A cross-process defect root cause trace back method and system

Abstract

The present application relates to the technical field of defect detection, in particular to a cross-process defect root cause tracing method and system. The present application forms a multi-dimensional data feature matrix by integrating process parameters, equipment states and quality detection information, so that the performance evaluation of each process is more comprehensive. By constructing a process relationship diagram, the interaction and influence path between processes can be clearly depicted. At the same time, the introduction of process attenuation factors provides basic data support for quantifying the relationship between processes. By analyzing the process relationship diagram, the shortest propagation path and propagation probability of defects can be accurately identified. Combined with the construction of a knowledge graph, the root cause tracing of cross-process defects becomes systematic. The knowledge graph can not only effectively integrate and display data, but also facilitate quick problem positioning. By using adaptive correlation analysis technology, the system can intelligently adjust the analysis model and path when facing new data, continuously optimizing the defect detection and tracing process.

Description

Technical Field

[0001] This invention relates to the field of defect detection technology, and in particular to a method and system for tracing the root causes of defects across processes. Background Technology

[0002] Cross-process defects refer to defects in one process that affect multiple subsequent processes and even the quality of the final product. These defects do not originate in a single process but spread across multiple production stages, thus affecting the quality of the entire production process. This is because in a multi-process production process, the output of each process is usually the input of the next process. If there is a defect in the preceding process, it is difficult for the subsequent processes to avoid being affected.

[0003] Current methods for tracing the root causes of cross-process defects rely on statistical process control (SPC) to monitor quality data in real time during production. Statistical methods are used to identify fluctuations and anomalies in the production process, and then analyze the root causes of cross-process quality problems. However, SPC typically performs statistical analysis on quality data from a single process. For the complex interactions between multiple processes, it may not be able to reveal the source of defects. In tracing the root causes of cross-process defects, the relationships between different processes are very complex, and simply relying on SPC analysis of data fluctuations in a single process is insufficient to effectively trace the root causes of cross-process defects. Summary of the Invention

[0004] The main objective of this invention is to provide a method for tracing the root causes of defects across processes, aiming to solve the technical problems in the prior art.

[0005] This invention proposes a method for tracing the root causes of defects across processes, comprising:

[0006] Acquire the full process chain data of the target product, and obtain the production quality characteristic parameters of each process based on the full process chain data, wherein the production quality characteristic parameters include process parameter information, equipment status information and quality inspection information;

[0007] Based on each of the aforementioned process parameter information, equipment status information, and quality inspection information, obtain the corresponding process data feature matrix;

[0008] Obtain the corresponding process attenuation factor based on the process data feature matrix of any two processes, and obtain the cross-process correlation strength based on the process attenuation factor;

[0009] A process relationship diagram is constructed based on each process and the cross-process correlation strength, and defect propagation data is obtained based on the process relationship diagram. The defect propagation data includes the shortest propagation path, defect propagation probability, and defect type.

[0010] A knowledge graph is constructed based on the full process chain data, the strength of multiple cross-process associations, the shortest propagation path, the probability of defect propagation, and the defect type. Cross-process root cause tracing is then performed based on the knowledge graph.

[0011] Preferably, the step of obtaining the process data feature matrix of the corresponding process based on each of the process parameter information, equipment status information, and quality inspection information includes:

[0012] Based on the process parameter information, obtain multiple process temperature data, process pressure data, and process speed data, and obtain the corresponding temperature skewness coefficient, pressure skewness coefficient, and speed skewness coefficient for each of the process temperature data, process pressure data, and process speed data;

[0013] Obtain the corresponding process stability feature vector based on each of the temperature skewness coefficient, pressure skewness coefficient, and velocity skewness coefficient;

[0014] Based on the equipment status information, obtain multiple equipment efficiency data and equipment load data, and obtain the corresponding efficiency skewness coefficient and load skewness coefficient based on each of the equipment efficiency data and equipment load data;

[0015] Obtain the corresponding device stability feature vector based on each of the efficiency skewness coefficients and load skewness coefficients;

[0016] Based on the quality inspection information, multiple dimensional deviation data, surface quality data, and defect area data are obtained, and corresponding dimensional skewness coefficients, quality skewness coefficients, and defect skewness coefficients are obtained based on each of the dimensional deviation data, surface quality data, and defect area data.

[0017] Obtain the corresponding quality stability feature vector based on each of the aforementioned size skewness coefficient, mass skewness coefficient, and defect skewness coefficient;

[0018] Normalize each of the quality stability feature vector, equipment stability feature vector, and process stability feature vector to obtain the corresponding standard quality stability feature vector, standard equipment stability feature vector, and standard process stability feature vector.

[0019] Construct a process data feature matrix for the corresponding process based on multiple standard quality stability feature vectors, standard equipment stability feature vectors, and standard process stability feature vectors.

[0020] Preferably, the step of obtaining the corresponding process attenuation factor based on the process data feature matrix of any two processes includes:

[0021] Obtain a difference matrix based on the difference values ​​of corresponding features in the process data feature matrices of any two processes, and obtain the probability of each difference value based on the difference matrix;

[0022] The frequency of each difference value is obtained based on the probability of multiple difference values, and the entropy value of the quality index is obtained based on the frequency of multiple difference values.

[0023] The mean and maximum difference values ​​are obtained from the difference matrix, and the equipment health index is obtained from the mean and maximum difference values.

[0024] The process influence feature matrix is ​​obtained by taking the partial derivatives of the corresponding features in the process data feature matrices of any two processes, and the process complexity is obtained based on the process influence feature matrix.

[0025] The process attenuation factor is calculated based on the process complexity, equipment health index, and quality index entropy value, wherein the calculation formula is:

[0026]

[0027] Wherein, G(SJ) represents the process attenuation factor, Z(BS) represents the quality index entropy value, G(FZ) represents the process complexity, and S(JZ) represents the equipment health index.

[0028] Preferably, the step of obtaining the cross-process correlation strength based on the process attenuation factor includes:

[0029] The first installation layout data and parameter fluctuation amount are obtained based on the process data feature matrix of the first process, and the quality fluctuation amount corresponding to the parameter fluctuation amount is obtained based on the process data feature matrix of the second process.

[0030] Based on the quality fluctuation and parameter fluctuation, obtain the parameter influence coefficients for any two processes.

[0031] The installation location data and layout data of the first product are obtained based on the first installation layout data.

[0032] The second installation layout data is obtained based on the process data feature matrix of the second process, and the second product installation location data and the second product layout data are obtained based on the second installation layout data.

[0033] The installation location correlation coefficient is obtained based on the first product installation location data and the second product installation location data; the product layout correlation coefficient is obtained based on the first product layout data and the second product layout data.

[0034] Obtain the process attenuation factor, and calculate the cross-process correlation strength based on the process attenuation factor, parameter influence coefficient, product layout correlation coefficient, and installation location correlation coefficient. The calculation formula is as follows:

[0035] K(GQ)=D(SY)*e -G(SJ)*[1-B(JX)*A(WX)] ;

[0036] Wherein, K(GQ) represents the cross-process correlation strength, G(SJ) represents the process attenuation factor, D(SY) represents the parameter influence coefficient, B(JX) represents the product layout correlation coefficient, and A(WX) represents the installation location correlation coefficient.

