Intelligent management and control system for furniture safety production process based on digital twinning

Abstract

The application discloses a furniture safety production process intelligent management and control system based on digital twinning, relates to the technical field of furniture manufacturing intelligent manufacturing, and comprises the following steps: collecting equipment operation data and process execution data to construct a digital twinning model, extracting a state change moment and a node execution moment to establish a propagation link, verifying a cause-effect relationship between equipment state change and process node execution, and locating an abnormal source state and an influence node. Based on the evolution track of the abnormal source state and the influence node, a state mutation point and a node response delay are identified, a time sequence correlation graph is constructed, and an evolution time difference of the state mutation point and a coupled state mutation point is calculated. A candidate adjustment scheme is generated and verified in the digital twinning model, an optimal scheme is selected as a control instruction for execution, and intelligent management and control of a furniture production process is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent manufacturing of furniture manufacturing, in particular to a furniture safety production process intelligent management and control system based on digital twinning. BACKGROUND

[0002] In recent years, the intelligent transformation and upgrading of the furniture manufacturing industry has put forward higher requirements for the safety management and control of the production process. The traditional furniture production management and control method mainly relies on manual experience to adjust production parameters and monitor the state, which is difficult to meet the real-time monitoring and intelligent management needs of modern production. Although some enterprises have introduced automatic control systems, due to the lack of deep correlation analysis of the state changes of the equipment and the execution of the process in the production process, it is difficult to accurately predict the propagation range of the abnormal state of the equipment, and the positioning of the abnormal source is not accurate enough, which affects the safety of production.

[0003] The abnormality detection method commonly used in the industry at present is mainly based on the threshold judgment of a single device or process node, which cannot effectively identify the correlation anomalies between multiple devices and multiple processes. The existing method lacks the ability to analyze the propagation path of the abnormal state, and it is difficult to discover potential safety hazards in time. The coupling relationship between the state of the equipment and the execution of the process is complex, and it is difficult to accurately capture the timing correlation characteristics between the state mutation points, which affects the accuracy of the abnormality warning.

[0004] Although the digital twinning technology has been applied in industrial production to some extent, the existing technology mainly focuses on the real-time monitoring of the running state of the equipment, and lacks in-depth analysis of the state mutation propagation characteristics. The existing control strategy is often lack of systematic verification of the adjustment scheme, which is easy to cause new abnormal state and affect the stability of production. SUMMARY

[0005] The purpose of the present application is to provide a furniture safety production process intelligent management and control system based on digital twinning, which can accurately identify the state mutation propagation rule of the equipment, accurately locate the abnormal source and automatically generate the optimal control strategy, so as to improve the safety and stability of the furniture production process.

[0006] The furniture safety production process intelligent management and control method based on digital twinning provided in the embodiment of the present application comprises the following steps:

[0007] Collecting the equipment running data and process execution data in the furniture safety production process, and constructing a digital twinning model;

[0008] Extracting the state change time from the equipment running data, extracting the node execution time from the process execution data, establishing a propagation link according to the state change time and the node execution time and marking the propagation direction;

[0009] Verify the causality between the device state change and the process node execution in the propagation direction, and locate the abnormal source state and the impact node when the causality verification fails;

[0010] Extract the evolution trajectory of the abnormal source state and the evolution trajectory of the impact node, identify the state mutation point and the node response delay in the evolution trajectory, and construct a time sequence correlation graph based on the time sequence correspondence of the state mutation point and the node response delay;

[0011] Identify the coupling state mutation point that has a time sequence coupling relationship with the state mutation point in the time sequence correlation graph, and calculate the occurrence time difference between the state mutation point and the coupling state mutation point as the evolution time difference;

[0012] Generate multiple candidate adjustment schemes in the digital twin model, propagate each candidate adjustment scheme in the propagation direction, and select the target adjustment scheme that eliminates the node response delay from the candidate adjustment scheme that does not produce a new state mutation point as the control instruction for execution.

[0013] Further, collecting device operation data and process execution data in the furniture safety production process, constructing a digital twin model includes:

[0014] Collecting device operation data in the furniture safety production process, segmenting the device operation data according to time sequence, extracting frequency domain features from segmented data, and constructing a device operation feature sequence;

[0015] Obtaining device state variables according to the device state change rule based on the device operation feature sequence, constructing a device state transition matrix based on the device state variables and the device operation feature sequence, and generating a device state prediction sequence using the device state transition matrix;

[0016] Collecting process execution data, time sequence matching process execution data and device state prediction sequence, establishing a process time sequence correlation table, and generating a process execution vector according to the process time sequence correlation table;

[0017] Analyzing the process connection time based on the process execution vector, calculating the process response time in combination with the device state prediction sequence, and constructing a process execution sequence through the process connection time and the process response time;

[0018] Bidirectional matching of device state prediction sequence and process execution sequence, calculation of state-process corresponding coefficient, and establishment of state space mapping table according to the state-process corresponding coefficient;

[0019] Constructing a virtual space model using the state space mapping table, establishing a bidirectional data exchange channel between the virtual space model and the actual production system, realizing virtual-real data synchronization, and constructing a digital twin model.

[0020] Further, the state change time is extracted from the equipment operation data, the node execution time is extracted from the process execution data, and the propagation link is established according to the state change time and the node execution time and the propagation direction is marked, which comprises:

[0021] The equipment operation data is wavelet decomposed to obtain decomposition coefficients, a time-frequency feature matrix is constructed according to the decomposition coefficients, an energy aggregation area is identified from the time-frequency feature matrix, a mutation feature of the energy aggregation area is calculated, and the state change time is determined;

[0022] The process execution data is divided into multiple time periods according to the state change time, the data change trend in each time period is calculated, the trend turning point is identified, and the time corresponding to the trend turning point is determined as the node execution time;

[0023] The time interval between adjacent state change times is calculated to obtain a state change sequence, the time interval between adjacent node execution times is calculated to obtain a node execution sequence, and a time sequence combination is constructed according to the corresponding relationship between the state change sequence and the node execution sequence;

[0024] The state change rule and the node execution rule are extracted in the time sequence combination, and the propagation link is determined according to the phase corresponding relationship of the state change rule and the node execution rule;

[0025] The transfer delay of adjacent times in the propagation link is calculated, and the propagation direction of the propagation link is marked according to the order of the transfer delay.

[0026] Further, the causal relationship between the equipment state change and the process node execution is verified along the propagation direction, and when the causal relationship verification fails, the abnormal source state and the impact node are located in reverse including:

[0027] The fluctuation feature of the equipment state change sequence and the response feature of the process node execution sequence are extracted along the propagation direction, the transfer entropy between the fluctuation feature and the response feature is calculated to obtain an information flow amount, the causal relationship strength is judged according to the information flow amount, the causal verification failure position is determined when the causal relationship strength is lower than a verification threshold, and the abnormal transfer area is divided based on the causal verification failure position;

[0028] The phase reconstruction of the equipment state change sequence and the process node execution sequence is performed in the abnormal transfer area, the delay feature after the phase reconstruction is extracted, the transfer abnormal point is identified, the state transition sequence is extracted based on the transfer abnormal point, and the abnormal transfer direction is obtained by analyzing the state transition sequence;

[0029] The conditional entropy of the state transition sequence is calculated in reverse along the abnormal transfer direction to obtain an entropy value distribution, the abnormal propagation path is determined according to the change trend of the entropy value distribution, and the abnormal source state is obtained by tracing back along the abnormal propagation path;

[0030] Extract the distribution characteristics of the abnormal source state, calculate the state probability deviation based on the distribution characteristics, extract the state deviation of each node on the abnormal propagation path, and determine the influencing node based on the degree of matching between the state deviation and the state probability deviation.

[0031] Furthermore, the evolutionary trajectories of the abnormal source state and the influencing nodes are extracted, and state abrupt change points and node response delays in the evolutionary trajectories are identified. Based on the temporal correspondence between state abrupt change points and node response delays, a temporal correlation graph is constructed, including:

[0032] Extract the evolution trajectory of the abnormal source state and calculate the local singular values ​​of the evolution trajectory to obtain dynamic features;

[0033] Extract the changing trends of dynamic features to identify topological mutation locations, determine the topological mutation locations as state mutation points, and obtain a time-series feature sequence based on the distribution of the state mutation points;

[0034] Extract the evolution trajectory of the influencing node, calculate the conditional mutual information entropy between the evolution trajectory of the influencing node and the evolution trajectory of the abnormal source state, obtain the node response delay based on the temporal distribution of the conditional mutual information entropy, and use the node response delay to perform time mapping on the temporal feature sequence to obtain the temporal correspondence.

