MES-based production data acquisition and analysis system

Abstract

The invention relates to the technical field of data analysis, and particularly discloses an MES-based production data acquisition and analysis system. Comprising a topology construction module used for collecting the operation rhythm of key equipment and constructing a collaborative topology baseline of an operation rhythm sequence; the feature fusion module is used for carrying out topological characterization and stability fusion on equipment collaborative fluctuation, generating a collaborative skewness index and a kurtosis collaborative coefficient, and forming collaborative modal features; the resonance analysis module is used for constructing a skewness-kurtosis phase diagram, judging systematic fluctuation resonance through a resonance triggering criterion, and identifying a fluctuation event cluster and a driving mode thereof; and the convergence regulation and control module carries out target node differentiation regulation and control based on a fluctuation amplification intensity link, updates a collaborative topology baseline, and realizes system adaptive convergence. According to the method, the fluctuation in production collaboration can be effectively identified, and the stability and controllability of a manufacturing system are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, and particularly relates to a production data acquisition and analysis system based on MES. BACKGROUND

[0002] As a core information system connecting enterprise planning layer and shop floor control layer, the manufacturing execution system (MES) of modern manufacturing industry can provide visual monitoring means for production management by acquiring real-time production site data, and becomes a key technology for improving manufacturing process transparency and execution efficiency.

[0003] At present, the production data acquisition based on MES can realize acquisition and storage of massive information such as equipment state, production progress and material consumption. The traditional analysis method is difficult to deeply mine and quantitatively represent the dynamic coordination relationship between multiple key devices on the production line. In actual production, the devices are not independently operated, but are tightly coupled through the process flow. The beat fluctuation of any device may be propagated and amplified in the system through complex dependence relationship, eventually leading to the decline of coordination efficiency of the whole production line, imbalance of production beat, and even causing chain stoppage.

[0004] Therefore, the present application provides a production data acquisition and analysis system based on MES. SUMMARY

[0005] The present application aims to provide a production data acquisition and analysis system based on MES to solve the above background problems.

[0006] The purpose of the present application can be achieved by the following technical solutions:

[0007] A production data acquisition and analysis system based on MES, comprising the following modules:

[0008] A topology construction module is used to acquire the running beat of the key device, establish a running beat sequence, perform production system coordination analysis based on the running beat sequence, and establish a coordination topology baseline between the key devices;

[0009] A feature fusion module is used to perform topology representation and coordination correction processing on the device coordination fluctuation based on the coordination topology baseline, obtain a coordination skewness index and a peak coordination coefficient respectively, and obtain a coordination modal feature through feature fusion;

[0010] A resonance analysis module is used to construct a skewness-peak state phase diagram and a corresponding resonance triggering criterion, determine whether the coordination modal feature has systematic fluctuation resonance, construct a fluctuation event cluster if there is, and analyze the amplification strength link and the corresponding driving mode of the fluctuation;

[0011] Convergence regulation module: based on the fluctuation event cluster and the amplification intensity link, the convergence target node is extracted, the skewness and the peak state of the convergence target node are differentiated and regulated, and the collaborative topology baseline is updated.

[0012] As a further scheme of the application: the manner of establishing the collaborative topology baseline is:

[0013] A dynamic time sequence dependent network is constructed, a community discovery algorithm is used to identify the device collaborative community in the network, and a core collaborative group in the community is screened out;

[0014] The node degree centrality value and the betweenness centrality value of the core collaborative group in the network under normal production state are obtained as characteristic parameters;

[0015] Based on the characteristic parameters, a collaborative topology baseline representing the collaborative state of the production system is established.

[0016] As a further scheme of the application: the manner of constructing the dynamic time sequence dependent network is:

[0017] The production time of the key device is divided into N equal length time windows;

[0018] For each time window, the running beat sequence of each key device in each sliding window is extracted, and the cross-correlation function between the running beat sequences of any two key devices of the production line is calculated;

[0019] Based on the cross-correlation functions of all time windows, the maximum value of the cross-correlation function calculated for each pair of key devices in all windows is obtained;

[0020] The T test is used to determine whether the mean value of the maximum value of the cross-correlation function is significantly positive, and the significance index P value is obtained;

[0021] If it is significantly positive, the cross-correlation function mean value is multiplied by the significance coefficient determined by the P value to generate a directed dependent edge weight;

[0022] The initial time sequence dependent graph is constructed by taking the key device as the node and the directed dependent edge as the connection.

[0023] As a further scheme of the application: the manner of performing topology representation and collaborative correction processing is:

[0024] A collaborative deviation sequence of the key device in the current time window is established;

[0025] The skewness coefficient of the collaborative deviation sequence is calculated, and the peak state collaborative correction is performed in combination with the network structure information of the topology baseline, to obtain a collaborative skewness index; the excess kurtosis of the collaborative deviation sequence is calculated, and the correction is performed in combination with the production beat stability, to obtain a peak state collaborative coefficient:

[0026] The skewness synergy index and the kurtosis synergy coefficient are taken as the synergy modality features.

[0027] As a further scheme of the present application, the peakness synergy correction is performed in the following manner:

[0028] The betweenness centrality value of each key device is obtained from the synergy topology baseline, and the number density of the dependent edges in the core synergy group corresponding to the current key device is obtained;

[0029] The skewness coefficient is multiplied by the betweenness centrality value, and the product result is summed with the skewness coefficient to obtain the betweenness skewness coefficient;

[0030] The betweenness skewness coefficient is multiplied by the number density to obtain the skewness synergy index.

