Production collaborative management method and system for digital factory

Abstract

The present invention provides a production collaboration management method and system for a digital factory. The method comprises: calculating an affinity matrix between devices based on their physical location, communication latency, device type, and historical collaboration relationships, and determining the collaboration weights of heterogeneous devices to obtain a device collaboration networking solution; then calculating a time series score based on each device group's task completion quality, energy consumption efficiency, device stability, and collaborative adaptability, and dynamically adjusting the weights in combination with environmental monitoring data to obtain a device credit score; then performing initial task allocation based on the device credit score, performing continuous bias analysis and resource balancing by analyzing data during task execution, and obtaining a dynamic task allocation strategy; finally, evaluating the collaborative effects of the device groups based on the task allocation strategy, and dynamically optimizing the device groups to obtain a collaborative evolution solution. Through dynamic evaluation and balancing mechanisms, the present invention effectively solves the problem of irrational device resource allocation.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to a production collaborative management method and system for a digital factory. Background Art

[0002] In digital factories, collaborative production management is crucial for ensuring production efficiency. Its core lies in the rational allocation and scheduling of equipment resources. Existing collaborative production management methods for digital factories primarily utilize task allocation strategies based on historical data. Specifically, the system prioritizes tasks for equipment with good historical performance based on their historical performance. This approach has demonstrated certain advantages in the early stages of a factory's digital transformation, ensuring basic production efficiency requirements in the short term.

[0003] However, as the digital factory operated over time, this historical data-based task allocation approach gradually revealed serious problems. High-performing equipment, as it continuously received more tasks, accumulated and optimized its performance data. However, equipment with average performance, with fewer tasks assigned, lacked opportunities for practice and improvement. This situation led to a typical "Matthew effect" in equipment resource allocation: the strong became stronger, while the weak became weaker. More seriously, this imbalance in resource allocation worsened over time, ultimately leaving some equipment in "optimization blind spots," unable to participate in efficient production collaboration and impacting the overall efficiency of the digital factory. This problem was particularly pronounced in complex digital factory environments with multiple areas and types of equipment. Summary of the Invention

[0004] The main purpose of the present invention is to solve the technical problem that in the existing digital factory production collaborative management method, task allocation based on historical performance data easily leads to unreasonable resource allocation;

[0005] A first aspect of the present invention provides a production collaborative management method for a digital factory, the production collaborative management method for a digital factory comprising:

[0006] Calculate the affinity matrix between devices based on the physical location, communication delay, device type, and historical collaboration relationship of devices in each area of ​​the digital factory. Then, determine the collaboration weights of heterogeneous devices based on the affinity matrix to obtain a device collaborative networking solution.

[0007] Calculating the device's time series score based on the task completion quality, energy consumption efficiency, device stability, and collaborative adaptability of each device group in the device collaborative networking solution, and dynamically weighting the time series score based on environmental monitoring data to obtain a device credit score;

[0008] Initial task allocation is performed on the device based on the device credit score, and continuous bias analysis is performed on the device based on the task allocation data collected during task execution. Device resources are balanced based on the analysis results to obtain a dynamic task allocation strategy;

[0009] According to the task allocation strategy, the production efficiency, energy utilization rate and task completion quality of the equipment group are evaluated for synergy effect, and based on the evaluation results, the member composition and capability complementarity of the equipment group are dynamically optimized to obtain a collaborative evolution plan for the equipment group.

[0010] Optionally, in a first implementation of the first aspect of the present invention, the device collaborative networking solution is obtained by calculating an affinity matrix between devices based on the physical location, communication delay, device type, and historical collaboration relationship of devices in each area of ​​the digital factory, and determining the collaboration weights of heterogeneous devices based on the affinity matrix. The method includes:

[0011] Perform Euclidean distance calculation on the physical location data of the device to obtain a location distance matrix, and perform sampling calculation on the round-trip delay of communication data packets between devices to obtain a communication delay matrix;

[0012] The compatibility of communication protocols of different types of devices is calculated to obtain the device type affinity matrix, and the historical collaboration data of devices is processed with a time decay function to obtain the collaboration intensity matrix.

[0013] Performing a multi-dimensional weighted fusion calculation based on the location distance matrix, the communication delay matrix, the device type affinity matrix, and the collaboration strength matrix to obtain an affinity matrix between devices;

[0014] Evaluating the communication efficiency between the heterogeneous device pairs in the affinity matrix to obtain an initial collaboration weight, and dynamically adjusting the initial collaboration weight according to the collaboration success rate of the heterogeneous device pairs under different working conditions to obtain a collaboration weight between the heterogeneous devices;

[0015] The devices are grouped and divided based on the collaboration weights to obtain a device collaborative networking solution.

[0016] Optionally, in a second implementation of the first aspect of the present invention, the device credit score is obtained by calculating the device time series score based on the task completion quality, energy consumption efficiency, device stability, and collaborative adaptability of each device group in the device collaborative networking solution, and dynamically weighting the time series score based on environmental monitoring data. The following comprises:

[0017] Performing least squares fitting calculation on the task position accuracy deviation, task completion time deviation, and task execution parameter deviation of each device group in the device collaborative networking solution to obtain a quality assessment matrix, and performing piecewise linear regression calculation on the quality assessment matrix to obtain task quality time series data;

[0018] Perform wavelet decomposition calculation on the device current fluctuation data, voltage fluctuation data, and power factor data of each device group in the device collaborative networking solution to obtain an energy consumption characteristic sequence, and perform cross-correlation calculation on the energy consumption characteristic sequence and the standard energy consumption model to obtain energy efficiency time series data;

[0019] Perform empirical mode decomposition on the equipment vibration spectrum data, temperature change data, and motor speed fluctuation data of each equipment group in the equipment collaborative networking solution to obtain state eigenvectors. Perform principal component analysis on the state eigenvectors to obtain stability time series data.

[0020] Performing Markov chain transition probability calculation on the task quality time series data, energy efficiency time series data, and stability time series data to obtain a time series score;

[0021] Principal component analysis is performed on the regional temperature field gradient data, humidity distribution data, and electromagnetic interference intensity distribution data in the environmental monitoring data to obtain an environmental feature vector. Neural network weight calculation is performed on the environmental feature vector and the time series score to obtain the equipment credit score.

[0022] Optionally, in a third implementation of the first aspect of the present invention, performing Markov chain transition probability calculation on the task quality time series data, energy efficiency time series data, and stability time series data to obtain a time series score includes:

[0023] Perform time series segmentation calculation on the task quality time series data to obtain a quality state sequence, perform frequency statistics on adjacent states in the quality state sequence to obtain a first-order transfer frequency matrix;

[0024] Normalizing the energy efficiency time series data and the stability time series data to obtain a standardized data sequence, and performing state quantization calculation on the standardized data sequence to obtain an energy efficiency state sequence and a stability state sequence;

[0025] Performing transition probability calculation on the quality state sequence, energy efficiency state sequence, and stability state sequence to obtain a Markov transition probability matrix, and performing stationary distribution calculation on the Markov transition probability matrix to obtain a state probability vector;

[0026] n-step transition probability calculation is performed according to the state probability vector and the Markov transition probability matrix to obtain a time series score.

