Method and system for pushing quality defect report handling measures based on historical experience

Abstract

The application discloses a push method and system of quality defect report processing measures based on historical experience. The method comprises the following steps: establishing a structured database; analyzing the text of the database by using natural language processing technology, extracting key technical features, and constructing a quality defect report keyword library; converting the features of the current quality defect report into a numerical vector, which comprises function position coding, key degree classification, safety classification, key sensitive device type, quality defect report filling condition, whether it is a nuclear grade device and affects its safety function, whether it is a major repair work order that needs technical discussion, and whether it is a common mode failure; calculating the similarity between the current quality defect report and the historical quality defect report, and selecting the top set value of historical quality defect reports with the highest similarity; calculating the comprehensive weight; and recommending the processing measures according to the comprehensive weight order. The application improves the processing efficiency, improves the effectiveness of the measures, iteratively accumulates the defect processing experience, optimizes the recommendation effect, and shortens the learning curve.

Description

Technical Field

[0001] This invention relates to the field of nuclear power plant quality management technology, and in particular to a method and system for pushing out quality defect report processing measures based on historical experience. Background Technology

[0002] A Quality Deficiency Report (QDR) is a report on equipment quality anomalies that occur unexpectedly during maintenance, inspection, and testing in nuclear power plants, where the quality of items fails to meet the original design requirements. Corrective actions must be promptly developed to restore the items to the design specifications. QDR is a quality management approach that integrates cause analysis and corrective action / plan management.

[0003] Regarding the formulation of QDR (Quality Defect Removal) handling measures, the measures for handling quality defects are a key link in ensuring the quality of repaired items. Traditional defect handling mainly relies on the personal experience of nuclear power plant technicians, which has the following problems: High dependence on experience: The formulation of treatment measures relies excessively on the experience and skill level of individual technical personnel; Difficulty in passing on effective experience: Excellent handling experience is hard to accumulate and pass on systematically; Low processing efficiency: Similar defects require repeated analysis, and historical solutions cannot be quickly reused; Low standardization: There is a lack of unified standards and best practice guidance for handling measures.

[0004] The existing nuclear power production management information system only allows manual selection of historical defect reports and historical QDR handling measures, lacking recommendation functions and unable to dynamically learn and optimize.

[0005] Therefore, there is an urgent need for a technical solution that can recommend handling measures based on past experience in handling defects, thereby improving handling efficiency and quality. Summary of the Invention

[0006] This invention aims to address the problems in existing technologies where QDR processing measures heavily rely on individual abilities and experience, making it difficult to obtain effective historical experience and resulting in low processing efficiency. It provides a method and system for pushing out quality defect report processing measures based on historical experience.

[0007] To address the aforementioned technical problems, this application provides a method for pushing out quality defect report processing measures based on historical experience, characterized by comprising: Step S1: Establish a structured database, which includes historical quality defect report handling measures, historical defect handling progress and results, historical defect report descriptions, and information on the consequences of the incidents and immediate actions taken. Step S2: Use natural language processing technology to analyze the text in the database, extract key technical features, and build a keyword library for quality defect reports; Step S3: Convert the features of the current quality defect report into a numerical vector. This numerical vector includes the functional location code, criticality level, safety level, type of critical sensitive equipment, quality defect report filling conditions, whether it is nuclear-grade equipment and affects its safety function, whether the overhaul work order requires technical discussion, and whether it is a common mode failure. Step S4: Calculate the similarity between the current quality defect report and historical quality defect reports, and select the historical quality defect reports with the highest similarity to the previous set value. Step S5: For the aforementioned set number of historical quality defect reports, calculate the comprehensive weight of each historical quality defect report from three dimensions: frequency of occurrence, implementation effect, and timeliness. The comprehensive weight = α × frequency of occurrence weight + β × implementation effect weight + γ × timeliness weight, where α + β + γ = 1. Step S6: Sort historical quality defect reports according to their comprehensive weight and recommend appropriate measures.