[0035] Evolutionary features are extracted based on temporal correspondence, evolutionary trajectory links are constructed, evolutionary directions are determined based on the temporal distribution of the evolutionary trajectory links, and the evolutionary directions are used as state propagation paths.

[0036] State evolution features are extracted along the state propagation path. The range of affected nodes is determined based on the state evolution features and the temporal correspondence. The propagation direction and correlation strength between state mutation points within the range of affected nodes are calculated to obtain the correlation matrix. The state mutation points are used as graph nodes, and the propagation direction and correlation strength in the correlation matrix are used as connection attributes to construct a temporal correlation graph.

[0037] Furthermore, in the temporal correlation map, coupled state abrupt change points that have a temporal coupling relationship with the state abrupt change points are identified, and the time difference between the occurrence of the state abrupt change point and the coupled state abrupt change point is calculated as the evolutionary time difference, including:

[0038] The energy distribution characteristics of the neighborhood of the state mutation point in the temporal correlation graph are calculated to obtain the mutation feature vector. The local topology of the state mutation point is analyzed based on the mutation feature vector, and the connection strength and propagation direction between the state mutation points are extracted to construct the local correlation feature matrix.

[0039] The information transmission amount between adjacent state mutation points is calculated based on the local correlation feature matrix to obtain the transmission entropy feature. The information flow trend between state mutation points is analyzed using the transmission entropy feature. The transmission direction and transmission intensity of the information flow trend are extracted. The transmission direction and transmission intensity are combined to construct a coupling evaluation index. Based on the coupling evaluation index, coupled state mutation points that have a temporal coupling relationship with the state mutation points are identified.

[0040] Extract the specific occurrence times of the state mutation point and the coupled state mutation point, calculate the time difference between the state mutation point and the coupled state mutation point, and determine the time difference as the evolution time difference.

[0041] Furthermore, multiple candidate adjustment schemes are generated in the digital twin model, and each candidate adjustment scheme is propagated and pre-performed along the propagation direction. The target adjustment scheme that eliminates node response delay is selected from the candidate adjustment schemes that have never generated a new state mutation point and is issued as a control command for execution, including:

[0042] In a digital twin model, the temporal characteristics of node response delay are analyzed, the changing trend of node response delay is extracted to construct an optimization objective, and adjustment parameters and adjustment timing are designed based on the optimization objective to generate multiple candidate adjustment schemes.

[0043] The current operating state is imported into the digital twin model as the initial condition. Based on the initial condition, the propagation pre-simulation of each candidate adjustment scheme is carried out along the propagation direction. The node state changes are extracted to obtain the evolution sequence. The newly added state mutation points in the evolution sequence are identified, and the node response delay changes are calculated to obtain the pre-simulation results.

[0044] Based on the initial conditions and the pre-simulation results, candidate adjustment schemes are screened, and candidate adjustment schemes that generate new state mutation points are eliminated. In the remaining candidate adjustment schemes, node response delay elimination features are extracted, scheme evaluation indicators are constructed, and the optimal results are obtained by calculating the scheme evaluation indicators.

[0045] From the preferred results, select the candidate adjustment scheme that eliminates node response delay as the target adjustment scheme, convert the target adjustment scheme into a control command, and issue the control command for execution.

[0046] This invention provides an intelligent control system for furniture safety production processes based on digital twins, the system comprising:

[0047] The data acquisition module is used to collect equipment operation data and process execution data during the safe production of furniture, and to build a digital twin model;

[0048] A propagation link construction module is configured to extract state change time from the equipment operation data, extract node execution time from the process execution data, and establish a propagation link and mark a propagation direction according to the state change time and the node execution time;

[0049] An abnormal source positioning module is configured to verify a causal relationship between the equipment state change and the process node execution along the propagation direction, and locate an abnormal source state and an affected node when the causal relationship verification fails;

[0050] A timing correlation graph construction module is configured to extract an evolution track of the abnormal source state and an evolution track of the affected node, identify a state mutation point and a node response delay in the evolution tracks, and construct a timing correlation graph based on a timing correspondence relationship between the state mutation point and the node response delay;

[0051] An evolution time difference calculation module is configured to identify a coupled state mutation point that has a timing coupling relationship with the state mutation point in the timing correlation graph, and calculate a time difference between the occurrence of the state mutation point and the coupled state mutation point as an evolution time difference;

[0052] A control strategy execution module is configured to generate a plurality of candidate adjustment schemes in the digital twin model, perform propagation pre-performance on each candidate adjustment scheme along the propagation direction, and select a target adjustment scheme that eliminates the node response delay from the candidate adjustment schemes that do not produce a new state mutation point as a control instruction to be executed.

[0053] A technical solution provided in an embodiment of the present application is an electronic device, which comprises a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the steps in any of the foregoing methods when executing the computer program.

[0054] A technical solution provided in an embodiment of the present application is a computer readable storage medium, which stores computer program instructions, and the computer program instructions are executed by a processor to implement the steps in any of the foregoing methods.

[0055] This invention utilizes a digital twin model to collect production data in real time, accurately reflecting equipment operating status and process execution, providing a reliable data foundation for state change analysis. Establishing a propagation link based on state change moments and node execution moments effectively tracks the propagation path of abnormal states, improving the accuracy of anomaly source location. By identifying state change points and node response delays to construct a time-series correlation graph, it accurately depicts the coupling relationship between equipment status and process execution. Employing evolutionary time difference analysis, it accurately predicts the propagation trend of state changes, identifying potential safety hazards in advance. Using the digital twin model to perform propagation simulations of candidate adjustment schemes avoids control strategies triggering new abnormal states, ensuring the safe and stable operation of the production process and significantly improving the intelligent management and control level of furniture production. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 A flowchart illustrating the intelligent control method for furniture safety production process based on digital twins provided in this embodiment of the invention;

[0058] Figure 2 This is a flowchart illustrating the abnormal state propagation analysis in an embodiment of the present invention.

[0059] Figure 3 This is a schematic diagram of the structure of a digital twin-based intelligent control system for safe furniture production processes, provided in an embodiment of the present invention. Detailed Implementation

[0060] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. In the following description relating to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements.

[0061] like Figure 1 As shown, Figure 1 A flowchart of an intelligent control method for furniture safety production process based on digital twins provided in this invention is shown. The method includes the following steps:

[0062] Collect equipment operation data and process execution data during the safe production of furniture, and build a digital twin model;

[0063] extracting a state change time from equipment operation data, extracting a node execution time from process execution data, establishing a propagation link according to the state change time and the node execution time and labeling a propagation direction;

[0064] verifying a causal relationship between the equipment state change and the process node execution in the propagation direction, and when the causal relationship verification fails, locating an abnormal source state and an impact node by backtracking;

[0065] extracting an evolution trajectory of the abnormal source state and an evolution trajectory of the impact node, identifying a state mutation point and a node response delay in the evolution trajectory, and constructing a time sequence correlation graph based on a time sequence correspondence relationship between the state mutation point and the node response delay;

[0066] identifying a coupled state mutation point having a time sequence coupling relationship with the state mutation point in the time sequence correlation graph, and calculating a time difference between the state mutation point and the coupled state mutation point as an evolution time difference;

[0067] generating a plurality of candidate adjustment schemes in the digital twin model, performing propagation simulation in the propagation direction for each candidate adjustment scheme, and selecting a target adjustment scheme that eliminates the node response delay from the candidate adjustment schemes that do not produce a new state mutation point as a control instruction for execution.

