A production quality management and control method and system for automobile parts

Abstract

The application discloses a production quality management and control method and system for automobile parts, and relates to the technical field of automatic control, which comprises the following steps: binding a unique identification of a part, collecting multi-source process data, screening a reference sample set, and obtaining a reference parameter set; collecting current part measured data, calculating a joint out-of-bound index, judging the quality state of the part, identifying an abnormal cluster, executing targeted fine tuning, generating statistical quadruples and a candidate process parameter set; obtaining a standardized statistical index set, determining an optimal control instruction through lexicographic optimization, receiving the optimal control instruction, and arranging it into effect; performing feedforward robustness judgment, executing online monitoring, generating a disposal decision, solidifying an effective control instruction, and updating the reference parameter set. The application enhances the adaptability of the control strategy, and realizes continuous optimization and accurate management and control of the production process.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, and in particular to a method and system for quality control in the production of automotive parts. Background Technology

[0002] In the field of modern automobile manufacturing, the quality management of automotive parts production relies on technical systems such as statistical process control, measurement system analysis, and manufacturing execution system. By collecting key process parameters, test data, and energy consumption signals, the production process is monitored and statistically analyzed. The fusion and synchronous processing of multi-source sensor data provides a technical foundation for realizing the visualization of production status and timely detection of process deviations.

[0003] In the complex automotive parts production scenario, conventional quality control methods still face two limitations. On the one hand, existing methods are insufficient in handling the temporal consistency of multi-source heterogeneous process data, making it difficult to achieve accurate data alignment and high robust stability judgment under dynamic production cycles, thus affecting the reliable establishment of benchmark parameters. On the other hand, most process control strategies rely on experience or a single threshold, lacking feedforward correction and dynamic verification mechanisms, resulting in delays in correcting quality deviations and making it difficult to achieve continuous optimization and adaptive control of the production process. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a production quality control method for automotive parts to solve the problems of insufficient real-time consistency of multi-source data and insufficient dynamic adaptability of control strategies in existing technologies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for quality control in the production of automotive parts, comprising, Bind unique component identifiers, collect multi-source process data, perform time alignment and data cleaning, filter reference sample sets, set scrolling windows, and obtain a set of benchmark parameters; Collect current measured data of components, calculate joint out-of-bounds index based on the benchmark parameter set, determine the quality status of components, identify abnormal clusters, perform targeted fine-tuning, verify the fine-tuning effect, and generate statistical quadruplets and candidate process parameter sets. Read and verify the statistical quadruple and candidate process parameter set, obtain the standardized statistical index set, perform hard constraint screening and feasibility judgment, obtain the preferred process parameter set, and determine the optimal control command by lexicographical selection. Receive the optimal control command, arrange and activate it, perform feedforward robustness judgment, execute it on the machine and monitor it online, perform short-window acceptability verification, generate disposal decision, solidify the effective control command, and update the baseline parameter set.

[0007] As a preferred embodiment of the production quality control method for automotive parts described in this invention, the steps of binding unique component identifiers, collecting multi-source process data, and performing time alignment and data cleaning are as follows: Bind the unique identifier of the component, establish a unique traceability master index key, and collect multi-source process data by association according to the unique identifier of the component to obtain the original multi-source process data; Based on a unified time reference, the original multi-source process data is synchronized and cleaned to obtain cleaned and aligned multi-source process data.

[0008] In a preferred embodiment of the production quality control method for automotive parts described in this invention, the steps of filtering the reference sample set, setting a scrolling window, and obtaining the benchmark parameter set are as follows: Based on the cleaning alignment multi-source process data, samples with continuous production and complete data are selected to obtain a reference sample set; Based on the reference sample set, a rolling statistical window is set, a size benchmark and reference curve are constructed, and a benchmark parameter set is generated.

[0009] As a preferred embodiment of the production quality control method for automotive parts described in this invention, the steps of collecting current measured data of the parts, calculating joint out-of-bounds indicators based on a set of benchmark parameters, determining the quality status of the parts, and identifying abnormal clusters are as follows: Using the unique identifier of the component as an index, collect the current measured data of the component, obtain the measured dataset, and calculate the joint boundary violation index by combining it with the set of benchmark parameters; Based on the joint boundary violation index, determine the quality status of the components, generate quality status labels, calculate the proportion of out-of-bounds components, and identify abnormal clusters.

[0010] As a preferred embodiment of the production quality control method for automotive parts described in this invention, the steps of performing targeted fine-tuning, verifying the fine-tuning effect, and generating statistical quadruplets and candidate process parameter sets are as follows: Based on the workstation and process characteristics corresponding to the abnormal clusters, targeted fine-tuning is carried out to obtain the measured data of the fine-tuned parts; Based on the measured data of the fine-tuned components, the joint out-of-bounds index is recalculated to determine whether the fine-tuning is effective, the conclusion of the fine-tuning effect is obtained, and a statistical quadruple and candidate process parameter set are generated.

[0011] As a preferred embodiment of the production quality control method for automotive parts described in this invention, the steps of reading and verifying statistical quadruplets and candidate process parameter sets, obtaining a standardized statistical index set, performing hard constraint screening and feasibility determination, and obtaining an optimal process parameter set are as follows: Read and verify the statistical quadruple and candidate process parameter set, perform robustness preprocessing, obtain the effective candidate process parameter set, and generate a standardized statistical index set. Hard constraint screening is performed on the standardized statistical index set, the feasibility index of each valid candidate process parameter is calculated, and the optimal set of process parameters is obtained.

[0012] As a preferred embodiment of the production quality control method for automotive parts described in this invention, the step of determining the optimal control instruction by lexicographical order selection refers to generating the optimal control instruction by selecting the process parameters with the best comprehensive performance based on the preferred set of process parameters and according to the lexicographical order rule.

[0013] As a preferred embodiment of the production quality control method for automotive parts described in this invention, the steps of receiving optimal control commands, arranging and activating them, performing feedforward robustness judgment, executing them on the machine and monitoring them online, and performing short-window acceptability verification are as follows: Based on the optimal control instruction, the optimal control instruction is written into the control instruction activation table at the production cycle boundary moment, the activation timestamp is recorded, and the activated optimal control instruction is generated. Based on the already effective optimal control instructions, combined with real-time environmental data, the environmental robustness coefficient is calculated, the optimal control instructions are adjusted, and the execution control instructions are generated. The execution control command is bound to the unique identifier of the component and written into the equipment control register. After execution on the machine, the measured data of the continuous production window is collected, the short window acceptance index is calculated, and the short window acceptance verification conclusion is obtained.

