Modular optimization method for industrial production process

Abstract

The invention provides a modular optimization method for an industrial production process, which belongs to the technical field of electrical digital data processing, and comprises the following steps: firstly, constructing a real-time requirement matrix to perform real-time grading on modules; a time-sensitive network architecture is deployed to ensure that key signal transmission delay is controllable; edge computing nodes are introduced to reduce data transmission delay; a low-delay communication network is constructed based on an industrial 5G technology; real-time deviation fluctuation matrix monitoring is implemented to capture delay abnormity; adaptive control compensation is realized by applying a real-time adjustment gain matrix; establishing a time deterministic task scheduler to optimize task execution; implementing a cross-module synchronous clock mechanism to eliminate a time synchronization error; a modular real-time optimization engine is constructed to realize self-tuning of the system, and the problem of real-time control failure caused by communication delay and scheduling uncertainty in modular industrial production is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electronic digital data processing, and in particular relates to a modular optimization method for industrial production processes. Background Art

[0002] Modularization of industrial production processes is a key development direction for modern intelligent manufacturing. Traditional modular process systems rely primarily on standard industrial communication protocols such as Modbus, PROFINET, and EtherCAT for inter-module communication and integration, with central control systems coordinating the operation of each module. These technologies are widely used in conventional production environments, meeting basic automation requirements and supporting process parameter monitoring and control.

[0003] However, with the increasing sophistication of industrial production, traditional modular process systems face serious challenges. First, traditional industrial communication protocols suffer from uncertain delays under high loads, preventing critical control instructions from reaching the execution units in a timely manner. Second, centralized computing architectures result in excessively long data transmission paths, increasing processing latency. Third, fixed-gain control strategies are unable to dynamically compensate for the effects of delay fluctuations, reducing the accuracy of process parameter control. Finally, insufficient clock synchronization accuracy between modules causes coordination errors in distributed control systems.

[0004] These issues are particularly acute in industrial scenarios requiring high precision and dynamic response, such as precision chemicals, semiconductor manufacturing, and aerospace component processing. Communication delays and scheduling uncertainties between modules can lead to failures in real-time control of key process parameters, which can in turn cause product quality fluctuations, reduced production efficiency, and even safety incidents. A modular process optimization approach that effectively addresses this real-time control failure is urgently needed. Summary of the Invention

[0005] In view of this, the present invention provides an industrial production process modular optimization method, which can solve the technical problem in the prior art that communication delay and scheduling uncertainty in the industrial production process modular system lead to real-time control failure.

[0006] The present invention is implemented as follows: The present invention provides a modular optimization method for industrial production processes, including constructing a real-time requirement matrix, dividing the real-time level of each module of the industrial production process, and assigning higher priority to modules with high real-time requirements; deploying a time-sensitive network architecture, and implementing deterministic Ethernet technology on the communication channel between modules; introducing edge computing nodes, and assigning real-time deviation matrix calculation tasks to edge servers close to the production site; constructing a low-latency communication network based on industrial 5G technology; implementing real-time deviation fluctuation matrix monitoring; applying a real-time adjustment gain matrix, and adaptively compensating process parameter control instructions based on the control compensation mechanism equation; establishing a time-deterministic task scheduler; implementing a cross-module synchronous clock mechanism; constructing a modular real-time optimization engine, and continuously optimizing the real-time adjustment gain matrix parameters through an adaptive feedback optimization function.

[0007] Among them, the real-time requirement matrix refers to the quantification of each process module in industrial production into a matrix form according to the response time requirements, which includes module identifiers and corresponding maximum allowable response times, and is used for system resource allocation and task priority sorting.

[0008] The real-time deviation matrix is ​​formed by monitoring the communication and control execution time of each module in real time and calculating the difference between the actual response time and the theoretical value in the real-time requirement matrix. It is used to evaluate the real-time performance of the system.

[0009] The real-time deviation fluctuation matrix refers to a multidimensional data structure that records the changing trend of the real-time deviation within a certain time window. By statistically analyzing the deviation change rate and fluctuation amplitude, the system performance degradation trend is predicted.

[0010] The real-time adjustment gain matrix refers to an adaptive compensation parameter matrix set for each control loop based on the real-time deviation and its fluctuation, which is used to dynamically adjust the control instructions to compensate for the impact of communication delay.

[0011] Among them, the control compensation mechanism equation refers to the mathematical expression for calculating the optimal compensation amount based on the communication delay time, the rate of change of process parameters, the system sensitivity coefficient, the response time constant and the process state variable, and outputs the real-time compensation amount of the control instruction to ensure that the process parameter control can still be maintained within the target range in the presence of communication delay.

[0012] Among them, the adaptive feedback optimization function refers to a mathematical function that dynamically adjusts the weights of each parameter in the real-time gain matrix according to the historical error accumulation value, current performance indicators, system feedback speed, time weighting factor and disturbance resistance coefficient, and outputs the optimized gain matrix parameter set.

[0013] Among them, the communication delay time refers to the time required for instructions or data to be transmitted from the source module to the target module in the industrial control network, which is obtained by monitoring the time-sensitive network architecture; the process parameter change rate refers to the speed at which key process parameters change over time in the industrial production process, which is calculated from the process parameter sequence collected by the real-time data bus.

[0014] Among them, the system sensitivity coefficient refers to the response degree of process parameters to changes in control instructions, which is obtained by analyzing the process parameter response curve through edge computing nodes; the response time constant refers to the characteristic time required for the process system to respond to the control signal and reach a steady state, which is provided by the real-time requirement matrix.

[0015] Among them, the process state variable refers to the set of key parameters that describe the current process state, which are collected from each process module through a low-latency communication network; the historical error accumulation value refers to the weighted accumulation of system control errors over a period of time in the past, which is calculated by the real-time deviation fluctuation matrix.

[0016] This paper establishes a comprehensive real-time control assurance system through a series of innovative technical approaches, including constructing a real-time requirement matrix, deploying a time-sensitive network architecture, and introducing edge computing nodes. This method accurately quantifies the real-time requirements of process modules and allocates network and computing resources accordingly, ensuring that modules with high real-time requirements receive priority processing.

[0017] By implementing this invention, industrial production systems can effectively reduce the uncertainty of inter-module communication delays and control critical signal transmission delays within preset thresholds. A real-time deviation fluctuation matrix monitoring mechanism promptly captures delay fluctuations, triggering adaptive compensation based on the control compensation mechanism equation, thereby reducing the negative impact of delays on process control accuracy. Furthermore, a time-deterministic task scheduler and a cross-module synchronized clock mechanism jointly ensure the coordination and determinism of distributed control.

[0018] Based on the above-mentioned technical means, the present invention successfully solves the technical problem of real-time control failure caused by communication delay and scheduling uncertainty in the modular system of industrial production processes. While maintaining the flexibility of the modular architecture, the system achieves reliable real-time control of highly dynamic and high-precision process parameters, providing a practical technical solution for contemporary high-end manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flow chart of the method of the present invention.

[0020] Figure 2 This is a diagram analyzing the real-time requirements of the modules in Example 2.

[0021] Figure 3 This is a fluctuation analysis diagram of the real-time deviation of the communication link in Example 2.

[0022] Figure 4 This is a comparison before and after the gain matrix optimization in Example 2. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0024] like Figure 1 FIG. 1 is a flow chart of a modular optimization method for an industrial production process provided by the present invention, and the method comprises the following steps:

[0025] S01. Build a real-time requirement matrix to classify the real-time requirements of each module in the industrial production process, and assign higher priority to modules with high real-time requirements.

