Product quality problem resolution auto-generation method, apparatus, medium, and product

Abstract

The application discloses a product quality problem solution automatic generation method, equipment, medium and product, and relates to the field of data processing.The method comprises the following steps: determining a fault code and product type information of a current product quality problem as target fault code and target product type information, the fault code comprising a plurality of fault identification information, and the fault identification information comprising failure part information, failure position information and failure mode information; determining a first recommended solution set from a historical database based on the target fault code and the target product type information; when the data amount of the first recommended solution set is insufficient, determining a second recommended solution set from the historical database based on the fault identification information, the data amount of the second recommended solution set being greater than that of the first recommended solution set; and finally generating a quality problem recommended solution based on the second recommended solution set.The application can alleviate the problem of low product problem solution development efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a product quality problem solution automatic generation method, device, medium and product. BACKGROUND

[0002] In industrial manufacturing, product quality problems are often inevitable. If a product quality problem occurs, the quality problem person in charge needs to manually write a solution according to the responsibility module of the quality problem, the problem severity level and the specific problem content when formulating a solution to a quality problem according to the traditional method.

[0003] However, in many cases, the person in charge does not necessarily have experience in handling a specific problem, and there is no clear solution guidance for the handling method and the involved department. In this case, the formulated solution has great limitations and inaccuracy, and there is a great probability of being rejected in the subsequent solution audit. Moreover, repeated rejection and repeated formulation of solutions often occur, which greatly affects the efficiency of solving quality problems. SUMMARY

[0004] To solve the above technical problems and defects, the purpose of the present application is to provide a product quality problem solution automatic generation method, device, medium and product, which can alleviate the problem of low efficiency of product problem solution formulation.

[0005] To achieve the above purpose, in a first aspect, the present application provides a product quality problem solution automatic generation method, comprising: determining the fault code and product model information of the current product quality problem as target fault code and target product model information, the fault code comprising a plurality of fault identification information, and the fault identification information comprising failure part information, failure position information and failure mode information; determining a first recommended solution set from a historical database based on the target fault code and the target product model information, the historical database comprising a plurality of problem solution records, and the problem solution record comprising the model information of the problem product, the fault code and the corresponding problem solution method; when the data amount of the problem solution records in the first recommended solution set is less than a preset threshold, determining a second recommended solution set from the historical database based on the fault identification information corresponding to the target fault code, the data amount of the problem solution records in the second recommended solution set being greater than the data amount of the problem solution records in the first recommended solution set; and generating a quality problem recommended solution according to the target fault code, the target product model information and the problem solution records in the second recommended solution set.

[0006] The application filters a highly relevant first recommended solution set from a historical database by accurately matching target fault codes and product model information, preferentially recommends a high-frequency verified effective solution measure under the same vehicle model, directly reuses mature experience, and reduces manual trial and error costs. Secondly, when the first recommended set data is insufficient, the failure mode is expanded to the same type of problem solving records across vehicle models and components, and the recommended data is supplemented through multi-level sorting to ensure the diversity and applicability of the solution. In addition, when generating the final recommended solution, the system dynamically adapts the general operation steps in the historical solution to the specific parts and positions of the current fault, which not only retains the effectiveness of historical experience, but also avoids logical conflicts caused by parameter differences. The generation efficiency and accuracy of the solution are improved.

[0007] Optionally, in some embodiments, the quality problem recommended solution is generated according to the target fault code, the target product model information, and the problem solving records in the second recommended solution set, including: determining a solution text replacement placeholder of each target problem solving record according to the target fault code, the target product model information, and the plurality of target problem solving records in the second recommended solution set, the solution text replacement placeholder including a product model placeholder, a failure part placeholder, a failure position placeholder, and a failure mode placeholder; generating a plurality of solution information according to the solution text replacement placeholder corresponding to each target problem solving record, the solution information being used to represent the text content of the quality problem recommended solution.

[0008] By defining placeholders (such as product model, failure part, etc.) in the solution text, the above technical solution realizes accurate adaptation of historical maintenance solutions to current fault scenarios. This method dynamically binds general solution templates with specific fault parameters (part model, position), avoiding manual repeated writing of similar cases and improving solution generation efficiency; at the same time, the placeholder mechanism ensures that the recommended content strictly matches the fault code analysis result, preventing misoperation caused by parameter misplacement, such as replacing “replace [failure part]” with “replace automatic transmission seal ring”, making the suggestion directly executable.

[0009] Optionally, in some embodiments, after generating a plurality of solution instruction information according to the solution text replacement placeholder corresponding to each target problem solving record, it further includes: displaying the plurality of solution information to the user; and in response to a solution confirmation instruction of the user, determining a final quality problem recommended solution from the plurality of solution information.

[0010] By introducing the user confirmation link, the multiple solutions are selected by human, balancing the automatic recommendation and human experience judgment. This avoids the potential deviation of single algorithm decision (such as the high-frequency but unsuitable solution is mistakenly promoted), and improves the solution reliability through human-computer cooperation; at the same time, the user decision data can optimize the recommendation model in reverse, for example, the solution frequently adopted obtains higher weight in subsequent ranking, forming a data-driven continuous improvement cycle.

[0011] Optionally, in some embodiments, after determining the final quality problem recommended solution from the multiple solution information in response to the user's solution confirmation instruction, it further includes: filling the final quality problem recommended solution into the solution maintenance page of the current product quality problem to obtain a problem solution recommendation page; and in response to the user's editing operation on the problem solution recommendation page, adjusting the content in the problem solution recommendation page.

[0012] By adopting the above technical scheme, the final solution is embedded in the standardized maintenance page and editing is supported, realizing the visual management and flexible adjustment of the solution. The editing function allows maintenance personnel to fine-tune the solution (such as modifying the torque parameter or supplementing the detection step) according to the actual conditions on site (such as spare parts inventory, tool limitations), enhancing the landing adaptability of the solution; the maintenance page as the only information source ensures that the version is controllable and the revision record is traceable, avoiding the execution confusion caused by multiple versions of the solution.

[0013] Optionally, in some embodiments, after adjusting the content in the problem solution recommendation page in response to the user's editing operation on the problem solution recommendation page, it further includes: in response to the user's audit pass operation on the problem solution recommendation page, generating a new problem solution record based on the problem solution recommendation page after the audit pass; and storing the new problem solution record to a historical database.

[0014] By adopting the above technical scheme, the problem solution record after the audit is stored back to the historical database, constructing a self-enhanced knowledge accumulation system. The new record not only contains the original fault data, but also integrates the user's correction content (such as the optimized operation steps), so that the subsequent recommendation can absorb practical experience, especially for rare faults or new equipment, the system can quickly expand the solution coverage, reducing the impact of cold start problems on the recommendation quality.

[0015] Optionally, in some embodiments, determining the first recommended solution set from the historical database based on the target fault code and the target product model information comprises: querying all historical problem solving records containing the same fault code from the historical database according to the target fault code and the target product model information to obtain an initial data set; filtering the initial data set based on the target product model information to retain historical records matching the target product model information to obtain a same model data set; sorting the same model data set according to the total number of occurrences of the fault code of each problem solving record in the historical database to obtain a sorting result; and selecting the top N problem solving records in the sorting result to obtain the first recommended solution set.

[0016] By adopting the technical solution, the generation logic of the first recommended set is clear, and the solutions to the same model and high-frequency faults are preferentially filtered. Through the rigid condition of “same product model + fault code complete match”, it is ensured that the initial recommendation is highly consistent with the current device characteristics (such as the exclusive sealing design of a certain vehicle transmission), and the compatibility risk of cross-model adaptation is reduced. The sorting according to the historical frequency implicitly assumes that “high-frequency faults = mature solutions”, which improves the usability of the initial recommendation and shortens the user decision-making time.

[0017] Optionally, in some embodiments, determining the second recommended solution set from the historical database based on the fault identification information corresponding to the target fault code comprises: retrieving problem solving records containing the same failure mode from the historical database based on the target failure mode information in the target fault code to obtain an expanded data set; filtering problem solving records matching the target product model information from the expanded data set as first priority data, and problem solving records not filtered as second priority data; sorting the second priority data according to the historical occurrence frequency of the target failure mode information to obtain sorted second priority data; and according to the difference between the preset threshold and the current data amount of the first recommended solution set, cutting off a corresponding number of problem solving records from the first priority data and the sorted second priority data to obtain the second recommended solution set.

[0018] By adopting the technical solution, the recommendation range is expanded when the data is insufficient, and the coverage problem of long-tail faults is solved through failure mode matching and cross-model priority sorting. For example, when there are not enough cases of oil leakage of a new model of car, the same mode fault solution of other models of cars (such as oil leakage of a truck tank) is introduced, and then the solutions are recommended according to the model relevance and failure mode universality, which not only ensures the solution relevance (same model priority), but also uses cross-scene experience (high-frequency failure mode solution) to make up for the data gap and enhance the robustness of the system under sparse data.

[0019] In a second aspect, an embodiment of the present application provides an electronic device, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is configured to store computer program codes, the computer program codes comprise computer instructions, the one or more processors invoke the computer instructions to enable the electronic device to perform the method described in the first aspect or the second aspect, and any possible implementation manner of the first aspect or the second aspect.

[0020] In a third aspect, the present application provides a computer-readable storage medium comprising instructions that, when executed on the electronic device, cause the electronic device to perform the method described in the first aspect or the second aspect, and any possible implementation manner of the first aspect or the second aspect.

