Vehicle production quality risk multi-source data collaborative evaluation full-chain optimization method

Abstract

The application discloses a vehicle production quality risk multi-source data collaborative evaluation full-chain optimization method, relates to the technical field of vehicle production quality risk, and comprises the following steps: in the case that it is confirmed that the production line data source is a supplementary production line data source, based on data source characteristic changes, dynamically adjusting a multi-source data fusion rule, standardizing the unit, precision, frequency and format of the supplementary production line data source through a preset data mapping rule, and synchronously adjusting data fusion weight parameters; after completing the multi-source data fusion rule adjustment, based on a data source characteristic adaptation degree, dynamically adjusting a quality risk evaluation model. The application solves the problem of quality risk evaluation failure caused by the heterogeneous nature of the supplementary production line and the main production line data, realizes efficient and accurate quality risk evaluation and production optimization through dynamic adjustment of the data fusion rule and the risk evaluation model, and improves the quality control capability of the production process.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle production quality risk, and in particular to a full-chain optimization method for collaborative evaluation of multi-source data on vehicle production quality risk. Background Art

[0002] The collaborative assessment and full-chain optimization of multi-source data for vehicle production quality risks is a systematic quality risk management solution for the entire automotive manufacturing process, based on information technology and intelligent analytical methods. This approach integrates multi-source, heterogeneous data from design, procurement, production, quality inspection, and other stages (such as sensor data, production logs, equipment status, quality inspection reports, and manual records) to build a unified data collaboration model. This approach then uses fusion analysis techniques to assess and predict potential quality risks across the entire manufacturing chain. Its core approach is to break down data silos and enable information interoperability and verification across all manufacturing stages. This allows for precise identification of the root causes and impact paths of risks, and generates strategic recommendations for system optimization. This approach is driven by the fact that traditional quality control relies heavily on single-point data or post-processing, making it difficult to promptly identify systemic quality issues caused by complex relationships. This leads to ineffective prevention of product defects, increased manufacturing costs, and frequent customer complaints. Through the collaborative assessment of multi-source data and the closed-loop execution of optimization strategies across the entire chain, this approach not only improves the accuracy of risk identification but also strengthens the controllability and responsiveness of the entire process. It is a key supportive tool for promoting intelligent and high-quality development in automotive manufacturing.

[0003] Existing technology for collaboratively assessing vehicle production quality risk through multi-source data and optimizing the entire supply chain primarily relies on building a data fusion and analysis system covering the entire automotive manufacturing process to accurately identify quality risks and implement closed-loop optimization. This technology first deploys data collection portals across key links, including design, supply chain, manufacturing, assembly, quality inspection, and after-sales service. This system collects heterogeneous data from multiple sources, including product structure parameters, process configuration information, raw material batches, equipment operating status, environmental monitoring, operational records, defect detection results, and user feedback. Subsequently, through data standardization, cleansing, and semantic fusion, a unified data representation model is constructed. Machine learning, graph neural networks, and knowledge graphs are then used to perform cross-domain correlation analysis on the data, identifying potential quality risk factors and their transmission paths. Based on risk identification, the system further generates targeted optimization recommendations. By integrating with systems such as MES (Manufacturing Execution Systems), ERP (Enterprise Resource Planning), and QMS (Quality Management Systems), it enables automated adjustments and closed-loop optimization of production strategies, process parameters, and material usage. The entire process covers six major links: data collection, preprocessing, fusion modeling, risk assessment, optimized decision-making, and execution feedback. It constitutes a dynamic, intelligent, and life-cycle quality risk management mechanism, effectively improving the predictability and controllability of vehicle manufacturing quality.

[0004] The existing technology has the following deficiencies:

[0005] During the vehicle production process, when supplementary production lines are temporarily activated for small-batch production due to market demand fluctuations or special order requirements, differences in equipment model, sensor type, and automation level may exist between the supplementary production lines and the main production lines. This can lead to inconsistencies in the accuracy, format, unit, and upload frequency of the data collected by the supplementary production lines. This lacks unified quality labels and standards, which in turn affects subsequent data fusion and risk assessment. Existing technologies are unable to dynamically adjust multi-source data fusion rules and risk assessment models based on the changes in data characteristics during the temporary connection of the supplementary production lines. As a result, the heterogeneity of the supplementary production line data is not effectively identified and optimized, which can lead to the loss of critical quality anomalies and the timely detection of production defects such as abnormal weld strength and poor paint adhesion. This can ultimately result in substandard products entering the market, leading to quality accidents, product recalls, customer complaints, and economic losses.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0007] The purpose of the present invention is to provide a full-chain optimization method for collaborative evaluation of multi-source data on vehicle production quality risks to solve the problems in the above-mentioned background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risks, specifically comprising the following steps:

[0009] Step 1: Identify and classify production line data sources in the vehicle production process in real time based on data source characteristics to determine whether the data source belongs to a supplementary production line data source. Data source characteristics include device model, sensor type, data format, and data upload frequency.

[0010] Step 2: If the production line data source is confirmed to be a supplementary production line data source, dynamically adjust the multi-source data fusion rules based on changes in data source characteristics. Standardize the unit, precision, frequency, and format of the supplementary production line data source through preset data mapping rules to ensure consistency with the data format of the main production line data source, and simultaneously adjust the data fusion weight parameters.

[0011] Step 3: After completing the adjustment of the multi-source data fusion rules, the quality risk assessment model is dynamically adjusted based on the adaptability of the data source characteristics. By calculating the similarity of process parameters, equipment characteristics, and historical defect characteristics, the model with the highest adaptability is selected or correction parameters are loaded to generate an adapted risk assessment model.

[0012] Step 4: Input the adjusted data fusion result, calculate the synergy parameter and risk parameter based on normalization, the synergy parameter represents the process consistency, and the risk parameter represents the fluctuation level, input the preset mapping function to generate the quality risk assessment coefficient, and determine the risk level of the production status based on the comparison result between the generated quality risk assessment coefficient and the preset threshold value and the preset risk level threshold interval;

[0013] Step five: Implement differentiated production optimization measures based on the risk level of the determined production status, and record and provide feedback on the risk assessment input data, risk assessment model adjustment path, risk level results, and production optimization execution results.

[0014] Preferably, step one specifically includes the following sub-steps:

[0015] Using raw production line data from the vehicle production process as input, a multi-dimensional data source feature information set is established, including device model, sensor type, data format, and data upload frequency.

[0016] Standardize each piece of raw feature information to form a structured data source feature vector set, including unit conversion, format unification, and time series completion;

[0017] Call the preset comparison rule set to compare the standardized data source feature vector with the main production line reference feature vector to generate a production line consistency score;

[0018] The production line data source category is confirmed in real time based on the production line consistency score and the set threshold. Specifically, when the production line consistency score is less than or equal to the set threshold, the production line data source is confirmed to be classified as a supplementary production line data source; when the production line consistency score is greater than the set threshold, the data source is confirmed to be a main production line data source.

[0019] And output the classification results for subsequent multi-source data fusion rule adjustment and risk assessment model optimization.