[0026] S02. Deploy a time-sensitive network architecture and implement deterministic Ethernet technology on inter-module communication channels to ensure that the transmission delay of critical signals does not exceed the preset threshold;

[0027] S03. Introduce edge computing nodes to distribute real-time deviation matrix calculation tasks to edge servers close to the production site to reduce data transmission delays.

[0028] S04. Build a low-latency communication network based on industrial 5G technology and establish a unified real-time data bus for all inter-module communications;

[0029] S05. Implement real-time deviation fluctuation matrix monitoring to capture inter-module communication delay fluctuations and trigger compensation mechanisms when the delay exceeds the threshold;

[0030] S06. Apply real-time adjustment of the gain matrix to adaptively compensate the process parameter control instructions based on the control compensation mechanism equation to reduce the impact of communication delay on process control accuracy;

[0031] S07. Establish a time-deterministic task scheduler, construct modular process control tasks into a directed acyclic graph structure, and allocate deterministic time slots for key process control processes based on a minimum delay critical path algorithm;

[0032] S08. Implement a cross-module synchronous clock mechanism based on the precise time protocol to ensure that all modules share a unified time base and eliminate time synchronization errors;

[0033] S09. Build a modular real-time optimization engine to continuously optimize and adjust the gain matrix parameters in real time through adaptive feedback optimization functions to achieve system self-tuning.

[0034] Among them, the real-time requirement matrix refers to the quantification of each process module in industrial production into a matrix form according to the response time requirements, which includes module identifiers and corresponding maximum allowable response times, and is used for system resource allocation and task priority sorting.

[0035] The real-time deviation matrix is ​​formed by monitoring the communication and control execution time of each module in real time and calculating the difference between the actual response time and the theoretical value in the real-time requirement matrix. It is used to evaluate the real-time performance of the system.

[0036] The real-time deviation fluctuation matrix refers to a multidimensional data structure that records the changing trend of the real-time deviation within a certain time window. By statistically analyzing the deviation change rate and fluctuation amplitude, the system performance degradation trend is predicted.

[0037] The real-time adjustment gain matrix refers to an adaptive compensation parameter matrix set for each control loop based on the real-time deviation and its fluctuation, which is used to dynamically adjust the control instructions to compensate for the impact of communication delay.

[0038] Among them, the time-sensitive network architecture refers to a deterministic network technology architecture that complies with the IEEE 802.1TSN family of standards, which ensures the determinism and low latency of critical data stream transmission through mechanisms such as time synchronization, traffic shaping, and frame preemption.

[0039] Among them, deterministic Ethernet technology refers to a collection of technologies that achieve deterministic transmission delay based on standard Ethernet, including time-aware scheduler, bandwidth reservation mechanism and traffic priority management function.

[0040] Among them, edge computing nodes refer to computing devices deployed close to the data source, which are responsible for executing real-time computing tasks that are sensitive to delays and avoiding additional delays caused by transmitting data to remote servers.

[0041] Among them, industrial 5G technology refers to the fifth-generation mobile communication technology optimized for industrial application scenarios. It has ultra-reliable and low-latency communication capabilities and supports millisecond-level end-to-end communication delay.

[0042] Among them, the Precision Time Protocol refers to the high-precision time synchronization protocol defined by the IEEE 1588 standard, which can achieve nanosecond-level clock synchronization accuracy between network devices.

[0043] Among them, the control compensation mechanism equation refers to the mathematical expression for calculating the optimal compensation amount based on the communication delay time, process parameter change rate, system sensitivity coefficient, response time constant and process state variable, and outputs the real-time compensation amount of the control instruction to ensure that the process parameter control can still remain within the target range in the presence of communication delay; among them, the communication delay time refers to the time required for instructions or data to be transmitted from the source module to the target module in the industrial control network, which is obtained by monitoring the time-sensitive network architecture; the process parameter change rate refers to the speed at which the key process parameters change over time in the industrial production process, which is calculated from the process parameter sequence collected by the real-time data bus; the system sensitivity coefficient refers to the degree of response of the process parameters to changes in control instructions, which is obtained by analyzing the process parameter response curve through the edge computing node; the response time constant refers to the characteristic time required for the process system to reach a steady state in response to the control signal, which is provided by the real-time requirement matrix; the process state variable refers to the set of key parameters that describe the current process state, which is collected from each process module through a low-latency communication network.

[0044] Among them, the directed acyclic graph structure refers to a mathematical structure that decomposes process control tasks into nodes and directed edges. The nodes represent control tasks, and the directed edges represent the dependencies between tasks.

[0045] Among them, the minimum delay critical path algorithm refers to an algorithm that finds the longest path from the source node to the target node in a directed acyclic graph to ensure that critical control tasks are scheduled first.

[0046] Among them, the deterministic time slot refers to a fixed time interval pre-allocated in the time deterministic task scheduler, which is used to ensure that the execution of critical tasks is not interfered with by other tasks.

[0047] Among them, the adaptive feedback optimization function refers to a mathematical function that dynamically adjusts the weights of each parameter in the real-time adjustment gain matrix according to the historical error accumulation value, current performance index, system feedback speed, time weighting factor and disturbance resistance coefficient, and outputs the optimized gain matrix parameter set; the historical error accumulation value refers to the weighted accumulation of system control errors in the past period of time, which is calculated by the real-time deviation fluctuation matrix; the current performance index refers to the performance evaluation value of the current operation of the process control system, which is calculated by the modular real-time optimization engine based on the process quality parameters; the system feedback speed refers to the rate at which the control system responds to deviations, which is obtained through the rate of change of the real-time deviation matrix; the time weighting factor refers to the coefficient that assigns different weights to data at different time points in the optimization calculation, which is provided by the time-deterministic task scheduler; the disturbance resistance coefficient refers to a quantitative indicator of the system's ability to resist external interference, which is obtained by analyzing the fluctuation pattern in the real-time deviation fluctuation matrix.

[0048] The specific implementation of the above steps is described in detail below. The specific implementation of step S01 is to construct a real-time requirement matrix for industrial production process modules using the Analytic Hierarchy Process (AHP). First, response time requirement data for each process module is collected, including module identifiers and maximum allowable response times. A judgment matrix is ​​then established, and the real-time requirements of each module are compared pairwise. Eigenvalues ​​and eigenvectors are calculated to obtain the real-time weight of each module. The modules are then classified into three real-time performance levels based on the weights: high real-time modules (response time requirements less than 10 milliseconds), medium real-time modules (response time requirements between 10 and 100 milliseconds), and low real-time modules (response time requirements greater than 100 milliseconds). Finally, the real-time requirement matrix is ​​formed, with high real-time modules assigned priority values ​​of 9 to 10, medium real-time modules assigned priority values ​​of 5 to 8, and low real-time modules assigned priority values ​​of 1 to 4. The purpose of this step is to achieve rational allocation of system resources and ensure that critical modules have sufficient computing and network resources.

[0049] The specific implementation of step S02 involves deploying a time-sensitive network architecture based on the IEEE 802.1TSN family of standards. First, a network topology design is performed to identify critical data flow paths and mark communication links between modules with high real-time requirements. Next, a time-aware shaper based on IEEE 802.1Qbv is configured to divide the network cycle into eight time slots, with four reserved for high-real-time data flows. Next, the IEEE 802.1Qbu frame preemption mechanism is implemented, allowing high-priority data frames to interrupt the transmission of lower-priority data frames, ensuring that the transmission delay of high-real-time signals does not exceed a preset threshold of 1 millisecond. IEEE 802.1Qci flow-based ingress policies are then configured to limit bandwidth usage by non-critical data flows. Finally, the IEEE 802.1Qch round-robin queuing and forwarding mechanism is deployed to achieve deterministic data transmission, ensuring network jitter is less than 100 microseconds. This step ensures that communications between key modules have deterministic latency characteristics, supporting high-real-time industrial control applications.