[0068] collecting equipment operation data and process execution data in a furniture safety production process, and constructing a digital twin model including:

[0069] collecting equipment operation data in a furniture safety production process, segmenting the equipment operation data according to time sequence, extracting frequency domain features from the segmented data, and constructing an equipment operation feature sequence;

[0070] analyzing equipment state change rules based on the equipment operation feature sequence to obtain equipment state variables, constructing an equipment state transition matrix based on the equipment state variables and the equipment operation feature sequence, and generating an equipment state prediction sequence using the equipment state transition matrix;

[0071] collecting process execution data, time sequence matching the process execution data with the equipment state prediction sequence, establishing a process time sequence correlation table, and generating a process execution vector according to the process time sequence correlation table;

[0072] analyzing process connection time based on the process execution vector, calculating process response time in combination with the equipment state prediction sequence, and constructing a process execution sequence through the process connection time and the process response time;

[0073] bidirectional matching the equipment state prediction sequence and the process execution sequence, calculating a state-process correspondence coefficient, and establishing a state space mapping table according to the state-process correspondence coefficient;

[0074] A state space mapping table is used to construct a virtual space model, a two-way data exchange channel between the virtual space model and the actual production system is established, virtual and real data are synchronized, and a digital twin model is constructed.

[0075] First, the sensor network collects equipment operation data, including wood cutting machine vibration frequency, sanding machine motor current, paint spraying equipment pressure, and other parameters. These equipment operation data are divided into multiple time periods according to the production shift, and each time period contains 8 hours of continuous data. The segmented data is subjected to frequency domain feature extraction, and the vibration signal is processed using fast Fourier transform to extract frequency energy distribution characteristic values, while calculating current fluctuation rate, pressure stability index, and other characteristic quantities. These characteristic quantities form the equipment operation feature sequence, reflecting the changes in the operation state of the equipment at different time periods.

[0076] Taking a wood cutting machine as an example, by analyzing its vibration frequency feature sequence, it is found that when the frequency distribution is in the 80-120Hz interval and the amplitude variation rate is less than 5%, the equipment is in a normal state, and when the frequency deviates from the interval or the amplitude variation rate exceeds 10%, it indicates that the equipment may be in an abnormal state. Accordingly, the equipment state variable is determined, and the equipment state is quantified into three categories: normal, slight abnormality, and serious abnormality. Based on the equipment operation feature sequence and the equipment state variable, a device state transition matrix is constructed to record the probability of the device changing from one state to another. Using this matrix and the Markov prediction algorithm, a 24-hour equipment state prediction sequence is generated.

[0077] Process execution data is collected from the production management system, including the execution time and parameter information of the processes such as cutting, notching, assembling, sanding, and painting. The process execution data and the equipment state prediction sequence are matched according to the time stamp to establish a process timing association table, which records the start time, completion time, and corresponding equipment state information of each process. According to the process timing association table, a process execution vector is generated, and the vector elements include process type code, execution time, resource occupancy rate, and other information.

[0078] By analyzing the process execution vector, the transition time between adjacent processes is calculated. For example, the time interval between the completion of board cutting and the start of notching is the transition time, which is normally 15 minutes. Combined with the equipment state prediction sequence, the response time of each process is calculated, which is the time required from the issuance of the process instruction to the actual execution. Through the process transition time and the process response time, a complete process execution sequence is constructed, forming a time axis model of the entire furniture production process.

[0079] The device state prediction sequence is bidirectionally matched with the process execution sequence, and a state-process correspondence coefficient is calculated. The correspondence coefficient represents the degree of influence of device state change on process execution or the degree of influence of process execution on device state. For example, when the polishing device changes from a normal state to a slightly abnormal state, the corresponding paint process quality influence coefficient is 0.85, indicating that the device state change will cause the paint quality to decrease by 15%. According to the state-process correspondence coefficient, a state space mapping table is established to comprehensively record the correspondence between device state and process execution.

[0080] A virtual space model is constructed using the state space mapping table, which includes a virtual mirror of the furniture production line and can simulate the device running state and process execution in the real production environment. A bidirectional data exchange channel is established between the virtual space model and the actual production system through the OPCUA communication protocol to realize virtual-real data synchronization. When the device state changes in the actual production system, the data updates the virtual model through the exchange channel; when the production parameters are optimized or the process is adjusted in the virtual model, the optimization results can be fed back to the actual production system. Through this bidirectional data exchange mechanism, a complete digital twin model of furniture safety production is constructed.

[0081] For example, by collecting vibration data of a wood cutting device through a sensor, it is found that the frequency characteristics show abnormal fluctuations during the afternoon production shift, and the state transition matrix predicts that the device may appear in a serious abnormal state the next morning. Matching this information with the process execution sequence, it is found that the two key processes of material feeding and mortising may cause production delay. The digital twin model automatically adjusts the production plan, allocates part of the chair production tasks to other cutting devices, and arranges for device maintenance in advance.

[0082] The present application can accurately predict the trend of device state change, discover potential safety hazards in advance, realize the correlation analysis of process and device state, ensure the stability of product quality, support the dynamic optimization and adjustment of production plan, reduce the impact of device failure on production, establish a virtual-real integrated management and control system, and improve the scientificity and timeliness of production decision-making. Through the comprehensive application of these technical means, the safety accident risk in the furniture production process is effectively reduced, the production efficiency and product quality are improved, and the digital and intelligent transformation of the furniture manufacturing process is realized, providing technical support for the safety production of the furniture industry.

[0083] The state change time is extracted from the device running data, the node execution time is extracted from the process execution data, and a propagation link is established according to the state change time and the node execution time and the propagation direction is labeled, including:

[0084] Wavelet decomposition is performed on the equipment operation data to obtain decomposition coefficients, a time-frequency feature matrix is constructed according to the decomposition coefficients, an energy aggregation area is identified from the time-frequency feature matrix, a mutation feature of the energy aggregation area is calculated, and a state change time is determined;

[0085] According to the state change time, the process execution data is divided into multiple time periods, the data change trend in each time period is calculated, the trend turning point is identified, and the time corresponding to the trend turning point is determined as the node execution time;

[0086] The time interval between adjacent state change times is calculated to obtain a state change sequence, the time interval between adjacent node execution times is calculated to obtain a node execution sequence, and a time sequence combination is constructed according to the corresponding relationship between the state change sequence and the node execution sequence;

[0087] The state change rule and the node execution rule are extracted in the time sequence combination, and the propagation link is determined according to the phase corresponding relationship of the state change rule and the node execution rule;

[0088] The transmission delay of adjacent times in the propagation link is calculated, and the propagation direction of the propagation link is marked according to the order of the transmission delay.

[0089] In the process of furniture safety production, sensors are installed on key equipment such as woodworking cutting machines, sanding machines, and paint spraying equipment to collect operation data, including vibration, temperature, current, sound, and other signals. The collected equipment operation data is processed by wavelet decomposition, using db4 wavelet basis function, the original signal is decomposed for five layers, and a series of decomposition coefficients are obtained. The decomposition coefficients include five levels of detail coefficients and one approximation coefficient. After wavelet decomposition of the vibration signal, decomposition coefficients of different frequency bands are obtained, the low-frequency coefficients reflect the overall operation state of the equipment, and the high-frequency coefficients reflect the local abnormal characteristics. According to these decomposition coefficients, a time-frequency feature matrix is constructed, the rows of the matrix represent time sequences, the columns represent frequency components, and the matrix element values represent the energy size of the corresponding time-frequency points.

[0090] An energy aggregation area is identified from the time-frequency feature matrix, and an adaptive threshold method is used to determine that the area with energy density exceeding twice the average value as the energy aggregation area. The mutation features of the energy aggregation area are calculated, including energy change rate, frequency offset degree, and duration. When the energy change rate exceeds 30%, the frequency offset degree is greater than the preset threshold 10Hz, and the duration exceeds 5s, it is determined that the equipment state has changed significantly, and the corresponding time is recorded as the state change time. When the nozzle is blocked, the high-frequency energy area will have a significant mutation, and the time when the mutation occurs is recorded as the state change time.

[0091] Process execution data is collected from the production execution system, including process start time, completion time, operating parameters, and other information. The process execution data is divided into multiple time periods according to the previously determined state change time. For each time period, the process data, such as paint thickness, coverage uniformity, and other parameters, is calculated to determine the trend. Using a sliding window method, the window size is 30 data points, and the step size is 5 data points. The linear regression slope of the data in the window is calculated to represent the trend of the data. When the slope of two adjacent windows changes in sign and the change amplitude is greater than the preset threshold 0.05, a trend turning point is identified. The time corresponding to the trend turning point is determined as the node execution time.

