An AI-based intelligent management method and system for satellite test quality

Abstract

This invention discloses an intelligent management method and system for satellite testing quality based on artificial intelligence. The method mainly includes: establishing a multi-dimensional structured classification system for standardized input of satellite testing quality issues, and linking with a historical knowledge base to recommend similar issues; executing a multi-stage closed-loop status flow including issue analysis, rectification, verification, and closure, and configuring a time-based hierarchical automatic reminder mechanism; extracting real-time status data, processing records, and key statistical indicators to construct structured prompt words, and calling a large language model service to perform trend analysis, root cause clustering, and risk warning; combining real-time data and AI warning items, and outputting quantitative conclusions for factory review according to preset quantitative criteria; and automatically generating and distributing standardized quality reports according to a preset cycle. This invention significantly improves the efficiency, objectivity, and full lifecycle traceability of quality management in satellite mass production and multi-model parallel testing scenarios.

Description

Technical Field

[0001] This invention relates to the field of spacecraft test quality management technology, and in particular to an intelligent management method and system for satellite test quality based on artificial intelligence. Background Technology

[0002] Satellite testing is a crucial step in the satellite development process, encompassing the entire assembly, integration, and testing process from individual components to the complete satellite. It primarily includes multiple stages such as desktop testing, integrated testing, and environmental testing. During satellite testing, various quality issues inevitably arise, including but not limited to hardware failures, software defects, interface incompatibility, documentation errors, and operational mistakes. Effective management of these issues directly impacts the satellite's development quality and launch safety.

[0003] With the rapid development of commercial spaceflight, mass production of satellites has placed higher demands on testing efficiency and quality management. Taking mass-produced satellites as an example, multiple satellites may be undergoing different testing phases simultaneously on the same production line, involving the collaborative work of multiple testing teams and responsible units, significantly increasing the number and complexity of quality issues. However, existing satellite testing quality issue management methods suffer from the following three core technical problems: In existing satellite testing processes, there is a lack of standardized and traceable management paths for quality issues from discovery to final closure. Specifically: In the issue recording stage, different testers record issues independently, some using electronic documents, some using paper records, and some only mentioning them in meeting minutes. The recording formats are inconsistent, the level of detail in descriptions varies, and the classification standards differ, resulting in fragmented issue data scattered across multiple documents. In the issue handling stage, there is a lack of standardized closed-loop processes. Once an issue is recorded, there is no clear responsibility allocation mechanism, making it unclear which unit or individual should be responsible. Processing progress relies mainly on manual follow-ups and verbal communication, easily leading to issues being overlooked, processing exceeding deadlines, and shirking of responsibility. In the verification stage, there is a lack of standardized confirmation mechanisms. Some issues are closed without sufficient verification, creating potential quality risks. In the statistical analysis stage, due to inconsistent data formats and scattered storage, multi-dimensional statistical analysis of issues requires a large amount of manual work, often taking several hours to complete a single comprehensive analysis, and it is difficult to discover hidden trends and patterns in the issue data. During the factory review process, the quality information that the review committee can obtain mainly comes from qualitative textual descriptions and manually compiled brief statistics. Key indicators such as whether all serious problems have been closed and whether the overall closure rate meets the standards lack automated inspection and quantitative presentation methods, and the review conclusions largely rely on subjective experience judgment.

[0004] The management bottlenecks in the aforementioned stages prevent quality issues from flowing along a clear and unified path from discovery to closure, resulting in low overall quality management efficiency. This is especially true in batch production scenarios involving multiple satellites undergoing parallel testing, where the number of issues is large and many units are involved, making manual management methods insufficient to meet the requirements.

[0005] The experience gained in handling problems during satellite testing—including troubleshooting approaches, root cause analysis methods, and effective corrective and preventative measures—is an extremely valuable technical asset. However, under the current management system, this experience largely remains in the memories of a few experienced testers, or is scattered across various documents, emails, and meeting minutes, lacking a unified outlet for subsequent personnel to access and learn from. When these experienced personnel are transferred, leave, or participate in other projects, their valuable experience is lost. When testers for subsequent models or new hires encounter similar problems, they often need to analyze and troubleshoot from scratch, unable to quickly draw upon the experience of predecessors. Especially in batch production, where the technical states of different models are highly similar, many problem-solving solutions are reusable. However, due to the lack of a systematic knowledge accumulation and retrieval mechanism, this reusable experience is not effectively utilized, leading to the recurrence of similar problems in different models and batches, wasting significant time and resources.

