Process parameter optimization and quality risk early warning method and device

Abstract

The invention discloses a process parameter optimization and quality risk early warning method and device, relates to the field of high-end equipment manufacturing, and is used for guaranteeing the product quality under a dynamic production link. The method comprises the following steps: acquiring multi-source production data of the flexible reconfigurable manufacturing system; key indexes influencing the product quality are identified from the multi-source production data; based on the key index test value, process parameters of the process section are adjusted and tested in a self-adaptive optimization mode; performing quality risk early warning based on the optimized process parameters; and carrying out iterative optimization on the process parameters according to a quality risk early warning result. According to the method, the key indexes influencing the product quality are accurately identified and optimized, the process parameter adjustment pertinence is improved, and the optimization efficiency is improved; the spec parameter range is shrunk in a self-adaptive mode, and the proportion of critical risk products is reduced; the cross-process-section rework cost is reduced through a grading early warning mechanism; the method is subjected to closed-loop iteration and continuous updating, and the continuous adaptation relation with the dynamic environment is kept.

Description

Technical Field

[0001] This invention relates to the field of high-end equipment manufacturing, and in particular to a method and apparatus for optimizing process parameters and providing early warning of quality risks. Background Technology

[0002] In the process of modern manufacturing transforming towards intelligent manufacturing, reconfigurable manufacturing systems (RMS) are gradually becoming the production systems that can cope with fluctuations in market demand and uncertainties in production. Under this type of manufacturing system, the production process is no longer a traditional linear and static flow, but a complex system full of dynamic changes, involving numerous aspects such as procurement and manufacturing outsourcing, production operations, after-sales service and maintenance, equipment measurement and maintenance, etc. It is influenced by the interaction of multiple factors including design, people, machines, materials, methods, environment, and measurement, exhibiting significant spatiotemporal uncertainties. In this context, appropriate process parameters become key factors in ensuring product quality and improving production efficiency.

[0003] As core factors influencing the forming quality, performance, efficiency, and cost of parts, the optimization of process parameters has always been a key research focus in the manufacturing industry. Each product's initial specifications (spec) are set based on theoretical calculations, historical data, and expected performance during the initial design phase. However, in actual production, due to factors such as material properties, processing technology, and environmental conditions, these initial specifications often fail to fully adapt to actual production needs. For example, in the debugging and testing phase of motherboard production, some products' functional test indicators often fall within the critical range of spec parameters. While these products may pass the current stage of testing, they may fail subsequent processing or acceptance tests due to minor disturbances, especially in some critical functional tests. Adding to the complexity, products often have dozens of functional test indicators, some of which have negligible impact on overall quality. Blindly adjusting all indicators not only increases optimization costs but may also lead to new quality problems due to excessive intervention. Therefore, accurately identifying key indicators affecting product quality and optimizing process parameters accordingly has become the core challenge in process parameter optimization.

[0004] Meanwhile, quality risk prediction, as a preliminary step in quality control, is becoming increasingly important. During product testing, monitoring subtle changes in performance indicators can capture early signs of quality problems. However, existing quality risk prediction methods largely rely on fixed thresholds and lack a linkage mechanism with process parameter optimization. Even after determining performance indicators and obtaining recommended spec parameters, it is difficult to control product quality, leading to a large number of risky products flowing into subsequent stages and causing a surge in rework costs across process stages. For example, a batch of motherboards had indicator values ​​close to the lower limit of the spec during the debugging stage, and although it was judged as qualified, the indicator deteriorated during subsequent high-temperature environmental testing, causing overall functional failure. The rework cost alone accounted for more than 15% of the production cost of that batch.