[0021] In a fourth aspect, the present application provides a computer program product comprising instructions that, when executed on the electronic device, cause the electronic device to perform the method described in the first aspect or the second aspect, and any possible implementation manner of the first aspect or the second aspect.

[0022] It can be understood that the electronic device provided in the second aspect, the storage medium provided in the third aspect, and the computer program product provided in the fourth aspect are all used to execute the method provided by the present application. Therefore, the beneficial effects that can be achieved are referred to the beneficial effects in the corresponding method, which will not be described here.

[0023] The one or more technical solutions provided by the present application have at least the following technical effects or advantages:

[0024] 1. Effectively solve the problem of data sparsity, and improve the coverage and practicality of the solution. Through the hierarchical progressive recommendation strategy (first accurate matching of the same type full-fault solution, and then expanding to the cross-type similar failure mode solution), the system can flexibly cope with the problem of insufficient data under new products or rare failure scenarios. For example, when a rare oil leakage fault occurs in a new model of passenger car, the system can call the general processing logic of the truck engine oil leakage, and replace the difference parameters (such as part model, installation position) by using the placeholder technology, which not only breaks through the limitation of historical data, but also avoids the failure of recommendation caused by no matching result, so that the solution coverage is improved.

[0025] 2. Significantly improve the efficiency and automation level of quality management. The technical solution connects the links of fault feature analysis, intelligent matching of historical data, dynamic template generation, etc., forming an end-to-end automated solution production chain. The process of manually searching the case library and writing adaptive solutions in the traditional mode is compressed into minutes of automatic output, with efficiency improved by more than 80%. For example, the system can analyze the failure part code in the fault code in real time, automatically associate the engine model and fuel system parameters, generate a maintenance solution containing part replacement steps and torque standards, and significantly shorten the enterprise fault response cycle.

[0026] 3. Enhance the consistency and transferability of technical logic across scenarios. Through structured placeholder technology and failure mode similarity determination rules, the system can accurately retain the core maintenance logic (such as seal installation process, pressure test method) when expanding cross-model solutions, and only replace model-related variables (such as part size, interface type). This "logic solidification, parameter variable" design not only avoids technical misuse risks in manual adaptation, but also enables the large-scale reuse of historical experience. For example, for different vehicle models with circuit short circuit problems, the system outputs solutions that include standard steps such as insulation detection and wire harness replacement, and only adjusts wire diameter specifications and fuse parameters according to the model, ensuring maintenance quality stability and reducing enterprise cross-model maintenance costs. BRIEF DESCRIPTION OF DRAWINGS

[0027] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and serve to explain the principles of the present application. It is readily apparent to one of ordinary skill in the art that the accompanying drawings, only some embodiments of which are depicted, can be used to obtain other drawings without deviating from the spirit of the present application. In the drawings:

[0028] Figure 1 is a flowchart of a product quality problem solution automatic generation method according to an embodiment of the present application;

[0029] Figure 2 is a schematic diagram of a solution text replacement placeholder according to an embodiment of the present application;

[0030] Figure 3 is a schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] The terminology used in the following description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the description of the invention, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It also will be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0032] The terms "first", "second", "third", "fourth", "fifth", "sixth", "seventh" and "eight" are used only to describe the purpose of the embodiments and are not intended to imply relative importance of the indicated technical features. Thus, the features defined with "first", "second", "third", "fourth", "fifth", "sixth", "seventh" and "eight" can explicitly or implicitly include one or more of the features, and the meaning of "a plurality of" is two or more, unless otherwise specified and limited in the description of the embodiments of the invention.

[0033] It should also be noted that, unless otherwise explicitly specified and limited, the terms "set", "connected", and the like in the embodiments of the invention should be interpreted broadly. For example, "connected" can be fixed connection, or detachable connection, or integral connection; can be mechanical connection, or electrical connection; can be direct connection, or indirect connection through intermediate medium; can be internal communication of two elements; can be wired communication connection, or wireless communication connection. For those skilled in the art, the specific meaning of the above terms in the invention can be understood according to the specific circumstances. The embodiments of the invention are described in detail below.

[0034] In the traditional method, the process of formulating a solution by the quality problem owner usually follows the following artificially dominated path: when a product quality problem occurs, the owner first obtains a description of the problem phenomenon (for example, "equipment oil leakage" "part fracture") through on-site feedback or detection reports, then manually classifies the problem into a responsible module (such as "transmission system" "sealing component") according to personal experience or simple classification rules, and divides the problem severity level (such as "urgent" "general") based on subjective judgment. Then, the owner needs to retrieve the processing records of similar problems one by one in scattered historical work order records, Excel tables or paper documents, for example, manually screening potential reference cases in unstructured text descriptions by keywords (such as "leakage" "seal"). If some relevant records are found, the owner needs to analyze the solution measures in the cases (such as "replace the seal ring model" "adjust the assembly torque") one by one, and manually modify the wording to adapt to the new situation combined with the equipment model, supplier batch and other information of the current problem. If there is a lack of direct reference cases, the owner may obtain scattered suggestions through cross-department meetings or oral inquiries from experienced colleagues, and then integrate these fragmented information into a preliminary solution. Since the historical data lacks unified structured storage (such as not classified by fault components, modes, environments, etc.), and the solution description is mostly natural language text, the owner has difficulty in quickly locating effective information, and the final solution often has problems such as incomplete measures (such as not associating specific part numbers), incomplete responsibility departments (such as missing quality inspection links) or logical contradictions (such as repair steps conflicting with equipment parameters), leading to frequent rejection in the review stage, forcing the owner to repeatedly revise or even redevelop the solution.

[0035] This process highly depends on individual experience and manual cross-validation, and cannot systematically reuse implicit knowledge, becoming a bottleneck for both efficiency and accuracy.

[0036] Therefore, the embodiments of the present application provide a product quality problem solution automatic generation method, which significantly improves the generation efficiency and accuracy of the quality problem solution by introducing structured data processing and intelligent retrieval mechanism, effectively solving the core defects of strong experience dependence, low historical data utilization, poor solution adaptability and other problems in the traditional manual method.

[0037] Firstly, based on the structured analysis of fault codes (failed parts, failure positions, failure modes), the fault characteristics originally scattered in natural language descriptions are converted into machine recognizable multi-dimensional labels, completely eliminating the information misplacement caused by semantic ambiguity or expression difference in the traditional method. The system accurately filters out the solutions of the same model and the same fault from the historical database by automatically matching the target fault code with the product model, forming the first recommended set, ensuring the high consistency of the recommended results and the current problem in equipment model and fault mechanism, and avoiding the omission caused by cognitive limitations or keyword selection bias in manual retrieval.

[0038] Secondly, when the first recommended set data is insufficient, the system automatically expands the cross-product model search with the common feature of the failure mode, mines the solutions with the same failure mechanism but different application scenarios from the historical data, breaks through the bottleneck of data scarcity in the same model in the traditional method through the dynamic supplement mechanism, and realizes the multi-dimensional reuse of implicit experience. For example, for the sealing failure problem of the new model product for the first time, the system can automatically associate the historical solutions caused by similar sealing structure defects in other models to provide cross-platform reference basis for the person in charge. In addition, by presetting the threshold to control the size of the recommended set, the system filters redundant information while ensuring the diversity of solutions, solving the problem of information overload or data fragmentation when manually processing.

[0039] Finally, the recommended solution deeply integrates the current fault characteristics and historical solution logic, outputs standardized solutions through automatic template replacement (such as adapting product model parameters and failure part specifications), significantly reducing logical contradictions or key parameter errors in manual writing. More importantly, the system forms a closed loop of "data accumulation-intelligent recommendation-audit feedback", continuously optimizing the recommendation rules and data quality, while the traditional method cannot realize such self-iteration upgrade.

[0040] Through the integration of the above technologies, the embodiment of the present application shortens the solution development cycle, improves the audit pass rate, and promotes the transformation of enterprise quality knowledge from individual experience to systematic assets.

[0041] The product quality problem solving solution automatic generation method of the embodiment will be described below in combination with Figure 1 The product quality problem solving solution automatic generation method of the embodiment can be applied to a product quality problem solving solution automatic generation system (hereinafter referred to as the system), and the method comprises the following steps:

[0042] Step 101, the fault code and product model information of the current product quality problem are determined as the target fault code and target product model information.

[0043] Among them, the fault code includes multiple fault identification information, and the fault identification information includes failure part information, failure position information and failure mode information. The failure part information refers to the specific part name or number (such as bearing, sealing ring) that fails in the quality problem; the failure position information describes the installation position of the part in the equipment or product (such as the 3rd section of the transmission shaft, the interface of the housing); and the failure mode information represents the type of physical or functional abnormality (such as oil leakage, fracture, poor contact).

[0044] For example, assume that the current product quality problem is that the No. 6 flat car leaks oil, and the fault code is 171T311101011000G02M15, in which 171T311101011000 represents the failed part (automatic transmission with retarder assembly), G02 represents the failure location (transmission assembly), and M15 represents the failure mode (leakage).

[0045] Then, the target product model information is the No. 6 flat car, the target fault code is 171T311101011000G02M15, the corresponding failed part information is 171T311101011000, the failure location information is G02, and the failure mode information is M15.

[0046] Step 102, based on the target fault code and the target product model information, determine a first recommended solution set from the historical database.

[0047] The historical database includes a plurality of problem solving records, and each problem solving record includes the model information of a problem product, a fault code, and a corresponding problem solving method, such as “replace the transmission assembly sealing ring and perform pressure test to verify the sealing property”.