[0092] The time interval between adjacent state change times is calculated to form a state change sequence. Taking a paint spraying device as an example, the time intervals of the device from normal operation to nozzle blockage, from nozzle blockage to cleaning recovery, and other state changes are recorded to form a device state change sequence. The time interval between adjacent node execution times is calculated to form a node execution sequence. For example, the time intervals of the paint spraying process from start to parameter adjustment, from parameter adjustment to completion, and other nodes are recorded to form a node execution sequence. According to the time correspondence between the state change sequence and the node execution sequence, a time sequence combination is constructed. The time sequence combination is represented by an adjacency matrix, and the matrix element value represents the time correlation strength of the state change and the node execution.

[0093] The state change rule and the node execution rule are extracted in the time sequence combination. The state change rule analysis uses a time series pattern mining method to identify the periodic change characteristics of the device state. The node execution rule analysis uses a process flow correlation analysis method to identify the sequence dependency of process execution. The propagation link is determined according to the phase correspondence of the state change rule and the node execution rule. The phase correspondence is obtained by calculating the cross-correlation function of the time series of the two rules, and the time delay corresponding to the correlation peak represents the phase difference of the two rules. When the correlation coefficient is greater than 0.7 and the phase difference is within a reasonable range, it is confirmed that there is a propagation relationship between the two time points, and these associated points are connected to form a propagation link.

[0094] The transfer delay between adjacent times in the propagation link is calculated, i.e., the time difference between two adjacent time points. The propagation direction of the propagation link is labeled according to the order of the transfer delay. The propagation direction points from the earlier occurring time point to the later occurring time point. For example, when the nozzle of the paint spraying device is blocked, it is found through propagation link analysis that the impact is first propagated to the paint thickness parameter, then to the surface quality detection link, and finally may lead to product rework. According to the propagation direction, it is determined that the problem root is at the device level rather than the process parameter setting.

[0095] The device operation data is decomposed by wavelet, and an energy aggregation area is identified in the high frequency part. The state change time is determined as 10:30 on the production day by calculating the mutation feature. The process execution data is divided according to the time, and the trend change is calculated to find that the paint thickness parameter turns at 10:33, which is determined as the node execution time. The state change sequence and the node execution sequence are calculated, the timing combination is constructed, and it is identified that there is a high correlation between the device vibration anomaly and the paint quality decline. The propagation link is determined through phase analysis, the transmission delay is calculated as 3 minutes, and the propagation direction is marked as from the device state to the process quality.

[0096] The application accurately identifies the device state change feature through wavelet decomposition technology, establishes the timing correlation between the device state and the process execution, constructs the propagation link and determines the propagation direction, realizes the early discovery and propagation prediction of the abnormal state in the furniture production process. Through the construction of the digital twin model, the accurate mapping of the virtual space and the physical space is realized, which greatly improves the safety and stability of furniture production, reduces the quality defect rate and the risk of production accidents, improves the production efficiency and resource utilization, and provides an effective way for the digital transformation of furniture manufacturing enterprises.

[0097] The causal relationship between the device state change and the process node execution is verified along the propagation direction. When the causal relationship verification fails, the abnormal source state and the impact node are located in reverse:

[0098] The fluctuation feature of the device state change sequence and the response feature of the process node execution sequence are extracted along the propagation direction, the transmission entropy between the fluctuation feature and the response feature is calculated to obtain the information flow amount, the causal relationship strength is judged according to the information flow amount, and the causal verification failure position is determined when the causal relationship strength is lower than the verification threshold, and the abnormal transmission area is divided based on the causal verification failure position;

[0099] The phase reconstruction of the device state change sequence and the process node execution sequence is carried out in the abnormal transmission area, the delay feature after phase reconstruction is extracted, the transmission abnormal point is identified, the state transition sequence is extracted based on the transmission abnormal point, and the abnormal transmission direction is analyzed according to the state transition sequence;

[0100] The conditional entropy of the state transition sequence is calculated in reverse along the abnormal transmission direction to obtain the entropy value distribution, the abnormal propagation path is determined according to the change trend of the entropy value distribution, and the abnormal source state is obtained by tracing back along the abnormal propagation path;

[0101] The distribution feature of the abnormal source state is extracted, the state probability deviation is calculated according to the distribution feature, the state deviation of each node on the abnormal propagation path is extracted, and the impact node is determined according to the matching degree of the state deviation and the state probability deviation.

[0102] In the furniture production process, the fluctuation features of the equipment state change sequence and the response features of the process node execution sequence are extracted along the determined propagation direction. The fluctuation features of the equipment state change sequence include amplitude, frequency and waveform characteristics, which are extracted by sliding window analysis. Taking a wood carving machine as an example, the motor vibration amplitude change rate, temperature fluctuation period and current pulsation characteristics are extracted as fluctuation features. The response features of the process node execution sequence are obtained by analyzing the process parameter change trend, including parameter change rate, response delay and stability index. For the furniture surface treatment process, the paint film thickness change rate, drying time deviation and surface smoothness fluctuation are extracted as response features.

[0103] The transfer entropy between the fluctuation features and the response features is calculated to obtain the information flow amount, and the transfer entropy is calculated by the conditional probability under the historical state condition. The equipment state change sequence and the process node execution sequence are divided into data windows with a length of 20, and the sliding step is 5. The influence degree of the fluctuation features on the response features in each window is calculated, which is quantified as the information flow amount. When the information flow amount is greater than 0.6, it indicates that there is a strong causal relationship between the two sequences; when the information flow amount is between 0.3 and 0.6, it indicates that there is a moderate causal relationship; when the information flow amount is less than 0.3, it indicates that the causal relationship is weak. According to the calculation result, the strength of the causal relationship is judged, and the verification threshold is set to 0.3. When the strength of the causal relationship is lower than the verification threshold, it is determined that the position is a causal verification failure position. Based on the causal verification failure position, the abnormal transfer area is divided, and the continuous occurrence of the causal verification failure position and the previous and subsequent 5 time points are classified into the same abnormal transfer area.

[0104] In the abnormal transfer area, the equipment state change sequence and the process node execution sequence are phase reconstructed. The delay coordinate embedding method is used to map the time sequence to a high-dimensional phase space, the delay time is selected as 5 sampling points, the embedding dimension is 4, and the phase space trajectory is constructed. By analyzing the geometric characteristics of the phase space trajectory, the delay features after phase reconstruction are extracted, including phase difference, phase synchronization degree and conversion delay. The singular points in the delay features are identified as transfer abnormal points, which are positions with sudden phase difference or significantly reduced synchronization degree. Based on the transfer abnormal points, the state transition sequence is extracted, and the changes of the equipment state and the process state before and after the abnormal points are recorded to form a state transition matrix. The dominant direction of state transition is determined as the abnormal transfer direction by analyzing the transition probability distribution in the state transition sequence.

[0105] The conditional entropy of the state transition sequence is calculated in the reverse direction of the abnormality transmission direction to obtain an entropy value distribution. The conditional entropy is calculated using a sliding window method with a history window length of 10 to calculate the uncertainty of the current state under the condition of a given history state. For example, for a furniture production line, the entropy value distribution is obtained by reverse calculation along the abnormality transmission direction from the edge sealing process to the panel saw, and it is found that the entropy value gradually increases near the panel saw position and reaches a peak value of 1.8 at the motor control unit. According to the trend of the entropy value distribution, the abnormality propagation path is determined, and the region with a high entropy value corresponds to the main path of abnormality propagation, and the position of the entropy value peak corresponds to the abnormality source. The abnormality source state is obtained by tracing back along the abnormality propagation path, which is manifested as the parameter abnormality of the motor control unit of the panel saw.

[0106] The distribution characteristics of the abnormality source state are extracted, including the mean, variance, skewness and kurtosis of the state value. By comparing the distribution characteristics of the abnormality source state and the historical normal state, the state probability deviation is calculated. The specific method is to construct a probability distribution model of the normal state and calculate the degree of deviation of the abnormality source state from the normal distribution. For the motor control unit of the panel saw, the current parameter of the abnormal state deviates from the normal distribution by 2.5 standard deviations, and the state probability deviation reaches 0.85. The state deviation of each node on the abnormality propagation path is extracted, and the difference between the state value and the historical normal value is calculated for each node. According to the matching degree of the state deviation and the state probability deviation, the influence nodes are determined, and the matching degree is calculated by the Pearson correlation coefficient, and the nodes with a correlation coefficient greater than 0.7 are determined as the key influence nodes.