[0008] As one aspect of the above method, the database is characterized by including quality defect report descriptions, functional location codes, functional location descriptions, geographical descriptions, criticality classifications, safety classifications, types of critical and sensitive equipment, application dates, quality defect report filling conditions, defect descriptions, whether it is nuclear-grade equipment and affects its safety functions, whether a major overhaul work order requires technical discussion, whether it is a common-mode failure, cause analysis, historical defect report descriptions, event consequences and immediate actions taken, historical technical solutions, implementation effects of historical technical solutions, handling measures, defect handling results and progress, and quality control verification results.

[0009] As another embodiment of the above method, the step S2 specifically includes: Step S2-1: Clean the data in the database; Step S2-2: Using natural language processing technology, the continuous text in the defect description, cause analysis, historical defect description, event consequences and immediate actions, historical technical solutions, implementation effects of historical technical solutions, and handling measures is segmented into word units, and key technical features are extracted. Step S2-3: Remove meaningless word units; Step S2-4: Using natural language processing technology, words with similar meanings are unified according to preset rules to construct a keyword library for quality defect reports.

[0010] As another method described above, the similarity calculation method in step S4 is as follows: comparing the features included in the current quality defect report and the historical quality defects, the similarity is determined by the proportion of the number of identical features in the functional location code, criticality level, safety level, type of critical sensitive equipment, quality defect report filling conditions, whether it is nuclear-grade equipment and affects its safety function, whether the overhaul work order requires technical discussion, and whether it is a common-mode fault.

[0011] As another embodiment of the above method, the occurrence frequency weight is characterized in that, in step S5, the occurrence frequency weight = the number of times the same processing measure occurs when the [functional location code] and [quality defect report description] are the same in the current quality defect report and the historical quality defect report / the total number of quality defect reports with the same [functional location code] and [quality defect report description]; The implementation effect weight = [QC verification result] value * key word assignment, where [Qualified] in [QC verification result] is 1 and [Unqualified] is 0; the key words contained in [Defect handling result and progress] have set assignments; The timeliness weight is calculated using the following formula: (1) Where W is the timeliness weight, and e is the natural constant. The attenuation coefficient is... The number of days since the quality defect report was submitted.

[0012] As another method described above, it is characterized by further including: step S7, updating the database with the processed current quality defect report as a historical quality defect report.

[0013] This application also provides a system for pushing out quality defect report processing measures based on historical experience, characterized in that it includes: The database module is used to collect quality defect report data and build and store a structured database; The text analysis module is used to analyze the text in the database using natural language processing technology, extract key technical features, and build a keyword library for quality defect reports. The feature conversion module is used to convert the features of the current quality defect report into a numerical vector. The numerical vector includes the functional location code, criticality level, safety level, type of critical sensitive equipment, quality defect report filling conditions, whether it is nuclear-grade equipment and affects its safety function, whether the overhaul work order requires technical discussion, and whether it is a common mode fault. The similarity calculation module is used to calculate the similarity between the current quality defect report and historical quality defect reports, and select the historical quality defect reports with the highest similarity to a previous set value. The comprehensive weight calculation module is used to calculate the comprehensive weight of each historical quality defect report from three dimensions: frequency of occurrence, implementation effect, and timeliness for the pre-set number of historical quality defect reports. The comprehensive weight is calculated as follows: comprehensive weight = α × frequency of occurrence weight + β × implementation effect weight + γ × timeliness weight, where α + β + γ = 1. The sorting module is used to sort historical quality defect reports according to a comprehensive weight and recommend handling measures.

[0014] As one aspect of the aforementioned system, the text analysis module is specifically used to perform the following steps; Step S2-1: Clean the data in the database; Step S2-2: Using natural language processing technology, the continuous text in the quality defect report, including defect description, cause analysis, description of historical defect sheets, event consequences and immediate actions, historical technical solutions, implementation effects of historical technical solutions, and handling measures, is segmented into word units, and key technical features are extracted. Step S2-3: Remove meaningless word units; Step S2-4: Using natural language processing technology, words with similar meanings are unified according to preset rules to construct a keyword library for quality defect reports.

