PCB production process management system and method based on intelligent tracing

Abstract

The invention discloses a PCB production process management system and method based on intelligent tracing, and relates to the technical field of intelligent tracing. Production links and a parallel line of each link are obtained in a process line by using production characteristics; constructing a flow line framework of a PCB generation process; the correlation degree between the production data in the adjacent production links is calculated, the correlation degree of different production data in all the adjacent production links is judged, and the production data with the maximum correlation degree serves as transmission data of the corresponding production link; automatically learning a Bayesian network structure through maximum likelihood estimation to obtain a potential causal path of each piece of transmission data; generating a chained quality evaluation model; and after the real-time PCB quality is judged to be unqualified by utilizing a qualified quality threshold, performing reverse tracing by utilizing a potential causal path of each production link in the chain type quality evaluation model.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent tracing, and in particular to a PCB board production process management system and method based on intelligent tracing. Background Art

[0002] The widespread use of electronic products in modern society has led to increasing requirements for their quality and reliability. As the core component of electronic products, the quality of PCB boards directly affects product performance. In order to ensure the quality of electronic products, it is necessary to strictly monitor and trace the production process of PCB boards so that problems can be quickly located and solved when they occur; with the development of manufacturing industry towards intelligence and digitalization, and the deepening of concepts such as intelligent manufacturing, enterprises need to use advanced technologies to realize the visualization, controllability and traceability of the production process. Intelligent traceability technology has become a key means for PCB board manufacturers to improve production efficiency, quality control level and corporate competitiveness. Entering the 21st century, with the rapid development of information technology, especially the rise of technologies such as the Internet, the Internet of Things, and big data, the intelligent traceability technology of PCB board production process has developed rapidly. Enterprises began to introduce systems such as MES and ERP to realize the information management of the production process and the integrated sharing of data, and can conduct a more comprehensive traceability of the entire production process of PCB boards from raw materials to finished products. However, when tracing abnormal multi-layer PCB boards, due to the numerous production links, there is no information exchange and communication between some production links in the PCB board production process, which greatly delays the tracing time when tracing unqualified PCB boards. Summary of the invention

[0003] The purpose of the present invention is to provide a PCB board production process management system and method based on intelligent tracing to solve the problems raised in the prior art.

[0004] To achieve the above object, the present invention provides the following technical solutions: A PCB production process management method based on intelligent tracing, the method comprising the following steps: S100, collect all process lines in the PCB board production process, analyze the production data in all process lines, and extract production characteristics; use the production characteristics to obtain the production links and parallel lines of each link in the process line; build a process line framework for the PCB generation process; Furthermore, the specific steps of constructing the process line framework of the PCB generation process are: S101. Collect all process lines in the PCB board production process, collect all production data in each process line, extract the data value of each production data at the beginning and end of the corresponding production line; calculate the change value of all production data in the corresponding production line, the formula is: ; In the formula, Sc represents the change value of each production data in the calculated production line, S after represents the data value of each extracted production data at the beginning of the corresponding production line, S before Indicates the data value of each extracted production data at the end of the corresponding production line; After calculating the change values ​​of all production data in the production line, extract the production data corresponding to the maximum change value as the production feature of the production line; calculate the production features of all production lines in the PCB production process in turn; use the production features to judge all production lines, when the production features are the same, judge the corresponding production lines as parallel lines, when the production features are different, judge the corresponding production lines as different production links; S102. After all the production links and the parallel lines of each production link in the PCB production process are determined, the time series of each production link in the PCB production process is extracted; the extracted time series is sorted using the bubble sort method, and the production links are sorted using the time sorting result to obtain the PCB production link chain; the parallel lines in each production link and the production link chain are used to construct the process line framework of the PCB generation process.

[0005] By building a process line framework, the resources required for each link can be clearly displayed, and users can accurately allocate manpower, material resources and equipment accordingly; combing the process line can accurately locate repeated and redundant operations. The framework sets standard specifications for each process, and operators operate according to the process to reduce efficiency fluctuations caused by differences in personal habits. New employees can also use the framework to quickly get started, reduce training costs and adaptation time, and ensure the overall production rhythm is stable. The framework also reflects the speed of material circulation and assists in setting a reasonable inventory level. For parts with fast turnover, maintain a low inventory and high-frequency replenishment strategy; for parts with slow turnover, accurately adjust the purchase volume to avoid backlogs of funds and reduce storage costs.