[0006] During satellite mass production, the testing of each satellite generates a large amount of quality issue data, which should be crucial input for driving product design iteration and improvement. However, under the current management system, issue data from different models and batches are isolated. After testing, issue records are archived and sealed, lacking the ability to aggregate and analyze data across models and batches. Managers struggle to identify from a holistic perspective which issues are specific to individual models and which are platform-level common defects recurring across multiple models. More importantly, even if design improvements are implemented for a particular model to address a specific type of issue, there is a lack of systematic means to track and verify whether the improvements are truly effective in subsequent models. Whether the improvements have eliminated the problem or introduced new problems relies on manual judgment rather than data-driven verification. This "test and be done" approach, where changes are made without verification, results in a lack of closed-loop feedback in product technology iteration. Similar design defects may persist across multiple batches, preventing true continuous improvement.

[0007] In summary, existing satellite testing quality management methods suffer from systemic deficiencies in three aspects: management path, experience transfer, and technology iteration. There is an urgent need for a method, system, equipment, and storage medium that can provide a unified quality management path, establish a systematic knowledge transfer outlet, and drive iterative improvement through data-driven technology, in order to comprehensively enhance the level and efficiency of satellite testing quality management. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an intelligent management method and system for satellite test quality based on artificial intelligence. By establishing a multi-dimensional structured problem classification system and a standardized closed-loop management process, and combining the semantic analysis and reasoning capabilities of a large language model, it achieves digital management of the entire lifecycle of quality problems, from discovery, analysis, rectification, acceptance to closure. At the same time, through automated statistical dashboards, AI deep analysis, quantitative assessment of factory conditions, and the construction of a problem knowledge base, it provides objective, quantitative, and traceable decision-making basis for aerospace test review, and realizes the automatic accumulation and intelligent reuse of test experience.

[0009] This invention provides an intelligent management method for satellite test quality based on artificial intelligence, comprising: Establish a multi-dimensional structured problem classification system, standardize the input of quality problems found during satellite testing, generate a structured problem dataset that includes problem classification, problem level, problem status, responsibility attribution, time node and cause analysis dimensions, and mark the initial status identifier; Based on the structured problem dataset and initial status identifiers, a multi-stage status transition is executed, including problem analysis and responsibility allocation, rectification implementation, verification and confirmation, and problem closure. The problem status identifiers are dynamically updated to obtain real-time status data and processing records for each problem. Based on real-time status data and processing records, preliminary statistical analysis results are automatically generated. Extract the real-time status data, processing records, and preliminary statistical analysis results. Construct prompt words including a data overview area, a problem list area, and an analysis requirement area according to a preset template. Send these prompt words to the large language model service via an interface to perform analysis and obtain analysis results including trend analysis, root cause clustering, recurring problem identification, risk warning, and improvement suggestions. Based on the real-time status data, processing records, and risk warning items in the analysis results, a comprehensive evaluation is conducted against preset quantitative criteria, and a quantitative evaluation conclusion is output for factory review decision-making.

[0010] In one embodiment of the present invention, it further includes: The system automatically extracts structured problem datasets, real-time status data, and analysis results from the system according to a preset cycle, calls a large language model to perform periodic summary analysis to generate standardized quality reports, and automatically distributes them to preset recipients via email protocols or enterprise communication tool interfaces. When the issue status indicator is switched to closed, the corresponding issue record is automatically archived to the issue knowledge base. The archived content includes at least complete structured data such as issue description, root cause, corrective measures and preventive measures. When standardizing the input of quality issues discovered during satellite testing, the text describing the issues to be input is semantically matched with the issue knowledge base, and the historical issues with the highest matching degree and their solutions are retrieved and pushed to the system in real time for reference.