[0015] As another method of the above system, the similarity calculation module calculates the similarity by comparing the features included in the current quality defect report and the historical quality defects, and taking the proportion of the number of the same features in the common mode fault to the total number of features as the similarity. In the comprehensive weight calculation module: the occurrence frequency weight = the number of times the same handling measure occurs when the [Functional Location Code] and [Quality Defect Report Description] are the same in the current quality defect report and the historical quality defect reports / the total number of quality defect reports with the same [Functional Location Code] and [Quality Defect Report Description]; The implementation effect weight = [QC verification result] value * key word assignment, where [Qualified] in [QC verification result] is 1 and [Unqualified] is 0; the key words contained in [Defect handling result and progress] have set assignments; The timeliness weight is calculated using the following formula:

[0016] Where W is the timeliness weight, and e is the natural constant, approximately 2.718. The attenuation coefficient is... The number of days since the quality defect report was submitted.

[0017] As a further embodiment of the above system, it is characterized by further comprising: The user interaction module is used to provide an interactive interface, display recommendation results, and collect user feedback. The feedback learning module is used to dynamically optimize the recommendation algorithm based on users' adoption of recommendation measures and the processing results.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. Automatic recommendations reduce analysis time and improve processing efficiency; 2. Based on effective historical experience, recommendations can be made to avoid repeating mistakes and improve the effectiveness of measures; 3. Transform personal experience into systematic knowledge to achieve iterative accumulation of defect handling experience; 4. By continuously using the system, optimize the recommendation effect and form a virtuous cycle; 5. Provide experience-based guidance to new employees to shorten the learning curve. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the method for pushing out quality defect report processing measures based on historical experience in a specific implementation. Detailed Implementation

[0020] The technical solutions provided in this application are further illustrated below with reference to the embodiments.

[0021] like Figure 1 As shown in the figure, as a specific embodiment of the present invention, the method for pushing out quality defect report processing measures based on historical experience includes the following steps: Step S1: Establish a structured database, which includes information such as historical quality defect report (QDR) handling measures, historical defect handling progress and results, historical defect form descriptions, and event consequences and immediate actions taken. More specifically, the database may include QDR descriptions, functional location codes, functional location descriptions, geographical descriptions, criticality classifications, safety classifications, SPV (Single Point Vulnerability) types, application dates, QDR reporting conditions, defect descriptions, whether it is nuclear-grade equipment and affects its safety functions, whether a major overhaul work order requires technical discussion, whether it is a common-mode failure, root cause analysis, historical defect form descriptions, event consequences and immediate actions taken, historical technical solutions, implementation effects of historical technical solutions, handling measures, defect handling results and progress, and QC (Quality Control) verification results. Step S2: Analyze the text in the database using Natural Language Processing (NLP) technology to extract key technical features such as defects like "water seepage" and repair results like "good results," and build a keyword library for quality defect reports. This is specifically divided into the following four steps. Step S2-1: Clean the data: Clean the data in the database; for example, cleaning rules include: deleting data such as special characters, spaces, and consecutive punctuation marks; Step S2-2: Segment word units. Using natural language processing technology, the continuous text in the quality defect report, including defect description, cause analysis, historical defect description, event consequences and immediate actions, historical technical solutions, implementation effects of historical technical solutions, and handling measures, is segmented into word units, and key technical features are extracted. Steps S2-3: Remove stop words and meaningless word units; for example, remove meaningless word units such as modal particles. Step S2-4: Standardize word forms. Using natural language processing technology, words with similar meanings are unified according to preset rules. For example, "qualified", "can be used normally", and "good effect" are unified into "good effect", and a keyword library for quality defect reports is constructed. Step S3: Convert the features of the current quality defect report into a numerical vector. This numerical vector includes features such as functional location code, criticality level, safety level, SPV type, QDR reporting conditions, whether it is nuclear-grade equipment and affects its safety functions, whether the overhaul work order requires technical discussion, and whether it is a common-mode fault. The features included in this numerical vector are all fields with known meanings in the nuclear power industry.

