Order Management Method and System Based on MES

By constructing a model of time interference relationships between processes and equipment process mechanisms, the prediction and parameter compensation of virtual advance orders are realized, which solves the problem of time lag in quality closed-loop control in traditional MES systems. This enables quality prevention and control of hidden defects in the production of automotive parts, reducing scrap rate and hidden quality risks.

CN123117929APending Publication Date: 2026-11-17江西软件职业技术大学
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611508856.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-28
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Traditional MES systems in automotive parts manufacturing suffer from problems such as delayed quality closed-loop control, insufficient consideration of physical constraints between processes, and lack of multi-process cascade compensation, leading to the generation of non-conforming products and hidden quality defects, making it difficult to meet stringent quality and reliability requirements.

Method used

The order management method based on the MES system realizes the prediction and parameter compensation of virtual advance orders by constructing the time interference relationship between processes and the equipment process mechanism model. Combined with multiphysics simulation and closed-loop control, it ensures the time interference and quality prediction between processes, and adopts standard industrial communication protocols for instruction issuance and model updates.

Benefits of technology

It avoids the generation of defective products from the source, effectively controls hidden quality defects, reduces scrap rate and economic losses, has strong compatibility and low deployment cost, and is suitable for automotive parts production lines.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123117929A_ABST
    Figure CN123117929A_ABST
Patent Text Reader

Abstract

This invention discloses an order management method and system based on a Manufacturing Execution System (MES), belonging to the field of intelligent manufacturing technology. The invention constructs a digital model of order physical constraints, including inter-process time interference relationships and equipment process mechanism models. In the digital space, it creates virtual advance orders that correspond one-to-one with physical orders, setting a virtual time lead to advance the virtual orders ahead of the physical orders. When a virtual order reaches a critical process, measured data is acquired and substituted into the equipment process mechanism model to perform virtual processing simulation and calculate quality indicators. If the prediction does not meet the requirements, the compensated process parameters are solved in reverse and sent to the physical equipment in advance. After processing is completed, measured data is collected to update the model. This invention achieves predictive process closed-loop control at the MES level, effectively solving the time lag problem in quality closed-loop control.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing, and in particular to an order management method and system based on an MES system. Background Technology

[0002] Intelligent manufacturing is the core direction for the transformation and upgrading of the automotive parts industry. Manufacturing Execution System (MES), as the core hub connecting enterprise resource planning and industrial site control, is a key support for ensuring the quality and delivery efficiency of mass production of automotive parts. There are strict hard physical timing constraints between processes, and the requirements for product precision and reliability are extremely high. Even a small parameter deviation or a timeout in process connection can not only directly lead to the scrapping of workpieces, but may also cause safety hazards to the entire vehicle and huge recall losses.

[0003] Traditional MES (Manufacturing Execution System) in automotive parts production generally adopts a post-production inspection + passive adjustment model. Quality inspection is mostly carried out after a single process or batch is completed. By the time quality problems are discovered, defective products have already been produced, and the loss from scrapping a single batch of high-value precision parts can reach hundreds of thousands of yuan. At the same time, traditional scheduling only considers logical constraints such as equipment capacity and material supply, completely ignoring hard physical constraints such as the aging window after heat treatment and the curing window after surface treatment. As a result, defects such as insufficient latent fatigue strength often only appear during the vehicle's use stage, making traceability extremely difficult. Existing equipment-level PLC closed-loop control can only realize local parameter adjustments for single equipment, and the MES-level statistical process control module can only issue statistical anomaly alarms. Neither can predict quality risks in advance and proactively compensate for processes, making it difficult to meet the stringent quality and reliability requirements of the automotive parts industry. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing MES systems, such as delayed quality closed-loop control, insufficient consideration of physical constraints between processes, and lack of multi-process cascade compensation, and to provide an order management method and system based on MES.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: The order management method based on the MES system includes the following steps: S1: Receives production orders from the ERP system, parses the core information contained in the order, such as the product BOM, process route, quality standards, and delivery cycle, and constructs a digital model of the physical constraints of the order. This model contains two inseparable core components: first, the time interference relationship between processes, defining hard time window constraints that must be satisfied between adjacent key processes. These upper and lower limits of the time window serve as insurmountable hard boundaries for the timing of virtual advance orders; second, the equipment process mechanism model, encapsulating the quantitative mapping relationship between the input parameters and output quality indicators of each key process.

[0006] The time interference relationships between processes include, but are not limited to: after the heat treatment process is completed, subsequent machining processes must be performed within a specified time window ΔT_window; otherwise, the workpiece hardness will exceed the allowable range, leading to machining failure; after the surface coating process is completed, assembly must be carried out within the curing time window; otherwise, the coating adhesion will not meet the standards; after the welding process is completed, flaw detection must be performed within the cooling time window; otherwise, residual stress will cause crack propagation. The upper and lower limits of the time windows are determined by a large amount of process test data and stored in the process knowledge base.

[0007] The equipment process mechanism model is customized for different processes. For example, for the precision grinding process, the model is encapsulated with the following mapping relationship: ; .

[0008] In the formula, H is the measured hardness of the workpiece, f is the feed rate, n is the grinding wheel speed, w is the amplitude of the spindle vibration, ε is the form and position error, and Ra is the surface roughness. , This is a deviation term that can be calibrated online, used to compensate for systematic errors between the model and the actual production process; and All of these are mapping functions constructed based on physical mechanisms, which are implemented through analytical functions or differentiable neural networks.

