Quality data evaluation method for multi-variety small-batch manufacturing process

Abstract

The embodiment of the invention discloses a quality data evaluation method for a multi-variety small-batch manufacturing process, and the method comprises the steps: 1, employing a tolerance coefficient method as a data conversion algorithm, and obtaining a tolerance utilization rate of a single monitoring characteristic; step 2, performing process capability evaluation on each monitoring characteristic, combining standardized data of multiple monitoring characteristics for single monitoring characteristics capable of accumulating the total sample number N and single monitoring characteristics incapable of accumulating the total sample number N so as to perform control process capability evaluation, and performing process control in an auxiliary single-value monitoring mode; and step 3, evaluating the control process capability in the step 2, and formulating a quality state evaluation criterion of multi-variety small-batch process control. According to the technical scheme provided by the embodiment of the invention, the problems of low accuracy of quality evaluation, difficulty in establishing an effective control model and control limit and the like of a quality control mode of multi-variety and small-batch products are solved.

Description

Technical Field

[0001] This invention relates to, but is not limited to, the fields of process monitoring and data processing evaluation, and particularly to a method for quality data evaluation in multi-variety, small-batch manufacturing processes. Background Technology

[0002] For discrete manufacturing industries with multiple varieties and small batches, it is necessary to evaluate and control product quality during the manufacturing process through process monitoring.

[0003] Existing product quality assessment methods include: CN103760814B, "A Method for Determining the Process Capability Index of Multi-variety Small-batch Production Parts Based on Features," published by Nanjing University of Aeronautics and Astronautics in 2016. This method involves data normalization, calculating the standard deviation and tolerance of quality characteristics for different nominal dimensions, and then normalizing and merging the sample by determining whether the ratio is a constant. CN114169704B, "Product Manufacturing Process Quality Control Method, Device, Equipment, and Storage Medium," published by Chengdu Aircraft Industry Group Co., Ltd. in 2024, discloses a product manufacturing process quality control method, mainly through the analysis of sample numbers... According to the standard normal transformation, based on the characteristic data labeling collection and control chart monitoring and early warning of product production process quality control, the focus is on the design and development of production process quality control devices; CN112465377B "A method for key process identification and cluster analysis for multi-variety small-batch manufacturing process" released by Shenyang University of Technology in 2024 adopts a key process identification method for multi-variety small-batch manufacturing process based on clear set and grey relational analysis, constructs a key process identification model for multi-variety small-batch manufacturing process, and on this basis, performs cluster analysis on the key processes of each variety based on hierarchical cluster analysis method, determines the resolution selection scheme, and expands the sample data volume.

[0004] However, in the process of quality control of multiple varieties and small batches, due to the small sample size and discrete data, the existing methods mentioned above generally have problems such as low accuracy of quality assessment and difficulty in establishing effective control models and control limits. Summary of the Invention

[0005] The purpose of this invention is to solve the above-mentioned technical problems. This invention provides a method for quality data evaluation in a multi-variety, small-batch manufacturing process, which addresses the problems of low accuracy in quality evaluation and difficulty in establishing effective control models and control limits due to the small sample size and discrete data in quality control methods for multi-variety, small-batch products.

[0006] The technical solution of the present invention: The embodiments of the present invention provide a method for quality data evaluation in a multi-variety, small-batch manufacturing process, comprising: Step 1, data standardization processing, including: using the tolerance coefficient method as the data transformation algorithm to obtain the tolerance utilization rate of individual monitoring characteristics; Step 2 involves evaluating the process capability of each monitoring characteristic, including: dividing the control process of multiple varieties and small batches into various scenarios according to the total number of samples N; for individual monitoring characteristics that can accumulate a total number of samples N, using conventional process capability evaluation methods by accumulating the total number of samples N, or merging standardized data from multiple monitoring characteristics to form a data volume that meets the requirements for implementing control process capability evaluation; for individual monitoring characteristics that cannot accumulate a total number of samples N, merging standardized data from multiple monitoring characteristics to implement control process capability evaluation, and using single-value monitoring as an auxiliary method for process control. Step 3: Evaluate the control process capability from Step 2, and formulate quality status assessment criteria for multi-variety, small-batch process control, including: If the total sample size N is too small to calculate the capability of this single control characteristic, and the capability obtained by merging the standardized data of each control characteristic increases the capability of this single control characteristic, then single-value monitoring or control charts should be used to assist in the control of this single control characteristic.

[0007] The beneficial effects of this invention are as follows: This invention provides a method for quality data assessment in multi-variety, small-batch manufacturing processes. Firstly, it employs the tolerance coefficient method as a data conversion algorithm, monitoring the tolerance utilization rate of a single measurement value as a single value, and converting the upper and lower limits of a single monitoring characteristic to ±1, thereby eliminating differences in units, dimensions, and tolerances between different datasets. Secondly, in evaluating the process capability of various monitoring characteristics, process control is categorized into multiple scenarios based on the total sample size N. Corresponding process control methods are adopted for different scenarios. For single monitoring characteristics that can be accumulated and have a small total sample size N, as well as single monitoring characteristics that cannot be accumulated and have a large total sample size N, process capability assessment can be implemented by merging standardized data from multiple monitoring characteristics. Thirdly, by formulating quality status assessment criteria for multi-variety, small-batch process control, for situations where the total sample size N is very small and the capability calculated by merging standardized data of various control characteristics increases the capability of a certain control characteristic, single-value monitoring or control charts are used to assist in the control characteristic, and specific judgment criteria for single-value monitoring or control charts are proposed. The technical solution provided by this invention has the following beneficial effects: First, an innovative data conversion algorithm based on the tolerance coefficient method is proposed, which can convert data of different distribution types into tolerance utilization rate through standardization, eliminate differences in units, dimensions and tolerances between data, and solve the complexity and limitations of non-normally distributed data conversion; Second, the technical solution provided by this invention combines international and industry standards, takes into account a variety of factors, and formulates a discrepancy criterion that conforms to multi-variety small-batch data samples. It proposes targeted discrepancy criteria for non-normal data and complex production scenarios, thereby improving the accuracy and reliability of discrepancy detection. Third, it innovatively proposes a single-value control chart monitoring method based on uncertainty, which takes into account various potential sources of measurement variation, including the accuracy of the measuring equipment, environmental factors, and differences in operators. This makes the study of measurement uncertainty more detailed than the analysis of the MSA measurement system, which helps to discover subtle problems that are overlooked by the conventional analysis of MSA and enhances the reliability and credibility of the measurement results. Fourth, in the proposed application of control charts for analysis, the maximum number of points on the control chart that are allowed to exceed the control limits is based on the number of binomial distribution subgroups. Attached Figure Description