[0022] Step S4: Calculate the similarity between the current quality defect report and historical quality defect reports, select the top N historical quality defect reports with the highest similarity, extract their handling measures, and generate a recommendation list; N is set to 10 for example.

[0023] The similarity calculation method is as follows: compare the similarity of features included in the current quality defect report and the historical quality defects, and take the proportion of the number of identical features in the functional location code, criticality level, safety level, SPV type, QDR filling conditions, whether it is nuclear-grade equipment and affects its safety function, whether the overhaul work order requires technical discussion, and whether it is a common mode fault as the proportion of the total number of features.

[0024] Step S5: For the first N historical quality defect reports, calculate the comprehensive weight of each report from three dimensions: frequency of occurrence, implementation effect, and timeliness. The comprehensive weight is calculated as: Comprehensive Weight = α × Frequency of Occurrence Weight + β × Implementation Effect Weight + γ × Timeliness Weight, where α + β + γ = 1. Considering all factors, we can take α = 0.2, β = 0.5, and γ = 0.3, which are the set coefficients for the frequency of occurrence weight, implementation effect weight, and timeliness weight, respectively.

[0025] The frequency weight is calculated as follows: (Number of times the same handling measure occurs when the [Functional Location Code] and [QDR Description] are identical in the current and historical quality defect reports) / (Total number of quality defect reports with identical [Functional Location Code] and [QDR Description]). Therefore, the higher the frequency, the higher the weighting ratio. The weighting for the three dimensions—frequency, implementation effectiveness, and timeliness—can all be based on a percentage system. " / " indicates division.

[0026] The implementation effect weight is calculated as follows: [QC Verification Result] value * Keyword Segmentation Value. In the [QC Verification Result], "Qualified" is assigned a value of 1, and "Unqualified" is assigned a value of 0. Keyword segments in the [Defect Handling Result and Progress] field, such as "Temporary Replacement," can be assigned a value of 0.5, and "Good Results" can be assigned a value of 1, etc. The implementation effect weight is calculated by combining the "Qualified" or "Unqualified" value in the [QC Verification Result] field with the keyword segmentation value. "*" indicates multiplication.

[0027] The formula for calculating the timeliness weight is as follows: Formula 1: (1) Where W is the timeliness weight, and e is the natural constant, approximately 2.718. The attenuation coefficient can be set to control the descent speed. The number of days since the quality defect report was submitted is considered; it is evident that the closer the quality defect report is to the present time, the higher its timeliness and the higher its weighting score.

[0028] Step S6: Sort historical quality defect reports according to their comprehensive weight and recommend appropriate measures.

[0029] For example, the measure with the highest weight is placed at the top of the list, and the list is sorted according to the overall weight. Preferably, different background colors are used to highlight recommended measures with different overall weights. The overall weight is determined by steps 4 and 5. Step 4 selects a group of historical defect reports with the highest similarity (usually 10 reports per group). Step 5 calculates the overall weight of each historical defect report and sorts the group of reports with the highest similarity from highest to lowest according to the overall weight.

[0030] Preferably, the method further includes step S7, which updates the database as a historical quality defect report by processing the current quality defect report.