[0009] S2: Perform multi-physics simulation rigid verification on the order execution paths that have completed the initial scheduling, verifying whether there are physical conflicts in the scheduling results from three dimensions: energy, logistics, and environment. Specifically, this includes verifying whether the rated power of the equipment meets the peak energy consumption requirements of the process, whether there is congestion in the logistics path, whether the ambient temperature and humidity meet the process requirements, and whether the equipment load exceeds the rated value. Only after the verification passes can the virtual advance order management process be entered; if the verification fails, return to the scheduling module and re-optimize the scheduling based on physical constraints.

[0010] S3: Create virtual advance orders in the digital space that correspond one-to-one with physical orders, assign independent virtual timelines to the virtual advance orders, and configure virtual time advance amount ΔT so that the virtual timeline of the virtual advance orders advances slightly faster than the physical timeline, and start the synchronous operation of virtual advance orders and physical orders.

[0011] The virtual time lead ΔT is determined according to the following formula to ensure that quality prediction and parameter compensation are completed before the physical order reaches the critical process: ; In the formula, The maximum computation time required to reverse-engineer the compensation parameters, Transmission time in industrial communication networks The response time for loading new parameters into the physical device controller. The default value range of ΔT is 30 minutes to 1 hour, and it can be dynamically configured according to the cycle length of different processes: for long cycle processes (such as heat treatment), ΔT can be increased appropriately; for short cycle processes (such as precision turning), ΔT can be decreased appropriately.

[0012] S4: When the virtual time of the virtual advance order reaches the start time of a certain key process, the latest measured input data corresponding to the process in the physical world is obtained through the data and physical perception layer. This includes the characteristics of the incoming workpiece (such as hardness, size, and surface condition), the real-time status of the equipment (such as spindle vibration, temperature, and tool wear), and environmental parameters (such as workshop temperature and humidity and air pressure). The measured data is then substituted into the equipment process mechanism model to perform virtual processing simulation and predict the quality indicators after the process is completed.

[0013] S5: Compare the predicted quality indicators with the quality standards specified in the order. If the predicted quality indicators meet the requirements, keep the original process parameters unchanged; if the predicted quality indicators do not meet the requirements, solve the compensated process parameters in reverse based on the same equipment process mechanism model to obtain the corrected process instructions.

[0014] The reverse solution preferentially employs a gradient descent iterative scheme, which offers both fast solution speed and high accuracy. ; In the formula, the loss function ; and These are the weighting coefficients for form and position error and surface roughness, respectively (which can be adjusted according to product quality requirements). For the allowable tolerance of form and position error, α represents the maximum allowable surface roughness, and α is the adaptive learning rate. When the equipment process mechanism model is not differentiable, the system automatically switches to a pre-built parameter compensation lookup table indexed by hardness and vibration for interpolation. The query response time is less than 1 millisecond, meeting the real-time requirements of industrial sites.

[0015] Furthermore, this method also includes a multi-process serial pre-simulation and cascade compensation step: when it is predicted that the i-th process needs parameter compensation, the impact of the compensation parameter on the input state of all subsequent processes is re-simulated, and a set of optimal compensation parameter sequences is generated by solving a constrained global optimization problem to ensure that the quality indicators of all processes meet the standards and do not violate the time interference relationship between processes.

[0016] S6: The revised process instructions are sent to the physical equipment controller via standard industrial communication protocols, and the instructions are verified and confirmed. When the physical order arrives at this process step, the physical equipment controller automatically executes the revised process parameters without manual intervention.

[0017] S7: After this process is completed, measured quality data is collected using a coordinate measuring machine, surface roughness meter, and online visual inspection equipment. The measured quality data is compared with the predicted quality data, and the deviation term of the equipment process mechanism model is updated online using the deviation data between the predicted and measured values. , By using the weight parameters of differentiable neural networks, the model can achieve self-learning and self-adaptation, effectively solving the model drift problem and ensuring the long-term accuracy of quality prediction.

[0018] Furthermore, the present invention also includes an order management system based on an MES system, the system specifically comprising: The data and physical sensing layer is used to collect real-time equipment status, actual material measurement data, and environmental parameters. The order physical constraint modeling layer is used to construct a digital model of order physical constraints that includes inter-process time interference relationships and equipment process mechanism models; The multiphysics simulation rigidity verification module is used to verify physical conflicts in the energy, logistics, and environmental dimensions of the order execution path; The virtual advance order execution and closed-loop control layer communicates with the aforementioned layers and includes: The virtual order instantiation and synchronization module is used to create virtual advance orders and set the virtual time lead time ΔT; The timing constraint management module is used to advance the virtual timeline according to the hard boundaries of the time interference relationship between processes; The virtual machining simulation engine is used to call the equipment process mechanism model to predict quality indicators; A process parameter inverse solver is used to solve for compensated process parameters when predicting deviations. The instruction issuing module is used to write compensation parameters into the physical device controller; The closed-loop feedback and online learning module is used to receive measured quality data and update the model.

[0019] Explanation of beneficial effects: 1. Prevent defective products from the source: By using a virtual advance order pre-simulation mechanism, quality control is transformed from post-event remediation to pre-event prevention, which completely solves the time lag problem of traditional MES quality closed-loop control. This can significantly reduce the scrap rate of core precision automotive parts and greatly reduce economic losses.