[0107] After the abnormality source state is determined, the parameter adjustment is performed on the motor control unit of the panel saw of the furniture production line, and the current limit parameter is adjusted from 8.5A to 7.2A, and the vibration frequency control threshold is adjusted from 120Hz to 105Hz. At the same time, the key influence nodes are maintained, and the worn transmission bearing is replaced, and the installation angle of the cutting tool is adjusted.

[0108] The present application realizes accurate tracing and intervention of furniture production abnormalities by in-depth analysis of the information transmission mechanism between equipment state changes and process execution. The method can accurately identify the position of failed cause and effect verification and divide the abnormality transmission area; the phase reconstruction technology is used to extract deep delay features to identify the transmission abnormality point; the conditional entropy analysis is used to locate the abnormality source state, and the key influence nodes are determined by matching the state deviation. This multi-dimensional abnormality propagation analysis method greatly improves the detection accuracy and tracing efficiency of abnormal states in the furniture production process, reduces the production downtime and quality defect rate.

[0109] As shown in Figure 2 The evolution trajectory of the abnormality source state and the evolution trajectory of the influence nodes are extracted, the state mutation points and node response delays in the evolution trajectory are identified, and a time sequence correlation map is constructed based on the time sequence correspondence relationship between the state mutation points and the node response delays, including:

[0110] extracting an evolution trajectory of the abnormal source state, calculating a local singular value of the evolution trajectory to obtain a dynamic feature;

[0111] extracting a change trend of the dynamic feature to identify a topological mutation position, determining the topological mutation position as a state mutation point, and obtaining a time sequence feature sequence according to a distribution of the state mutation point;

[0112] extracting an evolution trajectory of the influence node, calculating a conditional mutual information entropy of the evolution trajectory of the influence node and the evolution trajectory of the abnormal source state, obtaining a node response delay based on a time sequence distribution of the conditional mutual information entropy, and performing time mapping on the time sequence feature sequence by using the node response delay to obtain a time sequence correspondence;

[0113] extracting an evolution feature according to the time sequence correspondence, constructing an evolution trajectory link, determining an evolution direction based on a time sequence distribution of the evolution trajectory link, and taking the evolution direction as a state propagation path;

[0114] extracting a state evolution feature along the state propagation path, determining an influence node range according to the state evolution feature and the time sequence correspondence, calculating a propagation direction and a correlation strength between state mutation points in the influence node range to obtain a correlation matrix, taking the state mutation points as graph nodes, and taking the propagation direction and the correlation strength in the correlation matrix as connection attributes to construct a time sequence correlation graph.

[0115] The evolution trajectory of the abnormal source state is extracted by phase space reconstruction on time sequence data of the abnormal source device. Multidimensional parameters such as motor speed, vibration intensity, and temperature change are selected to construct a state space, and a time delay embedding method is used for reconstruction. The embedding dimension is 5 and the time delay is 3 sampling intervals. After phase space reconstruction, the local singular value of the evolution trajectory is calculated to obtain a dynamic feature. The local singular value is calculated by using a sliding window method, the window size is 50 sampling points, and the step size is 10 sampling points. A covariance matrix is constructed for the data in the window, singular value decomposition is performed, and the ratio of the maximum singular value to the second largest singular value is extracted as the dynamic feature. The dynamic feature reflects the local instability and complexity of the system.

[0116] The change trend of the dynamic feature is extracted to identify the topological mutation position. The differential method is used to calculate the change rate of the dynamic feature. When the change rate exceeds a preset threshold, it is determined that the topological structure has mutated. For the mill motor, the change rate threshold is set to 0.35, and when the change rates of three consecutive windows all exceed the threshold, it is confirmed that the position is a topological mutation position. The topological mutation position is determined as a state mutation point, and a time sequence feature sequence is formed according to the time distribution of the state mutation point.

[0117] The evolution trajectory of the influence node is extracted, and the influence node includes process parameters and equipment components directly related to the abnormal source. For example, the influence node of the sander includes the sand belt tension, the grinding wheel speed, the workpiece conveying belt speed, etc. The evolution trajectory of each influence node is constructed in the same way as the abnormal source. The conditional mutual information entropy of the evolution trajectory of the influence node and the evolution trajectory of the abnormal source state is calculated, which represents the degree of mutual dependence between two variables given a third variable. The kernel density estimation method is used to approximate the probability distribution, and the Gaussian kernel function is selected with a bandwidth of 0.2. The node response delay is obtained based on the time series distribution of the conditional mutual information entropy, which is determined by finding the time difference corresponding to the peak value of the mutual information entropy. For the sand belt tension node, the response delay is 5 minutes; for the workpiece conveying belt speed node, the response delay is 8 minutes. The time mapping of the time series feature sequence is performed using the node response delay to obtain the time series correspondence, that is, the mutation time of the abnormal source state and the response time of each influence node are corresponded.

[0118] According to the time series correspondence, the evolution features are extracted, including the state transition rate, the stability index and the complexity quantization value. The state transition rate is calculated by dividing the Euclidean distance between adjacent state points by the time interval; the stability index is represented by the local Lyapunov exponent approximation value; and the complexity quantization value is calculated by using permutation entropy. The evolution trajectory link is constructed, and the evolution trajectories of each influence node are connected in series to form a link structure according to the time series correspondence. The evolution direction is determined based on the time series distribution of the evolution trajectory link, and the state transition sequence from early to late constitutes the evolution direction. The evolution direction is taken as the state propagation path to describe how the abnormal state propagates from the source through each influence node. The state propagation path is: motor abnormality→sand belt tension abnormality→grinding wheel speed fluctuation→workpiece conveying belt speed change→machining surface quality decline.

[0119] Along the state propagation path, the state evolution features are extracted, and the local dynamic features of each node on the path are calculated, including the curvature, divergence and convergence index of the state space trajectory. According to the state evolution features and the time series correspondence, the influence node range is determined, and the nodes with significant time series correspondence with the abnormal source state are selected into the range. The propagation direction and the correlation strength between the state mutation points in the influence node range are calculated to obtain the correlation matrix. The propagation direction is determined by the time sequence, and the correlation strength is quantified by the normalized mutual information value. For the state mutation points in the influence node range of the sander, a 5x5 correlation matrix is calculated, and the matrix element value ranges from 0 to 1, representing the correlation degree between nodes. The state mutation points are taken as the graph nodes, and the propagation direction and the correlation strength in the correlation matrix are taken as the connection attributes to construct the time series correlation graph. The node size in the graph represents the influence range, the connection line thickness represents the correlation strength, and the arrow direction represents the propagation direction.

[0120] The application accurately captures the state mutation point and its propagation law through topological mutation analysis and time sequence feature mapping; the propagation mechanism of the abnormal state between different nodes is revealed through conditional mutual information entropy calculation and time sequence corresponding relationship construction; and finally, the visualization expression and quantitative analysis of the abnormal state propagation network are realized through the construction of the time sequence correlation graph. The method significantly improves the abnormal detection capability and traceability accuracy of the furniture production process, reduces the false positive rate and false negative rate, shortens the time from abnormal occurrence to positioning the root cause, and improves the stability of the production line and the consistency of product quality.

[0121] In the time sequence correlation graph, identify the coupling state mutation point that has a time sequence coupling relationship with the state mutation point, and calculate the time difference between the occurrence of the state mutation point and the coupling state mutation point as the evolution time difference, including:

[0122] Calculate the energy distribution characteristics of the state mutation point neighborhood in the time sequence correlation graph to obtain the mutation feature vector, analyze the local topological structure of the state mutation point according to the mutation feature vector, extract the connection strength and propagation direction between the state mutation points, and construct a local correlation feature matrix;

[0123] According to the local correlation feature matrix, calculate the information transfer amount between adjacent state mutation points to obtain the transfer entropy feature, analyze the information flow trend between the state mutation points using the transfer entropy feature, extract the transfer direction and transfer strength of the information flow trend, combine the transfer direction and transfer strength to construct a coupling evaluation index, and identify the coupling state mutation point that has a time sequence coupling relationship with the state mutation point based on the coupling evaluation index;

[0124] Extract the specific occurrence time of the state mutation point and the coupling state mutation point, calculate the time difference between the state mutation point and the coupling state mutation point, and determine the time difference as the evolution time difference.