[0027] Optionally, in a fourth implementation of the first aspect of the present invention, performing initial task allocation on devices based on the device credit score, performing continuous bias analysis on the devices based on task allocation data collected during task execution, and balancing device resources based on the analysis results to obtain a dynamic task allocation strategy includes:

[0028] Calculating the sample standard deviation of the device credit score to obtain score dispersion data, calculating the probability density distribution of the score dispersion data to obtain an initial task allocation probability matrix, and performing initial task allocation based on the initial task allocation probability matrix;

[0029] Performing time series acquisition and sliding time window calculation on the task allocation data after the initial task allocation to obtain a task allocation time series matrix;

[0030] Performing K-means clustering calculation on the task allocation time series matrix to obtain a device task density distribution map, calculating a skewness coefficient on the device task density distribution map, grouping and marking devices whose skewness coefficient exceeds a threshold, and obtaining an optimized blind spot device group;

[0031] Performing principal component analysis on the task execution data of the blind spot optimization equipment group to obtain a task feature matrix, and performing hierarchical clustering on the task feature matrix to obtain task difficulty level data;

[0032] A Bayesian update calculation is performed on the initial task allocation probability matrix according to the task difficulty level data to obtain a dynamic task allocation strategy.

[0033] Optionally, in a fifth implementation of the first aspect of the present invention, performing K-means clustering calculation on the task allocation timing matrix to obtain a device task density distribution map, performing skewness coefficient calculation on the device task density distribution map, grouping and marking devices whose skewness coefficient exceeds a threshold, and obtaining an optimized blind spot device group includes:

[0034] Performing Gaussian kernel density estimation calculation on each sampling point of the task allocation timing matrix to obtain a device task distribution probability map, and performing gradient descent clustering calculation on the device task distribution probability map to obtain a task density cluster center;

[0035] Performing Euclidean distance calculation on the task density cluster centers to obtain a cluster distance matrix, and performing K-means iterative calculation based on the cluster distance matrix to obtain a device task density distribution map;

[0036] Performing probability moment estimation calculation on the equipment task density distribution map to obtain a distribution matrix, and performing third-order moment calculation on the distribution matrix to obtain a skewness coefficient;

[0037] A dynamic threshold value is calculated for the devices according to the skewness coefficient to obtain a grouping threshold value, the grouping threshold value is compared with the skewness coefficient, and the devices whose skewness coefficient exceeds the grouping threshold value are grouped and marked to obtain an optimized blind spot device group.

[0038] Optionally, in a sixth implementation of the first aspect of the present invention, performing a synergistic effect evaluation on the production efficiency, energy utilization, and task completion quality of the equipment group according to the task allocation strategy, and dynamically optimizing the member composition and capability complementarity of the equipment group based on the evaluation results to obtain a collaborative evolution plan for the equipment group includes:

[0039] Time series data of the equipment group task throughput, the equipment group task response time and the equipment group task accuracy are collected during the execution of the task allocation strategy to obtain equipment group production efficiency evaluation data;

[0040] Performing weighted calculation on the equipment group energy consumption data, equipment group standby time data, and equipment group peak power data during the execution of the task allocation strategy to obtain equipment group energy utilization evaluation data;

[0041] Performing a hierarchical analysis calculation on the equipment group based on the production efficiency evaluation data and the energy utilization rate evaluation data to obtain an equipment group performance evaluation matrix, and performing an eigenvalue decomposition calculation on the equipment group performance evaluation matrix to obtain an equipment group capability eigenvector;

[0042] An optimal matching calculation is performed on the device group members according to the device group capability feature vector to obtain a device group member reorganization plan, and a device group collaborative efficiency in the device group member reorganization plan is predicted and calculated to obtain a collaborative evolution plan for the device group.

[0043] A second aspect of the present invention provides a production collaborative management system for a digital factory, the production collaborative management system for a digital factory comprising:

[0044] The networking module is used to calculate the affinity matrix between devices based on the physical location, communication delay, device type, and historical collaboration relationship of devices in each area of ​​the digital factory, and determine the collaboration weights of heterogeneous devices based on the affinity matrix to obtain a device collaboration networking solution;

[0045] a scoring module for calculating a time series score of a device based on the task completion quality, energy consumption efficiency, device stability, and collaborative adaptability of each device group in the device collaborative networking solution, and dynamically weighting the time series score based on environmental monitoring data to obtain a device credit score;

[0046] A scheduling module is used to perform initial task allocation to devices based on the device credit score, and to continuously analyze the bias of devices based on the task allocation data collected during task execution. Based on the analysis results, device resources are balanced to obtain a dynamic task allocation strategy.

[0047] The optimization module is used to evaluate the synergistic effect of the production efficiency, energy utilization and task completion quality of the equipment group according to the task allocation strategy, and dynamically optimize the member composition and capability complementarity of the equipment group based on the evaluation results to obtain a collaborative evolution plan for the equipment group.

[0048] The above-mentioned digital factory production collaborative management method and system calculates the affinity matrix between devices based on the physical location, communication delay, device type and historical collaborative relationship of the devices, and determines the collaborative weight of heterogeneous devices to obtain a device collaborative networking solution; then calculates the timing score based on the task completion quality, energy consumption efficiency, device stability and collaborative adaptability of each device group, and dynamically adjusts the weight in combination with environmental monitoring data to obtain the device credit score; then performs initial task allocation based on the device credit score, and continuously analyzes the bias and balances the data during the task execution process to obtain a dynamic task allocation strategy; finally, evaluates the collaborative effect of the device group based on the task allocation strategy, and dynamically optimizes the device group to obtain a collaborative evolution solution. Through dynamic evaluation and balancing mechanisms, the present invention effectively solves the problem of irrational device resource allocation.

[0049] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the present invention. The purposes and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram of a first embodiment of a production collaborative management method for a digital factory according to an embodiment of the present invention;

[0052] Figure 2 Schematic diagram of a production collaborative management system for a digital factory in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0054] The terms "including," "having," and any variations thereof, as used in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to the process, method, product, or device.

[0055] To facilitate understanding of this embodiment, a production collaborative management method for a digital factory disclosed in an embodiment of the present invention is first described in detail. Figure 1 As shown, this method includes the following steps:

[0056] 101. Calculate the affinity matrix between devices based on the physical location, communication delay, device type, and historical collaboration relationship of devices in each area of ​​the digital factory. Determine the collaboration weights of heterogeneous devices based on the affinity matrix to obtain a device collaborative networking solution.

[0057] In one embodiment of the present invention, the affinity matrix between devices is calculated based on the physical location, communication delay, device type and historical collaboration relationship of devices in each area of ​​the digital factory, and the collaboration weights of heterogeneous devices are determined based on the affinity matrix to obtain a device collaborative networking solution, including: performing Euclidean distance calculation on the physical location data of the devices to obtain a location distance matrix, and sampling and calculating the round-trip delay of communication data packets between devices to obtain a communication delay matrix; performing matching calculation on the communication protocol compatibility of different types of devices to obtain a device type affinity matrix, and performing time decay function processing on the historical collaboration data of the devices to obtain a collaboration intensity matrix; performing multi-dimensional weighted fusion calculation based on the location distance matrix, communication delay matrix, device type affinity matrix and collaboration intensity matrix to obtain an affinity matrix between devices; evaluating the communication efficiency between heterogeneous device pairs in the affinity matrix to obtain an initial collaboration weight, and dynamically adjusting the initial collaboration weight based on the collaboration success rate of heterogeneous device pairs under different working conditions to obtain a collaboration weight between heterogeneous devices; and grouping and dividing devices based on the collaboration weight to obtain a device collaborative networking solution.