[0050] The specific implementation method of step S03 is to introduce edge computing nodes based on a heterogeneous computing architecture. First, at least three edge servers are deployed according to the physical layout of the factory to ensure that the physical distance between each high real-time module and the nearest edge server does not exceed 100 meters. Then configure the edge node computing resources, and each node is equipped with at least an 8-core processor, 16GB of memory and a hard real-time operating system. Then develop a task allocation algorithm based on resource awareness to give priority to allocating real-time deviation matrix calculation tasks to edge nodes with a computing load of less than 50%. Then implement a task migration strategy based on computing affinity. When the edge node load exceeds 80%, non-critical computing tasks are automatically migrated to the cloud. Finally, a two-way data caching mechanism is established to achieve incremental data synchronization between edge nodes and central servers, reducing redundant data transmission by 95%. This step effectively reduces the delay in round-trip data transmission by delegating delay-sensitive computing tasks to the edge, thereby improving the real-time response capability of the system.

[0051] The specific implementation method of step S04 is to build a low-latency communication network based on industrial 5G technology. First, 5G base stations are deployed to achieve full coverage in the factory area, with the base station spacing not exceeding 200 meters. Then configure the 5G network slice, establish an independent URLLC (ultra-reliable low-latency communication) slice for industrial control traffic, and reserve at least 20% of network resources. Then implement wireless resource scheduling optimization, adopt a semi-persistent scheduling mechanism to reduce signaling overhead, and ensure that the uplink scheduling delay is less than 0.5 milliseconds. Then establish a real-time data bus based on a distributed bus architecture, adopt a publish / subscribe mode to realize data exchange between modules, support data transmission priority division and service quality assurance. Finally, configure the network quality monitoring mechanism to monitor the end-to-end delay in real time, and trigger link switching when the delay exceeds 3 milliseconds. This step provides a unified low-latency data exchange platform for all inter-module communications, achieving millisecond-level communication performance.

[0052] The specific implementation method of step S05 is to implement real-time deviation fluctuation matrix monitoring. First, deploy distributed monitoring probes, arrange timestamp collection points on the communication links between modules, and the sampling frequency is not less than 1000Hz. Then calculate the real-time deviation matrix, compare the actual response time with the theoretical value in the real-time requirement matrix, and form a deviation data set. Then build an exponentially weighted moving average model, calculate the deviation change trend within a 30-second sliding window, and identify abnormal fluctuation patterns. Then implement a threshold-based alarm mechanism. When the communication delay deviation exceeds the set threshold (1 millisecond for high real-time modules, 10 milliseconds for medium real-time modules, and 50 milliseconds for low real-time modules), the corresponding compensation mechanism is triggered. Finally, build a historical fluctuation database to store at least 7 days of fluctuation data to support long-term performance analysis and optimization. This step provides a data basis for subsequent compensation and optimization by capturing the communication delay fluctuations between modules in real time.

[0053] The specific implementation method of step S06 is to apply the real-time adjustment gain matrix for control compensation. First, a mathematical model is established based on the control compensation mechanism equation, and the input includes communication delay time, process parameter change rate, system sensitivity coefficient, response time constant and process state variable. Then, the optimal compensation amount is calculated, and feedforward compensation is performed on the process parameter control instructions to offset the impact of communication delay. Then, adaptive gain adjustment is implemented, and the compensation parameters are dynamically updated according to the real-time deviation matrix. The gain adjustment range is 0.8 to 1.2 times the nominal value. Then, a model predictive control framework is established to predict the impact of communication delay on process parameters and generate compensation strategies in advance. Finally, a control effect evaluation mechanism is implemented to calculate the control accuracy improvement before and after compensation, achieving at least 90% accuracy improvement for the high real-time module. This step reduces the negative impact of communication delay on process control accuracy through an intelligent compensation algorithm, thereby improving the overall performance of the system.

[0054] The specific implementation method of step S07 is to establish a time-deterministic task scheduler. First, the process control flow dependencies are analyzed and the modular process control tasks are mapped into a directed acyclic graph structure, where nodes represent control tasks and directed edges represent dependencies. Then, a topological sorting algorithm is applied to determine the order of task execution to avoid resource conflicts and deadlocks. Next, task optimization based on the minimum delay critical path algorithm is implemented to identify the critical path that affects the overall response time of the system and assign the highest execution priority to the tasks on the critical path. Then, a deterministic time slot allocation mechanism is constructed to divide the system scheduling cycle (usually 10 milliseconds) into multiple time slots, reserving at least 40% of time resources for critical process control flows. Finally, a real-time scheduling strategy based on rate monotonicity is implemented to ensure that high-frequency periodic tasks receive the highest priority and ensure scheduling feasibility. This step ensures the timely execution of critical control tasks and reduces system uncertainty through a deterministic task scheduling mechanism.

[0055] The specific implementation method of step S08 is to implement a cross-module clock synchronization mechanism. First, deploy the IEEE 1588 Precision Time Protocol network and select network devices that support hardware timestamps as the infrastructure. Then configure the master-slave clock hierarchy, select the edge node with the highest stability as the master clock source, and the other nodes as slave clocks. Next, implement the precise time synchronization algorithm to achieve nanosecond-level clock synchronization accuracy by measuring network path delay and clock offset, ensuring that the synchronization error is less than 100 nanoseconds. Then establish a clock offset monitoring mechanism to regularly compare the local time of each module with the master clock reference time and record the offset trend. Finally, implement time synchronization redundancy protection, configure at least one backup master clock source, and automatically switch to it when the master clock source fails to ensure high availability of time synchronization services. This step eliminates time synchronization errors between modules by establishing a unified time base, providing a time basis for deterministic control.

[0056] The specific implementation method of step S09 is to build a modular real-time optimization engine. First, an adaptive feedback optimization function is designed. The input includes the historical error accumulation value, the current performance index, the system feedback speed, the time weighting factor and the disturbance resistance coefficient, and the optimized gain matrix parameter set is output. Then, parameter optimization based on a genetic algorithm is implemented. The optimal gain matrix parameter combination is found by simulating the evolutionary process. The population size is set to 50, and the evolutionary generations are no less than 20. Then, a performance evaluation framework is established, defining a multi-objective evaluation function including control accuracy, response time and stability, and quantifying the overall performance of the system. Then, an online learning mechanism is implemented to continuously adjust the optimization model parameters based on historical operation data to adapt to process changes. The initial value of the learning rate is set to 0.05 and gradually decreases to 0.01 as the system stability increases. Finally, an optimization effect verification process is established to compare the improvement degree of real-time indicators before and after optimization to ensure that the system performance is improved by at least 20%. This step realizes system self-tuning by continuously optimizing the real-time adjustment gain matrix parameters, improving the real-time control capability and production efficiency of the overall industrial production process.

[0057] The mathematical model or calculation process involved in the present invention is described in detail below.