[0005] Existing technical solutions have significant shortcomings in addressing the aforementioned problems: First, they blindly optimize performance indicators, relying heavily on manual experience or simple statistical analysis to identify optimization targets, making it difficult to accurately pinpoint key factors from a massive dataset. Second, their fixed process parameter optimization models are incompatible with changing production conditions. Third, quality risk warnings are disconnected from process parameter optimization, easily leading to product quality control failures. Fourth, they lack the capacity to process multi-source heterogeneous production data, failing to effectively integrate information such as process parameters, test results, and environmental data scattered across different systems, resulting in difficulties in ensuring the integrity and consistency of algorithm input data. These shortcomings make traditional methods extremely unsuitable for flexible, reconfigurable manufacturing systems, necessitating a novel method for process parameter optimization and quality risk warning to meet the quality control needs of dynamic production environments. Summary of the Invention

[0006] The purpose of this invention is to provide a method and apparatus for optimizing process parameters and providing early warning of quality risks, addressing all or part of the problems mentioned above, so as to ensure product quality in dynamic production processes.

[0007] The technical solution adopted in this invention is as follows: A method for optimizing process parameters and providing early warning of quality risks, comprising: S1. Acquire multi-source production data from flexible and reconfigurable manufacturing systems; S2. Identify key indicators affecting product quality from the multi-source production data; S3. Adaptively optimize the process parameters of the test section based on the test values ​​of key indicators; S4. Conduct quality risk early warning based on optimized process parameters; S5. Iteratively optimize process parameters based on the results of quality risk warning.

[0008] On the other hand, the present invention also provides a process parameter optimization and quality risk early warning device, which includes a processor and a storage medium, wherein the storage medium stores a computer program, and the processor runs the computer program to execute the above-described process parameter optimization and quality risk early warning method.

[0009] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This application aggregates massive amounts of production data from multiple subsystems. It first identifies key indicators (KPIs) that significantly impact product quality based on this heterogeneous production data. Targeted optimization of these KPIs reduces unnecessary optimization work and improves the efficiency of process parameter optimization. Furthermore, this application uses optimized process parameters for quality risk early warning. Both are iterated in a synchronous closed loop, ensuring consistency between product risk warnings and process parameter optimization, and continuous adaptation to the dynamic environment, thus enhancing the ability to control product quality. This application dynamically optimizes process parameters based on KPI test values, improving the adaptability of process parameters to the real-time production environment and maintaining excellent optimization results at all times, thereby guaranteeing product quality. Simultaneously, this application implements tiered management of high-risk products, allowing for pre-emptive rework of high-risk products and reducing rework costs across process stages. Attached Figure Description

[0010] The present invention will be described by way of example and with reference to the accompanying drawings, wherein: Figure 1 This is a flowchart of a process parameter optimization and quality risk early warning method in one embodiment.

[0011] Figure 2 This is a flowchart of a process parameter optimization method in one embodiment.

[0012] Figure 3 This is a flowchart of an embodiment of a process parameter iterative optimization method. Detailed Implementation

[0013] All features disclosed in this specification, or all steps in all disclosed methods or processes, may be combined in any way, except for mutually exclusive features and / or steps.

[0014] Any feature disclosed in this specification (including any appended claims and abstract) may be replaced by other equivalent or similar features, unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is merely one example of a series of equivalent or similar features.

[0015] This application proposes a method for optimizing process parameters and providing early warning of quality risks, such as... Figure 1 As shown, it includes: S1. Obtain multi-source production data under the flexible reconfigurable manufacturing system RMS.

[0016] The RMS production line involves multiple subsystems, including a Product Lifecycle Management (PLM) system, a Manufacturing Execution System (MES) system, and a testing equipment management system. The production process data from these subsystems constitutes the multi-source production data under the RMS. For each product produced, the RMS production line generates corresponding production process data. This data can be linked across multiple subsystems using a unique product identifier (such as a serial number), and timestamps can be used to link the production process data of the same product at different stages, ensuring data traceability.

[0017] For example, multi-source production data includes: product identification, product model, specifications, design parameters, production time, equipment number, test result data (including the passability of each product indicator and the specific indicator test value), lower specification limit (LSL), lower specification limit (USL), etc.

[0018] As an optional implementation, step S1 includes the following sub-steps: S11, Data Acquisition Phase.

[0019] Step S11 is responsible for collecting multi-source production data. In this sub-step, based on the product's production timeline, production process data is collected from each subsystem of the RMS production line, and the production process data obtained from each subsystem is linked using the product identifier as the keyword.