[0020] Preferably, step 2 specifically includes the following sub-steps:

[0021] Based on the confirmed feature change information of the supplementary production line data source, the feature change information of the equipment model characteristics, sensor type characteristics, data format characteristics, and data upload frequency characteristics corresponding to the supplementary production line data source is extracted in real time to form a feature change information set of the supplementary production line data source;

[0022] Based on the feature change information set of the supplementary production line data source, the preset multi-dimensional data mapping rule set is called to standardize the supplementary production line data source, including differences in equipment model, sensor accuracy, data format, and data upload frequency, to ensure that the supplementary production line data source is consistent with the main production line data source in dimensions such as unit, accuracy, frequency, and format;

[0023] Based on the standardized transformed supplementary production line data source, a fusion weight adjustment parameter is calculated, the historical volatility index, stability index and reliability index of the supplementary production line data are counted, and a preset weighted scoring function is used to calculate the weight adjustment coefficient of the supplementary production line data source;

[0024] According to the calculated weight adjustment coefficient, the weight proportion of the supplementary production line data in the multi-source data fusion is dynamically adjusted, and under the guidance of the updated data fusion rule, a multi-source data real-time fusion operation is performed to fuse the supplementary production line data source and the main production line data source into a data set with unified format, unified standard and unified weight for subsequent quality risk assessment model.

[0025] Preferably, step three specifically includes the following sub-steps:

[0026] Based on the completed multi-source data fusion rule dynamically adjusted production line data fusion result, the multi-dimensional feature information of the supplementary production line data source and the main production line data source is extracted to form a production line feature information set of process parameter feature information, equipment feature information and historical defect feature information;

[0027] A multi-dimensional similarity calculation method is used to calculate the similarity of the supplementary production line data source and the main production line data source in the process parameter feature dimension, the equipment feature dimension and the historical defect feature dimension, wherein the process parameter feature similarity is calculated by the Euclidean distance function, the equipment feature similarity is calculated by the cosine similarity function, and the historical defect feature similarity is calculated by the weighted overlap coefficient;

[0028] Based on the process parameter feature similarity, the equipment feature similarity and the historical defect feature similarity, a preset feature weight distribution mechanism is used to fuse the three similarity values into a single comprehensive feature adaptation value by weighted average method;

[0029] The calculated comprehensive feature adaptation value is compared with a preset risk assessment model selection threshold value, when the comprehensive feature adaptation value is greater than or equal to the risk assessment model selection threshold value, the main production line risk assessment model is selected; when the comprehensive feature adaptation value is less than the threshold value, the correction parameter for the supplementary production line feature difference is loaded or a new model version is derived;

[0030] The dynamically adjusted risk assessment model is applied to the quality risk assessment process to generate a quality risk assessment coefficient, and combined with a preset risk level threshold value, the risk level category of the current production state is determined, and all evaluation data and model adjustment paths are recorded and fed back for subsequent model optimization and automatic learning.

[0031] Preferably, a multi-dimensional similarity calculation method is adopted to calculate the similarity of the process parameter feature dimension, the equipment feature dimension and the historical defect feature dimension of the supplementary production line data source and the main production line data source, specifically:

[0032] In the process parameter feature dimension, the key process parameters reflecting the process state are extracted, and the extracted key process parameters are standardized. The Euclidean distance function is used to calculate the process parameter similarity of the supplementary production line data source and the main production line data source, to reflect the closeness of the two production lines in the process control level.

[0033] In the equipment feature dimension, the technical performance indicators of the production equipment are extracted and converted into vector representation. The cosine similarity function is used to calculate the equipment feature similarity of the supplementary production line data source and the main production line data source, to quantify the consistency level between the equipment technical parameters.

[0034] In the historical defect feature dimension, the quality defect data of the supplementary production line and the main production line in the past production process are extracted, and a historical defect feature set is constructed. The weighted overlap coefficient method is used to calculate the similarity of the historical defect feature set, to comprehensively evaluate the similarity of the two production lines in defect performance, and generate the corresponding defect feature similarity.

[0035] Preferably, step four specifically includes the following sub-steps:

[0036] The multi-source data fusion result after dynamic adjustment is taken as input and normalized to eliminate the numerical scale difference of different production line data dimensions and ensure that the data of each dimension is compared and calculated under a unified standard. The unified data range is [0, 1];

[0037] After completing the normalization processing, the synergy parameter and the risk parameter are calculated. The synergy parameter is used to measure the difference in process consistency between the supplementary production line data source and the main production line data source. The process parameter difference is calculated by the Euclidean distance function. The risk parameter is used to measure the volatility and risk level of the supplementary production line data source compared with the historical process benchmark. The risk is quantified by calculating the fluctuation amplitude of the current process parameter and the historical process benchmark.

[0038] The calculated synergy parameter and risk parameter are taken as input and input into the preset mapping function for processing. The synergy parameter and the risk parameter are fused by weighted summation to generate a quality risk assessment coefficient, which is used to reflect the overall quality risk level of the current production process.

[0039] The generated quality risk assessment coefficient is compared with the preset risk level threshold interval, and the risk level of the production status is determined based on the comparison results. The specific comparison analysis is as follows: if the quality risk assessment coefficient exceeds the upper limit value of the preset risk level threshold interval, the risk level of the production status is high risk; if the quality risk assessment coefficient is between the upper limit value and the lower limit value of the preset risk level threshold interval, the risk level of the production status is medium risk; if the quality risk assessment coefficient is lower than the lower limit value of the preset risk level threshold interval, the risk level of the production status is low risk.

[0040] Preferably, the specific method of calculating the synergy parameter and the risk parameter is as follows:

[0041] Extract key process parameter data for the same process steps from the supplementary production line data source and the main production line data source, and normalize the extracted process parameter data to eliminate the impact of different units and dimensions, ensuring that all process parameters are compared under a unified standard;

[0042] Based on the normalized key process parameter data, the Euclidean distance function is used to calculate the numerical difference between the supplementary production line data source and the main production line data source in the key process parameter space. The Euclidean distance function obtains the corresponding distance value by summing the squares of the differences between each pair of corresponding process parameters and then taking the square root.

[0043] Based on the calculated distance value, a preset difference mapping function is called to convert the distance value into a collaborative parameter value. The collaborative parameter value is limited to the range of 0 to 1. The preset difference mapping function is executed based on the correspondence between the distance value and the set threshold value;

[0044] Extract the real-time monitoring data of each key process parameter in the current production cycle of the supplementary production line, and compare it with the corresponding process parameters in the historical process benchmark model. Use the preset fluctuation amplitude calculation function to calculate the deviation amplitude and change trend between the current process parameters and the historical process benchmark, and quantify them into risk parameters.

[0045] Preferably, step five specifically includes:

[0046] After determining the risk level of the production status, corresponding production optimization measures are implemented according to the determined risk level. Specifically, when the risk level of the production status is high risk, the production stop instruction and process parameter review instruction of the supplementary production line are executed; when the risk level of the production status is medium risk, the capacity restriction instruction and quality inspection enhancement instruction are executed; when the risk level of the production status is low risk, the normal production instruction is executed and the risk monitoring data is recorded;

[0047] During the implementation of each production optimization measure, all involved risk assessment input data, risk assessment model adjustment path, risk level results and production optimization execution results are recorded in detail and fed back into the quality risk assessment database to support the subsequent optimization and upgrading of the quality risk assessment model.

[0048] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0049] 1. The present invention effectively solves the heterogeneity problem between the supplementary production line data source and the main production line data source that are temporarily enabled due to market demand fluctuations or special order requirements by identifying and classifying production line data sources in real time. Through standardized processing of multi-dimensional data source feature information such as equipment model, sensor type, data format, upload frequency, etc., the system can accurately identify the data source category and adjust the data fusion rules when the supplementary production line is temporarily connected, ensuring consistency between data sources and providing an accurate data basis for subsequent quality risk assessment. The implementation of this process not only improves data consistency, but also eliminates the risk of inconsistency caused by differences in equipment and sensors, and solves the problem of evaluation failure of existing technologies when faced with supplementary production line data.

