Precast beam field production scheduling analysis management method and system based on large model

Abstract

The invention discloses a precast beam field production scheduling analysis management method and system based on a large model, and the method comprises the steps: combing precast beam production technological processes, constructing a beam field Internet of Things sensing network, and collecting production information in real time; defining a general engineering data format, establishing a unified standardized beam field production knowledge base, and performing multi-modal data fusion on production elements; analyzing feasibility constraint rule conditions, describing a precast beam production scheduling problem, and establishing a beam field self-organizing scheduling decision mechanism; defining production state characteristics, establishing a beam field scheduling decision process model, and training a large scheduling model based on a PPO algorithm; and developing a beam field scheduling analysis management platform, automatically generating a production scheduling scheme in combination with a real-time production state, and realizing dynamic management of beam field production scheduling. The problems that in the prior art, the resource utilization efficiency is low, scheduling decision making depends on artificial experience, and continuous learning and self-optimization capacity is lacked are solved, and the intelligent level and the management effect of the precast beam field are improved.

Description

Technical Field

[0001] The present invention relates to the technical field related to smart beam yard and intelligent construction, and specifically to a production scheduling analysis management method and system for a prefabricated beam yard based on a large model. Background Art

[0002] With the vigorous development of building industrialization and intelligent construction technology, the country is vigorously advocating the coordinated development of intelligent construction and building industrialization, and gradually introducing intelligent construction technology into the field of infrastructure and transportation, in order to achieve the transformation from labor-intensive production mode to technology-intensive production mode, and achieve the goal of empowering enterprises. The precast beam yard is the main production site for precast beams. It is a large-scale temporary project with the characteristics of large investment, short operation cycle, and high beam production requirements. However, the existing beam yard scheduling is difficult to fully utilize the complex constraint relationship between process, equipment, materials and other data, and the versatility and practicality of the relevant scheduling schemes are not high.

[0003] Intelligent scheduling technology based on large models has application capabilities such as data mining, machine learning, cognition and knowledge engineering, and intelligent computing. It can make full use of the production data of the prefabricated beam yard, such as production processes, equipment parameters, material management, etc., and establish a production knowledge system for the entire life cycle of prefabricated beams through deep learning and natural language processing technology. At the same time, the AI-based intelligent scheduling system can perceive changes in the status of production orders, production equipment, material inventory, etc. in real time, integrate multi-factor production constraints, and automatically produce the best production intelligent scheduling according to actual conditions. It can further store and optimize large-scale knowledge bases to form an adaptive scheduling decision-making plan with accuracy, real-time performance, and intelligence, which improves the automation and intelligence level of the smart beam yard and reduces resource consumption and production costs.

[0004] Therefore, a large-model-based precast beam yard production scheduling analysis and management method is urgently needed to solve the problems that the existing scheduling management method cannot generate a highly universal and profitable intelligent scheduling decision-making plan based on the actual situation of the beam yard under multiple resource constraints, as well as the low utilization rate of various business data of the precast beam yard. Summary of the invention

[0005] The purpose of the present invention is to solve the problem of how to use a large model to perform scheduling state analysis and scheduling decision evaluation in a self-organizing, self-learning and self-optimizing manner, thereby realizing agile response and intelligent decision-making in precast beam yard scheduling. In order to achieve the purpose of the present invention, the technical solution adopted is:

[0006] The production scheduling analysis and management method of the precast beam yard based on the large model includes the following steps:

[0007] Step S1, sort out the prefabricated beam production process, build a beam field IoT sensor network, and collect production data in real time;

[0008] Step S2: define a general engineering data format, establish a unified and standardized beam yard production knowledge base, and perform multimodal data fusion on the production data;

[0009] Step S3, analyzing the feasibility constraint rule conditions, describing the precast beam production scheduling problem, and establishing a beam yard self-organizing scheduling decision-making mechanism;

[0010] Step S4, defining production status characteristics, establishing a beam yard scheduling decision process model, and training a large scheduling model based on the PPO algorithm;

[0011] Step S5: Develop a beam yard scheduling analysis management platform, automatically generate a production scheduling plan based on the real-time production status, and realize dynamic management of beam yard production scheduling.