[0006] S200, after generating the process line framework of the PCB production process, extract the production data in each production link, calculate the correlation between each production data in adjacent production links, determine the correlation between different production data in all adjacent production links, and use the production data with the largest correlation as the transfer data of the corresponding production link; Furthermore, the specific steps of using the production data with the greatest correlation as the transfer data of the corresponding production link are: S201. After generating the process line framework of the PCB production process, collect the records of the historical PCB production process, extract the production data in each production link, set the number of each production data extracted in each production link as N, draw a curve graph using the production data of adjacent production links, use the linear regression algorithm to fit the two production data of adjacent production links in the curve graph, and calculate the linear regression equation of the two production data as follows: , where Sa represents the production data of the next production link in the adjacent production links, and Sb represents the production data of the previous production link in the adjacent production links; a represents the slope of the calculated linear regression equation, and b represents the intercept of the calculated linear regression equation; In a pair of adjacent production links, the linear regression equation of every two production data is calculated. Assuming that the number of production data types in each production link is m, the number of linear regression equations calculated in a pair of adjacent production links is m. 2 , the linear regression equation of production data of every two adjacent production links in the process line framework is calculated; S202, extract the slope of each linear regression equation as the correlation between the two production data, and compare m in every two adjacent production links. 2 The two production data corresponding to the maximum correlation are selected as the transfer data of adjacent production links, and the transfer data Sd of all adjacent production links are obtained in the process line framework.

[0007] S300, collect the transmission data of each production link when anomalies occur in the historical PCB board production process, use the transmission data of adjacent links as nodes to build a Bayesian network, automatically learn the Bayesian network structure through maximum likelihood estimation, and obtain the potential causal path of each transmission data; and mark it in the process line framework; Furthermore, the specific steps to obtain the potential causal path of each transmission data are: S301, collect the transmission data of each production link when an abnormality occurs in the historical PCB board production process, set the transmission data of the previous production link in the adjacent production links as Sdb, and the transmission data of the next production link as Sda; collect two data values ​​in the historical PCB production process to form sets A and B, where A represents the transmission data set of the next production link, and B represents the transmission data set of the previous production link; initially set an undirected graph, use sets A and B as nodes of the undirected graph, perform deletion, addition and steering operations on each edge in the undirected graph, calculate the network structure score of the undirected graph after each operation when performing the operation, and construct the calculation model formula as follows: ; In the formula, BIC represents the network structure score of the undirected graph after each operation, n represents the number of transfer data in the collected transfer data set, k represents the number of model parameters, Represents the maximum likelihood estimate of the model; by calculating the network structure score, select the optimal network structure of the undirected graph when the network structure score is the largest; when there is an edge from B to A in the optimal network structure, determine that there is a causal flow from B to A between the transmitted data B and the transmitted data A; S302. After determining that there is a causal flow between the transfer data B and the transfer data A, extract the linear regression equation of the transfer data B and A in S200 as the causal relationship, combine the causal relationship with the edges of the transfer data B and the transfer data A in the optimal network structure to form a potential causal path, and mark the generated potential causal path in the process line framework; repeatedly calculate and mark the potential causal paths of all adjacent production links in the process line framework.

[0008] S400, in adjacent production links in the process line framework, using the transfer data output by the previous production link as the input of the next production link, combined with the marked potential causal path, to generate a chain quality assessment model; using the chain quality assessment model to calculate the quality of the PCB board produced in real time; Furthermore, the specific steps of calculating the quality of the PCB board produced in real time using the chain quality assessment model are as follows: S401, connect the undirected graphs of every two adjacent production links according to the process line framework, construct the transfer data output by the previous production link as the input of the next production link, combine the marked potential causal path, and use the Bayesian algorithm to build a chain quality assessment model; collect production data from the historical PCB production process to build a training set, and input the training set into the chain quality assessment model to train the model; S402. When real-time PCB production is carried out, the transfer data in the first production link in the process line framework is input into the chain quality assessment model, and the chain quality assessment model is used for calculation until the transfer data of the last production link is calculated, and the transfer data of the last production link is used as the quality Z of the produced PCB.

[0009] Since the model follows the chain structure of the production process, once a product has a quality problem, the chain quality assessment model can accurately locate the specific production process and dig deep into the root cause of the defect. During production, the chain model continuously receives real-time data from each process and gives the quality assessment results immediately. If it is monitored that a process parameter begins to deviate from the normal range, and based on past rules, there is a high probability that the subsequent process will have quality problems, an early warning will be issued in time to reserve enough time for the company to respond. By analyzing the contribution of each process in the chain model to the final quality and the mutual influence relationship, the bottleneck link in the production process can be accurately found.

[0010] S500, when producing PCB boards in real time, record the transfer data of each production link, draw an internal code on each layer of the final PCB board, and the internal code records the transfer data and quality of all production links of the corresponding layer of PCB board; then draw an external code outside the entire PCB board, and the external code contains the location information of each internal code; Furthermore, the specific steps for setting the inner code and the outer code are: S501, when producing multi-layer PCB boards in real time, record the transfer data of each production link, draw an internal code on each layer of the final PCB board, and the internal code of each layer contains the transfer data of all production links of producing the corresponding layer of PCB boards; S502, after drawing the inner code in each layer of the multi-layer PCB board, extract the position of each inner code, draw the outer code outside the multi-layer PCB board, and the outer code is formed by drilling holes on the panel boards of the top and bottom layers of the multi-layer PCB board; the outer code contains the position information of each inner code.