[0050] 2. The present invention dynamically adjusts the multi-source data fusion rules and risk assessment model, and the technical solution can adjust the assessment model and weight coefficients in real time according to the changes in the characteristics of the supplementary production line data source. In particular, in the similarity calculation of process parameters, equipment characteristics and historical defect characteristics, the solution ensures the adaptability of the assessment model under different production line data characteristics by accurately quantifying process consistency and volatility. This dynamic adjustment mechanism can select the most appropriate risk assessment model or load correction parameters based on the feature adaptability, avoiding the assessment deviation caused by the inability of the traditional assessment mechanism to adapt to the differences in supplementary production lines. Through this mechanism, the solution achieves the accuracy and efficiency of quality risk assessment and ensures the accuracy of risk judgment.

[0051] 3. The present invention ensures precise control of the production process under different risk levels by implementing differentiated production optimization measures. When the production status is in the high-risk category, production stop and process review are immediately executed to prevent unqualified products from entering subsequent processes; the medium-risk category effectively reduces quality risks through capacity restrictions and enhanced quality inspections; the low-risk category maintains normal production while conducting continuous risk monitoring and data recording. All data records will be fed back to the quality assessment model to support the continuous optimization and intelligent updating of the model. Through this systematic feedback mechanism, the solution can continuously improve the accuracy of quality risk assessment and provide strong data support for production decisions, thereby effectively reducing the incidence of quality defects, reducing the risk of product recalls, and protecting the company's reputation and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0053] Figure 1 Schematic diagram of the process of the full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risks according to the present invention;

[0054] Figure 2 This is a mind map of the method for the full-chain optimization method of collaborative evaluation of multi-source data for vehicle production quality risks according to the present invention. DETAILED DESCRIPTION

[0055] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0056] The present invention provides Figure 1 and Figure 2 The full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risks shown in the figure specifically includes the following steps:

[0057] Step 1: Based on data source feature information, the production line data sources in the vehicle production process are identified and classified in real time. By comparing the collected equipment model information, sensor type information, data format information, and data upload frequency information, it is determined whether the production line data source is a supplementary production line data source;

[0058] In order to achieve real-time identification and classification of temporary access to supplementary production lines during vehicle production, it is necessary to dynamically confirm and effectively classify the data sources of each production line in the vehicle production process based on data source feature information. This process specifically includes:

[0059] The original production line data collected from each production process in the vehicle production process is taken as input to establish a multi-dimensional data source feature information set, which includes but is not limited to equipment model information, sensor type information, data format information, and data upload frequency information. Among them, the equipment model information refers to the uniqueness parameters of each production equipment in terms of manufacturer, equipment model, version code, factory identification, etc.; the sensor type information refers to the sensor categories, accuracy level, sampling rate, etc. indicators used for production data collection; the data format information refers to the data coding standard, numerical unit, data structure definition, etc. adopted by each production data; and the data upload frequency information refers to the real-time and periodic characteristics of various data uploaded to the data management platform in the production process. Through the complete collection of these four types of information, the basic features of the production line data source can be comprehensively obtained.

[0060] Based on the collected data source feature information set, a multi-dimensional data feature analysis operation is performed to convert each original feature information into a standardized feature index, forming a structured data source feature vector set. In this process, through the preset data unit conversion rules, format unification rules and time series completion rules, the data source features generated by different production line equipment are standardized and converted to eliminate the heterogeneous effects caused by different equipment manufacturers, different sensor performances, and different process system configurations, thereby ensuring the comparability and consistency of each feature index.

[0061] Based on the structured data source feature vector set, a preset data source feature comparison rule set is called to compare each production line data source standardized feature vector with the main production line reference feature vector. The comparison rules include equipment type comparison rules, sensor performance comparison rules, data format consistency comparison rules, and upload frequency stability comparison rules. In the comparison process, a weight weighting method is used to comprehensively evaluate each comparison result to generate a production line consistency score index for each production line data source. Through the quantitative results of the consistency score, the similarity degree of the current production line data source and the main production line data source in the overall features can be scientifically judged.

[0062] Based on the production line consistency score index, the production line data source classification operation is completed in real time according to the set identification threshold. When the consistency score is less than or equal to the preset threshold, the production line data source is automatically classified as a supplementary production line data source; when the consistency score is greater than the threshold, it is confirmed that the data source belongs to the main production line data source. After completing the classification of the production line data source, the classification result is output in real time and enters the subsequent multi-source data fusion rule adjustment step to realize the multi-source data fusion and risk assessment model adaptive optimization under the condition of temporary access of the supplementary production line.

[0063] This not only enables real-time and accurate identification of data source features when a supplementary production line is temporarily connected, but also effectively prevents heterogeneous data from the supplementary production line from being directly incorporated into existing risk assessment models without verification, thus avoiding the resulting failure of quality risk assessments. This implementation method, through the construction of a data source feature information set, standardized feature vector analysis, feature comparison scoring, and real-time classification decisions, achieves a dynamic, scalable, and human-free intelligent data source identification mechanism, providing a reliable data foundation for subsequent vehicle production quality risk assessments.

[0064] Step 2: If the production line data source is confirmed to be a supplementary production line data source, the multi-source data fusion rules are dynamically adjusted based on the characteristic changes of the production line data source. By calling the preset data mapping rules, the unit, precision, frequency, and format of the supplementary production line data source are standardized to ensure that the supplementary production line data source and the main production line data source have a consistent data expression format, and the data fusion weight parameters are adjusted synchronously;

[0065] In order to effectively solve the problem of multi-source data fusion rule failure caused by changes in the characteristics of the supplementary production line data when the production line data source is confirmed to be the supplementary production line data source, the dynamic adjustment of the multi-source data fusion rules is realized to ensure that the supplementary production line data can be smoothly integrated into the main production line data and support subsequent risk assessment. Specifically, the following are included:

[0066] Based on the completed identification results of the supplementary production line data source, feature change information corresponding to the supplementary production line data source is extracted in real time. This feature change information refers to the differences in data characteristics exhibited by the supplementary production line compared to the main production line during actual production. Specifically, this includes differences in equipment model characteristics, sensor type characteristics, data format characteristics, and data upload frequency characteristics. Through the systematic extraction of this feature change information, a feature change information set for the supplementary production line data source is formed. This feature change information set truly, comprehensively, and dynamically reflects the heterogeneous differences between the supplementary production line data source and the main production line data source in multiple dimensions such as process, equipment, sensors, and data structure.

[0067] Based on the feature change information set of the supplementary production line data source, the preset multi-dimensional data mapping rule set is called to perform standardized transformation on the supplementary production line data source. Specifically, for inconsistent data units caused by differences in equipment models, all data items are unified into the main production line standard unit system by calling the unit conversion rules; for inconsistent measurement accuracy caused by differences in sensor types, the data accuracy level is adjusted by calling the data accuracy compensation rules; for differences in data formats, the data structure is unified into the common format of the main production line by calling the format mapping rules; for differences in data upload frequency, data completion and smoothing are achieved by calling the time series compensation rules. Through the unified calling and execution of the above-mentioned multi-dimensional data mapping rules, a standardized supplementary production line data expression set is formed to ensure that the supplementary production line data has the same data dimension, data semantics and data structure as the main production line data, providing a consistency basis for subsequent multi-source data fusion.