[0031] As a further scheme of the present application, the skewness synergy index is established in the following manner:

[0032] For each key device of the core synergy group, a beat interval sequence is established;

[0033] The difference between the beat interval sequence and the synergy beat reference sequence in each corresponding beat cycle is calculated to obtain a beat cycle difference;

[0034] The sum of the dependent edge weights of the key device and other key devices in the core synergy group in the topology baseline is obtained, the beat cycle difference is normalized, and the skewness synergy sequence of the key device in the current time window is established.

[0035] As a further scheme of the present application, the determination of whether the synergy modality features exist systematic fluctuation resonance is performed in the following manner:

[0036] Based on the historical distribution of the skewness-kurtosis phase diagram, a dynamic confidence ellipse is established according to a preset confidence interval;

[0037] A resonance triggering criterion is constructed based on the dynamic confidence ellipse, if the feature point meets the resonance triggering criterion, it is determined that the synergy modality features exist systematic fluctuation resonance beyond the historical confidence interval, if it does not meet, it is determined that it does not meet the systematic fluctuation resonance beyond the historical confidence interval.

[0038] As a further scheme of the present application, the skewness-kurtosis phase diagram is constructed in the following manner:

[0039] The skewness synergy index is taken as the horizontal axis, and the kurtosis synergy coefficient is taken as the vertical axis to construct the skewness-kurtosis phase diagram, and the synergy modality features are taken as feature points on the skewness-kurtosis phase diagram.

[0040] As a further scheme of the present application, the process of analyzing the driving mode is as follows:

[0041] The number of independent fluctuation sources identified in the fluctuation event cluster is acquired, and the fluctuation propagation path structure of the current time window is compared with the fluctuation propagation paths of the previous M continuous time windows to calculate the path structure similarity;

[0042] Based on the path structure similarity and the number of independent fluctuation sources, the driving mode is determined, and the corresponding driving mode is identified.

[0043] The driving mode includes a source driving amplification path and a condition coupling fluctuation structure.

[0044] As a further scheme of the application, the mode of the skewness and peakness differentiation regulation is:

[0045] For all convergent target nodes, based on the amplification strength output by the resonance analysis module, an amplification strength gradient matrix is constructed.

[0046] For each convergent target node, the beat interval time sequence and the beat coefficient of variation defined in the feature fusion module are acquired.

[0047] The change rate of the convergent target node beat coefficient of variation in the continuous M time windows is calculated to obtain the beat instability rate.

[0048] If the beat instability rate is positive, it is marked as a positive diffusion node, otherwise it is a call-back node.

[0049] According to the amplification strength gradient of the amplification strength gradient matrix, the convergent priority function is formed in combination with the node betweenness centrality.

[0050] Based on the convergent priority function and the positive diffusion node, the skewness and peakness convergence regulation strategy is executed for different driving modes.

[0051] The beneficial effects of the application are as follows:

[0052] 1. By constructing a dynamic time sequence dependent network, the internal coordination structure and core coordination group of the production system are identified from the time sequence data. This is beneficial to reduce the dependence on prior process knowledge, and the coordination topology baseline is beneficial to represent the dependence relationship between devices, and the dynamic updating feature of the coordination topology baseline can capture the slow evolution of the system coordination relationship, providing a reference benchmark for subsequent state monitoring and abnormal diagnosis.

[0053] 2. By generating cooperative skewness indices and kurtosis cooperative coefficients, single-series deviation information is transformed into cooperative modal characteristics influenced by network structure. This enhances the dimensionality and depth of the cooperative state description, enabling the system to identify fluctuations in critically located devices within the topological network. By constructing skewness-kurtosis phase diagrams and multi-dimensional resonance triggering criteria, the system diagnoses cooperative instability risk modes. This helps determine whether the system deviates from its normal state; by analyzing the movement trajectories and spatial clustering of feature points, it distinguishes whether fluctuations originate from a single device or from conditional coupling between multiple devices.

[0054] 3. Achieve a closed loop from analysis to control. Extract convergence target nodes based on the fluctuation propagation path and execute differentiated control strategies for skewed and kurtosis. By distinguishing between forward diffusion nodes and reversible nodes and implementing corresponding beat adjustment or fluctuation smoothing strategies, reduce secondary problems that may be caused by control. Update the collaborative topology baseline to enable the system to have adaptive evolution capabilities, forming a loop that continuously maintains production collaborative stability in a dynamic environment. Attached Figure Description

[0055] The invention will now be further described with reference to the accompanying drawings.

[0056] Figure 1 This is a functional block diagram of a production data acquisition and analysis system based on MES according to the present invention;

[0057] Figure 2 This is a flowchart of the process for determining whether there is a systemic fluctuation resonance that exceeds the historical confidence interval in this invention;

[0058] Figure 3 This is a flowchart of a production data acquisition and analysis method based on MES in this invention. Detailed Implementation

[0059] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0060] Example 1

[0061] like Figure 1 As shown, please refer to Figure 1 As shown, this invention is a production data acquisition and analysis system based on MES, comprising the following modules:

[0062] Topology building module: used to obtain the operating cycle time of key equipment, establish the operating cycle time sequence, perform collaborative analysis of the production system based on the operating cycle time sequence, and establish a collaborative topology baseline between key equipment;

[0063] The method for obtaining the operating cycle time of key equipment and establishing the operating cycle time sequence is as follows:

[0064] Preferably, the operating cycle sequence of key equipment on the production line that has completed timing alignment is obtained through MES (Manufacturing Execution System);

[0065] It should be noted that the aforementioned operating cycle sequence refers to the time sequence formed by the periodic start times of each device completing a work unit (such as processing a part);

[0066] For example, key equipment includes stamping machines and welding robots, and the cycle sequence can be calculated from event logs such as "work order start" and "processing completed" provided by the equipment controller or MES system;

[0067] Meanwhile, the beat sequence of the first device in the production line to start performing core processing actions (such as the start signal of the first robot grabbing a part on the assembly line) is used as the main beat reference sequence of the production line;

[0068] The method for constructing the collaborative topology baseline is as follows:

[0069] The production equipment is divided into groups according to the process flow (such as feeding unit, processing unit, and testing unit). Each equipment group is used as an analysis unit to determine the unit to which each key piece of equipment belongs.