[0011] S600, the qualified quality threshold is calculated using the PCB board data in the normal production process of the historical PCB. The user scans the external code and the internal code to obtain the real-time PCB board quality. After the qualified quality threshold is used to judge whether the real-time PCB board quality is unqualified, the potential causal path of each production link in the chain quality assessment model is used for reverse tracing.

[0012] Furthermore, the specific steps of using the potential causal path of each production link in the chain quality assessment model for reverse tracing are as follows: S601, collect the transfer data of the last production link when the historical PCB board is produced qualified, set the number of collected historical PCB production qualified transfer data as G, calculate the average value and standard deviation of the collected G transfer data, and use the average value and standard deviation to calculate the qualified quality threshold when the PCB is produced qualified, the formula is: ; In the formula, In represents the calculated qualified quality threshold of PCB production, Fp represents the calculated average value of G transfer data, and Fst represents the calculated standard deviation of G transfer data; S602. The user obtains the quality of the multi - layer PCB board by scanning the code, and uses the qualified quality threshold to judge the quality during real - time PCB production. When Z≥In, it is judged that the quality of the PCB board produced in real - time is qualified; when Z<In, it is judged that the quality of the PCB board produced in real - time is unqualified. When judging the PCB board produced in real - time, first use the outer code of the multi - layer PCB to locate the position of the single - layer PCB board with unqualified quality, and then scan the inner code of the single - layer PCB board with unqualified quality. Use the potential causal path in the chain - type quality evaluation model for reverse tracing, compare the transfer data of each production link with the transfer data of the qualified PCB board, and find out the production link with anomalies.

[0013] The inner and outer code systems give the PCB board a fine identification from the outer layer to the inner layer. When anomalies occur, the outer code can be used to quickly access the tracing system, and then with the help of the inner code, it can be accurately located to the specific layer or module, avoiding large - scale investigations and saving time and labor costs. Then it is combined with the chain - type quality evaluation model constructed according to the sequence of the production process and the causal relationship between processes. When anomalies are found, it can trace back reversely along the "chain" and accurately identify the upstream process that causes the problem and locate the root cause of the failure.

[0014] A PCB board production process management system based on intelligent tracing. The PCB board production process management system includes a data collection module, a process line framework construction module, a transfer data search module, a potential causal path search module, a chain - type quality evaluation model construction module, a quality judgment module, and a tracing module; The data collection module is used to collect production data during the historical PCB production process; The process line framework construction module is used to analyze the production data in all process lines and construct a process line framework; The transfer data search module is used to extract the production data in each production link, calculate the correlation degree between each production data in adjacent production links, and take the production data with the largest correlation degree as the transfer data of the corresponding production link; The potential causal path search module is used to construct a Bayesian network with the transfer data of adjacent links as nodes, automatically learn the Bayesian network structure through maximum likelihood estimation, and obtain the potential causal path of each transfer data; The chain - type quality evaluation model construction module is used to take the transfer data output by the previous production link as the input of the next production link, and combine the marked potential causal path to generate a chain - type quality evaluation model; The quality judgment module is used to calculate the qualified quality threshold using the transfer data of the last link of the historical production PCB board, and use the qualified quality threshold to judge the quality of the real - time PCB board produced; The traceability module is used to trace abnormal production links by scanning codes and chain quality assessment models when the real-time PCB board quality is judged to be unqualified.

[0015] The potential causal path search module includes an optimal network structure unit and a potential causal path unit; The optimal network structure unit is used to calculate the network structure score when operating the set undirected graph, and select the network structure corresponding to the maximum network structure score as the optimal network structure; The potential causal path unit is used to utilize the linear regression equation of the transfer data of adjacent production links as the causal relationship and generate the potential causal path in combination with the optimal network structure.

[0016] The traceability module includes a code scanning traceability unit and a model traceability unit; The code scanning and tracing unit is used to determine the location of the single-layer PCB board with unqualified quality after the user scans the internal and external codes; The model tracing unit is used to trace the abnormal production links by using the chain quality assessment model after the location of the single-layer PCB board with unqualified quality is determined.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention can clearly display the resources required for each link by constructing a process line framework, so that users can accurately allocate manpower, material resources and equipment; combing the process lines can accurately locate repeated and redundant operations. The framework sets standard specifications for each process, and operators operate according to the process to reduce efficiency fluctuations caused by differences in personal habits.