[0125] On the basis of the constructed time sequence correlation graph, the energy distribution characteristics of the state mutation point neighborhood in the time sequence correlation graph are calculated to obtain the mutation feature vector. The state mutation point neighborhood refers to a subgraph structure with the mutation point as the center and a radius of topological distance 2. The energy distribution characteristics are calculated by weighted measurement. For each connection in the neighborhood, different weights are assigned according to its correlation strength, and the higher the correlation strength, the higher the weight. The weighted sum of all connections in the neighborhood is calculated to form the total energy value; then the distribution proportion of energy in different directions is calculated to form the energy distribution characteristics. The energy distribution of the state mutation point neighborhood of the wood cutting machine presents uneven characteristics, mainly concentrated in the cutting accuracy and feed speed correlation direction, accounting for 62% of the total energy. After normalizing the energy distribution characteristics by direction, the mutation feature vector is formed, and the vector dimension is the same as the number of neighborhood nodes. Each component represents the distribution proportion of energy in that direction.

[0126] The local topology structure of the state mutation point is analyzed according to the mutation feature vector, including centrality, connectivity and clustering coefficient. The centrality represents the importance of the node in the network, which is determined by calculating the number of edges directly connected to it; the connectivity represents the proportion of other nodes reachable from the node; and the clustering coefficient represents the connection tightness between the neighbors of the node. For the state mutation point of the cutting machine, the centrality is 4, the connectivity is 0.7, and the clustering coefficient is 0.5, indicating that the point is in a key connection position in the network, but the connection between the neighbor nodes is not tight enough. The local correlation feature matrix is constructed by extracting the connection strength and propagation direction between state mutation points. The element value of the matrix represents the connection strength between two nodes, and the element symbol represents the propagation direction, with positive value representing from the row node to the column node and negative value representing from the column node to the row node. The dimension of the matrix is equal to the number of nodes in the neighborhood, and the diagonal elements are zero.

[0127] The information transmission amount between adjacent state mutation points is calculated according to the local correlation feature matrix to obtain the transmission entropy feature. The transmission entropy calculation uses the sliding window method, with window length of 30 time points and step length of 5 time points. For each pair of adjacent state mutation points, the prediction contribution of the source point state to the future state of the target point is calculated, and the known historical state of the target point is subtracted to obtain the pure information transmission amount. The larger the transmission entropy value, the more significant the information transmission.

[0128] The information flow trend between state mutation points is analyzed using the transmission entropy feature, and the information flow diagram is drawn, with the arrow direction representing the dominant information flow direction and the line thickness representing the flow strength. The transmission direction and transmission strength of the information flow trend are extracted, with the transmission direction determined by the positive and negative values of the transmission entropy and the transmission strength determined by the absolute value of the transmission entropy. The coupling evaluation index is constructed by combining the transmission direction and transmission strength, which is the weighted sum of the transmission strength multiplied by the direction factor (1 for forward and -1 for reverse). The coupling evaluation threshold is set to 0.3, and when the absolute value of the coupling evaluation index between two state mutation points exceeds the threshold and there is a temporal relationship, it is determined that there is a time series coupling relationship between the two points. The coupling state mutation points with time series coupling relationship with the state mutation points are identified based on the coupling evaluation index.

[0129] The specific occurrence time of the state mutation point and the coupling state mutation point is extracted, which is accurately obtained through the timestamp record of the digital twin platform. The state mutation point time is determined by detecting the dynamic feature mutation, and the first time point with feature change rate exceeding the threshold is used as the reference. The time difference between the state mutation point and the coupling state mutation point is calculated, and the time difference is determined as the evolution time difference. The evolution time difference represents the time required for the abnormal state to propagate from one node to another node, reflecting the system response speed and propagation delay characteristics.

[0130] For example, this paper analyzes the quality fluctuation problem that occurs during the processing of panel furniture. A digital twin model is constructed to collect real-time status data of each piece of equipment on the production line. The sander is identified as a key state mutation point in the time-series correlation graph. Its neighborhood energy distribution characteristics are calculated, yielding a mutation feature vector [0.62, 0.15, 0.18, 0.05], indicating that energy is mainly concentrated in the sanding pressure direction. The connection strength and propagation direction between the sander and adjacent nodes are extracted, constructing a 4×4 local correlation feature matrix. The transfer entropy feature is calculated, revealing a significant information flow between the sander and the surface treatment equipment, with a transfer entropy value of 0.39. Based on the coupling evaluation index of 0.41, the surface treatment equipment is identified as the coupling state mutation point of the sander. The sander's state mutation point occurs at 108 minutes after production begins, and the surface treatment equipment's state mutation point occurs at 115 minutes, resulting in an evolution time difference of 7 minutes. By analyzing the evolution time difference and transmission characteristics, it was determined that abnormal belt tension in the sander was the root cause of the surface treatment effect fluctuation. Timely adjustment of the belt tension parameters effectively solved the quality fluctuation problem.

[0131] This invention achieves accurate identification of key anomalies in furniture production by calculating the energy distribution characteristics and local topology of state abrupt change points in a time-series correlation graph. Through information transmission entropy analysis and the construction of coupling evaluation indicators, it reveals the propagation mechanism and coupling relationship of abnormal states among different equipment nodes. Furthermore, by calculating evolutionary time difference, it quantifies the propagation speed and delay characteristics of abnormal states. This method overcomes the limitation of traditional anomaly detection focusing only on single-point anomalies, establishing a global anomaly propagation network model, thus improving the accuracy and predictive ability of anomaly detection.

[0132] Multiple candidate adjustment schemes are generated in the digital twin model. Each candidate adjustment scheme is propagated and pre-performed along the propagation direction. From the candidate adjustment schemes that have not generated any new state mutation points, the target adjustment scheme that eliminates the node response delay is selected as the control command for execution, including:

[0133] In a digital twin model, the temporal characteristics of node response delay are analyzed, the changing trend of node response delay is extracted to construct an optimization objective, and adjustment parameters and adjustment timing are designed based on the optimization objective to generate multiple candidate adjustment schemes.

[0134] The current operating state is imported into the digital twin model as the initial condition. Based on the initial condition, the propagation pre-simulation of each candidate adjustment scheme is carried out along the propagation direction. The node state changes are extracted to obtain the evolution sequence. The newly added state mutation points in the evolution sequence are identified, and the node response delay changes are calculated to obtain the pre-simulation results.

[0135] The candidate adjustment scheme is screened based on the initial condition and the pre-play result, the candidate adjustment scheme generating a new state mutation point is eliminated, the node response delay elimination feature is extracted from the remaining candidate adjustment scheme, the scheme evaluation index is constructed, and the preferred result is obtained by calculating the scheme evaluation index;

[0136] The candidate adjustment scheme eliminating the node response delay is selected from the preferred result as the target adjustment scheme, the target adjustment scheme is converted into a control instruction, and the control instruction is issued for execution.

[0137] The timing characteristics of the node response delay are analyzed in the digital twin model, which is a virtual mapping constructed by collecting sensing data of each device on the furniture production line. The node response delay refers to the time required from the source node state change to the target node response. The timing characteristics are extracted by time series analysis method, including trend component, periodic component and random component. In the furniture production line, the node response delay timing data between the wood cutting equipment and the polishing equipment shows a fluctuating upward trend, and the average delay gradually increases from 5s to 12s. The growth rate of the delay obtained by time series decomposition is 0.7s / h. The change trend of the node response delay is extracted to construct the optimization target, and the minimization of the delay time and the maximization of the delay stability are taken as the optimization target. According to the optimization target, adjustment parameters and adjustment timing are designed. The adjustment parameters include device operation parameters, process parameters and control strategy parameters, and the adjustment timing refers to the order and time interval of parameter adjustment. For the wood cutting and polishing process, the designed adjustment parameters include cutting speed, cutting depth, polishing pressure and polishing speed, and the adjustment timing is sorted according to the dependency relationship, that is, the cutting speed is adjusted first, the cutting depth is adjusted after the cutting is stable, the polishing pressure is adjusted, and finally the polishing speed is adjusted. Through parameter space grid search and timing arrangement combination, a plurality of candidate adjustment schemes are generated. The candidate schemes generated for the wood processing production line include high-speed light pressure scheme, medium-speed balanced scheme and low-speed heavy pressure scheme, each scheme contains different parameter combinations and adjustment timing, and a total of 15 specific candidate adjustment schemes are generated.