[0008] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of the present invention and do not constitute a limitation on the technical solutions of the present invention.

[0009] Figure 1 This is a flowchart illustrating a method for evaluating quality data in a multi-variety, small-batch manufacturing process, as provided in an embodiment of the present invention. Figure 2 This is a correlation curve between sample size and control chart sensitivity in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the calculation of capacity based on measurement data for four length characteristics A, B, C, and D in an embodiment of the present invention. Figure 4 This is a schematic diagram showing the standardization process of individual data in features A, B, C, and D in an embodiment of the present invention before and after the standardization process. Figure 4 Figure a in the diagram shows the data before transformation, and figure b shows the data after standardization. Figure 5 This is a schematic diagram illustrating the calculation of capabilities using standardized data for the four length characteristics A, B, C, and D in an embodiment of the present invention. Figure 6 This is a histogram comparison of the data distribution before and after the transformation of monitoring feature C in an embodiment of the present invention; Figure 6 Figure a in the diagram is the histogram of data distribution before the transformation, and Figure b is the histogram of data distribution after the transformation. Figure 7 This describes the calculation capabilities and distribution results of the merged data after transformation of characteristics A, B, C, and D in an embodiment of the present invention. Figure 8This is a schematic diagram comparing the distribution results of the capability indices (i.e., the distribution results) of the four length characteristics A, B, C, and D before and after the transformation in an embodiment of the present invention. Figure 9 This is a schematic diagram of the control chart types used in different time periods in an embodiment of the present invention; Figure 10 This invention provides a schematic diagram of seven consecutive points on one side of the control limit center in an embodiment of the invention; where diagram a represents n. Run Schematic diagram of probability calculation principle, Figure b is a schematic diagram of 7 consecutive points on one side of the control limit center; Figure 11 This is a schematic diagram of a continuous upward or downward trend of 7 points in an embodiment of the present invention; Figure a shows an example of the effect of 7 consecutive upward points, and Figure b is a control chart. Figure 12 This is a schematic diagram illustrating the application of the pre-control chart in an embodiment of the present invention; Figure 13 This is a control chart for monitoring the process of 100% full inspection of measurement uncertainty in an embodiment of the present invention; Figure 14 This is a schematic diagram of a case study using pre-control chart analysis in the application verification of this invention; Figure 15 This is a schematic diagram of an analysis case based on uncertainty. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

[0011] As explained in the background section, in the process of quality control of multiple varieties and small batches, due to the small sample size and discrete data, the accuracy of quality assessment is generally low, and it is difficult to establish effective control models and control limits.

[0012] To address the aforementioned problems, embodiments of the present invention provide a method for quality data assessment in multi-variety, small-batch manufacturing processes, comprising: data standardization processing, process capability evaluation, process monitoring, establishment of discrepancy criteria, and application of process control charts. This method is particularly suitable for quality assessment and control in multi-variety, small-batch discrete manufacturing industries. The quality data assessment method proposed in this invention effectively solves the problems of low accuracy of traditional statistical methods and difficulty in establishing effective control models and control limits in multi-variety, small-batch quality data assessment due to small and discrete sample sizes.

[0013] The present invention provides the following specific embodiments, which can be combined with each other. For the same or similar concepts or processes, they may not be described again in some embodiments.

[0014] Figure 1 This is a flowchart illustrating a method for quality data evaluation in a multi-variety, small-batch manufacturing process, as provided in an embodiment of the present invention. Figure 1 As shown in the embodiment of the present invention, the method for quality data evaluation in a multi-variety, small-batch manufacturing process includes the following steps: Step 1, data standardization processing, includes: using the tolerance coefficient method as the data transformation algorithm to obtain the tolerance utilization rate of individual monitoring characteristics.

[0015] In this step, the tolerance coefficient method is used as the data transformation algorithm. First, similar processing procedures are grouped into "typical procedures". The range and proportion of a single measurement value relative to the tolerance (i.e., the tolerance utilization rate of a single measurement value) are monitored as a single value. The single monitoring characteristic is added to the upper and lower limits of the tolerance and converted to ±1 to eliminate the differences in units, dimensions and tolerances between different datasets. That is, it is no longer necessary to consider the original data model, which improves the comparability between features.

[0016] It should be noted that in the embodiments of the present invention, the monitoring object of SPC is a single monitoring feature, and each sample data in each single monitoring feature is a single value collected for that monitoring feature.