[0031] As another specific implementation, to carry out the above method, a push system for quality defect report processing measures based on historical experience is provided accordingly. This system includes: The database module is used to collect data and build and store a structured database. This database includes information such as historical quality defect report (QDR) handling measures, historical defect handling progress and results, historical defect form descriptions, and event consequences and immediate actions taken. More specifically, the database may include QDR descriptions, functional location codes, functional location descriptions, geographical descriptions, criticality classifications, safety classifications, SPV (Single Point Vulnerability) types, application dates, QDR reporting conditions, defect descriptions, whether it is nuclear-grade equipment and affects its safety functions, whether it is a major overhaul work order requiring technical discussion, whether it is a common-mode failure, root cause analysis, historical defect form descriptions, event consequences and immediate actions taken, historical technical solutions, implementation effects of historical technical solutions, handling measures, defect handling results and progress, and QC (Quality Control) verification results. The database module can also be used to clean the collected defective data. Cleaning rules include, for example, deleting data such as special symbols, spaces, and consecutive punctuation marks to achieve data standardization. Text Analysis Module: Used to analyze the text in the database using NLP technology, extract key technical features such as defects like "water seepage" and repair results like "good results", and build a keyword library for quality defect reports; The feature conversion module is used to convert the features of the current quality defect report into a numerical vector. The numerical vector includes features such as functional location code, criticality level, safety level, SPV type, QDR filling conditions, whether it is nuclear-grade equipment and affects its safety function, whether the overhaul work order requires technical discussion, and whether it is a common mode failure. Similarity calculation module: used to calculate the similarity between the current quality defect report and historical quality defect reports, select the historical quality defect reports with the highest similarity (preset value N), extract their handling measures, and generate a recommendation list; N is set to 10 for example. The similarity calculation method is as follows: compare the features included in the current quality defect report and the historical quality defects, and take the proportion of the same features in the functional location code, criticality level, safety level, SPV type, QDR filling conditions, whether it is nuclear-grade equipment and affects its safety function, whether the overhaul work order requires technical discussion, and whether it is a common mode fault as the proportion of the total number of features. The comprehensive weight calculation module is used to calculate the comprehensive weight of each of the first N historical quality defect reports from three dimensions: frequency of occurrence, implementation effect, and timeliness. The comprehensive weight is calculated as follows: Comprehensive Weight = α × Frequency of Occurrence Weight + β × Implementation Effect Weight + γ × Timeliness Weight, where α, β, and γ are the set coefficients for the frequency of occurrence weight, implementation effect weight, and timeliness weight, respectively, and α + β + γ = 1. Considering all factors, α = 0.2, β = 0.5, and γ = 0.3 can be chosen. The frequency weight is calculated as follows: (Number of times the same handling measure occurs when the [Functional Location] and [QDR Description] are identical in the current quality defect report and historical quality defect reports) / (Total number of quality defect reports with identical [Functional Location] and [QDR Description]). Therefore, the higher the frequency, the higher the weighting ratio for frequency. The weighting for the three dimensions—frequency, implementation effectiveness, and timeliness—can all be based on a percentage system.

[0032] The implementation effect weight is calculated as follows: [QC Verification Result] value * Keyword segmentation value. In the [QC Verification Result], "Qualified" is assigned a value of 1, and "Unqualified" is assigned a value of 0. Keyword segments in the [Defect Handling Result and Progress] field, such as "Temporary Replacement," can be assigned a value of 0.5, and "Good Results" can be assigned a value of 1, etc. The implementation effect weight is calculated by combining the "Qualified" or "Unqualified" value in the [QC Verification Result] field with the keyword segmentation value. * indicates multiplication.

[0033] The formula for calculating the timeliness weight is as follows: Where W is the timeliness weight, and e is the natural constant, approximately 2.718. The attenuation coefficient can be set to control the descent speed. The number of days since the quality defect report was submitted is considered; it is evident that the closer the quality defect report is to the present time, the higher its timeliness and the higher its weighting score.

[0034] The sorting module is used to sort historical quality defect reports according to their comprehensive weight and recommend handling measures; for example, the measure with the highest weight is placed at the top of the list and listed in order of comprehensive weight; preferably, different background colors are used to highlight recommended measures with different comprehensive weights.

[0035] Furthermore, the text analysis module is specifically used to perform the following steps; Step S2-1: Clean the data in the database; Step S2-2: Using natural language processing technology, the continuous text in the quality defect report, including defect description, cause analysis, description of historical defect sheets, event consequences and immediate actions, historical technical solutions, implementation effects of historical technical solutions, and handling measures, is segmented into word units, and key technical features are extracted. Step S2-3: Remove meaningless word units; Step S2-4: Using natural language processing technology, words with similar meanings are unified according to preset rules to construct a keyword library for quality defect reports.