[0012] Preferably, in step S1, the beam yard IoT sensor network includes monitoring of key production equipment and intelligent equipment;

[0013] The key production equipment includes: steel bar processing equipment, mobile pedestal, hydraulic formwork, material placing machine, torpedo tank, steam curing kiln, tensioning grouting, gantry crane equipment that reflects the production status through production process parameters;

[0014] The intelligent equipment includes equipment that integrates video surveillance, environmental monitoring, and AI visual recognition functions through information technology;

[0015] The production data includes engineering information, component information, order information, personnel information, equipment information, material information, production process information, and environmental information.

[0016] Preferably, in step S2, the beam yard production knowledge base includes: a personnel data set, an equipment data set, a material data set, a process method professional data set and a scheduling rule data set;

[0017] The process and method professional data set includes enterprise-level precast beam production process standards specific to the work steps formed by project accumulation;

[0018] The scheduling rule data set includes an interactive rule set for logical reasoning and decision-making under multi-objective resource constraints implemented by a large model; the multimodal data fusion includes providing a unified data structure, extracting key data from the collected multimodal data, and realizing the unification of the multimodal data structure based on the preprocessing of the multimodal data.

[0019] Preferably, in step S3, the feasibility constraint rule conditions include, respectively, establishing rules and presetting target conditions based on the production scheduling needs of the precast beam yard under the conditions of satisfying material constraints, process constraints, equipment constraints, personnel constraints, and yard constraints, and using constraint entropy to obtain the optimal solution for the scheduling target, the formula is as follows:

[0020]

[0021] The constraint entropy includes the ratio of daily constraint conditions to total usage. In order to express the degree of balance in the use of constraint conditions, the larger the value, the more balanced the distribution of the constraint resource conditions. The smaller the value, the more unbalanced the distribution of the constraint resource conditions. It is necessary to reconsider the optimization resource scheduling scheme. x is the constraint entropy of the xth constraint, v xt is the usage of the xth constraint at time t, θ k is the specified construction period of the component; Z x The larger the value, the more balanced the distribution of the constraint resource conditions. x The smaller the value, the more unbalanced the distribution of the constraint resource condition is, and the resource scheduling optimization scheme needs to be reconsidered;

[0022] The self-organizing scheduling decision mechanism of the beam yard includes: after receiving the order information, the model task unit retrieves the corresponding production component parameters, process information, completion date and other information from the beam yard production knowledge base, generates the process to be produced, binds the process information with the component model, and then selects the production line with the production capacity of the process, and releases the production task to the execution unit; after accepting the task, the execution unit analyzes the process attributes of the prefabricated component production, and determines whether to add new production tasks according to the workload, production capacity and workstation occupancy of the production line and personnel; if the production line or production equipment is occupied or faulty, the evaluation result is that the new production task cannot be accepted; according to the real-time workstation production status of the beam yard production data, the updated production status is input into the planning unit and the optimal decision action is generated; the action is repeated until there is a workstation capable of accepting the new production task; when the production equipment completes the current production process, the execution unit sends a signal to the planning unit to request scheduling, inputs the current waiting process status into the planning unit and generates the optimal decision action; through the mutual transmission of data among the execution unit, the planning unit and the task unit, the model tracks the processing status of the component in real time.

[0023] Preferably, the beam yard scheduling decision process modeling in step S4 includes establishing a Markov decision process mathematical model based on the multi-resource constraint scheduling environment of the prefabricated beam yard, including designing a state space S, an action space A, a state transition P and a reward function R in the form of a quaternion, as shown in the formula:

[0024] r t =R(s t ,a t )

[0025] Among them, R() represents the transition from state s t With action at To reward t The state space reflects all the production scheduling information in the beam field production. The state setting needs to express the characteristics of the collaborative manufacturing environment as much as possible, and represent the values ​​of each state variable. The sample defines 6 dimensional features, including: task scheduling results, task start time, task end time, process start production time, process end production time; the action space is the set of actions decided by the model; the reward function is the feedback obtained after the model decides the action under certain conditions. The function setting research focuses on the four aspects of production minimization delay, production, transportation, and waiting costs.

