MES intelligent mistake proofing early warning system

Abstract

The application discloses an MES intelligent mistake-proofing early warning system and relates to the technical field of intelligent mistake-proofing early warning; the system comprises an input capturing module, a behavior modeling module, a business correlation module, an evolution monitoring module, a propagation prediction module and an intelligent intervention module; the input capturing module establishes a business session when an operator performs a business operation, collects personnel, task, process and real-time behavior data and assembles the data into operation data; the behavior modeling module constructs stable context, real-time context and task context and performs weighted fusion to generate operation context; the business correlation module establishes a business correlation network based on business dependency relationship and calculates the consistency index of each correlation edge; the evolution monitoring module tracks the time sequence change of the consistency index and locates a descending correlation edge; the propagation prediction module calculates an error propagation index; and the intelligent intervention module generates a hierarchical intervention strategy according to the error propagation index and executes the strategy.

Description

Technical Field

[0001] This invention relates to the field of intelligent error prevention and early warning technology, and more specifically, to an MES intelligent error prevention and early warning system. Background Technology

[0002] As manufacturing enterprises improve their digitalization levels, MES systems have been deeply integrated into key processes such as production order creation, BOM configuration, material requisition, and process parameter adjustment. In the actual operation of MES systems, a large amount of data still needs to be manually selected or entered by operators. Due to frequent production changes and differences in operator experience and attention spans, human error (such as incorrect version selection or incorrect code entry) can easily lead to material loss, rework, and even quality accidents.

[0003] However, existing MES systems generally use fixed verification rules, such as required fields and parameter ranges, for interception. But because this method relies on preset logical constraints, it can only identify explicit anomalies such as format violations or numerical limits exceeding limits. It cannot identify the subconscious operational habits of specific operators in specific scenarios. Although a few existing technologies introduce operator profiles for anomaly detection, they use offline static profiles and single-time matching with historical errors. They cannot perceive the dynamic changes in the current state of the operator, nor can they track the evolution trend of erroneous behavior. In essence, it is still a matter of checking errors after the fact or blocking errors in real time, which is difficult to eliminate the risk of human error at the root. Summary of the Invention

[0004] The main objective of this invention is to provide an intelligent error prevention and early warning system for MES to overcome the problems mentioned in the background art.

[0005] To achieve the above objectives, an intelligent error prevention and early warning system for MES is provided, comprising:

[0006] The input capture module is used to establish a business session when the operator performs business operations, collect personnel objects, task objects, process objects and real-time behavior objects, and assemble them into operation data;

[0007] The behavior modeling module is used to receive the operation data, construct a stable context, a real-time context, and a task context, and then weightedly fuse the three to generate a comprehensive operation context.

[0008] The business association module is used to establish a business association network based on the dependency relationship between business objects in the MES system, and to calculate the consistency index of each business association edge according to the comprehensive operation context.

[0009] The evolution monitoring module is used to track the temporal changes of the consistency index of each business-related edge, and to locate the path of consistency decline when it is determined that the consistency shows a continuous deterioration trend.

[0010] The propagation prediction module is used to map the descent associated edges in the consistency descent path to activated nodes, and predict the activation probability of the activated nodes propagating to other nodes along the business association chain based on the pre-built error propagation graph. Potential target nodes that meet preset conditions are used as predicted target nodes. The error propagation index is calculated based on the activation probabilities of the activated nodes and the predicted target nodes to generate a current activated path node set, which is the union of the activated nodes and the predicted target nodes.

[0011] The intelligent intervention module is used to generate and execute tiered intervention strategies based on the error propagation index.

[0012] Furthermore, the personnel object includes the operator's unique identifier, current job title, total number of historical operations, and job authorization level; the task object includes the order number, product model, task type, and operation type; the process object includes the process route number, BOM version number, material code, and nominal value of key process parameters; and the real-time behavior object includes the field input order, time interval between adjacent field inputs, number of times each field is modified, number of page returns, and number of BOM switches.

[0013] Furthermore, the stable context constructed by the behavior modeling module includes operation proficiency level, product coverage, process route usage preference, equipment usage preference, and standard input order. The stable context is updated at fixed intervals based on historical business operation records. The real-time context includes current input rhythm, field modification frequency, page jump frequency, BOM switching frequency, and parameter adjustment frequency. The real-time context is dynamically refreshed with each step of the operator's input. The task context includes production change status, order complexity coefficient, equipment change status, and production stage identifier. The task context is determined when a business session is established and remains unchanged during that business session.

[0014] Furthermore, the business association network consists of business objects as nodes and the dependencies between business objects as directed edges. The business objects include product model, BOM version, process route, equipment number, process parameters, and work report. The business association edges include five association edges: product model to BOM version, BOM version to process route, process route to equipment number, equipment number to process parameters, and process parameters to work report. The consistency index is calculated by the business association module according to the business constraint rules pre-stored in the MES system and is used to quantify the matching degree between the business objects at both ends of each association edge.

[0015] Furthermore, the evolution monitoring module tracks the business consistency index sequence of the most recent K monitoring times, calculates the consistency change rate between adjacent times, and determines whether the consistency shows a continuous deterioration trend based on the consistency continuous deterioration index. When it is determined to be a continuous deterioration trend, the consistency decline path is pushed to the propagation prediction module.

[0016] The evolution monitoring module tracks the business consistency index sequence of the most recent K monitoring times, calculates the consistency change rate of adjacent times, and determines whether the consistency shows a continuous deterioration trend based on the consistency deterioration index. The consistency deterioration index is the ratio of the number of times the consistency change rate is negative in the most recent K monitoring times to K. When the consistency deterioration index is greater than or equal to the preset deterioration trend threshold, it is determined to be a continuous deterioration trend, and the consistency decline path is pushed to the propagation prediction module.