[0058] Specifically, in the actual operational environment of a digital factory, the system first needs to obtain the physical coordinate data of each device from the device management system. This data, typically in the form of X, Y, and Z coordinates, accurately reflects the device's specific location in the factory's three-dimensional space. Using this location data, the system uses a spatial rectangular coordinate system to calculate the actual distance between each pair of devices, thereby constructing a position distance matrix that reflects the spatial relationship between the devices. Simultaneously, the system also needs to evaluate the communication performance between the devices. This process is accomplished by cyclically sending probe packets between the devices. Specifically, the system collects the round-trip transmission time of data packets during different time periods, such as peak production periods and stable production periods. Each time period is sampled at least 1000 times, with a sampling interval of 100 milliseconds. The system removes outliers from the collected latency data and calculates the average value, ultimately generating a communication latency matrix that represents the communication performance between the devices. For example, between two brands of AGVs deployed in a digital factory, the system records the latency of transmitting different types of data packets, such as task instructions, status information, and sensor data. This data fully captures the communication performance characteristics between the devices. By establishing the location distance matrix and communication delay matrix, the system obtains the basic data of physical distribution and communication performance between devices.

[0059] Specifically, in a digital factory environment, different types of devices often use different communication protocols, such as OPC UA, Modbus, and Profinet. The system requires in-depth analysis of these protocols, including aspects such as their data encapsulation format, transmission mechanisms, security policies, and error handling mechanisms. By establishing protocol feature vectors, the system can calculate the degree of compatibility between different protocols. For example, when two devices both support OPC UA, their protocol compatibility is assigned a higher value. However, when one device uses OPC UA and the other uses Modbus, the system assigns a lower value based on the complexity of the protocol conversion. These compatibility values ​​ultimately form a device type affinity matrix. Simultaneously, the system extracts device collaboration records from the production history database. These records contain information such as collaboration time, collaboration type, and collaboration results. By designing a time decay function, the system processes this historical collaboration data, assigning higher weights to newer collaboration records and lower weights to older collaboration records. This processing reflects the timeliness of device collaboration relationships, and the resulting collaboration strength matrix accurately represents the actual collaboration capabilities between devices. In practical applications, the time decay function adopts an exponential decay form to ensure that the impact of historical data gradually decreases over time.

[0060] Specifically, after obtaining four basic matrices reflecting the different characteristics of the devices, the system performs a multi-dimensional fusion of these matrices to produce a comprehensive affinity matrix. This fusion process first requires normalizing each matrix to ensure comparability across different dimensions. The normalized matrices are then multiplied by their corresponding weight coefficients, which reflect the importance of each dimension in device collaboration. For example, in some scenarios, communication performance may be more important than physical distance, so the weight coefficient for the communication latency matrix is ​​adjusted accordingly. The weighted results of the multiple matrices are linearly combined to produce the final affinity matrix, where each element quantitatively represents the overall affinity between the corresponding device pairs. To ensure the reliability of the calculation results, the system performs outlier detection and data cleansing on the raw data for each dimension, and considers data confidence during the fusion process. Furthermore, to accommodate dynamic changes in the factory environment, these weight coefficients are not fixed but are dynamically adjusted based on actual production conditions, ensuring that the affinity matrix accurately reflects the current state of device collaboration. This process utilizes a multi-dimensional weighted fusion algorithm that comprehensively considers the influence of multiple factors, including location distance, communication latency, protocol compatibility, and historical collaboration.

[0061] Specifically, in digital factories, heterogeneous devices primarily refer to AGVs of different brands and models. While these devices all possess the basic functions of automated navigation and task execution, they differ significantly in control systems, communication protocols, and performance characteristics due to differences in manufacturer. A specialized evaluation of the communication efficiency of these heterogeneous devices is required because data exchange between devices of different brands often requires protocol or data format conversion, a process that incurs additional communication overhead and potential compatibility issues. The system collects metrics such as data transmission rate, packet integrity, and protocol conversion time between heterogeneous devices to construct a quantitative communication efficiency evaluation system, thereby calculating the initial collaboration weight. In actual production, heterogeneous devices face a variety of operating conditions, such as routine task delivery under normal operating conditions, intensive scheduling under high-load conditions, and emergency avoidance under interference conditions. The system continuously collects and statistically analyzes the collaboration of heterogeneous devices under these different operating conditions, including data on task completion rate, collaboration response time, and error handling capabilities. This statistical data is used to calculate the collaboration success rate of heterogeneous devices under various operating conditions and to adjust the initial collaboration weight accordingly. Through this dynamic adjustment mechanism, the final collaborative weight can accurately reflect the collaborative capabilities of heterogeneous devices in actual production environments.

[0062] Specifically, after determining the collaboration weights between heterogeneous devices, the system begins the device grouping process. First, a device relationship graph is constructed based on the collaboration weights, where nodes represent devices and edge weights indicate the strength of the collaboration relationship between devices. Using a graph community discovery algorithm, the system identifies closely collaborating groups of devices. During the grouping process, the system prioritizes devices with high collaboration weights, while also considering the balance of the number of devices within the group. Furthermore, the system assesses the impact of the grouping scheme on the overall communication load to minimize inter-group communication requirements. The resulting device collaborative networking solution not only defines the membership of each device group but also the collaborative relationships between them. This solution includes communication routing strategies, load balancing strategies, and failover mechanisms for each device group. These strategies and mechanisms are developed based on a comprehensive set of pre-collected data. In practice, the system uses a graph partitioning algorithm to group devices. This algorithm ensures high collaboration while maintaining a balanced size across device groups, ultimately forming the final device collaborative networking solution.

[0063] 102. Calculate the device's time series score based on the task completion quality, energy consumption efficiency, device stability, and collaborative adaptability of each device group in the device collaborative networking solution, and dynamically adjust the time series score based on environmental monitoring data to obtain the device credit score;

[0064] In one embodiment of the present invention, the timing score of the device is calculated based on the task completion quality, energy consumption efficiency, device stability and collaborative adaptability of each device group in the device collaborative networking scheme, and the timing score is dynamically weighted based on the environmental monitoring data to obtain the device credit score, including: performing least squares fitting calculation on the task position accuracy deviation, task completion time deviation and task execution parameter deviation of each device group in the device collaborative networking scheme to obtain a quality evaluation matrix, performing piecewise linear regression calculation on the quality evaluation matrix to obtain task quality timing data; performing wavelet decomposition calculation on the device current fluctuation data, voltage fluctuation data and power factor data of each device group in the device collaborative networking scheme to obtain an energy consumption characteristic sequence, and performing the energy consumption characteristic sequence on the device collaborative networking scheme. The energy consumption characteristic sequence is cross-correlated with the standard energy consumption model to obtain energy efficiency time series data; the equipment vibration spectrum data, temperature change data and motor speed fluctuation data of each equipment group in the equipment collaborative networking solution are subjected to empirical mode decomposition to obtain state characteristic vectors, and the state characteristic vectors are subjected to principal component analysis to obtain stability time series data; the task quality time series data, energy efficiency time series data and stability time series data are subjected to Markov chain transfer probability calculation to obtain time series scores; the regional temperature field gradient data, humidity distribution data and electromagnetic interference intensity distribution data in the environmental monitoring data are subjected to principal component analysis to obtain environmental characteristic vectors, and the environmental characteristic vectors and time series scores are subjected to neural network weight calculation to obtain equipment credit scores.