[0026] Preferably, in step S4, the training of the scheduling model based on the PPO algorithm includes, when a scheduling event is triggered, sensing the production status and inputting it into the Actor-new network to obtain the Q value of the action, and converting the Q value into the action A through SoftMax calculation. t , the probability distribution, according to which the production line is selected for the component, and then a reward r is fed back t and the latest production status t+1 , loop multiple times, and schedule data (s 1 ,a 1 ,r 1 ......,s t ,a t ,r t ) is stored in the experience buffer pool;

[0027] Will s t Input to the Critic-NN network, the network will generate the value V(s) of the action t ); then, the discounted future cumulative reward R = {R0, R1, …, RT} is obtained through the discount transformation Rt, where γ represents the decay factor; the advantage function and the loss function of the critic network are calculated, and the critic network parameters are updated by back propagation;

[0028] Input the stored state s into the Actor-old and Actor-new neural networks to obtain the probability distribution and calculate the importance sampling rate;

[0029] Clip is used to limit the update amplitude of the Actor network during the important sampling process, the loss function of the Actor network is calculated, and the parameters of the old Actor-old network are replaced with the parameters of the Actor-new network.

[0030] Repeat until all training sets are completed; when the learning curve remains relatively stable, it indicates that the scheduling agent has learned appropriate scheduling rules according to changes in production status.

[0031] Preferably, in the method of the present invention, it is necessary to develop a beam yard scheduling analysis management platform, integrate the trained large model into the platform, automatically generate a production scheduling plan in combination with the real-time production status, create an intelligent AI scheduling assistant, and allow users to query production information through natural language input. The assistant can reply in the form of text, ensuring that the large model can quickly respond to user needs.

[0032] The present invention also provides a large-model-based precast beam yard production scheduling analysis and management system, including: a production data perception module, a production knowledge management module, a large-model adaptive scheduling module, and an intelligent AI scheduling assistant module;

[0033] The production data perception module is used to store production data such as component information, order information, material information, etc. monitored from key production equipment and intelligent equipment, and transmit it to the production knowledge management module;

[0034] The production knowledge management module is used to define a unified data structure, extract key data from the collected multimodal data, and build a unified and standardized beam production knowledge base for precast beams through knowledge sources, knowledge representation, entity recognition, and relationship extraction, providing data support for large model training;

[0035] The large model adaptive scheduling module: constructs a feature extraction network, establishes an AI scheduling large model based on reinforcement learning, designs an adaptive learning mechanism for the scheduling process, uses knowledge base data for big data training to further optimize the scheduling plan, completes algorithm optimization and iteration, and realizes adaptive scheduling under multi-resource constraints;

[0036] The intelligent AI scheduling assistant module: creates an intelligent AI scheduling assistant, automatically generates a production scheduling Gantt chart based on real-time production information, and further optimizes the production data perception module, production knowledge management module and large model adaptive scheduling module through accuracy evaluation and feedback of the generated data.

[0037] The present invention also provides a computer device, including a processor and a memory, wherein the memory can run a computer program on the processor, and when the processor executes the program, the steps of the prefabricated beam yard production scheduling analysis management method based on a large model are implemented.

[0038] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the production scheduling analysis and management method of a prefabricated beam yard based on a large model are implemented.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows: the production scheduling analysis and management method of the precast beam yard based on the large model provided by the present invention collects and analyzes production factor data in real time by deploying a sensor network, thereby improving the timeliness and accuracy of the data; by defining a common engineering data format, the multi-source heterogeneous data of the precast beam yard is stored in a normalized knowledge base, thereby improving the utilization rate of production resources; by establishing a self-organizing scheduling decision-making mechanism for the beam yard, more accurate project scheduling generation rules and resource optimization are achieved; by using a deep reinforcement learning model to perform scheduling state analysis and scheduling decision evaluation in a self-organizing, self-learning and self-optimizing manner, full use is made of on-site production big data, and the problem of on-site scheduling relying on manual experience rather than data-driven scheduling decision-making is solved; by developing a beam yard scheduling management platform and integrating a large scheduling model, the model's continuous learning and self-optimization capabilities are enhanced, effectively improving the intelligence level and management effectiveness of the precast beam yard. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. The drawings are only used to illustrate the preferred implementation modes and are not considered to limit the present invention.