[0020] 2. Effectively prevent and control hidden quality defects: Incorporate hard physical timing constraints between processes into the entire order process management, avoid material performance deterioration caused by process connection delays from the scheduling stage, and eliminate hidden safety hazards such as insufficient fatigue strength that may be exposed in automotive parts during the vehicle use stage.

[0021] 3. High compatibility and low deployment cost: The system adopts standard industrial communication protocols and seamlessly integrates with the existing MES system and equipment of automotive parts production lines. It does not require large-scale hardware replacement and can be quickly integrated into applications to meet the large-scale deployment needs of mass production in the automotive industry. Attached Figure Description

[0022] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the specific steps of the order management method based on the MES system of the present invention. Figure 2 This is an architecture diagram of the order management system based on the MES system of this invention. Detailed Implementation

[0023] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0024] Example 1 This embodiment uses a mass production line for core precision automotive parts as an application scenario to provide a detailed description of the order management method and system based on the MES system described in this invention. The production line applicable to this embodiment includes multiple basic processing steps, key processing steps, and final inspection and assembly steps. Among these key steps, there are strict hard physical timing constraints, and product quality directly affects the operational performance and safety reliability of the entire vehicle.

[0025] like Figure 1 As shown, the order management method based on the MES system described in this embodiment specifically includes the following steps: S1: Receive production orders and construct a digital model of the physical constraints of the orders.

[0026] In this step, the system first receives automotive parts production orders from the Enterprise Resource Planning (ERP) system via a standard interface. It automatically parses the order's Bill of Materials (BOM), standard process routes, quality acceptance standards for each process, order delivery cycle, and batch information. Based on the parsed core order information, the system constructs a unique digital model of the order's physical constraints at the order physical constraint modeling layer. This model consists of two core parts: the time interference relationship between processes and the equipment process mechanism model. These two parts together define the physical constraint boundaries that must be followed during order execution.

[0027] The inter-process time interference relationship describes the insurmountable hard time window constraint between adjacent critical processes caused by changes in the physicochemical properties of materials. This constraint serves as the hard boundary for the subsequent virtual advance order timing. In this embodiment, the most representative inter-process time interference relationship is as follows: after the workpiece blank undergoes material property modification treatment in the preceding critical process, the subsequent precision machining process must be performed within a specified time window. This is because after the preceding critical process, the material structure inside the workpiece is in an unstable transitional state. As time progresses, natural property changes occur, leading to a gradual change in the workpiece's mechanical properties and a redistribution of internal stress. If subsequent machining is performed beyond the specified time window, machining accuracy cannot be guaranteed, and the durability of the finished parts will be significantly reduced. The upper and lower limits of this time window are determined by statistical analysis of a large amount of process test data and stored in the system's process knowledge base, and can be dynamically adjusted according to the different materials of the parts.

[0028] Equipment process mechanism models are used to encapsulate the quantitative mapping relationships between input process parameters, environmental parameters, equipment status parameters, and output quality indicators for each key process. This embodiment constructs the following form of equipment process mechanism model for the core key processes: ; .

[0029] The parameters in the formula are defined as follows: The measured mechanical properties of the workpiece are obtained by online testing equipment after the completion of the previous key process, reflecting the processing adaptability of the workpiece material; The core machining feed parameters are one of the core process parameters of key processes, which directly affect machining efficiency and surface quality. The core machining speed parameter is also a core process parameter, which, together with the feed parameter, determines the amount of material processed per unit time. The vibration amplitude of the core components of the processing equipment is collected in real time by vibration sensors installed on the core components of the equipment, reflecting the operating status of the equipment. Excessive vibration can cause defects on the processed surface. The form and position errors of key features of a workpiece are a key indicator for measuring the machining accuracy of parts and directly affect the assembly accuracy and operational stability of the parts. The surface roughness parameters of key surfaces of a workpiece are core indicators for measuring surface processing quality and directly affect the wear resistance and service life of parts. , For mapping functions built based on physical mechanisms, the following two implementation methods can be adopted: Method 1 (Analytical Function Implementation): , These are analytical functions derived from the physical mechanisms of cutting mechanics and tribology. For precision grinding processes, f1 and f2 can be expressed in the following general functional form: ; ; Among them, a1, b1, c1, d1, e1, a2, b2, c2, d2, and e2 are model parameters to be calibrated, which are fitted and calibrated using the least squares method or nonlinear regression method based on historical production data. The above function form is derived from the physical mechanism of grinding: hardness H directly affects the difficulty of material removal and the quality of the machined surface; feed rate f and grinding wheel speed n jointly determine the amount of material removed per unit time and the generation of grinding heat; and spindle vibration amplitude w has an exponential influence on the waviness and shape error of the machined surface. The physical meaning and value range of each model parameter are determined through a limited number of process experiments based on specific grinding conditions (including grinding wheel characteristics, coolant conditions, and machine tool rigidity).