[0014] As a preferred embodiment of the production quality control method for automotive parts described in this invention, the specific steps of generating disposal decisions, solidifying effective control instructions, and updating the benchmark parameter set are as follows: Based on the short-window acceptability verification results, the validity of the execution control command is determined, a disposal decision is generated, and the disposal decision and the updated execution parameter set are obtained. Based on the decision and the updated set of execution parameters, the execution control instructions are solidified as valid control instructions. The solidified valid control instructions are obtained, and an updated set of baseline parameters is generated.

[0015] Secondly, this invention provides a production quality control system for automotive parts, comprising, The benchmark acquisition module is used to bind unique identifiers of parts, collect multi-source process data, perform time alignment and data cleaning, filter reference sample sets, set a scrolling window, and obtain a set of benchmark parameters. The quality stability assessment module is used to collect the current measured data of the components, calculate the joint out-of-bounds index based on the benchmark parameter set, judge the quality status of the components, identify abnormal clusters, perform targeted fine-tuning, review the fine-tuning effect, and generate statistical quadruplets and candidate process parameter sets. The process optimization module is used to read and verify the statistical quadruple and candidate process parameter set, obtain the standardized statistical index set, perform hard constraint screening and feasibility judgment, obtain the optimized process parameter set, and determine the optimal control command by lexicographical selection. The control execution module is used to receive optimal control commands, orchestrate their activation, perform feedforward robustness judgments, execute them on the machine and monitor them online, perform short-window acceptability verification, generate disposal decisions, solidify effective control commands, and update the baseline parameter set.

[0016] The beneficial effects of this invention are as follows: by calculating joint out-of-bounds indicators and identifying abnormal clusters, dynamic assessment of the quality status of parts and precise location of abnormal areas are achieved; by performing targeted fine-tuning and verifying the fine-tuning effect, rapid correction of production parameters based on real-time multi-source process data is achieved, improving the accuracy and response efficiency of quality stability judgment; by using hard constraint screening and feasibility determination, the optimal control command is determined, enhancing the adaptability of the control strategy and achieving continuous optimization and precise control of the production process. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Fig. 1 This is a flowchart of the production quality control methods for automotive parts.

[0019] Fig. 2 This is a schematic diagram of a production quality control system for automotive parts.

[0020] Fig. 3 A flowchart for generating statistical quadruples and candidate process parameter sets.

[0021] Fig. 4 A flowchart for generating optimal control commands. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figs. 1-4 As one embodiment of the present invention, this embodiment provides a method for quality control in the production of automotive parts, comprising the following steps: S1. Bind the unique identifier of the component, collect multi-source process data, perform time alignment and data cleaning, filter the reference sample set, set a scrolling window, and obtain the benchmark parameter set.

[0026] Bind the unique identifier of the component, establish a unique traceability master index key, and collect multi-source process data by association according to the unique identifier of the component to obtain the original multi-source process data.

[0027] Each component has a unique DataMatrix ECC200 QR code pre-etched on its surface as its unique identifier. Fixed barcode readers are installed at the entrance of each process on the production line to read the unique identifier of the component as it enters the workstation and synchronize the unique identifier of the component with the trigger signal of the PLC.

[0028] Collect multi-source process data, including process parameters, detection data, process curves, and energy consumption power sequence. Process parameters include, but are not limited to, spindle speed, feed rate, process pressure, humidity, and temperature. Detection data includes, but is not limited to, dimensional deviations. Process curves include, but are not limited to, tightening or pressing curves. Energy consumption power sequence includes, but is not limited to, energy consumption power and single-piece cycle time.

[0029] Using the unique identifier of each component as the index key, data from various sensors, testing equipment, and power meters are written into the same data chain in chronological order, generating data entries for component-process-parameter-test-energy consumption-cycle time. The data entry structure ensures complete backtracking of all process data for each component.

[0030] Based on a unified time reference, the original multi-source process data is synchronized and cleaned to obtain cleaned and aligned multi-source process data.

[0031] To ensure data synchronization from different acquisition terminals, the production line's control network uniformly adopts the IEEE-1588 precision clock synchronization protocol to synchronize the clocks of all PLCs, measuring instruments, power acquisition terminals, and cameras.

[0032] Residual time difference correction is performed on different data sources for the same component. Specifically, the time series of PLC trigger events of the process and the time series of events recorded by the corresponding detection equipment are obtained. The optimal time shift is calculated by adopting the principle of minimizing the median absolute deviation, so that the matching residuals of the PLC trigger event time series of the process and the event time series recorded by the corresponding detection equipment are aligned in a way that minimizes the overall error. All types of data are resampled according to a unified time axis.

[0033] The optimal time shift is expressed as: ; in, Indicates the optimal time shift amount. Represents a time-shifted variable. Represents the median operator. This indicates the index number of the time sequence of the PLC-triggered event. No. The timestamp of each PLC-triggered event Indicates and The timestamp of the event recorded by the most recent detection equipment. This represents the nearest neighbor matching function.

[0034] Data cleaning includes three aspects: time consistency, amplitude reasonableness, and field integrity verification. Specifically, in terms of time consistency, if the continuous sampling time is found to be out of order or the time interval is less than 0.5 times the device's set sampling period, outliers are deleted or merged. In terms of amplitude reasonableness, when the size deviation exceeds ±5 times the design tolerance or the power value suddenly increases to more than 3 times the average value, the sampling point is considered an abnormal measurement and is removed. When the tightening curve shows reverse angle micro-vibration, noise is filtered out by smoothing with 5-point movement.

[0035] It should be noted that setting the equipment sampling period to 0.5 times is an empirical value recommended by the international standard IEC60068 in sampling consistency testing. This ensures that duplicate data is removed without losing valid points, maintaining time accuracy while avoiding excessive cleaning. Setting the design tolerance to ±5 times covers more than 99.9% of the possible physical values, effectively eliminating sampling anomalies while retaining data on process boundary states. Setting the average power value to 3 times is a balance point between equipment inertial response and sampling noise suppression. Verified through multiple production line tests, setting the average power value to 3 times can stably eliminate more than 95% of abnormal peaks without mistakenly deleting normal points. Tightening or pressing curves may exhibit local short-cycle reverse fluctuations in the torque curve under mechanical vibration or sampling jitter. The curve sampling resolution is typically 0.2° / point, and a 5-point moving average is equivalent to a 1° sliding window, effectively filtering out sampling noise frequencies without weakening the main deformation characteristics.

[0036] It should be noted that if a component is missing necessary inspection items in a critical process, the component will be marked as having incomplete data and will not be included in the subsequent reference sample set.