[0048] Specifically, the system takes the complete fault code as the exact matching condition, and filters all problem solving records with the same fault code from the historical database (hereinafter referred to as records). In this process, the system preferentially retains records that completely match the target product model information (such as the No. 6 flat car), and counts the frequency of occurrence of the same fault code under the same platform vehicle. If there are multiple records of the same fault code for the same vehicle model (for example, the No. 7 flat car + fault code 171T311101011000G02M15 occurs 10 times), the top 10 records are sorted in descending order of frequency of occurrence, and the top 10 records are selected as the first recommended solution set; if there are data of the same fault code for different vehicle models (for example, the No. 8 flat car occurs 8 times), the vehicle model priority (target vehicle model > other vehicle models) and the frequency of occurrence are sorted to supplement up to 10 records. For example, the 10 high-frequency records of the No. 7 flat car will be included in the recommendation, and the 8 records of the No. 8 flat car may be partially or completely excluded due to low frequency and non-matching vehicle model.

[0049] The specific implementation process of this step is as follows:

[0050] First, the exact matching screening is performed to query all records of the same fault code in the historical database under the condition of complete fault code. For example, if the target fault code is 171T311101011000G02M15 (automatic transmission with retarder assembly oil leakage), only the records with the same fault code (such as the same fault code data in flatbed No. 7 and flatbed No. 8) are extracted. Then, the vehicle model priority is divided, and the records completely consistent with the target product model (flatbed No. 6) are preferentially retained in the matching results.

[0051] If there is a record of the same fault code in the target vehicle model (target product model), it is preferentially displayed; if there is no record of the target vehicle model, it is included in the matching data of other vehicle models. For example, if the target vehicle model is flatbed No. 6 but there is no matching data, and flatbed No. 7 and No. 8 have records of the same fault code (10 times and 8 times, respectively), then the next step of screening is entered.

[0052] Subsequently, the frequency sorting and interception are performed, and the historical records of the same fault code are sorted in descending order according to the total frequency of the problem in the same vehicle model. For example, the problem solving records of flatbed No. 7 + fault code 171T311101011000G02M15 appear 10 times, and the records of flatbed No. 8 + the same fault code appear 8 times, so the priority of the 10 records of flatbed No. 7 is higher. The system will sort the records from high to low according to the number of records, and intercept the first 10 as the first recommended solution set.

[0053] If the frequency records of the same vehicle model are sufficient (such as 10 records of flatbed No. 7), they are directly included in the recommended set, and other vehicle model data (such as 8 records of flatbed No. 8) are excluded due to low frequency and vehicle model mismatch; if the data of the same vehicle model is insufficient (such as only 3 records), the records with higher frequency in other vehicle models are supplemented, but the total number does not exceed 10.

[0054] When the data amount of the problem solving records in the first recommended solution set is less than the preset threshold (for example, 10), the first recommended solution set is recommended to the user for reference.

[0055] Among them, the first recommended solution set is generated according to the principle of target vehicle model priority and high frequency priority. For example, if the target vehicle model (flatbed No. 6) has no data, and flatbed No. 7 and No. 8 have 10 and 8 records respectively, the first recommended solution set will include all 10 high-frequency records of flatbed No. 7, and the data of flatbed No. 8 is not included due to insufficient frequency and vehicle model priority.

[0056] In this embodiment, the historical database is constructed as follows: First, a large number of records of past product quality problems are collected, which should include the model information of the problem product, the fault code and the corresponding problem solving method. Then, the database structure is designed to ensure that the above information can be stored and support accurate matching query by fault code and product model. Next, the collected data is entered into the database and indexed to improve query efficiency. At the same time, the database needs to be updated regularly, adding new problem solving records, and cleaning and verifying the data to ensure the accuracy and integrity of the data. In addition, data classification and labeling can also be considered to expand the query range through fuzzy matching or partial matching when the data volume is insufficient. Finally, data access and management mechanisms are established to ensure the security and availability of the database.

[0057] Step 103, when the data volume of problem solving records in the first recommended solution set is less than the preset threshold, determining a second recommended solution set from the historical database based on the fault identification information corresponding to the target fault code.

[0058] Among them, the data volume of problem solving records in the second recommended solution set is greater than that in the first recommended solution set. The data volume of the second recommended solution set is larger because the screening condition of the first recommended solution set is based on strict matching of complete fault code + same model, resulting in insufficient data volume after screening. Therefore, the screening condition is relaxed to only failure mode matching + cross-model compatibility, thereby including more historical records (such as solutions for different vehicle models, different components but the same failure mode), and supplementing the number through frequency sorting, so that the data volume of problem solving records in the second recommended solution set is more sufficient.

[0059] When the data volume of the first recommended set is insufficient, the system extracts the failure mode in the target fault code (such as M15 corresponding to "leakage"), relaxes the screening condition, and retrieves all records with the same failure mode from the historical database (without limiting platform vehicle model and complete fault code). For example, if the target failure mode is "leakage", records containing different vehicle models (flatbed No. 6, No. 7, No. 8) or different failure parts / locations but the same failure mode "leakage" are included in the candidate. Subsequently, the system performs multi-level sorting according to vehicle model matching degree and failure mode frequency: preferentially retaining records identical to the target vehicle model (flatbed No. 6), and then arranging other vehicle model records in descending order of failure mode frequency. Finally, the data is supplemented to the preset number (such as 10) according to the sorting result, for example, supplemented until the total amount meets the standard. If it is still insufficient, stop supplementing and only use the existing data for recommendation.

[0060] In some embodiments, the specific implementation process of this step is as follows:

[0061] When the first recommended set data amount is less than the preset threshold (e.g., less than 10), the system expands and filters a second recommended set from the historical database based on the failure mode in the target fault code (e.g., M15 corresponds to "oil leakage").

[0062] First, the failure mode is extracted, the target fault code is parsed, the last 3-bit failure mode code (e.g., M15) is extracted and mapped to a specific problem type (e.g., oil leakage). For example, the failure mode of the target fault code 171T311101011000G02M15 is "oil leakage".

[0063] Then, fuzzy matching filtering is performed, the failure mode is used as a condition, the filtering rules are relaxed, and all records with the same failure mode in the historical database are retrieved, without limiting the platform vehicle model and complete fault code. For example, if the failure mode is "oil leakage", records containing different vehicle models (flatbed No. 6 vehicle, No. 7 vehicle, No. 8 vehicle), different failure parts (e.g., chassis assembly, transmission assembly), or different failure positions (e.g., oil tank, sealing ring) but the failure mode is the same "oil leakage" are included in the candidate.

[0064] Next, multi-level priority sorting is performed, the first priority is the same vehicle model record, and the record with the same target vehicle model (flatbed No. 6 vehicle) is preferentially retained. For example, if there is a record of oil leakage of other components of flatbed No. 6 vehicle (e.g., engine oil leakage) in the historical library, even if the fault code is different (due to the difference in failure position or part), it will be preferentially extracted. The second priority is the descending order of failure mode frequency, and the records of non-target vehicle models are sorted in descending order of failure mode frequency in the historical library. For example, flatbed No. 8 vehicle has 3 records of oil leakage, and flatbed No. 7 vehicle has 2 records, so the 3 records of flatbed No. 8 vehicle are sorted more in front. Finally, data is supplemented and intercepted, after sorting according to the above priority rules, the system intercepts data from high to low until the preset number (e.g., 10) is supplemented.

[0065] For example, if the first recommended set has only 3 data, the second recommended set needs to be supplemented by 7: assuming that there are the following data in the historical library: ① one record of oil leakage of flatbed No. 6 vehicle (vehicle model matching, highest priority); ③ three records of oil leakage of flatbed No. 8 vehicle (frequency 3 times, second priority); ② two records of oil leakage of flatbed No. 7 vehicle (frequency 2 times, lowest priority).

[0066] The system is included in order of ranking ① > ③ > ②, and when 7 pieces are supplemented, 1 piece of ①, 3 pieces of ③, and 2 pieces of ② are included, and the remaining 1 piece of vacancy is not supplemented. The final second recommended set contains cross-model, cross-component, same-type failure mode solutions. For example, it may contain records such as "Flat No. 6 engine oil leakage: replace cylinder gasket" and "Flat No. 8 oil tank oil leakage: tighten the interface". Although these solutions are not completely consistent with the target fault code, they can be effective supplements because the failure modes are the same and have been prioritized. If it is still less than 10 pieces after supplementing, the system only outputs the current data and does not further expand.

[0067] In step 104, the quality problem recommended solution is generated according to the target fault code, the target product model information, and the problem solving records in the second recommended solution set.

[0068] Specifically, first, the target fault code is parsed to determine the failure part name (such as automatic transmission with retarder assembly), failure location name (such as transmission assembly), failure mode name (such as oil leakage), and target product model (such as Flat No. 6 car).

[0069] Subsequently, historical problem solving records are extracted from the second recommended solution set one by one, and content adaptation is performed on the solution text in each record: through text recognition technology, the general description fragments (such as "replace [component]" and "detect [problem type]") related to the failure part, failure location, and failure mode in the solution are identified, and the general terms (such as "[component]" and "[problem type]") in these fragments are replaced with the specific names of the current fault (such as "transmission assembly" and "oil leakage"). For example, if the solution in a certain historical record is "replace the oil tank and perform oil leakage detection", the system will identify "oil tank" as the same attribute of the failure location name (belonging to the "sealing component" category) through the pre-defined component-problem type mapping table, and then replace it with the current failure location name "transmission assembly" to form the adapted solution "replace the transmission assembly and perform oil leakage detection".