[0041] Figure 1 An overall flow chart of a large-model-based precast beam yard production scheduling analysis and management method provided for the first embodiment of the present invention.

[0042] Figure 2 An overall flow chart of a large-model-based precast beam yard production scheduling analysis and management system provided for the second embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the following further describes how the present invention is implemented in conjunction with the accompanying drawings and specific implementation methods.

[0044] Example 1

[0045] Reference Figure 1 , which is the first embodiment of the present invention, provides a precast beam yard production scheduling analysis management method based on a large model, including:

[0046] S1: Sort out the precast beam production process, build an IoT sensor network for beam yards, collect production information in real time, and realize data sharing and global perception throughout the entire life cycle of precast beam production.

[0047] The IoT sensor network of the beam yard includes monitoring of key production equipment and intelligent equipment; the key production equipment includes steel bar processing equipment, mobile pedestals, hydraulic formwork, fabricators, torpedo tanks, steam-curing kilns, tensioning and grouting, gantry cranes and other equipment that reflect the production status through production process parameters. The intelligent equipment includes equipment that integrates video surveillance, environmental monitoring, AI visual recognition and other functions through information technology.

[0048] The production information (Table 1) includes engineering information, component information, order information, personnel information, equipment information, material information, production process information, environmental information, etc.

[0049] Table 1:

[0050]

[0051]

[0052] S2: Define the general engineering data format, establish a unified and standardized beam yard production knowledge base, and perform multimodal data fusion on production factors.

[0053] The beam yard production knowledge base includes a personnel data set, an equipment data set, a material data set, a process and method professional data set, and a scheduling rule data set; the process and method professional data set includes enterprise-level prefabricated beam production process standards specific to work steps formed by project accumulation.

[0054] The scheduling rule data set includes an interactive rule set for logical reasoning and decision-making under multi-objective resource constraints implemented by a large model; the multimodal data fusion includes providing a unified data structure that can extract key data from the collected multimodal data, and realize the unification of the multimodal data structure based on the preprocessing of the multimodal data.

[0055] S3: Analyze the feasibility constraint rule conditions, describe the precast beam production scheduling problem, and establish a self-organizing scheduling decision-making mechanism for the beam yard.

[0056] The feasibility constraint rule conditions include, respectively, establishing rules and presetting target conditions based on the production scheduling needs of the prefabricated beam yard under the conditions of satisfying material constraints, process constraints, equipment constraints, personnel constraints, and yard constraints, and using constraint entropy to obtain the optimal solution for the scheduling target. The formula is as follows:

[0057]

[0058] The constraint entropy includes the ratio of daily constraint usage to total usage. In order to better express the degree of balance in the use of constraint conditions, the larger the value, the more balanced the distribution of the constraint resource conditions. The smaller the value, the more unbalanced the distribution of the constraint resource conditions. It is necessary to reconsider the optimization resource scheduling scheme. x is the constraint entropy of the xth constraint, v xt is the usage of the xth constraint at time t, θ k is the specified construction period of the component, when Z x The larger the value, the more balanced the distribution of the constraint resource conditions. x The smaller the value, the more unbalanced the distribution of the constraint resource condition is, and the resource scheduling optimization scheme needs to be reconsidered;

[0059] The self-organizing scheduling decision mechanism of the beam yard includes: after receiving the order information, the model task unit retrieves the corresponding production component parameters, process information, completion date and other information from the beam yard production knowledge base, generates the process to be produced, binds the process information with the component model, and then selects the production line with the production capacity of the process, and releases the production task to the execution unit; after accepting the task, the execution unit analyzes the process attributes of the prefabricated component production, and determines whether to add new production tasks according to the workload, production capacity and workstation occupancy of the production line and personnel; if the production line or production equipment is occupied or faulty, the evaluation result is that the new production task cannot be accepted; according to the real-time workstation production status of the beam yard production data, the updated production status is input into the planning unit and the optimal decision action is generated; the action is repeated until there is a workstation capable of accepting the new production task; when the production equipment completes the current production process, the execution unit sends a signal to the planning unit to request scheduling, inputs the current waiting process status into the planning unit and generates the optimal decision action; through the mutual transmission of data among the execution unit, the planning unit and the task unit, the model can track the processing status of the component in real time.