[0018] 2. The present invention implements internal and external codes and models for tracing. The internal and external code system gives the PCB board a fine mark from the outer layer to the inner layer. When an abnormality occurs, the external code can be used to quickly access the traceability system, and then the internal code can be used to accurately locate the specific level or module, avoiding large-scale investigations and saving time and labor costs; then the chain quality assessment model is combined with the sequence of the production process and the causal relationship between the processes. When an abnormality is found, it can be traced back along the "chain", accurately identify the upstream process that caused the problem, and locate the root cause of the fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a module distribution diagram of a PCB board production process management system based on intelligent tracing according to the present invention; Figure 2 The present invention is a schematic diagram of the steps of a PCB board production process management method based on intelligent traceability. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution. A PCB production process management method based on intelligent tracing, the method comprising the following steps: S100, collect all process lines in the PCB board production process, analyze the production data in all process lines, and extract production characteristics; use the production characteristics to obtain the production links and parallel lines of each link in the process line; build a process line framework for the PCB generation process; The specific steps to build the process line framework of the PCB generation process are: S101. Collect all process lines in the PCB board production process, collect all production data in each process line, extract the data value of each production data at the beginning and end of the corresponding production line; calculate the change value of all production data in the corresponding production line, the formula is: ; In the formula, Sc represents the change value of each production data in the calculated production line, S after represents the data value of each extracted production data at the beginning of the corresponding production line, S before Indicates the data value of each extracted production data at the end of the corresponding production line; After calculating the change values ​​of all production data in the production line, extract the production data corresponding to the maximum change value as the production feature of the production line; calculate the production features of all production lines in the PCB production process in turn; use the production features to judge all production lines, when the production features are the same, judge the corresponding production lines as parallel lines, when the production features are different, judge the corresponding production lines as different production links; S102. After all the production links and the parallel lines of each production link in the PCB production process are determined, the time series of each production link in the PCB production process is extracted; the extracted time series is sorted using the bubble sort method, and the production links are sorted using the time sorting result to obtain the PCB production link chain; the parallel lines in each production link and the production link chain are used to construct the process line framework of the PCB generation process.

[0022] By building a process line framework, the resources required for each link can be clearly displayed, and users can accurately allocate manpower, material resources and equipment accordingly; combing the process line can accurately locate repeated and redundant operations. The framework sets standard specifications for each process, and operators operate according to the process to reduce efficiency fluctuations caused by differences in personal habits. New employees can also use the framework to quickly get started, reduce training costs and adaptation time, and ensure the overall production rhythm is stable. The framework also reflects the speed of material circulation and assists in setting a reasonable inventory level. For parts with fast turnover, maintain a low inventory and high-frequency replenishment strategy; for parts with slow turnover, accurately adjust the purchase volume to avoid backlogs of funds and reduce storage costs.

[0023] S200, after generating the process line framework of the PCB production process, extract the production data in each production link, calculate the correlation between each production data in adjacent production links, determine the correlation between different production data in all adjacent production links, and use the production data with the largest correlation as the transfer data of the corresponding production link; The specific steps of using the production data with the greatest correlation as the transfer data of the corresponding production link are: S201. After generating the process line framework of the PCB production process, collect the records of the historical PCB production process, extract the production data in each production link, set the number of each production data extracted in each production link as N, draw a curve graph using the production data of adjacent production links, use the linear regression algorithm to fit the two production data of adjacent production links in the curve graph, and calculate the linear regression equation of the two production data as follows: , where Sa represents the production data of the next production link in the adjacent production links, and Sb represents the production data of the previous production link in the adjacent production links; a represents the slope of the calculated linear regression equation, and b represents the intercept of the calculated linear regression equation; In a pair of adjacent production links, the linear regression equation of every two production data is calculated. Assuming that the number of production data types in each production link is m, the number of linear regression equations calculated in a pair of adjacent production links is m. 2 , the linear regression equation of production data of every two adjacent production links in the process line framework is calculated; S202, extract the slope of each linear regression equation as the correlation between the two production data, and compare m in every two adjacent production links. 2 The two production data corresponding to the maximum correlation are selected as the transfer data of adjacent production links, and the transfer data Sd of all adjacent production links are obtained in the process line framework.