[0138] The current running state is imported into the digital twin model as the initial condition, including real-time state parameters and environmental parameters of each device. For the wood processing line, the initial conditions include cutting speed 25 m / min, cutting depth 8 mm, polishing pressure 0.8 MPa, and polishing speed 1200 rpm, and a response delay of 12 s between cutting and polishing. The propagation pre-play of each candidate adjustment scheme is performed based on the initial condition along the propagation direction, which is determined according to the process flow, simulating the state propagation step by step from the upstream to the downstream node. The propagation pre-play uses the discrete event simulation method, with a time step of 0.1 s and a simulation time of 3 times the working cycle period. For each candidate scheme, parameter adjustment is performed in the digital twin environment, and response changes are recorded. The evolution sequence is obtained by extracting the node state change, with time as the horizontal axis and node state as the vertical axis, forming a trajectory diagram of state change over time. The state change is calculated by the relative change rate of key parameters, and when the change rate exceeds the preset threshold, it is considered as a state change point. The newly added state mutation point in the evolution sequence is identified, which refers to the mutation point that does not exist in the original production process but appears in the pre-play process. The pre-play result is calculated by the node response delay change, the difference between the response delay before and after adjustment is compared, and the adjustment effect is evaluated.

[0139] Based on the initial condition and the pre-play result, the candidate adjustment scheme is screened, and a multi-level screening strategy is adopted. The candidate adjustment scheme that produces a new state mutation point is eliminated, and the scheme that may introduce new risks is directly excluded. In the remaining candidate adjustment schemes, the node response delay elimination features are extracted, including delay reduction, delay stability and delay consistency. The delay reduction is calculated by the absolute difference of the delay time before and after adjustment; the delay stability is calculated by the variance of the delay time sequence; and the delay consistency is calculated by the coefficient of variation of the delay time under different working conditions. The scheme evaluation index is constructed, and the features of the three dimensions are combined by weighting. The preferred result is obtained by calculating the scheme evaluation index, and the higher the evaluation index, the better the scheme.

[0140] The candidate adjustment scheme that eliminates node response delay is selected as the target adjustment scheme from the optimized results. Selection criteria include the scheme with the highest evaluation index and the lowest implementation difficulty. For the wood processing line case, scheme 6, with the highest evaluation index, is selected as the target adjustment scheme. The specific parameters of this scheme are: cutting speed adjusted to 22 m / min, cutting depth adjusted to 7 mm, grinding pressure adjusted to 0.75 MPa, and grinding speed adjusted to 1300 rpm. The adjustment sequence is as follows: first, reduce the cutting speed by 3 m / min; after waiting 2 minutes, reduce the cutting depth by 1 mm; after waiting 1 minute, reduce the grinding pressure by 0.05 MPa; and finally, increase the grinding speed by 100 rpm. The target adjustment scheme is converted into control commands, which include parameter setpoints, execution time points, and execution order. The control commands are encoded into a data format recognizable by the equipment and distributed to the corresponding equipment controllers via the industrial control network. For the wood processing line, the control commands are converted into executable program code, and the parameters are adjusted sequentially according to the set adjustment sequence. The system issues control commands for execution and monitors the execution effect in real time through feedback. When deviations from expectations are detected, alternative solutions can be triggered or the original configuration can be restored.

[0141] This invention achieves risk prediction and effect evaluation for production process adjustments through a multi-scheme generation and propagation pre-simulation mechanism; it ensures the scientific validity and feasibility of adjustment schemes through multi-dimensional evaluation indicators and intelligent screening algorithms; and it achieves seamless integration from virtual optimization to physical adjustment through an automated control command conversion and execution mechanism. This method overcomes the limitations of experience-based decision-making in traditional production management, establishing a data-driven intelligent decision-making system that significantly improves the stability and controllability of furniture production processes. This method can proactively identify and eliminate abnormal response delays in the production process, reduce equipment failures and quality defects, and improve product consistency and production efficiency.

[0142] like Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a digital twin-based intelligent control system for furniture safety production processes provided in an embodiment of the present invention. The system includes:

[0143] The data acquisition module is used to collect equipment operation data and process execution data during the safe production of furniture, and to build a digital twin model;

[0144] The propagation link construction module is used to extract the state change time from the equipment operation data, extract the node execution time from the process execution data, establish a propagation link based on the state change time and the node execution time, and mark the propagation direction.

[0145] The anomaly source localization module is used to verify the causal relationship between equipment state changes and process node execution along the propagation direction. When the causal relationship verification fails, it traces back to locate the anomaly source state and the affected node.

[0146] The time sequence correlation graph construction module is configured to extract the evolution track of the abnormal source state and the evolution track of the influence node, identify the state mutation point and the node response delay in the evolution track, and construct a time sequence correlation graph based on the time sequence corresponding relationship of the state mutation point and the node response delay.

[0147] The evolution time difference calculation module is configured to identify a coupled state mutation point that has a time sequence coupling relationship with the state mutation point in the time sequence correlation graph, and calculate the time difference between the occurrence time of the state mutation point and the coupled state mutation point as the evolution time difference.

[0148] The control strategy execution module is configured to generate a plurality of candidate adjustment schemes in the digital twin model, perform propagation pre-rehearsal on each candidate adjustment scheme along the propagation direction, and select a target adjustment scheme that eliminates the node response delay from the candidate adjustment schemes that do not produce new state mutation points as a control instruction for execution.

[0149] A technical solution provided in an embodiment of the present application is an electronic device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the steps in any of the foregoing methods when executing the computer program.

[0150] A technical solution provided in an embodiment of the present application is a computer readable storage medium, which stores a computer program, and the processor implements the steps in any of the foregoing methods when executing the computer program.

[0151] The specific embodiments described above are preferred embodiments of the present application, and do not limit the specific implementation range of the present application, and the scope of the present application includes but is not limited to the specific embodiments. Any equivalent changes made according to the shape and structure of the present application are within the protection scope of the present application.