[0017] Furthermore, the propagation prediction module constructs the error propagation graph during system initialization. The error propagation graph is a directed graph structure, where nodes represent business error types and directed edges have weights to characterize the propagation probability between error types. Upon receiving the consistency descent path, the propagation prediction module maps each descent-related edge in the consistency descent path to a corresponding activated node, and predicts the activation probability of each potential target node along the direction of the error propagation graph, starting from the activated node. The activation probabilities of the activated nodes and the predicted target nodes are weighted and summed to obtain the error propagation index.

[0018] Furthermore, the intelligent intervention module determines to perform one of the following interventions based on the preset interval of the error propagation index and the number of nodes in the current active path node set: prompt-level intervention, confirmation-level intervention, or correction-level intervention. The prompt-level intervention displays prompt information in a non-blocking manner, the confirmation-level intervention requires the operator to confirm each item in a modal dialog box, and the correction-level intervention is automatically performed by the system and requires the operator to confirm each item.

[0019] Furthermore, in the correction-level intervention, the intelligent intervention module performs automatic correction operations in sequence according to the priority order of product to BOM, BOM to process, process to equipment, equipment to parameters, and parameters to work report, based on the descending associated edges carried by the activated nodes in the current active path node set. The original value before correction and the new value after correction are recorded in the correction log for the operator to confirm item by item.

[0020] Furthermore, the intelligent intervention module will directly perform a correction-level intervention when any of the following situations occur: the consistency index between the product model and the BOM version is 0; the consistency index between the equipment number and the process parameters is 0 and the process parameters involved exceed the safe operating range of the equipment; the material code does not match the current BOM version and the material code is not in the alternative material list of this BOM version.

[0021] The beneficial effects of this invention are:

[0022] (1) This invention constructs a three-layer model of stable context, real-time context and task context and performs weighted fusion to fully characterize the complete context of each operation from three dimensions: the operator's long-term operating habits, the real-time behavior status in the current session and the task attributes of the current order. On this basis, the inherent business dependencies between products, BOM, process, equipment, parameters and work reports in the MES system are constructed into an association network. The consistency index is calculated for each association edge to realize the overall business consistency assessment of the operation data, and the error identification is upgraded from single-point field verification to chain-level structured analysis.

[0023] (2) By continuously tracking the temporal changes of the consistency index of each associated edge, the present invention calculates the decline rate and the proportion of continuous deterioration at adjacent times, and can identify the evolution trend when the business consistency just shows signs of decline but before an actual error is formed; further, the historical error records are constructed into an error propagation graph, and the probability of the current error tendency to propagate to subsequent business links along the business dependency chain is predicted based on the consistency decline path, realizing the leap from identifying the deviation that has occurred to predicting the error that will occur, so that the system can actively intervene before the error actually forms and block the path of the error into the production process from the root.

[0024] (3) Based on the risk range of the error propagation index and the breadth of the current active path node set, the present invention automatically determines to perform prompt-level, confirmation-level, or correction-level intervention. Prompt-level intervention provides reference information in a non-blocking manner without interrupting the normal operation process; confirmation-level intervention requires the operator to confirm each item in a modal dialog box to ensure that the operator is aware of all inconsistencies; correction-level intervention is automatically performed by the system according to business priority and records correction logs. The operator can only continue after confirming each correction item. The three-level progressive intervention strategy ensures smooth operation in low-risk situations and ensures forced intervention in high-risk situations, taking into account both error prevention effect and operation efficiency. Attached Figure Description

[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0026] Figure 1 This is a schematic diagram of the module connections of the present invention. Detailed Implementation

[0027] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of the invention described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0031] Application Scenario: Existing MES systems generally use fixed verification rules, such as required fields and parameter ranges, for interception. However, because this method relies on preset logical constraints, it can only identify explicit anomalies such as format violations or numerical limits exceeding limits. It cannot identify the subconscious operational habits of specific operators in specific scenarios. Although a few existing technologies introduce operator profiling for anomaly detection, they use offline static profiling and single-time matching with historical errors. This cannot perceive the dynamic changes in the operator's current state, nor can it track the evolution trend of erroneous behavior. Essentially, it is still a matter of post-event error checking or immediate error blocking, which is difficult to eliminate the risk of human error at its root. To solve the above technical problems, this embodiment proposes an intelligent error prevention and early warning system for MES. It is applied to the MES manufacturing execution system. During the process of operators performing production order creation, process route configuration, material requisition, production reporting, and process parameter adjustment, the system analyzes the operation behavior in real time, predicts the risk of error before the operation is submitted, and proactively provides intelligent early warning.

[0032] Please see Figure 1 As shown, the present invention is an intelligent error prevention and early warning system for MES, comprising: an input capture module, a business association module, an evolution monitoring module, a propagation prediction module, and an intelligent intervention module;

[0033] When the input capture module detects that an operator has received a work order and activated the operation interface, it immediately establishes a unique business session and captures personnel objects, task objects, process objects, and real-time behavior objects. Specifically, personnel objects include the operator's unique identifier, current job title, total number of historical operations, and job authorization level. The job authorization level is an integer, with a higher value indicating higher authority. In this embodiment, the job authorization level includes five levels from 1 to 5. Task objects include order number, product model, task type (i.e., normal, rework, and trial production), and operation type (changeover operation and non-changeover operation). Process objects include process route number, BOM version number, material code, and nominal values ​​of key process parameters. Common dimensions of process parameters are usually temperature, pressure, speed, duration, and location. Specifically, these are standard reference values ​​extracted from the process route or formula, used for consistency comparison with the real-time values ​​actually entered by the operator. Process parameters vary greatly across different industries, and different processes have corresponding nominal values ​​for key process parameters. Real-time behavior objects include field input order, time interval sequence between adjacent field inputs, number of modifications to each field, number of page returns, and number of BOM switches.