[0037] Preferably, the step of constructing a process relationship diagram based on each process and the cross-process correlation strength, and obtaining defect propagation data based on the process relationship diagram, includes:

[0038] Obtain all cross-process association strengths for each process and select the largest cross-process association strength as the dependency association strength;

[0039] The most dependent process is determined based on the strength of each dependency association and the corresponding process, thus obtaining the process dependency relationship;

[0040] Each process is treated as a node, the process dependency is treated as an edge, and the corresponding cross-process association strength is treated as the edge weight to construct a process relationship graph.

[0041] Obtain all propagation paths for each process based on the edges of the process relationship graph, and determine the shortest propagation path based on all propagation paths;

[0042] Obtain the historical defect probability for each process, and obtain the propagation probability based on the historical defect probability and the shortest propagation path;

[0043] The defect propagation probability is obtained based on the propagation probability and the historical defect probability.

[0044] Preferably, the building module includes:

[0045] The selection unit is used to obtain all cross-process association strengths for each process and select the largest cross-process association strength as the dependency association strength.

[0046] The first determining unit is used to determine the most dependent process based on the strength of each dependency association and the corresponding process, thereby obtaining the process dependency relationship;

[0047] The building unit is used to construct a process relationship graph by taking each process as a node, the process dependency as an edge, and the corresponding cross-process association strength as the edge weight.

[0048] The second determining unit is used to obtain all propagation paths for each process based on the edges of the process relationship diagram, and to determine the shortest propagation path based on all propagation paths.

[0049] The first acquisition unit is used to acquire the historical defect probability of each process and acquire the propagation probability based on the historical defect probability and the shortest propagation path.

[0050] The second acquisition unit is used to acquire the defect propagation probability based on the propagation probability and the historical defect probability.

[0051] This application also provides a cross-process defect root cause tracing system, including:

[0052] The first acquisition module is used to acquire the full process chain data of the target product, and to acquire the production quality characteristic parameters of each process based on the full process chain data, wherein the production quality characteristic parameters include process parameter information, equipment status information and quality inspection information;

[0053] The second acquisition module is used to acquire the process data feature matrix of the corresponding process based on each of the process parameter information, equipment status information and quality inspection information;

[0054] The third acquisition module is used to obtain the corresponding process attenuation factor based on the process data feature matrix of any two processes, and to obtain the cross-process correlation strength based on the process attenuation factor.

[0055] The construction module is used to construct a process relationship diagram based on each process and the cross-process correlation strength, and to obtain defect propagation data based on the process relationship diagram, wherein the defect propagation data includes the shortest propagation path, defect propagation probability, and defect type;

[0056] The traceability module is used to construct a knowledge graph based on the full process chain data, the strength of multiple cross-process associations, the shortest propagation path, the probability of defect propagation, and the defect type, and to perform cross-process root cause tracing based on the knowledge graph.

[0057] Preferably, the step of performing cross-process root cause tracing based on the knowledge graph includes:

[0058] Obtain the defect type, and select multiple related processes from the knowledge graph based on the defect type to obtain a process set;

[0059] The cross-process association strength of each process is obtained based on the process set, and the corresponding shortest propagation path is obtained by reverse inference from the knowledge graph based on each cross-process association strength.

[0060] The probability of defect propagation is inferred from each of the shortest propagation paths.

[0061] Obtain the entropy value of the quality index for each process in the process set, and obtain the probability of occurrence of the corresponding defect based on the entropy value of each quality index and the corresponding defect propagation probability;

[0062] The occurrence probabilities of the multiple defects are sorted in order of magnitude, and the process corresponding to the highest defect occurrence probability is selected as the root cause process.

[0063] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described cross-process defect root cause tracing method.

[0064] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described cross-process defect root cause tracing method.

[0065] The beneficial effects of this invention are as follows: By acquiring full-process chain data of the target product, this invention greatly enhances the understanding of the interactions between multiple processes. By integrating process parameters, equipment status, and quality inspection information, a multi-dimensional data feature matrix is ​​formed, making the performance evaluation of each process more comprehensive. By constructing a process relationship diagram, the interactions and influence paths between processes can be clearly depicted. This graphical approach makes complex interactions more intuitive, helping technicians better understand the potential causes of defects. Simultaneously, the introduction of process attenuation factors provides basic data support for quantifying the relationships between processes, avoiding the ambiguity and limitations of previous methods in analyzing correlation strength. By analyzing the process relationship diagram, the shortest propagation path and propagation probability of defects can be accurately identified, providing a more comprehensive understanding of the production process. Effective risk management, rather than reactive retrospective analysis, reduces economic losses and time costs associated with defects. The construction of knowledge graphs systematizes the root cause tracing of defects across processes. Knowledge graphs not only effectively integrate and display data but also facilitate rapid problem location, especially when facing complex multi-process interactions. They help technicians quickly identify the root cause and take appropriate measures. Adaptive correlation analysis technology enables the system to intelligently adjust its analysis model and path in response to new data, continuously optimizing the defect detection and tracing process. This flexibility is particularly important in rapidly changing production environments, effectively improving quality management efficiency. Through the fusion and analysis of multi-dimensional data, the relationships between complex processes are revealed, providing enterprises with efficient, accurate, and systematic defect investigation and management methods. Attached Figure Description

[0066] Figure 1 This is a schematic diagram of a method flow according to an embodiment of the present invention.

[0067] Figure 2 This is a schematic diagram of the system structure according to an embodiment of the present invention.

[0068] Figure 3 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application.

[0069] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0070] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0071] like Figure 1 As shown, this application provides a method for tracing the root cause of defects across processes, including:

[0072] S1. Obtain the full process chain data of the target product, and obtain the production quality characteristic parameters of each process based on the full process chain data, wherein the production quality characteristic parameters include process parameter information, equipment status information and quality inspection information;

[0073] S2. Obtain the process data feature matrix of the corresponding process based on each of the process parameter information, equipment status information and quality inspection information;

[0074] S3. Obtain the corresponding process attenuation factor based on the process data feature matrix of any two processes, and obtain the cross-process correlation strength based on the process attenuation factor;

[0075] S4. Construct a process relationship diagram based on each process and the cross-process correlation strength, and obtain defect propagation data based on the process relationship diagram, wherein the defect propagation data includes the shortest propagation path, defect propagation probability, and defect type;

[0076] S5. Construct a knowledge graph based on the full process chain data, the strength of multiple cross-process associations, the shortest propagation path, the probability of defect propagation, and the defect type, and perform cross-process root cause tracing based on the knowledge graph.