[0058] The real-time requirement matrix construction process in step S01 involves the analytic hierarchy process (AHP) calculation, which is specifically expressed as follows:

[0059] W=[w1,w2,...,w n ] T ;

[0060] Where W is the module real-time weight vector; w i is the real-time weight of the i-th module, i = 1, 2, ..., n; n is the total number of modules.

[0061] The weight vector calculation process is as follows:

[0062] A=[a ij ] n×n ;

[0063] Where A is the judgment matrix; a ij is the real-time importance ratio of module i to module j, a ij =1 / a ji , a ii =1.

[0064] AW=λ max W;

[0065] Where λ max is the maximum eigenvalue of the judgment matrix A.

[0066]

[0067] Where CI is the consistency index, which is used to evaluate the consistency of the judgment matrix.

[0068]

[0069] Where CR is the consistency ratio; RI is the random consistency index, whose value is obtained by looking up the table based on the matrix order. For example, when n = 3, RI = 0.58; when n = 4, RI = 0.90. When CR < 0.1, the matrix is ​​judged to have satisfactory consistency.

[0070] The real-time classification standards are as follows:

[0071]

[0072] Where, L i is the real-time level of module i; T i The maximum allowed response time of module i, in milliseconds.

[0073] The priority assignment rules are:

[0074]

[0075] Where, P i is the priority value of module i; H is the set of modules with high real-time requirements; M is the set of modules with medium real-time requirements; and L is the set of modules with low real-time requirements. This formula normalizes the weights within each real-time requirement level, ensuring that modules with high real-time requirements receive a priority of 9 to 10, modules with medium real-time requirements receive a priority of 5 to 8, and modules with low real-time requirements receive a priority of 1 to 4.

[0076] The final real-time requirement matrix is ​​expressed as:

[0077] R=[r ij ] n×2 ;

[0078] Where R is the real-time requirement matrix; r i1 is the identifier of module i; r i2 is the priority value P of module i i .

[0079] The calculation process of the real-time deviation fluctuation matrix in step S05 is as follows:

[0080] D=[d ij ] n×m ;

[0081] Where D is the real-time deviation matrix; d ij is the real-time deviation value of the i-th module at the j-th sampling moment; n is the total number of modules; m is the number of sampling points.

[0082]

[0083] Where, t ij is the actual response time of the i-th module at the j-th sampling moment; The theoretical maximum allowable response time of the i-th module in the real-time requirement matrix.

[0084] The real-time deviation fluctuation matrix is ​​calculated using the exponentially weighted moving average model:

[0085] V=[v ij ] n×k ;

[0086] Where V is the real-time deviation fluctuation matrix; v ij is the deviation fluctuation value of the i-th module in the j-th time window; k is the number of time windows.

[0087] v ij =α·d i，m-k+j +(1-α)·v i，j-1 ;

[0088] Where α is the smoothing factor, the range is [0.1, 0.3], and the default value is generally 0.2; v i，0 is the initial value, usually d i，m-k .

[0089] Fluctuation trend calculation:

[0090] Δv ij =v ij -v i，j-1 ;

[0091] Where Δv ij is the fluctuation rate of the i-th module in the j-th time window.

[0092] Abnormal fluctuation judgment criteria:

[0093]

[0094] Where A i is the abnormal fluctuation sign of module i; θ i is the deviation threshold of module i, which is determined according to the real-time level. It is 1 millisecond for high real-time modules, 10 milliseconds for medium real-time modules, and 50 milliseconds for low real-time modules; δ i is the fluctuation rate threshold of module i, usually set to θ i 10% to 20% of the total.

[0095] The control compensation mechanism equation in step S06 is expressed as follows:

[0096] C(t)=C0(t)+ΔC(t);

[0097] Where C(t) is the compensated control instruction; C0(t) is the original control instruction; ΔC(t) is the compensation amount.

[0098] Compensation calculation:

[0099]

[0100] Where G is the real-time adjustment gain matrix; τ is the communication delay time vector; is the process parameter change rate vector; s is the system sensitivity coefficient vector; T c is the response time constant vector; X is the process state variable vector; f is the compensation function.

[0101] The real-time adjustment gain matrix is ​​expressed as:

[0102] G=[g ij ] n×n ;

[0103] Where g ij is the compensation gain coefficient of the i-th control loop for the j-th process parameter. For independent control loops, G is usually a diagonal matrix, i.e. g ij =0(i≠j).

[0104] The specific form of the compensation function is:

[0105]

[0106] Where φ(X) is the process state adjustment function, which is used to adjust the compensation amount according to the current process state.

[0107]

[0108] Where x i is the i-th process state variable; is the standard operating point of the i-th process state variable; β i is the influence coefficient of the i-th state variable; q is the number of state variables.

[0109] Adaptive gain adjustment:

[0110] g ii (t+1)=g ii (t)·(1+γ·sign(e)·|e| α );

[0111] Where, e is the control error; γ is the learning rate, ranging from [0.01, 0.1]; α is the nonlinear factor, ranging from [0.5, 1.5], usually 1; sign is the sign function. Gain value g ii The adjustment range is [0.8, 1.2] times the nominal value.

[0112] The minimum delay critical path algorithm in step S07 involves the critical path calculation of the directed acyclic graph:

[0113] G = (V, E);

[0114] Where G is a directed acyclic graph; V is a set of nodes, representing control tasks; and E is a set of directed edges, representing the dependencies between tasks.

[0115] Earliest start time calculation:

[0116]

[0117] Where, ES(v j ) is the node v j The earliest starting time; t(v i ) is the node v i The execution time of the source node v s , ES(v s )=0.

[0118] Latest start time calculation:

[0119]

[0120] Where, LS(v i ) is the node v i The latest start time. For the sink node v t , LS(v t )=ES(v t ).

[0121] Time margin calculation:

[0122] Slack (v i )=LS(v i )-ES(v i );

[0123] Where, Slack(v i ) is the node v i The time slack of nodes on the critical path is 0.

[0124] Deterministic time slot allocation is based on a rate monotonicity algorithm:

[0125]

[0126] Where U is the processor utilization; C i is the execution time of task i; T i is the period of task i. When the above conditions are met, the system is schedulable.

[0127] Time slot allocation priority:

[0128]

[0129] Where, P i is the scheduling priority of task i. The shorter the period, the higher the priority.

[0130] The adaptive feedback optimization function in step S09 is expressed as follows:

[0131] G new =G old +ΔG;

[0132] Where G new is the optimized gain matrix; G old is the gain matrix before optimization; ΔG is the gain adjustment amount.

[0133] Gain adjustment calculation:

[0134]

[0135] Where η is the learning rate, which has an initial value of 0.05 and gradually decreases to 0.01 as the system stability improves; is the gradient of the performance index J with respect to the gain matrix G.

[0136] Performance indicator function:

[0137] J(G)=w1J1(G)+w2J2(G)+w3J3(G)+w4J4(G)+w5J5(G);

[0138] Where, J1(G) is the control accuracy index; J2(G) is the response time index; J3(G) is the stability index; J4(G) is the robustness index; J5(G) is the energy consumption index; w1 to w5 are weight coefficients, satisfying

[0139] Calculation of each sub-indicator:

[0140]

[0141] Where e(t) is the control error and T is the evaluation time window.

[0142]

[0143] Where, is the response time of the i-th key process parameter; n is the number of key process parameters.

[0144] J3(G)=λ max (A-BG);

[0145] Where λ max (A-BG) is the real part of the maximum eigenvalue of the system state matrix A-BG, which is used to evaluate the system stability.

[0146]

[0147] Where, ΔJ(G, Δp j ) is the system parameter p j Change Δp j The performance index changes caused by this; m is the number of system parameters.