[0070] Finally, all the adapted solutions are output according to the original ranking of the second recommended set (same model first, failure mode frequency descending), while filtering out logically conflicting content (such as original solutions containing operation steps unrelated to the current fault), to ensure that the recommended content strictly matches the failure mode (oil leakage) and component type (transmission related) of the target fault, and finally generate quality problem recommended solutions that can be directly applied.

[0071] The embodiment adopts the above method steps, and improves the efficiency and accuracy of solution generation by intelligently matching historical data and key features of current quality problems. First, by accurately matching the target fault code (including failed parts, failure location, failure mode) and product model information, a highly relevant first recommended solution set is screened from the historical database, and the high-frequency verified effective solution measures under the same vehicle type are preferentially recommended, the mature experience is directly reused, and the manual trial and error cost is reduced. Second, when the first recommended set data is insufficient, based on the failure mode (such as oil leakage), it is extended to the same type of problem solving records across vehicle types and components (such as engine oil leakage of Model 6 and oil tank oil leakage of Model 8), and through multi-level sorting (same vehicle type first, failure mode frequency descending order), the recommended data is supplemented to ensure the diversity and applicability of the solution. In addition, when generating the final recommended solution, the system dynamically adapts the general operation steps in the historical solution (such as "replace the sealing part" and "perform oil leakage detection") to the specific parts and locations of the current fault (such as the automatic transmission with retarder assembly), which not only retains the effectiveness of historical experience, but also avoids logical conflicts caused by parameter differences.

[0072] The embodiment not only effectively shortens the problem solving time, but also improves the first-time audit pass rate of the solution through data-driven priority rules, and continuously feeds back the accumulated solutions to the historical database, forming a closed-loop optimization, building a standardized and reusable quality problem knowledge base for the enterprise, and long-term reducing the quality control cost and promoting the product reliability improvement.

[0073] In some embodiments, the method of the embodiment specifically includes the following steps:

[0074] S201, the fault code and product model information of the current product quality problem are determined as the target fault code and target product model information.

[0075] This step can refer to the description of the foregoing embodiments, which will not be repeated here.

[0076] S202, according to the target fault code and the target product model information, all historical problem solving records containing the same fault code are queried from the historical database to obtain an initial data set.

[0077] Specifically, the system performs an exact query operation in the historical database according to the target fault code (such as 171T311101011000G02M15) and the target product model information (such as the flatbed No. 6 vehicle), extracts all historical problem solving records containing the same complete fault code, and forms an initial data set. For example, if the target fault code is "171T311101011000G02M15" (automatic transmission with retarder assembly oil leakage), only the records with the same fault code (such as the records of the flatbed No. 7 vehicle and the flatbed No. 8 vehicle) are retained, and other data that does not match the code is excluded.

[0078] This step is implemented by exact matching of the database (such as WHERE fault code = target value in SQL), which ensures that the initial data set is completely consistent with the technical features of the current fault, providing a high-relevance data basis for subsequent recommendations.

[0079] S203, based on the target product model information, the initial data set is screened to retain the historical records matching the target product model information, and a same model data set is obtained.

[0080] On the basis of the initial data set, the system further screens out historical records that completely match the target product model information (such as the flatbed No. 6 vehicle) to generate a same model data set. For example, if the initial data set contains records of the flatbed No. 6 vehicle, the flatbed No. 7 vehicle, and the flatbed No. 8 vehicle, only the record of the flatbed No. 6 vehicle is retained, and other vehicle type data is excluded.

[0081] This step is implemented by field matching (such as WHERE model information = target model), and the purpose is to preferentially reuse solutions for the same vehicle model, because the design, component configuration, and problem patterns of the same model product are closer, and the adaptability and success rate of the solutions are higher. If the same model data is empty (such as no matching record for the flatbed No. 6 vehicle), this step is skipped and the cross-model data processing is directly entered.

[0082] S204, according to the total number of times each problem solving record in the same model data set appears in the historical database, the sorting result is obtained.

[0083] For each record in the same model data set, the system can count the total number of times the corresponding fault code appears in the historical database, and sort them from high to low according to the number of times. For example, if the fault code 171T311101011000G02M15 of the flatbed No. 6 vehicle corresponds to a record that appears 8 times in the historical database, and another same model record appears 5 times, the record with 8 times is sorted more in front. This sorting rule is based on the assumption that "high frequency = high reliability", and believes that solutions that have been repeatedly verified are more likely to be effective.

[0084] Technically, the sorting is completed by pre-computed frequency statistics table or dynamic aggregation query (such as SQL COUNT and ORDER BY), ensuring real-time and accurate results.

[0085] S205, select the top N problem solving records in the sorting result to obtain the first recommended solution set.

[0086] Specifically, the system generates the first recommended solution set by selecting the top N (e.g. N=10) problem solving records according to the sorting result. For example, if the sorted same-model data set contains 15 records, the top 10 records with the highest frequency are selected; if the data volume is less than N (e.g. only 5), all are included in the first recommended solution set to ensure that the data volume in the first recommended solution set is sufficient.

[0087] This step is implemented by paging query (such as SQL LIMIT N). The final output of the recommended set meets the same model matching and high frequency priority principle, and takes into account the adaptability and reliability of the solution, providing high-quality candidate data for subsequent manual review or automatic processing.

[0088] When the data volume of problem solving records in the first recommended solution set is greater than or equal to the preset threshold, the system pushes the first recommended solution set to the user for reference.

[0089] When the data volume of problem solving records in the first recommended solution set is less than the preset threshold, step 206 is entered.

[0090] S206, based on the target failure mode information in the target fault code, retrieve the problem solving records containing the same failure mode from the historical database to obtain an expanded data set.

[0091] Specifically, the system extracts the failure mode information (such as M15 corresponding to "leakage") in the target fault code, and retrieves all records containing the same failure mode (e.g. "leakage") in the historical database to form an expanded data set.

[0092] This step is implemented by fuzzy matching: ignoring the failure parts and location information (such as the first 19 characters) in the fault code, and only using the failure mode code (last 3 characters) as the screening condition. For example, if the target failure mode is leakage (M15), the expanded data set contains all records with the same failure mode (leakage) of different vehicle models (such as flat plate No. 6, No. 7, No. 8), different failure parts (such as transmission, fuel tank), even if their complete fault codes are different (such as 171T311101012000A03M15).

[0093] S207, filter out the problem solving records matching the target product model information from the extended dataset as first priority data, and the rest as second priority data.

[0094] Specifically, from the extended dataset, the system filters out the records matching the target product model information (e.g. flatbed No. 6) as first priority data (e.g. records of flatbed No. 6 engine oil leakage), and the rest as second priority data (e.g. flatbed No. 7 fuel tank leakage, flatbed No. 8 seal ring leakage).

[0095] This step is achieved by field matching (e.g. WHERE model information = target model), ensuring that solutions for the same model are recommended first, as they are more similar in component configuration and design to the current problem, and are more adaptable. If the first priority data is empty (e.g. no oil leakage records for flatbed No. 6), proceed directly to the second priority processing.

[0096] S208, sort the second priority data according to the historical occurrence frequency of the target failure mode information, to obtain sorted second priority data.

[0097] For the second priority data (records of non-target vehicle models), the system can count the total frequency of each record's failure mode in the historical database and sort them from high to low frequency. For example, if the flatbed No. 8 oil leakage problem occurs 3 times and the flatbed No. 7 occurs 2 times, the 3 records of flatbed No. 8 are sorted more forward.

[0098] This sorting is based on the principle of "high frequency = high universality", considering that solutions that occur frequently across vehicle models may have more widespread effectiveness. Technically, this sorting is achieved through pre-computed frequency statistics or dynamic aggregation queries (such as SQL COUNT and ORDER BY).

[0099] S209, according to the difference between the preset threshold and the current data volume of the first recommended solution set, extract a corresponding number of problem solving records from the first priority data and the sorted second priority data, to obtain the second recommended solution set.

[0100] Specifically, the system extracts data according to the difference between the preset threshold (e.g. total recommended number 10) and the current data volume of the first recommended set (e.g. the first recommended set already has 3, and needs to be supplemented by 7), in priority order: first, all first priority data (e.g. 1 record of flatbed No. 6 oil leakage) is included, if still insufficient, then from the sorted second priority data, supplement the remaining number in descending order of frequency (e.g. take 3 records of flatbed No. 8 oil leakage, 2 records of flatbed No. 7 oil leakage, a total of 6 records, the total recommended set is 3+1+6=10 records). If the data volume is still insufficient, only the records obtained are retained.

[0101] This step is achieved through conditional judgment (such as IF…ELSE) and pagination query (such as LIMIT), ultimately ensuring that the data volume of the second recommended set is greater than the first recommended set, and the recommended content meets both vehicle model adaptability and failure mode experience reuse.

[0102] S210, according to the target fault code, the target product model information and the second recommended solution set, determine the solution text replacement placeholder of each target problem solving record.

[0103] Among them, the solution text replacement placeholder includes product model placeholder, failure part placeholder, failure location placeholder and failure mode placeholder.

[0104] The system replaces the placeholders in each historical solution text in the second recommended set according to the target fault code and the target product model information. First, parse the fault code to extract the specific name of the failure part, location and mode (for example, the failure part is “automatic transmission with retarder assembly”), and determine the current product model (such as flat plate No. 6 car); Then locate the general placeholder field (such as “[failure location]” and “[failure mode]”) in the historical solution text and replace it with the corresponding information of the target fault (such as “transmission assembly” and “oil leakage”), for example, “replace [failure location] sealing ring” is converted to “replace transmission assembly sealing ring”. This process is achieved through predefined replacement rules and string operations, ensuring that the solution content matches the current problem accurately.