[0077] As described in steps S1-S5 above, this invention acquires full-process chain data of the target product and obtains process parameter information, equipment status information, and quality inspection information for each process based on the full-process chain data. By collecting data from all processes throughout the entire production process, a complete picture of each process can be obtained, ensuring that all process information is covered, eliminating information blind spots, and providing a basis for cross-process analysis. Based on the full-process chain data, the process parameters, equipment status, and quality inspection information for each process can be extracted, allowing the data for each process to be refined to the level of process parameters, equipment status, and quality inspection. This enables precise capture of the specific sources of quality fluctuations, thereby allowing for more refined identification of potential issues in the process. The problem provides a basis for subsequent process optimization and quality improvement. By obtaining the process data feature matrix for each process step through information on process parameters, equipment status, and quality inspection, the process parameters, equipment status, and quality inspection information are integrated into a process data feature matrix. The data for each process step is represented in matrix form, providing a structured data format for subsequent analysis. Matrix processing makes the data from different processes structured and standardized, facilitating subsequent processing and calculation. Converting process data into matrix form allows for efficient subsequent calculations using numerical analysis methods, improving efficiency. The corresponding process attenuation factor is obtained from the process data feature matrices of any two processes, and the cross-process correlation strength is obtained based on the process attenuation factor. Specifically, the attenuation factor is used to obtain the cross-process correlation strength. Obtaining the corresponding process attenuation factor from the process data feature matrix of a process refers to calculating the corresponding process attenuation factor from the process data feature matrices of any two processes in the entire process chain of the target product. Each process has a corresponding process attenuation factor relative to multiple other processes. Through the process attenuation factor, a corresponding cross-process correlation strength can be obtained. Therefore, each process has multiple cross-process correlation strengths. By calculating the attenuation factor between any two processes, the degree of influence between processes is reflected. This factor calculation considers changes in process parameters, fluctuations in equipment status, and quality inspection results. The attenuation factor can quantify the degree of influence between different processes, revealing the weight and role of each process in the entire production process. The introduction of this method allows for a more precise understanding of the impact of a particular process on defects, providing crucial clues for root cause tracing. Calculating the cross-process correlation strength using attenuation factors helps reveal the mutual influence between different processes. Cross-process correlation strength provides a quantitative analysis tool for revealing the interactions between multiple processes, aiding in the identification of key inter-process correlations. Through cross-process correlation strength, it is possible to effectively identify which processes have a greater impact on defects, thus enabling effective tracing. A process relationship diagram is constructed using each process and its cross-process correlation strength. Based on this diagram, the shortest propagation path, defect propagation probability, and defect type of defect propagation data are obtained. The process relationship diagram, constructed using processes and their correlation strengths, illustrates the connections and influence paths between each process.Process relationship diagrams clearly illustrate the relationships between processes, making defect propagation paths and impact chains readily apparent. This facilitates intuitive understanding of complex interactions between processes, aiding in analysis and decision-making. Based on the process relationship diagram, the shortest path for defect propagation, the probability of defect propagation, and defect types are calculated. This data helps analyze how defects propagate during production and their potential consequences. Accurately calculating the path, probability, and type of defect propagation provides clear clues for root cause analysis. The shortest propagation path can quickly identify the root cause of defects, and the defect propagation probability further improves the accuracy of the analysis. A knowledge graph is constructed using full-process chain data, the strength of multiple cross-process associations, the shortest propagation path, the defect propagation probability, and the defect type. Furthermore, based on the knowledge graph... Cross-process root cause analysis aggregates data from the entire process chain, including cross-process correlation strength and defect propagation paths, to construct a comprehensive knowledge graph. This graph not only describes the relationships between processes but also includes the defect propagation process and prediction. The knowledge graph integrates complex cross-process information and provides comprehensive knowledge support for root cause analysis, systematically managing complex production data. Through a graph-based presentation, it provides data support and decision-making basis for production line optimization, defect prevention, and continuous improvement. Utilizing the constructed knowledge graph for cross-process defect root cause analysis allows for in-depth analysis of the relationships between processes and the source of defects. Cross-process root cause analysis can accurately identify defect sources originating across multiple processes, solving the problem of ineffective traceability in existing technologies.

[0078] In one embodiment, step S2, which involves obtaining the process data feature matrix of the corresponding process based on each of the process parameter information, equipment status information, and quality inspection information, includes:

[0079] S21. Obtain multiple process temperature data, process pressure data, and process speed data based on the process parameter information, and obtain the corresponding temperature skewness coefficient, pressure skewness coefficient, and speed skewness coefficient based on each of the process temperature data, process pressure data, and process speed data;

[0080] S22. Obtain the corresponding process stability feature vector based on each of the temperature skewness coefficient, pressure skewness coefficient, and velocity skewness coefficient;

[0081] S23. Obtain multiple device efficiency data and device load data based on the device status information, and obtain the corresponding efficiency skewness coefficient and load skewness coefficient based on each of the device efficiency data and device load data;

[0082] S24. Obtain the corresponding equipment stability feature vector based on each of the efficiency skewness coefficients and load skewness coefficients;

[0083] S25. Obtain multiple dimensional deviation data, surface quality data, and defect area data based on the quality inspection information, and obtain the corresponding dimensional skewness coefficient, quality skewness coefficient, and defect skewness coefficient based on each of the dimensional deviation data, surface quality data, and defect area data;

[0084] S26. Obtain the corresponding quality stability feature vector based on each of the size skewness coefficient, mass skewness coefficient, and defect skewness coefficient;

[0085] S27. Normalize each of the quality stability feature vector, equipment stability feature vector, and process stability feature vector to obtain the corresponding standard quality stability feature vector, standard equipment stability feature vector, and standard process stability feature vector.

[0086] S28. Construct a process data feature matrix for the corresponding process based on the multiple standard quality stability feature vectors, standard equipment stability feature vectors, and standard process stability feature vectors.

[0087] As described in steps S21-S28 above, this invention acquires multiple process temperature data, process pressure data, and process speed data through process parameter information. Based on each process's temperature, pressure, and speed data, it obtains corresponding temperature skewness coefficients, pressure skewness coefficients, and speed skewness coefficients. Through each temperature, pressure, and speed skewness coefficient, it obtains a corresponding process stability feature vector. By acquiring temperature, pressure, and speed data from multiple processes, key process parameters in the production process can be comprehensively recorded and monitored. By calculating the skewness coefficient of each process, the distribution characteristics of various process parameters in different processes can be revealed. The skewness coefficient helps to detect deviations and anomalies in process parameters, providing information for process stability. A deep understanding of the process is crucial for quality control of complex multi-process systems. The process stability feature vector obtained based on skewness coefficients is a mathematical abstraction of the state of each process. It provides a quantifiable standard for the correlation between multiple processes, making it easier to identify stability issues in process parameters by comparing the feature vectors of each process. Multiple equipment efficiency and load data are obtained through equipment status information, and corresponding efficiency and load skewness coefficients are obtained for each. The corresponding equipment stability feature vector is then derived from each efficiency and load skewness coefficient. Equipment efficiency and load data directly affect the quality and output of the production process. By collecting this data, it is possible to achieve… Dynamic monitoring of equipment status, including efficiency skewness and load skewness coefficients, reflects performance deviations during actual operation. This helps identify anomalies under different working conditions. By quantifying performance deviations, it provides data support for equipment maintenance, adjustment, and optimization, helping to reduce production instability caused by equipment problems. Calculating equipment stability feature vectors allows for the formation of a complete performance index system at the equipment level. Transforming equipment performance data into feature vectors not only enables real-time monitoring of equipment stability but also allows for comparison with other process or quality data. Multiple dimensional deviation data, surface quality data, and defect area data can be obtained through quality inspection information, and further analysis can be performed based on each dimensional deviation, surface quality, and defect area data. By acquiring corresponding dimensional skewness coefficients, quality skewness coefficients, and defect skewness coefficients from the area data, and obtaining corresponding quality stability feature vectors for each of these coefficients, we can gain a deeper understanding of various aspects of product quality through the collection of quality inspection information. Data such as dimensional deviations, surface quality, and defect area directly reflect the production precision and appearance quality of the product. These skewness coefficients can reflect the abnormal distribution of different quality characteristics (such as dimensions, surface quality, and defects) during the production process. By calculating these skewness coefficients, we can reveal the root causes of quality fluctuations, especially the quality impact across processes, helping to locate potential process or equipment problems.Quality stability feature vectors can transform different quality inspection information into standardized feature parameters, providing a unified reference for subsequent analysis. These stable feature vectors enable more accurate monitoring and prediction of quality, and provide a more scientific basis for tracing the root causes of defects across processes. By normalizing each quality stability feature vector, equipment stability feature vector, and process stability feature vector, corresponding standard quality stability feature vector, standard equipment stability feature vector, and standard process stability feature vector are obtained. A process data feature matrix for the corresponding process is constructed using multiple standard quality stability feature vectors, standard equipment stability feature vectors, and standard process stability feature vectors, and then normalized. This allows for the effective comparison and integration of feature vectors from different sources (processes, equipment, quality, etc.). Standardized feature vectors not only improve data comparability but also eliminate potential dimensional differences between data sources, providing a foundation for multi-dimensional, multi-process cross-domain analysis. By integrating multiple standardized feature vectors into a single process data feature matrix, the key characteristics of each process can be comprehensively displayed, providing data support for further analysis and prediction. This feature matrix not only reveals the relationships between processes but also provides an intuitive data framework for tracing the root causes of cross-process defects, greatly improving the accuracy and efficiency of inter-process interaction analysis.