[0011] In one embodiment of the present invention, the multi-dimensional structured problem classification system adopts a two-level classification structure. The first-level classification includes problems in the test preparation stage, hardware problems, software problems, interface documentation problems, overall design problems, and operational problems. Each first-level classification is further divided into corresponding second-level classifications based on the satellite model and test requirements. The standardized input includes text input or voice input. When voice input is used, the voice signal is converted into natural language descriptive text by the voice recognition engine. The large language model is called to perform semantic analysis on the descriptive text to automatically recommend primary category, secondary category and question level. The input is completed after manual confirmation or correction.

[0012] In one embodiment of the present invention, when a problem approaches or exceeds the planned closing date, a reminder notification mechanism is automatically triggered; Send reminders via email protocols or enterprise communication tool interfaces, and implement a tiered escalation strategy: send a reminder 3 days before the planned closing date, send daily reminders after the deadline, and escalate the notification to the superior manager 7 days after the deadline.

[0013] In one embodiment of the present invention, the preliminary statistical analysis results include the percentage of problem status, the distribution of problem levels, the distribution of problem categories, the closure rate of responsible units, the total number of problems, the overall closure rate, the number of serious and above problems, the number of overdue unclosed problems, and the average closure cycle.

[0014] In one embodiment of the present invention, the preset quantification criteria include: All critical and serious issues have been closed; The overall issue closure rate has reached the preset threshold; The number of overdue unclosed issues is zero; and No high-risk items that have not been eliminated were found in the analysis results; If any criterion is not met, the assessment conclusion will be "not passed" or an additional risk assessment explanation will be required. The assessment results will be presented in the form of quantitative data.

[0015] This invention also provides an intelligent management system for satellite test quality based on artificial intelligence, comprising: The structured data entry module is configured to generate a structured problem dataset and mark the initial state identifier; The closed-loop process management module is configured to perform multi-stage state transitions and responsibility allocation based on the structured problem dataset and initial state identifiers, dynamically update problem state identifiers, and obtain real-time state data and processing records. An automated statistical analysis module is configured to automatically generate preliminary statistical analysis results. The AI-assisted analysis module is configured to extract the real-time status data, process records and preliminary statistical analysis results, construct prompt words and call the large language model service to perform analysis, and generate analysis results including trend analysis, root cause clustering, risk warning and improvement suggestions; The factory evaluation module is configured to perform a comprehensive evaluation based on the real-time status data, processing records, and risk warning items in the analysis results, against preset quantitative criteria, and output a quantitative evaluation conclusion.

[0016] In one embodiment of the present invention, it further includes: The periodic report generation module is configured to automatically extract system operation data according to a preset period, call the AI ​​analysis interface to generate standardized quality reports, and automatically distribute the reports via email or enterprise communication tool interfaces. The question knowledge base module is configured to automatically perform experience archiving when a question's status changes to closed; and it is configured with a semantic retrieval interface to instantly trigger similar question retrieval and historical solution recommendations when the structured data entry module receives a new question. The cross-project correlation analysis module is configured to identify common problems and track improvements.

[0017] In one embodiment of the present invention, the structured data input module has a built-in speech recognition engine and a large language model semantic analysis interface, which is used to convert speech signals into text and automatically recommend question categories and levels.

[0018] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the program, implements the above-described intelligent management method for satellite test quality based on artificial intelligence.

[0019] The present invention has the following beneficial effects: (1) By establishing a structured classification system and introducing voice input and AI-assisted automatic classification, the problem of fragmented quality problem data was solved, the threshold for manual input and the error rate of classification were reduced, and testers could quickly complete the problem input on the operation site, thus realizing the unified management of problem data.

[0020] (2) By standardizing the closed-loop process and the automatic reminder mechanism, the omission of problems and the overdue processing are avoided, thus improving the efficiency of closed-loop management.

[0021] (3) Through automated statistical analysis, the manual statistical work that originally required 4-8 hours is shortened to real-time updates, which greatly improves the efficiency of data analysis.

[0022] (4) By introducing AI-assisted in-depth analysis, problem trend prediction and risk warning were realized, enabling quality management to shift from passive response to proactive prevention.

[0023] (5) By using a quantitative evaluation model, objective data support is provided for factory review, which changes the current situation of relying on subjective experience judgment.

[0024] (6) The automatic generation and distribution of periodic reports enabled timely sharing of quality information and improved the efficiency of multi-team collaboration.