[0017] Step 2 involves evaluating the process capability of each monitoring characteristic, including: dividing the control process of multiple varieties and small batches into various scenarios according to the total number of samples N; for individual monitoring characteristics that can accumulate a total number of samples N, using conventional process capability evaluation methods by accumulating the total number of samples N, or merging standardized data from multiple monitoring characteristics to form a data volume that meets the requirements for implementing control process capability evaluation; for individual monitoring characteristics that cannot accumulate a total number of samples N, merging standardized data from multiple monitoring characteristics to implement control process capability evaluation, and using single-value monitoring as an auxiliary method for process control. In this step, for example, the control process of multiple varieties in small batches is categorized into three scenarios according to the total number of samples N; For scenarios with a total sample size N≥125, the conventional process capability evaluation method is adopted; For scenarios where 30≤N<125, for monitoring characteristics with a total number of accumulative samples N, conventional process capability evaluation methods can be adopted after accumulating the total number of samples N to N≥125. Alternatively, data from multiple monitoring characteristics with similar features can be standardized and merged to form a data volume that meets the requirements for implementing control process capability evaluation. For a single monitoring characteristic with a total sample size N that cannot be accumulated, since the total sample size N < 30, the data after standardization of each monitoring characteristic is used to evaluate the control process capability by merging the data, and process control is carried out by assisting single-value monitoring.

[0018] Step 3: Evaluate the control process capability from Step 2, and formulate quality status assessment criteria for multi-variety, small-batch process control, including: If the total sample size N is too small to calculate the capability of this single control characteristic, and the capability obtained by merging the standardized data of each control characteristic increases the capability of this single control characteristic, then single-value monitoring or control charts should be used to assist in the control of this single control characteristic.

[0019] In one implementation of this invention, step 2 above, which involves merging standardized data from multiple monitoring characteristics to evaluate the control process capability, includes: Step 21: Use measurement data of individual monitoring characteristics to perform capability calculation and capability evaluation; Step 22: Use the tolerance utilization rate obtained from the conversion of each measurement data in the single monitoring characteristic to perform capability calculation and capability evaluation; Step 23: After merging the standardized data of each individual monitoring characteristic, perform capability calculation and capability evaluation on all the merged standardized data.

[0020] Based on the comparison and analysis of the combined capability and the standardized data capability of each individual control feature in step 23 above, if the capability of one control feature is less than the combined capability, the combined capability will be increased by the other control features, which will increase the capability of the individual control feature that is not high to begin with. In this case, it is necessary to combine the warning limit based on the individual value or apply an appropriate control chart for monitoring and evaluation.

[0021] Accordingly, step 3 employs single-value monitoring or control charts for auxiliary control of individual control characteristics, including: based on the capability calculations of each control characteristic in step 2, setting three levels of quality status assessment during the multi-variety, small-batch control process: Level 0: No control chart is checked. For levels where the capability calculated after merging standardized data reaches or exceeds the qualified indicators, periodic capability evaluation is adopted, and process control is assisted by single-value monitoring. Level 1: On-site real-time monitoring and control charts are used to assist in process control. For scenarios with a sample size N < 30, anomaly criteria are established for the control process of multiple varieties and small batches based on the allowable probability of anomaly occurrence. Level 2: Use analytical control charts to analyze the performance of the previous monitoring process.

[0022] The discrepancy criteria established in Level 1 above for the multi-variety, small-batch control process include: Guideline 1: Infringement of control boundaries is not permitted; Criterion 2: No nine consecutive points fall on one side of the center of the control chart; Criterion 3: There are no six consecutive points showing an upward or downward trend.

[0023] The following is an illustrative example illustrating the implementation of the quality data evaluation method for multi-variety, small-batch manufacturing processes provided by the present invention.

[0024] Implementation Example: This implementation example uses the quality data evaluation of a rotary automated processing unit as an example. The following are several solutions to illustrate this implementation example: (1) Purpose and scope of this implementation example In this implementation example, based on an automated machining production line for a rotary actuator of an airborne product, the quality data evaluation method provided by this invention is adopted. In this scenario, the quality status data is automatically collected, and the data collection is real-time and highly accurate. Moreover, the data samples in this scenario meet the multi-scenario application needs of this invention for insufficient samples in the discrete manufacturing industry of aerospace.

[0025] (2) Implementation process: The following is a detailed description of the implementation process of quality evaluation and quality control using the quality data evaluation method provided by this invention in the automated machining scenario of a certain rotary actuator of an airborne product.

[0026] Step 1: Define and clarify the application scenarios of the quality data assessment method provided by this invention. The data required for implementing process capability and control (SPC) in mass production needs to cover 20 production days. The total number of parts obtained through sampling is at least 125. Unlike traditional mass production, the total amount of quality data generated by multi-variety, small-batch quality data cannot meet the basic requirement of 125 samples. The biggest challenge is the insufficient amount of data. The core of its evaluation is to collect and analyze data from different products in order to monitor the quality status of the production process in real time. Through summary statistical analysis, the statistical indicators of each product are determined, thereby understanding the quality level and stability of the product.

[0027] Step 2: Standardization and transformation of quality data The biggest challenge in multi-variety, small-batch SPC applications is the lack of data, mainly reflected in the insufficient sample size n for each sampling and the insufficient total number of samples N; where N represents the total number of samples used for evaluation, which can be accumulated over a period of time.

[0028] In a multi-variety, small-batch production environment, the limited production quantity of each product significantly restricts the sample size (n) for each sampling. For example, a product might only be produced in batches of 20-30 units, meaning the sample size (n) for each sampling might be much smaller than n=5 in large-batch production. This reduced sample size directly impacts the accuracy and reliability of SPC analysis; a smaller sample size can lead to unstable statistical results and increase the risk of misjudgment. Figure 2 The figure shown is a correlation curve between sample size and control chart sensitivity in an embodiment of the present invention. Figure 2 In a control chart, the horizontal axis represents the offset of the process center, and the vertical axis represents the probability that the control chart can detect changes in the process, i.e., the detection rate. This allows the creation of a control chart efficacy curve. Figure 1 The correlation between sample size and control chart sensitivity can be clearly seen; at the same detection rate, the smaller the sample size n, the larger the offset, and the higher the probability of missed detection.