[0088] In one embodiment, step S3, which involves obtaining the corresponding process attenuation factor based on the process data feature matrix of any two processes, includes:

[0089] S31. Obtain a difference matrix based on the difference values ​​of corresponding features in the process data feature matrices of any two processes, and obtain the probability of each difference value based on the difference matrix.

[0090] S32. Obtain the frequency of each difference value based on the probability of multiple difference values, and calculate the entropy value of the quality index based on the frequency of multiple difference values. The calculation formula is as follows:

[0091]

[0092] Where Z(BS) represents the entropy value of the quality index, and C(PL) i This represents the frequency of the i-th difference value, where i represents the index of the frequency of the difference value, and N represents the number of frequencies of the difference value.

[0093] S33. Obtain the mean and maximum difference value based on the difference matrix, and obtain the equipment health index based on the mean and maximum difference value;

[0094] S34. Obtain the process influence feature matrix based on the partial derivatives of the corresponding features in the process data feature matrices of any two processes, and obtain the process complexity based on the process influence feature matrix.

[0095] S35. Calculate the process attenuation factor based on the process complexity, equipment health index, and quality index entropy value, wherein the calculation formula is:

[0096]

[0097] Wherein, G(SJ) represents the process attenuation factor, Z(BS) represents the quality index entropy value, G(FZ) represents the process complexity, and S(JZ) represents the equipment health index.

[0098] As described in steps S31-S35 above, the calculation formulas for the process attenuation factor and the entropy value of the quality index require normalization of their respective parameters to eliminate dimensional differences between different variables. This ensures all variables are on the same order of magnitude, making the calculation more stable and effective. The method for obtaining the difference matrix based on the difference values ​​of corresponding features in the process data feature matrices of any two processes is as follows: first, normalize the data features within the process data feature matrices of both processes; then, calculate the difference between the corresponding data features in the process data feature matrices of the two processes to obtain the difference value of each data feature. Multiple difference values ​​can then be arranged according to the format of the process data feature matrix to obtain the difference matrix. The method for obtaining the frequency of each difference value based on the probability of multiple difference values ​​is as follows: first, calculate the total probability by summing the probabilities of multiple difference values; then, obtain the frequency of each difference value by the ratio of the probability of each difference value to the total probability. The equipment health index is obtained by subtracting the mean from 1 and the maximum difference value. The frequency of the equipment health index is obtained by considering the differences within the process influence feature matrix. The complexity of a process is obtained by weighted summation of characteristic data (partial derivatives). This invention obtains a difference matrix by calculating the difference values ​​of corresponding features in the process data feature matrices of any two processes. By calculating the feature difference values ​​of any two processes in the process data feature matrix, the differences between different processes can be accurately captured, helping to identify potential sources of defects between different processes. Traditional methods often only analyze the data fluctuations of a single process, ignoring the complex relationships between processes. This invention effectively improves the accuracy and comprehensiveness of cross-process defect tracing by constructing a multi-process feature difference matrix. It obtains the probability of each difference value based on the difference matrix, and obtains the frequency of each difference value through the probabilities of multiple difference values. By calculating the probability of the difference values, the likelihood of each process feature difference occurring can be quantified. This makes cross-process defect tracing not only rely on qualitative analysis, but also combines probabilistic statistical methods, providing a more objective and quantitative analysis means. By statistically analyzing the frequency of occurrence of multiple difference values, it helps to assess which feature differences frequently occur between processes, further revealing the potential key factors of cross-process defects. This frequency analysis enhances the statistical significance of variance values, making defect source tracing more reliable. Traditional defect tracing methods often focus on fluctuations in single data points while neglecting recurring variance characteristics, potentially leading to the overlooking of certain underlying problems. This invention, through frequency analysis, can more comprehensively identify key factors that continuously affect process quality and calculate the entropy value of quality indicators based on the frequency of multiple variance values. This is because each process in production has certain characteristic value differences, reflecting various factors such as process fluctuations, equipment status, and personnel operation. These variance values ​​are usually random and exhibit a certain probability distribution. The frequency of multiple variance values ​​(i.e., the number of times or probability of different variance values ​​occurring) can help understand the prevalence or anomaly of these variances in the production process.The calculation of entropy values ​​for quality indicators reflects the overall quality volatility during the production process. It quantifies product quality uncertainty by analyzing the distribution of variance values. A more uniform distribution of variance values ​​results in a higher entropy value, indicating greater quality volatility. Conversely, a higher probability of certain variance values ​​leads to a lower entropy value, suggesting more stable quality. The calculation of frequency and entropy values ​​helps identify potential instability factors in the process and pinpoint the sources of quality fluctuations. Entropy values ​​quantify the uncertainty and complexity of product quality, providing a basis for subsequent quality optimization. As a measure of system uncertainty, the calculation of quality indicator entropy values ​​allows for the assessment of complexity and uncertainty between different processes. Furthermore, the calculation of quality indicator entropy values ​​provides a quantitative standard for defect tracing. This invention effectively measures potential abnormal fluctuations in the system during production. Traditional methods often fail to provide a holistic quality control perspective. The introduction of entropy makes quality monitoring of the entire production process more comprehensive and dynamic. Through entropy analysis, it is possible to more accurately determine whether there are abnormal fluctuations or potential defects in a process. By obtaining the mean and maximum difference values ​​through the difference matrix, the standard and extreme differences between processes can be clearly identified. This provides a direct basis for assessing the health status of processes and potential defects. Traditional methods typically focus on the average data analysis of processes, neglecting the impact of extreme data points. This invention, through the analysis of maximum difference values, more comprehensively reveals the key factors that may lead to defects. Equipment health indicators are obtained based on the mean and maximum difference values. Equipment health is an important indicator for assessing whether production equipment is in good working condition. The combination of mean and maximum difference values ​​can provide a basis for decision-making regarding equipment maintenance and optimization. By accurately calculating equipment health, it is possible to help detect equipment anomalies in a timely manner and reduce cross-process defects caused by equipment problems. Existing equipment health assessment technologies usually rely on a single data source, lacking comprehensive and detailed monitoring of equipment status. This invention comprehensively considers multiple data characteristics during equipment operation, providing a more scientific and reliable health assessment method. By obtaining the partial derivatives of corresponding features in the process data feature matrices of any two processes, the process influence feature matrix is ​​obtained. Partial derivative analysis can reveal... By identifying the sensitivity relationships between different process characteristics, this invention pinpoints key factors influencing process outcomes, providing a more detailed process impact analysis and revealing potential inter-process interactions. Existing methods typically rely on static data analysis, lacking a deep understanding of the dynamic impacts between processes. This invention, through partial derivative analysis, more accurately captures the interactions and influences between processes, enhancing the depth and precision of the analysis. Furthermore, it obtains process complexity based on the process impact feature matrix. Assessing process complexity helps identify complex processes that are difficult to control, allowing for timely improvement measures. This makes the entire process analysis more systematic and comprehensive, effectively preventing cross-process defects caused by complex processes. Existing technologies typically cannot quantify process complexity.This invention provides more targeted improvement suggestions through quantitative process complexity assessment. It calculates a process attenuation factor using process complexity, equipment health indicators, and quality index entropy values. The process attenuation factor is a key parameter for assessing process stability and sustainability. Process complexity is obtained through a process influence feature matrix, reflecting the inherent complexity of the process. Higher complexity means more parameters and variables need to be controlled, potentially leading to greater process fluctuations and impacting the stability of subsequent processes and product quality. Therefore, higher process complexity may have a greater impact on subsequent processes, and the attenuation factor may increase accordingly. Equipment health indicators directly affect process stability; equipment failure, aging, or instability can exacerbate process fluctuations, thus affecting the quality of the entire production process. Lower equipment health indicators indicate a lower impact on process stability. The greater the impact on the production process, the higher the attenuation factor typically is. The entropy value of the quality indicator reflects the degree of quality fluctuation. Processes with large quality fluctuations may indicate difficulties in quality control, and subsequent processes will also be affected by these fluctuations. A higher entropy value indicates greater quality volatility in the process, and the attenuation factor may be even higher, meaning that the impact of preceding processes on subsequent processes gradually increases. Process complexity, equipment health, and quality fluctuations (reflected by entropy values) can comprehensively reflect the influence relationships between processes from different perspectives. Calculating the attenuation factor using these factors allows for a comprehensive consideration of process complexity, equipment status, and quality fluctuations, resulting in a more accurate attenuation factor. By calculating the attenuation factor, it is possible to identify which processes have a greater impact on the entire production process, helping managers maintain, adjust, or optimize key processes and equipment.