[0068] Based on the standardized expression set of supplementary production line data, the weight adjustment parameters for multi-source data fusion are further calculated to achieve dynamic optimization of multi-source data fusion rules. Specific steps include: For the standardized supplementary production line data and main production line data, the historical volatility indicators, stability indicators, and credibility indicators of their key quality influencing factors are statistically analyzed. Based on a preset weighted scoring function, a weight adjustment coefficient for the supplementary production line data source is calculated. This weight adjustment coefficient dynamically determines the weight proportion of the supplementary production line data in the overall data fusion process, effectively suppressing the excessive impact of high volatility and low quality consistency of the supplementary production line data on the overall evaluation results, thereby improving the stability, reliability, and evaluation value of the fused data.

[0069] After completing the standardized transformation of the supplementary production line data and adjusting the fusion weights, based on the updated data fusion rules, a real-time fusion operation of multi-source data is performed. The supplementary production line data source and the main production line data source are fused under the guidance of unified rules to form a fused data set with a unified format, unified standards, and unified weights. This fused data set will serve as the input of the subsequent quality risk assessment model, ensuring that the supplementary production line data can be truly, effectively, and objectively involved in risk assessment and decision support without compromising the original stability of the assessment model.

[0070] Through the above, not only can we respond to changes in the characteristics of the supplementary production line data in real time and dynamically adjust the multi-source data fusion rules, but we can also ensure the comprehensiveness, accuracy and robustness of the fused data through the dual mechanisms of standardized transformation and weight adjustment, and effectively solve the problems of rigid fusion rules and invalid evaluation models when facing temporary supplementary production line data in existing technologies, thereby providing a solid and reliable data foundation and method support for the collaborative evaluation of multi-source data of vehicle production quality risks and full-chain optimization.

[0071] The multi-source data fusion rules and preset data mapping rules used must be strictly defined and clearly limited. Multi-source data fusion rules refer to a set of rules that are formulated to guide the data fusion process of different data sources based on the data structure, data unit, data accuracy, data upload frequency and data semantic differences between different production line data sources in the vehicle production process. The multi-source data fusion rules not only include standardized conversion requirements for different data sources, but also include data importance weight setting for different data sources, abnormal data identification and elimination standards, time series alignment rules and missing data completion rules, so as to ensure that multi-source heterogeneous data can be effectively integrated under a unified data system, avoiding fusion failure or evaluation deviation caused by differences in data sources.

[0072] Pre-set data mapping rules refer to a set of mapping relationships pre-established based on the characteristic standards of the main production line data source. They are used to unify the supplementary production line data source with the main production line data source across multiple dimensions, such as units, precision, format, and frequency. Specifically, these pre-set data mapping rules include, but are not limited to, unit conversion rules, data precision compensation rules, data format reconstruction rules, time series interpolation rules, and data consistency correction rules. Unit conversion rules use a unified data measurement system to ensure consistent conversion of data units generated by different devices and sensors; data precision compensation rules correct data of varying measurement precision by setting accuracy benchmarks; data format reconstruction rules standardize data structures through unified data encoding and structure definitions; time series interpolation rules smooth and complement data of varying sampling frequencies; and data consistency correction rules eliminate the impact of process parameter differences between different devices or production lines. By invoking and executing these multi-dimensional mapping rules, the supplementary production line data source can be fully aligned with the main production line data source across all key data dimensions, thus supporting the smooth implementation of multi-source data fusion rules and ensuring the accuracy and stability of quality risk assessments.

[0073] Step 3: After completing the dynamic adjustment of the multi-source data fusion rules, the quality risk assessment model is dynamically adjusted based on the feature compatibility between the supplementary production line data source and the main production line data source. By calculating the process parameter similarity, equipment feature similarity, and historical defect feature similarity of the production line data source, the risk assessment model version with the highest feature compatibility with the current data source is selected, or the supplementary production line correction parameters are loaded into the risk assessment model to generate a risk assessment model that is compatible with the current data features.

[0074] After completing the dynamic adjustment of the multi-source data fusion rules, in order to fully consider the multi-dimensional feature differences between the supplementary production line data source and the main production line data source, and then realize the dynamic adjustment of the quality risk assessment model, to ensure that the quality risk assessment model has flexible adaptability and high accuracy to the data characteristics of different production lines, the specific steps include:

[0075] Based on the completed multi-source data fusion rule dynamic adjustment of the production line data fusion result, the system comprehensively extracts the multi-dimensional feature information of the supplementary production line data source and the main production line data source, and forms a production line feature information set including process parameter feature information, equipment feature information and historical defect feature information. Among them, the process parameter feature information refers to various process parameters reflecting the operation process control level in the specific production process, such as welding temperature, assembly torque, coating thickness, etc.; the equipment feature information refers to the physical and performance characteristics of the specific manufacturing equipment on the production line, such as category, brand, model, accuracy level, load capacity, etc.; the historical defect feature information refers to the defect type, defect severity, defect occurrence frequency, defect repair effect and other information accumulated in the past production process of the production line. Through the extraction of the above multi-dimensional feature information, a comprehensive data basis can be provided for the subsequent accurate calculation of feature adaptation degree.

[0076] Based on the above extracted production line feature information set, a multi-dimensional similarity calculation method is used to perform similarity calculation on the process parameter feature dimension, equipment feature dimension and historical defect feature dimension between the supplementary production line data source and the main production line data source. The calculation of process parameter feature similarity is based on the standardized value of each process parameter, and the Euclidean distance function is used to measure the relative difference; the calculation of equipment feature similarity is by converting each performance index of the equipment into a vector, and using the cosine similarity function to quantify the consistency level between the equipment; the calculation of historical defect feature similarity uses a weighted overlap coefficient based method, which considers the defect type consistency, defect impact weight and defect duration impact. Through the above three independent similarity calculation methods, the similarity degree of the supplementary production line data source and the main production line data source in the key dimension can be scientifically and quantitatively reflected.

[0077] Based on the process parameter feature similarity, equipment feature similarity and historical defect feature similarity, a preset feature weight distribution mechanism is used to calculate the comprehensive feature adaptation degree value. The feature weight distribution mechanism presets the weight coefficients of the influence degree of different feature dimensions on the overall risk assessment, for example, the process parameter weight is set to 50%, the equipment feature weight is set to 30%, and the historical defect feature weight is set to 20%. Using the weighted average method, the above three similarity values are fused into a single comprehensive feature adaptation degree value. This comprehensive feature adaptation degree value serves as a quantitative basis for judging whether the current supplementary production line data source can directly adapt to the main production line risk assessment model, while avoiding the excessive influence of a certain feature dimension on the overall adaptation result, ensuring the scientificity and objectivity of the comprehensive judgment.

[0078] Based on the calculated comprehensive feature fitness value, it is compared with the preset risk assessment model selection threshold. When the comprehensive feature fitness value is greater than or equal to the risk assessment model selection threshold, the main production line risk assessment model version is directly selected to perform the quality risk assessment task; when the comprehensive feature fitness value is less than the risk assessment model selection threshold, the preset risk assessment model dynamic adjustment rule is called to dynamically generate a quality risk assessment model that is compatible with the data characteristics of the supplementary production line by loading correction parameters for the supplementary production line feature differences in the main production line risk assessment model or deriving a new model version. The risk assessment model dynamic adjustment rule defines parameter correction functions or model structure fine-tuning strategies to achieve flexible adjustment of the assessment model, ensuring that the model can reflect the data patterns, process characteristics and potential risk propagation paths unique to the supplementary production line.