[0034] The aforementioned personnel objects, task objects, process objects, and real-time behavior objects are assembled into unified operational data and sent to the business-related module; the specific assembly and unification process is as follows:

[0035] The original data collected from four types of objects underwent format standardization and type conversion. Specifically, for personnel objects: the operator's unique identifier was converted to a unified personnel ID format within the system, and the current job title was uniformly converted to a standard job code; the job authorization level was directly parameterized as an integer in subsequent analysis and calculation processes, without requiring additional conversion; the total number of historical cumulative operations was recorded as an integer, serving as the benchmark value for subsequent proficiency assessment. For task objects: the order number retained its original string format; the product model was uniformly converted to the standard model code in the product coding system; the task type was represented by an enumerated integer, assigning different code values ​​to normal orders, rework orders, and trial production orders; the operation type was represented by a Boolean value to indicate changeover operations and non-changeover operations. For process objects: the process route number, BOM version number, and material code all retained their original string format; the nominal values ​​of key process parameters were organized in the form of key-value pair sets, for example, recording the nominal temperature value for temperature parameters and the nominal pressure value for pressure parameters, serving as a reference benchmark for subsequent consistency comparison. The process parameters vary greatly across different industries, and each process has its own set of nominal values ​​for key process parameters; the regularization of real-time behavior objects: the field input order is recorded in the form of an ordered array to record the access order of each field; the time interval sequence between adjacent field inputs is recorded in the form of a floating-point array, with the unit being seconds; the number of field modifications, the number of page returns, and the number of BOM switches are all recorded in integer form;

[0036] After the above data is organized, the four types of objects are encapsulated into a complete operation data according to a predefined data structure. Specifically, this operation data consists of four core components: The first part is personnel information, which includes the operator's unique identifier, job code, total number of historical operations, and job authorization level. The job authorization level is an integer, with a higher value indicating higher authority. In this embodiment, five job authorization levels are set from 1 to 5, where level 1 corresponds to trainee operator, level 3 corresponds to skilled operator, and level 5 corresponds to team leader or senior technician. The second part is task information, which includes order number, product model code, task type code, and changeover flag. The task type code is used to distinguish between normal orders, rework orders, and trial production orders. The changeover flag is used to indicate whether the current operation involves product switching, and this flag provides key input for the construction of subsequent task context. The third part is process information, which includes process route number, BOM version number, material code, and a set of nominal values ​​for key process parameters. The set of nominal values ​​for key process parameters is organized in pairs, with parameter names and nominal values, covering general process dimensions such as temperature, pressure, speed, duration, and location. Specific dimensions are determined based on the actual configuration of the current process route. The fourth part is real-time operation information, which includes an array of field input order, a sequence of time intervals between adjacent field inputs, a cumulative value of the number of modifications to each field, a cumulative value of the number of page returns, and a cumulative value of the number of BOM switches. This data is dynamically updated as the operator continues to input data, reflecting the operator's real-time behavior in the current session. After the above four parts of information are encapsulated according to a fixed data hierarchy, a unique identifier for the current business session and a session creation timestamp are added to the operation data, forming a complete and independently transmittable set of operation data, which is then sent to the behavior modeling module.

[0037] The behavior modeling module is based on a three-layer context model. It models operational behavior from three dimensions: long-term stable features, real-time dynamic features, and current task features, to obtain stable context, real-time context, and task context. Based on these, a weighted fusion is performed to generate the operational context, specifically including:

[0038] The stable context reflects the operator's long-term, slowly changing operational habits and proficiency, including proficiency level, product coverage, process route usage preferences, equipment usage preferences, and standard input sequence. Specifically, the operator's unique identifier and total historical operation count are extracted from the personnel information in the operation data. Simultaneously, the operator's historical business operation records stored in the MES historical database are retrieved; in this embodiment, the time window is the most recent 90 days. A level mapping is performed based on the operator's total historical operation count: when the total number of operations is less than 50, the proficiency level is determined as Level 1, corresponding to a trainee operator; when the total number of operations is greater than or equal to 50 and less than 200, it is determined as Level 2, corresponding to a junior operator; when the total number of operations is greater than or equal to 200 and less than 500, it is determined as Level 3, corresponding to a skilled operator; when the total number of operations is greater than or equal to 500 and less than 1000, it is determined as Level 4, corresponding to a senior operator; and when the total number of operations is greater than or equal to 1000, it is determined as Level 5, corresponding to a technician-level operator. The system also counts all products handled by the operator in the past 90 days. Product Models: The deduplicated set of product models is used as the operator's product coverage. Frequency of each process route selected by the operator in the past 90 days is calculated, and the frequency of each process route is divided by the total number of operations to obtain a normalized process route preference probability distribution. Frequency of each device used by the operator in the past 90 days is also calculated and normalized to obtain a device preference probability distribution. Access order of each step when the operator completes business operations in the past 90 days is extracted, and the transition frequency between any two steps is calculated to form an inter-step transition probability matrix. The rows and columns of this matrix correspond to the input fields on the operation page, and the value of each element in the matrix represents the probability of jumping from the corresponding field in the row to the corresponding field in the column. The stable context is stored in a structured form in the cache of the behavior modeling module, and the update cycle is set to be recalculated every 24 hours. For new operators with less than 50 cumulative operations, the stable context is temporarily replaced by the average of the corresponding dimensions of three operators with similar authorization levels in the same position. Once the operator's cumulative operation count reaches 50, it is switched to their own data.