[0099] By combining process complexity, equipment health, and quality index entropy, the decline trend of a process can be comprehensively assessed, providing a feasible basis for prevention and repair. Existing technologies lack a systematic method for comprehensively evaluating multiple factors. This invention effectively predicts possible process decline through multi-factor analysis, preventing problems from occurring in advance. By introducing methods such as probability statistics, entropy analysis, and partial derivative analysis, this invention can more comprehensively and objectively assess abnormal factors in the production process, improving the accuracy and reliability of defect tracing.

[0100] In one embodiment, step S3, which involves obtaining the cross-process correlation strength based on the process attenuation factor, includes:

[0101] S36. Obtain the first installation layout data and parameter fluctuation amount based on the process data feature matrix of the first process, and obtain the quality fluctuation amount corresponding to the parameter fluctuation amount based on the process data feature matrix of the second process.

[0102] S37. Obtain the parameter influence coefficients for any two processes based on the quality fluctuation and parameter fluctuation.

[0103] S38. Obtain the first product installation location data and the first product layout data based on the first installation layout data;

[0104] S39. Obtain the second installation layout data based on the process data feature matrix of the second process, and obtain the second product installation position data and the second product layout data based on the second installation layout data.

[0105] S310. Obtain the installation location correlation coefficient based on the first product installation location data and the second product installation location data, and obtain the product layout correlation coefficient based on the first product layout data and the second product layout data.

[0106] S311. Obtain the process attenuation factor, and calculate the cross-process correlation strength based on the process attenuation factor, parameter influence coefficient, product layout correlation coefficient, and installation location correlation coefficient, wherein the calculation formula is:

[0107] K(GQ)=D(SY)*e -G(SJ)*[1-B(JX)*A(WX)] ;

[0108] Wherein, K(GQ) represents the cross-process correlation strength, G(SJ) represents the process attenuation factor, D(SY) represents the parameter influence coefficient, B(JX) represents the product layout correlation coefficient, and A(WX) represents the installation location correlation coefficient.

[0109] As described in steps S36-S311 above, this invention obtains the first installation layout data and parameter fluctuations through the process data feature matrix of the first process. By using the process data feature matrix to obtain the first installation layout data and parameter fluctuations, it can accurately capture data features related to installation layout and parameter fluctuations in the first process, providing basic data support for subsequent quality analysis and cross-process relationship establishment. By quantifying and extracting features in the process, it can effectively reveal key factors affecting product quality and lay a solid data foundation for defect tracing. The quality fluctuation corresponding to the parameter fluctuations is obtained based on the process data feature matrix of the second process. The quality fluctuation refers to the product's performance in the second process due to the influence of parameters in the first process. The fluctuation in product quality caused by fluctuations in parameters is analyzed by examining the process data feature matrix of the second process. This correlation between parameter fluctuations and quality fluctuations allows parameter fluctuations across different processes to be directly reflected in quality fluctuations, effectively identifying the impact of parameter changes on quality fluctuations. This provides a clear causal chain for cross-process defect analysis. Accurate prediction and quantification of quality fluctuations facilitates precise location of defect sources. The parameter influence coefficients for any two processes are obtained using quality and parameter fluctuations. In multiple processes, parameter fluctuations are often key factors leading to defects. Calculating these parameter influence coefficients quantitatively describes the mutual influence relationships between parameters in different processes, revealing... The interdependence between different processes provides a quantitative analysis tool for tracing the root causes of cross-process defects. First, installation layout data is obtained to acquire the installation location and layout data of the first product. Second, installation layout data is obtained through the process data feature matrix of the second process. Based on this second installation layout data, installation location and layout data of the second product are then acquired. By obtaining the product's installation location and layout data from the first installation layout data, the installation method can be accurately understood at the spatial and structural levels. This provides spatial and structural data support for subsequent cross-process defect analysis, facilitating a more comprehensive analysis of potential problems during installation. The installation layout data of the second process provides a different perspective than that of the first process. The installation data source, by extracting the installation location and layout data of the second product, allows for comparison of installation differences across different processes. This not only provides more multi-dimensional data support for cross-process defect analysis but also reveals the potential impact of installation methods and layout differences on the final product quality. The installation location correlation coefficient is obtained by using the installation location data of the first and second products. This coefficient quantitatively describes the similarity or difference in product installation locations across different processes, thereby assessing the impact of installation location changes on product quality. The introduction of the installation location correlation coefficient provides spatial correlation analysis for root cause analysis of defects across different processes, revealing quality problems caused by differences in installation location.Product layout correlation coefficients are obtained by analyzing first and second product layout data. These coefficients quantify the similarity or difference in product layout between two processes, revealing the impact of layout design on product quality. Especially when multiple processes are performed simultaneously, comparing layout differences across processes allows for a clearer identification of the contribution of layout issues to quality fluctuations. By obtaining process attenuation factors and calculating cross-process correlation strength based on these factors, parameter influence coefficients, product layout correlation coefficients, and installation location correlation coefficients, the attenuation effect of each process on product quality can be effectively considered. This, combined with the parameter influence coefficients, product layout correlation coefficients, and installation location correlation coefficients, further strengthens the correlation between product layout and quality. The correlation coefficient enables a comprehensive analysis of the correlation strength between multiple processes, improving the accuracy of cross-process quality analysis and allowing cross-process defect tracing to move beyond single-process limitations. By comprehensively considering the complex relationships between different processes, it can more fully reveal the root causes of quality fluctuations. Therefore, this invention uses multi-dimensional data matrix analysis to accurately extract the interactions and quality fluctuation factors between processes, effectively tracing the root causes of cross-process defects. Unlike traditional methods that rely solely on data from a single process, this method, through interconnected analysis of multiple processes, reveals complex dependencies between processes and accurately locates the source of defects. By calculating correlation coefficients and influence coefficients between processes, it can optimize process layout and installation. This method, which considers factors such as location, enhances quality control in the production process. It is suitable for complex production flows and allows for flexible adjustment and optimization for different processes and parameters, demonstrating broad application potential. By calculating cross-process correlation strength using process attenuation factors, parameter influence coefficients, product layout correlation coefficients, and installation location correlation coefficients, the aim is to comprehensively quantify and optimize the interactions and influences between different processes, thereby improving product quality and production efficiency. The parameter influence coefficient reflects the degree to which the input parameters of a certain process affect the output results of other processes (such as product quality and installation effect). Quantifying the influence of different parameters provides a deeper understanding of how the parameters of a certain process affect the quality fluctuations of subsequent processes. The product layout correlation coefficient describes... The layout design and installation position of products between different processes influence the interactions between processes. If the product layouts of preceding and subsequent processes are similar or related, their process and quality fluctuations may be more closely linked. The layout correlation coefficient shows that the influence between processes is not only based on the characteristics and parameters of process data, but also considers physical spatial factors. The installation position of the workpiece affects factors such as stress and temperature distribution during processing, thus affecting the degree of quality fluctuation. The installation position correlation coefficient reflects the similarity or difference in the installation positions of products between different processes. The installation position of the product may affect the interaction between processes. By considering the correlation of installation positions between processes, it helps to predict the spatial relationships between processes.Especially in multi-process operations, a change in one location can affect the execution of multiple processes. Therefore, the installation location correlation coefficient is used to reflect the impact of spatial relationships on the strength of cross-process correlation.

