MES system production process intelligent optimization method and system based on artificial intelligence

Abstract

The invention relates to the technical field of power system fault diagnosis, and particularly discloses an MES system production flow intelligent optimization method and system based on artificial intelligence, and the method comprises the steps: collecting the quality inspection data of all artware in each production batch in real time through an MES system in a production process; when it is found through quality inspection that a large number of crafts in a certain production batch are unqualified, the production batch is marked as a defective production batch, in the same way, the unqualified crafts are marked as defective crafts, and the MES automatically triggers a quality alarm; gathering and integrating the defective crafts in the defective production batches, combining the defective crafts in pairs to obtain a spatial-temporal characteristic matrix of each pair of defective crafts, and identifying all abnormal production time points of the defective crafts through a clustering algorithm; and finally, the fault source of the defective handicraft produced in each abnormal production time point is reversely traced and positioned, and the fault source is fed back to the corresponding management personnel of the factory, so that the fault source tracing efficiency of the defective handicraft is effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system fault diagnosis, and relates to an intelligent optimization method and system for the production process of an MES system based on artificial intelligence. Background Art

[0002] In modern manufacturing, quality control is a key link to ensure product qualification rate and customer satisfaction. By using MES (Manufacturing Execution System) to collect quality inspection data of each handicraft in each production batch in real time during the production process and triggering a quality alarm when a large number of handicrafts are found to be unqualified, enterprises can significantly improve the efficiency and accuracy of quality management.

[0003] It is important and necessary to trace the defect problems of defective handicrafts based on MES (Manufacturing Execution System). As the core system connecting production plans and shop floor execution, the MES system can collect, record, and track key data in the production process in real time, providing data support and analysis basis for tracing defect problems. In modern manufacturing, product quality directly affects the market competitiveness of enterprises, and the appearance of defective handicrafts often leads to customer complaints, increased costs, and even damage to brand reputation. Therefore, quickly and accurately finding the root cause of defect problems and making improvements is an indispensable part of enterprise production management. Summary of the Invention

[0004] In view of the above problems existing in the prior art, the present invention provides an intelligent optimization method and system for the production process of an MES system based on artificial intelligence to solve the above technical problems.

[0005] To achieve the above and other purposes, the technical solutions adopted by the present invention are as follows: On the one hand, the present invention provides an intelligent optimization method for the production process of an MES system based on artificial intelligence, and the method includes the following steps: During the production process, use the MES system to collect quality inspection data of each handicraft in each production batch in real time; when quality inspection finds that there are a large number of unqualified handicrafts in a certain production batch, mark this production batch as a defective production batch, and similarly mark the unqualified handicrafts as defective handicrafts, and the MES automatically triggers a quality alarm; Aggregate and integrate each defective handicraft in each defective production batch, combine them pairwise to obtain a spatio-temporal feature matrix of each pair of defective handicrafts, and identify all abnormal production time points of the defective handicrafts through a clustering algorithm; Finally, trace back in reverse to locate the fault sources of the defective handicrafts produced at each abnormal production time point, and feedback the fault sources to the corresponding management personnel in the factory.

[0006] When quality inspection discovers that a large number of handicrafts in a certain production batch are unqualified, that production batch is recorded as a defective production batch. The specific operation logic is as follows: Extract the measured values of each size parameter corresponding to each handicraft in each production batch from the quality inspection data of each handicraft in each production batch , where a is the number of each production batch, b is the number of each handicraft within the production batch, b ranges from 1 to M, c is the number of each size parameter, and c ranges from 1 to N; Then calculate the size compliance coefficient of the handicrafts within each production batch , where M and N respectively represent the total number of handicrafts within the production batch and the total number of size parameters, represents the weight of the c-th size parameter, reflecting the importance of this size parameter to the overall quality, and satisfies ; is the single compliance degree of the b-th handicraft corresponding to the c-th size parameter within the a-th production batch, and its specific calculation formula is , is the target value of the handicrafts corresponding to the c-th size parameter within the a-th production batch, respectively represent the effective tolerance range, lower tolerance, and upper tolerance of the c-th size parameter; Select the corresponding unilateral tolerance according to the measured value direction of each size parameter corresponding to each handicraft within each production batch: If , then ; If , then ; Compare the size compliance coefficient of the handicrafts within each production batch with the set handicraft size compliance threshold coefficient. If the size compliance coefficient of the handicrafts within a certain production batch is less than the set handicraft size compliance threshold coefficient, then that production batch is recorded as a defective production batch.

[0007] The spatio-temporal feature matrix of each defective handicraft within each defective production batch is specifically shown as: Obtain the production time point, equipment number, and operator ID of each defective handicraft within each defective production batch from the MES system; And calculate the process deviation degree of each size parameter of each defective handicraft within each defective production batch , is the measured value of the c-th size parameter corresponding to the f-th defective handicraft within the d-th defective production batch, d is the number of each defective production batch, and f is the number of each defective handicraft within the defective production batch; are respectively the upper limit permission value and the target value of the handicrafts corresponding to the c-th size parameter within the d-th defective production batch; Precisely determine the production time points of each defective handicraft within each defective production batch to the millisecond level, display them in the form of time stamps, and perform One-Hot encoding on the equipment number and operator ID; Thus, the spatio-temporal feature matrix of each defective handicraft within each defective production batch is obtained , which respectively represent the time stamp display form of the production time point and the set of One-Hot encodings of equipment information, and respectively represent the One-Hot encoding forms of the equipment number and operator ID.