[0020] S12, Data Preprocessing Stage.

[0021] Step S12 is responsible for preprocessing the collected production process data to facilitate subsequent steps.

[0022] In step S12, differentiated preprocessing strategies are adopted based on the characteristics of different data types. Specifically, for numerical data (such as test values, environmental parameters, etc.), outliers are identified by combining the 3 Sigma principle with the interquartile range method. For values ​​deviating from the reasonable range, production logs are used to determine whether the cause is equipment failure or human error. If it is the former, the data is corrected; if it is the latter, it is marked and removed. For missing values, if the missing percentage is less than 5% (or other values), the mean of the same indicator in the same batch is used to imput the missing values; if the missing percentage is high (reaching 5% or above), prediction and imputation are performed based on the values ​​of relevant indicators using random forest models. For textual data (such as test result descriptions), manual standardization is performed, unifying expressions such as "qualified" and "passed" corresponding to passing the test to "pass" (or other equivalent labels), and unifying expressions such as "unqualified" and "failed" corresponding to failing the test to "fail" (or other equivalent labels), to facilitate subsequent algorithm processing.

[0023] S2. Identify key indicators that affect product quality from multi-source production data.

[0024] Step S2 is responsible for evaluating the impact weight of multi-source production data (test result data) on product quality and screening out the test indicators (i.e. key indicators) that need to be focused on in order to determine the optimization targets.

[0025] As an optional implementation, step S2 includes the following sub-steps: S21, Candidate Indicator Identification Stage.

[0026] In step S21, test indicators that affect product quality are identified through correlation analysis and used as candidate indicators. For example, test indicators that cause product quality tests to fail due to poor test values.

[0027] Specifically, association analysis employs association rule algorithms (such as the Apriori algorithm) to mine frequent itemsets from the passability data of the test and the final test results (recorded in the scene2_firstpassyield table), and then to discover the association rules between the passability status of each indicator and the final product failure. By configuring indicators such as antecedent_support, consequent_support, and confidence, a strong association rule of "low passability of indicator (judged by a set threshold) → final product failure" is trained, and indicators that meet this association rule are selected as candidate indicators.

[0028] S22, Candidate indicator screening stage.

[0029] For the selected candidate indicators, the chi-square test was used to verify the statistical significance of each candidate indicator with the product quality test results. A p-value of 0.1 (or other values) was set, and the chi-square test value was compared with the p-value. When the chi-square test value of an indicator was less than the p-value, the indicator was considered to have a significant association with the product quality test results. Based on the calculated chi-square test values ​​of each candidate indicator, candidate indicators with chi-square test values ​​less than the p-value were selected.

[0030] S23, Candidate Indicator Determination Stage.

[0031] For candidate indicators selected by the chi-square test, the correlation between each candidate indicator is analyzed. For highly collinear candidate indicators (such as candidate indicators with an absolute correlation coefficient greater than 0.8 (or other values), one is retained and the rest are removed (such as removing candidate indicators with lower correlation ranking) to avoid redundancy.

[0032] By following the steps above, key indicators that significantly impact product quality can be identified from the candidate indicators. Optimization around these key indicators can effectively improve the targeting of process parameter adjustments. For ease of description, the operation in step S2 is defined as the key indicator identification model.

[0033] S3. Adaptively optimize the process parameters (spec parameters) of the test section based on the test values ​​of key indicators.

[0034] Step S3 employs a multi-dimensional spec parameter adjustment strategy. Under the premise of ensuring that the product defect rate is controllable (within the set threshold), the proportion of risky products is reduced by narrowing the range of spec parameters based on the real-time test values ​​of key indicators.

[0035] As an optional implementation method, such as Figure 2 As shown, step S3 includes the following sub-steps: S31, Process parameter optimization strategy design stage.