[0110] In one embodiment, step S4, which involves constructing a process relationship diagram based on each process and the cross-process correlation strength, and obtaining defect propagation data based on the process relationship diagram, includes:

[0111] S41. Obtain all cross-process association strengths for each process and select the largest cross-process association strength as the dependency association strength;

[0112] S42. Determine the most dependent process based on the strength of each dependency association and the corresponding process to obtain the process dependency relationship;

[0113] S43. Construct a process relationship graph by treating each process as a node, process dependencies as edges, and the corresponding cross-process association strength as edge weights.

[0114] S44. Obtain all propagation paths for each process based on the edges of the process relationship diagram, and determine the shortest propagation path based on all propagation paths;

[0115] S45. Obtain the historical defect probability for each process, and obtain the propagation probability based on the historical defect probability and the shortest propagation path;

[0116] S46. Obtain the defect propagation probability based on the propagation probability and the historical defect probability.

[0117] As described in steps S41-S46 above, the propagation path is obtained using a graph algorithm through the edges in the process relationship graph. Specifically, a depth-first search can be used to find all possible paths starting from a certain process. This invention obtains all cross-process association strengths for each process and selects the largest cross-process association strength as the dependency association strength. By obtaining all cross-process association strengths for each process, the interaction relationships between different processes can be accurately identified, avoiding the limitation of traditional methods that can only analyze a single process. Selecting the largest cross-process association strength as the dependency association strength ensures that the traceability process focuses on the most critical inter-process influences, thereby improving traceability accuracy. The most critical inter-process influence is determined by each dependency association strength and its corresponding process. By identifying process dependencies, the relationship between each process and its most dependent process can be simplified, simplifying complex inter-process dependency networks and facilitating further analysis and operation. This helps determine which processes have a significant impact on defect propagation, facilitates the discovery of potential defect sources, and optimizes monitoring and prevention measures in subsequent production processes. A process relationship graph is constructed by treating each process as a node, process dependencies as edges, and the corresponding cross-process association strength as edge weights. Transforming process relationships into a graph model vividly illustrates the interrelationships between processes and allows for effective analysis using graph theory methods. This modeling approach offers high visualization and operability, facilitating the intuitive identification of defect propagation paths between processes. The connection strength, used as the edge weight, further amplifies the magnitude of the influence between different processes, providing a more accurate basis for subsequent defect tracing. By obtaining all propagation paths for each process through the edges of the process relationship graph, and determining the shortest propagation path based on all paths, the interference of redundant information can be reduced by analyzing all propagation paths and selecting the shortest path. This accurately pinpoints the core path of defect propagation, improving the speed and accuracy of defect location. Determining the shortest propagation path simplifies the defect tracing process, enabling the tracing method to still execute efficiently when facing complex interactions between processes. By obtaining the historical defect probability for each process and calculating the propagation probability based on the historical defect probability and the shortest propagation path, the historical defect probability of each process can be incorporated into the process. This system can make predictions based on past data, thereby improving the accuracy of defect propagation probability predictions and providing a reference for future production processes. By combining historical defect probabilities and propagation paths, dynamic analysis of defect propagation in each process can be achieved, making it more timely and targeted, and facilitating the development of preventative measures. By obtaining defect propagation probabilities through propagation probabilities and historical defect probabilities, and combining these probabilities, the generation and propagation mechanisms of defects can be comprehensively considered, providing a comprehensive quantitative analysis of the mutual influence between processes. Through accurate defect propagation probability prediction, enterprises can identify high-risk processes in advance and take corresponding preventative measures, reducing the occurrence and propagation of defects, improving production quality and efficiency, and enhancing the accuracy of defect traceability.It can also help companies better control quality during the production process and reduce losses caused by defects.

[0118] In one embodiment, step S5, which involves cross-process root cause tracing based on the knowledge graph, includes:

[0119] S51. Obtain the defect type, and select multiple related processes from the knowledge graph according to the defect type to obtain a process set;

[0120] S52. Obtain the cross-process association strength of each process according to the process set, and reverse-engineer the corresponding shortest propagation path from the knowledge graph based on each cross-process association strength;

[0121] S53. Based on each of the shortest propagation paths, the corresponding defect propagation probability is calculated.

[0122] S54. Obtain the entropy value of the quality index for each process in the process set, and obtain the corresponding defect occurrence probability based on the entropy value of each quality index and the corresponding defect propagation probability;

[0123] S55. Sort the multiple defects in order of probability, and select the process corresponding to the highest defect probability as the root cause process.