[0070] The production time of key equipment is divided into N equal-length time windows, with N ranging from 500 to 2000, preferably N=1000;

[0071] For each time window, extract the operating cycle sequence of each key device within each sliding window, and calculate the cross-correlation function between the operating cycle sequences of any two key devices on the production line (denoted as device A and device B);

[0072] It should be noted that the sliding step size is 5% to 20% of the window length.

[0073] The maximum value of the cross-correlation function for each pair of critical devices (device A and device B) is calculated across all time windows based on the cross-correlation function of all time windows.

[0074] Use the T-test to determine whether the mean of the maximum values ​​of the cross-correlation functions is significantly positive;

[0075] It should be noted that the method for determining whether a value is significantly positive is as follows: calculate the mean of the maximum values ​​of all cross-correlation functions, i.e., the P-value; if the P-value is less than the preset significance level (usually 0.05), then it is determined to be significantly positive.

[0076] If the significance is positive, the mean of the cross-correlation function is multiplied by the significance coefficient determined by the p-value to generate the directed dependency edge weights.

[0077] Using key equipment as nodes and directed dependency edges as connections, an initial temporal dependency graph is constructed. Dynamically updated weights are then added to the initial temporal dependency graph to establish a dynamic temporal dependency network.

[0078] Preferably, the method for establishing a dynamic temporal dependency network is as follows: after the temporal dependency graph is constructed, the time window is slid backward by one step (preferably, the sliding step is 10% of the window length, that is, when N=1000, 100 windows are slid), and the above calculation process is repeated to realize the dynamic update of dependency weights and form a dynamic temporal dependency network.

[0079] Based on the dynamic temporal dependency network, the time period recorded in the historical database as normal production is selected, and the arithmetic mean and variance of the weights of each dependency edge in the network during the normal production time period are calculated as the features corresponding to the steady-state dependency mode of the system.

[0080] The community discovery algorithm identifies device collaboration clusters in the network and filters out the core collaboration groups within the clusters.

[0081] Preferably, the Louvain method in the community detection algorithm is adopted. By optimizing the modularity index of the network, the network is divided into several communities with tight internal connections and sparse external connections. The resolution parameter in the Louvain method is optimized and determined by combining the grid search method with the modularity index.

[0082] The method for selecting the core collaborative groups within a community is as follows: from the identified communities, communities with an average weight of dependent edges within the community that is greater than a preset threshold (e.g., 0.7) and a weight variance that is less than a preset threshold (e.g., 0.1) are selected and defined as the core collaborative groups.

[0083] It should be noted that, based on historical normal production data statistics: the core collaborative group dependency edge weight data of 1000 normal production time windows were collected, and their statistical distribution was calculated; the average weight threshold was set to the 75th percentile (example 0.7) to ensure that closely collaborative groups were selected; the variance threshold was set to the 25th percentile (example 0.1) to ensure stable collaboration within the group;

[0084] The number of dependency edges directly connected to key equipment nodes in the dynamic temporal dependency network of the core collaboration group under normal production conditions is used as the node degree centrality value.

[0085] Based on the dynamic temporal dependency network and the selected core cooperative group, a local network structure is formed with key devices in the core cooperative group as nodes and directed dependency edges between devices as connections. The betweenness centrality value of a node is obtained by calculating the proportion of the number of shortest paths in the local network that pass through the current key device node to the total number of shortest paths.

[0086] Obtain the node degree centrality and betweenness centrality values ​​of the core collaborative group network under normal production conditions as feature parameters;

[0087] Based on characteristic parameters, a collaborative topology baseline characterizing the collaborative state of the production system is established;

[0088] Understandably, the purpose of constructing a collaborative topology baseline is:

[0089] Objective 1: To quantitatively characterize the core structural features of equipment collaborative networks. The degree centrality value of a node reflects the strength of the direct collaborative association between equipment and other equipment, while the betweenness centrality value reflects the intermediary role of the equipment in the collaborative link. Together, they constitute the collaborative structure benchmark under normal production conditions, providing a basis for comparing whether the collaborative relationship between equipment deviates from the steady state during real-time monitoring.

[0090] Objective 2: To improve the targeting of convergence control, during the control phase, the priority of convergence target nodes can be determined based on the steady-state values ​​of the two parameters in the baseline (such as prioritizing the control of hub nodes with high betweenness centrality), ensuring that the control actions focus on the equipment that has the greatest impact on the coordinated stability, and reducing the interference of blind control on the system.

[0091] Feature fusion module: used to perform topological characterization and collaborative correction processing on equipment collaborative fluctuations based on collaborative topological baseline, to obtain collaborative skewness index and kurtosis collaborative coefficient, and to obtain collaborative modal features through feature fusion;

[0092] The process of identifying the cooperative modal characteristics of the equipment operating cycle sequence based on the cooperative topology baseline is as follows:

[0093] Based on the core collaborative group determined by the collaborative topology baseline, the start time point of each working cycle of each key device in the group is identified from the synchronous operation status signal, and the difference between adjacent start time points is calculated to obtain the complete cycle interval time series of each key device.