[0025] (7) Through the knowledge base construction and entry process, similar problems are recommended in real time, enabling testers to obtain historical processing experience as soon as a problem is discovered, which promotes the systematic inheritance and reuse of testing experience and reduces the repetitive analysis of similar problems.

[0026] (8) Through cross-project correlation analysis, common platform-level problems can be identified, promoting systematic improvement and continuous iterative optimization of product design. Attached Figure Description

[0027] Figure 1 A flowchart of an intelligent satellite test quality management method based on artificial intelligence is shown in one embodiment of the present invention; Figure 2 A block diagram of an artificial intelligence-based intelligent management system for satellite test quality is shown in one embodiment of the present invention; Figure 3 A flowchart of closed-loop management of problems in one embodiment of the present invention is shown; Figure 4 A flowchart of AI-assisted analysis in one embodiment of the present invention is shown. Detailed Implementation

[0028] In the following description, the invention is described with reference to various embodiments. However, those skilled in the art will recognize that the embodiments may be practiced without one or more specific details or with other alternatives and / or additional methods, materials, or components. In other instances, well-known structures, materials, or operations are not shown or described in detail so as not to obscure the inventive points of the invention. Similarly, for illustrative purposes, specific quantities, materials, and configurations are set forth to provide a comprehensive understanding of embodiments of the invention. However, the invention is not limited to these specific details.

[0029] In this invention, the various embodiments are merely intended to illustrate the solutions of the invention and should not be construed as limiting.

[0030] In this specification, references to "an embodiment" or "this embodiment" mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment of the invention. The phrase "in one embodiment" appearing throughout this specification does not necessarily refer to the same embodiment in all instances.

[0031] Furthermore, the numbering of the steps in the methods of the present invention does not limit the execution order of the method steps. Unless otherwise specified, the method steps may be executed in different orders.

[0032] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0033] Figure 1 A flowchart of an intelligent satellite test quality management method based on artificial intelligence is shown in one embodiment of the present invention.

[0034] like Figure 1 As shown, in this embodiment, the intelligent management method for satellite test quality based on artificial intelligence includes: Step S1: Structured Problem Entry. Establish a multi-dimensional structured problem classification system, standardize the entry of quality problems found during satellite testing, and generate a structured problem dataset containing problem classification, problem level, problem status, responsibility attribution, time node, and root cause analysis dimensions, and mark the initial status identifier. The classification system includes the following dimensions: Problem classification dimension, adopting a two-level classification structure. The first-level classification includes at least test preparation phase problems, hardware problems, software problems, interface documentation problems, overall design problems, and operational problems. Each first-level classification can have corresponding second-level classifications set according to the satellite model and test requirements; Problem severity dimension, including four levels: critical, severe, minor, and slight; Problem status dimension, including pending analysis, under rectification, under verification, and closed; Responsibility attribution dimension, including the responsible unit and responsible person; Time dimension, including the discovery date, planned closure date, and actual closure date; Root cause analysis dimension, including the root cause, corrective measures, and preventive measures. Each problem record includes at least the structured fields corresponding to the above dimensions. During the data entry process, testers can provide a problem description via text or voice input. When voice input is used, the system converts the voice signal into natural language descriptive text using a speech recognition engine. The system then calls a large language model to perform semantic analysis on the descriptive text, automatically recommending the primary, secondary, and problem levels. After manual confirmation or correction, the data entry is completed. Simultaneously, the system automatically retrieves historical similar problems from the problem knowledge base based on the descriptive text, recommending the historical problems with the highest matching degree and their solutions to the data entry personnel for reference.

[0035] Step S2: Standardized Closed-Loop Management. Based on the structured problem dataset and initial status identifiers, a multi-stage status flow is executed, including problem analysis and responsibility allocation, rectification implementation, verification and confirmation, and problem closure. Problem status identifiers are dynamically updated to obtain real-time status data and processing records for each problem. When a problem approaches or exceeds its planned closure date, a reminder notification is automatically sent via email or enterprise communication tools, and is tiered according to escalation strategies: a 3-day advance reminder, daily reminders for overdue issues, and an escalation notification to superiors for issues overdue by 7 days.