[0065] Specifically, after the device collaborative networking plan is established, the task execution performance of the devices within each device group needs to be analyzed. First, task position accuracy deviation data is collected. This refers to the error between the actual docking position of the AGV during a transport task and the preset position. This error data includes X-axis deviation, Y-axis deviation, and angular deviation. Task completion time deviation is also recorded, including time parameters such as task start delay, path execution time difference, and task completion confirmation delay. Furthermore, task execution parameter deviations are collected, such as motion parameters such as speed fluctuation, acceleration change, and steering compensation during the AGV task execution. The system uses a least-squares fitting method to calculate these multi-dimensional deviation data. By constructing a set of error equations and solving the optimal fit curve, a quality assessment matrix reflecting the device's task execution capability is generated. Next, the quality assessment matrix is ​​divided into multiple intervals based on the time series. Linear regression analysis is performed within each interval to obtain segmented linear fit results. This segmented processing approach accounts for the nonlinear temporal variations of device performance. The resulting task quality time series data reflects the performance trends of the device over different time periods.

[0066] Specifically, the analysis of equipment energy consumption begins with three key electrical parameters: current fluctuation data reflects changes in equipment energy consumption under different load conditions, achieved by collecting real-time current values ​​and recording their fluctuation range. Voltage fluctuation data characterizes power supply quality and equipment power stability, requiring monitoring of both instantaneous and effective voltage values. Power factor data reflects the equipment's energy efficiency, obtained by measuring the ratio of active power to apparent power. These raw data often contain multiple frequency components. The system uses wavelet decomposition to perform multi-scale analysis, decomposing the signal into sub-signals in different frequency bands. By selecting appropriate wavelet basis functions and decomposition levels, the system extracts the most representative energy consumption signatures. These signature sequences are cross-correlated with a pre-established standard energy consumption model, which is constructed based on typical energy consumption data under normal equipment operation. By calculating the cross-correlation coefficient between the signature sequences and the standard model, energy efficiency time series data reflecting the equipment's current energy efficiency are generated. This data clearly documents how the equipment's energy consumption level changes over time.

[0067] Specifically, evaluating equipment stability requires analyzing a variety of mechanical and thermal parameters. The system first collects the equipment's vibration spectrum data. This data, collected from vibration sensors installed at key locations on the AGV, contains information on the vibration amplitude and frequency of the equipment during operation. Simultaneously, the system monitors the equipment's temperature changes, using temperature sensors to record temperature changes in core components such as the motor, battery, and controller. Furthermore, it collects motor speed fluctuation data, which reflects the operational stability of the AGV drive system. The system processes this multi-source, heterogeneous monitoring data using empirical mode decomposition (EMD). This method adaptively decomposes complex nonlinear and nonstationary signals into a series of intrinsic mode functions, each representing a characteristic component of the original signal. By analyzing these modal functions, eigenvectors reflecting the equipment's operating status are derived. Principal component analysis is then performed on these high-dimensional eigenvectors to extract the most representative feature combinations, ultimately generating stability time series data reflecting the equipment's operational stability.

[0068] Specifically, after the system collects three sets of time series data representing different aspects of device performance, it uses Markov chain transition probability analysis to establish dynamic correlations between these data. First, each set of time series data is divided into several state intervals based on its numerical range, with each interval representing a performance level. The frequency of state transitions between adjacent moments is then statistically analyzed to construct a state transition frequency matrix. By normalizing the transition frequencies, a transition probability matrix between states is obtained. This probability matrix describes the evolution of the device's performance state, with each element representing the probability of the device transitioning from one performance state to another. Based on this transition probability matrix, the system predicts the device's future performance state and uses the degree of fit between the predicted results and the actual observed values ​​as the basis for calculating the time series score. This Markov chain-based scoring method takes into account the dynamic characteristics of device performance.

[0069] Specifically, environmental factors significantly impact the operating status of equipment, necessitating comprehensive monitoring of workshop environmental parameters. The system's deployed environmental monitoring network collects regional temperature field gradient data, which reflects the temperature distribution in different areas of the workshop. It also monitors humidity distribution data, recording the spatial distribution characteristics of air humidity. It also collects electromagnetic interference intensity distribution data, which originates from electromagnetic field intensity measurement points within the workshop. These environmental monitoring data typically have high dimensionality and require dimensionality reduction through principal component analysis. The system extracts the primary features of this data and constructs an environmental feature vector, which comprehensively reflects the overall state of the workshop environment. The environmental feature vector and the previously obtained time-series score are input into a pre-trained neural network model, which learns how environmental factors influence equipment performance. The neural network calculates the final equipment credit score, which comprehensively considers both the equipment's own performance and environmental impacts.

[0070] Furthermore, the Markov chain transition probability calculation of the task quality time series data, energy efficiency time series data and stability time series data to obtain a time series score includes: performing time series segmentation calculation on the task quality time series data to obtain a quality state sequence, performing frequency statistics on adjacent states in the quality state sequence to obtain a first-order transition frequency matrix; normalizing the energy efficiency time series data and the stability time series data to obtain a standardized data sequence, performing state quantization calculation on the standardized data sequence to obtain an energy efficiency state sequence and a stability state sequence; performing transfer probability calculation on the quality state sequence, energy efficiency state sequence and stability state sequence to obtain a Markov transfer probability matrix, performing stationary distribution calculation on the Markov transfer probability matrix to obtain a state probability vector; performing n-step transfer probability calculation based on the state probability vector and the Markov transfer probability matrix to obtain a time series score.

[0071] Specifically, when processing task quality time series data, the data sequence is first divided into several time periods according to a time window, with each period set to 30 minutes. Within each time period, the system calculates the mean and standard deviation of the task's position accuracy deviation, completion time deviation, and execution parameter deviation, and uses these statistics as the state features of that time period. Based on the distribution of these feature values, the state space is divided into multiple discrete state intervals. For example, task execution quality is classified into four levels: "Excellent," "Good," "Fair," and "Poor." By mapping the feature values ​​of each time period to the corresponding state interval, a state sequence is generated that reflects the changes in task quality. For this state sequence, the system calculates the number of transitions between states at adjacent moments, records the frequency of transitions from one state to another, and ultimately constructs a first-order transition frequency matrix. The rows and columns of this matrix correspond to the states in the state space, and the matrix elements record the number of transitions between corresponding state pairs.

[0072] Specifically, energy efficiency and stability time series data contain indicators with different dimensions and numerical ranges, necessitating data normalization for comparability. For energy efficiency time series data, the system uses maximum and minimum value normalization to map the data to the [0, 1] interval. For stability time series data, the system uses Z-score normalization to ensure the data conforms to a standard normal distribution. The normalized data series preserves the original data's trend while eliminating the effects of dimension and scale. Next, the system performs state quantization on the normalized data series, using K-means clustering to partition continuous values ​​into discrete states. Energy efficiency data is clustered into three states: "efficient," "normal," and "inefficient," while stability data is divided into three states: "stable," "fluctuating," and "abnormal." Through this quantization process, the system generates energy efficiency and stability state series, which describe the device's operating status at different moments in time using discrete state values.