[0008] All abnormal production time points of defective handicrafts are identified through a clustering algorithm, including: Define the spatio-temporal feature composite distance for each pair of defective handicrafts , where α, β, and γ are collectively referred to as spatio-temporal process weight coefficients and satisfy α + β + γ = 1. α is specifically the set time dimension weight, reflecting the impact of production time point continuity on clustering; β is specifically the set space dimension weight; γ is specifically the set process dimension weight; j is the number of each pair of defective handicrafts; is the time distance for each pair of defective handicrafts, which is corrected using a Gaussian kernel function, , where exp(·) represents the exponential function with the natural constant e as the base, respectively represent the time stamps of the production time points corresponding to the first defective handicraft and the second defective handicraft within the j-th pair of defective handicrafts; σ is the set time decay factor, with a default value set to 3 times the duration of a single production beat, in ms; is the space distance for each pair of defective handicrafts, which is corrected and calculated using a mixture of the discrete feature jaccard similarity and the continuous coordinate Euclidean distance: , respectively represent the sets of One-Hot encodings of the equipment information corresponding to the first defective handicraft and the second defective handicraft within the j-th pair of defective handicrafts; respectively represent the equipment topological coordinates corresponding to the equipment numbers of the first defective handicraft and the second defective handicraft within the j-th pair of defective handicrafts, displayed in the form of three-dimensional vectors; R is the set space coordinate normalization radius, taking the maximum physical span of the production line; is the process distance for each pair of defective handicrafts, , respectively represent the process deviation degrees of the first defective handicraft and the second defective handicraft within the j-th pair of defective handicrafts corresponding to the c-th dimensional parameter, is the global standard deviation of the c-th dimensional parameter; Combine the timestamps of the production time points corresponding to the first defective handicraft and the second defective handicraft within the jth pair of defective handicrafts to obtain the production period of the jth pair of defective handicrafts; Use the elbow method of the K-distance curve to calculate the K-distance of the production period corresponding to each pair of defective handicrafts, and then draw the K-distance curve to find the elbow point of the production period corresponding to each pair of defective handicrafts. Record the K-distance corresponding to the elbow point of the production period corresponding to each pair of defective handicrafts as the base radius of each pair of defective handicrafts ; For the production period of each pair of defective handicrafts, count the number of neighbors within the base radius and record it as ; and select the largest number of neighbors from the number of neighbors within the base radius of the production period corresponding to each pair of defective handicrafts ; Thus, calculate the search neighborhood radius of each pair of defective handicrafts ; If the spatio-temporal feature composite distance of a certain pair of defective handicrafts is less than or equal to the search neighborhood radius corresponding to this pair of defective handicrafts and satisfies that the absolute value of the production time point of the first defective handicraft within this pair of defective handicrafts minus the production time point of the second defective handicraft is less than or equal to the set maximum allowable time interval, then determine that the first defective handicraft and the second defective handicraft within this pair of defective handicrafts are in the same cluster. Otherwise, determine that the first defective handicraft and the second defective handicraft within this pair of defective handicrafts are not in the same cluster and reject cross-period merging; Obtain and summarize the production time points of all defective handicrafts in the same cluster to obtain all abnormal production time points of the defective handicrafts.

[0009] Locate the fault sources of the defective handicrafts produced at each abnormal production time point. The specific location process is as follows: Arbitrarily select a defective handicraft from each defective handicraft produced at each abnormal production time point as the target defective handicraft produced at each abnormal production time point, and obtain the spatio-temporal feature matrix of the target defective handicraft produced at each abnormal production time point; Furthermore, obtain all event record features of each upstream node before the occurrence of each abnormal production time point, and summarize them into the corresponding event matrix of each upstream node, where the event matrix includes the occurrence time point of each event and the event information; Sort all the events of each upstream node before the occurrence of each abnormal production time point in sequence according to the time stamp of each event, and number them according to the sorting order. Thus, number each upstream event as 1, 2,... i,... Q in sequence; The equipment association strength, operator association strength, and material batch association strength between each upstream event and the defective handicrafts produced at each abnormal production time point are obtained; they are respectively denoted as , where i is the number of each upstream event, i = 1, 2... Q, and v is the number of each abnormal production time point; Thus, the reverse propagation probability from the defective handicrafts produced at each abnormal production time point to each upstream event is defined , where α1, β1, and γ1 respectively represent the propagation weight coefficients of equipment, personnel, and materials, and satisfy α1 + β1 + γ1 = 1; λ is the time decay factor, and by default λ = 0.05 / hour, which is used to control the decay rate of the influence of historical events on the current propagation, represents the time interval between the v-th abnormal production time point and the i-th upstream event; Thus, the path weight from the defective handicrafts produced at each abnormal production time point to each upstream event is calculated ; is the process parameter coupling degree from the defective handicrafts produced at each abnormal production time point to each upstream event; Thus, the upstream event with the largest path weight is selected as the fault source of the defective handicrafts produced at the corresponding abnormal production time point.