[0036] like Figure 3 As shown, in this embodiment, the closed-loop management process includes 5 stages: Phase 1 (Problem Discovery and Entry): After discovering a problem during testing, testers enter it in a standardized manner according to the structured classification system, filling in the required fields such as problem title, category, and level. The system automatically records the discovery date and generates a unique problem number.

[0037] Phase Two (Problem Analysis and Responsibility Assignment): Quality management personnel analyze and confirm the entered problems, verify the problem level, identify the responsible unit and person, set a planned closure date, and the problem status changes from "Pending Analysis" to "In Progress".

[0038] Phase 3 (Implementation of Rectification): The responsible unit conducts investigation and rectification based on the nature of the problem, and fills in the root cause analysis, corrective measures and preventive measures.

[0039] Phase 4 (Verification and Confirmation): After rectification is completed, the issue status changes to "Verification in Progress," where testers verify the effectiveness of the rectification. If verification passes, the process moves to the closure phase; if verification fails, the process returns to the rectification phase.

[0040] Phase 5 (Issue Closure): After successful verification, the issue status changes to "Closed", and the system automatically records the actual closure date.

[0041] The automated reminder notification mechanism sends email notifications via the SMTP protocol and pushes messages to enterprise communication tools via a Webhook interface. The reminder escalation strategy is as follows: a reminder notification is sent 3 days before the planned closing date, daily reminder notifications are sent after the deadline, and an escalation notification is sent to superiors 7 days after the deadline.

[0042] Step S3: Automated Statistical Analysis. Based on real-time status data and processing records, preliminary statistical analysis results are automatically generated, including: problem status statistics (quantity and percentage), problem level distribution, problem category distribution analysis, statistics of responsible units and closure rate, and key indicator dashboards (including total number of problems, closure rate, number of serious problems, number of overdue unclosed problems, and average closure cycle).

[0043] The system automatically calculates and displays the following 5 statistical dimensions: (1) Problem status statistics: The number and percentage of problems in the four statuses of pending analysis, rectification, verification and closure are counted respectively.

[0044] (2) Problem level distribution: count the number of problems at four levels: fatal, serious, general, and minor.

[0045] (3) Problem Classification and Distribution Analysis: Count the number and proportion of problems according to the primary classification to identify high-frequency problem categories. In practice, Pareto analysis and other methods can be used to sort the classification data and calculate the cumulative proportion.

[0046] (4) Statistics of responsible units: The total number of problems, the number of closed problems, and the closure rate are counted by responsible unit.

[0047] (5) Key Indicator Dashboard: Displays the total number of issues, overall closure rate, number of serious and above issues, number of overdue issues, and average closure cycle.

[0048] Step S4: AI-assisted deep analysis. Extract the real-time status data, processing records, and preliminary statistical analysis results. Construct prompt words according to a preset template, including a data overview area, a problem list area, and an analysis requirement area. Send these prompts to the large language model service via an interface for analysis, obtaining analysis results including trend analysis, root cause clustering, recurring problem identification, risk warning, and improvement suggestions.

[0049] like Figure 4 As shown, in this embodiment, AI-assisted deep analysis includes the following steps: Step 1 (Data Extraction): Extract the structured field data of all problem records from the structured database, as well as the key statistical indicators generated in step S3.

[0050] Step Two (Prompt Term Construction): Organize the data according to the preset prompt term template. The template contains three areas: a data overview area, where you fill in summary statistics such as the total number of questions, closure rate, and number of questions at each level; a question list area, where you fill in the structured field information for each question; and an analysis requirements area, which clearly requires the large language model to perform five analysis tasks: trend analysis, root cause clustering, duplicate question identification, risk warning, and improvement suggestion generation.

[0051] Step 3 (AI Analysis Execution): Send the constructed prompt words to the large language model service via the HTTPRESTAPI interface, set appropriate model parameters (such as temperature parameters, maximum output length, etc.), and obtain the analysis results.

[0052] Step 4 (Report Generation): Format the AI ​​analysis results according to the preset report template chapter structure to generate a standardized report containing chapters such as data overview, trend analysis, root cause analysis, risk warning, and improvement suggestions.

[0053] Step 5 (Report Distribution): Send the report as an email attachment to relevant personnel via the SMTP protocol, and simultaneously push a report summary to the company's communication tools via a Webhook interface. Periodic automatic execution can be achieved through cron scheduling.