[0060] S4: Define the production status characteristics, establish the beam yard scheduling decision process model, and train the scheduling model based on the PPO algorithm.

[0061] The beam yard scheduling decision process modeling includes establishing a Markov decision process mathematical model based on the multi-resource constraint scheduling environment of the prefabricated beam yard. The Markov decision process includes that the Markov decision process MDP model is a sequential decision mathematical model that can enable the intelligent agent to learn an action selection strategy that can maximize the cumulative reward through interactive learning, including designing the state space S, action space A, state transition P and reward function R in the form of a quaternary as shown in the formula:

[0062] r t =R(s t ,at )

[0063] Among them, R() represents the transition from state s t With action a t To reward t The state space reflects all the production scheduling information in the beam field production. The state setting needs to express the characteristics of the collaborative manufacturing environment as much as possible, and represent the values ​​of each state variable. The sample defines 6 dimensional features, including: task scheduling results, task start time, task end time, process start production time, process end production time; the action space is the set of actions decided by the model; the reward function is the feedback obtained after the model decides the action under certain conditions. The function setting research focuses on the four aspects of production minimization delay, production, transportation, and waiting costs.

[0064] The PPO algorithm-based scheduling model training includes: when a scheduling event is triggered, sensing the production status and inputting it into the Actor-new network to obtain the Q value of the action, and converting the Q value into the action A through SoftMax calculation. t , the probability distribution, according to which the production line is selected for the component, and then a reward r is fed back t and the latest production status t+1 , loop multiple times, and schedule data (s 1 ,a 1 ,r 1 ......,s t ,a t ,r t ) is stored in the experience buffer pool;

[0065] Will s t Input to the Critic-NN network, the network will generate the value V(s) of the action t ). Then, by discounting the transformation R t Get discounted future cumulative reward R = {R 0 ,R 1 ,…,R T}, where γ represents the attenuation factor. Calculate the advantage function and the loss function of the critic network, and back-propagate to update the critic network parameters;

[0066] Input the stored state s into the Actor-old and Actor-new neural networks to obtain the probability distribution and calculate the importance sampling rate;

[0067] Clip is used to limit the update amplitude of the Actor network during the important sampling process, the loss function of the Actor network is calculated, and the parameters of the old Actor-old network are replaced with the parameters of the Actor-new network. Repeat until all training sets are completed.

[0068] The adaptive learning mechanism of the scheduling process includes the scheduling state, scheduling action, reward and training algorithm design in the deep reinforcement learning modeling process. The scheduling state is defined as production data related to the equipment, materials, etc.

[0069] The local state variable equipment utilization U of the sorting agent at decision time t k (t), the average completion rate of all production CR(t), the machine load rate AT k (t), estimated delay rate And the actual delay rate The above five state variables reflect the production status of the entire beam yard. The data tuple corresponding to the five observation value features is as follows:

[0070]

[0071] Defining CT h (t) is the completion time of the last operation of device h at time t, Indicates that operation J of process i has been completed on equipment k, where n i (t) represents the remaining workload. The utilization rate of equipment h at time t is U k (t) is calculated as:

[0072]

[0073] According to the completion rate CJ of task i at time t i (t), then the average completion rate of all tasks is calculated as follows:

[0074]

[0075] According to the load rate of all machines at time t, ATk(t), the total processing time of the process in the equipment queue is calculated as follows:

[0076]

[0077] (t) represents the estimated delay rate of the workpiece at time t, expressed as the ratio of the remaining processing time of the job to the slack time, (t) represents the actual delay rate of the workpiece at time t, and the formula is as follows:

[0078]

[0079] The scheduling action is a collection of scheduling decisions under different states. At decision time t, when there is an idle device requesting scheduling, the idle production device should be given priority:

[0080] argmin k∈Ωij (AT k (t))

[0081] The reward function is usually defined based on the advantages and disadvantages of the state space during the operation. Since the goal of the reinforcement learning algorithm is to maximize the cumulative reward, its increasing direction should be consistent with the decreasing direction of the optimization target value. The global reward function at time t is calculated based on the cumulative lead time r. lead and machine load balancing load The compound reward is expressed as follows: if the expected completion time of the process is earlier than the actual completion time, then r lead is positive; otherwise it is negative; such as the formula:

[0082]

[0083] Among them, EC i (t) is the workpiece J at time t i The estimated completion time is calculated using the scheduling time t and the remaining processing time of the workpiece, as shown in the formula:

[0084]

[0085] S5: Develop a beam yard scheduling analysis and management platform to automatically generate production scheduling plans based on real-time production status, and realize dynamic management of beam yard production scheduling.

[0086] Going a step further, we created an intelligent AI scheduling assistant that allows users to query production information through natural language input, and the assistant can reply in the form of text, ensuring that the big model can quickly respond to user needs.

[0087] Example 2

[0088] Reference Figure 2 , is an embodiment of the present invention, and provides a precast beam yard production scheduling analysis and management system based on a large model, which is characterized by: a production data perception module, a production knowledge management module, a large model adaptive scheduling module, and an intelligent AI scheduling assistant module;

[0089] The production data perception module is used to store production data such as component information, order information, material information, etc. monitored from key production equipment and intelligent equipment, and transmit it to the production knowledge management module;

[0090] The production knowledge management module is used to define a unified data structure, extract key data from the collected multimodal data, and build a unified and standardized beam production knowledge base for precast beams through knowledge sources, knowledge representation, entity recognition, and relationship extraction, providing data support for large model training;

[0091] The large model adaptive scheduling module: constructs a feature extraction network, establishes an AI scheduling large model based on reinforcement learning, designs an adaptive learning mechanism for the scheduling process, uses knowledge base data for big data training to further optimize the scheduling plan, completes algorithm optimization and iteration, and realizes adaptive scheduling under multi-resource constraints;

[0092] The intelligent AI scheduling assistant module: creates an intelligent AI scheduling assistant, automatically generates a production scheduling Gantt chart based on real-time production information, and further optimizes the production data perception module, production knowledge management module and large model adaptive scheduling module through accuracy evaluation and feedback of the generated data.

[0093] Example 3

[0094] An embodiment of the present invention is different from the first two embodiments in that:

[0095] If the functions are expressed in the form of software functional units and used or sold as independent products, they can be stored in a computer-readable storage medium. Based on this, the technical solution of the present invention can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: a mobile hard disk, a USB flash drive, a read-only memory, an optical disk, and other media that can store program codes.

[0096] The logic and / or steps represented in the flowchart can be considered as an ordered list of instructions for implementing logical functions, which can be specifically implemented in a computer-readable medium for use by an instruction execution system, device or equipment, or used in combination with these instruction execution systems, devices or equipment.

[0097] More specific examples of computer readable media include the following: installation media, such as CD-ROM, floppy disk or tape device; computer system memory or random access memory, such as DRAM, SRAM, EDO RAM, etc. The storage medium includes two or more storage media that can reside in different locations. The storage medium can store program instructions that can be executed by one or more processors.