[0029] Traditional SPC requires accumulating at least 25 sets of samples, with each set typically having a sample size n of 5, resulting in N = 25 * 5. This generates a total of no less than 125 measurements, and the resulting power curves can usually convey information about the process performance characteristics.

[0030] To address the two issues of insufficient data volume mentioned above, the evaluation method proposed in this embodiment of the invention determines the sample size n and the total sample data N in the following manner: Sample size n: This implementation example follows Shewhart's reasonable subgrouping principle. The control charts in this example are based on this principle, meaning that for n values ​​between 3 and 7, if a sample size of n=3 is selected, it cannot guarantee that intra-group fluctuations are solely due to chance, while inter-group fluctuations are mainly caused by abnormal factors. Therefore, in this implementation example, only one item is actually drawn at a time, i.e., n=1. A moving subgroup is introduced, where a new measurement is added to a subgroup, and the first measurement in the previous subgroup is deleted, resulting in a new subgroup with a sample size of n=3. For example, if the current sample is i, and the current subgroups are i-1, i, and i+1, then the next sample's subgroups will be i, i+1, and i+2.

[0031] Total sample size N: For individual monitoring characteristics, combine historical measurement data or new measurement data from continued production and use conventional control charts for control; for monitoring characteristics without historical or new measurement data, use the tolerance coefficient method to convert each measurement value into the corresponding tolerance utilization rate to eliminate differences in units, dimensions and tolerances between different datasets and simplify data interpretation and analysis.

[0032] Step 3, Capability Analysis In this implementation example, by analyzing the process method, four similar length characteristics (e.g., the length characteristics of four types of products) of four types of parts with similar processing forms in the production line are selected and labeled as A, B, C, and D, which are the four monitoring characteristics.

[0033] 3.1 Perform capability calculation and capability evaluation for individual monitoring characteristics; By identifying factors affecting process quality during control, a process similarity analysis model is constructed. Similar processes are grouped with the original process to form "typical processes." The tolerance coefficient method is used to convert measurement data into standardized data, i.e., tolerance utilization rate. Hypothesis testing and other data verification methods are then applied to the converted standardized data to determine if the sample conforms to statistical sample patterns, thereby verifying the sample's correctness. For example... Figure 3 The diagram shown illustrates how the capacity of four length characteristics (A, B, C, and D) is calculated using measurement data in an embodiment of the present invention. Figure 3 The diagram illustrates the capabilities and distribution of four characteristics, A, B, C, and D, before transformation.

[0034] Figure 3 In S, T, and S represent the mean, standard deviation, and tolerance of the corresponding individual monitoring characteristic, respectively. Columns 2 to 5 are calculated from the individual values ​​of each monitoring characteristic. Potential Capability Index P p The values ​​were distributed in the range [1.44, 1.98], with an inter-individual variation of 37.5%; the key ability index P... pk The values ​​are distributed in the interval [1.39, 1.71], with an inter-individual variation of 23%. Among them, models A, B, and C are judged to be mixed distributions and log-normal distributions, and are basically approximating normal distributions. Characteristic D has a large number of data points at lower values; if this influence is eliminated, the model will also be a symmetrical unimodal distribution. Models A, B, and C show little difference, while model D has two peaks. The overall evaluation conclusion is: using an ability index of 1.33 as the evaluation standard, green indicates passing the evaluation, meaning all ability indices meet the requirements.

[0035] 3.2 Perform capability calculation and capability evaluation on the tolerance utilization rate obtained from the conversion of individual monitoring characteristics; The above similar characteristics are converted using the tolerance utilization method. The conversion method is as follows: ; in: A i Each piece of data representing characteristic A, T m It is the tolerance center. USL , LSL These represent the upper and lower tolerance limits, respectively. A iTr express A i The standardized data after conversion; the conversion methods for characteristics B, C, and D are the same as described above. For example... Figure 4 The diagram shown illustrates the standardization process of individual data points in features A, B, C, and D in an embodiment of the present invention. Figure 5 The diagram shown illustrates how the four length characteristics A, B, C, and D are used to calculate capabilities based on standardized data after conversion, in an embodiment of the present invention. Figure 5 The diagram illustrates the capabilities and distribution results of the four characteristics A, B, C, and D after transformation.

[0036] contrast Figure 3 and Figure 5 By comparing the data before and after the conversion, it can be seen that the capability index after the conversion... P p , P pk Almost no change (only characteristics A, B, and C show a negligible difference of 0.01 due to the issue of decimal place retention), and the distribution of characteristics A, B, C, and D changes only in that characteristic C changes from a "log-normal distribution" to a "normal distribution". For example... Figure 6 The image shows a histogram comparison of the data distribution before and after the transformation of monitoring characteristic C in an embodiment of the present invention. The transformation of other characteristics has little impact on the data distribution.

[0037] 3.3 After merging individual monitoring characteristics, perform capability calculation and capability evaluation; The standardized data after transformation of data from various monitoring characteristics are merged, that is, the data after transformation of characteristics A, B, C, and D are merged, and then a comprehensive summary analysis is performed, such as... Figure 7 The figure shows the calculation capability and distribution results of the merged data after transformation of characteristics A, B, C, and D in an embodiment of the present invention.