[0054] Step S5: Quantitative assessment of factory / launch conditions. Based on the real-time status data, processing records, and risk warning items in the analysis results, a comprehensive assessment is conducted against preset quantitative criteria, and a quantitative assessment conclusion is output for factory review decision-making.

[0055] The quantitative assessment of factory / launch conditions is based on the following four criteria: (1) Have all critical and serious issues been closed? If there are any unclosed critical or serious issues, the assessment conclusion is "fail".

[0056] (2) Whether the overall problem closure rate reaches the preset threshold (e.g., 95%) - If the closure rate is lower than the threshold, the evaluation conclusion is "not passed".

[0057] (3) Is the number of overdue unclosed issues zero? If there are overdue unclosed issues, the reasons and risk assessments must be explained for each one.

[0058] (4) Are there any high-risk items that have not been eliminated in the AI ​​risk warning? If there are high-risk items, risk assessment and mitigation measures must be provided.

[0059] The evaluation results are presented in the form of quantitative data, providing the factory review committee with an objective basis for decision-making.

[0060] Step S6: Automatic generation of periodic reports. The system automatically extracts structured problem datasets, real-time status data, and analysis results from the system according to a preset cycle. It then calls a large language model to perform periodic summary analysis to generate standardized quality reports, which are automatically distributed to preset recipients via email protocols or enterprise communication tool interfaces. Step S7: Problem Knowledge Base Construction and Similar Problem Recommendation. When the problem status indicator transitions to "closed," the corresponding problem record is automatically archived to the problem knowledge base. The archived content includes at least complete structured data such as problem description, root cause, corrective measures, and preventive measures. When standardizing the input of quality issues discovered during satellite testing, the text describing the issues to be input is semantically matched with the issue knowledge base, and the historical issues with the highest matching degree and their solutions are retrieved and pushed to the system in real time for reference.

[0061] Figure 2 A block diagram of an artificial intelligence-based intelligent management system for satellite test quality is shown in one embodiment of the present invention.

[0062] like Figure 2 As shown, in this embodiment, the AI-based intelligent management system for satellite test quality includes: The structured data entry module 10 is configured to generate a structured question dataset and mark the initial state identifier. It has a built-in speech recognition engine and a large language model semantic analysis interface to convert speech signals into text and automatically recommend question categories and levels. The closed-loop process management module 20 is configured to perform multi-stage state transitions and responsibility allocation based on the structured problem dataset and initial state identifiers, dynamically update problem state identifiers, and obtain real-time state data and processing records. The automated statistical analysis module 30 is configured to automatically generate preliminary statistical analysis results; AI-assisted analysis module 40 is configured to extract the real-time status data, process records and preliminary statistical analysis results, construct prompt words and call the large language model service to perform analysis, and generate analysis results including trend analysis, root cause clustering, risk warning and improvement suggestions; The factory evaluation module 50 is configured to perform a comprehensive evaluation based on the real-time status data, processing records and risk warning items in the analysis results, and compare them with preset quantitative criteria, and output a quantitative evaluation conclusion. The periodic report generation module 60 is configured to automatically extract system operation data according to a preset period, call the AI ​​analysis interface to generate standardized quality reports, and automatically distribute the reports through email or enterprise communication tool interfaces. The problem knowledge base module 70 is configured to automatically perform experience archiving operations when the problem status changes to closed; and is configured with a semantic retrieval interface to trigger similar problem retrieval and historical solution recommendation in real time when the structured data entry module 10 receives a new problem; The cross-project correlation analysis module 80 is configured to identify common problems and track improvements.

[0063] In this embodiment, the system supports cross-model and cross-batch problem data aggregation and correlation analysis, specifically including: (1) Problem distribution comparison: Compare the number of problems, classification distribution and grade distribution of different models and batches to identify trend changes.

[0064] (2) Common problem identification: Through root cause clustering analysis, platform-level common problems that repeatedly occur in multiple models are identified.

[0065] (3) Improvement effect tracking: After a certain model implements design improvements for a certain type of problem, track the occurrence of such problems in subsequent models to verify whether the improvement measures are effective.