[0037] Based on the cleaning alignment multi-source process data, samples with continuous production and complete data are selected to obtain a reference sample set.

[0038] From the historical data of the stable process stage, 400 consecutive qualified parts with complete data are selected as the initial reference sample set. The rolling statistical window is set to 80 parts. During the operation, the statistical characteristics in the window are updated once every 80 parts are processed to maintain the dynamic benchmark.

[0039] It should be noted that a sample size of 400 pieces can cover minor drifts before and after tool changes, making the reference statistics more robust; a rolling statistical window of 80 pieces can generate a new statistical snapshot within 10 to 20 minutes, providing real-time response and avoiding excessive jitter.

[0040] Based on the reference sample set, a rolling statistical window is set, a size benchmark and reference curve are constructed, and a benchmark parameter set is generated.

[0041] Extract the dimensional deviations of all components from the reference sample set and calculate the median dimensional deviation. and median absolute deviation The dimensional reference range is [ , This can effectively suppress the impact of extreme outliers on the overall interval.

[0042] Record the process capability index for critical dimensions to ensure that the production line meets the requirement of a process capability index greater than or equal to the lower quality limit of 1.67 under steady state.

[0043] It should be noted that a process capability index of 1.67 or higher is a value determined based on international automotive industry quality standards, statistical reliability requirements, and engineering steady-state verification. It represents that the production process has reached a highly reliable and stable state, complies with industry standards, and ensures the statistical robustness and feasibility of subsequent stability assessment and control strategies.

[0044] A reference curve is constructed using phase registration and point-by-point median method. Specifically, each measured process curve in the reference sample set is phase registered and aligned with the previous reference curve. The optimal phase offset is determined by minimizing the integral squared difference between the two curves. The phase-registered and aligned measured process curve is obtained. The median value of the phase-registered and aligned measured process curve is taken at each angle sampling point to generate a new reference curve. The phase registration and point-by-point median process is repeated until the reference curve converges to obtain the final reference curve.

[0045] The optimal phase offset is expressed as: ; in, Indicates the first The optimal phase offset of the measured process curve of each component relative to the reference curve. This represents the phase offset variable to be calculated. and Indicates the starting and ending angles of the integral. Indicates the first The measured process curves of each component, with phase offset variable applied. The function after that, Indicates the first Reference curve for each iteration, Indicates the iteration number of the reference curve. This indicates the rotation angle of the tightening shaft or press head throughout the process. This indicates the component index.

[0046] It should be noted that the initial reference curve is a visually typical measured curve selected from the reference sample set. The curve has no obvious burrs and clear key feature points, and is used as the reference curve for round 0.

[0047] For each measured process curve aligned by phase registration in the reference sample set, the relative deviation value of the curve is calculated using the normalized L2 norm, and the 95th percentile of the relative deviation value is taken as the curve deviation threshold.

[0048] Specifically, the relative deviation of the curve is expressed as: ; in, This represents the relative deviation of the curve. The current tightening curve of the components, This represents the reference curve.

[0049] It should be noted that the 95th percentile of the relative deviation value of the curve was selected as the curve deviation threshold through statistical analysis and engineering verification of historical sample data during the stable production phase. Specifically, when the production line is in a stable operation phase, the tightening or pressing curve samples of 1,000 qualified parts are collected, the relative deviation value of the curve for each part is calculated, and the relative deviation values ​​are sorted and statistically analyzed. The deviation value at the 95th percentile corresponds to the process stability boundary. Using the 95th percentile value as the curve deviation threshold can ensure that 95% of the samples are judged normally, while maintaining high sensitivity for the 5% of abnormal samples. The false positive rate and false negative rate are optimally balanced, which meets the process capability control standards for automotive parts manufacturing.

[0050] Furthermore, the benchmark parameters are solidified into fixed thresholds to obtain a set of benchmark parameters. These fixed thresholds include a size threshold, a curve deviation threshold, a single-unit energy consumption threshold, and a single-unit cycle time lower limit threshold. The size threshold is defined as follows: , The dimensional deviation boundary is used as the boundary; the curve deviation threshold is used as the upper limit of the tightening or pressing curve deviation; the 80th percentile of the energy consumption of a single unit in the reference sample is used as the energy consumption threshold of a single unit; the 20th percentile of the cycle time of a single unit is used as the lower limit threshold of the cycle time of a single unit. All fixed thresholds are derived from real measurable data, and the quantile method ensures statistical stability.

[0051] To obtain the energy consumption of a single component, specifically, the power signal of the production station is read, the start and end times of processing of each component are marked, the energy consumption integration is started with the station's carry signal and ended with the carry signal, the area under the power curve is calculated and the unit is converted, and multiple samples of the same process are repeatedly collected to obtain the energy consumption value of each component. After removing outliers, the median value is taken as the energy consumption of a single component for the reference sample.

[0052] The cycle time of a single part is obtained by synchronously comparing the time signals of the same part at the inlet and outlet of the process. Specifically, the start time is recorded when the barcode scanning signal at the station inlet triggers the start time record, and the end time is recorded when the finished product inspection or unloading signal at the station outlet triggers the end time record. The time difference between the inlet and outlet of the process is used as the cycle time of the single part.

[0053] The set of reference parameters includes, but is not limited to, median and median absolute deviation of dimensional deviations, reference curves, curve deviation thresholds, single-unit energy consumption thresholds, and single-unit cycle time lower limit thresholds.

[0054] It should be noted that the dimensional deviation boundaries are determined based on statistical robustness and process tolerance control principles. Specifically, [ , It can cover the vast majority of normal processing fluctuation samples, eliminate the influence of extreme outliers on the judgment results. Compared with the mean and standard deviation methods, the median and median absolute deviation are not sensitive to outliers and can better reflect the true degree of dispersion under stable process conditions. By setting the upper and lower three median absolute deviations as the dimensional deviation boundaries, it ensures that the dimensional fluctuations of qualified parts are included, timely identifies abnormal parts that deviate from the normal processing range, and achieves effective constraints on the dimensional stability of parts.

[0055] The single-component energy consumption threshold is selected from the 80th percentile of the reference sample single-component energy consumption. This is determined by comprehensively considering the stability of energy consumption distribution and the control of process upper limits. Energy consumption in the production process is distributed to the right with respect to equipment wear, ambient temperature and processing load. The energy consumption of most parts is concentrated in the lower range. Under a few abnormal conditions, energy consumption increases significantly. Taking the 80th percentile as the single-component energy consumption threshold can cover most normal operating conditions and effectively identify abnormal energy consumption parts, avoiding sudden increases in energy consumption that could lead to equipment overload or energy waste.