[0024] S300, collect the transmission data of each production link when anomalies occur in the historical PCB board production process, use the transmission data of adjacent links as nodes to build a Bayesian network, automatically learn the Bayesian network structure through maximum likelihood estimation, and obtain the potential causal path of each transmission data; and mark it in the process line framework; The specific steps to obtain the potential causal path for each transmission data are: S301, collect the transmission data of each production link when an abnormality occurs in the historical PCB board production process, set the transmission data of the previous production link in the adjacent production links as Sdb, and the transmission data of the next production link as Sda; collect two data values ​​in the historical PCB production process to form sets A and B, where A represents the transmission data set of the next production link, and B represents the transmission data set of the previous production link; initially set an undirected graph, use sets A and B as nodes of the undirected graph, perform deletion, addition and steering operations on each edge in the undirected graph, calculate the network structure score of the undirected graph after each operation when performing the operation, and construct the calculation model formula as follows: ; In the formula, BIC represents the network structure score of the undirected graph after each operation, n represents the number of transfer data in the collected transfer data set, k represents the number of model parameters, Represents the maximum likelihood estimate of the model; by calculating the network structure score, select the optimal network structure of the undirected graph when the network structure score is the largest; when there is an edge from B to A in the optimal network structure, determine that there is a causal flow from B to A between the transmitted data B and the transmitted data A; S302. After determining that there is a causal flow between the transfer data B and the transfer data A, extract the linear regression equation of the transfer data B and A in S200 as the causal relationship, combine the causal relationship with the edges of the transfer data B and the transfer data A in the optimal network structure to form a potential causal path, and mark the generated potential causal path in the process line framework; repeatedly calculate and mark the potential causal paths of all adjacent production links in the process line framework.

[0025] S400, in adjacent production links in the process line framework, using the transfer data output by the previous production link as the input of the next production link, combined with the marked potential causal path, to generate a chain quality assessment model; using the chain quality assessment model to calculate the quality of the PCB board produced in real time; The specific steps of calculating the quality of PCB boards produced in real time using the chain quality assessment model are as follows: S401, connect the undirected graphs of every two adjacent production links according to the process line framework, construct the transfer data output by the previous production link as the input of the next production link, combine the marked potential causal path, and use the Bayesian algorithm to build a chain quality assessment model; collect production data from the historical PCB production process to build a training set, and input the training set into the chain quality assessment model to train the model; S402. When real-time PCB production is carried out, the transfer data in the first production link in the process line framework is input into the chain quality assessment model, and the chain quality assessment model is used for calculation until the transfer data of the last production link is calculated, and the transfer data of the last production link is used as the quality Z of the produced PCB.

[0026] Since the model follows the chain structure of the production process, once a product has a quality problem, the chain quality assessment model can accurately locate the specific production process and dig deep into the root cause of the defect. During production, the chain model continuously receives real-time data from each process and gives the quality assessment results immediately. If it is monitored that a process parameter begins to deviate from the normal range, and based on past rules, there is a high probability that the subsequent process will have quality problems, an early warning will be issued in time to reserve enough time for the company to respond. By analyzing the contribution of each process in the chain model to the final quality and the mutual influence relationship, the bottleneck link in the production process can be accurately found.

[0027] S500, when producing PCB boards in real time, record the transfer data of each production link, draw an internal code on each layer of the final PCB board, and the internal code records the transfer data and quality of all production links of the corresponding layer of PCB board; then draw an external code outside the entire PCB board, and the external code contains the location information of each internal code; The specific steps for setting internal and external codes are: S501, when producing multi-layer PCB boards in real time, record the transfer data of each production link, draw an internal code on each layer of the final PCB board, and the internal code of each layer contains the transfer data of all production links of producing the corresponding layer of PCB boards; S502, after drawing the inner code in each layer of the multi-layer PCB board, extract the position of each inner code, draw the outer code outside the multi-layer PCB board, and the outer code is formed by drilling holes on the panel boards of the top and bottom layers of the multi-layer PCB board; the outer code contains the position information of each inner code.

[0028] S600, the qualified quality threshold is calculated using the PCB board data in the normal production process of the historical PCB. The user scans the external code and the internal code to obtain the real-time PCB board quality. After the qualified quality threshold is used to judge whether the real-time PCB board quality is unqualified, the potential causal path of each production link in the chain quality assessment model is used for reverse tracing.

[0029] The specific steps for backward tracing using the potential causal paths of each production link in the chain - type quality assessment model are as follows: S601. Collect the transfer data of the last production link when the historical PCB board is produced qualified. Let the number of collected transfer data of qualified historical PCB production be G. Calculate the average value and standard deviation of the G collected transfer data, and calculate the qualified quality threshold when the PCB is produced qualified using the average value and standard deviation. The formula is: ; In the formula, In represents the calculated qualified quality threshold of PCB production, Fp represents the average value of the G calculated transfer data, and Fst represents the standard deviation of the G calculated transfer data; S602. The user obtains the quality of the multi - layer PCB board by scanning the code, and judges the quality during real - time PCB production using the qualified quality threshold. When Z≥In, it is judged that the quality of the PCB board produced in real - time is qualified; when Z < In, it is judged that the quality of the PCB board produced in real - time is unqualified. When judging the PCB board produced in real - time, first use the outer code of the multi - layer PCB to locate the position of the single - layer PCB board with unqualified quality, and then scan the inner code of the single - layer PCB board with unqualified quality. Use the potential causal path in the chain - type quality assessment model for backward tracing, compare the transfer data of each production link with the transfer data of the qualified PCB board, and find the production link with anomalies.