[0066] To improve the accuracy and stability of large language model analysis, this embodiment features a detailed design for the prompt word engineering. The system employs a strategy combining structured instructions with contextual constraints. In the data overview area, the system dynamically fills in macro-level indicators for the current testing phase; in the issue list area, the system anonymizes and formats the raw data, retaining only core fields relevant to root cause analysis to prevent context overflow; in the analysis requirements area, the system injects explicit constraint instructions, requiring the model to output trend identification, root cause clustering, duplicate issue prompts, risk warnings, and improvement suggestions. After the model returns, the system parses the returned structure, mapping the text content to front-end visual charts for an intuitive presentation of AI insights. This design ensures that AI output is not limited to textual descriptions but can directly drive management decisions.

[0067] In one embodiment of the invention, the quantitative criteria for factory evaluation are not static and rigid; the system supports dynamic baseline adjustments based on historical data. If the testing complexity of a certain satellite model is extremely high, the system allows the quality manager to fine-tune the closure rate threshold and add specific explanations, but "zeroing out fatal / serious problems" and "mitigation of high-risk items" are inviolable red lines. The vector retrieval algorithm of the knowledge base module adopts a hybrid retrieval strategy, combining keyword retrieval and semantic vector cosine similarity calculation. For technical terms, keyword matching is emphasized, while for phenomenon descriptions, semantic matching is emphasized. This hybrid strategy significantly improves the recall and accuracy of cross-model problem recommendations. Simultaneously, the system records the adoption rate and rectification effectiveness feedback for each recommended solution, used to continuously optimize the ranking weights of the knowledge base.

[0068] The computer device described in this invention can employ high-performance servers, edge computing nodes, or cloud virtualization instances. The processor can utilize a multi-core CPU paired with an AI accelerator card to support localized large-model inference or high-concurrency data computation. Storage includes, but is not limited to, RAM, solid-state drive arrays, and distributed cloud storage. The computer program is deployed using a microservice architecture and runs containerized within a cluster, ensuring high availability and elastic scalability. The computer-readable storage medium can be a server hard drive, solid-state drive, USB flash drive, optical disc, or cloud object storage bucket, persistently storing the instruction code, model weight files, structured database files, and prompt word template configuration files for the aforementioned steps. When the program on the storage medium is loaded and executed by the processor, the entire quality intelligent management process described in this invention can be fully reproduced.

[0069] Those skilled in the art should understand that the technical features in the above embodiments can be equivalently replaced or combined according to actual engineering needs. For example, the speech recognition engine can be replaced with any commercially available service; the large language model can be replaced with an open-source model and deployed in a local private environment to meet the data security and confidentiality requirements of the aerospace field; the reminder notification channel can be replaced with an SMS gateway or an intranet system; and the quantitative evaluation threshold can be configured in stages according to the technological maturity of different satellite models. As long as the core logic, data flow relationship, and final technical effect are consistent with the present invention, they all fall within the protection scope of the present invention.

[0070] Although various embodiments of the invention have been described above, it should be understood that they are presented by way of example only and not as limitations. It will be apparent to those skilled in the art that various combinations, modifications, and alterations can be made without departing from the spirit and scope of the invention. Therefore, the breadth and scope of the invention disclosed herein should not be limited by the exemplary embodiments disclosed above, but should be defined solely by the appended claims and their equivalents.

Claims

1. A satellite test quality intelligent management method based on artificial intelligence, characterized in that, include: Establish a multi-dimensional structured problem classification system, standardize the input of quality problems found during satellite testing, generate a structured problem dataset that includes problem classification, problem level, problem status, responsibility attribution, time node and cause analysis dimensions, and mark the initial status identifier; Based on the structured problem dataset and initial status identifiers, a multi-stage status transition is executed, including problem analysis and responsibility allocation, rectification implementation, verification and confirmation, and problem closure. The problem status identifiers are dynamically updated to obtain real-time status data and processing records for each problem. Based on real-time status data and processing records, preliminary statistical analysis results are automatically generated. Extract the real-time status data, processing records, and preliminary statistical analysis results. Construct prompt words including a data overview area, a problem list area, and an analysis requirement area according to a preset template. Send these prompt words to the large language model service via an interface to perform analysis and obtain analysis results including trend analysis, root cause clustering, recurring problem identification, risk warning, and improvement suggestions. Based on the real-time status data, processing records, and risk warning items in the analysis results, a comprehensive evaluation is conducted against preset quantitative criteria, and a quantitative evaluation conclusion is output for factory review decision-making.