[0056] The lower limit threshold for single-piece cycle time is selected from the 20th percentile value of the single-piece cycle time. This is determined comprehensively based on the needs of production efficiency and lower limit control of cycle time. Under stable production conditions, the single-piece cycle time of qualified parts is statistically analyzed. The single-piece cycle time follows a slightly left-skewed distribution, with most samples concentrated around the average value. About 20% of the samples have significantly shorter cycle times. Samples with excessively short single-piece cycle times may have potential quality risks such as excessively fast feeding or insufficient cooling. The 20th percentile value of the single-piece cycle time is used as the lower limit threshold for single-piece cycle time. This represents the high efficiency level that the process can achieve under stable conditions, preventing excessively fast cycle times from causing dimensional deviations and sudden increases in energy consumption. As a dynamic lower limit for production rhythm, the single-piece cycle time lower limit threshold, together with the energy consumption threshold, constrains the process balance in multi-objective coordinated control, achieving stable coordination between production capacity and quality.

[0057] The set of baseline parameters is stored in a read-only database with timestamps and process version numbers, and a version record file is generated synchronously to solidify the set of optimal time shift and optimal phase offset.

[0058] S2. Collect the current measured data of the components, calculate the joint out-of-bounds index based on the benchmark parameter set, determine the quality status of the components, identify abnormal clusters, perform targeted fine-tuning, verify the fine-tuning effect, and generate statistical quadruplets and candidate process parameter sets.

[0059] Using the unique identifier of the component as an index, collect the current measured data of the component, obtain the measured dataset, and calculate the joint boundary violation index by combining it with the set of benchmark parameters.

[0060] When a component with a unique identifier enters the designated inspection or assembly station through the production line, the production line automatically triggers a synchronous data acquisition action. Specifically, the dimensional deviation of the component is collected by online measuring equipment, and each measuring point is stored with the component's unique identifier as an index; the tightening or pressing curve is collected by a torque sensor installed on a tightening gun or pressing equipment; the power meter records the equipment's operating power signal in real time, and the controller automatically calculates the energy consumption of the component; the start and end times of the process are recorded by the PLC, and the cycle time of the component is calculated.

[0061] The current measured data of the component is bound to the component's unique identifier as the primary key, uploaded to the quality data platform, and a measured dataset of the current component is generated.

[0062] The baseline parameter set is retrieved from the database to calculate the dimensional overrun and relative curve deviation of the current component.

[0063] Calculate the dimensional deviation amount, which measures the relative degree of deviation of the current component's dimensional deviation from the median of dimensional deviations in the set of reference parameters.

[0064] Specifically, the dimensional overrun is expressed as: ; in, Indicates the current number The dimensional deviation of a component Indicates the current number Dimensional deviations of individual components.

[0065] Furthermore, the relative deviation value of the curve is calculated using the normalized L2 norm, and the joint boundary violation index is calculated using the relative deviation value of the curve and the dimensional boundary violation amount.

[0066] Specifically, the joint boundary crossing indicator is expressed as: ; in, Indicates the current number The combined out-of-bounds indicators of individual components, Indicates the current number The relative deviation value of the curve of each component.

[0067] Based on the joint boundary violation index, determine the quality status of the components, generate quality status labels, calculate the proportion of out-of-bounds components, and identify abnormal clusters.

[0068] When the combined out-of-bounds index is less than or equal to 1, the overall quality of the component is stable, and the quality status label is stable; when the combined out-of-bounds index is greater than 1, the component is judged as an out-of-bounds part, and the quality status label is out-of-bounds.

[0069] It should be noted that the dividing point 1 between quality stability and deviation is determined statistically by selecting a batch of qualified and stable operating component samples at the beginning of production. Specifically, the dimensional deviation and curve relative deviation values ​​of the qualified and stable operating component samples are standardized to obtain standardized dimensional deviation and standardized curve deviation. The geometric mean of the standardized dimensional deviation and standardized curve deviation of each qualified and stable operating component sample is calculated. The median of the geometric mean is selected as the comprehensive quality benchmark. The distribution of the joint deviation index of the qualified and stable operating component samples is analyzed. The joint deviation index is generally concentrated below 1. For components with abnormal quality, the joint deviation index is significantly greater than 1. Using 1 as the dividing point between stability and deviation can ensure a unified judgment standard under different operating conditions, tolerate process fluctuations, and identify abnormalities in a timely manner.

[0070] To identify potential batch or continuous anomalies, a sliding window is established based on the joint out-of-bounds index and the time sequence of the unique identifier of the parts. Specifically, with the current part as the center, 30 parts before and after it are selected to form the sliding window, and the proportion of out-of-bounds parts within the sliding window is calculated.

[0071] It should be noted that the sliding window size is set to 30 pieces, which is based on the production rhythm of parts and the sensitivity of process changes. This ensures that potential abnormal trends are effectively captured among a sufficient number of parts, balancing real-time performance and data stability. When problems are detected, timely responses can be made without excessive fluctuations.

[0072] When the proportion of out-of-bounds parts exceeds 5%, it is determined that an abnormal cluster has appeared at the current workstation. Further analysis is performed on the equipment information corresponding to the abnormal cluster, including tool number, program number, and fixture number, to generate an abnormal location information package. If multiple workstations have abnormal clusters at the same time, they are sorted according to workstation priority, and the critical processes are handled first.

[0073] It should be noted that the out-of-bounds parts ratio exceeding 5% is derived from statistical analysis under normal production conditions. By observing a large amount of historical data, when the out-of-bounds parts ratio is below 5%, there are usually no batch or systemic problems. When the out-of-bounds parts ratio is above 5%, it indicates that there may be relatively concentrated quality fluctuations and potential batch or continuous abnormal risks. Setting 5% as the out-of-bounds parts ratio threshold can help detect problems in a timely manner and avoid excessive intervention or misjudgment.

[0074] Based on the workstations and process characteristics corresponding to the abnormal clusters, targeted fine-tuning is carried out to obtain the measured data of the fine-tuned parts.

[0075] Specifically, fine-tuning is performed based on the types of anomalous clusters, which include size deviation anomalies, curve deviation anomalies, and energy consumption anomalies.

[0076] Among them, for abnormal dimensional deviation, the tool geometry is adjusted and a compensation value is automatically generated in the radial direction. The compensation value is written into the machine tool parameter area by the tool compensation actuator and takes effect immediately; for abnormal curve deviation, the relative deviation value of the curve of the current part is greater than the curve deviation threshold, the tightening termination angle of the tightening program is automatically reduced or the pressing and holding time is increased at the pressing station; for abnormal energy consumption, the single-piece energy consumption of the current part is greater than the single-piece energy consumption threshold, the spindle speed is reduced while the feed rate ratio remains unchanged to reduce overload energy consumption.