[0148]

[0149] Where u(t) is the control input; R is the positive definite weight matrix.

[0150] The parameter optimization process based on genetic algorithm includes:

[0151] 1. Encoding: Encode the non-zero elements of the gain matrix G into chromosomes.

[0152] 2. Fitness function: Where ∈ is a small positive number to prevent the denominator from being zero.

[0153] 3. Selection: Use the roulette wheel method to select individuals, and the selection probability is proportional to the fitness.

[0154] 4. Crossover: Use arithmetic crossover, G new =αG1+(1-α)G2, where α∈[0,1] is the cross coefficient.

[0155] 5. Mutation: with probability p m Perform Gaussian mutation on chromosomes. Where N(0, σ 2 ) has a mean of 0 and a variance of σ 2 Gaussian random variable.

[0156] Genetic algorithm parameter settings: population size is 50, evolutionary generations are not less than 20, crossover probability p c =0.8, mutation probability p m =0.1, standard deviation of variation σ = 0.05.

[0157] Optionally, the time-sensitive network architecture configuration in step S02 involves frame transmission delay calculation:

[0158] D total =D prop +D trans +D proc +D queue ;

[0159] Where D total is the total transmission delay; D prop is the propagation delay, which is related to the link length and signal propagation speed; D trans is the transmission delay, which is equal to the frame size divided by the link bandwidth; D proc To process delay, it is related to the processing capacity of network equipment; D queue is the queuing delay, which is related to the network load and scheduling algorithm.

[0160] Optional, time-sensitive network key performance indicator calculation:

[0161] J net =max(D total )-min(D total );

[0162] Where, J net Network jitter refers to the variation range of transmission delay. TSN network requires it to be less than 100 microseconds.

[0163] Optionally, the industrial 5G network slice resource allocation calculation in step S04:

[0164] R URLLC =α·R total ;

[0165] Where R URLLC The amount of resources allocated for the URLLC slice; R total is the total network resources; α is the resource allocation ratio, which is at least 0.2.

[0166] Optionally, the precision time protocol synchronization error calculation in step S08 is:

[0167]

[0168] Where ΔT is the clock offset; δ is the inherent deviation between the master and slave clocks; t1 is the timestamp when the master clock sends the synchronization message; t2 is the timestamp when the slave clock receives the synchronization message; t3 is the timestamp when the slave clock sends the delay request; t4 is the timestamp when the master clock receives the delay request; ∈ is the measurement error, which is usually less than 10 nanoseconds.

[0169] Optional, clock synchronization accuracy requirements:

[0170] |ΔT|<100ns;

[0171] Where |ΔT| is the absolute value of the clock offset, which the system requires to be less than 0.1 microsecond.

[0172] Specifically, the principle behind this invention is as follows: The core principle behind this invention's solution to the real-time control failure problem in modular industrial production process systems lies in constructing a multi-layered real-time assurance mechanism that, through synergy, eliminates the impact of communication delays and scheduling uncertainty on the control system. First, a real-time requirement matrix quantifies and ranks each module based on its response time requirements, providing a scientific basis for system resource allocation and ensuring that limited resources prioritize the needs of critical modules. Second, a time-sensitive network architecture, combined with deterministic Ethernet technology, achieves network-level latency determinism, keeping data transmission delays within a predictable range.

[0173] In terms of computing architecture, this invention introduces edge computing nodes, deploying latency-sensitive computing tasks close to the data source, significantly reducing data transmission paths and fundamentally reducing processing latency. The application of industrial 5G technology provides a high-bandwidth, low-latency wireless channel for inter-module communication, enhancing system deployment flexibility. The real-time deviation matrix and the real-time deviation fluctuation matrix constitute the system performance monitoring layer, achieving closed-loop control by capturing delay fluctuations and triggering compensation mechanisms.

[0174] The innovation of this invention lies in the design of a real-time adjustment gain matrix and control compensation mechanism equation. Based on multi-dimensional data such as communication delay time, process parameter change rate, and system sensitivity coefficient, it calculates the optimal compensation amount, enabling control instructions to adaptively compensate for the effects of communication delay. This control strategy, based on a combination of prediction and compensation, maintains process parameter control accuracy despite communication delays.

[0175] The time-deterministic task scheduler constructs process control tasks as a directed acyclic graph. Using a minimum-delay critical path algorithm, it identifies critical control processes and assigns them deterministic time slots to ensure uninterrupted execution. The cross-module synchronized clock mechanism achieves nanosecond-level clock synchronization through a precise time protocol, resolving timing issues in the coordinated control of distributed systems.

[0176] Finally, a modular real-time optimization engine continuously adjusts system parameters through an adaptive feedback optimization function, enabling the system to self-adjust as the environment changes, maintaining optimal real-time control performance. This multi-level, closed-loop optimization technology architecture conforms to the principles of cybernetics and can effectively solve real-time control problems in industrial production.

[0177] A specific embodiment 1 of the present invention is provided below. The specific implementation of each step in this embodiment 1 is described in detail as follows.

[0178] The specific implementation of step S01 is to construct the real-time requirement matrix of industrial production process modules through the analytic hierarchy process (AHP). First, the response time requirement data of each process module is collected, including the module identifier and the maximum allowed response time. Then, a judgment matrix A = [a ij ] n×n , where A is the judgment matrix, a ij is the real-time importance ratio of module i to module j, a ij =1 / a ji , a ii =1. Then calculate the eigenvalue and eigenvector to obtain the module real-time weight vector W = [w1, w2, ..., w n ] T , satisfying AW=λ max W, where λ max is the maximum eigenvalue of the judgment matrix A. To ensure the consistency of the judgment matrix, the consistency index is calculated and consistency ratio Where RI is the random consistency index, which is obtained by looking up the table based on the matrix order, and requires CR<0.1. Then, based on the weight, the modules are divided into three levels of real-time performance: Where L i is the real-time level of module i, T i is the maximum allowed response time of module i. Finally, according to the priority allocation rule Forming the real-time requirement matrix R = [r ij ] n×2 , where r i1 is the identifier of module i, r i2 is the priority value P of module i i H represents the set of modules with high real-time requirements, M represents the set of modules with medium real-time requirements, and L represents the set of modules with low real-time requirements. This priority allocation rule ensures that modules with high real-time requirements receive a priority of 9 to 10, modules with medium real-time requirements receive a priority of 5 to 8, and modules with low real-time requirements receive a priority of 1 to 4. This step uses the Analytic Hierarchy Process (AHP) to precisely quantify the real-time requirements of each module, achieving a reasonable allocation of system resources and ensuring that critical modules receive sufficient computing and network resources.

[0179] The specific implementation of step S02 is to deploy a time-sensitive network architecture based on the IEEE 802.1TSN family of standards. First, a network topology design is performed to identify critical data flow paths and mark the communication links between high-real-time modules. Then, a time-aware shaper based on IEEE 802.1Qbv is configured to divide the network cycle into eight time slots, of which four are reserved exclusively for high-real-time data flows. The IEEE 802.1Qbu frame preemption mechanism is then implemented to allow high-priority data frames to interrupt the transmission of low-priority data frames. The frame transmission delay is calculated using formula D total =D prop +D trans +D proc +D queue , where D total is the total transmission delay, D prop is the propagation delay, which is related to the link length and signal propagation speed, D trans is the transmission delay, which is equal to the frame size divided by the link bandwidth, D proc To process the delay, it is related to the processing capacity of the network equipment. queue It is the queuing delay, which is related to the network load and scheduling algorithm. By controlling each delay component, it ensures that the high real-time signal transmission delay does not exceed the preset threshold of 1 millisecond. net =max(D total )-min(D total ), where J net Network jitter represents the variation in transmission delay, ensuring it is less than 100 microseconds. IEEE 802.1Qci flow-based ingress policies are then configured to limit bandwidth usage by non-critical data flows. Finally, the IEEE 802.1Qch round-robin queuing and forwarding mechanism is deployed to achieve deterministic data transmission. This step utilizes time-sensitive networking technology to ensure deterministic latency in communications between critical modules, supporting high-real-time industrial control applications.