[0039] The real-time context reflects the operator's current real-time behavioral state, dynamically refreshing with each input step in the current business session. The real-time context includes the current input rhythm, field modification frequency, page jump frequency, BOM switching frequency, and parameter adjustment frequency. Specifically, all valid interval values ​​in the sequence of adjacent field input time intervals are extracted, and their arithmetic mean is calculated. This average is used as the quantification value of the current input rhythm, in seconds. A smaller current input rhythm value indicates a faster operator speed; values ​​that are too large or too small indicate a possible abnormal state. The total number of input fields on the current page and the cumulative number of modifications for each field are extracted from the operation data. The sum of the modification counts for all fields is divided by the total number of input fields to obtain the field modification frequency. The field modification frequency reflects the operator's uncertainty about the entered content in the current session. The cumulative number of page returns is used as the raw count for page jump frequency, which records the total number of times the operator returns to the previous page or previous step in the current session. The cumulative number of BOM switching times is extracted from the operation data, and the number of operation steps completed in the current session is also counted. Dividing the cumulative number of BOM switching times by the number of completed operation steps yields the BOM switching frequency, which reflects the operator's hesitation or uncertainty in BOM version selection. The total number of parameters contained in the set of nominal values ​​of key process parameters is obtained from the process information of the operation data. Simultaneously, the cumulative number of modifications to each process parameter field is counted from the real-time operation information. The sum of the modification counts for all process parameter fields is divided by the total number of parameters to obtain the parameter adjustment frequency. It should be noted that the real-time context is updated after each field input by the operator.

[0040] The task context reflects the inherent characteristics of the current operation scenario, including changeover status, order complexity coefficient, equipment switching status, and production stage identifier. Specifically, the changeover flag in the operation data is directly taken; a value of 1 indicates that the current operation involves product switching, and a value of 0 indicates no product switching. The order complexity is determined based on the task type code and product model. When the task type code corresponds to a normal order, the base value of the complexity coefficient is 1.0; when it corresponds to a rework order, the base value is 1.3; and when it corresponds to a trial production order, the base value is 1.5. Furthermore, the average number of process steps for this product model in historical production records is queried. If the average number of steps is greater than the median average number of steps for all product models in the workshop, the complexity coefficient is increased by 0.2 from the base value. After the above calculations, a value greater than... A floating-point number equal to 1.0 is used as the order complexity coefficient; the equipment number in the process information of the current operation data is compared with the equipment number used in the operator's previous completed order. If they are different, the equipment switching status value is 1; otherwise, it is 0. If the operator has no completed orders in this shift, the equipment switching status value is set to 0. The production stage is determined according to the task type code. When the task type code corresponds to a normal order, the current life cycle stage of the product model is further queried. When it is in the small batch verification stage, it is marked as trial production; when it is in the batch production stage, it is marked as mass production; when the task type code corresponds to a rework order, it is marked as rework; when it corresponds to a trial production order, it is marked as trial production. It should be noted that the task context is calculated once when the operator's business session is established and does not change during the continuous operation of this task.

[0041] The stable context, real-time context, and task context are weighted and fused to generate the operational context. Specifically, the feature data of each dimension in the three types of contexts are uniformly converted into a numerical vector format. The five dimensions of the stable context are converted into a five-dimensional vector, and the values ​​of each dimension have been normalized to the range of 0 to 1. The five dimensions of the real-time context are also converted into a five-dimensional vector, where the input rhythm is taken as its deviation rate relative to the operator's historical average, and the other four dimensions directly use the ratios calculated in the previous steps. The four dimensions of the task context are converted into four-dimensional vectors, where the production changeover status and equipment switching status directly use values ​​of 0 or 1, and the order complexity coefficient minus 1.0 is used as the value of this dimension. The production stage identifier is mapped to 0.2 according to trial production, mass production, finalization, and rework, respectively. The values ​​are 5, 0.75, 0.50, and 0.30. The three vectors are weighted and summed according to preset weights. In this embodiment, the weight of the stable context is 0.40, the weight of the real-time context is 0.35, and the weight of the task context is 0.25, with a sum of 1. The fused operational context is a fourteen-dimensional numerical vector, where the first five dimensions come from the stable context, the middle five from the real-time context, and the last four from the task context. Each dimension maintains a continuous value within the range of 0 to 1. It should be noted that the operator's long-term behavioral pattern is the most reliable benchmark for judging operational consistency, therefore the stable context is given the highest weight; real-time behavioral changes are most sensitive to changes in operational status, so they are given the second highest weight; the task scenario plays a corrective and regulatory role, so it is given a lower weight.

[0042] The business association module constructs a business dependency chain based on the dependencies between various businesses within the system, and analyzes the state of the associated edges accordingly, specifically including:

[0043] A business association network is established based on the inherent business dependencies within the MES system. In the MES system, there are defined dependency chains between business objects involved in production operations: the product model determines the available BOM version, the BOM version determines the applicable process route, the process route limits the selectable equipment types, the equipment type constrains the range of configurable process parameters, and the correctness of the process parameters directly affects the validity of subsequent work reporting data. Based on these business dependencies, the business association network is defined as a graph structure composed of nodes and directed edges. Nodes represent business objects, and directed edges represent the dependencies between business objects. In this embodiment, the node set of the business association network contains six nodes: product model node, BOM version node, process route node, equipment number node, process parameter node, and work reporting node. The work reporting node is not activated until the current operation is submitted; it is automatically added by the system after the operator completes the work reporting operation. The set of directed edges contains five dependency edges: the edge from the product model node to the BOM version node, the edge from the BOM version node to the BOM version node, ... product model node to the BOM version node, the edge from the BOM version node to the BOM version node, and the edge from the BOM version node to the BOM version node. Edges pointing to process route nodes, process route nodes pointing to equipment number nodes, equipment number nodes pointing to process parameter nodes, and process parameter nodes pointing to work reporting nodes are all linked. For each associated edge (directed edge), a consistency index is calculated based on the corresponding dimension data in the operation context vector and the pre-stored business rule base within the MES system. A consistency index closer to 1 indicates a higher degree of matching between the business objects at both ends of the associated edge, while a consistency index closer to 0 indicates a lower degree of matching. Specifically: the product model code and BOM version number of the current operation are extracted from the operation context. The list of all currently effective BOM versions for that product model is queried in the MES system. If the currently selected BOM version number is in the effective version list, the consistency index from product model to BOM version is 1. If the currently selected BOM version number is not in the effective version list but exists in the historical version list of that product model, the consistency index from product model to BOM version is 0.3, reflecting an intermediate state where the version has expired but was still valid. If the currently selected BOM version number is neither in the effective version list nor in the historical version list, the consistency index from product model to BOM version is 0.