[0077] After the fine-tuning is executed, the measured data of the next 10 parts are collected to verify the fine-tuning effect. The execution time, adjustment parameters and affected workstation information of each adjustment action are written into the fine-tuning record table for traceability and comparison.

[0078] The reason for selecting to collect actual measurement data from the subsequent 10 components is to ensure that the fine-tuning effect can be fully verified in the production process and to avoid the accidental interference of a single component. Selecting 10 components can effectively reflect the long-term impact of fine-tuning on the production process, and data collection and verification can be completed relatively quickly under the production rhythm.

[0079] Based on the measured data of the fine-tuned components, the joint out-of-bounds index is recalculated to determine whether the fine-tuning is effective, the conclusion of the fine-tuning effect is obtained, and a statistical quadruple and candidate process parameter set are generated.

[0080] To ensure stability after fine-tuning, the joint out-of-bounds index of 10 components was continuously monitored after the fine-tuning was executed, and the geometric mean stability index was calculated.

[0081] Specifically, the geometric mean stability index is expressed as: ; in, Indicators of geometric mean stability This indicates the first continuous production after fine-tuning. Sequential index of each component This indicates the first continuous production after fine-tuning. Joint cross-boundary indicators for components and parts.

[0082] When the geometric mean stability index is less than or equal to 0.9, the fine-tuning is effective, and the current process parameter settings are solidified as the new version of the execution control parameter file; when the geometric mean stability index is greater than 0.9, the fine-tuning is insufficient or in the wrong direction, and it automatically reverts to the previous stable execution control parameter file, generates maintenance prompts, and the review process does not require additional manual intervention.

[0083] Specifically, when fine-tuning is effective, the process parameters (including but not limited to spindle speed, feed rate, press-holding time, tightening termination angle, and cooling time) and related environmental factors (including but not limited to temperature, humidity, and energy consumption level) in the new version of the execution control parameter file are recorded as feasible process parameter combinations. Based on the component quality data corresponding to each feasible process parameter combination, the process capability index, defect rate, average energy consumption per piece, and average cycle time per piece for critical dimensions are calculated, and statistical quadruplets corresponding to each feasible process parameter combination are generated. The feasible process parameter combinations and their corresponding statistical quadruplets are stored in a structured manner to generate a candidate process parameter set.

[0084] It should be noted that the geometric mean stability index threshold of 0.9 was determined based on statistical analysis of historical process data and accumulated engineering experience. Retrospective analysis of the effects of multiple fine-tuning experiments showed that when the geometric mean stability index is less than or equal to 0.9, in most cases, fine-tuning has effectively improved process quality, and process stability meets production requirements. Selecting 0.9 as the geometric mean stability index threshold ensures that after fine-tuning, the quality fluctuation of parts during production is small, avoiding excessive fine-tuning that could lead to production instability. Furthermore, extensive data verification has shown that it can effectively balance stability and production efficiency, reducing unnecessary adjustments and ensuring the efficient and stable operation of the process. When the geometric mean stability index is greater than 0.9, it indicates that the fine-tuning effect is insufficient and has not met the expected production quality requirements, necessitating a revert to the previous stable execution control parameter file.

[0085] Within the set scrolling window, the production and inspection data of all components are statistically summarized, the process capability index of critical dimensions is calculated to reflect the stability of the production process, the proportion of out-of-bounds parts is calculated as the defect rate, the arithmetic mean of the energy consumption of each component is calculated as the average energy consumption of each component, and the arithmetic mean of the cycle time of each component is calculated as the average cycle time of each component.

[0086] The process capability index of critical dimensions, the proportion of out-of-bounds parts, the average energy consumption per piece, and the average cycle time per piece are combined into a statistical quadruple. This quadruple is then categorized and summarized according to different equipment, programs, or tool combinations to generate a set of candidate process parameters. Each candidate process parameter corresponds to an independent set of process parameters and statistical characteristics.

[0087] S3. Read and verify the statistical quadruple and candidate process parameter set, obtain the standardized statistical index set, perform hard constraint screening and feasibility judgment, obtain the preferred process parameter set, and determine the optimal control command by lexicographical selection.

[0088] Read and verify the statistical quadruples and candidate process parameter sets, perform robustness preprocessing, obtain the effective candidate process parameter set, and generate a standardized statistical index set.

[0089] Retrieve the statistical quadruple and candidate process parameter set from the previous scrolling window. For each candidate process parameter, check the completeness of the statistical quadruple to ensure that all four statistical indicators have valid values.

[0090] If any indicator in the statistical quadruple is missing, the candidate process parameter will be marked as having incomplete data and will not be included in subsequent calculations.

[0091] To prevent data update delays from causing interference, a consistency check is performed on the timestamps during the verification process, retaining only the statistical data within the current production batch. After verification is completed, a set of verified candidate process parameters is generated.

[0092] From the verified set of candidate process parameters, read the statistical quadruples corresponding to each candidate process parameter and perform a robustness preprocessing. Specifically, calculate the median and absolute deviation of the process capability index of the critical dimension. When the difference between the process capability index of a certain candidate process parameter and the median value exceeds three times the median absolute deviation, the candidate process parameter is removed, and the set of valid candidate process parameters is obtained.

[0093] The median value and distribution range of the four statistical indicators of the effective candidate process parameter set are calculated, and dimensionless processing is performed to unify the relative scale of each statistical indicator.

[0094] Specifically, dimensionless statistical indicators are expressed as: ; in, This represents a dimensionless statistical indicator. This represents the raw value of the statistical indicator. This represents the median value of a statistical indicator. Indicates the distribution range of statistical indicators.

[0095] Each statistical indicator is mapped to [-1, 1]. The process capability index, defect rate, average energy consumption per unit, and average cycle time per unit of the critical dimensions of different process parameters are comparable. The dimensionless quadruple data are organized into a standardized statistical indicator set in a unified format and indexed by process parameter number and timestamp.

[0096] Hard constraint screening is performed on the standardized statistical index set, the feasibility index of each valid candidate process parameter is calculated, and the optimal set of process parameters is obtained.

[0097] The constraint values ​​in the benchmark parameter set are used as the screening benchmark, and a multi-condition hard constraint judgment is performed. Specifically, the hard constraints include quality constraints, defect rate constraints, energy consumption constraints, and cycle time constraints. Among them, the quality constraint is that the process capability index of the critical dimension is greater than or equal to 1.67; the defect rate constraint is that the defect rate is less than or equal to the maximum allowable defect rate; the energy consumption constraint is that the average energy consumption per unit is less than or equal to the energy consumption threshold per unit; and the cycle time constraint is that the average cycle time per unit is less than or equal to 1.1 times the lower limit threshold of the cycle time per unit.

