicators, judgment standards and related parameters of each professional category are extracted. Combined with the key node information and data format requirements of different inspection links, the specification clauses are transformed into extraction rules to build a key field extraction rule library. A standardized structured dataset is constructed based on the editable text, normalized files, and extracted images by combining named entity recognition and rule matching.

4. The method according to claim 1, characterized in that, A training knowledge base is constructed based on the aforementioned standardized structured dataset and railway engineering specifications and historical cases, specifically as follows: The standardized structured dataset is used as the training sample input to a pre-set large AI model based on the Transformer architecture, and a training knowledge base is constructed by combining relevant standards, historical problem cases, and risk level classification rules related to railway engineering quality inspection.

5. The method according to claim 1, characterized in that, The target AI large model is obtained based on the training knowledge base and the preset multimodal large model, specifically as follows: Based on a multimodal large model, structured numerical data is transformed into text descriptions and input into a text encoder along with the canonical text in the knowledge base. Image features are extracted using a visual encoder and mapped to a text embedding space to achieve alignment between visual and text features. Adversarial training techniques are employed to obtain diverse training data, and the multimodal large model is optimized through fine-tuning training methods. This enables the multimodal large model to learn the relationship between perturbation data and quality inspection standards and specifications, thereby gradually improving its adaptability and judgment accuracy for different quality inspection tasks. By simulating changes in the field environment, enhanced image data is generated, and the defect recognition ability of the multimodal large model under complex visual conditions is trained, so that the multimodal large model can enhance its adaptability to data fluctuations and uncertainties in the field environment during the learning process. The target AI large model is obtained.

6. The method according to claim 1, characterized in that, The preprocessed business file is input into the target AI large model for multi-dimensional logical verification, and the multi-dimensional logical verification results are obtained as follows: The form integrity of the preprocessed business file is verified based on the target AI large model, and a first verification result is obtained; The data compliance of the preprocessed business file is verified based on the target AI large model, and a second verification result is obtained; Based on the target AI large model, the logical correlation and data consistency between different preprocessed business files are verified to obtain a third verification result; The first verification result, the second verification result, and the third verification result are used as the multi-dimensional logical verification result.

7. The method according to claim 1, characterized in that, Based on the multi-dimensional logical verification results, risk information is identified and the risk level is assessed. The target AI model is then used to predict risk development trends and generate targeted response suggestions, specifically: Based on the multi-dimensional logical verification results and combined with historical problem cases and risk level classification rules in the training knowledge base, risk information is identified, the risk information is classified and the risk level of the risk information is evaluated. The classification results include form-type problems, data-type problems and correlation-type problems, and the risk levels include high risk, medium risk and low risk. Based on the target AI big model, the development trend of the risk information is predicted, and the prediction result is obtained; Based on the prediction results, classification results, and risk levels, risk information will be pushed to the corresponding responsible parties and corresponding handling opinions will be generated.

8. A quality inspection management risk analysis and feedback device, characterized in that, The device includes: The preprocessing module is used to construct a multi-source heterogeneous railway engineering quality inspection business document database, and to preprocess the multi-format files in the multi-source heterogeneous railway engineering quality inspection business document database to generate a standardized structured dataset. The construction module is used to build a training knowledge base based on the standardized structured dataset, railway engineering specifications and historical cases, and to obtain the target AI large model based on the training knowledge base and the preset multimodal large model. The verification module is used to input the preprocessed business file into the target AI model for multi-dimensional logical verification and obtain the multi-dimensional logical verification results. The identification module is used to identify risk information and assess risk level based on the multi-dimensional logical verification results, link the target AI big model to predict the risk development trend, and generate targeted handling suggestions.

9. A computing device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the quality inspection management risk analysis and feedback method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer-executable instructions, which, when executed by a processor, implement the steps of the quality inspection management risk analysis and feedback method according to any one of claims 1 to 7.

Images