[0030] The inner and outer code systems assign fine identifiers to the PCB board from the outer layer to the inner layer. When anomalies occur, the outer code can be used to quickly access the traceability system, and then with the help of the inner code, it can be accurately located to the specific layer or module, avoiding large - scale investigations and saving time and labor costs; then it is constructed in combination with the chain - type quality assessment model according to the sequence of the production process and the causal relationship between processes. When anomalies are found, it can trace back reversely along the "chain" and accurately identify the upstream process that causes the problem and locate the root cause of the fault.

[0031] A PCB board production process management system based on intelligent tracing. The PCB board production process management system includes a data collection module, a process line framework construction module, a transfer data search module, a potential causal path search module, a chain - type quality assessment model construction module, a quality judgment module, and a traceability module; The data collection module is used to collect production data during historical PCB production; The process line framework construction module is used to analyze the production data in all process lines and construct a process line framework; The transfer data search module is used to extract the production data in each production link, calculate the correlation between each production data in adjacent production links, and use the production data with the largest correlation as the transfer data of the corresponding production link; The potential causal path search module is used to construct a Bayesian network using the transfer data of adjacent links as nodes, automatically learn the Bayesian network structure through maximum likelihood estimation, and obtain the potential causal path of each transfer data; The chain quality assessment model building module is used to use the transfer data output from the previous production link as the input of the next production link, combined with the marked potential causal path, to generate a chain quality assessment model; The quality judgment module is used to calculate the qualified quality threshold value by using the transmission data of the last link of the historical production of PCB boards, and judge the quality of the real-time PCB boards produced by using the qualified quality threshold value; The traceability module is used to trace abnormal production links by scanning codes and chain quality assessment models when the real-time PCB board quality is judged to be unqualified.

[0032] The potential causal path search module includes an optimal network structure unit and a potential causal path unit; The optimal network structure unit is used to calculate the network structure score when operating the set undirected graph, and select the network structure corresponding to the maximum network structure score as the optimal network structure; The potential causal path unit is used to utilize the linear regression equation of the transfer data of adjacent production links as the causal relationship and generate the potential causal path in combination with the optimal network structure.

[0033] The traceability module includes a code scanning traceability unit and a model traceability unit; The code scanning and tracing unit is used to determine the location of the single-layer PCB board with unqualified quality after the user scans the internal and external codes; The model tracing unit is used to trace the abnormal production links by using the chain quality assessment model after the location of the single-layer PCB board with unqualified quality is determined.

[0034] Embodiment: Now, a factory producing multi-layer PCB boards is judged for quality, and the total number of process routes collected is 6. Through analysis, the different production links are obtained as link 1-link 2-link 3; there are two parallel lines in link 1, and there are three parallel lines in link 2, and finally a process line framework is constructed; Extract the production data in each link, and calculate that the transfer data of link 1 is material purity, the transfer data of link 2 is etching angle, and the transfer data of link 3 is line impedance; The chain quality assessment model of two adjacent production links is calculated by using the Bayesian network algorithm as ZU=1.5×SH+4 and SH=2.35×CH+8.9; The internal code and external code are drawn for the multi-layer PCB board, and the quality judgment is performed on the 6-layer PCB board. It is found that the quality of the second layer is unqualified. The user determines the position of the second layer by scanning the external code, and then uses the chain quality assessment model to trace back and find the abnormality in link 2.

[0035] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A PCB production process management method based on intelligent traceability, characterized by: The method comprises the following steps: S100, collect all process lines in the PCB board production process, analyze the production data in all process lines, and extract production characteristics; use the production characteristics to obtain the production links and parallel lines of each link in the process line; build a process line framework for the PCB generation process; S200, after generating the process line framework of the PCB production process, extract the production data in each production link, calculate the correlation between each production data in adjacent production links, determine the correlation between different production data in all adjacent production links, and use the production data with the largest correlation as the transfer data of the corresponding production link; S300, collect the transmission data of each production link when anomalies occur in the historical PCB board production process, use the transmission data of adjacent links as nodes to build a Bayesian network, automatically learn the Bayesian network structure through maximum likelihood estimation, and obtain the potential causal path of each transmission data; and mark it in the process line framework; S400, in adjacent production links in the process line framework, using the transfer data output by the previous production link as the input of the next production link, combined with the marked potential causal path, to generate a chain quality assessment model; using the chain quality assessment model to calculate the quality of the PCB board produced in real time; S500, when producing PCB boards in real time, record the transfer data of each production link, draw an internal code on each layer of the final PCB board, and the internal code records the transfer data and quality of all production links of the corresponding layer of PCB board; then draw an external code outside the entire PCB board, and the external code contains the location information of each internal code; S600, the qualified quality threshold is calculated using the PCB board data in the normal production process of the historical PCB. The user scans the external code and the internal code to obtain the real-time PCB board quality. After the qualified quality threshold is used to judge whether the real-time PCB board quality is unqualified, the potential causal path of each production link in the chain quality assessment model is used for reverse tracing.