[0124] As described in steps S51-S55 above, this invention obtains a defect type and selects multiple related processes from a knowledge graph based on the defect type to obtain a process set. By accurately identifying the defect type, multiple processes associated with the defect can be selected from the knowledge graph. This not only improves the depth of understanding of the defect but also accurately identifies potentially affected process groups, avoiding the problem of incomplete or missed traceability caused by traditional methods that only consider a single process. This enables comprehensive cross-process analysis, providing more comprehensive and accurate data support for subsequent steps. The cross-process association strength of each process is obtained through the process set. The cross-process association strength reveals the interrelationships and dependencies between different processes. The analysis of cross-process correlation strength data provides a quantitative basis for subsequent root cause tracing. By quantifying the cross-process correlation strength, it is possible to effectively identify which processes have close relationships, thereby helping to locate the possible propagation paths and sources of defects. Furthermore, based on the cross-process correlation strength, the corresponding shortest propagation path is derived from the knowledge graph. Traditional methods often rely on intuitive assumptions or single data points for analysis, while this step obtains the shortest propagation path through reverse inference. Deriving the shortest propagation path from the knowledge graph effectively reduces unnecessary path analysis, ensuring the efficiency and accuracy of the tracing process. It reduces computational complexity, avoids interference from irrelevant factors, and provides a clear defect propagation trajectory. The shortest propagation path is derived from the knowledge graph. By inferring the corresponding defect propagation probability and then reverse-engineering the probability of defect propagation for each path, the probability of defect propagation is quantified. This not only enables cross-process correlation analysis but also allows for more accurate prediction of the propagation probability of defects between processes. This method effectively compensates for the shortcomings of existing technologies that cannot accurately quantify the propagation process, thereby improving the accuracy of prediction and the ability to identify abnormal processes. By obtaining the entropy value of the quality index for each process in the process set and obtaining the corresponding defect occurrence probability based on the entropy value of each quality index and the corresponding defect propagation probability, the introduction of the quality index entropy value provides the system with a quantitative analysis of the quality fluctuations of each process. This allows for a more scientific assessment of the quality differences between different processes, combined with defect propagation... Using probability to calculate defect occurrence probability comprehensively considers the inherent quality characteristics of a process and its role in defect propagation, thus forming a comprehensive quality control framework. This effectively improves the accuracy of root cause analysis. By sorting multiple defect occurrence probabilities in order of magnitude and selecting the process with the highest probability as the root cause process, this sorting and selection step is a crucial step in prioritizing all possible root causes. It effectively identifies the most likely root cause process, helping relevant personnel quickly focus on the process that may be causing the problem. This avoids redundant analysis and over-speculation that may exist in traditional methods, improving problem-solving efficiency. Furthermore, it utilizes quantified cross-process correlation strength and the shortest propagation path.It can complete defect propagation analysis between multiple processes in a short time. By combining quality index entropy values ​​and defect propagation probabilities, it can perform high-precision defect prediction in multivariate environments, avoiding blind analysis in traditional methods. Process sequencing and root cause selection make the traceability process simpler, more intuitive, easier to operate and apply, and suitable for a variety of complex manufacturing processes.

[0125] like Figure 2 As shown, this application also provides a cross-process defect root cause tracing system, including:

[0126] The first acquisition module is used to acquire the full process chain data of the target product, and to acquire the production quality characteristic parameters of each process based on the full process chain data, wherein the production quality characteristic parameters include process parameter information, equipment status information and quality inspection information;

[0127] The second acquisition module is used to acquire the process data feature matrix of the corresponding process based on each of the process parameter information, equipment status information and quality inspection information;

[0128] The third acquisition module is used to obtain the corresponding process attenuation factor based on the process data feature matrix of any two processes, and to obtain the cross-process correlation strength based on the process attenuation factor.

[0129] The construction module is used to construct a process relationship diagram based on each process and the cross-process correlation strength, and to obtain defect propagation data based on the process relationship diagram, wherein the defect propagation data includes the shortest propagation path, defect propagation probability, and defect type;

[0130] The traceability module is used to construct a knowledge graph based on the full process chain data, the strength of multiple cross-process associations, the shortest propagation path, the probability of defect propagation, and the defect type, and to perform cross-process root cause tracing based on the knowledge graph.

[0131] In one embodiment, the building module includes:

[0132] The selection unit is used to obtain all cross-process association strengths for each process and select the largest cross-process association strength as the dependency association strength.

[0133] The first determining unit is used to determine the most dependent process based on the strength of each dependency association and the corresponding process, thereby obtaining the process dependency relationship;

[0134] The building unit is used to construct a process relationship graph by taking each process as a node, the process dependency as an edge, and the corresponding cross-process association strength as the edge weight.

[0135] The second determining unit is used to obtain all propagation paths for each process based on the edges of the process relationship diagram, and to determine the shortest propagation path based on all propagation paths.

[0136] The first acquisition unit is used to acquire the historical defect probability of each process and acquire the propagation probability based on the historical defect probability and the shortest propagation path.

[0137] The second acquisition unit is used to acquire the defect propagation probability based on the propagation probability and the historical defect probability.

[0138] It should be noted that each module and unit in the cross-process defect root cause tracing system corresponds one-to-one with the steps in the cross-process defect root cause tracing method.

[0139] like Figure 3 As shown, this application also provides a computer device, which can be a server, and its internal structure can be as follows: Figure 3 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores all data required for the cross-process defect root cause tracing method. The network interface communicates with external terminals via a network connection. The computer program is executed by the processor to implement the cross-process defect root cause tracing method.

[0140] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment on which the present application is applied.

[0141] An embodiment of this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-described cross-process defect root cause tracing methods.

[0142] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media provided in this application and in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0143] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0144] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for tracing the root causes of defects across production processes, characterized in that, include: Acquire the full process chain data of the target product, and obtain the production quality characteristic parameters of each process based on the full process chain data, wherein the production quality characteristic parameters include process parameter information, equipment status information and quality inspection information; Based on each of the aforementioned process parameter information, equipment status information, and quality inspection information, obtain the corresponding process data feature matrix; The corresponding process attenuation factor is obtained based on the process data feature matrix of any two processes, including: Obtain a difference matrix based on the difference values ​​of corresponding features in the process data feature matrices of any two processes, and obtain the probability of each difference value based on the difference matrix; The frequency of each difference value is obtained based on the probability of multiple difference values, and the entropy value of the quality index is obtained based on the frequency of multiple difference values. The calculation formula is as follows: ； Where Z(BS) represents the entropy value of the quality index, and C(PL) i This represents the frequency of the i-th difference value, where i represents the index of the frequency of the difference value, and N represents the number of frequencies of the difference value. The mean and maximum difference values ​​are obtained from the difference matrix, and the equipment health index is obtained from the mean and maximum difference values. The process influence feature matrix is ​​obtained by taking the partial derivatives of the corresponding features in the process data feature matrices of any two processes, and the process complexity is obtained based on the process influence feature matrix. The process attenuation factor is calculated based on the process complexity, equipment health index, and quality index entropy value, wherein the calculation formula is: ； Wherein, G(SJ) represents the process attenuation factor, Z(BS) represents the quality index entropy value, G(FZ) represents the process complexity, S(JZ) represents the equipment health index, and the cross-process correlation strength between any two processes is obtained based on the process attenuation factor. A process relationship diagram is constructed based on the full process chain data and cross-process correlation strength of each process, and defect propagation data is obtained based on the process relationship diagram. The defect propagation data includes the shortest propagation path, defect propagation probability, and defect type. A knowledge graph is constructed based on the full process chain data, the strength of multiple cross-process associations, the shortest propagation path, the probability of defect propagation, and the defect type. Cross-process root cause tracing is then performed based on the knowledge graph.