[0038] In comparison with the above Figure 3 , Figure 5 and Figure 7 The following conclusions can be drawn: (1) The capability index calculated by combining the length characteristics A, B, C, and D after standardization and transformation. P p , P pk All are less than the original capability indices of characteristics A, B, and C, and greater than the original capability index of characteristic D (regardless of whether before or after the conversion, because data conversion has almost no impact on capability values). (2) After the length characteristics A, B, C and D are standardized and converted, the histogram and distribution obtained by merging are very similar to those of characteristics A, B and C. However, characteristic D is different from the merged histogram because there is an abnormal increase in the value of a certain point.

[0039] It should be noted that the capability index calculated by combining characteristics A through D... P p , P pk If all values ​​are greater than (or less than) their original capability indices, control charts are not needed; the combined capability indices can be used. P p , P pk Directly represents the capability of characteristics A through D. For example... Figure 8 The diagram shown is a comparison of the capability index, i.e., the distribution results, before and after the transformation of the four length characteristics A, B, C, and D in an embodiment of the present invention.

[0040] Verification revealed that the calculation of the capability index after standardization of individual characteristics A, B, C, and D is effective. Figure 8 As shown, after merging the individual monitoring characteristics, the following situations arise: If there are individual monitoring characteristics with particularly small data volumes, such as a total sample size N ≤ 10, capability cannot be calculated. If a single value is near the upper or lower specification limits, after standardization, if other individual control characteristics perform well, it will inadvertently inflate the capability of the already weak individual characteristic. Therefore, in this case, it is necessary to combine warning limits based on individual values ​​or apply appropriate control charts for monitoring and evaluation. That is, for characteristic D above, auxiliary process control is also required through control charts.

[0041] Step four involves using a multi-variety, small-batch control chart to assist in monitoring the control process. This step is implemented through the following process: Control chart anomaly detection methods need to be tailored to specific application requirements, distinguishing between analytical control charts and control charts for general purposes. Figure 9 The diagram shown illustrates the control chart types used in different time periods in an embodiment of the present invention. This embodiment, tailored to the practical application scenario of multi-variety, small-batch manufacturing, sets the stability anomaly detection at three levels: (1) Level 0, no control chart check; (2) Level 1, control charts are used for real-time on-site monitoring; (3) Level 2, control charts are used to analyze the performance of previous monitoring processes; in the quality status assessment of multi-variety small-batch control processes, the number of times the control limits can be violated is determined based on the binomial distribution theory and the amount of data.

[0042] The monitoring schemes for the above levels are explained as follows: Level 0: No control charts are used for monitoring, and no feedback is provided. For capabilities calculated after merging standardized data that meet or exceed the qualified capability index (e.g., capability requirement 1.33, actual achievement 1.67), periodic capability evaluation is adopted, and process control is assisted by single-value monitoring.

[0043] Level 1, control chart used for real-time on-site monitoring; For scenarios with a sample size N < 30, based on the allowable probability of anomalies, anomaly detection criteria are established for multi-variety, small-batch control processes, including: Guideline 1: Infringement of control boundaries is not permitted; Criterion 2: No nine consecutive points fall on one side of the center of the control chart; Criterion 3: There are no six consecutive points showing an upward or downward trend.

[0044] The above criteria in Level 1 are determined as follows: The principle of control chart control is that a low-probability event will not occur in a single trial, but when there are enough trials, i.e., enough points on the quality control chart, an inevitable event will occur that exceeds the control limit. This takes into account the influence of manufacturing elements such as manpower, machinery, materials, methods, environment, and measurement.

[0045] (a) Set the permissible probability for quality control of multiple varieties and small batches, based on the premise that the control limits are not allowed to be violated.

[0046] (b) Increase the number of consecutive points from 7 to 9 on one side of the control limit center; An n Run If a chain with 7 or more points appears above or below the center line, it indicates that the mean of the process has shifted due to special reasons; n Run The probability that all 7 values ​​appear on the center line at once is extremely small. The probability is calculated as follows: We can obtain n Run = 0.78%; like Figure 10 As shown, this is a schematic diagram of seven consecutive points on one side of the control limit center in an embodiment of the present invention; where diagram a represents n. Run The probability calculation principle diagram, Figure b is a schematic diagram of 7 consecutive points on one side of the control limit center.

[0047] In multi-variety, small-batch manufacturing processes, it is normal for the control center to be too large or too small for a certain stage or batch of production. This is because adjusting the personnel, machinery, materials, methods, environment, and measurement is very responsible and costly, and there are many uncertainties. Therefore, without affecting the ability to meet the standards, the control center can be relaxed to 9 points.

[0048] (c) Tighten the threshold of 7 consecutive points for a rise or fall to 6 points.

[0049] A chain of 7 points showing a continuous increase or decrease in value is called an upward or downward trend. For seven values, there are a total of n! = 7! = 1*2*3*4*5*6*7 = 5040 possible permutations. Of these 5040 possible permutations, only one sequence shows a continuous increase in value. The probability of an upward trend is calculated as follows: ; like Figure 11 The diagram illustrates a continuous upward or downward trend of seven points in an embodiment of the present invention; Figure a shows an example of the effect of seven consecutive upward points, and Figure b is a control chart. Since the probability of such anomalies occurring is relatively small, but when they are detected, a single value may already have a significant long-term trend, especially for control charts of multiple varieties and small batches, stricter measures are implemented in practical applications.

[0050] Level 2, used for analyzing control charts from previous monitoring processes. For analysis control charts, data points may exceed the upper and lower control limits, but these limits are restricted using the following algorithm. In the auxiliary control process of multi-variety, small-batch analysis control charts, calculations are performed based on the number of data points in the control chart and the binomial distribution theory, considering confidence intervals: the maximum allowed limits for points on the control chart to exceed the upper and lower control limits, respectively. The calculation method is as follows: ; g(k) represents the probability that exactly k samples exceed the control limit; k is the number of samples in the control chart that exceed the control limit; p represents the proportion of the population that exceeds the control limit; Assume p is a known quantity; n is the sample size; p represents the probability of a (binary) certain event X occurring; for example, "exceeding the control limit".