[0030] Method 2 (implemented via differentiable neural network): , The network is a differentiable neural network pre-trained with historical production data. This differentiable neural network is a multi-layer feedforward neural network (BP neural network). Its input layer contains four nodes, corresponding to the measured workpiece hardness H, feed rate f, grinding wheel speed n, and spindle vibration amplitude w, respectively. The output layer contains two nodes, corresponding to the form and position error ε and surface roughness Ra, respectively. There are one to two hidden layers, each containing five to ten nodes. The activation function for the hidden layers is either a sigmoid function or a hyperbolic tangent function (Tanh), and the activation function for the output layers is a purelin function. The network is trained using measured data pairs collected during historical production as training samples. The mean squared error (MSE) between the model's predicted values ​​and the measured values ​​is used as the loss function. The back propagation (BP) algorithm combined with the Adam adaptive moment estimator is used for training. An early stopping strategy on the validation set is employed to prevent overfitting until the model's prediction accuracy on the validation set reaches a predetermined threshold. The aforementioned neural network structure and training method are conventional techniques in this field. Those skilled in the art can construct and train a prediction model that meets the accuracy requirements by following the instructions in the specification and combining it with production data under specific grinding conditions.

[0031] , The system bias term, which can be calibrated online, is used to compensate for the inherent systematic errors between model assumptions and actual production processes, as well as model drift caused by slow time-varying factors such as equipment wear and environmental changes.

[0032] The digital model of the physical constraints of the order, which is constructed in this step, will be transmitted to the multiphysics simulation rigid verification module and the virtual advance order execution and closed-loop control layer, serving as the basis for all subsequent processes.

[0033] S2: Perform multiphysics simulation rigidity verification on the order execution path that has completed the initial scheduling.

[0034] In this step, the system first calls the production scheduling module of the MES system to generate the initial execution path and scheduling plan for the order based on conventional logical constraints such as equipment capacity, personnel scheduling, and material supply. Subsequently, the initial scheduling results are input into the multiphysics simulation rigidity verification module to perform physical conflict verification on the scheduling results from three core dimensions: energy, logistics, and environment, to ensure that the scheduling plan is physically executable.

[0035] In terms of energy, the verification module will simulate the energy consumption curves of each process during production to verify whether the rated power of the equipment can meet the peak energy consumption requirements of the process, thus avoiding processing interruptions or quality defects due to insufficient power. In terms of logistics, the verification module will simulate the transfer process of workpieces between processes to verify whether the transportation capacity of automated logistics equipment can meet the scheduling requirements, thus avoiding process connection delays caused by logistics congestion. In terms of environment, the verification module will verify whether environmental parameters such as temperature, humidity, and cleanliness in the workshop can meet the process requirements of each key process. For example, precision machining processes require the temperature fluctuations in the workshop to be controlled within a small range; otherwise, thermal expansion and contraction will affect the machining accuracy.

[0036] If the verification passes, the system will generate a verification report and transmit the initial scheduling plan and the digital model of the order's physical constraints to the virtual advance order execution and closed-loop control layer to initiate the virtual advance order management process. If the verification fails, the system will generate a detailed physical conflict analysis report and return the report to the production scheduling module to guide the scheduling module to re-optimize the scheduling based on physical constraints until the verification passes.

[0037] S3: Create a virtual advance order and set the virtual time advance amount.

[0038] This step is performed by the virtual order instantiation and synchronization module in the virtual advance order execution and closed-loop control layer. The engine creates virtual advance orders in the digital space that correspond one-to-one with physical orders, allocates independent virtual timelines to the virtual advance orders, and configures virtual time advance amounts. This allows the virtual timeline of virtual advance orders to advance slightly faster than the physical timeline, thus enabling virtual orders to run synchronously with physical orders ahead of them.

[0039] Virtual time advance The value of must satisfy the following constraints: ; The physical meanings of each parameter in the formula are defined as follows: The maximum computation time for reverse engineering of process parameters, including the total time for model calling, iterative calculation, and result output; The transmission time of the industrial communication network includes the bidirectional transmission time from the MES system to the physical device controller, and the time for the physical device controller to return an acknowledgment signal. Response time of the physical equipment controller to load new process parameters and complete preparation work.

[0040] The derivation logic of this formula lies in the fact that the virtual time lead must reserve a sufficient time window to ensure that the system can complete the entire process of quality prediction, parameter inverse solving, instruction issuance, and equipment preparation before the physical order reaches the start time of the critical process, thereby achieving predictive process compensation. In this embodiment, The value range is set to 30 minutes to 1 hour. The reason for this range is: if If the size is too small, it may not allow enough time to complete the above process, resulting in compensation instructions not being issued in a timely manner; if If the value is too large, the deviation between the measured data used in the virtual order and the actual processing status of the physical order will increase, reducing the accuracy of quality prediction. Considering the computing power, communication speed, and equipment response characteristics of this production line, this embodiment selects an intermediate value as... The optimal value is one that ensures timely compensation while maximizing the timeliness of the predicted data. For long-cycle critical processes, the value can be appropriately increased. Up to the upper limit of the range; for short-cycle critical processes, the range can be appropriately reduced. Up to the lower limit of the range.

[0041] Virtual advance orders and physical orders will proceed synchronously according to their respective timelines, with virtual orders always remaining in place. The system uses a unified timestamp service to achieve precise synchronization between the two timelines, ensuring that the timing of virtual orders strictly corresponds to the actual execution status of physical orders.

[0042] S4: Virtual processing simulation and quality index prediction.