[0010] On the other hand, the present invention provides an intelligent optimization system for the production process of an MES system based on artificial intelligence, including a quality inspection module, a time point identification module, and a fault source tracing module. The above-mentioned modules are connected by wired and / or wireless connection methods to realize data transmission between the modules; Quality inspection module: During the production process, the MES system is used to collect the quality inspection data of each handicraft in each production batch in real time; when it is found through quality inspection that there are a large number of unqualified handicrafts in a certain production batch, the production batch is recorded as a defective production batch, and similarly, the unqualified handicrafts are recorded as defective handicrafts, and the MES automatically triggers a quality alarm; Time point identification module: The defective handicrafts in each defective production batch are aggregated and combined in pairs to obtain the spatio-temporal feature matrix of each pair of defective handicrafts, and all abnormal production time points of the defective handicrafts are identified through a clustering algorithm; Fault source tracing module: Finally, it traces back in reverse to locate the fault source of the defective handicrafts produced at each abnormal production time point, and feeds back the fault source to the corresponding management personnel in the factory.

[0011] As described above, the intelligent optimization method and system for the production process of the MES system based on artificial intelligence provided by the present invention has at least the following beneficial effects: The present invention uses a clustering algorithm to analyze the spatio-temporal feature matrix of defective handicrafts, and can accurately identify all abnormal production time points of defective handicrafts. This method summarizes and integrates the defective handicrafts within each defective production batch, and combines them pairwise to obtain the spatio-temporal feature matrix of each pair of defective handicrafts. By analyzing these matrices through the clustering algorithm, the system can identify all abnormal production time points of defective handicrafts. This accurate identification method helps enterprises deeply understand the occurrence time and characteristics of quality problems, thereby providing data support and analysis basis for subsequent fault source location; Secondly, reverse tracing and fault source location are important links to ensure the closed-loop of quality control. By identifying abnormal production time points, enterprises can further reverse trace the fault sources of all defective handicrafts produced within these time points. The fault sources may include equipment failures, operator errors, material quality problems, etc. The system feeds these fault sources back to the corresponding management personnel in the factory, enabling them to take corrective measures in a timely manner to prevent similar problems from occurring again. This method of reverse tracing and fault source location not only improves the efficiency of solving quality problems, but also provides a scientific basis for enterprise improvement, promoting the continuous optimization of the production process; In summary, by using the MES system to collect quality inspection data in real time during the production process, trigger quality alerts, use the clustering algorithm to identify abnormal production time points, and finally reverse trace the fault sources and feedback them to the management personnel, enterprises can significantly improve the efficiency and accuracy of quality management. This method can not only quickly respond to and solve quality problems, but also continuously improve the production process through data analysis and feedback mechanisms, improving product quality and production efficiency. In the current manufacturing environment, this method has important benefits and necessities, and can significantly enhance the market competitiveness and customer satisfaction of enterprises. Description of the Drawings

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0013] Figure 1 It is a schematic connection diagram of each step of the method of the present invention.

[0014] Figure 2 It is a schematic connection diagram of each module of the system of the present invention. Detailed Embodiments

[0015] The following will combine the embodiments of the present invention. The above content is only an example and explanation of the concept of the present invention. Those skilled in the art of the present technology can make various modifications or supplements to the described specific embodiments or use similar methods for substitution, as long as they do not deviate from the concept of the invention or exceed the scope defined by this claim book, they should fall within the protection scope of the present invention.

[0016] Embodiment 1 Please refer to Figure 1 As shown, for the intelligent optimization method of the production process of the MES system based on artificial intelligence, this method includes the following steps: During the production process, use the MES system to collect the quality inspection data of each handicraft in each production batch in real time; when it is found through quality inspection that there are a large number of unqualified handicrafts in a certain production batch, mark this production batch as a defective production batch, and similarly mark the unqualified handicrafts as defective handicrafts, and the MES automatically triggers a quality alarm; When it is found through quality inspection that there are a large number of unqualified handicrafts in a certain production batch, mark this production batch as a defective production batch, and the specific operation logic is: Extract the measured values of the corresponding size parameters of each handicraft in each production batch from the quality inspection data of each handicraft in each production batch , where a is the number of each production batch, b is the number of each handicraft in the production batch, the value range of b is from 1 to M, and c is the number of each size parameter, the value range of c is from 1 to N; N represents the total number of size parameters to be detected for a single handicraft (including but not limited to length, aperture, thickness, etc.).