[0036] In step S31, for each key indicator, a multi-dimensional adjustment strategy was designed to optimize the process parameters: Adjustment Strategy 1: Proportional Adjustment Strategy. Based on the distribution of key performance indicators (i.e., spec parameter distribution) of qualified products in historical multi-source production data, the upper and lower limits of the spec parameters are contracted according to predetermined proportions (such as the lower limit adjustment rate LSL_ratio and the upper limit adjustment rate USL_ratio). For example, the original LSL is adjusted upward by 10%, and the original USL is adjusted downward by 10%.

[0037] Adjust strategy two, Adjust the strategy. Centering on the mean of the key performance indicator (KPI) test values ​​of qualified products from historical multi-source production data, set LSL and USL to the mean minus three times the standard deviation (the standard deviation of the KPI test values ​​of qualified products from historical multi-source production data). The mean increased by three times the standard deviation (i.e.) ,if The adjustment strategy is to use a factor of 6 (standard deviations). This utilizes the characteristics of a normal distribution to control the non-conforming rate.

[0038] Adjustment strategy three: IQR (Interquartile Range) adjustment strategy. Based on the IQR of key performance indicator (KPI) test values ​​of qualified products from historical multi-source production data, set LSL to the lower quartile minus 1.5 times (or other nearby values) of the IQR, and set USL to the upper quartile plus 1.5 times the IQR. This adjustment strategy is particularly suitable for process parameters that are not normally distributed.

[0039] Each adjustment strategy records the adjusted LSL and USL (referred to as adjust_LSL and adjust_USL, respectively), as well as the estimated number of nonconforming products based on adjust_LSL and adjust_USL, to estimate the change in the product nonconforming rate. For example, fail_num_ratio, fail_num_3sigma, and fail_num_IQR represent the estimated number of nonconforming products after adjustment by the above three adjustment strategies, respectively.

[0040] S32, Effect Evaluation Phase.

[0041] Step S32 evaluates the changes in product quality caused by each adjustment strategy based on the process parameters adjusted in each dimension in step S31.

[0042] Because the specifications are narrowing towards the middle, when evaluating the process adjustment instructions based on the adjusted process parameters, products that were originally within the critical range of the (original) process parameters may be determined to be non-conforming after the adjustment, thus increasing the product non-conforming rate. This increased non-conforming rate will increase the number of products reworked in the adjustment process, increasing rework costs. However, the increased number of reworked products in the adjustment process will reduce the final product failure rate (which can be understood as the product recall rate). Products that ultimately fail require recall and repair, therefore, a lower "final product failure rate" will reduce repair costs. The cost changes caused by process parameter optimization include increased rework costs (the former) and reduced repair costs (the latter). In terms of cost, if the latter is higher than the former, then optimizing process parameters is meaningful. If we favor the final failure rate, then regardless of whether the latter is higher than the former, optimizing process parameters is meaningful as long as the final product failure rate is reduced.

[0043] In step S32, a multi-objective evaluation model is constructed, with the "increase in product defect rate" of the adjusted test process segment and the "decrease in final product failure rate" of the complete machine as the core indicators, to quantitatively evaluate the above three adjustment strategies.

[0044] For example, calculations show that a proportional adjustment strategy increases the product defect rate by 3% and reduces the final product failure rate by 25%. Adjusting the strategy resulted in a 5% increase in the product defect rate, but a 30% decrease in the final product fail rate. Adjusting the IQR strategy resulted in a 4% increase in the product defect rate, but a 28% decrease in the final product fail rate.

[0045] Different application scenarios have different requirements for the adjusted test results, and the results of the various adjustment strategies mentioned above may not all meet the requirements. You can use one or more of the following criteria—product defect rate (or increase) or final product failure rate (or decrease)—as screening standards to select adjustment strategies from the various options that meet these criteria. For example, as an optional implementation method, you can select adjustment strategies whose results meet the tolerance threshold (e.g., no more than 5%) set for the increase in the product defect rate based on the application scenario.

[0046] S33, Process parameter determination stage.

[0047] In step S33, based on the desired product performance indicators, the final process parameters are determined from the selected adjustment strategies.