[0098] It should be understood that each functional unit of the present invention can be implemented by software, hardware, firmware or a combination thereof. In the above-mentioned real-time method, it is implemented by hardware, and in another embodiment, it can be implemented by any one of the following techniques known in the art or a combination thereof. If implemented by software or firmware, multiple steps or methods can be implemented by controlling them with system execution commands stored in a memory.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for production scheduling analysis and management of precast beam yard based on a large model, characterized in that: The following steps are involved: Step S1, sort out the prefabricated beam production process, build a beam field IoT sensor network, and collect production data in real time; Step S2: define a general engineering data format, establish a unified and standardized beam yard production knowledge base, and perform multimodal data fusion on the production data; Step S3, analyzing the feasibility constraint rule conditions, describing the precast beam production scheduling problem, and establishing a beam yard self-organizing scheduling decision-making mechanism; Step S4, defining production status characteristics, establishing a beam yard scheduling decision process model, and training a large scheduling model based on the PPO algorithm; Step S5: Develop a beam yard scheduling analysis management platform, automatically generate a production scheduling plan based on the real-time production status, and realize dynamic management of beam yard production scheduling.

2. The method for production scheduling, analysis and management of precast beam yard based on large model according to claim 1 is characterized in that: In step S1, the beam yard IoT sensor network includes monitoring of key production equipment and intelligent equipment; The key production equipment includes: steel bar processing equipment, mobile pedestal, hydraulic formwork, material placing machine, torpedo tank, steam curing kiln, tensioning grouting, gantry crane equipment that reflects the production status through production process parameters; The intelligent equipment includes equipment that integrates video surveillance, environmental monitoring, and AI visual recognition functions through information technology; The production data includes engineering information, component information, order information, personnel information, equipment information, material information, production process information, and environmental information.

3. The method for production scheduling, analysis and management of precast beam yard based on large model according to claim 1 is characterized in that: In step S2, the beam yard production knowledge base includes: a personnel data set, an equipment data set, a material data set, a process method professional data set and a scheduling rule data set; The process and method professional data set includes enterprise-level precast beam production process standards specific to the work steps formed by project accumulation; The scheduling rule data set includes an interactive rule set for logical reasoning and decision-making under multi-objective resource constraints implemented by a large model; the multimodal data fusion includes providing a unified data structure, extracting key data from the collected multimodal data, and realizing the unification of the multimodal data structure based on the preprocessing of the multimodal data.

4. The method for production scheduling, analysis and management of precast beam yard based on large model according to claim 1 is characterized in that: In step S3, the feasibility constraint rule conditions include establishing rules and presetting target conditions based on the production scheduling needs of the precast beam yard under the conditions of satisfying material constraints, process constraints, equipment constraints, personnel constraints, and yard constraints, and using constraint entropy to obtain the optimal solution for the scheduling target. The formula is as follows: The constraint entropy includes the ratio of daily constraint conditions to total usage. In order to express the degree of balance in the use of constraint conditions, the larger the value, the more balanced the distribution of the constraint resource conditions. The smaller the value, the more unbalanced the distribution of the constraint resource conditions. It is necessary to reconsider the optimization resource scheduling scheme. x is the constraint entropy of the xth constraint, v xt is the usage of the xth constraint at time t, θ k is the specified construction period of the component; Z x The larger the value, the more balanced the distribution of the constraint resource conditions. x The smaller the value, the more unbalanced the distribution of the constraint resource condition is, and the resource scheduling optimization scheme needs to be reconsidered; The self-organizing scheduling decision mechanism of the beam yard includes: after receiving the order information, the model task unit retrieves the corresponding production component parameters, process information, completion date and other information from the beam yard production knowledge base, generates the process to be produced, binds the process information with the component model, and then selects the production line with the production capacity of the process, and releases the production task to the execution unit; after accepting the task, the execution unit analyzes the process attributes of the prefabricated component production, and determines whether to add new production tasks according to the workload, production capacity and workstation occupancy of the production line and personnel; if the production line or production equipment is occupied or faulty, the evaluation result is that the new production task cannot be accepted; according to the real-time workstation production status of the beam yard production data, the updated production status is input into the planning unit and the optimal decision action is generated; the action is repeated until there is a workstation capable of accepting the new production task; when the production equipment completes the current production process, the execution unit sends a signal to the planning unit to request scheduling, inputs the current waiting process status into the planning unit and generates the optimal decision action; through the mutual transmission of data among the execution unit, the planning unit and the task unit, the model tracks the processing status of the component in real time.