[0073] Specifically, the system combines the quality state sequence, energy efficiency state sequence, and stability state sequence into a multidimensional state space, and calculates the state transition probability on this basis. First, the transition frequency between each state combination in the multidimensional state space is counted, and then the transition frequency of each state is normalized to obtain the Markov transition probability matrix. This matrix describes the probability distribution of the system transitioning from the current state to the next state, where the sum of the elements in each row of the matrix is ​​1, representing the sum of the probabilities of transitioning from a certain state to all possible states. This transition probability matrix is ​​subjected to eigenvalue decomposition and iterative calculation. When the state distribution tends to be stable, the resulting probability distribution is the stationary distribution of the Markov chain. This stationary distribution represents the frequency of occurrence of each state of the system in the long-term evolution process in the form of a probability vector, reflecting the long-term performance of the equipment performance state.

[0074] Specifically, after obtaining the state probability vector and the Markov transition probability matrix, the system needs to predict the performance state distribution of the device in the next n time steps. By performing n-th power operation on the Markov transition probability matrix, an n-step transition probability matrix is ​​obtained, which describes the probability of the system transitioning from the current state to each target state after n time steps. Multiply the current state probability vector by the n-step transition probability matrix to obtain the state probability distribution after n steps. The system compares this probability distribution with the preset performance standards, calculates the weighted average value based on the weight coefficient of each state, and finally obtains a time series score that reflects the expected level of device performance. This score comprehensively considers the historical performance and future trends of the equipment in the three dimensions of quality, energy efficiency, and stability, providing a quantitative basis for equipment performance evaluation.

[0075] 103. Initially assign tasks to devices based on their credit scores, and continuously analyze the bias of devices based on the task allocation data collected during task execution. Balance device resources based on the analysis results to obtain a dynamic task allocation strategy.

[0076] In one embodiment of the present invention, the initial task allocation to the device according to the device credit score, and the continuous bias analysis of the device based on the task allocation data collected during the task execution process, and the device resource balancing according to the analysis results to obtain a dynamic task allocation strategy include: calculating the sample standard deviation of the device credit score to obtain score dispersion data, performing probability density distribution calculation on the score dispersion data to obtain an initial task allocation probability matrix, and performing initial task allocation according to the initial task allocation probability matrix; performing time series collection and sliding time window calculation on the task allocation data after the initial task allocation to obtain a task allocation time series matrix; performing K-means clustering calculation on the task allocation time series matrix to obtain a device task density distribution map, performing skewness coefficient calculation on the device task density distribution map, grouping and marking devices whose skewness coefficient exceeds a threshold to obtain an optimized dead-end device group; performing principal component analysis calculation on the task execution data of the optimized dead-end device group to obtain a task feature matrix, and performing hierarchical clustering calculation on the task feature matrix to obtain task difficulty level data; performing Bayesian update calculation on the initial task allocation probability matrix according to the task difficulty level data to obtain a dynamic task allocation strategy.

[0077] Specifically, when processing device credit scores, the system collects scoring data for all devices within the last 24 hours, recording it every 10 minutes. The score at each point in time encompasses comprehensive performance information such as the device's task completion, energy efficiency, and device status. For example, for 10 AGVs in a factory, the system would collect 144 scores for each device over a 24-hour period. The standard deviation of each device's score is then calculated. This calculation takes the average of the 144 scores, then the difference between each score and the average. These differences are squared, summed, and then divided by the number of samples, taking the square root to obtain the standard deviation. The resulting 10 standard deviations constitute the score dispersion data. To calculate the probability density distribution of this dispersion data, kernel density estimation is used, smoothing each data point using a Gaussian kernel function. Specifically, every 0.1 unit within the data range is selected, and the sum of the contributions of all sample points at that point, calculated using the Gaussian kernel function, is calculated to obtain the probability density of that point. The bandwidth parameter is determined by calculating the standard deviation and interquartile range of the samples. A smaller bandwidth yields finer distribution details, while a larger bandwidth reduces the impact of noise. Finally, the calculated probability density is combined with each device's historical performance on different task types to establish an initial task assignment probability matrix. For example, if there are three types of handling tasks within a factory, the system will generate a 10-row, 3-column probability matrix, where each value represents the probability of the corresponding device being assigned a specific type of task.

[0078] Specifically, after initial task assignment, the system uses a distributed task monitoring network to collect task execution data in real time. Each AGV is equipped with a task execution status recorder, which records information about the currently executing task every five minutes, including task number, task type identifier, task start timestamp, estimated completion time, actual completion time, and task execution quality indicators. This collected data is transmitted to a central data processing server via industrial Ethernet. The server sets a 60-minute time window, which slides forward every five minutes, and processes the data within the window. Within each time window, the system calculates metrics such as the number of tasks per device, the proportion of task types, average execution time, and task completion rate. These statistics are arranged in chronological order by device number to form a task allocation time series matrix. For example, with 10 AGVs, after eight hours of system operation, a 10-row, 96-column time series matrix is ​​generated. Each row represents a device, and each column represents the statistical results for a time window. The matrix elements contain detailed task execution data for each device within that time window.

[0079] Specifically, when analyzing the task allocation time series matrix, the system first preprocesses the data, normalizing different types of metric data to the same scale. It then executes the K-means clustering algorithm, setting the number of clusters to three, representing high load, normal load, and low load conditions. During the clustering process, the system randomly selects three initial cluster centers, calculates the distance from each device's task allocation data to these three centers, and assigns the device to the cluster closest to them. The system then recalculates the center of each cluster, repeating this process until the cluster center no longer changes significantly. The clustering results are presented as a device task density distribution graph, with time on the horizontal axis and task density on the vertical axis. Curves of different colors represent devices with different load conditions. To calculate the skewness coefficient for this distribution graph, the data points for each curve are first sorted by size, and the median and quartiles are calculated. The degree of skewness of the distribution is determined based on the central tendency of the data. The system sets a skewness threshold of ±1.5. If a device's skewness coefficient exceeds this range, it is marked as an optimization blind spot device, indicating that the device is either chronically overloaded or chronically underloaded.

[0080] Specifically, the system collects detailed task execution data from the optimized blind spot device group, including multiple indicators such as navigation positioning accuracy, path planning time, task completion time, battery consumption rate, task switching frequency, and number of error handling. These raw data are first cleaned to remove outliers and missing values. Then, principal component analysis is performed. First, the correlation coefficient matrix between each indicator is calculated, and its eigenvalues ​​and eigenvectors are solved. The eigenvectors with the largest eigenvalues ​​are selected as principal components. A task feature matrix is ​​constructed using these principal components. Each row of the matrix represents a task, and each column represents a principal component feature. Then, a hierarchical clustering algorithm is used. First, each task is regarded as a separate class, and the Euclidean distance between classes is calculated. The two classes with the closest distance are merged each time to form a task clustering tree. Appropriate split levels are selected in the clustering tree to divide the tasks into different difficulty levels, usually divided into three levels: basic, advanced, and expert.

[0081] Specifically, the system uses the Bayesian update method to dynamically adjust the task allocation strategy. First, a priori model is constructed based on the historical work data of the equipment, recording the success rate of each device on tasks of different difficulty levels. When new task execution data is generated, these probability values ​​are updated using the Bayesian formula. During the update process, more weight is assigned to the most recent execution data, and less weight is assigned to older data. The updated probability value is multiplied by the initial task allocation probability matrix to obtain a new task allocation strategy. For example, after an AGV device completes an expert-level task, the system will increase the probability of assigning the device to high-difficulty tasks, while correspondingly reducing the probability of assigning tasks of other difficulty levels. This dynamic update mechanism ensures that the task allocation strategy can promptly reflect changes in the actual working capabilities of the equipment. The final dynamic task allocation strategy is stored in the system in the form of a probability matrix to guide the subsequent task allocation process.