[0044] Extract the current BOM version number and process route number from the operation context vector, and query the set of available process routes associated with this BOM version in the MES system. If the current process route number is in the set of available process routes, the consistency index from product model to BOM version is 1. If the current process route number is not in the set of available process routes, but the process route has the same process sequence in terms of step composition as a standard process route in the set, the consistency index from product model to BOM version is 0.5. If the current process route is completely unrelated to the BOM version, the consistency index from product model to BOM version is 0.

[0045] Extract the current process route number and equipment number from the operation context vector, query the constraints of the process route on equipment capabilities in the MES system, including indicators such as equipment accuracy level, processing capability range and equipment type matching degree, and compare the current equipment capability parameters with the capability parameters required by the process route item by item. If all indicators meet the requirements, the consistency index from the process route to the equipment number is 1; if one indicator does not meet the requirements, the consistency index from the process route to the equipment number is 0.5; if two or more indicators do not meet the requirements, the consistency index from the process route to the equipment number is 0.

[0046] Extract the current device number and the set of nominal values ​​of process parameters from the operation context vector. Query the capability range of the device in the MES system. The capability range includes the acceptable minimum and maximum values ​​of each process parameter. For each process parameter, compare its nominal value with the device capability range. If the nominal value falls within the preferred range of the device capability range, the consistency contribution value of the parameter is 1; if the nominal value falls within the acceptable range but exceeds the preferred range, the consistency contribution value of the parameter is 0.7; if the nominal value exceeds the device capability range, the consistency contribution value of the parameter is 0. Sum the consistency contribution values ​​of all process parameters and divide by the total number of process parameters to obtain the consistency index of the association edge from the device number to the process parameter.

[0047] In this embodiment, the consistency index of process parameters to work reporting is not calculated before the current operation is submitted. After the operator performs the work reporting operation, the key fields in the work reporting data are checked to see if they are completely filled. If all are filled, the consistency index of process parameters to work reporting is 1. If any key field such as work reporting quantity or quality status is missing, the consistency index of process parameters to work reporting is 0.6. If two or more fields are missing, the consistency index of process parameters to work reporting is 0.

[0048] After the consistency index of the above five associated edges is calculated, the consistency indices of all associated edges are organized into the corresponding off-diagonal elements of the business association matrix to form the business association matrix. The association matrix is ​​a square matrix of six rows and six columns, with rows and columns arranged in the order of product model, BOM version, process route, equipment number, process parameters, and work report. The diagonal elements of the matrix are always 1, indicating that the nodes themselves are completely consistent. For rows and columns of work report nodes that have not been calculated, they will be automatically added after the work report is completed. After the business association matrix is ​​constructed, it is pushed to the evolution monitoring module.

[0049] The evolution monitoring module, based on the business association matrix, identifies whether business consistency exhibits a continuously deteriorating trend by tracking the temporal changes of the consistency index within the matrix as the operational process progresses. Specifically, this includes:

[0050] After each operator completes input for a field, all calculated valid consistency indices are extracted from the business association matrix. If there are q associated edges at the current time, the corresponding consistency index values ​​are denoted as L1 to Lq. The consistency indices are summed and then divided by the number of associated edges q to obtain the business consistency index at the current time. The business consistency index ranges from 0 to 1. The closer the value is to 1, the higher the overall matching degree between the current business objects and the better the business consistency. The lower the value, the more conflicts or mismatches have occurred between the business objects.

[0051] The business consistency index of the most recent K moments is extracted to form a time series, where K is a positive integer. In this embodiment, K is 10, meaning the business consistency index is tracked for the most recent 10 monitoring moments. Based on the time series, the consistency change rate between two adjacent moments is calculated, which is the difference between the business consistency changes of two adjacent moments and divided by the time interval between the two moments. When the consistency change rate is negative, it indicates that the business consistency has decreased within the current monitoring interval. The number of consistency change rates less than zero is counted and divided by K to obtain the consistency deterioration index. The consistency deterioration index ranges from 0 to 1, representing the proportion of monitoring times with decreased consistency in the most recent K monitoring sessions out of the total number of monitoring sessions. When the consistency deterioration index is greater than or equal to a preset deterioration trend threshold, it is determined to be a continuous deterioration trend. In this embodiment, the deterioration trend threshold is set to 0. A value of 0.6 indicates that the consistency index has decreased in more than 6 out of the last 10 monitoring sessions, indicating a clear and continuous deterioration trend in business consistency. Once this trend is confirmed, the decreasing correlation edges leading to the consistency decline are further located. All business consistency indices in the current business correlation matrix are traversed, and decreasing correlation edges that meet the following two conditions are added to the decreasing path set: first, the business consistency index of the edge is less than a preset consistency alarm threshold, which is 0.5 in this embodiment; second, the business consistency index of the edge is less than the business consistency index at the current moment than at the previous moment. Each decreasing correlation edge in the decreasing path set represents a business relationship in the current operation that has shown significant inconsistency and is continuously deteriorating. The decreasing path set is then pushed to the propagation prediction module as the starting basis for error propagation prediction.