5. The method for production scheduling, analysis and management of precast beam yard based on large model according to claim 1 is characterized in that: Step S4: The beam yard scheduling decision process model includes establishing a Markov decision process mathematical model based on the multi-resource constraint scheduling environment of the prefabricated beam yard, including designing a state space S, an action space A, a state transition P, and a reward function R in the form of a quaternary, as shown in the formula: r t =R(s t ,a t ) Among them, R() represents the transition from state s t With action a t To reward t The state space reflects all the production scheduling information in the beam yard production. The state setting needs to express the characteristics of the collaborative manufacturing environment as much as possible, and represent the numerical values ​​of each state variable. The sample defines 6 dimensional features, including: task scheduling results, task start time, task end time, process start production time, process end production time; the action space is the set of actions decided by the model; the reward function is the feedback obtained after the model decides the action under certain conditions. The function setting research focuses on the four aspects of production minimization delay, production, transportation, and waiting costs.

6. The method for production scheduling, analysis and management of precast beam yard based on large model according to claim 1 is characterized in that: In step S4, the PPO algorithm-based scheduling model training includes, when a scheduling event is triggered, sensing the production status and inputting it into the Actor-new network to obtain the Q value of the action, and converting the Q value into the action A through SoftMax calculation. t , the probability distribution, according to which the production line is selected for the component, and then a reward r is fed back t and the latest production status t+1 , loop multiple times, and schedule data (s1, a1, r1......, s t ,a t ,r t ) is stored in the experience buffer pool; Will s t Input to the Critic-NN network, the network will generate the value V(s) of the action t ); then, the discounted future cumulative reward R = {R0, R1, …, RT} is obtained through the discount transformation Rt, where γ represents the decay factor; the advantage function and the loss function of the critic network are calculated, and the critic network parameters are updated by back propagation; Input the stored state s into the Actor-old and Actor-new neural networks to obtain the probability distribution and calculate the importance sampling rate; Clip is used to limit the update amplitude of the Actor network during the important sampling process, the loss function of the Actor network is calculated, and the parameters of the old Actor-old network are replaced with the parameters of the Actor-new network. Repeat until all training sets are completed; when the learning curve remains relatively stable, it indicates that the scheduling agent has learned appropriate scheduling rules according to changes in production status.

7. The method for production scheduling, analysis and management of precast beam yard based on large model according to claim 6 is characterized in that: Develop a beam yard scheduling analysis management platform and integrate the trained large model into the platform. Combined with the real-time production status, the production scheduling plan is automatically generated and an intelligent AI scheduling assistant is created. This allows users to query production information through natural language input. The assistant can reply in text form, ensuring that the large model can quickly respond to user needs.

8. The large model-based precast beam yard production scheduling analysis and management system according to any one of claims 1 to 7, characterized in that: include: Production data perception module, production knowledge management module, large model adaptive scheduling module, intelligent AI scheduling assistant module; The production data perception module is used to store production data such as component information, order information, material information, etc. monitored from key production equipment and intelligent equipment, and transmit it to the production knowledge management module; The production knowledge management module is used to define a unified data structure, extract key data from the collected multimodal data, and build a unified and standardized beam production knowledge base for precast beams through knowledge sources, knowledge representation, entity recognition, and relationship extraction, providing data support for large model training; The large model adaptive scheduling module: constructs a feature extraction network, establishes an AI scheduling large model based on reinforcement learning, designs an adaptive learning mechanism for the scheduling process, uses knowledge base data for big data training to further optimize the scheduling plan, completes algorithm optimization and iteration, and realizes adaptive scheduling under multi-resource constraints; The intelligent AI scheduling assistant module: creates an intelligent AI scheduling assistant, automatically generates a production scheduling Gantt chart based on real-time production information, and further optimizes the production data perception module, production knowledge management module and large model adaptive scheduling module through accuracy evaluation and feedback of the generated data.

9. A computer device comprising a processor and a memory, wherein the memory can run a computer program on the processor, characterized in that: When the processor executes the program, the steps of the prefabricated beam yard production scheduling analysis management method based on a large model are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the prefabricated beam yard production scheduling analysis management method based on a large model are implemented.

Images