Spinning enterprise multi-order scheduling method and device based on thinking tree structure

Abstract

The invention relates to a spinning enterprise multi-order scheduling method and device based on a thinking tree structure, and belongs to the field of spinning enterprise order scheduling. According to the invention, through intelligent data collection and preprocessing, accurate analysis of production data is realized; complex production information is effectively organized and managed by constructing a scheduling thinking tree; resource allocation is optimized through an evolutionary computing method, and the resource utilization rate is increased; by making a production scheduling plan, the reliability of the order delivery period is ensured; and through real-time monitoring and adjustment, the capability of dealing with emergencies is enhanced. On the company level, through an intelligent and automatic scheduling system, the production efficiency is remarkably improved, manual intervention and errors are reduced, and the production cost is reduced. On the industrial level, the digital transformation of the textile industry and the improvement of the intelligent manufacturing level are promoted, and the market competitiveness of enterprises is enhanced. According to the invention, resource optimization configuration of the textile industry can be realized, and sustainable development of the whole industry can be promoted.

Description

Technical Field

[0001] The present invention belongs to the field of order scheduling for spinning enterprises, and in particular relates to a method and device for scheduling multiple orders for spinning enterprises based on a mind tree structure. Background Art

[0002] The textile industry is an important part of the light industry and has an important impact and contribution to the national economy. As a traditional industry, the textile industry not only plays a significant role in absorbing labor, but also occupies an important position in export earnings. As the basic link of textile production, the spinning industry is crucial to the operation of the entire textile industry chain. The textile industry occupies a pivotal position in my country's national economy and has made great contributions to economic growth, employment promotion and technological progress over the years. In recent years, with the advancement of science and technology and changes in market demand, the textile industry has continuously carried out industrial upgrading and technological innovation to adapt to the needs of global competition and sustainable development. In addition, the textile industry has also actively responded to national policies, promoted the development of green environmental protection and intelligent manufacturing, and further enhanced the international competitiveness of the industry.

[0003] As an important part of the textile industry, the multi-order scheduling problem of spinning enterprises has increasingly become a key factor affecting production efficiency and corporate competitiveness. Due to the wide variety of textile products and complex production processes, there are significant differences between orders in terms of process flow, delivery time and quality requirements. Therefore, how to effectively arrange and dispatch multiple orders under limited production resources has become an important challenge facing textile enterprises. Traditional scheduling methods often rely on experience and manual operations, which are difficult to adapt to the changing market demand and personalized customization trends in modern textile production. In addition, the continuous increase in orders and the shortening of delivery time have further aggravated the complexity of scheduling, resulting in problems such as waste of resources, production delays and inventory backlogs in the production process.

[0004] At present, spinning enterprises are facing the following major problems in the process of multi-order scheduling: First, the order complexity is high, the number of orders is large and the types are various. The traditional manual scheduling method is difficult to handle efficiently, resulting in low scheduling efficiency. Second, the resource utilization rate is uneven. Equipment is often idle or overloaded during the production process, and existing resources cannot be fully utilized. Third, it is difficult to guarantee the delivery date of orders. Due to unreasonable scheduling, production delays often occur, resulting in the inability to deliver customer orders on time. Fourth, there is a lack of real-time monitoring and adjustment capabilities. Once an emergency occurs during the production process, it is difficult to adjust the production plan in time, resulting in a disconnection between the production plan and actual production. These problems have seriously affected the production efficiency and market competitiveness of enterprises.

[0005] To this end, textile companies urgently need an efficient and intelligent multi-order scheduling method to optimize resource allocation, improve production efficiency, meet the diverse needs of customers, and maintain their advantages in the fierce market competition. Summary of the invention

[0006] The purpose of the present invention is to solve the complex and changeable production scheduling problem of spinning enterprises, and to provide a multi-order scheduling method and device for spinning enterprises based on a mind tree structure. Through intelligent data collection and preprocessing, accurate analysis of production data is achieved; by constructing a scheduling mind tree, complex production information is effectively organized and managed; resource allocation is optimized through evolutionary computing methods to improve resource utilization; by formulating a production scheduling plan, the reliability of order delivery is ensured; through real-time monitoring and adjustment, the ability to respond to emergencies is enhanced. The present invention greatly improves the scheduling efficiency and production management level of spinning enterprises, ensures the on-time delivery of orders, optimizes resource allocation, and improves the overall competitiveness of enterprises.

