CN-121212856-B - Digital MES management system based on intelligent manufacturing

Abstract

The invention relates to the technical field of intelligent manufacturing, in particular to a digital MES management system based on intelligent manufacturing, which is used for solving the problems that the prior art cannot accurately integrate symbol reasoning and multi-agent reinforcement learning, cannot combine a self-adaptive multi-objective evolutionary algorithm with a three-level cooperative architecture of equipment, workshops and factories, cannot dynamically optimize resource scheduling under the condition of meeting the constraint of timeliness, efficiency and energy consumption, and reduces the intelligence, robustness and expandability of scheduling; according to the invention, symbol reasoning and multi-agent reinforcement learning are fused through the self-evolution intelligent scheduling module, the interpretability of rules and the self-adaptability of learning are considered, the self-adaptive multi-objective evolutionary algorithm and three-level cooperative architecture of equipment, workshops and factories are combined, the resource scheduling is dynamically optimized under the condition that the constraint of timeliness, efficiency and energy consumption is met, and the strategy closed-loop iteration is realized through executing feedback, so that the intelligence, robustness and expandability of the scheduling are remarkably improved.

Inventors

• LI XIAOCHUN

Assignees

• 杭州友成科技有限公司

Dates

Publication Date

20260512

Application Date

20251106

Claims (6)