[0082] Furthermore, the K-means clustering calculation is performed on the task allocation timing matrix to obtain a device task density distribution map, the skewness coefficient is calculated on the device task density distribution map, and the devices whose skewness coefficient exceeds the threshold are grouped and marked to obtain an optimized dead-end device group, including: performing Gaussian kernel density estimation calculation on each sampling point of the task allocation timing matrix to obtain a device task distribution probability map, performing gradient descent clustering calculation on the device task distribution probability map to obtain a task density cluster center; performing Euclidean distance calculation on the task density cluster center to obtain a cluster distance matrix, performing K-means iterative calculation based on the cluster distance matrix to obtain a device task density distribution map; performing probability moment estimation calculation on the device task density distribution map to obtain a distribution matrix, performing third-order moment calculation on the distribution matrix to obtain a skewness coefficient; performing dynamic threshold calculation on the devices based on the skewness coefficient to obtain a grouping threshold, comparing the grouping threshold with the skewness coefficient, grouping and marking the devices whose skewness coefficient exceeds the grouping threshold, and obtaining an optimized dead-end device group.

[0083] Specifically, when processing the task allocation time-series matrix, the system performs a Gaussian kernel density estimation on the data at each sampling point. The system first treats the task allocation data at each time point in the time-series matrix as a sample point, selects an appropriate bandwidth parameter, and applies a Gaussian kernel smoothing function to each sample point. In practice, the time-series matrix contains data sampled every five minutes. Each sampling point records information such as the number of tasks assigned to a device, task type, and execution duration. The system then performs kernel density estimation on each of these sampling points, calculating the probability density of the task distribution at each time point. By processing a large number of sampling points, the system generates a continuous probability distribution curve, forming a device task distribution probability map. Gradient descent clustering is then performed on this probability map, iteratively searching for local maxima. These points serve as cluster centers for the task density. The system sets a convergence condition during the clustering process, terminating the iteration when the gradient change falls below a preset threshold. The resulting cluster centers reflect the dominant pattern in task allocation.

[0084] Specifically, after obtaining the task density cluster centers, the system calculates the Euclidean distance between these centers. This calculation takes into account multiple dimensions, including the number of tasks, the distribution of task types, and execution time. The distance between each pair of cluster centers is recorded in a cluster distance matrix, which is a symmetric matrix with zero diagonal elements. For 10 cluster centers, a 10×10 distance matrix is ​​formed. Based on this distance matrix, the system performs a K-means iterative calculation, which consists of two steps: assignment and update. In the assignment step, each data point is assigned to the cluster center with the closest distance; in the update step, the center point position of each cluster is recalculated. This process is repeated until the clustering results stabilize. The resulting device task density distribution diagram shows the distribution of tasks assigned to different devices. Each curve in the diagram represents the changing trend of the task load of a device.

[0085] Specifically, the system further analyzes the device task density distribution map, first performing a probability moment estimation calculation. During the estimation process, the mean, variance, and other statistics are calculated for each task density curve, and these statistics form a distribution matrix. Each row in the distribution matrix corresponds to a device, and each column corresponds to a statistical feature. During data processing, the system centralizes the raw data to facilitate subsequent moment calculations. The processed distribution matrix is ​​then subjected to a third-order moment calculation, which reflects the degree of asymmetry in the data distribution. During the third-order moment calculation, the system cubes the deviation of each data point from the mean, then calculates the average value, and finally divides it by the cube of the standard deviation to obtain the skewness coefficient. This coefficient reflects the degree of skewness in the task distribution, with a positive value indicating a right-skewed distribution and a negative value indicating a left-skewed distribution.

[0086] Specifically, after obtaining the skewness coefficient, the system needs to determine an appropriate grouping threshold. This process uses a dynamic threshold calculation method. First, the skewness coefficient distribution of all devices is statistically analyzed, and the mean and standard deviation of the skewness coefficient are calculated. Based on historical operating experience, the system sets the threshold to the mean plus or minus two standard deviations. This setting effectively identifies devices with abnormal task allocation. If a device's skewness coefficient falls outside this range, the system marks it as a device for optimization. During the comparison process, the system considers both positive and negative skewness. Excessively positive skewness indicates an over-concentration of tasks on a device, while excessively negative skewness indicates an under-allocation of tasks. Ultimately, these marked devices are assigned to a group of devices with poor optimization potential, and these devices will be prioritized for optimization in subsequent task allocation. In this way, the system can promptly detect and correct imbalances in task allocation.

[0087] 104. According to the task allocation strategy, the production efficiency, energy utilization rate and task completion quality of the equipment group are evaluated for synergistic effects, and based on the evaluation results, the member composition and capability complementarity of the equipment group are dynamically optimized to obtain the collaborative evolution plan of the equipment group.

[0088] In one embodiment of the present invention, the synergistic effect evaluation of the production efficiency, energy utilization rate and task completion quality of the equipment group is performed according to the task allocation strategy, and the member composition and capability complementarity of the equipment group are dynamically optimized based on the evaluation results to obtain a collaborative evolution plan of the equipment group, including: time series acquisition of the equipment group task throughput data, equipment group task response time data and equipment group task accuracy data during the execution of the task allocation strategy to obtain equipment group production efficiency evaluation data; weighted calculation of the equipment group energy consumption data, equipment group standby time data and equipment group peak power data during the execution of the task allocation strategy to obtain equipment group energy utilization evaluation data; hierarchical analysis calculation of the equipment group according to the production efficiency evaluation data and energy utilization evaluation data to obtain an equipment group performance evaluation matrix, eigenvalue decomposition calculation of the equipment group performance evaluation matrix to obtain an equipment group capability eigenvector; optimal matching calculation of the equipment group members according to the equipment group capability eigenvector to obtain an equipment group member reorganization plan, and predictive calculation of the equipment group collaborative efficiency in the equipment group member reorganization plan to obtain a collaborative evolution plan of the equipment group.

[0089] Specifically, during the execution of the task allocation strategy, the system monitors the operational status of the equipment group in real time through a distributed data collection network. Each equipment group is equipped with a dedicated data collection module that collects task throughput data at 10-second intervals, recording the number of tasks completed per unit time. For example, a group consisting of five AGVs will generate 2,880 throughput sampling points during an 8-hour operation cycle. The system also collects task response time data, including information such as the wait time from receiving the task instruction to the start of execution and the actual time taken for task execution. For task accuracy data, the system records metrics such as position accuracy and posture accuracy after task completion. After preprocessing, this raw data is arranged in chronological order to form a multidimensional time series. The system then applies sliding window processing to this time series data, with a window size of 30 minutes and a sliding period of 10 minutes. Statistics such as the mean and standard deviation are calculated within each window, ultimately generating evaluation data reflecting the overall production efficiency of the equipment group.