[0079] The dynamically adjusted risk assessment model is applied to the quality risk assessment process based on multi-source data fusion results. The quality risk assessment coefficient for the current production status is generated in real time. Combined with the preset risk level threshold, the risk level category of the current supplementary production line's production status is determined. At the same time, the production line feature information, similarity calculation results, comprehensive fitness value, risk assessment model adjustment path, and risk level determination results involved in this risk assessment process are fully recorded in the assessment model continuous optimization database. This serves as a training data source for subsequent automatic model evolution and dynamic learning, gradually improving the assessment model's adaptability and accuracy to different production line change scenarios.

[0080] Through the above, not only can the multi-dimensional feature adaptability calculation and comprehensive evaluation of the supplementary production line data source and the main production line data source be realized, but also the dynamic adjustment method of the feature-driven risk assessment model can be used to solve the problems in the existing technology of single assessment model, lack of flexible adjustment, and inability to adapt to temporary production line data feature changes, effectively improving the intelligence level of vehicle production quality risk assessment and the whole-chain optimization capability.

[0081] In order to achieve scientific quantification and accurate evaluation of the differences between the supplementary production line data source and the main production line data source, it is first necessary to perform multi-dimensional similarity calculations on the process parameter feature dimension, equipment feature dimension, and historical defect feature dimension based on the production line feature information set extracted from the supplementary production line data source and the main production line data source, so as to quantify the degree of similarity of different production line data sources in each key dimension.

[0082] Specifically, in the process parameter feature dimension, by extracting various key process parameters reflecting the process state, including but not limited to welding temperature, pressure control value, coating thickness, assembly torque and heat treatment time, and standardizing the extracted process parameter values, the data scales, unit differences and dimensions of different process parameters are unified in the same evaluation system. Based on the standardized process parameter features, the Euclidean distance function is used to calculate the similarity of the supplementary production line data source and the main production line data source in the process parameter dimension, which can reflect the distance relationship between the two production lines in the multi-dimensional process parameter space, thereby quantifying the closeness of the process control level. The smaller the Euclidean distance value, the closer the two production lines are in the process control level, and the higher the similarity.

[0083] In the equipment feature dimension, by extracting the technical performance indicators of various production equipment used by the production line, including equipment brand, model, production capacity, running accuracy, sensor sensitivity, energy efficiency, etc., and converting the above equipment feature indicators into vector representation. Based on the vector representation, the cosine similarity function is used to calculate the similarity of the equipment features of the supplementary production line data source and the main production line data source. The cosine similarity function can reflect the consistency level between the technical parameters of the equipment by calculating the cosine value of the angle between the two equipment feature vectors. The closer the cosine value is to 1, the closer the equipment performance is, and the higher the similarity is.

[0084] In the historical defect feature dimension, by extracting the quality defect data of the supplementary production line and the main production line in the past production process, including defect type, defect level, defect frequency, defect impact range, defect repair cycle, etc., a historical defect feature set is constructed. The weighted overlap coefficient method is used to calculate the similarity of the historical defect feature set. This method sets the weight values of different types of defects and combines multiple factors such as defect frequency, defect severity, and defect duration to comprehensively evaluate the similarity of the two production lines in defect performance. The higher the weighted overlap coefficient, the higher the similarity of the production line in the historical defect distribution and evolution trend.

[0085] Through the above process parameter feature similarity calculation, equipment feature similarity calculation and historical defect feature similarity calculation, the differences between the supplementary production line data source and the main production line data source can be fully quantified from different dimensions and different levels, and a multi-dimensional feature similarity index system is constructed, which provides data support for the dynamic adjustment of the subsequent risk assessment model.

[0086] In order to scientifically fuse the similarity values of three different dimensions into a single index that can be used for risk assessment model selection decision, a feature weight distribution mechanism is adopted to calculate the comprehensive feature adaptation value. The feature weight distribution mechanism is a multi-dimensional fusion method based on preset influence weight coefficients. First, weight coefficients are set for process parameter features, equipment features and historical defect features. The setting principle of the weight coefficients is determined based on the influence degree of each feature dimension on the accuracy of vehicle production quality risk assessment, for example: the influence of process parameter features on risk assessment results is the most significant, so the weight is set to 50%; the influence of equipment features on risk assessment results is larger, but not as significant as process control level, so the weight is set to 30%; the influence of historical defect features on risk assessment results is relatively small, so the weight is set to 20%.

[0087] After determining the weight coefficients of each feature dimension, the process parameter feature similarity, equipment feature similarity and historical defect feature similarity are fused and calculated by weighted average method to generate the comprehensive feature adaptation value. The weighted average method adopts the mathematical function: comprehensive feature adaptation value = (process parameter feature similarity x 50%) + (equipment feature similarity x 30%) + (historical defect feature similarity x 20%). Through this method, multiple dimensional similarity indexes can be converted into a unified standard comprehensive evaluation value, effectively avoiding the excessive influence of a single feature dimension on the overall evaluation result, and ensuring the balance and scientificity of the evaluation decision.

[0088] The comprehensive feature adaptation value is used as the key decision basis for subsequent risk assessment model dynamic adjustment, and its value can accurately reflect whether the current supplementary production line data source can directly adapt to the main production line risk assessment model, or whether the risk assessment model needs to be corrected or reconstructed. Through this innovative technical means of multi-dimensional similarity calculation and weight fusion, the adaptability and dynamic response ability of the quality risk assessment model to multi-source heterogeneous data can be effectively improved, ensuring that the vehicle production quality risk collaborative evaluation process has high flexibility, intelligence and reliability.

[0089] Step four, taking the multi-source data fusion result after dynamic adjustment as input, calculating the collaborative parameter and the risk parameter based on normalization processing, the collaborative parameter is used to represent the process consistency between the production line data sources, and the risk parameter is used to represent the risk fluctuation level of the production line data source compared with the historical process benchmark, inputting the collaborative parameter and the risk parameter into the preset mapping function to generate the quality risk assessment coefficient, and based on the comparison result of the quality risk assessment coefficient and the preset risk level threshold interval, determining the risk level category of the current production state;

[0090] In order to calculate and generate quality risk assessment coefficients through the dynamically adjusted multi-source data fusion results, accurately classify the risk level of the current production status, and take reasonable production optimization measures for different risk levels, this can be achieved through the following methods:

[0091] The results of the production line data source, which have been dynamically adjusted using multi-source data fusion rules, are used as input for normalization. Normalization is a crucial step in data preprocessing, aiming to eliminate differences in numerical scales across different data dimensions and ensure that data from each dimension can be compared and calculated using a unified standard. Specifically, normalization maps key parameters from different production line data sources (such as process parameters, equipment performance data, and production process data) to a unified numerical range (e.g., [0, 1]), ensuring that all data items fall within the same dimensional range. This ensures that subsequent calculations are not biased by larger or smaller data values ​​in a particular dimension, thereby ensuring a balanced impact of all data types on the evaluation results.

[0092] After normalization, the normalized data is further calculated to generate synergy parameters and risk parameters. Synergy parameters measure the process consistency between the supplementary line data source and the main line data source. Process consistency refers to whether the process parameters (such as temperature, pressure, and time) used by the two lines remain consistent when performing the same process steps. Synergy parameters are calculated by comparing the differences in process parameter values ​​for the supplementary line and the main line during the same process steps, using the Euclidean distance function (i.e., calculating the straight-line distance between two data points) to quantify the similarity between the two. A smaller distance indicates greater consistency in process control between the two lines and a higher synergy parameter value. Risk parameters measure the volatility and risk level of the supplementary line data source when compared to a historical process benchmark. Specifically, risk parameters quantify the stability of the production process by calculating the fluctuation range between current process parameters and historical process benchmark data. When current process parameters fluctuate significantly, the risk parameter value is higher, indicating that the production process may be subject to greater risk.