Claims

1. A digital-twin-based intelligent management and control method for a furniture safety production process, characterized in that, The method comprises the following steps: Collecting equipment operation data and process execution data in the production process of furniture safety, and constructing a digital twin model; Extracting state change moments from the equipment operation data and node execution moments from the process execution data, establishing a propagation link according to the state change moments and the node execution moments, and marking the propagation direction; Verifying the causal relationship between the equipment state change and the process node execution along the propagation direction, and locating the abnormal source state and the impact node when the causal relationship verification fails; Extracting the evolution track of the abnormal source state and the evolution track of the impact node, identifying the state mutation point and the node response delay in the evolution track, and constructing a time sequence correlation graph based on the time sequence correspondence relationship of the state mutation point and the node response delay; Identifying coupled state mutation points that have a time sequence coupling relationship with the state mutation points in the time sequence correlation graph, calculating the occurrence time difference between the state mutation points and the coupled state mutation points as the evolution time difference, and quantifying the propagation speed of the abnormal state and the node delay characteristics through the evolution time difference; Generating multiple candidate adjustment schemes in the digital twin model, performing propagation pre-performance on each candidate adjustment scheme along the propagation direction, and selecting a target adjustment scheme that eliminates the node response delay from the candidate adjustment schemes that do not produce new state mutation points as a control instruction for execution; Extracting the evolution track of the abnormal source state and the evolution track of the impact node, identifying the state mutation point and the node response delay in the evolution track, and constructing a time sequence correlation graph based on the time sequence correspondence relationship of the state mutation point and the node response delay comprises: Extracting the evolution track of the abnormal source state, calculating the local singular value of the evolution track to obtain a dynamic feature; Extracting the change trend of the dynamic feature to identify a topological mutation position, determining the topological mutation position as a state mutation point, and obtaining a time sequence feature sequence according to the distribution of the state mutation point; Extracting the evolution track of the impact node, calculating the conditional mutual information entropy of the evolution track of the impact node and the evolution track of the abnormal source state, obtaining the node response delay based on the time sequence distribution of the conditional mutual information entropy, and time mapping the time sequence feature sequence using the node response delay to obtain a time sequence correspondence relationship; Extracting an evolution feature according to the time sequence correspondence relationship, constructing an evolution track link, determining an evolution direction based on the time sequence distribution of the evolution track link, and taking the evolution direction as a state propagation path; Extracting a state evolution feature along the state propagation path, determining an impact node range according to the state evolution feature and the time sequence correspondence relationship, calculating the propagation direction and the correlation strength between the state mutation points in the impact node range to obtain a correlation matrix, taking the state mutation points as graph nodes, and taking the propagation direction and the correlation strength in the correlation matrix as connection attributes to construct a time sequence correlation graph; Identifying coupled state mutation points that have a time sequence coupling relationship with the state mutation points in the time sequence correlation graph, and calculating the occurrence time difference between the state mutation points and the coupled state mutation points as the evolution time difference comprises: The energy distribution characteristics of the neighborhood of the state mutation point in the temporal correlation graph are calculated to obtain the mutation feature vector. The local topology of the state mutation point is analyzed based on the mutation feature vector, and the connection strength and propagation direction between the state mutation points are extracted to construct the local correlation feature matrix. The information transmission amount between adjacent state mutation points is calculated based on the local correlation feature matrix to obtain the transmission entropy feature. The information flow trend between state mutation points is analyzed using the transmission entropy feature. The transmission direction and transmission intensity of the information flow trend are extracted. The transmission direction and transmission intensity are combined to construct a coupling evaluation index. Based on the coupling evaluation index, coupled state mutation points that have a temporal coupling relationship with the state mutation points are identified. Extract the specific occurrence times of the state mutation point and the coupled state mutation point, calculate the time difference between the state mutation point and the coupled state mutation point, and determine the time difference as the evolution time difference; Multiple candidate adjustment schemes are generated in the digital twin model. Each candidate adjustment scheme is propagated and pre-performed along the propagation direction. From the candidate adjustment schemes that have not generated any new state mutation points, the target adjustment scheme that eliminates the node response delay is selected as the control command for execution, including: In a digital twin model, the temporal characteristics of node response delay are analyzed, the changing trend of node response delay is extracted to construct an optimization objective, and adjustment parameters and adjustment timing are designed based on the optimization objective to generate multiple candidate adjustment schemes. The current operating state is imported into the digital twin model as the initial condition. Based on the initial condition, the propagation pre-simulation of each candidate adjustment scheme is carried out along the propagation direction. The node state changes are extracted to obtain the evolution sequence. The newly added state mutation points in the evolution sequence are identified, and the node response delay changes are calculated to obtain the pre-simulation results. Based on the initial conditions and the pre-simulation results, candidate adjustment schemes are screened, and candidate adjustment schemes that generate new state mutation points are eliminated. In the remaining candidate adjustment schemes, node response delay elimination features are extracted, scheme evaluation indicators are constructed, and the scheme evaluation indicators are calculated to obtain the results. From the results, select the candidate adjustment scheme that eliminates node response delay as the target adjustment scheme, convert the target adjustment scheme into a control command, and issue the control command for execution.

2. The digital-twin-based intelligent management and control method for furniture safety production processes according to claim 1, characterized in that, Collecting equipment operation data and process execution data during the safe production of furniture, and constructing a digital twin model includes: Collect equipment operation data during the safe production process of furniture, segment the equipment operation data according to the time series, extract frequency domain features from the segmented data, and construct equipment operation feature sequence; Based on the analysis of the equipment operation characteristic sequence, the equipment state change pattern is obtained to obtain the equipment state variables. Based on the equipment state variables and the equipment operation characteristic sequence, an equipment state transition matrix is ​​constructed, and the equipment state transition matrix is ​​used to generate an equipment state prediction sequence. Collect process execution data, perform time-series matching between process execution data and equipment status prediction sequence, establish process time sequence association table, and generate process execution vector based on the process time sequence association table; Based on the process execution vector analysis of process connection time, the process response time is calculated by combining the equipment status prediction sequence, and the process execution sequence is constructed by the process connection time and the process response time. The device state prediction sequence is bidirectionally matched with the process execution sequence, a state-process correspondence coefficient is calculated, and a state space mapping table is established according to the state-process correspondence coefficient; A virtual space model is constructed by using the state space mapping table, a bidirectional data exchange channel between the virtual space model and the actual production system is established, virtual and actual data are synchronized, and a digital twin model is constructed.

3. The digital-twin-based intelligent management and control method for furniture safety production processes according to claim 1, characterized in that, The state change time is extracted from the device operation data, the node execution time is extracted from the process execution data, a propagation link is established according to the state change time and the node execution time, and the propagation direction is labeled, including: The device operation data is wavelet-decomposed to obtain decomposition coefficients, a time-frequency feature matrix is constructed according to the decomposition coefficients, an energy aggregation area is identified from the time-frequency feature matrix, a mutation feature of the energy aggregation area is calculated, and the state change time is determined; The process execution data is divided into multiple time periods according to the state change time, the data change trend in each time period is calculated, the trend turning point is identified, and the time corresponding to the trend turning point is determined as the node execution time; The time interval between adjacent state change times is calculated to obtain a state change sequence, the time interval between adjacent node execution times is calculated to obtain a node execution sequence, and a time sequence combination is constructed according to the correspondence relationship between the state change sequence and the node execution sequence; The state change rule and the node execution rule are extracted in the time sequence combination, and the propagation link is determined according to the phase correspondence relationship between the state change rule and the node execution rule; The transmission delay of adjacent times in the propagation link is calculated, and the propagation direction of the propagation link is labeled according to the order of the transmission delay.

4. The digital-twin-based intelligent management and control method for furniture safety production processes according to claim 1, characterized in that, The causal relationship between the device state change and the process node execution is verified along the propagation direction, and when the causal relationship verification fails, the abnormal source state and the impact node are located in reverse, including: The fluctuation feature of the device state change sequence and the response feature of the process node execution sequence are extracted along the propagation direction, the transmission entropy between the fluctuation feature and the response feature is calculated to obtain an information flow amount, the causal relationship strength is judged according to the information flow amount, the causal verification failure position is determined when the causal relationship strength is lower than a verification threshold, and the abnormal transmission area is divided based on the causal verification failure position; The phase reconstruction of the device state change sequence and the process node execution sequence is performed in the abnormal transmission area, the delay feature after the phase reconstruction is extracted, the transmission abnormal point is identified, the state transition sequence is extracted based on the transmission abnormal point, and the abnormal transmission direction is analyzed according to the state transition sequence; The conditional entropy of the state transition sequence is calculated in reverse along the abnormal transmission direction to obtain an entropy value distribution, the abnormal propagation path is determined according to the change trend of the entropy value distribution, and the abnormal source state is obtained by tracing back along the abnormal propagation path; The distribution feature of the abnormal source state is extracted, the state probability deviation is calculated according to the distribution feature, the state deviation of each node on the abnormal propagation path is extracted, and the impact node is determined according to the matching degree of the state deviation and the state probability deviation.

5. The intelligent management and control system for furniture safety production process based on digital twinning, for realizing the method of any one of claims 1-4, characterized in that, The system comprises: A data acquisition module is configured to acquire device operation data and process execution data in a furniture safety production process, and construct a digital twin model. The propagation link construction module is configured to extract a state change time from the equipment operation data, extract a node execution time from the process execution data, establish a propagation link according to the state change time and the node execution time, and mark a propagation direction; The anomaly source positioning module is configured to verify a causal relationship between the equipment state change and the process node execution along the propagation direction, and when the causal relationship verification fails, backtrack to locate an abnormal source state and an impact node; The timing correlation graph construction module is configured to extract an evolution track of the abnormal source state and an evolution track of the impact node, identify a state mutation point in the evolution track and a node response delay, and construct a timing correlation graph based on a timing corresponding relationship of the state mutation point and the node response delay; The evolution time difference calculation module is configured to identify a coupled state mutation point that has a timing coupling relationship with the state mutation point in the timing correlation graph, and calculate a time difference between the occurrence of the state mutation point and the coupled state mutation point as an evolution time difference; The control strategy execution module is configured to generate a plurality of candidate adjustment schemes in the digital twin model, perform a propagation preview on each candidate adjustment scheme along the propagation direction, and select a target adjustment scheme that eliminates the node response delay from the candidate adjustment schemes that do not produce a new state mutation point as a control instruction to be executed.

6. An electronic device, comprising: The computer readable storage medium stores computer program instructions, and the computer program instructions are executed by the processor to implement the steps in the method of any one of claims 1 to 4. The computer readable storage medium stores computer program instructions, and the computer program instructions are executed by the processor to implement the steps in the method of any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, ​

Images