2. According to a PCB board production process management method based on intelligent tracing according to claim 1, it is characterized in that: The specific steps of constructing the process line framework of the PCB generation process in S100 are: S101, collecting all process lines in the PCB board production process, collecting all production data in each process line, and extracting the data value of each production data at the beginning and end of the corresponding production line; Calculate the change value of all production data in the corresponding production line. The formula is: ； In the formula, Sc represents the change value of each production data in the calculated production line, S after represents the data value of each extracted production data at the beginning of the corresponding production line, S before Indicates the data value of each extracted production data at the end of the corresponding production line; After calculating the change values ​​of all production data in the production line, the production data corresponding to the maximum change value is extracted as the production feature of the production line; Calculate the production characteristics of all production lines in the PCB production process in turn; use the production characteristics to judge all production lines. When the production characteristics are the same, the corresponding production lines are judged to be parallel lines; when the production characteristics are different, the corresponding production lines are judged to be different production links; S102. After all the production links and the parallel lines of each production link in the PCB production process are determined, the time series of each production link in the PCB production process is extracted; the extracted time series is sorted using the bubble sort method, and the production links are sorted using the time sorting result to obtain the PCB production link chain; the parallel lines in each production link and the production link chain are used to construct the process line framework of the PCB generation process.

3. The PCB board production process management method based on intelligent tracing according to claim 1 is characterized in that: The specific steps of using the production data with the greatest correlation as the transmission data of the corresponding production link in S200 are: S201. After generating the process line framework of the PCB production process, collect the records of the historical PCB production process, extract the production data in each production link, set the number of each production data extracted in each production link as N, draw a curve graph using the production data of adjacent production links, use the linear regression algorithm to fit the two production data of adjacent production links in the curve graph, and calculate the linear regression equation of the two production data as follows: , where Sa represents the production data of the next production link in the adjacent production links, and Sb represents the production data of the previous production link in the adjacent production links; a represents the slope of the calculated linear regression equation, and b represents the intercept of the calculated linear regression equation; In a pair of adjacent production links, the linear regression equation of every two production data is calculated. Assuming that the number of production data types in each production link is m, the number of linear regression equations calculated in a pair of adjacent production links is m. 2 , the linear regression equation of production data of every two adjacent production links in the process line framework is calculated; S202, extract the slope of each linear regression equation as the correlation between the two production data, and compare m in every two adjacent production links. 2 The two production data corresponding to the maximum correlation are selected as the transfer data of adjacent production links, and the transfer data Sd of all adjacent production links are obtained in the process line framework.

4. The PCB production process management method based on intelligent tracing according to claim 3 is characterized in that: The specific steps of obtaining the potential causal path of each transfer data in S300 are: S301, collect the transmission data of each production link when an abnormality occurs in the historical PCB board production process, set the transmission data of the previous production link in the adjacent production links as Sdb, and the transmission data of the next production link as Sda; collect two data values ​​in the historical PCB production process to form sets A and B, where A represents the transmission data set of the next production link, and B represents the transmission data set of the previous production link; initially set an undirected graph, use sets A and B as nodes of the undirected graph, perform deletion, addition and steering operations on each edge in the undirected graph, calculate the network structure score of the undirected graph after each operation when performing the operation, and construct the calculation model formula as follows: ； In the formula, BIC represents the network structure score of the undirected graph after each operation, n represents the number of transfer data in the collected transfer data set, k represents the number of model parameters, Represents the maximum likelihood estimate of the model; by calculating the network structure score, select the optimal network structure of the undirected graph when the network structure score is the largest; when there is an edge from B to A in the optimal network structure, determine that there is a causal flow from B to A between the transmitted data B and the transmitted data A; S302. After determining that there is a causal flow between the transfer data B and the transfer data A, extract the linear regression equation of the transfer data B and A in S200 as the causal relationship, combine the causal relationship with the edges of the transfer data B and the transfer data A in the optimal network structure to form a potential causal path, and mark the generated potential causal path in the process line framework; repeatedly calculate and mark the potential causal paths of all adjacent production links in the process line framework.