[0048] In step S33, the selected adjustment strategies are further refined. Monte Carlo simulations are used to predict the changes in product quality risk (pass / fail) and cost (rework cost before and after process parameter optimization) under different initially selected adjustment strategies. Based on the preferred requirements for one or more product indicators (product defect rate, final product failure rate, or cost change), the optimal adjustment strategy, i.e., the optimal spec parameters, is determined from the initially selected adjustment strategies. For example, based on the cost priority principle, the process parameters optimized by the adjustment strategy with the largest cost reduction before and after optimization are used as the final process parameters; based on the quality priority principle, the process parameters optimized by the adjustment strategy with the largest reduction in the final product failure rate are used as the final process parameters; based on the balance priority principle, the cost reduction and the final product failure rate reduction are weighted and summed according to the set weights, and the process parameters optimized by the adjustment strategy with the highest sum value are used as the final process parameters; based on the low rework priority principle, the process parameters optimized by the adjustment strategy with the smallest increase in the product defect rate before and after optimization are used as the final process parameters. Obviously, the final process parameters can also be determined based on other combinations of these indicators.

[0049] For example, in the optimization of motherboard ZY process parameters, the final selection The strategy was adjusted, reducing the LSL from 30 to 32.11 and the USL from 90 to 85.45. This adjustment increased the product defect rate by 4.2%, falling below the tolerance threshold of 5%. It is projected that the final product failure rate will decrease by 28%. These adjusted process parameters will be used as the new process parameters for the testing and adjustment phase.

[0050] S4. Conduct quality risk early warning based on optimized process parameters.

[0051] Step S4, based on the optimized process parameters, enables real-time identification and risk level classification control of risky products.

[0052] As an optional implementation, step S4 includes the following sub-steps: S41, Key Indicator Risk Threshold Determination Stage.

[0053] In step S41, the risk threshold of key indicators is determined to identify the risk level of key indicators.

[0054] As an optional implementation method, combining optimized spec parameters with product business knowledge, product risks (reflected by indicator risks) are divided into three risk ranges from high to low: Level 1 Risk: The test value of the key indicator exceeds the optimized spec parameter, i.e., it is lower than adjust_LSL or higher than adjust_USL. Products at Level 1 Risk (i.e., products with Level 1 Risk indicators, the same applies below) are directly judged as unqualified.

[0055] Level 2 Risk: Key indicator test values ​​fall within the critical range of the optimized spec parameters, specifically between adjust_LSL and adjust_LSL+5% (adjust_USL-adjust_LSL), or between adjust_USL-5% (adjust_USL-adjust_LSL) and adjust_USL. These four points constitute the Level 2 risk threshold. The critical range (5%) is adjustable. Products at Level 2 risk are considered to have a high risk of quality defects.

[0056] Level 3 Risk: Although the key indicator test values ​​are within the optimized spec parameter range and not within the critical range (i.e., within the Level 2 risk threshold range), they still deviate from the distribution center of historical high-performing product key indicator test values ​​by a certain distance. The Level 3 risk threshold is defined as the distance the distribution center deviates outward from the predetermined distance. Clearly, the Level 3 risk threshold is closer to the distribution center than the Level 2 risk threshold. For example, a deviation from the average key indicator test values ​​of historical high-performing products exceeding [a certain value]... (or other values). Products at level three risk require enhanced follow-up testing.

[0057] Products not classified as being in the above three risk levels are considered products that do not require intervention. The risk thresholds for classifying the three risk levels can be dynamically adjusted according to the judgment rules corresponding to the product type. For example, for medical equipment with extremely high reliability requirements, the proportion of the critical range in the second-level risk level can be reduced from 5% to 3%.

[0058] S42, Quality Risk Early Warning Stage.

[0059] A quality risk early warning model is constructed, with key indicator test values ​​as input features and product risk indicators as output features. The test values ​​of each key indicator are determined to have a risk level based on a set risk threshold. Finally, the quality risk early warning model outputs the key indicators that are at risk (i.e., key indicators whose test values ​​fall within the risk threshold range) and their corresponding risk levels.

[0060] S43, Risk Product Classification and Disposal Stage.