[0105] For example, as shown in Figure 2 The solution text is “flat plate No. 8 car chassis assembly oil tank leaks, suggest replacing the oil tank and adjusting the sealing”, the system replaces the placeholders by the following rules:

[0106] Placeholder A (product model placeholder): replace the platform model “flat plate No. 8 car” with placeholder A;

[0107] Placeholder B (failure part placeholder): replace the failure part name “oil tank” with placeholder B;

[0108] Placeholder C (failure location placeholder): replace the failure location name “chassis assembly” with placeholder C;

[0109] Placeholder D (failure mode placeholder): according to the context, infer the failure mode as “leakage”, and replace it with placeholder D.

[0110] S211, according to the solution text replacement placeholder corresponding to each target problem solving record, generate multiple pieces of solution information, which are used to represent the text content of the recommended solution of the quality problem.

[0111] After the placeholder replacement is completed, the system sorts the processed multiple pieces of solution text according to the priority of the second recommended set (same type priority, failure mode frequency descending order) and integrates the sorted solution text into a final recommended solution list.

[0112] For example, the generated solutions "adjust the transmission assembly bolt to solve the oil leakage" and "replace the automatic transmission with a retarder assembly seal" are output according to the original data sorting, and the content conflicting with the current failure logic (such as the operation steps involving irrelevant failure modes) is automatically filtered. The generated solution list is directly associated with the target failure characteristics, reduces manual secondary modification, and realizes standardized and rapid output of solutions.

[0113] For example, after the placeholder replacement is completed, the solution information can be "ABCD, suggest replacing C and adjusting the sealing".

[0114] This embodiment realizes the generalization and structured reuse of the solution template. Through the standardization of the placeholders (A = platform vehicle, B = failed part, C = failure location, D = failure mode), the system abstracts the specific failure description (such as "flat plate No. 8 vehicle chassis assembly oil tank leakage") into a symbolic combination with clear logical relationship (ABCD), so that the same solution template can be dynamically adapted to different scenarios. For example, when processing "flat plate No. 6 vehicle transmission assembly seal ring oil leakage", only the placeholders need to be replaced with A (flat plate No. 6 vehicle), B (seal ring), C (transmission assembly), and D (oil leakage), and the adaptive solution "flat plate No. 6 vehicle transmission assembly seal ring oil leakage, suggest replacing the transmission assembly and adjusting the sealing" can be generated. This design not only retains the core operation logic of the historical solution (replace the failed part at the failure location + adjust the sealing), but also isolates the specific parameters through the placeholders, avoiding the non-reusable template caused by the differences in part models and locations, thereby significantly improving the cross-scenario applicability of the solution and reducing the cost of manual writing of repetitive content.

[0115] S212, display multiple pieces of solution information to the user.

[0116] Specifically, the system sorts the processed multiple pieces of solution information according to the priority and displays them to the user. For example, the solution text "flat plate No. 8 vehicle chassis assembly oil tank leakage, suggest replacing the oil tank and adjusting the sealing" is replaced by the placeholder to generate the general template "ABCD, suggest replacing C and adjusting the sealing".

[0117] When dealing with the current quality problem (such as the transmission assembly seal ring oil leakage of the flatbed No. 6 vehicle), the system replaces the placeholders A / B / C / D in the solution information with the target information (A = flatbed No. 6 vehicle, B = seal ring, C = transmission assembly, D = oil leakage) respectively, generates an adaptive solution: "The transmission assembly seal ring of the flatbed No. 6 vehicle leaks oil, and it is recommended to replace the seal ring and conduct a sealing adjustment".

[0118] At the same time, the system extracts other similar failure mode (such as oil leakage) templates (for example "adjust C connecting bolt to solve D") from the historical database, and after replacement, generates the second solution: "The transmission assembly seal ring of the flatbed No. 6 vehicle leaks oil, and it is recommended to adjust the connecting bolt of the transmission assembly to solve the oil leakage".

[0119] Finally, the system displays multiple solutions in the form of a list to the user according to the rules of the same model priority and failure mode frequency descending order.

[0120] S213, in response to the user's solution confirmation instruction, determines the final quality problem recommended solution from the multiple solution information.

[0121] Specifically, after the user selects or confirms one of the solutions through the interface, the system executes the final solution binding according to the user's instruction. For example, if the user selects the first solution "replace the seal ring and conduct a sealing adjustment", the system associates the current fault code (171T311101011000G02M15), product model (flatbed No. 6 vehicle) and operation steps (replace the seal ring, sealing detection) of the solution, and generates a standardized solution document containing complete fault analysis (failed parts, location, mode) and specific operation process. At the same time, the system appends this confirmation record to the historical database and updates the frequency statistics of the fault code to provide data support for subsequent recommendations.

[0122] S214, fill the final quality problem recommended solution into the solution maintenance page of the current product quality problem to obtain a problem solution recommendation page.

[0123] Specifically, the system fills the structured template of the final recommended solution into the solution maintenance page of the product quality problem. For example, the original problem "the oil tank of the flatbed No. 8 vehicle chassis assembly leaks oil, and it is recommended to replace the oil tank and conduct a sealing adjustment" generates a general template "A's D appears B at C, and it is recommended to replace B and conduct a sealing adjustment" after replacing the placeholders (A = platform vehicle placeholder, B = failed part placeholder, C = failure location placeholder, D = failure mode placeholder).

[0124] At this time, the system dynamically binds the placeholders to specific parameters according to the analysis results of the current quality problem (such as the oil leakage of the transmission assembly seal ring of truck No. 6), replaces A with "truck No. 6", B with "seal ring", C with "transmission assembly", and D with "oil leakage", generates the adapted solution text "The oil leakage of truck No. 6 occurs at the seal ring of the transmission assembly, and it is recommended to replace the seal ring and perform sealing adjustment", and automatically fills this content into the edit box of the solution maintenance page for the user to view and confirm.

[0125] S215, in response to the user's editing operation on the problem solution recommendation page, adjusts the content in the problem solution recommendation page.

[0126] Specifically, when the user edits the recommended content on the solution maintenance page (for example, modifies the text to "replace the transmission assembly seal ring and perform three pressure tests"), the system listens to the user's input actions in real time, synchronizes the modified content to the memory variable through front-end event capture (such as onChange event), and triggers the re-rendering of the page DOM element to display the updated text.

[0127] If the user adjusts the format (such as adding a segment symbol or highlighting a keyword), the system parses the operation instructions through a rich text editor component (such as Quill.js) and stores the format markers (such as HTML tags or CSS styles) persistently. All editing operations will not affect the logical binding relationship of the original placeholders (such as A / B / C / D still associated with the fault code analysis results), only the display layer text is personalized adjusted, and the final user confirmed solution will contain the edited content and the original fault parameters, forming a complete record that can be traced.

[0128] When the user edits the problem solution recommendation page (for example, modifies the text description, adjusts the step order, or supplements the operation details), the system captures the user's input behavior in real time through the event listening module of the front-end interface, synchronizes the modified content to the temporary data structure (such as JSON object) in memory, and dynamically updates the page DOM element to display the adjusted effect immediately; if the user clicks Save, the system associates the edited complete solution text with the fault code of the current quality problem and the product model information, generates a new version record containing the original recommended solution and the user's modified content, while retaining the modification log (such as modification time, operator, and changed content difference), and returns the updated data to the server through the API interface, and stores it persistently in the "user modified solution" independent partition of the historical database, ensuring that subsequent queries or statistics can distinguish between system recommendations and manual optimization versions, while keeping the analysis results of the original fault parameters (failed parts, location, mode) unchanged, and only adjusting the operation description layer of the solution adaptively.

[0129] S216, in response to the user's audit pass operation on the problem solution recommendation page, generating a new problem solution record based on the audited problem solution recommendation page.

[0130] Specifically, when the user clicks the audit pass button, it can be considered that the user has audited the problem solution recommendation page, and the system extracts the complete content in the problem solution recommendation page (including the target fault code, product model, user-edited solution text, timestamp, operator information), and structures and encapsulates the difference part (such as the modified operation step "perform three pressure tests after replacing the sealing ring") of the original recommended solution, to generate a new problem solution record.

[0131] For example, the system combines the current fault code (171T311101011000G02M15), product model (Flat 6 car), final solution text ("replace the transmission assembly sealing ring and perform three pressure tests"), audit status (passed), version number (V2.1), etc. Field combination into JSON data object, and generate unique record identification (such as SHA-256 value) through hash algorithm, to ensure data integrity and traceability.

[0132] S217, store the new problem solution record to the historical database.

[0133] Specifically, the system stores the newly generated problem solution record to the specified data table (such as quality_solution_records) of the historical database through the database API interface (such as RESTful POST request).

[0134] When storing, the system performs the following operations:

[0135] 1) Data cleaning: check the fault code format (16+3+3), product model naming specification (such as "Flat X car");

[0136] 2) Field mapping: align the keys of the JSON object with the column names of the database table (such as solution_text mapped to solutions.content field);

[0137] 3) Index optimization: establish a joint index on the fault code (fault_code) and product model (product_model) fields to speed up subsequent queries;

[0138] 4) Version control: if there is already a historical record for the same fault code, add a new record and associate the version chain (point to the original record through the parent_id field).