[0052] The propagation prediction module, based on the set of descent paths and a pre-built error propagation graph, predicts the probability that currently activated error types will propagate along the business-related chain to other business processes, and calculates the error propagation index, specifically including:

[0053] An error propagation graph is constructed during system initialization. This graph is a directed graph structure, where nodes represent business error types and directed edges represent the possibility of an error propagating from one node to another. In this embodiment, the error propagation graph contains six error type nodes: product error, BOM error, process error, equipment error, parameter error, and reporting error. The propagation relationship between each error type is determined based on the reverse inference of the business dependency chain: product errors may lead to BOM errors, BOM errors may lead to process errors, process errors may lead to equipment errors or parameter errors, and parameter errors may lead to reporting errors. The direction of the above propagation relationship is consistent with the direction of the business dependency chain, that is, the error of the upstream node will propagate downstream along the business chain.

[0054] The weight of each directed edge in the error propagation graph is a conditional probability, representing the probability that one error type will occur after another has occurred. The initial values ​​of the edge weights are calculated statistically based on the occurrence sequence of historically confirmed errors in the MES system over the past 24 months. Specifically, for any two error types i and j, the module counts the number of times error j occurs in subsequent operations after error i has occurred in all error records. Divide by the total number of times error i occurs We obtain the conditional probability of error i propagating to error j. It should be noted that the conditional probability is statistically significant after 24 months of data statistics. As new error records continue to be generated during the operation of the system, the above edge weights are recalculated every month so that the propagation graph can reflect the latest error evolution pattern.

[0055] A mapping relationship is established between each associated edge in the descent path set and the error type nodes in the error propagation graph. Specifically, when the consistency between product model and BOM version decreases, the two nodes of product error and BOM error are activated; when the consistency between BOM version and process route decreases, the two nodes of BOM error and process error are activated; when the consistency between process route and equipment number decreases, the two nodes of process error and equipment error are activated; when the consistency between equipment number and process parameters decreases, the two nodes of equipment error and parameter error are activated; when the consistency between process parameters and work report decreases, the two nodes of parameter error and work report error are activated. The set of all activated nodes is denoted as A. For each potential target node j in the error propagation graph that is not activated but has a directed edge from an activated node to it, the probability of activation of the potential target node j is calculated. Since all activated nodes may propagate errors to the predicted target node through their respective propagation edges, and each propagation path is assumed to be independent, the activation probability of the potential target node is the complement of the failure probabilities of each path:

[0056]

[0057] Where A is the set of nodes with activated error types. For error type i to error type j

[0058] The propagation edge weights, all activated error nodes have the opportunity to propagate the error to the potential target node, and the probability that the potential target node j is eventually activated is equal to the probability that at least one path has successfully propagated; it should be noted that for any activated node i in A, along the directed edge direction of the error propagation graph, all nodes that can be reached from i through one or more propagation edges constitute the potential target node set B. Only potential target nodes that meet the following two inclusion conditions are identified as predicted target nodes: (1) Path distance condition: The propagation path distance between the predicted target node and at least one activated node must not exceed 2, that is, starting from an activated node, along the direction of the directed propagation edge, it can reach the node after no more than two steps of propagation; Statistical analysis in actual production sites shows that after the error propagates along the business dependency chain for more than two steps, the propagation probability has decayed to a low level, and the reference value for real-time intervention decision-making of the current operation is limited. Including distant nodes that are more than two steps away will introduce noise; (2) Activation probability threshold condition: The activation probability of the predicted target node j The activation probability must be greater than or equal to a preset activation probability threshold. In this embodiment, the activation probability threshold is set to 0.1. When the activation probability of the potential target node... When Pj is less than 0.1, the probability of activation is low, and even if it is included in the summation calculation, its contribution to error propagation is extremely limited. To improve computational efficiency, it is only included in the set of predictable target nodes when Pj reaches 0.1 or above. A comprehensive evaluation is performed on the currently activated error propagation paths, and the error propagation index E is calculated according to the following formula:

[0059]

[0060] in, Let be the activation probability of the i-th node in the current propagation path, and n be the total number of activated nodes and the predicted target node in the current propagation path. To assign a hazard weight to each node in error propagation, the preset values ​​for each node's hazard weight in this embodiment are: product error 0.20, BOM error 0.25, process error 0.20, equipment error 0.10, parameter error 0.15, and reporting error 0.10, with the sum of these weights being 1. The allocation of hazard weights is based on the following criteria: BOM errors have the widest impact on subsequent production chains, therefore they are assigned the highest weight of 0.25; product errors and process errors have the next widest impact, each assigned 0.20; the impact of parameter errors on quality needs to be considered in conjunction with the importance of specific parameters, and is therefore assigned a comprehensive weight of 0.15; the impact of equipment errors and reporting errors is relatively limited and mostly immediate, each assigned 0.10; the error propagation index E ranges from 0 to 1. An error propagation index E of 0 indicates no current error propagation risk; the larger the error propagation index E, the wider the propagation stage of the error, and the higher the risk of subsequent business processes being affected.

[0061] The intelligent intervention module generates tiered intervention strategies based on the error propagation index and the set of currently activated path nodes. The set of currently activated path nodes C is the union of the set of already activated nodes A and the set of predicted target nodes B; specifically, it includes:

[0062] A preset error propagation range is defined. In this embodiment, the upper limit of the error propagation range is set to 0.7 and the lower limit is set to 0.4. When the error propagation index is less than 0.4, the current operation is determined to be in a low-risk range, no intervention action is triggered, and the operation is directly submitted and released.