[0017] Example: If a handicraft needs to detect 3 parameters (length, width, height), then N = 3; Then calculate the size compliance coefficient of the handicrafts in each production batch , where M and N respectively represent the total number of handicrafts in the production batch and the total number of size parameters, represents the weight of the c-th size parameter, reflecting the importance of this size parameter to the overall quality, and satisfies ; is the single compliance of the b-th handicraft corresponding to the c-th size parameter in the a-th production batch, and its specific calculation formula is , is the target value of the handicrafts in the a-th production batch corresponding to the c-th size parameter, respectively represent the effective tolerance range, lower tolerance and upper tolerance of the c-th size parameter; Select the corresponding unilateral tolerance according to the measured value direction of the corresponding size parameters of each handicraft in each production batch: If , then ; If , then ; is the upper tolerance of the c-th dimensional parameter, that is, the maximum allowable positive deviation. Similarly is the minimum allowable positive deviation; , which means the measured value is equal to the target value, indicating perfect compliance; , the measured value exceeds the tolerance range, indicating complete non-compliance; , the measured value is within the tolerance range but deviates from the target value; Compare the dimensional compliance coefficients of the handicrafts within each production batch with the set dimensional compliance threshold coefficient for handicrafts. If the dimensional compliance coefficient of the handicrafts within a certain production batch is less than the set dimensional compliance threshold coefficient for handicrafts, then mark this production batch as a defective production batch.

[0018] Obtain the spatio-temporal feature matrices of each defective handicraft within each defective production batch, and identify all abnormal production time points of the defective handicrafts through a clustering algorithm. Furthermore, locate the propagation paths of each other unshipped handicraft within each defective production batch, and thus stop the shipment of each other unshipped handicraft within each defective production batch; The spatio-temporal feature matrices of each defective handicraft within each defective production batch are specifically shown as: Obtain the production time points, equipment numbers, and operator IDs of each defective handicraft within each defective production batch from the MES system; And calculate the process deviation degrees of each dimensional parameter of each defective handicraft within each defective production batch , is the measured value of the c-th dimensional parameter corresponding to the f-th defective handicraft in the d-th defective production batch. d is the number of each defective production batch, and f is the number of each defective handicraft within the defective production batch; are respectively the upper limit permission value and the target value of the c-th dimensional parameter corresponding to the handicrafts in the d-th defective production batch; The numerator part of the above calculation formula represents the absolute difference between the measured value and the target value; The denominator part represents the difference between the target value and the upper limit value, which is used to normalize the absolute difference; And accurately represent the production time points of each defective handicraft within each defective production batch in millisecond format, display them in the form of time stamps, and perform One-Hot encoding on the equipment numbers and operator IDs; Thus, obtain the spatio-temporal feature matrices of each defective handicraft within each defective production batch , respectively represent the time stamp display form of the production time point and the set of One-Hot encodings of the equipment information, They respectively represent the One-Hot encoding forms of the device number and the operator ID.

[0019] All abnormal production time points of defective handicrafts are identified therefrom through a clustering algorithm, including: Integrate all defective handicrafts within each defective production batch, combine them pairwise, and obtain the spatio-temporal feature matrix of each pair of defective handicrafts; among them, duplicate combinations of defective handicrafts and the first and second defective handicrafts within each pair of defective handicrafts are different need to be excluded; Define the spatio-temporal feature composite distance of each pair of defective handicrafts , where α, β, and γ are collectively called spatio-temporal process weight coefficients, and satisfy α + β + γ = 1. α is specifically the set time dimension weight, reflecting the impact of production time point continuity on clustering; β is specifically the set space dimension weight; γ is specifically the set process dimension weight; j is the number of each pair of defective handicrafts; is the time distance of each pair of defective handicrafts, which is corrected using a Gaussian kernel function, , exp(·) represents the exponential function with the natural constant e as the base, respectively represent the timestamps of the production time points corresponding to the first and second defective handicrafts within the jth pair of defective handicrafts; σ is the set time decay factor, with a default value set to 3 times the duration of a single production beat, in units of ms, usually with a default value set to 1800 ms; is the space distance of each pair of defective handicrafts, which is corrected and calculated using a mixture of the discrete feature jaccard similarity and the continuous coordinate Euclidean distance: , respectively represent the One-Hot encoding sets of the device information corresponding to the first and second defective handicrafts within the jth pair of defective handicrafts; respectively represent the device topological coordinates corresponding to the device numbers of the first and second defective handicrafts within the jth pair of defective handicrafts, shown in the form of three-dimensional vectors; R is the set space coordinate normalization radius, taking the maximum physical span of the production line; is the process distance of each pair of defective handicrafts, , respectively represent the process deviation degrees of the first and second defective handicrafts within the jth pair of defective handicrafts corresponding to the cth dimensional parameter, is the global standard deviation of the cth dimensional parameter; Combine the timestamps of the production time points corresponding to the first defective handicraft and the second defective handicraft within the j-th pair of defective handicrafts to obtain the production period of the j-th pair of defective handicrafts; Use the elbow method of the K-distance curve to calculate the K-distance of the production period corresponding to each pair of defective handicrafts, and then draw the K-distance curve to find the elbow point of the production period corresponding to each pair of defective handicrafts. Denote the K-distance corresponding to the elbow point of the production period corresponding to each pair of defective handicrafts as the base radius of each pair of defective handicrafts ; For the production period of each pair of defective handicrafts, count the number of its neighbors within the base radius and denote it as ; And screen out the largest number of neighbors from the number of neighbors within the base radius of the production period corresponding to each pair of defective handicrafts ; Calculate the search neighborhood radius of each pair of defective handicrafts accordingly ; If the composite distance of the spatio-temporal features of a certain pair of defective handicrafts is less than or equal to the search neighborhood radius corresponding to this pair of defective handicrafts and the absolute value of the production time point of the first defective handicraft within this pair of defective handicrafts minus the production time point of the second defective handicraft is less than or equal to the set maximum allowable time interval, then it is determined that the first defective handicraft and the second defective handicraft within this pair of defective handicrafts belong to the same cluster. Otherwise, it is determined that the first defective handicraft and the second defective handicraft within this pair of defective handicrafts do not belong to the same cluster, and cross-period merging is rejected; Obtain and summarize the production time points of all defective handicrafts belonging to the same cluster to obtain all abnormal production time points of the defective handicrafts.