[0098] It should be noted that in the hard constraint screening process, the maximum allowable defect rate is set according to industry standards and the reliability requirements of the production process. Under normal circumstances, in a high-precision production environment, the maximum allowable defect rate is 2%. There are slight fluctuations in the production process, and a certain percentage of defective products are allowed, not exceeding 2% to ensure the overall product quality. If the defect rate exceeds 2%, it will affect the stability of the production line and the product qualification rate.

[0099] The 1.1 times single-piece cycle time lower limit threshold is determined based on the balance between production efficiency and product quality. The single-piece cycle time lower limit threshold is set based on process optimization and the maximum load capacity of the equipment. During the production process, the workload and fault tolerance range of the equipment and workers are considered to give the production line a certain buffer space to deal with occasional process fluctuations or minor equipment downtime and adjustments. Setting the 1.1 times single-piece cycle time lower limit threshold as the upper limit of the average cycle time per piece can ensure that the production line is in a state of efficient and stable operation most of the time, while avoiding the increase in quality risks caused by excessively fast cycle time.

[0100] Construct a feasibility index, denoted as: ; in, This represents the feasibility index. Indicates the index number of the valid candidate process parameter. Indicates the first The process capability index of the critical dimension of the effective candidate process parameters. Indicates the first The failure rate of each valid candidate process parameter Indicates the first Average energy consumption per unit for each valid candidate process parameter Indicates the first The average cycle time per unit for each valid candidate process parameter. This indicates the maximum allowable non-conforming rate. Indicates the energy consumption threshold for a single item. This represents the lower limit threshold for the cycle time of a single item.

[0101] If the feasibility index is greater than or equal to 1, the effective candidate process parameters fully satisfy the hard constraints and are in the fully feasible range. When the feasibility index is between 0.8 and 1, the effective candidate process parameters are in the boundary feasible range and are determined to be boundary feasible process parameters. If the feasibility index is less than 0.8, the effective candidate process parameters are determined to be infeasible and eliminated to obtain the preferred candidate process parameter set.

[0102] The standard of setting 1.0 as the fully feasible range is based on the industry standard of process capability index. In industrial production, when the process capability index reaches 1.67, it means that the production capacity can meet the Six Sigma quality requirements and the non-conformity rate in the production process is very low. A feasibility index of 1.0 indicates that the effective candidate process parameters fully meet all constraints and are fully feasible.

[0103] Setting 0.8 as the starting value for the boundary feasible interval is determined by comprehensively considering the tolerance range of statistical distribution and the redundancy control of the production process. Through statistical analysis of a large amount of historical production data, when the feasibility index of the effective candidate process parameter is between 0.8 and 1, it indicates that the effective candidate process parameter is still within a reasonable range and some minor fluctuations can be accepted. Specifically, setting 0.8 as the starting value for the boundary feasible interval takes into account both the stability and flexibility of the production process, allowing the effective candidate process parameter to operate in the short term. However, it needs to be specially monitored to prevent long-term deviations from the baseline value. If the effective candidate process parameter with a feasibility index between 0.8 and 1 can maintain a high pass rate within a specific production cycle, adjustments are required if deviations continue to occur.

[0104] Furthermore, the statistical indicators of each candidate process parameter in the preferred candidate process parameter set are re-processed to be dimensionless, and the relative scale of each statistical indicator is unified. Indicators with different dimensions are comparable in the same comparison domain. The process capability index, defect rate, average energy consumption per unit, average cycle time per unit, and feasibility index of the key dimensions of each preferred candidate process parameter are recorded in a structured form as preferred process parameter entries to generate the preferred process parameter set.

[0105] Based on the optimized set of process parameters, the process parameters with the best overall performance are selected according to the lexicographical order rule, and the optimal control command is generated.

[0106] Furthermore, based on the preferred set of process parameters, a lexicographical selection process is executed. Specifically, a multi-objective decision-making logic is adopted, prioritizing quality, then cycle time, and finally energy consumption. From the preferred set of process parameters, one or more sets of process parameters Ω1 with the largest process capability index for the critical dimension are selected. If Ω1 includes multiple process parameters, then the set Ω2 with the smallest average cycle time per unit is selected within Ω1. If Ω2 still contains multiple process parameters, then the process parameter with the smallest average energy consumption per unit is selected within Ω2 as the optimal process parameter. The control parameter corresponding to the optimal process parameter is the optimal control command.

[0107] To compare the overall performance of different process parameters and verify the rationality of the lexicographical sorting results, auxiliary selection indexes were calculated.

[0108] Specifically, the auxiliary selection index is expressed as follows: ; in, Indicates the first Auxiliary selection index for candidate process parameters.

[0109] It should be noted that a smaller auxiliary selection index value indicates better overall performance. The auxiliary selection index is only used for ranking reference and does not participate in actual control calculations. The process parameter with the smallest auxiliary selection index is selected as the control command for the production process, which is triggered at the component cycle time boundary to achieve real-time process parameter optimization.

[0110] S4. Receive the optimal control command, arrange and activate it, perform feedforward robustness judgment, execute it on the machine and monitor it online, perform short-window acceptability verification, generate disposal decision, solidify the effective control command, and update the baseline parameter set.

[0111] Based on the optimal control instruction, the optimal control instruction is written into the control instruction activation table at the production cycle boundary moment, the activation timestamp is recorded, and the activated optimal control instruction is generated.

[0112] The optimal control instructions are transmitted to the production line control terminal. These instructions include specific process execution parameters, such as spindle speed, feed rate, press-holding time, and tightening termination angle. In the control terminal, a control instruction activation table is established with timestamp as the primary key to record the loading time and corresponding activation window range of each set of optimal control instructions.

[0113] When the production cycle signal reaches the process switching boundary, the control terminal reads the optimal control instruction at the corresponding time from the control instruction effective table, writes the optimal control instruction into the device register through the standard industrial communication bus, records the effective time of the optimal control instruction, and marks the status of the optimal control instruction as effective.

[0114] Based on the already effective optimal control instructions and combined with real-time environmental data, the environmental robustness coefficient is calculated, the optimal control instructions are adjusted, and the execution control instructions are generated.

[0115] The optimal control command that has taken effect is subjected to feedforward robustness judgment. Specifically, the feedforward robustness judgment is performed in combination with real-time collected environmental data, including the current ambient temperature and current ambient humidity. The deviation between the environmental conditions and the reference environment is quantified by calculating the environmental robustness coefficient.