[0051] The following describes the application of the control chart for real-time on-site monitoring in Level 1 of the present invention.

[0052] When implementing the multi-variety, small-batch process control in this invention, data judgment logic is formed according to sampling inspection n=2 / n=3 and 100% inspection n=1.

[0053] In the sampling inspection n=2 / n=3 mode, the control chart selection type is set in combination with the two scenarios of N≥30 and N<30, and the maximum number of control points allowed to exceed the control limit is calculated based on the binomial distribution.

[0054] In the 100% verification n=1 mode, a single-value control chart based on measurement uncertainty is set up. Various potential sources of measurement variation are comprehensively considered. By quantifying and evaluating the measurement uncertainty, the confidence interval of the measured value is fed back. The process monitoring of the single value is extended on the basis of the existing control chart.

[0055] In one alternative implementation, the control chart for real-time on-site monitoring in this embodiment of the invention can be based on traditional control charts, such as Shewhart control charts, Pearson control charts, and extended Shewhart control charts.

[0056] In another alternative implementation, the control chart for real-time on-site monitoring in this embodiment of the invention can be the following control chart: Method 1: Establish a pre-control chart - rainbow chart Scenario: An algorithm for controlling the process based on specification limits assumes that the characteristics of the products in the production process are measurable and adjustable, without assuming a certain distribution or stable distribution. The data is partitioned in advance, and small batches of parts are judged to fall into different tolerance areas, thereby enabling real-time monitoring and early warning.

[0057] Taking the raw measurement data (or standardized data) of monitoring characteristics A, B, C, and D as an example, such as Figure 12 The diagram shown is a schematic representation of the application of the pre-control chart in an embodiment of the present invention.

[0058] Because the general process capability requirement is C p and C pk ≥1.33; if the minimum requirement is met, C p and C pk =1.33, which means the data is normally distributed and symmetrical about the tolerance center. Therefore, we can deduce: ; C p C is the potential ability index. pk This refers to the key competency index; it should be noted that C here... p and C pk P corresponding to the above p and P pk The physical meaning is consistent, the region is P p and P pk This indicates short-term ability, taking into account both between-group and within-group variation, i.e., all unstable variations in the process; C p and C pk This indicates long-term capability, which is considered as a stable process. Therefore, the calculation of within-group variation is reduced, and the focus is on considering between-group variation.

[0059] Therefore, we find that 6s (process discreteness) accounts for 75% of T (tolerance). So, the warning line for single-value monitoring is generally set at 75%-80% (yellow) of the tolerance center. Based on the above analysis, upper and lower warning limits are set for internal control. If the upper or lower warning limits are exceeded, pre-control begins.

[0060] Method 2: Establish a single-value monitoring chart based on measurement uncertainty Scenario: When implementing 100% full inspection, this invention proposes a method to incorporate measurement uncertainty into process monitoring. As shown in the pre-control chart of Method 1, if the process capability requirement is 1.33, the resulting process variation requirement must not exceed 75% of the tolerance width. This is because it considers the measurement system's Cg / Cgk and %GRR. Since %GRR cannot be eliminated, a percentage is set as an acceptance limit. Based on this limit, there will be ambiguity near the boundary. If the process random discrete range requirement is set to 100% of the tolerance width, it will obviously bring very high risks. Based on this logic, this invention proposes to incorporate measurement uncertainty into product and process monitoring.

[0061] Based on the results of expanded uncertainty, applying them to the monitoring process for individual values, that is, after taking into account the variation or uncertainty Uc of the measurement system itself (95% confidence level or higher), the monitoring of characteristics enters a new level and a new height.

[0062] Taking the raw data of monitoring characteristics A, B, C, and D as an example, consider the application of process monitoring for full inspection with measurement uncertainty, such as... Figure 13 As shown, Figure 13 This is a control chart for monitoring the process of 100% full inspection of measurement uncertainty in an embodiment of the present invention.

[0063] The greatest advantage of this method lies in the verification of the measurement system and process. Regardless of the process and characteristics being evaluated, it is essential to first verify whether the measurement uncertainty is compatible with the predefined maximum permissible variation capability target. This is a necessary step before process verification and monitoring. Simultaneously, the variation identified in the earlier verification of the measurement process is organically integrated into product monitoring. This means that the two quality tools, MSA and SPC, are no longer independent verifications in stages, but are instead integrated into the final product.

[0064] (III) Application Validation The following specific examples demonstrate the application of the quality data evaluation method for multi-variety, small-batch manufacturing processes provided by the embodiments of the present invention in the production process: This evaluation standard has been applied to over 2000 characteristics of 26 parts, including XXD9AZ-20-01-01, XXD8AZ-30-01-01, and XXD10BZ-20-01-01, in the applicant company's automated machining unit for gear blanks. A sample of 35 unstable characteristics of three types of parts was analyzed. Through process monitoring and fluctuation early warning, and subsequent improvements, the process capability index increased by 47.51%. For example... Figure 14 , 15 As shown in Tables 1, 2, and 3, Figure 14 This is a schematic diagram of a case study using pre-control chart analysis in the application verification of this invention. Figure 15 This is a schematic diagram of an analysis case based on uncertainty.