2. The cross-process defect root cause tracing method according to claim 1, characterized in that, The step of obtaining the process data feature matrix of the corresponding process based on each of the process parameter information, equipment status information, and quality inspection information includes: Based on the process parameter information, obtain multiple process temperature data, process pressure data, and process speed data, and obtain the corresponding temperature skewness coefficient, pressure skewness coefficient, and speed skewness coefficient for each of the process temperature data, process pressure data, and process speed data; Obtain the corresponding process stability feature vector based on each of the temperature skewness coefficient, pressure skewness coefficient, and velocity skewness coefficient; Based on the equipment status information, obtain multiple equipment efficiency data and equipment load data, and obtain the corresponding efficiency skewness coefficient and load skewness coefficient based on each of the equipment efficiency data and equipment load data; Obtain the corresponding device stability feature vector based on each of the efficiency skewness coefficients and load skewness coefficients; Based on the quality inspection information, multiple dimensional deviation data, surface quality data, and defect area data are obtained, and corresponding dimensional skewness coefficients, quality skewness coefficients, and defect skewness coefficients are obtained based on each of the dimensional deviation data, surface quality data, and defect area data. Obtain the corresponding quality stability feature vector based on each of the aforementioned size skewness coefficient, mass skewness coefficient, and defect skewness coefficient; Normalize each of the quality stability feature vector, equipment stability feature vector, and process stability feature vector to obtain the corresponding standard quality stability feature vector, standard equipment stability feature vector, and standard process stability feature vector. Construct a process data feature matrix for the corresponding process based on multiple standard quality stability feature vectors, standard equipment stability feature vectors, and standard process stability feature vectors.

3. The cross-process defect root cause tracing method according to claim 1, characterized in that, The step of obtaining the cross-process correlation strength based on the process attenuation factor includes: The first installation layout data and parameter fluctuation amount are obtained based on the process data feature matrix of the first process, and the quality fluctuation amount corresponding to the parameter fluctuation amount is obtained based on the process data feature matrix of the second process. Based on the quality fluctuation and parameter fluctuation, obtain the parameter influence coefficients for any two processes. The installation location data and layout data of the first product are obtained based on the first installation layout data. The second installation layout data is obtained based on the process data feature matrix of the second process, and the second product installation location data and the second product layout data are obtained based on the second installation layout data. The installation location correlation coefficient is obtained based on the first product installation location data and the second product installation location data; the product layout correlation coefficient is obtained based on the first product layout data and the second product layout data. Obtain the process attenuation factor, and calculate the cross-process correlation strength based on the process attenuation factor, parameter influence coefficient, product layout correlation coefficient, and installation location correlation coefficient. The calculation formula is as follows: K(GQ)=D(SY)*e -G(SJ)*[1-B(JX)*A(WX)] ; Wherein, K(GQ) represents the cross-process correlation strength, G(SJ) represents the process attenuation factor, D(SY) represents the parameter influence coefficient, B(JX) represents the product layout correlation coefficient, and A(WX) represents the installation location correlation coefficient.

4. The cross-process defect root cause tracing method according to claim 1, characterized in that, The step of constructing a process relationship diagram based on the full process chain data and cross-process correlation strength of each process, and obtaining defect propagation data based on the process relationship diagram, includes: Obtain all cross-process association strengths for each process and select the largest cross-process association strength as the dependency association strength; The most dependent process is determined based on the strength of each dependency association and the corresponding process, thus obtaining the process dependency relationship; Each process is treated as a node, the process dependency is treated as an edge, and the corresponding cross-process association strength is treated as the edge weight to construct a process relationship graph. Obtain all propagation paths for each process based on the edges of the process relationship graph, and determine the shortest propagation path based on all propagation paths; Obtain the historical defect probability and historical defect type for each process, and obtain the propagation probability based on the historical defect probability and the shortest propagation path; The defect propagation probability is obtained based on the propagation probability and the historical defect probability.

5. The cross-process defect root cause tracing method according to claim 1, characterized in that, The step of performing cross-process root cause tracing based on the knowledge graph includes: Obtain the defect type, and select multiple related processes from the knowledge graph based on the defect type to obtain a process set; The cross-process association strength of each process is obtained based on the process set, and the corresponding shortest propagation path is obtained from the knowledge graph based on each cross-process association strength. The corresponding defect propagation probability is obtained based on each of the shortest propagation paths; Obtain the entropy value of the quality index for each process in the process set, and obtain the probability of occurrence of the corresponding defect based on the entropy value of each quality index and the corresponding defect propagation probability; The occurrence probabilities of the multiple defects are sorted in order of magnitude, and the process corresponding to the highest defect occurrence probability is selected as the root cause process.

6. A cross-process defect root cause tracing system, characterized in that, include: The first acquisition module is used to acquire the full process chain data of the target product, and to acquire the production quality characteristic parameters of each process based on the full process chain data, wherein the production quality characteristic parameters include process parameter information, equipment status information and quality inspection information; The second acquisition module is used to acquire the process data feature matrix of the corresponding process based on each of the process parameter information, equipment status information and quality inspection information; The third acquisition module is used to obtain the corresponding process attenuation factor based on the process data feature matrix of any two processes, including: Obtain a difference matrix based on the difference values ​​of corresponding features in the process data feature matrices of any two processes, and obtain the probability of each difference value based on the difference matrix; The frequency of each difference value is obtained based on the probability of multiple difference values, and the entropy value of the quality index is obtained based on the frequency of multiple difference values. The calculation formula is as follows: ； Where Z(BS) represents the entropy value of the quality index, and C(PL) i This represents the frequency of the i-th difference value, where i represents the index of the frequency of the difference value, and N represents the number of frequencies of the difference value. The mean and maximum difference values ​​are obtained from the difference matrix, and the equipment health index is obtained from the mean and maximum difference values. The process influence feature matrix is ​​obtained by taking the partial derivatives of the corresponding features in the process data feature matrices of any two processes, and the process complexity is obtained based on the process influence feature matrix. The process attenuation factor is calculated based on the process complexity, equipment health index, and quality index entropy value, wherein the calculation formula is: ； Wherein, G(SJ) represents the process attenuation factor, Z(BS) represents the quality index entropy value, G(FZ) represents the process complexity, S(JZ) represents the equipment health index, and the cross-process correlation strength between any two processes is obtained based on the process attenuation factor. The construction module is used to construct a process relationship diagram based on each process and the cross-process correlation strength, and to obtain defect propagation data based on the process relationship diagram, wherein the defect propagation data includes the shortest propagation path, defect propagation probability, and defect type; The traceability module is used to construct a knowledge graph based on the full process chain data, the strength of multiple cross-process associations, the shortest propagation path, the probability of defect propagation, and the defect type, and to perform cross-process root cause tracing based on the knowledge graph.

7. The cross-process defect root cause tracing system according to claim 6, characterized in that, The building module includes: The selection unit is used to obtain all cross-process association strengths for each process and select the largest cross-process association strength as the dependency association strength. The first determining unit is used to determine the most dependent process based on the strength of each dependency association and the corresponding process, thereby obtaining the process dependency relationship; The building unit is used to construct a process relationship graph by taking each process as a node, the process dependency as an edge, and the corresponding cross-process association strength as the edge weight. The second determining unit is used to obtain all propagation paths for each process based on the edges of the process relationship diagram, and to determine the shortest propagation path based on all propagation paths. The first acquisition unit is used to acquire the historical defect probability of each process and acquire the propagation probability based on the historical defect probability and the shortest propagation path. The second acquisition unit is used to acquire the defect propagation probability based on the propagation probability and the historical defect probability.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images