[0043] When the virtual time of a virtual advance order reaches the start time of a critical process, the timing constraint management module will trigger the virtual processing pre-simulation engine. The pre-simulation engine first sends a data request to the data and physical perception layer to obtain the latest measured input data corresponding to that process in the physical world, including workpiece material characteristic data (such as mechanical properties, dimensional deviations, and surface conditions), real-time equipment status data (such as core component vibration, temperature, and tool wear), and environmental parameter data (such as workshop temperature, humidity, and air pressure).

[0044] The data and physical sensing layer performs preprocessing operations such as filtering, denoising, and standardization on the collected raw data before transmitting it to the virtual machining simulation engine via industrial Ethernet. The engine then inputs the preprocessed measured data into a pre-built equipment process mechanism model to perform virtual machining simulation. The simulation relies on lightweight model optimization and parallel computing technology, making its simulation speed tens of times faster than physical machining. It can simulate the entire machining process in a very short time and predict the quality indicators (such as form and position errors in this embodiment) after the machining process is completed. With surface roughness ).

[0045] S5: Inverse solution of process parameters and multi-process cascade compensation.

[0046] After the virtual machining simulation is completed, the system compares the predicted quality indicators with the quality standards specified in the order. If the predicted quality indicators meet the requirements, the original process parameters remain unchanged, and the system continues to advance the virtual advance order to the next process. If the predicted quality indicators do not meet the requirements, the process parameter inverse solver is triggered, and the compensating process parameters that can make the quality indicators meet the standards are solved in reverse based on the same equipment process mechanism model.

[0047] In this embodiment, the inverse solution preferentially adopts the gradient descent iterative algorithm, and its iterative format is as follows: ; The physical meanings of each parameter in the formula are defined as follows: , : No. The core machining feed parameters and rotation speed parameters during the next iteration; , : No. The core machining feed parameters and rotation speed parameters during the next iteration; Adaptive learning rate, used to control the step size of iterations; Loss function Gradient vectors relating to process parameters.

[0048] loss function It is used to measure the degree of deviation between predicted quality indicators and quality standards, and is defined as follows: ; In the formula: The permissible tolerances for form and position errors are specified in the order quality standards; The maximum permissible value for surface roughness is specified in the order quality standards; , : These are the weighting coefficients for form and position error and surface roughness, respectively, used to reflect the importance of different quality indicators.

[0049] In this embodiment, the weighting coefficient and The values ​​are all set to a range of 0.3 to 0.7. The reason for this range is that different types of automotive parts have different requirements for form and position errors and surface roughness. For example, high-speed moving parts have higher requirements for surface roughness, while heavy-duty load-bearing parts have higher requirements for form and position errors. Therefore, it is necessary to set an adjustable weight range to adapt to the needs of different products. For the general-purpose automotive core precision parts in this embodiment, the influence of form and position errors and surface roughness on product performance is comparable. Therefore, equal weights are selected as the optimal values.

[0050] Adaptive learning rate The value range is set to 0.01 to 0.1. The reason for this range is that if the learning rate is too small, the iterative convergence speed will be too slow, failing to meet real-time requirements; if the learning rate is too large, the iteration process is prone to oscillations, failing to converge to the optimal solution. Considering both convergence speed and stability, this embodiment selects the intermediate value as... The optimal value.

[0051] When the equipment process mechanism model adopts a purely analytical function form and is not differentiable, the system will automatically switch to a pre-built parameter compensation lookup table for interpolation. This lookup table uses the workpiece mechanical property parameters... Vibration amplitude of core components of equipment For indexing, the optimal compensation process parameters under different operating conditions are pre-stored, and fast query is achieved through hash indexing technology. The query response time is less than 1 millisecond, which fully meets the real-time requirements of industrial sites.

[0052] Furthermore, when it is predicted that parameter compensation is needed for a certain process, the system will automatically initiate a multi-process sequential pre-simulation and cascaded compensation process. This process will re-simulate the impact of the compensation parameter on the input states of all subsequent processes. For example, adjusting the process parameters of a previous critical process will change the machining allowance of the workpiece, thereby affecting the process parameters of subsequent processes. The system generates a set of optimal compensation parameter sequences by solving a constrained global optimization problem (in this embodiment, the Sequential Quadratic Programming (SQP) algorithm) to ensure that the quality indicators of all processes meet the standards and do not violate the hard constraints of time interference relationships between processes.

[0053] S6: Issuance and execution of revised process instructions.

[0054] After obtaining the optimal compensation process parameters, the system generates corrected process instructions and transmits them to the physical device controller via the instruction delivery module. This embodiment uses a unified architecture standard industrial communication protocol for instruction transmission. This protocol offers advantages such as high security, strong interoperability, and reliable data transmission, and is widely used in the automotive manufacturing industry.

[0055] Before sending an instruction, the instruction issuing module performs a CRC check on the instruction content to ensure that the instruction is not tampered with during transmission. After receiving the instruction, the physical equipment controller returns an acknowledgment signal to the system. If the system does not receive an acknowledgment signal within a specified time, it will automatically resend the instruction to ensure reliable delivery. After loading the revised process parameters and completing the preparation work, the physical equipment controller will enter a waiting state. When the physical order reaches the start time of that process, it will automatically execute the revised process parameters without manual intervention.

[0056] S7: Feedback of measured quality data and online model updates.

[0057] After this process is completed, the data and physical sensing layer will collect the measured quality data of the workpiece through high-precision online inspection equipment and transmit the measured data to the closed-loop feedback and online learning module. The module will compare the measured quality data with the previously predicted quality data and calculate the deviation between the two.