[0139] At the same time, the system writes user operation logs (such as "User ID 123 passed the audit on 2023-09-15 14:30") into a separate operation audit table (audit_logs), enabling full-link tracking of data modifications.

[0140] The implementation process of the method of the present embodiment will be described below in conjunction with a specific application example as follows:

[0141] (1) The key positioning content of the quality problem is the platform vehicle model (product model) + fault code (the fault code has a fixed splicing rule: failure part information (16 bits) + failure location information (3 bits) + failure mode information (3 bits)). The fault code is determined offline, which clearly indicates which component has a problem at what location.

[0142] For example: the platform vehicle model of a certain quality problem is flatbed No. 6, and the fault code is 171T311101011000G02M15. Then the fault code can be split into the following fault identification information according to the rules:

[0143] Failure part: 171T311101011000 (automatic transmission with retarder assembly);

[0144] Failure location: G02 (transmission assembly);

[0145] Failure mode: M15 (oil leakage).

[0146] Note: The failure parts, failure locations, and failure modes mentioned in the following text are obtained by intercepting according to this rule, and will not be described again.

[0147] The fault code is a pre-maintained basic data in the system, and there is a separate account page. You can query the corresponding code and name and other related data according to the splicing content.

[0148] According to the above platform vehicle model and fault code information, it can be initially confirmed that the problem is: flatbed No. 6 automatic transmission with retarder assembly transmission assembly oil leakage.

[0149] (2) Because different platform vehicles may have the same problem, i.e., different platform vehicles have the same fault code. At this time, the system will query the historical quality problem data (problem solving records) of the same fault code in the historical database based on the fault code of the current quality problem, and use the solution of the queried historical quality problem data as the recommended solution (the first recommended solution set) and perform the top-up operation (the recommended solution candidate data is up to 10).

[0150] Note: If the same fault code of the same platform vehicle model appears multiple times, the system will sort them by frequency, and the more frequent ones will be easier to be recommended.

[0151] For example:

[0152] Flatbed No. 7 vehicle + fault code 171T311101011000G02M15 appears 10 times,

[0153] Flatbed No. 8 vehicle + fault code 171T311101011000G02M15 appears 8 times,

[0154] So the 10 data of flatbed No. 7 vehicle + fault code 171T311101011000G02M15 will be recommended as the solution for subsequent operation, and other data will not be recommended.

[0155] (3) After the previous operation, if the number of recommended solution set is less than 10, the system will obtain the failure mode of the current problem according to the fault code splicing rule, query the historical quality problem data of the same failure mode, sort the obtained data by platform vehicle model > failure mode frequency, and obtain the problem solving record data as the recommended data. The recommended solution set will be supplemented to 10.

[0156] For example: the current quality problem platform vehicle is flatbed No. 6 vehicle, and the failure mode is oil leakage. The system obtains the data containing the following data by taking the failure mode as the query condition:

[0157] ① The historical quality problem data of flatbed No. 6 vehicle with failure mode of oil leakage (because of different failure positions, the failure code is different from the current quality problem, so it is not matched in the previous operation), the number is 1.

[0158] ② The historical quality problem data of flatbed No. 7 vehicle with failure mode of oil leakage, the number is 2.

[0159] ③ The historical quality problem data of flatbed No. 8 vehicle with failure mode of oil leakage, the number is 3.

[0160] Because the platform vehicle of ① is the same as the current problem, the data contained in ① is sorted as the optimal recommendation, and the frequency of ③ is higher than that of ②, so the data priority of ③ is higher than that of ②. According to the priority rule, the final priority sorting is ①>③>②, and the system will supplement the data with higher priority into the recommended solution set (at most 10) according to the sorting to obtain the second recommended solution set.

[0161] Note: If the number of candidate data in the second recommended solution set is less than 10 after this step, no further supplement will be performed, and only the currently obtained data will be used as recommended data for subsequent operations.

[0162] (4) The system will finally obtain the recommended solution data (up to 10) corresponding to the solution content processing, according to the fault code corresponding to each piece of data Split the obtained failure part code / name, failure location code / name, failure mode code / name and the platform vehicle corresponding to the problem, and replace the current quality problem platform vehicle, failure part code / name, failure location code / name, failure mode code / name information in the corresponding solution content.

[0163] For example:

[0164] a. The system obtains the data corresponding to the current quality problem (the quality problem for which the recommended solution is required):

[0165] Platform vehicle: flat plate No. 6 car;

[0166] Fault code: 171T311101011000G02M15 (according to the fault code splicing rule and fault code information query, the following data is obtained);

[0167] Failure part code: 171T311101011000, failure part name: automatic transmission with retarder assembly;

[0168] Failure location code: G02, failure location name: transmission assembly;

[0169] Failure mode code: M15, failure mode name: oil leakage.

[0170] b. The system obtains the data corresponding to one piece of quality problem in the recommended solution data:

[0171] Platform vehicle: flat plate No. 8 car;

[0172] Solution content: flat plate No. 8 car chassis assembly oil tank oil leakage, suggest replacing the oil tank and adjusting the sealing;

[0173] Fault code: 171T311101012000A03M15; according to the fault code splicing rule and fault code information query, the following data is obtained:

[0174] Failure part code: 171T311101012000, failure part name: chassis assembly;

[0175] Failure location code: A03, failure location name: oil tank;

[0176] Failure mode code: M15, failure mode name: oil leakage;

[0177] As Figure 2 shown: the system intercepts and removes the failure part, failure location, and failure mode information contained in the solution content.

[0178] Obtain content: ABCD, recommend replacing C and adjusting the sealing. (ABCD is the corresponding content placeholder explained in Figure 2 )

[0179] c. Replace the placeholders in the content obtained in step b with the current quality problem data obtained in step a, as follows:

[0180] A (platform vehicle placeholder) → flatbed No. 6 vehicle;

[0181] B (failure part placeholder) → automatic transmission with retarder assembly;

[0182] C (failure location placeholder) → transmission assembly;

[0183] D (failure mode placeholder) → oil leakage;

[0184] Obtain the final solution content: flatbed No. 6 vehicle automatic transmission with retarder assembly transmission assembly oil leakage, recommend replacing the transmission assembly and adjusting the sealing.

[0185] (5) When the user fills in the solution, the system will intelligently push the 10 recommended data obtained after the above operations (application pop-up window). The user can view the solution content obtained after the system's intelligent processing of each recommended data and the corresponding quality problem data information. After determining the appropriate recommended solution, the user can fill in the solution content obtained after the system's processing of the selected data into the current problem solution maintenance page by clicking the application button, and can manually adjust and supplement the filled-in solution content.

[0186] (6) When the solution to this problem is approved, the system will record the problem and the corresponding solution to the system's historical database. After data accumulation over time, the format of the solution content will be more template-based. When a new problem is recommended for intelligent solution, the screening data will be faster, the recommended solution will be more accurate, and the solution content will be more standardized.

[0187] In some embodiments, after generating a quality problem recommendation solution, the embodiment can also convert the quality problem recommendation solution into an AR (Augmented Reality, Augmented Reality) operation guide, superimpose virtual instructions (such as part disassembly sequence, torque parameter, dangerous area marking) through smart glasses or mobile devices. The specific implementation process is as follows:

[0188] Deploy SLAM (Simultaneous Localization and Mapping, Simultaneous Localization and Mapping) space positioning engine: Integrate visual-inertial SLAM algorithm in AR device (such as HoloLens), build three-dimensional point cloud map of maintenance environment in real time through camera and IMU data, accurately track the position and attitude of the device in the physical space (accuracy ±2mm). Among them, the visual-inertial SLAM algorithm is a real-time positioning and mapping technology that combines camera visual data and inertial measurement unit (IMU) information. It captures environmental feature points (such as device outlines, bolt positions) through the camera and calculates their relative motion, while using the angular velocity and acceleration data of the IMU to predict the instantaneous pose change of the device. Then, through Kalman filtering or graph optimization algorithm, the two types of data are aligned in space and time, making up for the limitations of a single sensor (such as tracking loss of the camera when moving quickly or lack of light, and cumulative error drift of the IMU). Finally, high-precision three-dimensional space positioning (error less than 2mm) and dynamic environment mapping of the device in the maintenance scene are realized, ensuring that AR virtual instructions (such as disassembly arrows, torque prompts) can be accurately superimposed on the real surface of the physical device, even in complex working conditions (such as personnel walking, tool blocking) to maintain stable and reliable virtual-real fusion effect.

[0189] Load device CAD model library: Convert the CAD engineering model of the target part (such as the transmission) into lightweight grid data (such as glTF format), associate metadata (part name, disassembly sequence, torque parameter) and preload to the local storage of the AR device.

[0190] Real-time CAD-physical space registration: Based on the point cloud map generated by SLAM, align the key feature points (such as bolt hole position, edge contour) of the CAD model and the physical part through the ICP (Iterative Closest Point) algorithm, to achieve millimeter-level fitting of virtual model and real object.

[0191] Generate dynamic AR superimposed instructions: Fuse real-time camera image and registered CAD model, superimpose dynamic three-dimensional arrows (indicate disassembly direction), thermal color temperature labels (such as red display for overloaded components) and floating text prompts (such as "Bolt M8: 25N·m±5%") on the surface of the physical device.

[0192] Embedded voice instruction interaction module: integrated voice recognition engine (such as Whisper), maintenance personnel can describe the operation of the abnormality (such as "bolt jam") through natural language, the system analyzes and triggers the corresponding prompt (such as "switch to impact wrench mode") in real time.