[0094] Based on the currently set analysis time window (e.g., the window length is T=1000 main beat cycles), extract the segment corresponding to the time window from the beat interval time series as the beat interval sequence of the key equipment in the current window;

[0095] Obtain the node degree centrality values ​​of each key device in the topology baseline;

[0096] The beat interval sequence of all key equipment in the core coordination group in the current time window is weighted and summed (the node degree centrality of each key equipment in the topology baseline is used as the weight) to generate a coordination beat reference sequence representing the overall rhythm of the core coordination group.

[0097] For each key device in the core coordination group, the difference between the beat interval sequence and the coordination beat reference sequence at each corresponding beat cycle is calculated to obtain the beat cycle difference;

[0098] Obtain the sum of the dependency edge weights of key equipment and other key equipment in the core collaboration group in the topology baseline, normalize the cycle difference, and establish the collaboration deviation sequence of key equipment in the current time window;

[0099] Feature offset recognition processing is performed based on the cooperative deviation sequence to obtain cooperative modal features;

[0100] The method for performing feature offset recognition is as follows:

[0101] The skewness coefficient of the cooperative deviation sequence is calculated and the kurtosis cooperative correction is performed by combining the network structure information of the topological baseline to obtain the kurtosis cooperative coefficient.

[0102] The method for performing peak-state collaborative correction is as follows:

[0103] Obtain the betweenness centrality value of each critical device from the collaborative topology baseline, as well as the number density of dependency edges in the core collaborative group corresponding to the current critical device;

[0104] The number density is obtained by calculating the ratio of the actual number of dependent edges within the core collaborative group to the maximum number of dependent edges.

[0105] The skewness coefficient is multiplied by the betweenness centrality value, and the product is summed with the skewness coefficient to obtain the betweenness skewness coefficient.

[0106] Multiplying the betweenness skewness coefficient by the number density yields the cooperative skewness index;

[0107] Calculate the excess kurtosis (i.e., actual kurtosis - 3) of the coordinated deviation sequence, and correct it based on production cycle stability to obtain the kurtosis coordination coefficient:

[0108] The method for correction based on production cycle stability is as follows:

[0109] Calculate the production cycle variation coefficient of the key equipment in the current time window, which is the ratio of the standard deviation of the cycle interval time to the average value. Take the reciprocal of the production cycle variation coefficient and add one to obtain the cycle reference coefficient, which is used to characterize the stability of the current production cycle. The higher the stability, the larger the production cycle variation coefficient.

[0110] It should be noted that the reciprocal plus 1 is used in the calculation to avoid the fact that the coefficient of variation is meaningless when it is 0;

[0111] The collaborative stability factor is obtained from the collaborative topology baseline. The collaborative stability factor is calculated as the ratio of the actual collaborative degree of the collaborative group in the current time window to the average collaborative degree during the historical normal production period.

[0112] The kurtosis coordination coefficient is obtained by multiplying the coordination stability factor, excess kurtosis, and beat reference coefficient.

[0113] It should be noted that the physical meaning of the cooperative skewness index is: an indicator that quantifies the direction of asymmetric fluctuations in a production system and their impact on topological transmission. By using the skewness coefficient of the cooperative deviation sequence, the dominant direction of the fluctuation deviation from the baseline is determined (positive values ​​indicate positive deviation, such as beat lag, and negative values ​​indicate negative deviation, such as beat lead). Combined with the betweenness centrality value of the cooperative topological baseline (which amplifies the weight of fluctuations in key equipment) and the number density of dependent edges (which reflects the deviation diffusion effect under close cooperation), it reflects which direction the fluctuations deviate from and how many related equipment will be affected by the deviation.

[0114] The physical meaning of the kurtosis coordination coefficient is: an indicator that quantifies the peak concentration of coordinated fluctuations and extreme risks in a production system. It takes the excess kurtosis of the coordinated deviation sequence as the core. A positive value indicates that the fluctuation contains a large number of extreme amplitudes (such as sudden beat deviations) and has a peak distribution. When it is close to 0, the fluctuation is smooth. Combined with the beat stability and coordinated stability factor correction, it reflects the proportion of extreme fluctuations and the level of risk of the fluctuations causing system coordination imbalance.

[0115] The cooperative skewness index and kurtosis cooperative coefficient are used as cooperative modal features.

[0116] Example 2

[0117] Please see Figure 1 As shown, the present invention is a production data acquisition and analysis system based on MES, which also includes the following modules:

[0118] Resonance Analysis Module: Used to construct skewed-kurtosis phase diagrams and corresponding resonance triggering criteria, and determine whether there is systematic wave resonance in the cooperative mode characteristics; if so, it constructs wave event clusters and analyzes the amplification intensity links and corresponding driving modes of the waves; if not, it continuously monitors the cooperative mode characteristics.

[0119] The method for constructing the skewed-kurtosis phase diagram and the corresponding resonance triggering criterion is as follows:

[0120] With the cooperative skewness index as the horizontal axis and the kurtosis cooperative coefficient as the vertical axis, a skewed-kurtosis phase diagram is constructed, and the cooperative mode features are regarded as feature points on the skewed-kurtosis phase diagram;

[0121] During normal production periods, all historical characteristic points of all critical equipment are plotted on a skewed-kurtosis phase diagram.

[0122] Calculate the distribution center and dispersion of historical feature points (such as the covariance matrix) to form a normal cooperative state cloud. Based on this, construct a dynamic confidence ellipse (such as a 95% confidence interval). The inside of the ellipse is the stable region. The outside region of the ellipse is further divided into warning zone and risk zone.