[0058] Based on the calculated deviations, the module uses the incremental stochastic gradient descent (SGD) algorithm to update the parameters of the equipment process mechanism model online, including the system deviation term. , And the weight parameters of the differentiable neural network. The advantage of incremental learning algorithms is that they do not require retraining the entire model; only minor adjustments to the model parameters are needed based on the latest measured data. Therefore, the computation speed is fast, enabling real-time online updates of the model. Through this closed-loop feedback mechanism, the system can continuously correct the deviation between the model and the actual production process, effectively solving the model drift problem caused by factors such as equipment wear and tear, environmental changes, and tool replacement, and ensuring the long-term accuracy of quality prediction.

[0059] The system architecture described in this embodiment is as follows: Figure 2 As shown, it includes a data and physical perception layer, an order physical constraint modeling layer, a multiphysics simulation rigid verification module, and a virtual advance order execution and closed-loop control layer. These layers communicate through standard interfaces, and data flows and transforms between them in an orderly manner according to the steps described above, collectively achieving predictive process closed-loop control at the MES level.

[0060] The functions and roles of each core module are as follows: Data and Physical Sensing Layer: Serving as the interface between the system and the physical production site, this layer is primarily responsible for collecting all the raw data required for method execution, including real-time equipment operating status, workpiece incoming material characteristics, environmental parameters, and measured quality data after processing. This layer supports the integration of various industrial sensors such as vibration, temperature, and pressure sensors, as well as online inspection equipment like coordinate measuring machines and surface roughness testers. It is also compatible with mainstream standard industrial communication protocols such as OPCUA and Modbus TCP, allowing seamless integration with existing physical equipment controllers. After basic preprocessing such as filtering, noise reduction, and standardization, the collected raw data is transmitted in real-time to the upper-layer modules, providing data support for order physical constraint model construction, virtual simulation, and online model updates.

[0061] Order Physical Constraint Modeling Layer: This is the digital foundation module of the entire system. This module receives production orders from the ERP system, parses core information such as the product BOM, process route, and quality standards, and retrieves the corresponding product's inter-process time window parameters and equipment process mechanism model templates from the process knowledge base to construct a digital model of order physical constraints that corresponds one-to-one with the physical order. The model also encapsulates insurmountable hard time constraints between processes, as well as the quantitative mapping relationship between input parameters and output quality indicators. The generated model is synchronously transmitted to the multiphysics simulation verification module and the virtual advance order control layer, serving as the basis for the execution of all subsequent processes.

[0062] The Multiphysics Simulation Rigid Verification Module is responsible for verifying the physical feasibility of the initial scheduling plan generated by the MES system. This module receives the initial scheduling results and a digital model of the order's physical constraints, simulating the order execution process from three core dimensions: energy, logistics, and environment. It verifies for physical conflicts such as insufficient equipment power, logistics congestion, and unmet environmental parameters. If the verification passes, the order proceeds to the virtual advance order management process; if the verification fails, a conflict analysis report is generated and returned to the MES scheduling module to guide it in re-optimizing the schedule based on physical constraints.

[0063] Virtual advance order execution and closed-loop control layer: This layer coordinates the entire lifecycle management and predictive process control of virtual advance orders. Its sub-modules include the following functions: Virtual order instantiation and synchronization module: Creates virtual advance orders in the digital space that correspond one-to-one with physical orders, assigns them independent virtual timelines, calculates and configures the virtual time lead ΔT according to the formula, and achieves precise synchronization between the virtual timeline and the physical timeline through a unified timestamp service, ensuring that the virtual orders always run ahead of the physical orders.

[0064] The timing constraint management module strictly follows the hard boundaries of time interference between processes defined in the order physical constraint model, advances the virtual timeline of virtual advance orders, monitors timing compliance in real time, and immediately triggers an alarm and initiates the adjustment process if a timing conflict occurs.

[0065] Virtual processing simulation engine: When the virtual time reaches the start time of the key process, it calls up the latest measured data from the data and physical perception layer, substitutes it into the equipment process mechanism model to perform virtual processing simulation, and quickly predicts the quality indicators of the process.

[0066] Inverse solver for process parameters: When the predicted quality is substandard, the gradient descent algorithm is used first to solve for the compensation process parameters in reverse. If the model is not differentiable, it automatically switches to a pre-built parameter compensation lookup table for interpolation. It also supports multi-process cascade compensation, generating a sequence of compensation parameters that meets the quality and timing constraints of all processes through global optimization.

[0067] Command issuance module: Converts the obtained modified process commands into a format recognizable by the physical equipment controller, issues them to the corresponding equipment through standard industrial communication protocols, and completes command verification and confirmation to ensure reliable transmission and execution of commands.

[0068] Closed-loop feedback and online learning module: Receives measured quality data after processing, compares it with predicted values ​​to calculate deviation, and uses an incremental learning algorithm to update the deviation terms and network weights of the equipment process mechanism model online, solving the model drift problem and continuously improving the accuracy of quality prediction.