[0020] It should be added that in the DBSCAN clustering algorithm, whether two data points belong to the same cluster depends on whether the distance between them is within the specified neighborhood radius. If the distance between two data points is greater than, then they are considered not to belong to the same cluster; Not belonging to the same cluster specifically means: 1. Too far distance: Under the composite distance metric of spatio-temporal features, the distance between period A1 and period A2 exceeds the neighborhood radius of period A, indicating that they are not close enough in spatio-temporal features; 2. Clustering separation: DBSCAN will divide data points into different clusters, and only those points within each other's neighborhood radius will be grouped into the same cluster. The distance between period A1 and period A2 exceeds, so they will not be grouped into the same cluster; 3. Anomaly detection: If period A1 and period A2 belong to different clusters respectively, then they may represent different abnormal production periods. The periods within each cluster have similar spatio-temporal features, indicating that they may be affected by the same abnormal factors; In DBSCAN, if the number of neighbors of a point is greater than or equal to the minimum number of points, then the point is called a core point. Core points form the core of clusters; Border point: If a point is not a core point but is within the neighborhood of a core point, then the point is called a border point; Noise point: If a point is neither a core point nor within the neighborhood of any core point, then the point is regarded as a noise point or an outlier; If neither time period A1 nor time period A2 has enough neighbor points to make them core points, or if their distance exceeds the neighborhood radius of each other, then they will not be in the same cluster; That time period A1 and time period A2 do not belong to the same cluster means that they are not similar enough in spatio-temporal characteristics to be classified as the same type of abnormal production time period.

[0021] Finally, trace back in reverse to locate the fault sources of the defective handicrafts produced at each abnormal production time point, and feedback the said fault sources to the corresponding management personnel in the factory.

[0022] To locate the fault sources of the defective handicrafts produced at each abnormal production time point, the specific location process is as follows: Select any defective handicraft from the defective handicrafts produced at each abnormal production time point as the target defective handicraft produced at each abnormal production time point, and obtain the spatio-temporal feature matrix of the target defective handicraft produced at each abnormal production time point; Furthermore, obtain all the event record features of each upstream node before each abnormal production time point occurs, and summarize them into the corresponding event matrix of each upstream node, where the event matrix includes the occurrence time points of each event and the event information; Sort all the events of each upstream node before each abnormal production time point occurs in sequence according to the time stamps of each event, and number them according to the sorting order, so that the upstream events are numbered 1, 2,... i,... Q in sequence; Thus, obtain the equipment association strength, operator association strength, and material batch association strength between each upstream event and the target defective handicraft produced at each abnormal production time point; denote them respectively as , where i is the number of each upstream event, i = 1, 2... Q, and v is the number of each abnormal production time point; ; ; ; Thus, define the reverse propagation probability from the target defective handicraft produced at each abnormal production time point to each upstream event , α1, β1, and γ1 represent the propagation weight coefficients of equipment, personnel, and materials respectively, and satisfy α1 + β1 + γ1 = 1; λ is the time decay factor, with a default value of λ = 0.05 / hour, which is used to control the decay rate of the impact of historical events on the current propagation. represents the time interval between the time point of the v-th abnormal production time point and the i-th upstream event; Upstream node This is a specific node in the production process, which may be a production process, equipment, operation step, or a certain monitoring point. At this node, important production-related data is usually recorded; represents the time interval between the time point of the v-th abnormal production time point and the i-th upstream event. If the occurrence time point of the upstream node event is 8:00, this refers to 8:00 in the morning of the same day when an event or an operation occurred at this upstream node. For example, this time point may be the equipment startup, the start of a certain process, the measurement time of an important parameter, or the completion time of an operation. In quality traceability and defect analysis, the event corresponding to the time point 8:0 is very important because it can help track and analyze whether abnormal situations occurred in the production process after this time point. If quality problems are found in the subsequent production process, by tracing back to the event at this time point, it can help identify the cause of the problem and take corresponding measures for correction. If a large batch of unqualified products are found in a production batch, by analyzing the event that occurred at the upstream node at 8:00, it can be determined whether the operation or event at this time point led to subsequent quality problems.

[0023] Thus, calculate the path weights from the production target defective handicrafts at each abnormal production time point to each upstream event ; is the process parameter coupling degree from the production target defective handicrafts at each abnormal production time point to each upstream event; The process parameter coupling degree from the production target defective handicrafts at each abnormal production time point to each upstream event is usually obtained by calculating the mutual information (MI) from the process parameters to evaluate the correlation of key parameters between the upstream node and the abnormal cluster. The greater the mutual information, the stronger the correlation between the two variables. If = 2.3, then this value represents the process parameter correlation between the upstream event and the abnormal production time point. Specifically, it can be a score from 0 to 10, indicating that there is a 23% correlation between the process parameters of these two nodes; These values are usually obtained based on historical data and statistical analysis. For example, by analyzing the production parameter data at different time points and using statistical models to estimate the mutual information between parameters of different nodes.