[0180] The specific implementation of step S03 is the same as above and will not be described in detail here.

[0181] The specific implementation of step S04 is to build a low-latency communication network based on industrial 5G technology. First, 5G base stations are deployed to achieve full coverage in the factory area, with the base station spacing not exceeding 200 meters. Then, 5G network slices are configured to establish independent URLLC (Ultra-Reliable Low-Latency Communication) slices for industrial control traffic, and resource allocation is based on the formula R URLLC =α·R total , where R URLLC The amount of resources allocated for the URLLC slice, R totalWhere α is the total network resources and α is the resource allocation ratio, which must be at least 0.2, meaning that at least 20% of network resources are reserved. Next, wireless resource scheduling optimization is implemented, using a semi-persistent scheduling mechanism to reduce signaling overhead and ensure uplink scheduling latency is less than 0.5 milliseconds. A real-time data bus based on a distributed bus architecture is established, using a publish / subscribe model for inter-module data exchange, supporting data transmission prioritization and ensuring quality of service. Finally, a network quality monitoring mechanism is configured to monitor end-to-end latency in real time and trigger link switching when latency exceeds 3 milliseconds. This step provides a unified low-latency data exchange platform for all inter-module communications, achieving millisecond-level communication performance.

[0182] The specific implementation of step S05 is to implement real-time deviation fluctuation matrix monitoring. First, deploy distributed monitoring probes and arrange timestamp collection points on the communication links between modules with a sampling frequency of not less than 1000Hz. Then calculate the real-time deviation matrix D = [d ij ] n×m , where D is the real-time deviation matrix, d ij is the real-time deviation value of the i-th module at the j-th sampling moment, n is the total number of modules, and m is the number of sampling points. The real-time deviation value is calculated as Where t ij is the actual response time of the i-th module at the j-th sampling moment, is the theoretical maximum allowable response time of the ith module in the real-time requirement matrix. Then construct the real-time deviation fluctuation matrix V = [v ij ] n×k , where V is the real-time deviation fluctuation matrix, v ij is the deviation fluctuation value of the i-th module in the j-th time window, and k is the number of time windows. The exponentially weighted moving average model is used to calculate v ij =α·d i，m-k+j +(1-α)·v i，j-1 , where α is the smoothing factor, ranging from [0.1, 0.3], and generally defaults to 0.2. The fluctuation trend is calculated as Δv ij =v ij -v i，j-1 , where Δv ij is the fluctuation change rate of the i-th module in the j-th time window. Then implement the threshold-based alarm mechanism, and the abnormal fluctuation judgment standard is Where A i is the abnormal fluctuation sign of module i, θ i is the deviation threshold of module i, which is determined according to the real-time level. The high real-time module is 1 millisecond, the medium real-time module is 10 milliseconds, and the low real-time module is 50 milliseconds. i is the fluctuation rate threshold of module i, usually set to θ iThe delay is 10% to 20%. When the communication delay deviation exceeds the set threshold, the corresponding compensation mechanism is triggered. Finally, a historical fluctuation database is built, storing at least 7 days of fluctuation data to support long-term performance analysis and optimization. This step captures inter-module communication delay fluctuations in real time, providing a data foundation for subsequent compensation and optimization.

[0183] The specific implementation of step S06 is to apply real-time adjustment of the gain matrix for control compensation. First, a mathematical model is established based on the control compensation mechanism equation. The control instruction calculation adopts the formula C(t)=C0(t)+ΔC(t), where C(t) is the compensated control instruction, C0(t) is the original control instruction, and ΔC(t) is the compensation amount. The compensation amount is calculated as Where G is the real-time adjustment gain matrix, τ is the communication delay time vector, is the process parameter change rate vector, s is the system sensitivity coefficient vector, T c is the response time constant vector, X is the process state variable vector, and f is the compensation function. The real-time adjustment gain matrix is ​​expressed as G = [g ij ] n×n , where g ij is the compensation gain coefficient of the i-th control loop for the j-th process parameter. For independent control loops, G is usually a diagonal matrix, i.e. g ij =0(i≠j). The specific form of the compensation function is Where φ(X) is the process state adjustment function, which is calculated as Where x i is the i-th process state variable, is the standard operating point of the i-th process state variable, β i is the influence coefficient of the i-th state variable, and q is the number of state variables. Then, adaptive gain adjustment is implemented, using formula g ii (t+1)=g ii (t)·(1+γ·sign(e)·|e| α ), where e is the control error, γ is the learning rate, the range is [0.01, 0.1], α is the nonlinear factor, the range is [0.5, 1.5], usually 1, and sign is the sign function. Gain value g ii The adjustment range is [0.8, 1.2] times the nominal value. A model predictive control framework is then established to predict the impact of communication delay on process parameters and generate compensation strategies in advance. Finally, a control effectiveness evaluation mechanism is implemented to calculate the control accuracy improvement before and after compensation, achieving at least a 90% improvement in accuracy for high-real-time modules. This step uses an intelligent compensation algorithm to reduce the negative impact of communication delay on process control accuracy, improving overall system performance.

[0184] The specific implementation of step S07 is to establish a time-deterministic task scheduler. First, the process control flow dependencies are analyzed and the modular process control tasks are mapped into a directed acyclic graph structure G = (V, E), where G is a directed acyclic graph, V is a node set representing the control tasks, and E is a directed edge set representing the dependencies between tasks. Then, a topological sorting algorithm is applied to determine the order of task execution to avoid resource conflicts and deadlocks. Next, task optimization based on the minimum delay critical path algorithm is implemented to calculate the earliest start time. Where ES(v j ) is the node v j The earliest starting time, t(v i ) is the node v i The execution time of the source node v s , ES(v s )=0. Calculate the latest start time Where LS(v i ) is the node v i The latest start time. For the sink node v t , LS(v t )=ES(v t ). Calculate the time margin Slack(v i )=LS(v i )-ES(v i ), where Slack(v i ) is the node v i The time margin of the nodes on the critical path is 0. Identify the critical path that affects the overall response time of the system and assign the highest execution priority to the tasks on the critical path. Then build a deterministic time slot allocation mechanism to divide the system scheduling cycle (usually 10 milliseconds) into multiple time slots, reserving at least 40% of the time resources for the critical process control process. Finally, implement a real-time scheduling strategy based on rate monotonicity, and the processor utilization meets Where U is the processor utilization, C i is the execution time of task i, T i is the period of task i. When the above conditions are met, the system is schedulable. The task scheduling priority is calculated as Where P i is the scheduling priority of task i. The shorter the period, the higher the priority. This ensures that high-frequency periodic tasks receive the highest priority and ensures scheduling feasibility. This step uses a deterministic task scheduling mechanism to ensure the timely execution of key control tasks and reduce system uncertainty.