[0069] Example 2

[0070] This embodiment uses a batch processing production line for drive shaft spline shafts of an automotive parts manufacturing company as a practical application scenario. Based on the company's actual production measurement data, the application process and actual effects of the method of this invention are verified through an instance. This production line produces 200 heavy-duty truck drive shaft spline shafts per batch. The core critical processes are, in sequence, controlled atmosphere carburizing heat treatment, precision cylindrical grinding, and magnetic particle inspection. Among them, after the heat treatment is completed, the grinding process must be started within a 120-minute hard timing window; otherwise, the decrease in surface hardness of the workpiece will lead to substandard form and position accuracy and surface quality.

[0071] S1: Receive production orders and construct a digital model of the physical constraints of the orders.

[0072] Received a production order for 200 drive shafts and splined shafts from the ERP system, and analyzed the product BOM, process route, and quality acceptance standards: allowable tolerances for form and position errors. =0.008mm, maximum permissible surface roughness =0.8μm. Construct the corresponding digital model of the physical constraints of the order: Inter-process time interference: Hard timing window from heat treatment to grinding The time should be set to 10min~120min (grinding can begin 10min after heat treatment, but should not exceed 120min at the latest). Exceeding the upper limit will cause the workpiece hardness to decrease, leading to processing failure. Equipment process mechanism model: For the MK1320 CNC cylindrical grinding machine on the production line, a differentiable neural network mapping model was trained based on 12 months of historical production data of the enterprise. The input items are the measured hardness of the workpiece H, feed speed f, grinding wheel speed n, and spindle vibration amplitude w. The output items are the form and position error ε and surface roughness Ra. The initial prediction average relative error of the model is 4.2%.

[0073] S2: Rigidity verification of multiphysics simulation.

[0074] The initial scheduling results generated by MES are input into the multiphysics simulation rigidity verification module to complete rigidity verification from three dimensions: energy dimension verifies that the peak power of the heat treatment furnace is 380kW, which meets the equipment's rated power requirement of 400kW; logistics dimension verifies that the AGV single-batch transfer time is 15 minutes with no path congestion; environmental dimension verifies that the temperature fluctuation in the grinding workshop is ±0.8℃, which meets the requirements of precision machining process. Upon successful verification, the virtual advance order management process begins.

[0075] S3: Create a virtual advance order and set the virtual time advance amount.

[0076] Create virtual advance orders in the digital space that correspond one-to-one with physical orders, and configure independent virtual timelines. Based on actual field measurement parameters, calculate the maximum computation time using reverse engineering. =1.2min, Industrial Ethernet communication transmission time =0.3min, device controller parameter loading response time =0.5min, according to the formula The calculated minimum lead time is 2.0 min. Considering the 8-minute cycle time for grinding a single workpiece, the final configuration is ΔT=30 min, ensuring timely compensation while maintaining the timeliness of the measured data.

[0077] S4: Virtual processing simulation and quality index prediction.

[0078] When the virtual time reaches the start time of the grinding process, the latest measured data from the physical site are collected: average workpiece hardness H=62HRC, initial feed rate f=0.25mm / r, grinding wheel speed n=1800r / min, and spindle vibration amplitude w=0.12μm. The data is then substituted into the equipment's process mechanism model for simulation, which takes 0.8 minutes. The predicted results are: average form and position error ε=0.0092mm, average surface roughness Ra=0.91μm for this batch of workpieces; both indicators exceed the quality standards.

[0079] S5: Inverse solution of process parameters and multi-process cascade compensation.

[0080] Trigger the process parameter inverse solver to solve the compensation parameters using a gradient descent iterative algorithm. Set the form and position error weights. =0.6, surface roughness weight With a learning rate of α=0.05 and an adaptive learning rate of 0.4, the optimal compensation process parameters were obtained after 12 iterations: feed rate f=0.18mm / r and grinding wheel speed n=2100r / min. A multi-process cascaded simulation was simultaneously initiated to verify that the grinding time per workpiece increased by 2 minutes after parameter adjustment, but the entire batch still met the 120-minute time window constraint, and there was no adverse impact on subsequent flaw detection processes.

[0081] S6: Issuance and execution of revised process instructions.

[0082] The compensated process parameters are sent to the grinding machine PLC controller via the OPCUA standard industrial communication protocol. After CRC verification and equipment confirmation, the command is reliably delivered. When the physical order arrives at the grinding process, the equipment automatically loads the compensated parameters and performs the machining, requiring no manual intervention throughout the entire process.

[0083] S7: Feedback of measured quality data and online model updates.

[0084] After the processing of this batch of 200 workpieces was completed, online inspection was performed using a coordinate measuring machine and a surface roughness tester. The measured quality data were: average form and position error ε = 0.0068 mm, average surface roughness Ra = 0.72 μm, all meeting the quality acceptance standards. The measured data was fed back to the closed-loop feedback module, and the incremental stochastic gradient descent algorithm was used to update the model bias term and network weights online. After the update, the average relative error of the model prediction was reduced to 3.7%.

[0085] Ten consecutive batches totaling 2000 products from this production line were selected, and the core quality and efficiency indicators of the traditional MES post-processing control model and the model of this invention were statistically analyzed. The comparison results are shown in the table below: Indicator Item Traditional MES control mode Method mode of the present invention Relative improvement Process scrap rate 1.85% 0.35% 81.1% Roundness deviation rate 2.10% 0.40% 81.0% Surface roughness deviation rate 1.65% 0.30% 81.8% Process connection timeout rate 3.20% 0% 100% Average quality loss per batch (yuan) 7280 1360 81.3% Model prediction mean relative error - 3.7% - The test data shows that the method of the present invention, through virtual pre-simulation and pre-process compensation, transforms quality control from post-remediation to pre-prevention, which greatly reduces the scrap rate and quality loss of the process; at the same time, through hard time constraint control between processes, it completely eliminates the hidden quality defects caused by the timeout of process connection, verifying the actual beneficial effect of the present invention.