[0024] Thus, the upstream event with the largest path weight is selected as the fault source for the defective handicrafts produced at the corresponding abnormal production time point.

[0025] During the production process, the usage duration and fault records of equipment are very important data sources. By analyzing the usage time and failure rate of equipment, the equipment association strength can be calculated. Suppose the shared duration and failure rate of equipment M1 can be obtained from the equipment management system.

[0026] The operation records and proficiency ratings of operators can be obtained from the human resource management system or the production management system. By analyzing the operation batches and proficiency of operators, the operator association strength can be calculated.

[0027] The batch records of materials and the risk scores of suppliers can be obtained from the material management system. By analyzing the number of defects in material batches and the risk scores of suppliers, the material association strength can be calculated.

[0028] The time records during the production process are key data and can be obtained from the production management system. By analyzing the time interval between the abnormal cluster and the upstream node, the impact of the time decay factor on the propagation probability can be calculated.

[0029] The coupling degree of process parameters can be calculated through the historical data of key parameters (such as temperature, pressure, etc.) during the production process. Through statistical analysis, such as mutual information (MI), the process parameter correlation between the upstream event and the abnormal production time point can be evaluated.

[0030] Through the time reverse tracing algorithm, starting from the time point when the defect is detected, the possible propagation paths can be traced in the reverse time direction. This helps to quickly locate the source of the defect, shorten the troubleshooting time, and improve production efficiency. The algorithm comprehensively considers various factors such as equipment, personnel, and materials. Through weight assignment and association strength calculation, it can analyze the defect propagation path more comprehensively and accurately. This multi-factor analysis method can avoid the limitations of single-factor analysis and provide a more comprehensive quality control solution. The introduction of the time decay factor enables the algorithm to dynamically adjust the weights, focus on recent key events, and avoid the interference of historical irrelevant events. This ability of dynamic adjustment and real-time analysis is of great significance for quickly responding to production anomalies and improving production quality. By calculating the weights and optimal paths of the propagation paths, the algorithm provides clear path interpretability. This not only helps to understand the process of defect propagation but also provides a scientific basis for subsequent improvement measures.

[0031] During the production process, historical events may have a certain impact on current quality issues, but not all historical events are of equal importance. By introducing a time decay factor, the algorithm reduces the impact of historical events on the current propagation over time, thus focusing on recent key nodes. Specifically, the time decay factor weights historical events in the form of an exponential term, avoiding interference from irrelevant historical events. This dynamic decay method ensures that when analyzing the defect propagation path, the algorithm focuses on key events that have a direct impact on current quality issues. For example, in a certain production batch, if a device malfunctioned recently, the weight of this malfunction event will be higher than that of a malfunction event several months ago, thus occupying a more important position in the defect propagation path. This method of focusing on recent key nodes helps enterprises quickly identify and solve current quality problems, improving production flexibility and response speed; Finally, path interpretability is a major highlight of the time reverse tracing algorithm. By calculating the weights of the defect propagation paths, the algorithm can identify the path with the maximum weight and directly map it to the process flow chart. This method of path interpretability enables enterprises to quickly locate the fault source, understand the process of defect propagation, and take corresponding improvement measures. Specifically, the calculation result of the path with the maximum weight not only provides the detailed path of defect propagation, but also helps enterprises visually see the location where the problem occurred and the propagation process by comparing it with the process flow chart. For example, in a certain production batch, if the path with the maximum weight shows problems with the operation of a certain device and operator at a specific time point, then the enterprise can directly find the corresponding location on the process flow chart for detailed inspection and analysis. This method of path interpretability not only improves the efficiency of fault troubleshooting, but also provides a scientific decision-making basis for enterprises, supporting the rapid location and solution of quality problems in production; In summary, the time reverse tracing algorithm, through technical advantages such as multi-factor quantification, dynamic decay, and path interpretability, helps enterprises achieve more accurate and efficient quality control and fault troubleshooting. This method can not only comprehensively reflect the actual situation in the production process, focus on recent key nodes, but also visually display the defect propagation path, providing a scientific decision-making basis for enterprises and improving production efficiency and product quality.

[0032] Please refer to Figure 2 As shown, the intelligent optimization system for the production process of the AI-based MES system includes a quality inspection module, a time point identification module, and a fault source tracing module. The above-mentioned modules are connected by wired and / or wireless connection methods to achieve data transmission between the modules; Quality inspection module: During the production process, the MES system is used to collect the quality inspection data of each handicraft in each production batch in real time; when a large number of handicrafts in a certain production batch are found to be unqualified during quality inspection, the production batch is recorded as a defective production batch, and similarly, the unqualified handicrafts are recorded as defective handicrafts, and the MES automatically triggers a quality alarm; Time point identification module: Aggregate and integrate each defective handicraft in each defective production batch, and combine them in pairs to obtain the spatio-temporal feature matrix of each pair of defective handicrafts, and identify all abnormal production time points of the defective handicrafts through a clustering algorithm; Fault source tracing module: Finally, trace back in reverse to locate the fault sources of the defective handicrafts produced at each abnormal production time point, and feedback the fault sources to the corresponding management personnel in the factory.