[0116] The environmental robustness coefficient is expressed as: ; in, Indicates the environmental robustness coefficient. Indicates the current ambient temperature. Indicates the reference ambient temperature. Indicates the maximum permissible temperature deviation. Indicates the current ambient humidity. Indicates the reference ambient humidity. This indicates the maximum permissible humidity deviation.

[0117] It should be noted that, specifically, to obtain the reference ambient temperature, when the production line is in a stable operating phase, the air temperature in the production area is continuously monitored by temperature sensors. The actual measured values ​​of the ambient temperature are recorded over multiple consecutive production batches, and the median value of the temperature change is statistically obtained. This median value of the temperature change is used as the reference ambient temperature. To obtain the maximum permissible temperature deviation, specifically, the range of ambient temperature change is calculated during the stable production phase, and the 95th percentile of the temperature fluctuation amplitude of qualified parts over the corresponding time period is used as the maximum permissible temperature deviation.

[0118] To obtain the reference ambient humidity, specifically, relative humidity data of the air is collected by a humidity sensor within the same time period, and the median value of the humidity measurement during the continuous qualified batch production is used as the reference ambient humidity; to obtain the maximum permissible humidity deviation, specifically, in order to statistically analyze the fluctuation range of humidity measurement data, the 95th percentile value of the fluctuation range of humidity within the normal qualified production range is determined as the maximum permissible humidity deviation.

[0119] If the environmental robustness coefficient is greater than or equal to 0.8, the current environmental conditions are consistent with the reference environment, and the optimal control command can be executed directly; if the environmental robustness coefficient is less than 0.8, slight parameter corrections are performed, such as reducing the spindle speed or extending the press-holding time, to generate robust execution control commands.

[0120] It should be noted that the environmental robustness coefficient threshold of 0.8 was determined based on a large amount of historical production data. By analyzing the trends of component qualification rate and process capability changes under different environmental temperature and humidity fluctuation conditions, when the environmental robustness coefficient is lower than 0.8, the product qualification rate drops significantly and fluctuates violently. 0.8 is used as the dividing point between environmental stability and fluctuation risk to distinguish between control commands that can be directly executed and those that require fine-tuning.

[0121] The execution control command is bound to the unique identifier of the component and written into the equipment control register. After execution on the machine, the measured data of the continuous production window is collected, the short window acceptance index is calculated, and the short window acceptance verification conclusion is obtained.

[0122] Furthermore, when the next cycle begins, the control terminal writes the execution parameter group in the control command into the equipment control register through standard industrial communication. During the equipment execution process, it collects and records the component size deviation, tightening or pressing curve, single-piece energy consumption and actual production cycle in real time, and records them synchronously with the component's unique identifier and stores them in the execution log.

[0123] Real-time monitoring of components within the continuous production window is performed to calculate the joint boundary crossing index and obtain the short window acceptability index.

[0124] Specifically, the short window acceptability index is expressed as: ; in, This indicates the short window acceptability index. This indicates the number of components within the short window.

[0125] When the short window acceptance index is less than or equal to 0.05, it means that the quality fluctuation, energy consumption level and cycle deviation of the components within the short window are all within the allowable range, and the short window acceptance verification conclusion is acceptable; when the short window acceptance index is greater than 0.05, it means that the executed control command deviates from the benchmark under the current conditions, and the short window acceptance verification conclusion is deviation.

[0126] It should be noted that the short window acceptability index threshold of 0.05 was determined based on historical data statistics during the stable production phase. By analyzing the quality fluctuations, energy consumption changes, and cycle deviations of continuous short window samples, when the short window acceptability index exceeds 0.05, the stability of the production process decreases and the defect rate increases. 0.05 is used as the judgment boundary between stable production and deviation from the benchmark.

[0127] Based on the short-window acceptability verification results, the validity of the execution control instructions is determined, a disposal decision is generated, and the disposal decision and the updated execution parameter set are obtained.

[0128] When the short-window acceptability verification conclusion is "acceptable," the execution control command remains unchanged and is marked as a valid control command. When the short-window acceptability verification conclusion is "deviation," a handling decision is automatically generated. Specifically, if the deviation is due to dimensional deviation, tool geometry compensation is performed; if the deviation is due to an abnormal tightening curve, the tightening termination angle is reduced; if the deviation is due to excessive energy consumption, the spindle speed is reduced; if the deviation is due to cycle time extension, the cooling time is ended early. All handling decisions are recorded in the handling log, and the updated execution parameter group is automatically loaded and verified in the next production cycle.

[0129] Based on the decision and the updated set of execution parameters, the execution control instructions are solidified as valid control instructions. The solidified valid control instructions are obtained, and an updated set of baseline parameters is generated.

[0130] When the short-window acceptability verification result is accepted three times in a row, the corresponding execution control instruction is solidified into a new valid control instruction and written into the control parameter library. Based on the three consecutive stable production data, the main statistical parameters in the benchmark parameter set are updated to generate an updated benchmark parameter set.

[0131] It should be noted that if three consecutive decision-making actions are corrections in the same direction, the upper or lower limit of the allowable range of the corresponding baseline parameter will be automatically expanded to adapt to the new production stability range. After the update is completed, the new effective control instructions and the updated baseline parameter set will be used as the default control parameter set for the next production cycle.

[0132] This embodiment also provides a production quality control system for automotive parts, including: The benchmark acquisition module is used to bind unique identifiers of parts, collect multi-source process data, perform time alignment and data cleaning, filter reference sample sets, set a scrolling window, and obtain a set of benchmark parameters.

[0133] The quality stability assessment module is used to collect the current measured data of the components, calculate the joint out-of-bounds index based on the benchmark parameter set, judge the quality status of the components, identify abnormal clusters, perform targeted fine-tuning, review the fine-tuning effect, and generate statistical quadruplets and candidate process parameter sets.

[0134] The process optimization module is used to read and verify the statistical quadruple and candidate process parameter sets, obtain the standardized statistical index set, perform hard constraint screening and feasibility judgment, obtain the optimized process parameter set, and determine the optimal control command by lexicographical selection.

[0135] The control execution module is used to receive optimal control commands, orchestrate their activation, perform feedforward robustness judgments, execute them on the machine and monitor them online, perform short-window acceptability verification, generate disposal decisions, solidify effective control commands, and update the baseline parameter set.