[0123] Based on the historical distribution of the skewed-kurtosis phase diagram, a dynamic confidence ellipse is established according to the preset confidence interval;

[0124] Preferably, the confidence interval is 95%;

[0125] like Figure 2 As shown, a resonance triggering criterion is constructed. If the feature point meets the resonance triggering criterion, it is determined that the cooperative modal feature has a systematic fluctuation resonance that exceeds the historical confidence interval. If it does not meet the criterion, it is determined that it does not meet the systematic fluctuation resonance that exceeds the historical confidence interval.

[0126] Resonance triggering criteria include:

[0127] Resonance triggering criterion one: The current feature point falls outside the confidence ellipse;

[0128] Resonance triggering criterion two: Continuously track feature points in adjacent time windows, calculate the rate of change of the cooperative skewness index and kurtosis cooperative coefficient within the feature points in the Euclidean distance between two adjacent time windows, and use this as the moving speed;

[0129] Obtain N1 sets of movement speeds and calculate the acceleration of the movement speeds (N1 is the number of consecutive sets of movement speeds obtained, with a value of 3-5, preferably N1=3, that is, select 3 consecutive sets of feature point movement speeds in adjacent time windows to calculate the acceleration).

[0130] When the direction of movement points outside the stable region and the acceleration is positive, it is determined that the system is accelerating away from the normal state.

[0131] Resonance Triggering Criterion 3: Based on the cooperative topology baseline, if multiple pairs of strongly dependent devices (dependency edge weight greater than the threshold, dependency edge weight > 0.6) have feature points that move synchronously out of the stable zone or form spatial clusters in the warning zone / risk zone, then cooperative instability is determined to have occurred.

[0132] Among them, the dependency edge weight threshold is obtained by collecting the 70th percentile of all dependency edge weights during normal production. Devices with weights higher than this value are considered to be strongly dependent devices.

[0133] If a feature point satisfies any one of the resonance triggering criteria, it is determined that the cooperative modal features exhibit systematic fluctuation resonance that exceeds the historical confidence interval.

[0134] It is understandable that the purpose of determining the existence of systematic fluctuation resonances in collaborative modal features that exceed historical confidence intervals is as follows:

[0135] Firstly, it enables early warning of production coordination risks, reducing the cascading failures caused by the spread of fluctuations. By determining systemic fluctuation resonance through dynamic confidence ellipses and resonance triggering criteria, early warnings can be issued before fluctuations evolve into substantial production failures, reserving a critical time window for regulation.

[0136] Secondly, it provides a targeted basis for differentiated regulation, reducing the interference of blind regulation on the system. After resonance is determined, high-impact nodes are locked based on the amplification intensity link, and positive diffusion nodes and reversible nodes are distinguished by the node beat instability rate, reducing new collaborative imbalances caused by traditional blind regulation;

[0137] The method for constructing wave event clusters and analyzing the amplification intensity links and corresponding driving modes of waves is as follows:

[0138] Based on feature points that satisfy the resonance triggering criterion, the corresponding key equipment identifiers, fluctuation occurrence time windows, cooperative skewness index, and kurtosis cooperative coefficient are extracted.

[0139] All micro-fluctuation events that trigger resonance within the same time window are clustered using K-Means clustering based on device dependencies in the collaborative topology baseline to form a cluster of fluctuation events.

[0140] Those skilled in the art will understand that, using micro-fluctuation events that trigger resonance as clustering samples, the feature dimensions of each sample are selected as "the cooperative topological position corresponding to the key equipment identifier (based on the core cooperative group number), the fluctuation occurrence time window deviation (the difference between the current window and the resonance start window), the cooperative skewness index, and the kurtosis cooperative coefficient". The features are first standardized by Z-score. The number of clusters K is determined based on the number of core cooperative groups in the cooperative topological baseline (K is equal to the number of core cooperative groups to ensure that the fluctuations within the cluster match the cooperative structure). The elbow rule is used to verify the rationality of the K value. After initializing K cluster centroids, the similarity between each micro-fluctuation event and each centroid is calculated by Euclidean distance. The events are assigned to the clusters with the closest distance, and the cluster centroids are iteratively updated until the change in the centroid position is less than a preset threshold (e.g., 0.01) or the number of iterations reaches the upper limit (e.g., 50 times). Each cluster formed in the end is a fluctuation event cluster, which contains micro-fluctuation events with similar time, close cooperative relationship and similar fluctuation characteristics.

[0141] Based on the network structure of the collaborative topology baseline, wave propagation paths are identified from wave event clusters;

[0142] Starting from the critical equipment node with the largest fluctuation amplitude, i.e. the highest cooperative skewness index, trace the fluctuation source in reverse along the direction of the dependency edge;

[0143] Calculate the amplification intensity between adjacent critical equipment nodes along the wave propagation path:

[0144] It should be noted that the wave propagation path is obtained as follows: the downstream node collaborative skewness index and the upstream node collaborative skewness index of the wave propagation path are collected; the product of the upstream node collaborative skewness index and the dependent edge weight is calculated; the downstream node collaborative skewness index is then compared with the product to obtain the amplification intensity.

[0145] Select link segments along the wave propagation path where the amplification intensity of adjacent critical equipment nodes is higher than 1, and use these as amplification intensity links.

[0146] Based on the spatiotemporal distribution characteristics of wave event clusters and combined with amplification intensity links, the driving mode is identified:

[0147] Preferably, the method for identifying the driving mode is as follows:

[0148] Obtain the number of independent fluctuation sources identified in the fluctuation event cluster;

[0149] It should be noted that the method for determining an independent fluctuation source is as follows: In the network structure of the cooperative topology baseline, if a critical device node triggers resonance due to its own cooperative skewness index, and none of its upstream dependent device nodes trigger resonance, or the fluctuation amplitude of the upstream nodes is much lower than that of the node (for example, lower than 50% of its amplitude), then the node is determined to be an independent fluctuation source.