[0193] Record maintenance deviation events: when the voice description or sensor data (such as torque not meeting the standard) detects deviation, the current AR picture (including superimposed guidance), timestamp and operation context are automatically intercepted, a structured deviation log (such as "step 3: bolt torque 18N·m < 25N·m | reason: jam | voice note: suggest lubricating threads") is generated.

[0194] Dynamic adjustment of guidance logic: based on the deviation log, a reinforcement learning model is trained, when the same type of deviation occurs repeatedly (such as a certain bolt frequently jamming), pre-operation guidance (such as "spray rust remover in advance") is automatically inserted before the corresponding step, and the risk area is highlighted.

[0195] This embodiment realizes the efficiency, standardization and safety of the maintenance process through intuitive guidance of virtual and real fusion: the AR superimposed part disassembly sequence animation and torque parameter labeling (such as "bolt M8: 25N·m") can accurately guide the maintenance personnel to operate according to the standard process, avoiding misassembly or parameter deviation caused by lack of experience; real-time danger area labeling (such as high-temperature component red mask) and tactile feedback (such as device vibration when torque is too high) can actively warn potential risks and reduce the probability of injury; at the same time, AR guidance converts abstract text schemes into spatially visualized interaction steps (such as arrow guide tool path), making novice maintenance efficiency improve by more than 60%, greatly shortening the fault handling cycle and equipment downtime loss, ultimately achieving multiple goals of controllable maintenance quality, cost optimization and personnel safety.

[0196] In some embodiments, the dynamic adjustment of guidance logic specifically includes:

[0197] Collect maintenance deviation logs: the system records deviation events (such as "bolt jam") in the maintenance process in real time, including operation steps, sensor data (torque value), voice notes and AR picture screenshots, and generates a structured deviation log database.

[0198] Identify high-frequency similar deviation patterns: analyze the deviation logs through clustering algorithms (such as DBSCAN) to count the frequency of similar deviations (such as a certain bolt jamming) and context association (such as when the ambient humidity is greater than 80%, the frequency increases by 50%), and mark high-frequency risk nodes.

[0199] Dynamic insertion of pre-operation guidance: when the maintenance process triggers a high-frequency risk node, call the adaptive maintenance pre-operation reinforcement learning model to output pre-operation instructions (such as "step 3 before: spray rust remover"), and highlight the associated components (such as the red flashing bolt) and display operation animations in the AR view.

[0200] Closed-loop verification and optimization: After the pre-operation, the system tracks the actual deviation occurrence rate (such as the number of stalls), feeds back the results to the reinforcement learning model to update the policy weights, forming a self-evolution loop of "risk prediction-intervention-effect verification".

[0201] The construction and training process of the adaptive maintenance pre-operation reinforcement learning model includes the following stages:

[0202] (1) Data preparation stage:

[0203] Extract structured deviation logs from the historical database, each log record contains maintenance stages (such as disassembly, installation), device sensor data (torque, temperature), environmental parameters (humidity, dust), pre-operation actions (such as spraying rust remover), and result indicators (deviation occurrence, time consumption, cost).

[0204] Perform NLP processing on unstructured data (maintenance personnel voice notes), extract keywords (such as "rusty" "loose") and encode them into binary features (such as "rusty = 1"), clean invalid or conflicting data, and build training and test sets.

[0205] (2) State and action space design stage:

[0206] The state space is composed of four-dimensional features: 1) One-Hot encoding of maintenance stages (such as 10 stages corresponding to 10 dimensions); 2) Device working conditions (normalized temperature, vibration, etc. Sensor data); 3) Environmental parameters (humidity, dust concentration); 4) Historical deviation frequency of the same type (sliding window statistics). The action space defines 20 pre-operation instructions, such as A0 (no operation), A1 (spray rust remover), A2 (preheat tools to 80°C), A3 (clean contact surface), etc. Each action corresponds to a unique operation code and execution parameters.

[0207] (3) Reward function design stage:

[0208] Design a multi-objective weighted reward function R: basic deviation reward R 偏差 (deviation does not occur +1.0, occurs -0.5), efficiency reward R 效率 (every 1 minute saved +0.2, overtime -0.1), cost penalty R 成本 (consumed cost is converted into a penalty value in a certain proportion), exploration incentive R 探索 (first action performed +0.3).

[0209] Reward function R = 0.6R 偏差 + 0.2R 效率 + 0.1R 成本 + 0.1R 探索, by dynamically adjusting the weight balance short-term benefits and long-term strategy stability.

[0210] (4) Algorithm implementation phase:

[0211] The double deep Q network (DDQN) architecture is adopted: the main network (input 64-dimensional state, output 20 action Q values) and the target network (delayed synchronization parameters), the hidden layer is 128→64 node full connection layer (ReLU activation), the output layer is linearly activated. The experience replay pool stores 100,000 samples, and samples according to the TD error priority (error high sample extraction probability + 30%), to prevent overfitting caused by data correlation. The optimizer uses Adam (learning rate 0.001), and the loss function is mean square error (MSE).

[0212] (5) Training process phase:

[0213] During offline pre-training, 512 historical data are extracted from each batch to calculate the Q value of the main network and the Q value of the target network target =r+γ⋅max TargetNet(s'). Where: Qtarge represents the target Q value, representing the expected long-term return of the current state action pair, used to calculate the training loss; r represents the immediate reward, that is, the direct reward obtained after executing the action (such as +1.0 if no deviation occurs); γ represents the discount factor, which can be 0.95, used to balance the importance of current rewards and future potential rewards (future rewards are attenuated to 95% of the current value); TargetNet(s') represents the Q value prediction of the target network for all possible actions of the next state s'; max represents selecting the maximum Q value in the next state s', representing the expected return of the optimal action in the future.

[0214] Update the main network parameters, and synchronize the target network every 1000 steps. In the online fine-tuning phase, new maintenance data is received in real time, and if the TD error > 0.5, incremental training is triggered: add new data to the experience pool and sample 256 according to the 1:3 new-old ratio, and perform gradient descent; retrain the whole amount once a week to deal with equipment aging or process changes.

[0215] (6) Strategy verification and deployment phase:

[0216] The model effect is verified on an independent test set (30% historical data) with a requirement of a bias reduction rate of ≥40%, a cost increase of ≤15%, and an expert evaluation of action interpretability (≥8 / 10 points). After verification, the model is packaged as a REST API, which receives real-time state vectors (such as the current maintenance stage, humidity 80%, and bolt torque 18 N·m) every second and returns the top-3 pre-operation recommendations (such as A1 confidence 82% and A3 confidence 75%). The AR system automatically inserts guidance (such as highlighting the bolt and displaying a paint rust remover animation) based on a confidence threshold (>70%), otherwise it prompts manual decision-making. The execution results are fed back to the database to drive continuous optimization of the model.

[0217] In some embodiments, after dynamically adjusting the guidance logic, the following steps are further included:

[0218] Synchronous remote expert collaboration: upload the bias log and AR screen to the cloud in real time, and the expert views the synchronous perspective with superimposed guidance through the WebAR interface and marks the correction opinion (such as "use a heat gun to soften the thread glue"), and the local AR view is updated immediately.

[0219] Generate intelligent maintenance report: integrate standard operation guidance, bias record, and expert correction content, generate interactive report on timeline, click step to play AR guidance animation, and export PDF / 3DPDF format for quality traceability.

[0220] Closed-loop optimization of CAD-scene mapping: based on the SLAM positioning data and manual correction record of each maintenance, continuously optimize the feature matching rules of CAD model and physical device (such as adaptive adjustment of ICP weight parameter), improve the initialization speed and registration accuracy of subsequent AR guidance.

[0221] This embodiment realizes the precision and intelligence of maintenance guidance through augmented reality (AR) technology, based on high-precision spatial registration of visual-inertial SLAM algorithm and CAD model, real-time superimposition of virtual operation guidance (such as part disassembly path, torque parameter annotation) on the surface of physical equipment, guiding maintenance personnel to operate according to standardized process; combined with voice interaction and sensor feedback (such as tactile vibration warning of over-limit torque), real-time capture and record operation bias (such as "bolt not reaching 25 N·m"), dynamically adjust guidance logic (such as insert lubrication pre-operation step); at the same time, remote experts can guide complex problems through AR annotation, forming a closed-loop maintenance network of "local operation-cloud collaboration", ultimately improving the efficiency of novice maintenance by 60%, reducing operation error rate by 45%, reducing the cost of expert resource calling by 70%, and continuously optimizing guidance strategy based on historical bias data, continuously improving maintenance quality and equipment reliability.

[0222] The method provided by the above embodiment can be executed by a product quality problem solution automatic generation system composed of an electronic device. The electronic device in the embodiment of the present application is described from the perspective of hardware processing. Please refer to Figure 3 FIG. 1 is a schematic diagram of an entity device structure of the electronic device in the embodiment of the present application.

[0223] It should be noted that, Figure 3 The structure of the electronic device shown is only an example and should not limit the functions and use range of the embodiment of the present application.