[0065] Table 1 Comparison of Capability Data for XXD9AZ-20-01-01

[0066] Table 2 Comparison of Capability Data for XXD8AZ-30-01-01

[0067] Table 3 Comparison of Capability Data for XXD10BZ-20-01-01

[0068] (iv) Propose corresponding countermeasures for potential risks. 4.1 Risk Analysis a. Risk of Changes in Production Conditions: Over time, production conditions may change, such as equipment upgrades, process improvements, and changes in raw material suppliers. These changes may affect the distribution and quality characteristics of the data, leading to situations where this invention is not suitable for application.

[0069] b. Challenges from New Technologies and Methods: New technologies and methods are constantly emerging in the field of quality control, and these new technologies may challenge the advancement and effectiveness of the findings of this study. Failure to keep abreast of and apply these new technologies may put us at a disadvantage in market competition.

[0070] 4.2 Countermeasures a. Continuous Monitoring and Optimization: In applying the quality data evaluation method for multi-variety, small-batch manufacturing processes provided by this invention, a continuous internal management system can be established to regularly analyze and evaluate production data, promptly identifying the impact of changes in production conditions on quality control. Based on the monitoring results, data transformation algorithms, outlier detection criteria, and control charts can be optimized and adjusted to ensure they always adapt to changes in production.

[0071] b. Strengthen technological research and development and innovation: Pay close attention to new technologies and methods in the field of quality control, actively carry out technological innovation, and improve our own quality control level by continuously introducing new technologies and methods to maintain a competitive advantage in the market.

[0072] c. Strengthen data management and security: Establish a sound data management system, strengthen the management and monitoring of data collection, storage, transmission, and use processes to ensure data integrity, accuracy, and security. At the same time, enhance employee data security awareness training to increase employees' understanding and respect for data privacy and security.

[0073] The above embodiments of the present invention implement multiple data distribution models, promptly eliminating and resolving the U-distribution transformation process for non-normally distributed data. This is particularly important when data is insufficient to confirm the model's distribution status, fully demonstrating the advantages of the tolerance coefficient method for data transformation proposed in this invention. Furthermore, this embodiment, through data transformation algorithms, regresses typical distribution model variations, fully exposing the diverse scenarios of actual production processes, providing strong support for proposing combined application methods. Moreover, the data volume of this embodiment covers various production scenarios that enterprises may encounter, comprehensively validating the comprehensiveness and rationality of the evaluation method.

[0074] 4.3 Key Discoveries and Technological Breakthroughs a. A quality data evaluation standard for multiple varieties and small batches has been established.

[0075] Based on the current situation and application needs of multi-variety, small-batch manufacturing, the quality data evaluation criteria and innovations of this invention are of vital importance for improving production efficiency, ensuring product quality, and enhancing competitiveness.

[0076] b. A data conversion and processing method for tolerance utilization rate is proposed. Traditional methods of merging multiple small batches of data after normalization transformation are limited to normal or similar distributions. According to a 1999 study by Daimler-Chrysler & Ford, only 2% of the data in actual production exhibits a normal distribution. Therefore, the data processing method of normalization transformation and merging has many limitations. This invention proposes a data transformation algorithm that successfully avoids this problem.

[0077] c. Single-value control chart monitoring based on measurement uncertainty study The single-value monitoring based on measurement uncertainty proposed in this invention is a pioneering approach in multi-variety, small-batch SPC. It deeply considers various potential sources of measurement variation, including measurement equipment accuracy, environmental factors, and personnel differences, making measurement uncertainty research more detailed than MSA measurement system analysis. This helps to uncover subtle issues that might be overlooked by conventional MSA analysis. It enhances the reliability and credibility of measurement results. By quantifying and evaluating measurement uncertainty, it provides a clear understanding of the confidence interval of measured values, offering a more solid basis for decision-making, thus making the control chart monitoring results of 100% inspection results more convincing. It is particularly suitable for our company's current full-parameter inspection mode or production mode with 100% inspection of critical characteristic processes.

[0078] d. Stability of quality control charts based on binomial distribution Because the principle of quality control charts is based on low-probability events, what might not occur in a single trial will eventually exceed the control limits if there are a sufficient number of trials (i.e., a sufficient number of points on the quality control chart). As long as the number of overdue alarms is not exceeded, exceeding the control limits is generally acceptable. Based on the number of points on the control chart, and given a 99% or 99.73% confidence level, the maximum number of points on the control chart allowed to exceed the control limits for a specific subgroup is calculated using a Bernoulli binomial distribution. This method allows for a clearer understanding of the past processing status during the analysis phase.

[0079] While the embodiments disclosed in this invention are as described above, they are merely illustrative of the embodiments to facilitate understanding of the invention and are not intended to limit the invention. Any person skilled in the art to which this invention pertains may make any modifications and variations in the form and details of the implementation without departing from the spirit and scope disclosed herein; however, the scope of patent protection for this invention shall still be determined by the scope defined in the appended claims.

Claims

1. A method for quality data evaluation in multi-variety, small-batch manufacturing processes, characterized in that, include: Step 1, data standardization processing, including: using the tolerance coefficient method as the data transformation algorithm to obtain the tolerance utilization rate of individual monitoring characteristics; Step 2 involves evaluating the process capability of each monitoring characteristic, including: dividing the control process of multiple varieties and small batches into various scenarios according to the total number of samples N; for individual monitoring characteristics that can accumulate a total number of samples N, using conventional process capability evaluation methods by accumulating the total number of samples N, or merging standardized data from multiple monitoring characteristics to form a data volume that meets the requirements for implementing control process capability evaluation; for individual monitoring characteristics that cannot accumulate a total number of samples N, merging standardized data from multiple monitoring characteristics to implement control process capability evaluation, and using single-value monitoring as an auxiliary method for process control. Step 3: Evaluate the control process capability from Step 2, and formulate quality status assessment criteria for multi-variety, small-batch process control, including: If the total sample size N is too small to calculate the capability of this single control characteristic, and the capability obtained by merging the standardized data of each control characteristic increases the capability of this single control characteristic, then single-value monitoring or control charts should be used to assist in the control of this single control characteristic.