[0061] For products with different risk levels, different handling methods are set in advance, and the products are handled in accordance with the corresponding handling methods based on the product risk level.

[0062] For example, products with Level 1 risk are reworked directly; products with Level 2 risk are marked as having higher quality risks and the frequency of testing is increased; and products with Level 3 risk are further tested. This reduces rework costs across different testing and commissioning stages.

[0063] S5. Iteratively optimize process parameters based on the results of quality risk warning.

[0064] After assessing the risk indicators based on the test values ​​of key indicators, step S5 iteratively optimizes the method based on the assessed risk indicators and makes adjustments by periodically evaluating the effectiveness of each stage.

[0065] As an optional implementation method, such as Figure 3 As shown, step S5 includes the following sub-steps: S51, Effectiveness Evaluation Phase.

[0066] The effectiveness of process parameter optimization and quality risk early warning can be quantitatively evaluated. This evaluation can be conducted at a set frequency (e.g., monthly) or periodically. The effectiveness of process parameter optimization is measured by indicators such as "reduction in final product failure rate," "increase in product non-conforming rate," or "cost change." For example, a larger reduction in the final product failure rate, a larger cost reduction, and a tolerable increase in the product non-conforming rate indicate a better process parameter optimization effect. The effectiveness of quality risk early warning is measured by indicators such as "risk identification accuracy," "false alarm rate," or "missed alarm rate." For example, a higher risk identification accuracy, a lower false alarm rate, and a lower missed alarm rate indicate a better quality risk early warning effect.

[0067] When the evaluation results (e.g., a negative cost reduction, or a false alarm rate of 12% for a batch of products with estimated secondary risk) meet the update conditions, process parameter optimization is triggered.

[0068] S52, Parameter Correction Stage.

[0069] When the evaluation results of the optimization effect of process parameters or the effect of product risk warning trigger the update conditions, the weights of key indicators and the corresponding process parameters are adjusted.

[0070] For example, if the actual impact of a key indicator is lower than expected after evaluation, its weight in product quality is reduced, meaning the key indicator and its corresponding process parameters are redefined. If the optimized spec parameters cause the product defect rate to exceed the tolerance threshold, the specification threshold range is appropriately widened, meaning the percentage reduction in each specification indicator is decreased, so that the product defect rate remains within the tolerance threshold.

[0071] S53, Model Update Phase.

[0072] Updates to process parameters will trigger updates to risk thresholds, and the product quality early warning model will be updated accordingly. The quality risk early warning model will be retrained using the latest recorded multi-source production data (i.e., risk thresholds will be redefined) to incorporate the latest process changes and environmental factors.

[0073] For example, when production equipment is updated, the operating data of the new equipment should be promptly added to the training set to ensure the model's adaptability to new production conditions. Furthermore, data processing workflows can be optimized based on user feedback, such as adding an automatic adaptation module for data on new product models to reduce manual intervention. Alternatively, if the false alarm rate of the risk level estimated by the quality risk warning model is too high (e.g., reaching the upper limit of the threshold), the quality risk warning model should be retrained.

[0074] Compared to existing technologies, this application, rather than relying on manual methods based on experience or knowledge, utilizes in-depth mining of multi-source production data, combined with statistical analysis and machine learning methods, to more accurately identify key functional indicators affecting product quality, avoiding ineffective optimization and improving the targeting of process parameter adjustments. Furthermore, by reasonably narrowing the range of spec parameters, the proportion of critically risky products is reduced, constructing a more reasonable dynamic adaptive process parameter optimization mechanism. Simultaneously, this application establishes a quality risk early warning model linked to process parameter optimization, achieving precise control of risky products through a tiered early warning mechanism, reducing rework costs across process stages. Finally, a closed-loop management system is formed: "key indicator identification - process parameter optimization - quality risk early warning - data feedback iteration." By linking early warning results with optimization strategies, the weights of key indicators and the rules for adjusting spec parameters are continuously revised, enabling the method to possess self-learning capabilities and gradually improve its adaptability and accuracy in dynamic production environments.