[0036] Furthermore, the system for pushing out quality defect report processing measures based on historical experience may also include: User interaction module: Used to provide an intuitive interactive interface, display recommendation results and collect user feedback; Feedback learning module: Used to dynamically optimize the recommendation algorithm based on user adoption and processing results. For example, one optional dynamic optimization recommendation algorithm is to update the structured database with the processed current quality defect report as a historical quality defect report.

[0037] As can be seen from the above detailed description of this application: 1. Automatic recommendations reduce analysis time and improve processing efficiency; 2. Based on effective historical experience, recommendations can be made to avoid repeating mistakes and improve the effectiveness of measures; 3. Transform personal experience into systematic knowledge to achieve iterative accumulation of defect handling experience; 4. By continuously using the system, optimize the recommendation effect and form a virtuous cycle; 5. Provide experience-based guidance to new employees to shorten the learning curve.

[0038] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for pushing out quality defect report processing measures based on historical experience, characterized in that, include: Step S1: Establish a structured database, which includes historical quality defect report handling measures, historical defect handling progress and results, historical defect report descriptions, and information on the consequences of the incidents and immediate actions taken. Step S2: Use natural language processing technology to analyze the text in the database, extract key technical features, and build a keyword library for quality defect reports; Step S3: Convert the features of the current quality defect report into a numerical vector. This numerical vector includes the functional location code, criticality level, safety level, type of critical sensitive equipment, quality defect report filling conditions, whether it is nuclear-grade equipment and affects its safety function, whether the overhaul work order requires technical discussion, and whether it is a common mode failure. Step S4: Calculate the similarity between the current quality defect report and historical quality defect reports, and select the historical quality defect reports with the highest similarity to the previous set value. Step S5: For the aforementioned set number of historical quality defect reports, calculate the comprehensive weight of each historical quality defect report from three dimensions: frequency of occurrence, implementation effect, and timeliness. The comprehensive weight = α × frequency of occurrence weight + β × implementation effect weight + γ × timeliness weight, where α + β + γ = 1, and α, β, and γ are set coefficients. Step S6: Sort historical quality defect reports according to their comprehensive weight and recommend appropriate measures.

2. The method for pushing out quality defect report processing measures according to claim 1, characterized in that, The database includes quality defect report descriptions, functional location codes, functional location descriptions, geographical descriptions, criticality levels, safety levels, types of critical and sensitive equipment, application dates, quality defect report filling conditions, defect descriptions, whether it is nuclear-grade equipment and affects its safety functions, whether the overhaul work order requires technical discussion, whether it is a common mode failure, cause analysis, descriptions of historical defect reports, consequences of the incident and immediate actions taken, historical technical solutions, implementation effects of historical technical solutions, handling measures, defect handling results and progress, and quality control verification results.

3. The method for pushing out quality defect report processing measures according to claim 1 or 2, characterized in that, Step S2 specifically includes: Step S2-1: Clean the data in the database; Step S2-2: Using natural language processing technology, the continuous text in the defect description, cause analysis, historical defect description, event consequences and immediate actions, historical technical solutions, implementation effects of historical technical solutions, and handling measures is segmented into word units, and key technical features are extracted. Step S2-3: Remove meaningless word units; Step S2-4: Using natural language processing technology, words with similar meanings are unified according to preset rules to construct a keyword library for quality defect reports.

4. The method for pushing out quality defect report processing measures according to claim 1 or 2, characterized in that, In step S4, the similarity calculation method is as follows: compare the features included in the current quality defect report and the historical quality defects, and take the proportion of the number of the same features in the functional location code, criticality level, safety level, type of critical sensitive equipment, quality defect report filling conditions, whether it is nuclear-grade equipment and affects its safety function, whether the overhaul work order requires technical discussion, and whether it is a common mode fault as the proportion of the total number of features.