[0150] The determination of an independent fluctuation source is based on the upstream amplitude being less than 50%, which is derived from historical fault data. When the upstream amplitude is less than 50% of the downstream amplitude, the upstream is not a fluctuation source in more than 90% of cases. Therefore, 50% is taken as the threshold.

[0151] Obtain the wave propagation path structure of the current time window and compare it with the wave propagation paths of the previous M consecutive time windows (preferably M=5) to calculate the path structure similarity.

[0152] It should be noted that the path structure similarity is calculated as follows:

[0153] In identifying the wave propagation path from the wave event cluster, a complete wave propagation path is identified, and the ordered sequence of key equipment nodes traversed by the wave propagation path is stored as a path node set.

[0154] Preferably, the system maintains a historical path database to store all path node sets identified within the most recent M consecutive time windows (M=5); the Jaccard similarity coefficient is used to calculate the ratio of the intersection and union of the current path node set and each historical path node set, and the average value is taken as the path structure similarity index.

[0155] Build driver mode determination rules:

[0156] Judgment Rule 1: If the number of identified independent fluctuation sources is 1, and the path structure similarity is greater than the preset threshold (preferably, the threshold is 0.75), then the current fluctuation mode is determined to be a source-driven amplification path.

[0157] Judgment Rule 2: If the number of identified independent fluctuation sources is greater than or equal to 2, or the path structure similarity is less than the preset judgment threshold (preferably, the threshold is 0.5), then the current fluctuation mode is judged to be a conditionally coupled fluctuation structure.

[0158] It should be noted that the preset judgment threshold is set based on the following: collecting path structure similarity data of a large number of time windows under the historical normal production state of the system, calculating its statistical distribution, setting the upper quartile (approximately 75th quartile) of the similarity distribution as the high similarity threshold (e.g., 0.75) in judgment rule one, and using it to identify stable links; setting the lower quartile (approximately 25th quartile) of the similarity distribution as the low similarity threshold (e.g., 0.5) in judgment rule two, and using it to identify significant drift;

[0159] Based on the determined driving mode, output the corresponding propagation graph:

[0160] If the path is determined to be a source-driven amplification path, the output is a chain amplification spectrum that starts from a single wave source and propagates along the amplification intensity chain; the wave source nodes and the amplification intensity values ​​of each path segment are marked in the spectrum.

[0161] If the structure is determined to be a conditionally coupled wave structure, the output will display a network graph showing multiple wave source nodes, their coupling relationships, and path conditional drift characteristics.

[0162] It should be noted that the path condition drift feature is labeled as follows: different colors or line types are used to mark the nodes or links that differ between the current path and the historical path in the graph; the driving mode determination results, fluctuation source information, amplification intensity links and corresponding propagation graphs are associated, stored and visualized.

[0163] Convergence control module: Based on the cluster of fluctuating events and the amplification intensity link, the convergence target node is extracted, and the convergence target node is subjected to differentiated control of skewness and kurtosis. The collaborative topology baseline is updated to realize the adaptive convergence regression of the analysis system.

[0164] The method for extracting the convergence target node based on the fluctuation event cluster and amplification intensity link is as follows:

[0165] Based on the wave event clusters identified by the resonance analysis module and the corresponding amplification intensity links, path segments with amplification intensity greater than 1 are extracted;

[0166] For each amplification path, obtain the cooperative skewness index and kurtosis cooperative coefficient corresponding to its upstream and downstream nodes;

[0167] If the cooperative skewness index of a downstream node is more than K times higher than that of an upstream node, and its feature point is located outside the confidence ellipse, then the node is marked as the convergence target node, and the dependency edge weight and betweenness centrality value of the node in the cooperative topology baseline are recorded.

[0168] Preferably, K=1.2;

[0169] The method for differentially controlling the skewness and kurtosis of the convergence target node and updating the collaborative topology baseline is as follows:

[0170] For all convergent target nodes, construct the amplification intensity gradient matrix based on the amplification intensity output by the resonance analysis module:

[0171] Understandably, the amplification intensity gradient matrix describes the amplification trend of wave energy along the topology link; a larger value indicates more significant wave diffusion. The rows and columns of the amplification intensity gradient matrix correspond to key device nodes in the cooperative topology baseline, and the elements in the matrix are labeled as follows: , represents the amplification intensity of the wave propagating from node i to node j. If there is no direct dependency edge between the two nodes or the amplification intensity does not exceed the threshold (e.g., 1), the corresponding matrix element is set to 0.

[0172] For each convergence target node, obtain the time series of beat intervals and the beat variation coefficient defined in the feature fusion module;

[0173] Calculate the rate of change of the beat variation coefficient of the convergence target node within M consecutive time windows (preferably M=5) to obtain the beat instability rate;

[0174] It should be noted that the beat instability rate = (current window beat variation coefficient - previous window beat variation coefficient) / previous window beat variation coefficient;

[0175] If the beat instability rate is positive, it indicates an increase in beat instability, and it is marked as a positive diffusion node; otherwise, it is a callback node.

[0176] Based on the amplified intensity gradient matrix, and combined with the node betweenness centrality, a convergence priority function is formed:

[0177] Preferably, by formula: Construct a convergence priority function, where Let i be the amplification intensity gradient at node i. The value of betweenness centrality, For the beat instability rate, , , The coefficients are weighted by experience; nodes with higher priority values ​​are executed parameter adjustments first.