[0007] To achieve the above object, the technical solution of the present invention is: a multi-order scheduling method for a spinning enterprise based on a thinking tree structure, comprising:

[0008] Data collection and preprocessing: Collect relevant data and perform cleaning and preprocessing;

[0009] Constructing a scheduling mind tree: Using the mind tree method to construct a scheduling mind tree, converting production demand into specific production tasks, and decomposing and clarifying the specific tasks and processes of each production line layer by layer;

[0010] Resource allocation optimization: Optimize resource allocation through evolutionary computing to ensure the best utilization of equipment, manpower and raw materials;

[0011] Formulate production scheduling plan: According to production tasks and resource allocation, determine the production tasks for each time period layer by layer, and formulate a preliminary production scheduling plan;

[0012] Real-time monitoring and adjustment: Monitor the progress and resource usage of each process in real time, and adjust tasks and resource allocation in a timely manner based on monitoring data to ensure the smooth implementation of the production plan.

[0013] In one embodiment of the present invention, the data collection and preprocessing are specifically implemented as follows:

[0014] The collection includes equipment data, raw material data, including raw material type, product data, order data, production process data, human resources data, energy consumption data, quality control data, logistics data, cost data, environmental data, production plan history data, enterprise management data, and compliance data;

[0015] Clean and preprocess the collected data.

[0016] In one embodiment of the present invention, equipment data includes the number and type of spinning machines, the number of spindles of each machine, the maximum operating speed and the actual operating speed, equipment efficiency and utilization rate, and equipment maintenance and servicing plan; raw material data includes raw material type and its proportion, inventory, quality parameters, supplier information and delivery cycle; product data covers yarn varieties, specifications, quality requirements and finished product inventory; order data includes customer order information, order priority and historical order data; production process data includes the production capacity and output of each process, the connection time between processes, the yarn change time and the production line balance data; human resource data includes the number of operators, skills, etc. The data includes production level, shift schedule and personnel efficiency data; energy consumption data, including energy consumption of each equipment and peak and valley electricity price information; quality control data, including online detection data, quality abnormality records, rework rate and scrap rate; logistics data, including transportation time, storage capacity and location information of raw materials and finished products; cost data, covering raw material cost, labor cost and energy cost; environmental data, including workshop temperature and humidity and air quality parameters; production plan historical data, including past scheduling plans, plan execution status and deviation analysis; enterprise management data, including enterprise resource planning system data and key performance indicators; compliance data, including industry standards and regulatory requirements and environmental protection indicators.

[0017] In one embodiment of the present invention, the preprocessed data is subjected to data, knowledge, and information extraction through technologies including data mining, machine learning, and natural language processing, and the textile industry big model is fine-tuned in the form of JSON questions and answers to construct the company-level spinning-specific big model of the enterprise; by using the mind tree framework, a scheduling mind tree is constructed to guide the company-level spinning-specific big model of the enterprise to solve complex scheduling problems.

[0018] In one embodiment of the present invention, the scheduling mind tree is constructed by taking the spinning production demand of the current month as the root node, with the first layer being different spinning types, the second layer being different production batches, the third layer being different processes, and the fourth layer being a hierarchical structure of different production lines; the specific construction method of the scheduling mind tree is as follows:

[0019] Decompose the spinning production demand of the month into multiple sub-nodes; each sub-node represents a specific product category; these sub-nodes include cotton yarn production demand, blended yarn production demand, dyed yarn production demand and special yarn production demand;

[0020] Each product category sub-node is further broken down into different production batches;

[0021] For each production batch, continue to decompose it into specific process steps, and each process step is used as a child node;

[0022] Each process step is assigned to each production line to ensure that the task of each production line is clear.