[0090] Specifically, the equipment group generates a large amount of energy-related data during operation. The system first collects real-time energy consumption data, including the instantaneous power consumption and cumulative power consumption of each device. This data is recorded every minute using an industrial-grade energy collector. The system also records the equipment's standby time—the period during which the equipment is idle but still consuming energy. Furthermore, the system monitors the equipment group's peak power data, recording power peaks and their duration during production. These three types of data have varying levels of importance, and the system assigns weights based on their impact on energy efficiency. Typically, energy consumption data is weighted 0.5, standby time data 0.3, and peak power data 0.2. These weighted data are normalized to the same numerical range and then linearly combined to generate a comprehensive assessment of the equipment group's energy efficiency. This assessment objectively reflects the overall energy efficiency performance of the equipment group.

[0091] Specifically, after obtaining production efficiency and energy utilization evaluation data, the system uses a hierarchical analysis method to conduct a comprehensive evaluation of the equipment group. First, an evaluation index system is established, consisting of two major categories: a production efficiency index group and an energy efficiency index group. The production efficiency index group includes three secondary indicators: task throughput, response time, and accuracy; the energy efficiency index group includes three secondary indicators: energy consumption level, standby time, and peak management. Expert scoring is used to determine the weights of each level of indicators, and a judgment matrix is ​​constructed. The judgment matrix is ​​then subjected to a consistency check to ensure the reliability of the evaluation results. On this basis, an equipment group performance evaluation matrix is ​​obtained, which comprehensively reflects the performance of the equipment group across different indicators. Eigenvalue decomposition is then performed on the performance evaluation matrix to obtain an eigenvector that characterizes the core capabilities of the equipment group. This eigenvector contains the main performance characteristics of the equipment group and provides a basis for subsequent optimization.

[0092] Specifically, based on the equipment group capability feature vector, the system begins to perform member optimization matching calculations. First, a device capability model is constructed, which includes the degree of expertise of each device in different task types. For example, in an equipment group that mixes AGVs of multiple brands, some devices excel at precision positioning tasks, while other devices perform better in long-distance transportation tasks. The system conducts a complementary analysis based on these specialties to find the optimal equipment combination. The optimization process uses a genetic algorithm to encode the combination of equipment into chromosomes and find the optimal solution through multiple generations of evolution. For each recombination plan obtained, the system will build a prediction model based on historical data to calculate the expected collaborative efficiency of the plan in the future. Finally, the plan with the highest expected efficiency is selected as the collaborative evolution plan for the equipment group. This plan specifies in detail the member composition, task division strategy and collaboration mechanism of the equipment group.

[0093] In this embodiment, the affinity matrix between devices is calculated based on the physical location, communication delay, device type, and historical collaborative relationship of the devices, and the collaborative weights of heterogeneous devices are determined to obtain a device collaborative networking solution. Then, a timing score is calculated based on the task completion quality, energy consumption efficiency, device stability, and collaborative adaptability of each device group, and the weight is dynamically adjusted in combination with environmental monitoring data to obtain a device credit score. Then, initial task allocation is performed based on the device credit score, and continuous bias analysis and resource balancing are performed by analyzing the data during task execution to obtain a dynamic task allocation strategy. Finally, the collaborative effect of the device group is evaluated based on the task allocation strategy, and the device group is dynamically optimized to obtain a collaborative evolution solution. The present invention effectively solves the problem of unreasonable device resource allocation through a dynamic evaluation and balancing mechanism.

[0094] The above describes the production collaborative management method of the digital factory in the embodiment of the present invention. The following describes the production collaborative management system of the digital factory in the embodiment of the present invention. Figure 2 An embodiment of the production collaborative management system of a digital factory in an embodiment of the present invention includes:

[0095] The networking module 201 is configured to calculate an affinity matrix between devices based on the physical location, communication delay, device type, and historical collaboration relationship of devices in each area of ​​the digital factory, and determine the collaboration weights of heterogeneous devices based on the affinity matrix to obtain a device collaborative networking solution;

[0096] Scoring module 202, configured to calculate a time series score of a device based on the task completion quality, energy consumption efficiency, device stability, and collaborative adaptability of each device group in the device collaborative networking solution, and dynamically adjust the weight of the time series score based on environmental monitoring data to obtain a device credit score;

[0097] Scheduling module 203, configured to perform initial task allocation to devices based on the device credit scores, perform continuous bias analysis on the devices based on the task allocation data collected during task execution, balance device resources based on the analysis results, and obtain a dynamic task allocation strategy;

[0098] The optimization module 204 is used to evaluate the synergistic effect of the production efficiency, energy utilization and task completion quality of the equipment group according to the task allocation strategy, and dynamically optimize the member composition and capability complementarity of the equipment group based on the evaluation results to obtain a collaborative evolution plan for the equipment group.

[0099] In an embodiment of the present invention, the production collaborative management system of the digital factory runs the production collaborative management method of the digital factory described above. The production collaborative management system of the digital factory calculates the affinity matrix between devices based on the physical location, communication delay, device type and historical collaborative relationship of the devices, and determines the collaborative weight of heterogeneous devices to obtain a device collaborative networking solution; then, a timing score is calculated based on the task completion quality, energy consumption efficiency, device stability and collaborative adaptability of each device group, and the weight is dynamically adjusted in combination with environmental monitoring data to obtain a device credit score; then, initial task allocation is performed based on the device credit score, and continuous bias analysis and resource balancing are performed by analyzing the data during the task execution process to obtain a dynamic task allocation strategy; finally, the collaborative effect of the device group is evaluated according to the task allocation strategy, and the device group is dynamically optimized to obtain a collaborative evolution solution. The present invention effectively solves the problem of unreasonable device resource allocation through a dynamic evaluation and balancing mechanism.

[0100] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0101] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0102] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A production collaborative management method for a digital factory, characterized in that: The production collaborative management method of the digital factory includes: Calculate the affinity matrix between devices based on the physical location, communication delay, device type, and historical collaboration relationship of devices in each area of ​​the digital factory. Then, determine the collaboration weights of heterogeneous devices based on the affinity matrix to obtain a device collaborative networking solution. Calculating the device's time series score based on the task completion quality, energy consumption efficiency, device stability, and collaborative adaptability of each device group in the device collaborative networking solution, and dynamically weighting the time series score based on environmental monitoring data to obtain a device credit score; Initial task allocation is performed on the device based on the device credit score, and continuous bias analysis is performed on the device based on the task allocation data collected during task execution. Device resources are balanced based on the analysis results to obtain a dynamic task allocation strategy; According to the task allocation strategy, the production efficiency, energy utilization rate and task completion quality of the equipment group are evaluated for synergy effect, and based on the evaluation results, the member composition and capability complementarity of the equipment group are dynamically optimized to obtain a collaborative evolution plan for the equipment group.

2. The production collaborative management method of a digital factory according to claim 1, characterized in that: The device collaborative networking solution includes calculating the affinity matrix between devices based on the physical location, communication delay, device type, and historical collaboration relationship of devices in each area of ​​the digital factory, and determining the collaboration weights of heterogeneous devices based on the affinity matrix. Perform Euclidean distance calculation on the physical location data of the device to obtain a location distance matrix, and perform sampling calculation on the round-trip delay of communication data packets between devices to obtain a communication delay matrix; The compatibility of communication protocols of different types of devices is calculated to obtain the device type affinity matrix, and the historical collaboration data of devices is processed with a time decay function to obtain the collaboration intensity matrix. Performing a multi-dimensional weighted fusion calculation based on the location distance matrix, the communication delay matrix, the device type affinity matrix, and the collaboration strength matrix to obtain an affinity matrix between devices; Evaluating the communication efficiency between the heterogeneous device pairs in the affinity matrix to obtain an initial collaboration weight, and dynamically adjusting the initial collaboration weight according to the collaboration success rate of the heterogeneous device pairs under different working conditions to obtain a collaboration weight between the heterogeneous devices; The devices are grouped and divided based on the collaboration weights to obtain a device collaborative networking solution.