[0178] It should be noted that, , , Value range: (satisfy Preferred =0.5、 =0.4、 =0.1), the value range is determined based on the weight of the parameter's influence on the system's control priority: corresponding to the amplification intensity gradient, it directly determines the risk of fluctuation spread, and has the highest weight. Corresponding betweenness centrality, reflecting the role of collaborative hubs, The corresponding beat instability rate only reflects the local state and has the lowest weight.

[0179] Based on the convergence priority function, skewed and kurtosis convergence control strategies are implemented for different driving modes.

[0180] Preferably, the method for implementing convergence control strategies for different driving modes is as follows:

[0181] S401, Skewed Convergence Control: For forward diffusion nodes, reduce the weight parameters of their upstream dependent edges to suppress the propagation of abnormal offsets; at the same time, adjust the running cycle parameters of the node in the MES system so that the cycle interval approaches the average cycle of the core coordination group.

[0182] For example, the dependency edge weights between the press and the feeding robot are obtained (e.g., weight 0.82), and the adjustment ratio is determined according to the convergence priority function (e.g., reduced by 10%). After updating, the weights are 0.74 to suppress the propagation speed of abnormal offsets. At the same time, the changes in dependency edge weights before and after the callback are recorded for subsequent network stability assessment. The average cycle time of the press in the current window is obtained as 2.35 seconds, while the average cycle time of the core coordination group is 2.10 seconds. The system issues a cycle time adjustment command in the MES, setting the target cycle time of the press to 2.15 seconds, with a cycle time correction range of approximately 8.5%, so that the cycle time approaches the steady-state value. After adjustment, the monitoring cycle time variation coefficient decreases by approximately 12%, and the coordination skewness index decreases by approximately 9%. The maximum value of the cross-correlation function of the press node is recalculated in the next time window.

[0183] S402, Peak convergence control: For nodes with peak coordination coefficients higher than the upper limit of the normal range, the fluctuation amplitude of local accumulation is reduced by increasing the buffer scheduling frequency or adjusting the parallel rhythm ratio.

[0184] Based on feedback from the collaborative stability factor, the beat reference coefficient is dynamically adjusted to smooth out local excessive fluctuations.

[0185] For example, in a detection robot node, the peak-state coordination coefficient is 20% higher than the upper limit of the normal range, while the cycle time stability decreases. The system executes the following peak-state convergence control: the buffer capacity of the detection robot is 12 items, and the average dwell time is 8.4 seconds; the system increases the task rotation frequency in the buffer (from 3 times / minute to 4 times / minute), reducing the local task backlog by about 15% and reducing the peak-state superposition effect; the current parallel operation ratio (number of parallel devices / total number of tasks) is 0.65, and the system adjusts it to 0.55 to balance the parallel rhythm and weaken the superposition of fluctuation peaks; at this time, the peak-state coordination coefficient decreases by about 13%, and the cycle time variation coefficient decreases by about 10%; based on the coordination stability factor (current coordination degree / historical steady-state coordination degree = 0.92), the system calculates the cycle time reference coefficient correction value, that is, the cycle time reference coefficient correction value = cycle time reference coefficient × coordination stability factor:

[0186] After performing parameter tuning, the system recalculates the maximum value of the cross-correlation function of each dependent edge in the cooperative topology baseline and updates the dynamic temporal dependency network.

[0187] If the standard deviation of the updated dependency edge weight fluctuation is less than 10% of the original steady-state variance, and the corresponding cooperative mode feature point re-enters the confidence ellipse interval, then the system is determined to have reached a convergent steady state.

[0188] If the convergent steady state is not reached after 3 adjustments (the standard deviation of the deviation of the dependent edge weight is greater than or equal to 10% of the original steady state variance or the feature point does not return to the confidence ellipse), then a manual intervention warning will be triggered and automatic adjustment will be stopped.

[0189] At this point, the adjusted dependency edge weights and cycle parameters are updated to the historical steady-state database, forming a new cooperative topology baseline, i.e., the real analysis system. The system's adaptive convergent regression.

[0190] Example ES Production According to the data collection and analysis methods, the following steps are included: running the cycle sequence to perform collaborative analysis of the production system and establishing a collaborative topology baseline between key equipment;

[0191] S20. Based on the cooperative topology baseline, the cooperative fluctuations of the equipment are topologically characterized and cooperatively corrected to obtain the cooperative skewness index and kurtosis cooperative coefficient, and the cooperative modal features are obtained by feature fusion.

[0192] S30. Construct a skewed-kurtosis phase diagram and the corresponding resonance triggering criteria to determine whether there is systematic wave resonance in the cooperative mode characteristics; if so, construct a wave event cluster and analyze the amplification intensity link and the corresponding driving mode of the wave; if not, continue to monitor the cooperative mode characteristics.

[0193] S40. Based on the cluster of fluctuating events and the amplification intensity link, extract the convergence target node, perform skewness and kurtosis differential control on the convergence target node, and update the collaborative topology baseline to achieve adaptive convergence regression of the analysis system.

[0194] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A production data acquisition and analysis system based on MES, characterized in that: Includes the following modules: Topology building module: used to obtain the operating cycle time of key equipment, establish the operating cycle time sequence, perform collaborative analysis of the production system based on the operating cycle time sequence, and establish a collaborative topology baseline between key equipment; Feature fusion module: Based on the cooperative topology baseline, the cooperative fluctuations of the equipment are topologically characterized and cooperatively corrected to obtain the cooperative skewness index and kurtosis cooperative coefficient, respectively, and the cooperative modal features are obtained through feature fusion. Resonance Analysis Module: Used to construct skewed-kurtosis phase diagrams and corresponding resonance triggering criteria to determine whether there is systematic fluctuation resonance in the cooperative modal characteristics; If they exist, construct a cluster of wave events and analyze the amplification intensity links and corresponding driving modes of the waves; Convergence control module: Based on the fluctuation event cluster and amplification intensity link, the convergence target node is extracted, and the convergence target node is subjected to differentiated control of skewness and kurtosis, and the collaborative topology baseline is updated.