[0093] The synergy parameter and risk parameter are input into a pre-set mapping function for processing. This mapping function is a mathematical model whose purpose is to fuse two independent parameters (synergy parameter and risk parameter) into a unified quality risk assessment coefficient. The design of the mapping function takes into account the relative importance of synergy parameters and risk parameters in assessing production quality risks. Therefore, weighted averaging, nonlinear weighting, or other adaptive algorithms can be used for parameter fusion. This mapping function ensures that the synergy parameter and risk parameter are effectively combined to generate a unified quality risk assessment coefficient. This coefficient represents the overall quality risk level of the current production process and reflects potential quality risks in the production process.

[0094] The generated quality risk assessment coefficient is compared with the preset risk level threshold interval, and the current production state can be classified as high risk, medium risk or low risk. If the quality risk assessment coefficient exceeds the upper limit value of the preset risk level threshold interval, the risk level of the production state is high risk, indicating that there is a serious quality problem in the current production state, and emergency control measures need to be taken immediately; if the quality risk assessment coefficient is between the upper limit value and the lower limit value of the preset risk level threshold interval, the risk level of the production state is medium risk, indicating that there is a certain quality problem, but the problem is lighter, which can be solved by strengthening quality monitoring and adjusting production rhythm; if the quality risk assessment coefficient is lower than the lower limit value of the preset risk level threshold interval, the risk level of the production state is low risk, indicating that the production process is relatively stable and the quality risk is low.

[0095] Based on the determination result, differential production optimization measures are taken. According to the determined risk level, the following corresponding optimization measures are implemented:

[0096] When the determination result is high risk, emergency measures such as immediate stop of production, recalibration of equipment, in-depth process inspection, etc. are taken to prevent unqualified products from entering subsequent processes or market;

[0097] When the determination result is medium risk, measures such as enhancing quality monitoring, increasing production detection frequency, focusing on monitoring key process links, etc. are started to ensure the stability of the production process;

[0098] When the determination result is low risk, normal production continues, but continuous monitoring and recording of production data are carried out for future risk assessment and optimization reference.

[0099] The execution process, results and adjustment measures involved in all production optimization measures are recorded and fed back to the data pool for subsequent quality risk assessment model optimization and continuous learning. This feedback mechanism can ensure that all data in the production process are fully utilized to provide accurate basis for future production decisions.

[0100] Through the above implementation, accurate quality risk assessment of the supplementary production line and main production line data sources can be performed, and differential and scientific production optimization measures can be taken based on the assessment results, significantly improving the quality control level in the production process, reducing the occurrence of quality defects, and ensuring the final quality of vehicle products.

[0101] The synergy parameter and the risk parameter are two core calculation indexes for realizing accurate quality risk assessment, and they respectively quantify and represent the difference and volatility of the supplementary production line data source and the main production line data source from different dimensions.

[0102] The synergy parameter is used to quantify the consistency between the supplementary line data source and the main line data source during process execution. Process consistency refers to the degree of difference in key process parameters between the supplementary line and the main line when executing the same production process step. Key process parameters include temperature setpoints, pressure control values, assembly gaps, welding currents, coating thicknesses, cooling times, and other process variables closely related to process control. To quantify process consistency, key process parameter data for the same process step is first extracted from the supplementary line and main line data sources. This extracted process parameter data is then normalized to standardize the different data units and dimensions into standardized values ​​to ensure comparability. Next, based on the normalized key process parameter data, the Euclidean distance function is used to calculate the difference between the supplementary line and the main line in the key process parameter space. The Euclidean distance function calculates the numerical distance between the supplementary line and the main line in the key process parameter space by taking the square root of the sum of the squares of the differences between each pair of corresponding process parameters. Subsequently, based on the calculated Euclidean distance value, the preset difference mapping function is called to map the distance value into a collaborative parameter value limited to the numerical range of 0 to 1, wherein the difference mapping function realizes numerical conversion based on the distance value and the set mapping relationship. The collaborative parameter value is used to reflect the consistency level of the supplementary production line data source and the main production line data source during the process execution, and participates in the subsequent quality risk level determination as an input parameter of the quality risk assessment model. This embodiment fully realizes the scientific quantification and consistency evaluation of collaborative parameters through the four steps of feature extraction, standardization processing, difference calculation and parameter mapping, and effectively supports the accuracy and reliability of multi-source data fusion and risk assessment models.

[0103] A preset difference mapping function converts the calculated Euclidean distance values ​​into synergy parameter values, achieving standardized expression and quantified output of process parameter variability. This preset difference mapping function is designed based on a predefined mapping relationship determined through extensive historical process data, quality fluctuation data, and production line stability analysis, ensuring good adaptability and reliability. The preset difference mapping function can be implemented using either linear or nonlinear mapping functions. Linear mapping functions achieve numerical conversion by setting a fixed proportional relationship between Euclidean distance values ​​and synergy parameter values, while nonlinear mapping functions employ mathematical methods such as exponential, logarithmic, or piecewise functions to achieve sensitive response and convergence control for extreme difference values. To ensure controllability of the mapping results, the preset difference mapping function constrains the synergy parameter values ​​to always be within the range of 0 to 1. Smaller Euclidean distance values ​​are mapped to larger synergy parameter values, while larger Euclidean distance values ​​are mapped to smaller synergy parameter values. The specific function type, parameter settings, and mapping range of the preset difference mapping function are determined through model training and simulation validation before system deployment. This effectively supports the dynamic assessment and continuous optimization of process consistency levels using multi-source data fusion and quality risk assessment models.

[0104] On the other hand, the risk parameter measures the volatility and risk level of the supplementary production line's data source compared to the historical process benchmark. The historical process benchmark is a reference model constructed based on the optimal ranges, control standards, and statistical characteristics of process parameters accumulated during long-term stable production. The risk parameter calculation method involves two steps: First, real-time monitoring data for each key process parameter during the supplementary production line's current production cycle is extracted and compared with the corresponding parameters in the historical process benchmark model. Second, based on a preset fluctuation amplitude calculation function, the magnitude and trend of the deviation of the current process parameter from the historical benchmark are statistically analyzed. For example, methods such as mean square error, standard deviation, or maximum deviation within a sliding window can be used to quantify volatility. When the deviation is small and the fluctuation trend is stable, the risk parameter value is low; when the deviation is large and the fluctuation trend is severe, the risk parameter value is high. The risk parameter value is also normalized to the range of 0 to 1 for unified measurement. A high risk parameter value indicates significant fluctuation or anomalies in the current production process, indicating process instability that may pose a quality risk.

[0105] Through the dual-dimensional calculation of collaborative parameters and risk parameters, it can not only reflect the level of process consistency between production lines, but also simultaneously identify the volatility risks that may arise during process execution, thereby providing multi-angle, data-driven support for comprehensive quality risk assessment.

[0106] In order to fuse the risk indicators of the two dimensions of the synergy parameter and the risk parameter into a unified quality risk assessment coefficient, a scientific and reasonable mapping function system needs to be designed to realize the organic integration of multi-source risk information.