[0033] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not mean the order of execution, and the order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0034] It should be understood that determining B based on A does not mean determining B only based on A, and B can also be determined based on A and / or other information.

[0035] As mentioned above, the above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all of them should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0036] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent optimization method for the production process of an MES system based on artificial intelligence, characterized in that, Including: During the production process, use the MES system to collect the quality inspection data of each handicraft in each production batch in real time; When a large number of handicrafts in a certain production batch are found to be unqualified during quality inspection, mark this production batch as a defective production batch. Similarly, mark the unqualified handicrafts as defective handicrafts, and the MES automatically triggers a quality alarm; Summarize and integrate each defective handicraft in each defective production batch, combine them in pairs to obtain the spatio-temporal feature matrix of each pair of defective handicrafts, and identify all abnormal production time points of the defective handicrafts through a clustering algorithm; Finally, trace back in reverse, locate the fault sources of the defective handicrafts produced at each abnormal production time point, and feedback the fault sources to the corresponding management personnel in the factory.

2. The intelligent optimization method for the production process of the AI-based MES system according to claim 1, wherein When a large number of handicrafts in a certain production batch are found to be unqualified during quality inspection, mark this production batch as a defective production batch. The specific operation logic is as follows: The measured values of the corresponding size parameters of each handicraft within each production batch are obtained by extracting the quality inspection data of each handicraft within each production batch. , where a is the number of each production batch, b is the number of each handicraft within the production batch, the value range of b is from 1 to M, c is the number of each size parameter, and the value range of c is from 1 to N; Then, the compliance coefficient of the handicraft size within each production batch is calculated , where M and N respectively represent the total number of handicrafts and the total number of size parameters within the production batch represents the weight of the c-th size parameter, reflecting the importance of this size parameter to the overall quality, and satisfies ; is the single compliance of the b-th handicraft corresponding to the c-th size parameter in the a-th production batch, and its specific calculation formula is , is the target value of the handicraft corresponding to the c-th size parameter in the a-th production batch respectively represent the effective tolerance range, lower tolerance, and upper tolerance of the c-th size parameter Select the corresponding unilateral tolerance according to the measured value direction of each handicraft corresponding to each size parameter within each production batch If , then ; If , then ; Compare the handicraft size compliance coefficient in each production batch with the set handicraft size compliance threshold coefficient. If the handicraft size compliance coefficient in a certain production batch is less than the set handicraft size compliance threshold coefficient, then mark this production batch as a defective production batch.

3. The intelligent optimization method for the production process of the AI-based MES system according to claim 1, wherein The spatio-temporal feature matrix of each defective handicraft in each defective production batch is specifically shown as: Obtain the production time points, equipment numbers, and operator IDs of each defective handicraft in each defective production batch from the MES system; And calculate the process deviation degree of each dimensional parameter of each defective handicraft within each defective production batch , is the measured value of the c-th dimensional parameter corresponding to the f-th defective handicraft in the d-th defective production batch, d is the number of each defective production batch, and f is the number of each defective handicraft within the defective production batch; are respectively the upper limit allowable value and the target value of the c-th dimensional parameter corresponding to the handicrafts in the d-th defective production batch; Precisely convert the production time points of each defective handicraft in each defective production batch to the millisecond form and display them in the form of time stamps. Perform One-Hot encoding on the equipment numbers and operator IDs; The spatio-temporal feature matrix of each defective handicraft within each defective production batch is thus obtained , respectively represent the timestamp display form of the production time point and the One-Hot encoding set of equipment information respectively represent the One-Hot encoding forms of the equipment number and the operator ID 4. The intelligent optimization method for the production process of the AI-based MES system according to claim 3, wherein, Identify all abnormal production time points of the defective handicrafts through a clustering algorithm, including: Define the spatio-temporal feature composite distance for each pair of defective handicrafts , where α, β, and γ are collectively referred to as spatio-temporal process weight coefficients, and satisfy α + β + γ = 1. α is specifically the set weight in the time dimension, reflecting the influence of the continuity of production time points on clustering; β is specifically the set weight in the space dimension; γ is specifically the set weight in the process dimension; j is the number of each pair of defective handicrafts; For the time distance between each pair of defective handicrafts, it is corrected using a Gaussian kernel function, , where exp(·) represents the exponential function with the natural constant e as the base, respectively represent the timestamps of the production time points corresponding to the first defective handicraft and the second defective handicraft within the j-th pair of defective handicrafts; σ is the set time decay factor, and the default value is set to 3 times the duration of a single production beat, with the unit of ms; For the spatial distance between each pair of defective handicrafts, a hybrid of discrete feature Jaccard similarity and continuous coordinate Euclidean distance is used for correction calculation: , respectively represent the One - Hot encoding sets corresponding to the device information of the first defective handicraft and the second defective handicraft within the j - th pair of defective handicrafts; respectively represent the device topological coordinates corresponding to the device numbers of the first defective handicraft and the second defective handicraft within the j - th pair of defective handicrafts, shown in the form of three - dimensional vectors; R is the set normalization radius of the spatial coordinates, taking the maximum physical span of the production line; is the process distance between each pair of defective handicrafts, , respectively represent the process deviation degrees of the first defective handicraft and the second defective handicraft in the j-th pair of defective handicrafts corresponding to the c-th dimensional parameter, is the global standard deviation of the c-th dimensional parameter.