2. The MES-based production data acquisition and analysis system according to claim 1, characterized in that: The method for establishing the collaborative topology baseline is as follows: A dynamic temporal dependency network is constructed, and a community discovery algorithm is used to identify device collaboration clusters in the network and filter out the core collaboration groups within the clusters. Obtain the node degree neutrality value and betweenness centrality value of the core collaborative group network under normal production conditions as feature parameters; Based on characteristic parameters, a collaborative topology baseline is established to characterize the collaborative state of the production system.

3. The MES-based production data acquisition and analysis system according to claim 2, characterized in that: The dynamic temporal dependency network is constructed as follows: Divide the production time of key equipment into N equal-length time windows; For each time window, extract the operating cycle sequence of each key device within each sliding window, and calculate the cross-correlation function between any two operating cycle sequences of key devices on the production line; The maximum value of the cross-correlation function for each pair of key devices is calculated across all time windows based on the cross-correlation function of all time windows. The T-test was used to determine whether the mean of the maximum values ​​of the cross-correlation functions was significantly positive, and the significance index P value was obtained. If the significance is positive, the mean of the cross-correlation function is multiplied by the significance coefficient determined by the p-value to generate the directed dependency edge weights. Construct an initial temporal dependency graph using key equipment as nodes and directed dependency edges as connections.

4. The MES-based production data acquisition and analysis system according to claim 1, characterized in that: The method for performing topological representation and collaborative correction is as follows: Establish a sequence of coordination deviations of key equipment within the current time window; The skewness coefficient of the cooperative deviation sequence is calculated and combined with the network structure information of the topological baseline to perform kurtosis cooperative correction, thus obtaining the cooperative skewness index; Calculate the excess kurtosis of the coordination deviation sequence and correct it based on production cycle stability to obtain the kurtosis coordination coefficient: The cooperative skewness index and kurtosis cooperative coefficient are used as cooperative modal features.

5. The MES-based production data acquisition and analysis system according to claim 4, characterized in that: The method for performing the aforementioned peak-state collaborative correction is as follows: Obtain the betweenness centrality value of each critical device from the collaborative topology baseline, as well as the number density of dependency edges in the core collaborative group corresponding to the current critical device; The skewness coefficient is multiplied by the betweenness centrality value, and the product is summed with the skewness coefficient to obtain the betweenness skewness coefficient. Multiplying the betweenness skewness coefficient by the number density yields the cooperative skewness index.

6. The MES-based production data acquisition and analysis system according to claim 4, characterized in that: The deviation sequence is established as follows: For each key piece of equipment in the core coordination group, establish a cycle interval sequence; The difference between the beat interval sequence and the coordinated beat reference sequence at each corresponding beat period is calculated to obtain the beat period difference; Obtain the sum of the dependency edge weights of key equipment and other key equipment in the core collaboration group in the topology baseline, normalize the cycle difference, and establish the collaboration deviation sequence of key equipment in the current time window.

7. The MES-based production data acquisition and analysis system according to claim 1, characterized in that: The method for determining whether the cooperative modal features exhibit systematic fluctuation resonance is as follows: Based on the historical distribution of the skewed-kurtosis phase diagram, a dynamic confidence ellipse is established according to the preset confidence interval; Based on the dynamic confidence ellipse, a resonance triggering criterion is constructed. If the feature point meets the resonance triggering criterion, it is determined that the cooperative modal feature has a systematic fluctuation resonance that exceeds the historical confidence interval. If it does not meet the criterion, it is determined that it does not meet the systematic fluctuation resonance that exceeds the historical confidence interval.

8. The MES-based production data acquisition and analysis system according to claim 7, characterized in that: The skewed-kurtosis phase diagram is constructed as follows: A skewed-kurtosis phase diagram is constructed with the cooperative skewness index as the horizontal axis and the kurtosis cooperative coefficient as the vertical axis, and the cooperative mode features are regarded as feature points on the skewed-kurtosis phase diagram.

9. The MES-based production data acquisition and analysis system according to claim 1, characterized in that: The process of analyzing the driving mode is as follows: Obtain the number of independent wave sources identified in the wave event cluster, as well as the wave propagation path structure of the current time window, and compare it with the wave propagation paths of the previous M consecutive time windows to calculate the path structure similarity. The driving mode is determined based on path structure similarity and the number of independent fluctuation sources, and the corresponding driving mode is identified. The driving modes include: source-driven amplification path and condition-coupled wave structure.

10. A production data acquisition and analysis system based on MES according to claim 1, characterized in that: The method for performing the aforementioned skewness and kurtosis differential regulation is as follows: For all convergent target nodes, construct the amplification intensity gradient matrix based on the amplification intensity output by the resonance analysis module: For each convergence target node, obtain the time series of beat intervals and the beat variation coefficient defined in the feature fusion module; Calculate the rate of change of the beat variation coefficient of the convergence target node over M consecutive time windows to obtain the beat instability rate; If the beat instability rate is positive, it is marked as a positive diffusion node; otherwise, it is a callback node. Based on the amplified intensity gradient matrix, and combined with the node betweenness centrality, a convergence priority function is formed: Based on the convergence priority function and the forward diffusion node, skewed and kurtosis convergence control strategies are implemented for different driving modes.

Images