5. A PCB board production process management method based on intelligent tracing according to claim 4, characterized in that: The specific steps of calculating the quality of the PCB produced in real time by using the chain quality assessment model in S400 are: S401, connect the undirected graphs of every two adjacent production links according to the process line framework, construct the transfer data output by the previous production link as the input of the next production link, combine the marked potential causal path, and use the Bayesian algorithm to build a chain quality assessment model; collect production data from the historical PCB production process to build a training set, and input the training set into the chain quality assessment model to train the model; S402. When real-time PCB production is carried out, the transfer data in the first production link in the process line framework is input into the chain quality assessment model, and the chain quality assessment model is used for calculation until the transfer data of the last production link is calculated, and the transfer data of the last production link is used as the quality Z of the produced PCB.

6. The PCB production process management method based on intelligent tracing according to claim 1 is characterized in that: The specific steps for setting the inner code and the outer code in S500 are: S501, when producing multi-layer PCB boards in real time, record the transfer data of each production link, draw an internal code on each layer of the final PCB board, and the internal code of each layer contains the transfer data of all production links of producing the corresponding layer of PCB boards; S502. After drawing the internal codes in each layer of the multi-layer PCB board, extract the positions of each internal code, and draw the external codes outside the multi-layer PCB board. The external codes are formed by drilling holes on the Panel boards of the topmost and bottommost layers of the multi-layer PCB board; the external codes contain the position information of each internal code.

7. A PCB board production process management method based on intelligent tracing according to claim 6, characterized in that: The specific steps of using the potential causal paths of each production link in the chain quality assessment model for reverse tracing in S600 are as follows: S601. Collect the transfer data of the last production link when the historical PCB board is produced qualified. Let the number of collected historical PCB production qualified transfer data be G, calculate the average value and standard deviation of the G collected transfer data, and calculate the qualified quality threshold when the PCB is produced qualified using the average value and standard deviation. The formula is: ； In the formula, In represents the calculated qualified quality threshold of PCB production, Fp represents the average value of the G calculated transfer data, and Fst represents the standard deviation of the G calculated transfer data; S602. The user obtains the quality of the multi-layer PCB board by scanning the code, and uses the qualified quality threshold to judge the quality during real-time PCB production. When Z≥In, it is judged that the quality of the PCB board produced in real time is qualified; when Z<In, it is judged that the quality of the PCB board produced in real time is unqualified; when judging the PCB board produced in real time, first use the external code of the multi-layer PCB to locate the position of the single-layer PCB board with unqualified quality, and then scan the internal code of the single-layer PCB board with unqualified quality, and use the potential causal path in the chain quality assessment model for reverse tracing, compare the transfer data of each production link with the transfer data of the qualified PCB board, and find the production link with anomalies.

8. A PCB board production process management system based on intelligent traceability, characterized by: The PCB board production process management system includes a data collection module, a process line framework construction module, a transfer data search module, a potential causal path search module, a chain quality assessment model construction module, a quality judgment module, and a traceability module; The data collection module is used to collect production data during the historical PCB production process; The process line framework construction module is used to analyze the production data in all process lines and construct a process line framework; The transfer data search module is used to extract the production data in each production link, calculate the correlation degree between each production data in adjacent production links, and use the production data with the largest correlation degree as the transfer data of the corresponding production link; The potential causal path search module is used to construct a Bayesian network with the transfer data of adjacent links as nodes, automatically learn the Bayesian network structure through maximum likelihood estimation, and obtain the potential causal path of each transfer data; The chain quality assessment model construction module is used to use the transfer data output by the previous production link as the input of the next production link, and combine the marked potential causal path to generate a chain quality assessment model; The quality judgment module is used to calculate the qualified quality threshold using the transfer data of the last link of the historical production PCB board, and use the qualified quality threshold to judge the quality of the real-time PCB board produced; The traceability module is used to trace the abnormal production link by the user using code scanning and the chain quality assessment model when it is judged that the quality of the real-time PCB board is unqualified.

9. A PCB board production process management system based on intelligent traceability according to claim 8, characterized in that: The potential causal path search module includes an optimal network structure unit and a potential causal path unit; The optimal network structure unit is used to calculate the network structure score when operating the set undirected graph, and select the network structure corresponding to the maximum network structure score as the optimal network structure; The potential causal path unit is used to utilize the linear regression equation of the transfer data of adjacent production links as the causal relationship and generate the potential causal path in combination with the optimal network structure.

10. The PCB production process management system based on intelligent traceability according to claim 8, characterized in that: The traceability module includes a code scanning traceability unit and a model traceability unit; The code scanning and tracing unit is used to determine the location of the single-layer PCB board with unqualified quality after the user scans the internal and external codes; The model tracing unit is used to trace the abnormal production links by using the chain quality assessment model after the location of the single-layer PCB board with unqualified quality is determined.

Images