[0063] When the error propagation index is greater than or equal to 0.4 and less than 0.7, the system is considered to be in a medium-risk state, triggering a prompt-level intervention. This intervention involves displaying a prompt message in the sidebar of the current operating interface in a non-blocking manner. Specifically, the prompt message content is generated based on the trigger descent path information carried in the current active path node set C. Specifically, for nodes in the current active path node set C that are already active, their corresponding descent associated edges are extracted, and readable prompt text is generated for these edges. For example, when a product error node is activated due to a decrease in product-to-BOM consistency, a prompt is generated: "A new version of the currently selected BOM is available; it is recommended to confirm whether to switch." When a device error node is activated due to a decrease in process-to-device consistency, a prompt is generated: "The current device capability does not meet the process route requirements; it is recommended to replace the device." The prompt message is displayed with a semi-transparent background in the sidebar or bottom status bar of the operating interface, without obscuring any input controls. The operator can completely ignore the prompt and continue submitting the operation without requiring any confirmation response. The purpose of this level of intervention is to provide reference information to the operator without interrupting the operation flow, assisting the operator in discovering and correcting potential errors independently.

[0064] When the error propagation index is greater than or equal to 0.7 and the number of nodes in the current active path node set C is less than or equal to 3, it is judged as a high-risk but limited propagation range, triggering a confirmation-level intervention. The specific execution method is as follows: a modal confirmation dialog box pops up in the center of the current operation interface. The dialog box lists the business consistency issues corresponding to the nodes in the current active path node set C that are in the activated state one by one. For nodes in the predicted target state, the dialog box is displayed in the form of a prediction prompt, clearly informing the operator that if the current inconsistency issues are not corrected, it is expected to further affect the business processes. Each activated issue occupies a separate line, and the content includes the name of the issue's associated edge, the current actual value, and the system's suggested correct value. The suggested correct value is extracted from the MES system's business rule base according to the type of the descending associated edge: for the descending associated edge from the product to the BOM, the latest BOM version currently in effect for that product is taken. This level of intervention involves several steps: For descending edges from the BOM to the process, a standard process route compatible with the current BOM is selected; for descending edges from the process to the equipment, idle equipment that meets the process requirements is selected. The dialog box has two buttons at the bottom: "Accept All Suggestions Item by Item" and "Confirm and Continue Item by Item." After the operator clicks "Accept All Suggestions Item by Item," all suggested correction values ​​are automatically filled into the corresponding input fields, and the dialog box closes. The operator can then continue with subsequent operations. Clicking "Confirm and Continue Item by Item" highlights each issue in the dialog box, requiring the operator to choose to accept the correction or retain the current value for each issue. The dialog box can only be closed after all items have been selected. This level of intervention requires the operator to actively confirm before continuing with subsequent operations. There is no automatic timeout to ensure that the operator is aware of all detected business inconsistencies and has made a clear choice before submission.

[0065] When the error propagation index is greater than or equal to 0.7 and the number of nodes in the current active path node set C is greater than 3, it is judged as a high-risk interval with a wide propagation range, triggering a correction-level intervention. The specific execution method is as follows: based on the descent association edge information carried by the nodes in the current active path node set C that are in an activated state, automatic correction operations are performed one by one according to the following priority order: the highest priority is the product-to-BOM association edge, where the correction operation is to replace the current BOM version with the latest version number currently in effect for that product model; the second highest priority is the BOM-to-process association edge, where the correction operation is to replace the current process route with a standard process route that is compatible with the current BOM version and is in an activated state; the third highest priority... The first priority is the process-to-equipment association edge, where the correction operation is to replace the current equipment number with equipment that meets the process route capacity requirements and is currently idle. The fourth priority is the equipment-to-process parameter association edge, where the correction operation is to adjust the current value of each process parameter to the system-recommended value. The system-recommended value is obtained by: firstly querying the historical production data of the parameter under the current equipment conditions; if there are numerical records positively correlated with product quality, then taking the average value of the parameter corresponding to the historical best quality as the correction value; if there is no such historical data, then taking the nominal value of the parameter as the correction value. The lowest priority is the process parameter-to-reporting association edge, where the correction operation is to fill in the missing values ​​of key fields required for reporting. Each item is automatically... After the correction operation is executed, the original value before correction and the new value after correction are recorded in the correction log in pairs. For the predicted target node in the current active path node set C, no direct correction is performed, but the operator will be informed of the potential propagation path that has been blocked by the current correction in the form of a prediction prompt in the subsequent correction report. After all automatic corrections are completed, a correction report pops up in the operation interface. The correction report lists the associated edge name, original value, correction value, and correction basis for each correction operation in tabular form. The correction report also displays the predicted target node information in the current active path node set C. For example, if the BOM version is not corrected, it is expected to lead to process route errors. This correction simultaneously blocks the propagation path. The correction report will be displayed for at least 10 seconds to ensure that operators have enough time to review all corrections. During the display, all input controls on the interface will be in read-only mode, and operators will not be able to perform any input operations. After the correction report is displayed, a final confirmation dialog box will pop up. For correction-level interventions, a single confirmation button will not be provided. Instead, operators will be required to confirm each automatically corrected item in the dialog box, i.e., confirm each item and continue. Operators must click to confirm each item to execute the correction before they can continue. If an operator believes that a correction is incorrect, they can choose to roll back the correction. Only that field will be restored to its original value before the correction, while other corrections will be retained.

[0066] It is important to note that when any of the following scenarios are included in the set of consistency degradation paths, the intervention level will be directly escalated to the correction level: The first scenario is a consistency degradation from product model to BOM version with a consistency index of 0 for that associated edge, meaning the current BOM version number and product model are completely mismatched; the second scenario is a consistency degradation from equipment number to process parameters with the current input value of any critical process parameter exceeding the equipment's safe operating range, which is pre-set by the equipment's factory parameters or the company's equipment management system; the third scenario is a mismatch between the material code and the current BOM version, and the material code is not in the alternative material list for that BOM version. These three mandatory escalation rules ensure that the error types most severely impacting production safety and product quality will not be abandoned by the system based on probability judgments under any circumstances.