3. The production collaborative management method of a digital factory according to claim 1, characterized in that: The device credit score is obtained by calculating the device's time series score based on the task completion quality, energy consumption efficiency, device stability, and collaborative adaptability of each device group in the device collaborative networking solution, and dynamically adjusting the time series score based on environmental monitoring data. The following includes: Performing least squares fitting calculation on the task position accuracy deviation, task completion time deviation, and task execution parameter deviation of each device group in the device collaborative networking solution to obtain a quality assessment matrix, and performing piecewise linear regression calculation on the quality assessment matrix to obtain task quality time series data; Perform wavelet decomposition calculation on the device current fluctuation data, voltage fluctuation data, and power factor data of each device group in the device collaborative networking solution to obtain an energy consumption characteristic sequence, and perform cross-correlation calculation on the energy consumption characteristic sequence and the standard energy consumption model to obtain energy efficiency time series data; Perform empirical mode decomposition on the equipment vibration spectrum data, temperature change data, and motor speed fluctuation data of each equipment group in the equipment collaborative networking solution to obtain state eigenvectors. Perform principal component analysis on the state eigenvectors to obtain stability time series data. Performing Markov chain transition probability calculation on the task quality time series data, energy efficiency time series data, and stability time series data to obtain a time series score; Principal component analysis is performed on the regional temperature field gradient data, humidity distribution data, and electromagnetic interference intensity distribution data in the environmental monitoring data to obtain an environmental feature vector. Neural network weight calculation is performed on the environmental feature vector and the time series score to obtain the equipment credit score.

4. The production collaborative management method of a digital factory according to claim 3, characterized in that: The Markov chain transition probability calculation is performed on the task quality time series data, energy efficiency time series data and stability time series data to obtain the time series score, which includes: Perform time series segmentation calculation on the task quality time series data to obtain a quality state sequence, perform frequency statistics on adjacent states in the quality state sequence to obtain a first-order transfer frequency matrix; Normalizing the energy efficiency time series data and the stability time series data to obtain a standardized data sequence, and performing state quantization calculation on the standardized data sequence to obtain an energy efficiency state sequence and a stability state sequence; Performing transition probability calculation on the quality state sequence, energy efficiency state sequence, and stability state sequence to obtain a Markov transition probability matrix, and performing stationary distribution calculation on the Markov transition probability matrix to obtain a state probability vector; n-step transition probability calculation is performed according to the state probability vector and the Markov transition probability matrix to obtain a time series score.

5. The production collaborative management method of a digital factory according to claim 1, characterized in that: The initial task allocation is performed on the device according to the device credit score, and the device is continuously biased based on the task allocation data collected during the task execution process. The device resources are balanced according to the analysis results to obtain a dynamic task allocation strategy including: Calculating the sample standard deviation of the device credit score to obtain score dispersion data, calculating the probability density distribution of the score dispersion data to obtain an initial task allocation probability matrix, and performing initial task allocation based on the initial task allocation probability matrix; Performing time series acquisition and sliding time window calculation on the task allocation data after the initial task allocation to obtain a task allocation time series matrix; Performing K-means clustering calculation on the task allocation time series matrix to obtain a device task density distribution map, calculating a skewness coefficient on the device task density distribution map, grouping and marking devices whose skewness coefficient exceeds a threshold, and obtaining an optimized blind spot device group; Performing principal component analysis on the task execution data of the blind spot optimization equipment group to obtain a task feature matrix, and performing hierarchical clustering on the task feature matrix to obtain task difficulty level data; A Bayesian update calculation is performed on the initial task allocation probability matrix according to the task difficulty level data to obtain a dynamic task allocation strategy.

6. The production collaborative management method of a digital factory according to claim 5, characterized in that: The K-means clustering calculation is performed on the task allocation time series matrix to obtain a device task density distribution map, the skewness coefficient is calculated on the device task density distribution map, and the devices whose skewness coefficient exceeds a threshold are grouped and marked to obtain the optimized blind spot device group. Performing Gaussian kernel density estimation calculation on each sampling point of the task allocation timing matrix to obtain a device task distribution probability map, and performing gradient descent clustering calculation on the device task distribution probability map to obtain a task density cluster center; Performing Euclidean distance calculation on the task density cluster centers to obtain a cluster distance matrix, and performing K-means iterative calculation based on the cluster distance matrix to obtain a device task density distribution map; Performing probability moment estimation calculation on the equipment task density distribution map to obtain a distribution matrix, and performing third-order moment calculation on the distribution matrix to obtain a skewness coefficient; A dynamic threshold value is calculated for the devices according to the skewness coefficient to obtain a grouping threshold value, the grouping threshold value is compared with the skewness coefficient, and the devices whose skewness coefficient exceeds the grouping threshold value are grouped and marked to obtain an optimized blind spot device group.

7. The production collaborative management method of a digital factory according to claim 1, characterized in that: The collaborative effect evaluation of the production efficiency, energy utilization rate and task completion quality of the equipment group is performed according to the task allocation strategy, and the member composition and capability complementarity of the equipment group are dynamically optimized based on the evaluation results to obtain the collaborative evolution plan of the equipment group, including: Time series data of the equipment group task throughput, the equipment group task response time and the equipment group task accuracy are collected during the execution of the task allocation strategy to obtain equipment group production efficiency evaluation data; Performing weighted calculation on the equipment group energy consumption data, equipment group standby time data, and equipment group peak power data during the execution of the task allocation strategy to obtain equipment group energy utilization evaluation data; Performing a hierarchical analysis calculation on the equipment group based on the production efficiency evaluation data and the energy utilization rate evaluation data to obtain an equipment group performance evaluation matrix, and performing an eigenvalue decomposition calculation on the equipment group performance evaluation matrix to obtain an equipment group capability eigenvector; An optimal matching calculation is performed on the device group members according to the device group capability feature vector to obtain a device group member reorganization plan, and a device group collaborative efficiency in the device group member reorganization plan is predicted and calculated to obtain a collaborative evolution plan for the device group.

8. A production collaborative management system for a digital factory, characterized by: The production collaborative management system of the digital factory includes: The networking module is used to calculate the affinity matrix between devices based on the physical location, communication delay, device type, and historical collaboration relationship of devices in each area of ​​the digital factory, and determine the collaboration weights of heterogeneous devices based on the affinity matrix to obtain a device collaboration networking solution; a scoring module for calculating a time series score of a device based on the task completion quality, energy consumption efficiency, device stability, and collaborative adaptability of each device group in the device collaborative networking solution, and dynamically weighting the time series score based on environmental monitoring data to obtain a device credit score; A scheduling module is used to perform initial task allocation to devices based on the device credit score, and to continuously analyze the bias of devices based on the task allocation data collected during task execution. Based on the analysis results, device resources are balanced to obtain a dynamic task allocation strategy. The optimization module is used to evaluate the synergistic effect of the production efficiency, energy utilization and task completion quality of the equipment group according to the task allocation strategy, and dynamically optimize the member composition and capability complementarity of the equipment group based on the evaluation results to obtain a collaborative evolution plan for the equipment group.

Images