2. The method for quality data evaluation in multi-variety, small-batch manufacturing processes according to claim 1, characterized in that, Step 1 includes: Step 11: Use the tolerance coefficient method as the data conversion algorithm to group similar processing procedures into typical procedures, and monitor the tolerance utilization rate of a single measurement value as a single value. Step 12: Convert the upper and lower control limits of the tolerance for individual monitoring characteristics to ±1, thereby eliminating differences in units, dimensions and tolerances between different datasets.

3. The method for quality data evaluation in multi-variety, small-batch manufacturing processes according to claim 1, characterized in that, Step 2 includes: The control process of multiple varieties in small batches is categorized into three scenarios according to the total number of samples N. For scenarios where the total number of samples N ≥ the first threshold, the conventional process capability evaluation method is adopted; For scenarios where the second threshold ≤ N < the first threshold, for monitoring characteristics with a total number of accumulative samples N, conventional process capability evaluation methods are adopted after accumulating the total number of samples N to N≥ the first threshold. Alternatively, data from multiple monitoring characteristics with similar features are standardized and then merged to form a data volume that meets the requirements for implementing control process capability evaluation. For a single monitoring characteristic with a total sample size N that cannot be accumulated, since the total sample size N of the single monitoring characteristic is less than the second threshold, the data after standardization of each monitoring characteristic is used to evaluate the control process capability by merging the data, and process control is carried out by assisting single-value monitoring.

4. The method for quality data evaluation in multi-variety, small-batch manufacturing processes according to claim 3, characterized in that, Step 2, which involves merging standardized data from multiple monitoring characteristics to evaluate control process capability, includes: Step 21: Use measurement data of individual monitoring characteristics to perform capability calculation and capability evaluation; Step 22: Use the tolerance utilization rate obtained from the conversion of each measurement data in the single monitoring characteristic to perform capability calculation and capability evaluation; Step 23: After merging the standardized data of each individual monitoring characteristic, perform capability calculation and capability evaluation on all the merged standardized data.

5. The method for quality data evaluation in a multi-variety, small-batch manufacturing process according to any one of claims 1 to 4, characterized in that, Step 3 involves using single-value monitoring or control charts to assist in the control of individual control characteristics. This includes setting three levels of quality status assessment during the multi-variety, small-batch control process, based on the capability calculations for each control characteristic in Step 2. Level 0: No control chart is checked. For levels where the capability calculated after merging standardized data reaches or exceeds the qualified indicators, periodic capability evaluation is adopted, and process control is assisted by single-value monitoring. Level 1: On-site real-time monitoring and control charts are used to assist in process control. For scenarios where the sample size N < the second threshold, anomaly criteria are established for the control process of multiple varieties and small batches based on the allowable probability of anomaly occurrence. Level 2: Use analytical control charts to analyze the performance of the previous monitoring process.

6. The method for quality data evaluation in multi-variety, small-batch manufacturing processes according to claim 5, characterized in that, The discrepancy criteria established in Level 1 for the multi-variety, small-batch control process include: Guideline 1: Infringement of control boundaries is not permitted; Criterion 2: No nine consecutive points fall on one side of the center of the control chart; Criterion 3: There are no six consecutive points showing an upward or downward trend.

7. The method for quality data evaluation in multi-variety, small-batch manufacturing processes according to claim 5, characterized in that, The method of using analytical control charts for auxiliary process control in Level 2 is as follows: In the auxiliary control process of control charts for multi-variety, small-batch analysis, calculations are performed based on the number of data points in the control chart and the binomial distribution theory, allowing data points on the control chart to exceed the upper and lower control limits.

8. The method for quality data evaluation in multi-variety, small-batch manufacturing processes according to claim 7, characterized in that, The probability of data points in the control chart exceeding the upper and lower control limits is calculated as follows: ； g(k) represents the probability that exactly k samples exceed the control limit; k is the number of samples in the control chart that exceed the control limit; p represents the proportion of the population that exceeds the control limit; Assume p is a known quantity; n is the sample size; p represents the probability of a certain event X occurring.

9. The method for quality data evaluation in multi-variety, small-batch manufacturing processes according to claim 5, characterized in that, In the process control of multiple varieties and small batches, data judgment logic is formed according to sampling inspection n=2 / n=3, or 100% inspection n=1; respectively: In the sampling inspection n=2 / n=3 mode, the control chart selection type is set in combination with the two scenarios of N≥30 and N<30, and the maximum number of control points allowed to exceed the control limit is calculated based on the binomial distribution. In the 100% verification n=1 mode, a single-value control chart based on measurement uncertainty is set up. Taking into account various sources of measurement variation, the measurement uncertainty is quantified and evaluated, and then fed back to the confidence interval of the measured value. The process monitoring of the single value is extended on the basis of the existing control chart.

10. The method for quality data evaluation in multi-variety, small-batch manufacturing processes according to claim 5, characterized in that, The control chart used for real-time on-site monitoring in Level 1 is as follows: Control Chart 1: Pre-control Chart - Rainbow Chart, applied in control processes based on specification limits. By setting data partitions, it determines whether small batches of products fall into different tolerance areas, thereby enabling real-time monitoring and early warning; and the warning line for single-value monitoring is set at 75%-80% of the tolerance center. Control Chart 2: Single-value monitoring chart based on measurement uncertainty. When implementing 100% full inspection, measurement uncertainty is incorporated into process monitoring and applied to the monitoring process for single values.

Images