5. The method for pushing out quality defect report processing measures according to claim 1 or 2, characterized in that, In step S5, the occurrence frequency weight = the number of times the same handling measure occurs when the "functional location code" and "quality defect report description" are the same in the current quality defect report and the historical quality defect report / the total number of quality defect reports with the same "functional location code" and "quality defect report description"; The implementation effect weight is defined as "QC verification result" value * key word assignment, where "qualified" in "QC verification result" is assigned 1 and "unqualified" is assigned 0; the key words contained in "defect handling results and progress" have set assignments. The timeliness weight is calculated using the following formula: (1) Where W is the timeliness weight, and e is the natural constant. The attenuation coefficient is... The number of days since the quality defect report was submitted.

6. The method for pushing out quality defect report processing measures according to claim 1 or 2, characterized in that, It also includes step S7, which updates the database as a historical quality defect report by processing the current quality defect report.

7. A system for pushing out quality defect report processing measures based on historical experience, characterized in that, include: The database module is used to collect quality defect report data and build and store a structured database; The text analysis module is used to analyze the text in the database using natural language processing technology, extract key technical features, and build a keyword library for quality defect reports. The feature conversion module is used to convert the features of the current quality defect report into a numerical vector. The numerical vector includes the functional location code, criticality level, safety level, type of critical sensitive equipment, quality defect report filling conditions, whether it is nuclear-grade equipment and affects its safety function, whether the overhaul work order requires technical discussion, and whether it is a common mode fault. The similarity calculation module is used to calculate the similarity between the current quality defect report and historical quality defect reports, and select the historical quality defect reports with the highest similarity to a previous set value. The comprehensive weight calculation module is used to calculate the comprehensive weight of each historical quality defect report from three dimensions: frequency of occurrence, implementation effect, and timeliness for the aforementioned set value of historical quality defect reports. The comprehensive weight is calculated as follows: comprehensive weight = α × frequency of occurrence weight + β × implementation effect weight + γ × timeliness weight, where α + β + γ = 1, and α, β, and γ are set coefficients. The sorting module is used to sort historical quality defect reports according to a comprehensive weight and recommend handling measures.

8. The push system for quality defect report processing measures according to claim 7, characterized in that, The text analysis module is specifically used to perform the following steps; Step S2-1: Clean the data in the database; Step S2-2: Using natural language processing technology, the continuous text in the quality defect report, including defect description, cause analysis, description of historical defect sheets, event consequences and immediate actions, historical technical solutions, implementation effects of historical technical solutions, and handling measures, is segmented into word units, and key technical features are extracted. Step S2-3: Remove meaningless word units; Step S2-4: Using natural language processing technology, words with similar meanings are unified according to preset rules to construct a keyword library for quality defect reports.

9. The push system for quality defect report processing measures according to claim 7 or 8, characterized in that, The similarity calculation module calculates similarity by comparing the features included in the current quality defect report and historical quality defects, and using the proportion of the number of identical features in the functional location code, criticality level, safety level, type of critical sensitive equipment, quality defect report filling conditions, whether it is nuclear-grade equipment and affects its safety function, whether the overhaul work order requires technical discussion, and whether it is a common-mode fault as the proportion of the total number of features. In the comprehensive weight calculation module: the frequency weight = the number of times the same handling measure occurs when the "functional location code" and "quality defect report description" are the same in the current quality defect report and the historical quality defect report / the total number of quality defect reports with the same "functional location code" and "quality defect report description"; The implementation effect weight is defined as "QC verification result" value * key word assignment, where "qualified" in "QC verification result" is assigned 1 and "unqualified" is assigned 0; the key words contained in "defect handling results and progress" have set assignments. The timeliness weight is calculated using the following formula: Where W is the timeliness weight, and e is the natural constant, approximately 2.

718. The attenuation coefficient is... The number of days since the quality defect report was submitted.

10. The push system for quality defect report processing measures according to claim 7 or 8, characterized in that, Also includes: The user interaction module is used to provide an interactive interface, display recommendation results, and collect user feedback. The feedback learning module is used to dynamically optimize the recommendation algorithm based on users' adoption of recommendation measures and the processing results.

Images