[0185] The specific implementation of step S08 is to implement a cross-module clock synchronization mechanism. First, deploy the IEEE 1588 precision time protocol network and select network devices with hardware timestamp support as the infrastructure. Then configure the master-slave clock hierarchy, select the edge node with the highest stability as the master clock source, and the other nodes as slave clocks. Then implement the precise time synchronization algorithm, and calculate the clock offset using the formula Where ΔT is the clock offset, δ is the inherent deviation between the master and slave clocks, t1 is the timestamp when the master sends a synchronization message, t2 is the timestamp when the slave receives the synchronization message, t3 is the timestamp when the slave sends a delay request, and t4 is the timestamp when the master receives the delay request. ∈ is the measurement error, typically less than 10 nanoseconds. The system requires clock synchronization accuracy |ΔT| < 100ns, where |ΔT| is the absolute value of the clock offset. By measuring network path delay and clock offset, nanosecond-level clock synchronization accuracy is achieved, ensuring synchronization error is less than 100 nanoseconds. A clock offset monitoring mechanism is then established to regularly compare each module's local time with the master clock reference time and record offset trends. Finally, time synchronization redundancy is implemented by configuring at least one backup master clock source, which automatically switches to another in the event of failure, ensuring high availability of the time synchronization service. This step eliminates time synchronization errors between modules by establishing a unified time base, providing a time foundation for deterministic control.

[0186] The specific implementation of step S09 is to build a modular real-time optimization engine. First, the adaptive feedback optimization function is designed, and the gain matrix optimization adopts the iterative formula G new =G old +ΔG, where G new is the optimized gain matrix, G old is the gain matrix before optimization, and ΔG is the gain adjustment amount. The gain adjustment amount is calculated as Where η is the learning rate, which has an initial value of 0.05 and gradually decreases to 0.01 as the system stability improves. is the gradient of the performance index J with respect to the gain matrix G. The performance index function is J(G) = w1J1(G) + w2J2(G) + w3J3(G) + w4J4(G) + w5J5(G), where J1(G) is the control accuracy index, J2(G) is the response time index, J3(G) is the stability index, J4(G) is the robustness index, J5(G) is the energy consumption index, and w1 to w5 are weight coefficients that satisfy The calculation of each sub-indicator is as follows: (control accuracy index), (Response time index), J3(G) = λ max (A-BG)(stability index), (robustness index), (Energy consumption index). Then the parameter optimization based on genetic algorithm is implemented, and the fitness function is Where ∈ is a small positive number to prevent the denominator from being zero. The roulette wheel method is used to select individuals, and the arithmetic crossover operation is G new =αG1+(1-α)G2, where α∈[0,1] is the cross coefficient. The Gaussian mutation operation is Where N(0,σ 2 ) has a mean of 0 and a variance of σ 2 Gaussian random variable. Genetic algorithm parameter settings: population size is 50, evolutionary generations are not less than 20 generations, crossover probability p c =0.8, mutation probability p m =0.1, standard deviation of variation σ=0.05. Then, a performance evaluation framework is established, and a multi-objective evaluation function including control accuracy, response time and stability is defined to quantify the overall performance of the system. Then, an online learning mechanism is implemented to continuously adjust the optimization model parameters based on historical operation data to adapt to process changes. The initial value of the learning rate is set to 0.05 and gradually decreases to 0.01 as the system stability improves. Finally, an optimization effect verification process is established to compare the improvement of real-time indicators before and after optimization to ensure that the system performance is improved by at least 20%. This step realizes system self-tuning by continuously optimizing the real-time adjustment gain matrix parameters, thereby improving the real-time control capability and production efficiency of the overall industrial production process.

[0187] To better understand and implement the present invention, Example 2 of a specific application scenario of the present invention is provided below: In a certain intelligent manufacturing laboratory, researchers implemented a modular real-time optimization method for industrial production processes for a six-axis industrial robot control system. The robot control system consists of multiple functional modules, including a motion planning module, a trajectory generation module, a joint control module, a visual perception module, a force feedback module, and an operation status monitoring module. Due to communication delays between the various modules of the system, the robot's control accuracy is insufficient when performing high-precision assembly tasks. In particular, when coordinated control of vision and force feedback is required, response delays seriously affect the operation effect.

[0188] First, the researchers used the analytic hierarchy process to construct a real-time requirement matrix. By analyzing and evaluating the response time requirements of each module, they established the judgment matrix shown in Table 1:

[0189] Table 1 Judgment matrix of real-time importance of each module

[0190]

[0191] The calculated weight vector W = [0.042, 0.093, 0.362, 0.172, 0.276, 0.055] T, the consistency ratio CR = 0.026 < 0.1, meeting the consistency requirements. After response time analysis, joint control and force feedback were determined to be high real-time modules (response time requirements of 0.5 milliseconds and 2 milliseconds, respectively), visual perception and trajectory generation were determined to be medium real-time modules (response time requirements of 15 milliseconds and 30 milliseconds, respectively), and motion planning and operation status monitoring were determined to be low real-time modules (response time requirements of 100 milliseconds and 200 milliseconds, respectively). The final real-time requirement matrix is ​​shown in Table 2:

[0192] Table 2 Real-time requirement matrix results

[0193] Module identifier Real-time level Maximum allowed response time (ms) Priority value Joint control high 0.5 9.57 force feedback high 2 9.43 Visual Perception middle 15 7.26 Trajectory Generation middle 30 5.74 Motion Planning Low 100 1.43 Operation status monitoring Low 200 1.57

[0194] Figure 2 The comparison of real-time requirements of six different modules in the industrial robot control system is shown. The chart adopts a dual Y-axis design. The blue bar chart represents the maximum allowable response time of each module (in milliseconds, logarithmic scale), and the red line chart represents the priority value. The modules are arranged from high to low according to the real-time level: joint control, force feedback (high real-time), visual perception, trajectory generation (medium real-time), motion planning, and operation status monitoring (low real-time). The figure clearly marks the real-time level of each module (with different color boxes) and shows the inverse relationship between response time and priority. The researchers then deployed a time-sensitive network architecture based on the IEEE 802.1TSN standard. The network topology design is shown in Table 3:

[0195] Table 3 TSN network configuration parameters

[0196]

[0197]

[0198] Next, three edge computing nodes were deployed, located near the industrial robot control cabinet, vision system, and force feedback system, with a maximum physical distance of 30 meters. The edge node computing resource configuration is shown in Table 4:

[0199] Table 4 Edge computing node configuration

[0200]

[0201] The researchers used industrial 5G networks to build a low-latency communication network, configured URLLC network slicing, reserved 30% of network resources for control traffic, and achieved communication performance with an end-to-end average latency of 0.8 milliseconds.

[0202] During implementation, the system monitored and collected real-time deviation data from the communication links between modules, generating a real-time deviation matrix. Using an exponentially weighted moving average model (smoothing factor α = 0.2), the deviation fluctuation matrix was calculated, revealing significant periodic delay fluctuations in the communication link from joint control to force feedback. The peak deviation reached 0.8 milliseconds, exceeding the 0.5 millisecond threshold. Figure 3 This figure shows the fluctuations in real-time performance deviations for four key communication links in the system, with time (seconds) on the horizontal axis and communication latency (milliseconds) on the vertical axis. The different colored curves represent the communication links between different modules: joint control → force feedback (red), visual perception → trajectory generation (green), trajectory generation → joint control (blue), and force feedback → force control execution (purple). The figure uses dotted lines to mark the latency thresholds for each link, and a red semi-transparent area highlights the portion of the "joint control → force feedback" link where the threshold is exceeded.