[0086] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An order management method based on an MES system, characterized in that, Includes the following steps: S1: Receive a production order and construct a digital model of the order's physical constraints. The model includes the time interference relationship between processes and the equipment process mechanism model. The time interference relationship between processes defines the hard time window constraints that must be satisfied between adjacent key processes. The upper and lower limits of the time window constraints serve as the insurmountable hard boundaries for the timing advancement of virtual advance orders. S2: Perform multi-physics simulation rigidity verification on the order execution path that has completed the initial scheduling. Only after the verification is passed can the virtual advance order management process be entered. S3: Create virtual advance orders in the digital space that correspond one-to-one with physical orders, set a virtual time advance amount ΔT, so that the virtual time axis of the virtual advance orders advances slightly faster than the physical time axis, and start the synchronous operation of virtual advance orders and physical orders; S4: When the virtual time of the virtual advance order reaches the start time of the key process, obtain the latest measured input data corresponding to the process in the physical world, substitute it into the equipment process mechanism model to perform virtual processing simulation, and calculate the quality index of the process. S5: If the quality indicators do not meet the requirements, the compensated process parameters are solved in reverse based on the same equipment process mechanism model to obtain the corrected process instructions; S6: The revised process instructions are sent to the physical equipment controller via the standard industrial communication protocol. When the physical order arrives at the process, the equipment controller automatically executes the revised process parameters. S7: After processing is completed, the measured quality data is collected and fed back to update the process mechanism model of the equipment online.

2. The order management method based on the MES system according to claim 1, characterized in that, The time interference relationship between processes mentioned in step S1 includes: after the heat treatment process is completed, it must be within a specified time window. The subsequent machining process is executed within the specified time window; otherwise, the workpiece hardness will exceed the allowable range, resulting in machining failure. The upper and lower limits of the time window serve as hard constraints for the advancement of virtual advance orders between adjacent processes.

3. The order management method based on the MES system according to claim 1, characterized in that, The virtual time lead ΔT mentioned in step S3 is determined according to the following formula: ; in The maximum computation time required to reverse-engineer the compensation parameters, For communication transmission time, The response time for loading parameters to the physical device controller; the value of ΔT ranges from 30 minutes to 1 hour and can be dynamically configured according to the process cycle.

4. The order management method based on the MES system according to claim 1, characterized in that, The equipment process mechanism model described in step S1 encapsulates the following mapping relationship for the precision grinding process: ; ; In the formula Here, denoted by , H is the measured hardness of the workpiece after machining, f is the feed rate, n is the grinding wheel speed, w is the spindle vibration amplitude, and Ra is the surface roughness. and All of these are mapping functions constructed based on physical mechanisms, which are implemented through analytical functions or differentiable neural networks; , This is a deviation term that can be calibrated online.

5. The order management method based on the MES system according to claim 1, characterized in that, The reverse solution described in step S5 uses a gradient descent iterative scheme: ; loss function in the formula ; in, For the first The feed rate at the nth iteration; is the feed rate at the th iteration. Grinding wheel speed at the next iteration; For the first Feed rate during the next iteration; For the first Grinding wheel speed at the next iteration; Adaptive learning rate; loss function Gradient vector of process parameters; This is the weighting coefficient for the form and position error; This is the weighting coefficient for surface roughness; Allowable tolerances for form and position errors; The maximum allowable value for surface roughness is given. When the model is not differentiable, the interpolation solution is obtained by switching to a pre-built parameter compensation lookup table indexed by hardness and vibration. The query response time is less than 1 millisecond. The lookup table is pre-built using process test data.

6. The order management method based on the MES system according to claim 1, characterized in that, It also includes multi-process serial simulation and cascade compensation steps: when it is predicted that the i-th process needs parameter compensation, the impact of the compensation on the input state of all subsequent processes is re-simulated. A set of compensation parameter sequences is generated by solving the global optimization problem to ensure that the quality of all processes meets the standards and does not violate the time interference relationship between processes.

7. An order management system based on an MES system, characterized in that, include: The data and physical sensing layer is used to collect real-time equipment status, actual material measurement data, and environmental parameters. The order physical constraint modeling layer is used to construct a digital model of order physical constraints that includes inter-process time interference relationships and equipment process mechanism models; The multiphysics simulation rigidity verification module is used to verify physical conflicts in the energy, logistics, and environmental dimensions of the order execution path; The virtual advance order execution and closed-loop control layer communicates with the aforementioned layers and includes: The virtual order instantiation and synchronization module is used to create virtual advance orders and set the virtual time lead time ΔT; The timing constraint management module is used to advance the virtual timeline according to the hard boundaries of the time interference relationship between processes; The virtual processing simulation module is used to call the equipment process mechanism model to predict quality indicators; A process parameter inverse solver is used to solve for compensated process parameters when predicting deviations. The instruction issuing module is used to write compensation parameters into the physical device controller; The closed-loop feedback and online learning module is used to receive measured quality data and update the model.