[0075] This invention is not limited to the specific embodiments described above. The invention extends to any new feature or combination disclosed in this specification, as well as any new method or process step or combination disclosed herein.

Claims

1. A method for optimizing process parameters and providing early warning of quality risks, characterized in that, include: S1. Acquire multi-source production data from flexible and reconfigurable manufacturing systems; S2. Identify key indicators affecting product quality from the multi-source production data; S3. Adaptively optimize the process parameters of the test section based on the test values ​​of key indicators; S4. Conduct quality risk early warning based on optimized process parameters; S5. Iteratively optimize process parameters based on the results of quality risk warning.

2. The method for optimizing process parameters and providing early warning of quality risks as described in claim 1, characterized in that, Key indicators affecting product quality were identified from the multi-source production data, including: S21. Identify candidate indicators that have an impact on product quality through correlation analysis; S22. Use the chi-square test method to screen the candidate indicators; S23. Deduplicatively analyze the selected candidate indicators through correlation analysis.

3. The method for optimizing process parameters and providing early warning of quality risks as described in claim 1, characterized in that, The process parameters of the test and adjustment process section are adaptively optimized based on the test values ​​of key indicators, including: S31. Optimize process parameters by adopting a multi-dimensional adjustment strategy for each key indicator; S32. Evaluate the changes in product quality caused by each adjustment strategy, and preliminarily screen adjustment strategies based on the evaluation results; S33. Based on the preferred requirements for product indicators, determine the optimal adjustment strategy from the initially screened adjustment strategies.

4. The method for optimizing process parameters and providing early warning of quality risks as described in claim 3, characterized in that, The optimization of process parameters using a multi-dimensional adjustment strategy includes: Adopt a proportional adjustment strategy Optimize process parameters by adjusting strategies and interquartile range (IQR) adjustment strategies.

5. The method for optimizing process parameters and providing early warning of quality risks as described in claim 4, characterized in that, Adopt a proportional adjustment strategy Adjustment strategies and interquartile range (IQR) adjustment strategies optimize process parameters, including: Based on the distribution of key indicator test values ​​of qualified products in multi-source production data, proportional adjustment strategies were adopted respectively. Optimize process parameters by adjusting strategies and interquartile range (IQR) adjustment strategies.

6. The method for optimizing process parameters and providing early warning of quality risks as described in claim 3, characterized in that, Based on the preferred requirements for product metrics, the optimal adjustment strategy is determined from the initially screened adjustment strategies, including: Monte Carlo simulations were used to predict the magnitude of product quality risk and cost changes under different initial screening adjustment strategies. Based on the preferred requirement for at least one of the following: product non-conformity rate, product recall rate, or cost change rate, the optimal adjustment strategy is determined from the initially screened adjustment strategies.

7. The method for optimizing process parameters and providing early warning of quality risks as described in claim 1, characterized in that, Quality risk warning based on optimized process parameters, including: S41. Determine the risk threshold for key indicators; S42. Based on the risk threshold, identify risk indicators according to the test values ​​of key indicators. S43. Dispose of the products according to the identified risk indicators and the corresponding disposal methods.

8. The method for optimizing process parameters and providing early warning of quality risks as described in claim 7, characterized in that, Determine the risk thresholds for key indicators, including: The optimized process parameter boundaries are used as the first-level risk threshold; The boundary of the predetermined critical range, which is narrowed inward from the optimized process parameter boundary, is used as the secondary risk threshold. The third-level risk threshold is defined as the deviation of the distribution center of key indicator test values ​​of historical high-quality products from a predetermined distance.

9. The process parameter optimization and quality risk early warning method as described in claim 1, comprising iterative optimization of process parameters based on the quality risk early warning result, including: S51. Quantitatively evaluate the effectiveness of process parameter optimization and quality risk early warning; S52. When the evaluation results meet the update conditions, process parameter optimization is triggered.

10. A process parameter optimization and quality risk early warning device, characterized in that, It includes a processor and a storage medium, the storage medium storing a computer program, the processor running the computer program to perform the process parameter optimization and quality risk early warning method as described in any one of claims 1-9.

Images