[0107] The mapping function is a function system based on mathematical modeling, and its core goal is to transform the two risk indicators of different sources, different meanings and different scales, synergy parameter and risk parameter, into a single quality risk assessment coefficient that can be directly used for risk level determination through quantitative fusion mechanism. In order to achieve this goal, first of all, the relative importance of the synergy parameter and the risk parameter needs to be set. This weight setting process is usually based on historical quality data statistical analysis, process expert experience or multi-dimensional regression analysis. For example, when the statistical results show that the process consistency accounts for a higher proportion of the impact on quality fluctuation, the weight of the synergy parameter can be set to 60%, and the weight of the risk parameter can be set to 40%; vice versa.

[0108] After determining the weights, the mapping function linearly fuses the two normalized parameters through weighted average calculation. The calculation formula of the mapping function can be expressed as: quality risk assessment coefficient = (synergy parameter x weight one) + (risk parameter x weight two). Wherein, the sum of weight one and weight two is 1. In addition, in order to improve the sensitivity of the mapping function to extreme risk changes, the mapping function can introduce a nonlinear mapping mechanism, such as an exponential weighting function, a logistic regression function or a dynamic adjustment model based on support vector machine, thereby enhancing the response ability and accuracy to risk change trend.

[0109] The generated quality risk assessment coefficient is also normalized to the interval [0, 1], where the value closer to 1 indicates a higher risk level of the current production process, and the value closer to 0 indicates a lower risk level of the current production process. The quality risk assessment coefficient not only provides an immediate assessment result of the risk state of the production process, but also has the characteristics of quantization, continuity and controllability, which facilitates efficient comparison with risk level threshold, and further drives subsequent differentiated production optimization decisions.

[0110] Through this mapping function system, scientific transformation of multi-dimensional risk information into a single risk index is realized, ensuring that the evaluation results can not only fully reflect the actual process and quality fluctuation, but also support the real-time decision-making needs of the production site, improving the intelligent level and application value of quality risk assessment.

[0111] Step five: Implement differentiated production optimization measures based on risk level categories. When the risk level category is high risk, execute the supplementary production line production stop instruction and process parameter review instruction. When the risk level category is medium risk, execute the capacity restriction instruction and quality inspection enhancement instruction. When the risk level category is low risk, execute normal production instructions and record risk monitoring data. At the same time, record and feedback the risk assessment input data, risk assessment model adjustment path, risk level results and production optimization execution results to achieve continuous optimization of the quality risk assessment model.

[0112] In order to implement differentiated production optimization measures based on production risk levels, ensure that quality risks in the production process are promptly and effectively controlled, and provide real-time response plans for quality problems that may arise in the production process, the following are specifically included:

[0113] After determining the quality risk level of the production process, appropriate production optimization measures are implemented based on the risk level category. When the production status is assessed as high-risk, a production stop instruction for the supplementary production line is immediately executed. This means that in high-risk situations, all risk-related production links will be suspended to prevent unqualified products from further entering subsequent processes or the market. At the same time, process parameter review instructions are executed for process links that may pose high risks to reassess the rationality and feasibility of the current process parameter settings and ensure that the process parameters are within a safe range. This step may include recalibrating production equipment, adjusting production parameters, retesting environmental conditions, etc., to minimize potential risks in process execution.

[0114] When the production risk assessment result is a medium-risk category, the production process will not be completely stopped, but more flexible and targeted optimization measures will be taken. First, capacity restriction instructions will be implemented to appropriately reduce production in production links where quality risks may exist, thereby reducing the generation and spread of substandard products. Capacity restrictions can be achieved not only by limiting the production quantity per hour, but also by adjusting the production rhythm and increasing the frequency of process inspections to improve product quality stability. In addition, quality inspection enhancement instructions will also be implemented, that is, strengthening quality inspection efforts on key process links and quality control points, increasing the frequency and accuracy of quality inspections, and ensuring that potential quality problems are discovered and corrected in a timely manner to prevent increased risks.

[0115] When the production status is assessed as low-risk, it indicates that there are no significant quality risks in the current production process and normal production processes can continue. However, although the production process is relatively stable, ongoing risk monitoring and data recording are still required. Therefore, normal production instructions should be implemented and the existing production process maintained, while all key quality indicators should be monitored and recorded in real time to ensure that any potential problems in the production process can be promptly identified and reported. This step also requires the establishment of an automatic alarm and early warning system to immediately issue an alarm and activate an emergency response mechanism when anomalies in quality monitoring data are detected.

[0116] During each production optimization implementation, detailed records must be kept of all risk assessment input data, risk assessment model adjustment paths, risk level results, and production optimization execution results. The purpose of recording this data is to provide data support and feedback for subsequent optimization and upgrades of the quality risk assessment model. Recorded data should include not only the various parameters used in the risk assessment but also specific response measures for each risk level and their implementation results, to facilitate analysis of the actual effectiveness and feasibility of different optimization measures.

[0117] All recorded risk assessment data, model adjustment paths, risk level results, and optimization action execution results are fed back into the quality risk assessment database. This continuous feedback mechanism not only optimizes the current risk assessment model but also provides valuable historical data support for future quality risk predictions. Each feedback feed automatically adjusts the model, making it more predictive and accurate, further enhancing quality risk management capabilities throughout the entire production process.

[0118] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0119] The above embodiments can be implemented in whole or in part via software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. A computer program product comprises one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions according to the embodiments of the present application are fully or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means (e.g., infrared, wireless, microwave, etc.). A computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0120] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0121] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0122] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative. For example, the division of units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0123] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0124] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0125] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risks, characterized by: The specific steps include: Step 1: Identify and classify production line data sources in the vehicle production process in real time based on data source characteristics to determine whether the data source belongs to a supplementary production line data source. Data source characteristics include device model, sensor type, data format, and data upload frequency. Step 2: If the production line data source is confirmed to be a supplementary production line data source, dynamically adjust the multi-source data fusion rules based on changes in data source characteristics. Standardize the unit, precision, frequency, and format of the supplementary production line data source through preset data mapping rules to ensure consistency with the data format of the main production line data source, and simultaneously adjust the data fusion weight parameters. Step 3: After completing the adjustment of the multi-source data fusion rules, the risk assessment model is dynamically adjusted based on the adaptability of the data source characteristics. By calculating the similarity of process parameters, equipment characteristics, and historical defect characteristics, the risk assessment model with the highest adaptability is selected or the correction parameters are loaded to generate an adapted risk assessment model. Step 4: Input the adjusted data fusion results, calculate the synergy parameter and risk parameter based on normalization, the synergy parameter represents process consistency, and the risk parameter represents the fluctuation level, input the preset mapping function to generate the quality risk assessment coefficient, and determine the risk level of the production status based on the comparison result between the generated quality risk assessment coefficient and the preset risk level threshold range; Step five: Implement differentiated production optimization measures based on the risk level of the determined production status, and record and provide feedback on the risk assessment input data, risk assessment model adjustment path, risk level results, and production optimization execution results.

2. The full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risk according to claim 1 is characterized in that: Step 1 specifically includes the following sub-steps: Using raw production line data from the vehicle production process as input, a multi-dimensional data source feature information set is established, including device model, sensor type, data format, and data upload frequency. Standardize each piece of raw feature information to form a structured data source feature vector set, including unit conversion, format unification, and time series completion; Call the preset comparison rule set to compare the standardized data source feature vector with the main production line reference feature vector to generate a production line consistency score; The production line data source category is confirmed in real time based on the production line consistency score and the set threshold. Specifically, when the production line consistency score is less than or equal to the set threshold, the production line data source is confirmed to be classified as a supplementary production line data source; when the production line consistency score is greater than the set threshold, the data source is confirmed to be a main production line data source. And output the classification results for subsequent multi-source data fusion rule adjustment and risk assessment model optimization.