5. The intelligent optimization method for the production process of the AI-based MES system according to claim 4, wherein, Identify all abnormal production time points of the defective handicrafts through a clustering algorithm, and also include: Combine the time stamps of the production time points corresponding to the first defective handicraft and the second defective handicraft in the jth pair of defective handicrafts to obtain the production time period of the jth pair of defective handicrafts; Calculate the K-distance of each pair of defective handicrafts corresponding to the production period by using the elbow method of the K-distance curve, and then draw the K-distance curve to find the elbow points of each pair of defective handicrafts corresponding to the production period. Record the K-distance corresponding to the elbow points of each pair of defective handicrafts corresponding to the production period as the base radius of each pair of defective handicrafts ; For each pair of defective handicrafts in the production period, count the number of their neighbors within the said basic radius, and record it as ; and select the maximum number of neighbors from the number of neighbors within the basic radius of the corresponding production period of each pair of defective handicrafts ; Calculate the search neighborhood radius of each pair of defective handicrafts accordingly ; If the spatio-temporal feature composite distance of a certain pair of defective handicrafts is less than or equal to the search neighborhood radius corresponding to this pair of defective handicrafts and satisfies that the absolute value of the production time point of the first defective handicraft in this pair of defective handicrafts minus the production time point of the second defective handicraft is less than or equal to the set maximum allowable time interval, then determine that the first defective handicraft and the second defective handicraft in this pair of defective handicrafts are in the same cluster. Otherwise, determine that the first defective handicraft and the second defective handicraft in this pair of defective handicrafts are not in the same cluster and reject cross-time period merging; Obtain and summarize the production time points of all defective handicrafts in the same cluster to obtain all abnormal production time points of the defective handicrafts.

6. The intelligent optimization method for the production process of the MES system based on artificial intelligence according to claim 1, characterized in that Locate the fault sources of the defective handicrafts produced at each abnormal production time point. The specific location process is as follows: Arbitrarily select a defective handicraft from each defective handicraft produced at each abnormal production time point as the target defective handicraft produced at each abnormal production time point, and obtain the spatio-temporal feature matrix of the target defective handicraft produced at each abnormal production time point; Furthermore, all event record features before each abnormal production time point of each upstream node are obtained and consolidated into an event matrix corresponding to each upstream node, where the event matrix includes the occurrence time point of each event and the event information; All events of each upstream node before each abnormal production time point are sorted in sequence according to the time stamps of each event occurrence, and numbered according to the sorting order, so that each upstream event is numbered 1, 2,... i,... Q in sequence; Thus, the equipment association strength, operator association strength, and material batch association strength between each upstream event and the defective handicrafts produced at each abnormal production time point are obtained; they are respectively denoted as , where i is the number of each upstream event, i = 1, 2... Q, and v is the number of each abnormal production time point; Thus, the reverse propagation probability of the defective handicrafts produced at each abnormal production time point to each upstream event is defined. , where α1, β1, and γ1 represent the propagation weight coefficients of equipment, personnel, and materials, respectively, and satisfy α1 + β1 + γ1 = 1; λ is the time decay factor, with a default value of λ = 0.05 / hour, which is used to control the decay rate of the influence of historical events on the current propagation. represents the time interval between the v-th abnormal production time point and the i-th upstream event. Calculate the path weights from the production target defective handicrafts at each abnormal production time point to each upstream event accordingly. ; The process parameter coupling degree from the production target defective handicrafts at each abnormal production time point to each upstream event; Thus, the upstream event with the largest path weight is selected as the fault source of the defective handicraft produced at the corresponding abnormal production time point.

7. An intelligent optimization system for the production process of an MES system based on artificial intelligence, characterized in that, It is implemented based on the intelligent optimization method for the production process of the MES system based on artificial intelligence according to any one of claims 1-6, and includes a quality inspection module, a time point identification module and a fault source tracing module. The above-mentioned modules are connected by wired and / or wireless connection methods to realize data transmission between the modules; Quality inspection module: During the production process, the MES system is used to collect the quality inspection data of each handicraft in each production batch in real time; when it is found through quality inspection that there are a large number of unqualified handicrafts in a certain production batch, the production batch is recorded as a defective production batch. Similarly, the unqualified handicrafts are recorded as defective handicrafts, and the MES automatically triggers a quality alarm; Time point identification module: The defective handicrafts in each defective production batch are aggregated and combined in pairs to obtain a spatio-temporal feature matrix of each pair of defective handicrafts, and all abnormal production time points of the defective handicrafts are identified through a clustering algorithm; Fault source tracing module: Finally, trace back in reverse to locate the fault source of the defective handicrafts produced at each abnormal production time point, and feedback the fault source to the corresponding management personnel in the factory.

Images