[0136] In summary, this invention achieves dynamic assessment of component quality status and precise location of abnormal areas by calculating joint out-of-bounds indicators and identifying abnormal clusters; it enables rapid correction of production parameters based on real-time multi-source process data by performing targeted fine-tuning and verifying the fine-tuning effect, thereby improving the accuracy and response efficiency of quality stability judgment; and it determines the optimal control command through hard constraint screening and feasibility judgment, enhancing the adaptability of the control strategy and achieving continuous optimization and precise control of the production process.

[0137] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for quality control in the production of automotive parts, characterized in that: include, Bind unique component identifiers, collect multi-source process data, perform time alignment and data cleaning, filter reference sample sets, set scrolling windows, and obtain a set of benchmark parameters; Collect current measured data of components, calculate joint out-of-bounds index based on the benchmark parameter set, determine the quality status of components, identify abnormal clusters, perform targeted fine-tuning, verify the fine-tuning effect, and generate statistical quadruplets and candidate process parameter sets. Read and verify the statistical quadruple and candidate process parameter set, obtain the standardized statistical index set, perform hard constraint screening and feasibility judgment, obtain the preferred process parameter set, and determine the optimal control command by lexicographical selection. Receive the optimal control command, arrange and activate it, perform feedforward robustness judgment, execute it on the machine and monitor it online, perform short-window acceptability verification, generate disposal decision, solidify the effective control command, and update the baseline parameter set; The specific steps for performing targeted fine-tuning, verifying the fine-tuning effect, and generating statistical quadruples and candidate process parameter sets are as follows: Based on the workstation and process characteristics corresponding to the abnormal clusters, targeted fine-tuning is carried out to obtain the measured data of the fine-tuned parts; Based on the measured data of the fine-tuned components, the joint boundary crossing index is recalculated to determine whether the fine-tuning is effective, the fine-tuning effect judgment conclusion is obtained, and a statistical quadruple and candidate process parameter set are generated. The specific steps are as follows: receiving the optimal control command, arranging and activating it, performing feedforward robustness judgment, executing it on the machine and monitoring it online, and performing short-window acceptability verification. Based on the optimal control instruction, the optimal control instruction is written into the control instruction activation table at the production cycle boundary moment, the activation timestamp is recorded, and the activated optimal control instruction is generated. Based on the already effective optimal control instructions, combined with real-time environmental data, the environmental robustness coefficient is calculated, the optimal control instructions are adjusted, and the execution control instructions are generated. The execution control command is bound to the unique identifier of the component and written into the equipment control register. After execution on the machine, the measured data of the continuous production window is collected, the short window acceptance index is calculated, and the short window acceptance verification conclusion is obtained. When the short-window acceptability verification result is accepted three times in a row, the corresponding execution control instruction is solidified into a new valid control instruction and written into the control parameter library. Based on the three consecutive stable production data, the main statistical parameters in the benchmark parameter set are updated to generate an updated benchmark parameter set.

2. The method for quality control in the production of automotive parts as described in claim 1, characterized in that: The process involves binding unique identifiers for components, collecting multi-source process data, performing time alignment and data cleaning, and the specific steps are as follows: Bind the unique identifier of the component, establish a unique traceability master index key, and collect multi-source process data by association according to the unique identifier of the component to obtain the original multi-source process data; Based on a unified time reference, the original multi-source process data is synchronized and cleaned to obtain cleaned and aligned multi-source process data.

3. The method for quality control in the production of automotive parts as described in claim 2, characterized in that: The specific steps for filtering the reference sample set, setting a scrolling window, and obtaining the benchmark parameter set are as follows: Based on the cleaning alignment multi-source process data, samples with continuous production and complete data are selected to obtain a reference sample set; Based on the reference sample set, a rolling statistical window is set, a size benchmark and reference curve are constructed, and a benchmark parameter set is generated.

4. The method for quality control in the production of automotive parts as described in claim 3, characterized in that: The specific steps for collecting current measured data of components, calculating joint out-of-bounds indices based on a set of benchmark parameters, determining the quality status of components, and identifying anomaly clusters are as follows: Using the unique identifier of the component as an index, collect the current measured data of the component, obtain the measured dataset, and calculate the joint boundary violation index by combining it with the set of benchmark parameters; Based on the joint boundary violation index, determine the quality status of the components, generate quality status labels, calculate the proportion of out-of-bounds components, and identify abnormal clusters.

5. The method for quality control in the production of automotive parts as described in claim 4, characterized in that: The steps for reading and verifying the statistical quadruplets and candidate process parameter sets, obtaining a standardized statistical index set, performing hard constraint screening and feasibility determination, and obtaining the optimal process parameter set are as follows: Read and verify the statistical quadruple and candidate process parameter set, perform robustness preprocessing, obtain the effective candidate process parameter set, and generate a standardized statistical index set. Hard constraint screening is performed on the standardized statistical index set, the feasibility index of each valid candidate process parameter is calculated, and the optimal set of process parameters is obtained.

6. The method for quality control in the production of automotive parts as described in claim 5, characterized in that: The process of determining the optimal control command by lexicographical order selection refers to selecting the process parameters with the best overall performance based on the preferred set of process parameters and according to lexicographical order rules, and generating the optimal control command.

7. The method for quality control in the production of automotive parts as described in claim 6, characterized in that: The specific steps for generating disposal decisions, solidifying effective control commands, and updating the baseline parameter set are as follows: Based on the short-window acceptability verification results, the validity of the execution control command is determined, a disposal decision is generated, and the disposal decision and the updated execution parameter set are obtained. Based on the decision and the updated set of execution parameters, the execution control instructions are solidified as valid control instructions. The solidified valid control instructions are obtained, and an updated set of baseline parameters is generated.

8. A production quality control system for automotive parts, based on the production quality control method for automotive parts according to any one of claims 1 to 7, characterized in that: include, The benchmark acquisition module is used to bind unique identifiers of parts, collect multi-source process data, perform time alignment and data cleaning, filter reference sample sets, set a scrolling window, and obtain a set of benchmark parameters. The quality stability assessment module is used to collect the current measured data of the components, calculate the joint out-of-bounds index based on the benchmark parameter set, judge the quality status of the components, identify abnormal clusters, perform targeted fine-tuning, review the fine-tuning effect, and generate statistical quadruplets and candidate process parameter sets. The process optimization module is used to read and verify the statistical quadruple and candidate process parameter set, obtain the standardized statistical index set, perform hard constraint screening and feasibility judgment, obtain the optimized process parameter set, and determine the optimal control command by lexicographical selection. The control execution module is used to receive optimal control commands, orchestrate their activation, perform feedforward robustness judgments, execute them on the machine and monitor them online, perform short-window acceptability verification, generate disposal decisions, solidify effective control commands, and update the baseline parameter set.

Images