2. The method according to claim 1, characterized in that, Also includes: The system automatically extracts structured problem datasets, real-time status data, and analysis results from the system according to a preset cycle, calls a large language model to perform periodic summary analysis to generate standardized quality reports, and automatically distributes them to preset recipients via email protocols or enterprise communication tool interfaces. When the issue status indicator is switched to closed, the corresponding issue record is automatically archived to the issue knowledge base. The archived content includes at least complete structured data such as issue description, root cause, corrective measures and preventive measures. When standardizing the input of quality issues discovered during satellite testing, the text describing the issues to be input is semantically matched with the issue knowledge base, and the historical issues with the highest matching degree and their solutions are retrieved and pushed to the system in real time for reference.

3. The method according to claim 1, characterized in that, The multi-dimensional structured problem classification system adopts a two-level classification structure. The first-level classification includes problems in the test preparation phase, hardware problems, software problems, interface documentation problems, overall design problems, and operational problems. Each primary category is further divided into corresponding secondary categories based on the satellite model and testing requirements. The standardized input includes text input or voice input. When voice input is used, the voice signal is converted into natural language descriptive text by the voice recognition engine. The large language model is called to perform semantic analysis on the descriptive text to automatically recommend primary category, secondary category and question level. The input is completed after manual confirmation or correction.

4. The method according to claim 1, characterized in that, When the issue approaches or exceeds the planned closure date, an automatic reminder notification mechanism is triggered. Send reminders via email protocols or enterprise communication tool interfaces, and implement a tiered escalation strategy: send a reminder 3 days before the planned closing date, send daily reminders after the deadline, and escalate the notification to the superior manager 7 days after the deadline.

5. The method according to claim 1, characterized in that, The preliminary statistical analysis results include the percentage of problem statuses, the distribution of problem levels, the distribution of problem categories, the closure rate of responsible units, the total number of problems, the overall closure rate, the number of serious and above problems, the number of problems that have not been closed within the time limit, and the average closure cycle.

6. The method according to claim 1, characterized in that, The preset quantification criteria include: All critical and serious issues have been closed; The overall issue closure rate has reached the preset threshold; The number of overdue unclosed issues is zero; and No high-risk items that have not been eliminated were found in the analysis results; If any criterion is not met, the assessment conclusion will be "not passed" or an additional risk assessment explanation will be required. The assessment results will be presented in the form of quantitative data.

7. An intelligent management system for satellite test quality based on artificial intelligence, characterized in that, include: The structured data entry module is configured to generate a structured problem dataset and mark the initial state identifier; The closed-loop process management module is configured to perform multi-stage state transitions and responsibility allocation based on the structured problem dataset and initial state identifiers, dynamically update problem state identifiers, and obtain real-time state data and processing records. An automated statistical analysis module is configured to automatically generate preliminary statistical analysis results. The AI-assisted analysis module is configured to extract the real-time status data, process records and preliminary statistical analysis results, construct prompt words and call the large language model service to perform analysis, and generate analysis results including trend analysis, root cause clustering, risk warning and improvement suggestions; The factory evaluation module is configured to perform a comprehensive evaluation based on the real-time status data, processing records, and risk warning items in the analysis results, against preset quantitative criteria, and output a quantitative evaluation conclusion.

8. The system according to claim 7, characterized in that, Also includes: The periodic report generation module is configured to automatically extract system operation data according to a preset period, call the AI ​​analysis interface to generate standardized quality reports, and automatically distribute the reports via email or enterprise communication tool interfaces. The question knowledge base module is configured to automatically perform experience archiving when a question's status changes to closed; and it is configured with a semantic retrieval interface to instantly trigger similar question retrieval and historical solution recommendations when the structured data entry module receives a new question. The cross-project correlation analysis module is configured to identify common problems and track improvements.

9. The system according to claim 7, characterized in that, The structured data entry module has a built-in speech recognition engine and a large language model semantic analysis interface, which is used to convert speech signals into text and automatically recommend question categories and levels.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the intelligent management method for satellite test quality based on artificial intelligence as described in any one of claims 1 to 6.

Images