1. 1. A digital MES management system based on intelligent manufacturing integrated in an intelligent manufacturing management platform, comprising: The data acquisition processing module is used for acquiring production operation parameters in the manufacturing execution process, sequentially carrying out signal filtering, missing value processing, abnormal value primary screening, data standardization and data compression aggregation on the acquired production operation parameters at the edge side, and automatically uploading the data to the intelligent manufacturing management platform after structured encapsulation; The self-evolution intelligent scheduling module adopts a multi-agent reinforcement learning and symbol reasoning engine, models a production unit as an agent and combines rule knowledge to generate scheduling decisions, integrates a self-adaptive multi-objective evolution optimization algorithm and a cross-level architecture, and coordinates scheduling tasks of an equipment layer, an interlayer and a factory layer according to production operation parameters in a manufacturing execution process; The process of combining multi-agent reinforcement learning with a symbolic reasoning engine and modeling a production unit as an agent by the self-evolution intelligent scheduling module and combining rule knowledge to generate scheduling decisions comprises the following steps: Acquiring the state of a current production unit, checking rules one by a symbol reasoning engine according to a preset sequence, wherein each rule comprises conditions and operations, when the conditions of a certain rule are completely matched with the current state, the symbol reasoning engine takes the operations contained in the rule as scheduling decisions, skips the processes except the subsequent agent decisions and table updating, directly executes scheduling decision output, and activates an agent corresponding to the current production unit if the symbol reasoning engine does not find any match after checking all the rules; the intelligent agent selects an operation from available operations according to the current state, takes the operation as a scheduling decision, executes the determined scheduling decision, starts equipment, enters the next state, acquires a feedback value, and sends the determined scheduling decision to an execution unit; the self-evolution intelligent scheduling module integrates a self-adaptive multi-objective evolution optimization algorithm and a cross-hierarchy architecture, and the process of coordinating scheduling tasks of a device layer, an interlayer and a factory layer according to production operation parameters in the manufacturing execution process comprises the following steps: The method comprises the steps that a factory layer reads a work order to be scheduled, the factory layer determines the execution sequence of the work order in the factory range, the factory layer sends work order control instructions to a workshop layer of a workshop to which the work order belongs, the workshop layer extracts an unassigned task from the work order, the workshop layer reads available states of all devices in the workshop, the workshop layer selects one device from available devices to execute the task according to a preset device grading rule, and the workshop layer sends task scheduling instructions to a device layer of the selected device; the equipment layer starts equipment to execute tasks, the equipment layer reads the running state in the task execution process, the equipment layer sends the task execution state to the workshop layer, the workshop layer calculates the task completion time, the equipment use efficiency and the energy consumption, the workshop layer sends a task completion report consisting of three indexes to the factory layer, and the factory layer compares the task completion time with the planned end time, the equipment use efficiency and the efficiency threshold value and the upper limit of the energy consumption; If the task completion time exceeds the planned end time, or the equipment use efficiency is lower than the efficiency threshold, or the energy consumption exceeds the energy consumption upper limit, the factory layer generates a parameter adjustment instruction, the factory layer sends the parameter adjustment instruction to the workshop layer, the workshop layer updates the equipment scoring rule, the workshop layer checks whether an unallocated task exists in a work order, if so, the next unallocated task is extracted, the equipment selection is continuously executed until the report sending process is finished, and if not, the scheduling coordination of the work order is finished; The intelligent quality traceability module is used for fusing production operation parameters, carrying out quality defect identification and diagnosis through deep learning and natural language processing, constructing a quality causal model by combining a causal graph neural network, recording traceability data by using a blockchain, and generating a quality intervention strategy based on digital twin; And the intelligent energy-saving management module monitors equipment energy consumption based on production operation parameters, adjusts equipment operation modes, cooperatively schedules multiple energy sources and predicts energy requirements in a production stage.
2. 2. The system of claim 1, wherein the intelligent quality traceability module fuses production operating parameters and performs quality defect identification and diagnosis by deep learning and natural language processing comprises: Reading current production operation parameters, reading an operation text corresponding to the current production operation parameters, calling a natural language processing model, processing the operation text, fusing the processed operation text with the production operation parameters, inputting the fused operation text into a pre-trained deep neural network, outputting the probability of quality defects of the current product by the network, judging that the current product has the defects if the probability is larger than a preset threshold, and judging that the current product has the defects if the probability is not smaller than the preset threshold; And calling a deep learning model, judging whether a quality defect exists or not based on the fusion result, reading a defect identification result output by the deep learning model, if the defect identification result is the defect, starting a diagnosis flow by the module, otherwise, skipping diagnosis, calling the diagnosis model, generating a defect reason based on the fusion result, outputting the defect identification result, and if the defect exists, outputting the defect reason.
3. 3. The system of claim 2, wherein the intelligent quality traceback module is configured to construct a quality causal model in conjunction with a causal graph neural network, comprising: Acquiring process condition data, acquiring intervention setting data, acquiring quality response data, calling a causal graph neural network, generating a causal structure based on the process condition data, the intervention setting data and the quality response data, identifying the process condition and the quality response with causal relation in the causal graph neural network, calculating causal effect intensity for each pair of the process condition and the quality response with causal relation, writing the calculated causal effect intensity into the causal structure, and outputting a quality causal model comprising the causal structure and the causal effect intensity.
4. 4. A digital MES management system based on intelligent manufacturing according to claim 3, wherein the process of the intelligent quality traceback module using blockchain record traceback data and generating quality intervention policies based on digital twinning comprises: Acquiring a piece of tracing data, reading the identifier of the latest block in the blockchain, combining the tracing data with the identifier of the latest block to generate new block content, generating a new block identifier according to the new block content, and writing the new block content and the new block identifier into the blockchain; The method comprises the steps of obtaining a current production state, inputting the current production state into a digital twin system, executing future production evolution deduction in the digital twin system, reading a future production evolution result output by the digital twin system, generating a quality intervention action sequence according to the future production evolution result, and outputting the quality intervention action sequence.
5. 5. The intelligent manufacturing-based digital MES management system according to claim 1, wherein the process of the intelligent energy saving management module monitoring equipment energy consumption and adjusting equipment operation modes based on production operation parameters comprises: Acquiring a current load value of the equipment, acquiring a power value corresponding to the current load value of the equipment from power characteristic configuration, acquiring a current time stamp, writing the power value and the current time stamp into energy consumption data storage as an energy consumption record, reading the last accumulated total energy consumption from the energy consumption data storage, reading the time stamp of the last energy consumption record from the energy consumption data storage, and calculating new energy consumption according to the time interval between the time stamp of the last energy consumption record and the current time stamp and combining the current power value; Accumulating the new energy consumption to the latest accumulated energy consumption total to obtain updated energy consumption total, acquiring an operation mode corresponding to the current load value of the equipment from the operation mode configuration, generating an equipment operation mode switching instruction, and sending the equipment operation mode switching instruction to the equipment.
6. 6. The intelligent manufacturing-based digital MES management system according to claim 5, wherein the process of performing multi-energy collaborative scheduling and predicting energy demand in a production phase by the intelligent energy saving management module includes: acquiring a production load value of a future period, acquiring an equipment running mode number of the future period, calling an energy demand prediction table, inputting the production load value and the equipment running mode number, outputting energy demand values at all times in the future, acquiring a total power demand value of a current dispatching cycle, acquiring an upper limit of output power of a first energy supply unit, acquiring a unit power dispatching cost of the first energy supply unit, acquiring an upper limit of output power of a second energy supply unit, and acquiring a unit power dispatching cost of the second energy supply unit; And calling a power distribution rule, inputting a total power requirement value, an upper limit of the output power of each energy supply unit and a unit power scheduling cost, generating an output power instruction of each energy supply unit, sending the output power instruction to the first energy supply unit, and sending the output power instruction to the second energy supply unit.