[0067] The above are merely embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A MES intelligent error prevention and early warning system, characterized in that, include: The input capture module is used to establish a business session when the operator performs business operations, collect personnel objects, task objects, process objects and real-time behavior objects, and assemble them into operation data; The behavior modeling module is used to receive the operation data, construct a stable context, a real-time context, and a task context, and then weightedly fuse the three to generate a comprehensive operation context. The business association module is used to establish a business association network based on the dependency relationship between business objects in the MES system, and to calculate the consistency index of each business association edge according to the comprehensive operation context. The evolution monitoring module is used to track the temporal changes of the consistency index of each business-related edge, and to locate the path of consistency decline when it is determined that the consistency shows a continuous deterioration trend. The propagation prediction module is used to map the descent associated edges in the consistency descent path to activated nodes, and predict the activation probability of the activated nodes propagating to other nodes along the business association chain based on the pre-built error propagation graph. Potential target nodes that meet preset conditions are used as predicted target nodes. The error propagation index is calculated based on the activation probabilities of the activated nodes and the predicted target nodes to generate a current activated path node set, which is the union of the activated nodes and the predicted target nodes. The intelligent intervention module is used to generate and execute tiered intervention strategies based on the error propagation index.

2. The MES intelligent error prevention and early warning system according to claim 1, characterized in that, The personnel objects include the operator's unique identifier, current job title, total number of historical operations, and job authorization level; the task objects include the order number, product model, task type, and operation type; the process objects include the process route number, BOM version number, material code, and nominal values ​​of key process parameters. The real-time behavior objects include the field input order, the time interval between adjacent field inputs, the number of times each field is modified, the number of times the page is returned, and the number of times the BOM is switched.

3. The MES intelligent error prevention and early warning system according to claim 1, characterized in that, The stable context constructed by the behavior modeling module includes operation proficiency level, product coverage, process route usage preference, equipment usage preference, and standard input order. The stable context is updated at fixed intervals based on historical business operation records. The real-time context includes current input rhythm, field modification frequency, page jump frequency, BOM switching frequency, and parameter adjustment frequency. The real-time context is dynamically refreshed with each step of the operator's input. The task context includes production change status, order complexity coefficient, equipment switchover status, and production stage identifier. The task context is determined when a business session is established and remains unchanged during the business session.

4. The MES intelligent error prevention and early warning system according to claim 1, characterized in that, The business association network consists of business objects as nodes and the dependencies between business objects as directed edges. The business objects include product model, BOM version, process route, equipment number, process parameters, and work report. The business association edges include five association edges: product model to BOM version, BOM version to process route, process route to equipment number, equipment number to process parameters, and process parameters to work report. The consistency index is calculated by the business association module according to the business constraint rules pre-stored in the MES system and is used to quantify the matching degree between the business objects at both ends of each association edge.

5. The MES intelligent error prevention and early warning system according to claim 1, characterized in that, The evolution monitoring module tracks the business consistency index sequence of the most recent K monitoring times, calculates the consistency change rate between adjacent times, and determines whether the consistency shows a continuous deterioration trend based on the consistency continuous deterioration index. When it is determined to be a continuous deterioration trend, the consistency decline path is pushed to the propagation prediction module. The evolution monitoring module tracks the business consistency index sequence of the most recent K monitoring times, calculates the consistency change rate of adjacent times, and determines whether the consistency shows a continuous deterioration trend based on the consistency deterioration index. The consistency deterioration index is the ratio of the number of times the consistency change rate is negative in the most recent K monitoring times to K. When the consistency deterioration index is greater than or equal to the preset deterioration trend threshold, it is determined to be a continuous deterioration trend, and the consistency decline path is pushed to the propagation prediction module.

6. The MES intelligent error prevention and early warning system according to claim 1, characterized in that, The propagation prediction module constructs the error propagation graph during system initialization. The error propagation graph is a directed graph structure, where nodes represent business error types and directed edges have weights to characterize the propagation probability between error types. Upon receiving the consistency descent path, the propagation prediction module maps each descent-related edge in the consistency descent path to a corresponding activated node, and predicts the activation probability of each potential target node along the direction of the error propagation graph, starting from the activated node. The activation probabilities of the activated nodes and the predicted target nodes are weighted and summed to obtain the error propagation index.

7. The MES intelligent error prevention and early warning system according to claim 6, characterized in that, The intelligent intervention module determines whether to perform a prompt-level intervention, a confirmation-level intervention, or a correction-level intervention based on the preset interval where the error propagation index is located and the number of nodes in the current active path node set. The prompt-level intervention displays prompt information in a non-blocking manner, the confirmation-level intervention requires the operator to confirm each item in a modal dialog box, and the correction-level intervention is automatically performed by the system and requires the operator to confirm each item.

8. The MES intelligent error prevention and early warning system according to claim 7, characterized in that, In the correction-level intervention, the intelligent intervention module performs automatic correction operations in sequence according to the priority order of product to BOM, BOM to process, process to equipment, equipment to parameters, and parameters to work report, based on the descending associated edges carried by the activated nodes in the current active path node set. The original value before correction and the new value after correction are recorded in the correction log for the operator to confirm item by item.

9. The MES intelligent error prevention and early warning system according to claim 1, characterized in that, The intelligent intervention module will directly perform a corrective intervention when any of the following situations occur: the consistency index between the product model and the BOM version is 0; the consistency index between the equipment number and the process parameters is 0 and the process parameters involved exceed the safe operating range of the equipment. The material code does not match the current BOM version and is not in the alternative material list for this BOM version.

Images