3. The full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risk according to claim 1 is characterized in that: Step 2 specifically includes the following sub-steps: Based on the confirmed feature change information of the supplementary production line data source, the feature change information of the equipment model characteristics, sensor type characteristics, data format characteristics, and data upload frequency characteristics corresponding to the supplementary production line data source is extracted in real time to form a feature change information set of the supplementary production line data source; Based on the feature change information set of the supplementary production line data source, the preset multi-dimensional data mapping rule set is called to standardize the supplementary production line data source, including differences in equipment model, sensor accuracy, data format, and data upload frequency, to ensure that the supplementary production line data source is consistent with the main production line data source in terms of unit, accuracy, frequency, and format. Based on the standardized supplementary production line data source, the fusion weight adjustment parameter is calculated. By statistically analyzing the historical volatility index, stability index, and credibility index of the supplementary production line data, the weight adjustment coefficient of the supplementary production line data source is calculated using a preset weighted scoring function. According to the calculated weight adjustment coefficient, the weight ratio of the supplementary production line data in the multi-source data fusion is dynamically adjusted, and under the guidance of the updated data fusion rules, the real-time fusion operation of multi-source data is performed to merge the supplementary production line data source and the main production line data source into a data set with unified format, unified standard and unified weight for use in subsequent risk assessment models.

4. The full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risk according to claim 1 is characterized in that: Step 3 specifically includes the following sub-steps: Based on the production line data fusion results dynamically adjusted by the completed multi-source data fusion rules, multi-dimensional feature information of the supplementary production line data source and the main production line data source is extracted to form a production line feature information set of process parameter feature information, equipment feature information and historical defect feature information; A multi-dimensional similarity calculation method is used to calculate the similarity between the supplementary production line data source and the main production line data source in the process parameter feature dimension, equipment feature dimension, and historical defect feature dimension. The process parameter feature similarity is calculated using the Euclidean distance function, the equipment feature similarity is calculated using the cosine similarity function, and the historical defect feature similarity is calculated using the weighted overlap coefficient. Based on the similarity of process parameter features, equipment features, and historical defect features, a preset feature weight allocation mechanism is used to merge the three similarity values ​​into a single comprehensive feature fitness value through the weighted average method; The calculated comprehensive feature fitness value is compared with the preset risk assessment model selection threshold. When the comprehensive feature fitness value is greater than or equal to the risk assessment model selection threshold, the main production line risk assessment model is selected; when the comprehensive feature fitness value is less than the threshold, correction parameters for the supplementary production line feature differences are loaded or a new model version is derived; The dynamically adjusted risk assessment model is applied to the quality risk assessment process to generate a quality risk assessment coefficient. Combined with the preset risk level threshold, the risk level category of the current production status is determined. At the same time, all assessment data and risk assessment model adjustment paths are recorded and fed back for subsequent risk assessment model optimization and automatic learning.

5. The full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risk according to claim 4 is characterized in that: A multi-dimensional similarity calculation method is used to calculate the similarity between the supplementary production line data source and the main production line data source in the process parameter feature dimension, equipment feature dimension, and historical defect feature dimension. Specifically: In the process parameter feature dimension, key process parameters reflecting the process status are extracted and standardized. The Euclidean distance function is used to calculate the similarity of process parameters between the supplementary production line data source and the main production line data source to reflect the degree of similarity in process control levels between the two production lines. In the equipment feature dimension, the technical performance indicators of production equipment are extracted and converted into vector representations. The cosine similarity function is used to calculate the similarity of equipment features between the supplementary production line data source and the main production line data source to quantify the consistency level between the equipment technical parameters. In the dimension of historical defect features, we extract the quality defect data of the supplementary production line and the main production line in the past production process, and construct a historical defect feature set. We use the weighted overlap coefficient method to calculate the similarity of the historical defect feature set, comprehensively evaluate the similarity of the defect performance of the two production lines, and generate the corresponding defect feature similarity.

6. The full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risk according to claim 1 is characterized in that: Step 4 specifically includes the following sub-steps: The dynamically adjusted multi-source data fusion results are used as input and normalized to eliminate the differences in the numerical scales of different production line data dimensions and ensure that the data of each dimension is compared and calculated under a unified standard, with a unified data range of [0,1]. After normalization, the synergy parameter and risk parameter are calculated. The synergy parameter is used to measure the difference in process consistency between the supplementary production line data source and the main production line data source. The process parameter difference is calculated using the Euclidean distance function. The risk parameter is used to measure the volatility and risk level of the supplementary production line data source compared to the historical process benchmark. The risk is quantified by calculating the fluctuation range of the current process parameters compared to the historical process benchmark. The calculated synergy parameters and risk parameters are used as input to the preset mapping function for processing. The synergy parameters and risk parameters are fused through weighted summation to generate a quality risk assessment coefficient, which is used to reflect the overall quality risk level of the current production process. The generated quality risk assessment coefficient is compared with the preset risk level threshold interval, and the risk level of the production status is determined based on the comparison results. The specific comparison analysis is as follows: if the quality risk assessment coefficient exceeds the upper limit value of the preset risk level threshold interval, the risk level of the production status is high risk; if the quality risk assessment coefficient is between the upper limit value and the lower limit value of the preset risk level threshold interval, the risk level of the production status is medium risk; if the quality risk assessment coefficient is lower than the lower limit value of the preset risk level threshold interval, the risk level of the production status is low risk.

7. The full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risk according to claim 6 is characterized in that: The specific method for calculating the synergy parameter and risk parameter is as follows: Extract key process parameter data for the same process steps from the supplementary production line data source and the main production line data source, and normalize the extracted key process parameter data to eliminate the impact of different units and dimensions, ensuring that all process parameters are compared under a unified standard; Based on the normalized key process parameter data, the Euclidean distance function is used to calculate the numerical difference between the supplementary production line data source and the main production line data source in the key process parameter space. The Euclidean distance function obtains the corresponding distance value by summing the squares of the differences between each pair of corresponding process parameters and then taking the square root. Based on the calculated distance value, a preset difference mapping function is called to convert the distance value into a collaborative parameter value. The collaborative parameter value is limited to the range of 0 to 1. The preset difference mapping function is executed based on the correspondence between the distance value and the set threshold value; Extract the real-time monitoring data of each key process parameter in the current production cycle of the supplementary production line, and compare it with the corresponding process parameters in the historical process benchmark model. Use the preset fluctuation amplitude calculation function to calculate the deviation amplitude and change trend between the current process parameters and the historical process benchmark, and quantify them into risk parameters.

8. The full-chain optimization method for collaborative assessment of multi-source data on vehicle production quality risk according to claim 1 is characterized in that: Step 5 specifically includes: After determining the risk level of the production status, corresponding production optimization measures are implemented according to the determined risk level. Specifically, when the risk level of the production status is high risk, the production stop instruction and process parameter review instruction of the supplementary production line are executed; when the risk level of the production status is medium risk, the capacity restriction instruction and quality inspection enhancement instruction are executed; when the risk level of the production status is low risk, the normal production instruction is executed and the risk monitoring data is recorded; During the implementation of each production optimization measure, all involved risk assessment input data, risk assessment model adjustment path, risk level results and production optimization execution results are recorded in detail and fed back into the quality risk assessment database to support subsequent optimization and upgrading of the risk assessment model.

Images