[0224] As Figure 3 shown, the electronic device includes a central processing unit (CPU) 401 which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 402 or programs loaded from a storage portion 408 into a random access memory (RAM) 403, such as performing the method described in the above embodiment. Various programs and data required for system operation are also stored in the random access memory (RAM) 403. The central processing unit (CPU) 401, the read-only memory (ROM) 402, and the random access memory (RAM) 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0225] The following components are connected to the input / output (I / O) interface 405: an input portion 406 including an audio input device, a button switch, and the like; an output portion 407 including a display and an audio output device, an indicator, and the like; a storage portion 408 including a hard disk and the like; and a communication portion 409 including a network interface card such as a LAN (Local Area Network) card, a modem, and the like. The communication portion 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output (I / O) interface 405 as needed. A removable medium 411 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is mounted on the drive 410 as needed so that a computer program read therefrom is installed in the storage portion 408 as needed.

[0226] In particular, the processes described above with reference to the flowcharts can be implemented as a computer software program in accordance with embodiments of the present application. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising computer instructions for performing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication section 409, and / or installed from the removable media 411. When the computer program is executed by the central processing unit (CPU) 401, various functions defined in the present application are performed.

[0227] Note that specific examples of computer readable storage media can include without limitation: an electrical connection having one or more wires; a portable computer diskette; a hard disk; a random access memory (RAM); a read-only memory (ROM); an erasable programmable read only memory (EPROM); a flash memory; an optical fiber; a portable compact disc read-only memory (CD-ROM); an optical storage device; a magnetic storage device; or any suitable combination of the foregoing. In the present application, a computer readable storage medium can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0228] The flowcharts and block diagrams in the attached drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowcharts or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved.

[0229] In particular, the electronic device of the present embodiment includes a processor and a memory coupled to the one or more processors, the memory storing computer program code comprising computer instructions to be invoked by the one or more processors to cause the electronic device to perform the method provided by the above-described embodiments.

[0230] As another aspect, the present application also provides a computer readable storage medium, which can be included in the electronic device described in the above embodiments, or can exist separately without being assembled into the electronic device. The storage medium carries one or more computer programs, which, when executed by a processor of the electronic device, cause the electronic device to implement the method provided in the above embodiments.

[0231] The above described embodiments are merely used to illustrate the technical solutions of the present application, but not for limiting the present application; even though the present application has been described in detail with reference to the foregoing embodiments, those ordinarily skilled in the art should understand: the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced equivalently; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0232] In the above embodiments, according to the context, the term "when" can be interpreted as meaning "if" or "after" or "in response to determining" or "in response to detecting". Similarly, according to the context, the phrase "upon determining" or "if detecting (the stated condition or event)" can be interpreted as meaning "if determining" or "in response to determining" or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)".

[0233] Those ordinarily skilled in the art can understand that all or part of the processes in the above method embodiments can be implemented by a computer program instructing relevant hardware, and the program can be stored in a computer readable storage medium, and when the program is executed, the processes of the above method embodiments can be included. The foregoing storage medium includes ROM or random storage memory RAM, magnetic disc or optical disc, and various program code storage media.

Claims

1. A product quality problem resolution solution auto-generation method, characterized by, The method comprises the following steps: Determine the target fault code and the target product model information from the current product quality problem fault code and product model information, the fault code comprising a plurality of fault identification information, the fault identification information comprising failure part information, failure location information and failure mode information; Determine a first recommended solution set from a historical database based on the target fault code and the target product model information, the historical database comprising a plurality of problem solving records, the problem solving records comprising product model information, fault code and corresponding problem solving method; When the data amount of the problem solving records in the first recommended solution set is less than a preset threshold, determine a second recommended solution set from the historical database based on the fault identification information corresponding to the target fault code, the data amount of the problem solving records in the second recommended solution set being greater than that in the first recommended solution set; Generate a quality problem recommended solution according to the target fault code, the target product model information and the problem solving records in the second recommended solution set; Convert the quality problem recommended solution into an AR operation guide, superimpose a virtual instruction including a part disassembly sequence, a torque parameter and a dangerous area mark through an intelligent glasses or a mobile device, and the specific implementation process is as follows: Deploy a SLAM space positioning engine: integrate a visual-inertial SLAM algorithm in the AR device, construct a three-dimensional point cloud map of the maintenance environment in real time through camera and IMU data, and accurately track the position and attitude of the device in the physical space; Load a device CAD model library: convert the CAD engineering model of the target part into lightweight grid data, associate metadata and preload to the local storage of the AR device; Real-time CAD-physical space registration: based on the point cloud map generated by SLAM, align the key feature points of the CAD model and the physical part through the ICP algorithm, and realize the millimeter-level fitting of the virtual model and the real object; Generate dynamic AR superimposed guide: fuse the real-time camera picture and the registered CAD model, superimpose dynamic three-dimensional arrows, thermal color temperature marks and floating text prompts on the surface of the physical device; Embed a voice instruction interaction module: integrate a voice recognition engine, and the maintenance personnel can describe the operation anomaly through natural language, and the system can analyze and trigger the corresponding prompt in real time; Record the maintenance deviation event: when the voice description or sensor data detects the deviation, automatically intercept the current AR picture, record the timestamp and operation context, and generate a structured deviation log; Dynamic adjustment of guide logic: based on the deviation log, train a reinforcement learning model, and when the same type of deviation repeatedly occurs, automatically insert a pre-operation guide before the corresponding step and highlight the risk area; The dynamic adjustment of guide logic comprises: collecting maintenance deviation logs; identifying high-frequency similar deviation patterns; dynamically inserting pre-operation guides; when the maintenance process triggers a high-frequency risk node, calling an adaptive maintenance pre-operation reinforcement learning model to output a pre-operation instruction, highlighting the associated components in the AR view and displaying the operation animation; closed loop verification and optimization. The construction and training process of the adaptive maintenance pre-operation reinforcement learning model includes the following stages: a data preparation stage, a state and action space design stage, a reward function design stage, an algorithm implementation stage, a training process stage, and a strategy verification and deployment stage. In the state and action space design stage, the state space is composed of four-dimensional features, including maintenance stage One-Hot encoding, equipment working condition, environmental parameters, and historical similar deviation frequency. The action space defines 20 pre-operation instructions, each action corresponds to a unique operation code and execution parameter. In the algorithm implementation stage, a double deep Q network architecture is used for the main network and the target network. The input of the main network is 64-dimensional state, and the output is 20 action Q values. The hidden layer is a 128→64 node full connection layer, and the output layer is linearly activated. The experience replay pool stores 100,000 samples, and samples are taken according to the TD error priority to prevent overfitting caused by data correlation. The optimizer uses Adam, and the loss function is mean square error.

2. The method of claim 1, wherein, The quality problem recommended solution is generated according to the target fault code, the target product model information, and the second recommended solution set, including: determining the solution text replacement placeholder of each target problem solving record according to the target fault code, the target product model information, and the second recommended solution set. The solution text replacement placeholder includes a product model placeholder, a failed part placeholder, a failure location placeholder, and a failure mode placeholder; generating multiple pieces of solution information according to the solution text replacement placeholder corresponding to each target problem solving record. The solution information is used to represent the text content of the quality problem recommended solution.

3. The method of claim 2, wherein, After generating multiple pieces of solution information according to the solution text replacement placeholder corresponding to each target problem solving record, the following steps are further included: Displaying multiple pieces of solution information to a user; In response to a solution confirmation instruction of the user, determining a final quality problem recommended solution from the multiple pieces of solution information.

4. The method of claim 3, wherein, After determining the final quality problem recommended solution from the multiple pieces of solution information in response to the solution confirmation instruction of the user, the following steps are further included: Filling the final quality problem recommended solution into the solution maintenance page of the current product quality problem to obtain a problem solution recommendation page; In response to an editing operation of the user on the problem solution recommendation page, adjusting the content in the problem solution recommendation page.

5. The method of claim 4, wherein, After adjusting the content in the problem solution recommendation page in response to the editing operation of the user on the problem solution recommendation page, the following steps are further included: In response to an audit pass operation of the user on the problem solution recommendation page, generating a new problem solving record based on the problem solution recommendation page after the audit pass; Storing the new problem solving record in the historical database.

6. The method according to any one of claims 1 to 5, characterized in that, determining a first recommended solution set from the historical database based on the target fault code and the target product model information, comprising: querying all historical problem solving records containing the same fault code from the historical database according to the target fault code and the target product model information to obtain an initial data set; filtering the initial data set based on the target product model information to retain historical records matching the target product model information to obtain a same model data set; sorting the same model data set according to the total number of occurrences of the fault code of each problem solving record in the historical database to obtain a sorting result; selecting the top N problem solving records in the sorting result to obtain the first recommended solution set.

7. The method according to any one of claims 1 to 5, characterized in that, determining a second recommended solution set from the historical database based on the fault identification information corresponding to the target fault code, comprising: retrieving problem solving records containing the same failure mode from the historical database based on the target failure mode information in the target fault code to obtain an expanded data set; filtering problem solving records matching the target product model information from the expanded data set as first priority data, and problem solving records not filtered as second priority data; sorting the second priority data according to the historical occurrence frequency of the target failure mode information to obtain sorted second priority data; cutting off a corresponding number of problem solving records from the first priority data and the sorted second priority data according to the difference between the preset threshold and the current data volume of the first recommended solution set to obtain the second recommended solution set.

8. An electronic device, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is configured to store computer program code, the computer program code comprising computer instructions, and the one or more processors are configured to invoke the computer instructions to cause the electronic device to perform the method of any one of claims 1-7.

9. A computer readable storage medium storing computer instructions, characterized in that, When the computer instructions run on the electronic device, the electronic device is caused to perform the method of any one of claims 1-7.

10. A computer program product, characterised in that, When the computer program product runs on the electronic device, the electronic device is caused to perform the method of any one of claims 1-7.

Images