Description

Digital MES management system based on intelligent manufacturing Technical Field The invention relates to the technical field of intelligent manufacturing, in particular to a digital MES management system based on intelligent manufacturing. Background Under the background of the acceleration of global manufacturing industry towards intelligent manufacturing, the existing MES system is difficult to meet the requirements of production management on automation, digitization, networking and intelligent deep fusion, and is low in integration level with ERP and a bottom control system, so that planning and execution are disjointed, equipment data acquisition is difficult, the production is difficult to deal with complex dynamic constraint due to experience rules, quality control is focused on post-inspection, advanced prevention and in-process control are not available, data acquisition is incomplete, inaccurate and untimely, analysis capability is weak, effective support decision cannot be made, system function solidification and expansibility are poor, and the system is difficult to adapt to the individual requirements of enterprises. The invention discloses an enterprise-based intelligent manufacturing digital integrated analysis system, which is used for carrying out coordination control on the work of enterprise operation management by establishing an ERP system in an enterprise operation layer and carrying out production collaborative management by establishing an MES system in a production management layer, and can safely realize data intercommunication, reduce the data acquisition difficulty and clearly reflect the execution condition of an enterprise production plan, the purchase condition of materials, the operation condition of equipment, the quality problem of products and the feedback of after-sales problems; However, the above-mentioned reference patent discloses that the enterprise full-link data is cooperatively developed through ERP and MES to realize safe intercommunication and efficient collection, comprehensively reflect the production, materials, equipment, quality and after-sales conditions, support transparent management, scientific decision, automatic execution and efficient operation, promote the core competitiveness of enterprises, but cannot precisely integrate symbol reasoning and multi-agent reinforcement learning, cannot combine an adaptive multi-objective evolutionary algorithm with a three-level cooperative architecture of equipment, workshops and factories, cannot dynamically optimize resource scheduling under the constraint of aging, efficiency and energy consumption, reduce the intelligence, robustness and expandability of scheduling, cannot realize high-precision defect identification and automatic diagnosis, cannot combine a causal graph neural network to construct an interpretable causal model, cannot utilize digital twin deduction production evolution and dynamically generate intervention strategies, and cannot form an 'identification, diagnosis, attribution and intervention' closed loop, thereby reducing the intelligent, traceability and active prevention capability of quality control. To this end, we propose a digital MES management system based on intelligent manufacturing, aiming at the above problems. Disclosure of Invention The invention aims to provide a digital MES management system based on intelligent manufacturing, which solves the problems that the prior art cannot accurately integrate symbol reasoning and multi-agent reinforcement learning, cannot combine a self-adaptive multi-target evolutionary algorithm with three-level cooperative architecture of equipment, workshops and factories, cannot dynamically optimize resource scheduling under the constraint of timeliness, efficiency and energy consumption, reduce the intelligence, robustness and expandability of scheduling, cannot realize high-precision defect identification and automatic diagnosis, cannot combine a causal graph neural network to construct an interpretable causal model, cannot utilize digital twin deduction production evolution and dynamically generate an intervention strategy, and cannot form a closed loop of identification, diagnosis, attribution and intervention, and reduce the intelligence, traceability and active prevention capability of quality control. The aim of the invention is achieved by the following technical scheme: A digital MES management system based on intelligent manufacturing integrated in an intelligent manufacturing management platform, comprising: The data acquisition processing module is used for acquiring production operation parameters in the manufacturing execution process, sequentially carrying out signal filtering, missing value processing, abnormal value primary screening, data standardization and data compression aggregation on the acquired production operation parameters at the edge side, and automatically uploading the data to the intelligent manufacturing management