[0203] To address this issue, researchers applied real-time adjustment of the gain matrix for control compensation. Based on the system identification results, an initial gain matrix was constructed as shown in Table 5:

[0204] Table 5 Initial gain matrix

[0205]

[0206]

[0207] By establishing a time-deterministic task scheduler, researchers mapped robot control tasks into a directed acyclic graph (DAG) structure and identified the critical path: visual perception → trajectory generation → joint control → force feedback → force control execution. They prioritized resource allocation for control tasks along this critical path, reserving 50% of the time for critical tasks. Using the IEEE 1588 Precision Time Protocol, they achieved cross-module clock synchronization with a stable synchronization accuracy of less than 75 nanoseconds.

[0208] Finally, a modular real-time optimization engine was constructed, using a genetic algorithm to optimize the gain matrix parameters. The population size was set to 50, the number of evolutionary generations was 30, the crossover probability was 0.8, and the mutation probability was 0.1. The optimization results showed that the gain coefficient for joint 3 control was adjusted from 1.12 to 1.16, and the gain coefficient for force control Fz was adjusted from 1.20 to 1.25. Other parameters were also adjusted accordingly. Figure 4 A side-by-side bar chart compares the gain matrix parameter changes before and after optimization. The horizontal axis represents the nine control parameters, and the vertical axis represents the gain coefficient values. Blue bars represent the values ​​before optimization, and red bars represent the values ​​after optimization. The specific values ​​are labeled above each bar. The percentage of parameter change is indicated between the two groups of bars (green indicates an increase).

[0209] The system performance test results are shown in Table 6:

[0210] Table 6 Performance comparison before and after system optimization

[0211] Performance indicators Before optimization After optimization Improvement Average control accuracy error (mm) 0.28 0.06 78.6％ Maximum control accuracy error (mm) 0.65 0.15 76.9％ Average response time (ms) 12.5 4.8 61.6％ Maximum response time (ms) 25.3 8.2 67.6％ System stability index 0.76 0.92 21.1％ Tracking error (mm) 0.42 0.11 73.8％ Force control error (N) 0.58 0.16 72.4％ Anti-interference ability index 0.63 0.87 38.1％

[0212] Traditional real-time control methods for industrial robots primarily rely on increasing hardware performance, simplifying control algorithms, or reducing system complexity to meet real-time requirements. However, these approaches often come at the expense of system functionality and control accuracy. For example, common communication optimization methods include adopting hard real-time operating systems, using fieldbuses, and simplifying control models. While these methods can improve real-time performance to a certain extent, they cannot effectively address communication delays in modular systems, especially in complex systems where different modules have different real-time requirements.

[0213] In contrast, the modular real-time optimization method for industrial production processes proposed in this paper addresses this problem through several innovative approaches: first, the analytic hierarchy process (AHP) is used to quantify the real-time requirements of each module, enabling precise resource allocation. Second, a deterministic communication infrastructure is constructed using time-sensitive networking and edge computing technologies. Furthermore, a real-time deviation fluctuation matrix monitoring and a real-time adjustment gain matrix compensation mechanism are introduced to not only monitor communication delays but also intelligently compensate for their effects. Finally, adaptive feedback optimization enables continuous system self-tuning. Implementation results demonstrate that this method significantly improves the real-time performance and control accuracy of industrial control systems without increasing hardware costs, achieving technological advancements such as a 75% reduction in control error and a 65% reduction in response time. This method provides a systematic solution for the modularization and real-time optimization of industrial production processes.

[0214] It should be noted that the variables involved in the present invention are explained in detail as shown in Tables 7 and 8 below.

[0215] Table 7 Variable Explanation Table (Part 1)

[0216]

[0217]

[0218] Table 8 Variable Explanation Table (Part 2)

[0219]

[0220]

[0221] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A modular optimization method for industrial production processes, characterized in that: This includes building a real-time requirement matrix, classifying the real-time levels of each module in the industrial production process, and assigning higher priority to modules with high real-time requirements; Deploy a time-sensitive network architecture and implement deterministic Ethernet technology on inter-module communication channels; introduce edge computing nodes to distribute real-time deviation matrix calculation tasks to edge servers close to the production site; build a low-latency communication network based on industrial 5G technology; and implement real-time deviation fluctuation matrix monitoring. Apply real-time adjustment of the gain matrix to adaptively compensate process parameter control instructions based on the control compensation mechanism equation; establish a time-deterministic task scheduler; Implement a cross-module synchronous clock mechanism; build a modular real-time optimization engine to continuously optimize and adjust the gain matrix parameters in real time through adaptive feedback optimization functions.

2. The industrial production process modular optimization method according to claim 1, characterized in that: The real-time requirement matrix refers to the quantification of each process module in industrial production according to the response time requirements into a matrix form, which includes module identifiers and corresponding maximum allowable response times, and is used for system resource allocation and task priority sorting.

3. The industrial production process modular optimization method according to claim 2, characterized in that: The real-time deviation matrix is ​​formed by monitoring the communication and control execution time of each module in real time and calculating the difference between the actual response time and the theoretical value in the real-time requirement matrix. It is used to evaluate the real-time performance of the system.

4. The industrial production process modular optimization method according to claim 3, characterized in that: The real-time deviation fluctuation matrix refers to a multidimensional data structure that records the real-time deviation change trend within a certain time window. By statistically analyzing the deviation change rate and fluctuation amplitude, the system performance degradation trend is predicted.

5. The industrial production process modular optimization method according to claim 4, characterized in that: The real-time adjustment gain matrix refers to an adaptive compensation parameter matrix set for each control loop based on the real-time deviation and its fluctuation. It is used to dynamically adjust the control instructions to compensate for the impact of communication delay.

6. The industrial production process modular optimization method according to claim 5, characterized in that: The control compensation mechanism equation refers to the mathematical expression that calculates the optimal compensation amount based on the communication delay time, process parameter change rate, system sensitivity coefficient, response time constant and process state variable, and outputs the real-time compensation amount of the control instruction to ensure that the process parameter control can still be maintained within the target range in the presence of communication delay.

7. The industrial production process modular optimization method according to claim 6, characterized in that: The adaptive feedback optimization function refers to a mathematical function that dynamically adjusts the weights of each parameter in the gain matrix in real time according to the historical error accumulation value, current performance index, system feedback speed, time weighting factor and disturbance resistance coefficient, and outputs the optimized gain matrix parameter set.

8. The industrial production process modular optimization method according to claim 7, characterized in that: The communication delay time refers to the time required for instructions or data to be transmitted from the source module to the target module in the industrial control network, which is monitored by the time-sensitive network architecture; the process parameter change rate refers to the speed at which key process parameters change over time in the industrial production process, which is calculated from the process parameter sequence collected by the real-time data bus.

9. The industrial production process modular optimization method according to claim 8, characterized in that: The system sensitivity coefficient refers to the degree of response of process parameters to changes in control instructions, which is obtained by analyzing the process parameter response curve through edge computing nodes; the response time constant refers to the characteristic time required for the process system to respond to the control signal and reach a steady state, which is provided by the real-time requirement matrix.

10. The industrial production process modular optimization method according to claim 9, characterized in that: Process state variables refer to a set of key parameters that describe the current process state, which are collected from each process module through a low-latency communication network; the historical error accumulation value refers to the weighted accumulation of system control errors over a period of time, which is calculated by the real-time deviation fluctuation matrix.

Images