[0023] In one embodiment of the present invention, a genetic algorithm is used to optimize resource allocation through evolutionary computing, which is specifically implemented as follows:

[0024] The scheduling scheme is represented by a chromosome, which is composed of a series of genes; the gene encoding method is as follows:

[0025] Each gene i Represents the resource allocation of a task, including equipment R equip 、Human Resources human 、Raw materials material and its start time T start and end time T end :

[0026] g i =(R equip ,R human ,R material ,T start ,T end )

[0027] A chromosome C contains the resource allocation and scheduling information of all tasks:

[0028] C=[g1,g2,…,g n ]

[0029] Where n is the total number of tasks; the encoding method needs to ensure that the chromosome can meet resource constraints and task dependencies;

[0030] The initial population is the starting point of genetic algorithm optimization, which is composed of a group of randomly generated chromosomes. Each chromosome represents a feasible scheduling solution. The initialization process needs to ensure that: a) all tasks meet the production order constraints; b) the total amount of allocated resources does not exceed the available resources; the population size N is adjusted according to the complexity of the problem;

[0031] The fitness function F(C) is used to evaluate the quality of each chromosome, i.e., the scheduling scheme. It is defined based on production objectives and constraints in the form of:

[0032]

[0033] Where T total : Total processing time, defined as the latest end time of all tasks:

[0034] T total =max(T end,1 ,T end,2 ,…,T end,,n )

[0035] The goal is to minimize;

[0036] U resources : Resource utilization, calculated as the ratio of actual resource usage to available resources, taking into account equipment, manpower, and raw materials:

[0037]

[0038] The goal is to maximize;

[0039] P violations : Constraint violation penalty term, used to penalize solutions that do not meet the constraints:

[0040] P violations =λ1P resource +λ2P time

[0041] Where: P resource is the total amount of resource over-allocation; is the number of violations of the task dependency order; λ1 and λ2 are penalty coefficients;

[0042] The weights w1, w2, and w3 are adjusted according to actual needs to balance the importance of different goals;

[0043] The selection operation selects chromosomes to enter the next generation according to the fitness value, and the selection probability of excellent chromosomes is higher; the selection operation methods include: roulette method, tournament method, crossover operation, and mutation operation;

[0044] The genetic algorithm terminates when any of the following conditions are met:

[0045] 1) Reach the required algebra G max ;

[0046] 2) The fitness of the population converges, that is, the maximum fitness change is less than the threshold ∈;

[0047] After termination, the chromosome with the highest fitness value is selected from the population as the optimal scheduling solution C*, and its corresponding task allocation and scheduling information is output.

[0048] In an embodiment of the present invention, each selection operation method is specifically implemented as follows:

[0049] 1) Roulette method: Allocate selection probability according to the fitness value ratio;

[0050] 2) Tournament method: Randomly select multiple chromosomes to compete, select the one with higher fitness; select the retained chromosomes to form the intermediate population P t ;

[0051] 3) The crossover operation generates new offspring chromosomes by combining some genes of two chromosomes, simulating the recombination of genetic information; including:

[0052] 3.1) Single-point crossover: randomly select a crossover point and exchange the genes of the two chromosomes at the rear section of the crossover point;

[0053] 3.2) Multi-point crossover: select multiple crossover points and alternately exchange chromosome segments;

[0054] For two chromosomes:

[0055] C1=[g1,g2,g3,g4],C2=[g5,g6,g7,g8]

[0056] After crossover at the second gene, it generates:

[0057] C′1=[g1,g2,g7,g8],C′2=[g5,g6,g3,g4]

[0058] Crossover probability P c Take 0.6-0.9;

[0059] 4) The mutation operation randomly changes the gene values ​​in the chromosome to increase the diversity of the population and prevent the algorithm from falling into the local optimum; including:

[0060] 4.1) Randomly adjust the task start time T start ;

[0061] 4.2) Change the resource allocation of the task, including changing R equip or R human ;

[0062] Chromosome C = [g1, g2, …, g n ] before the gene g2 mutates:

[0063] g2=(R equip =1,R human =2,R material =3,T start =5,T end =10)

[0064] After mutation:

[0065] g′2=(R equip =2,R human =1,R material =4,T start =6,T end =11)

[0066] Mutation probability R m Take 0.01-0.1.

[0067] In one embodiment of the present invention, a particle swarm algorithm PSO is used to optimize resource allocation through evolutionary computing, which is specifically implemented as follows:

[0068] A scheduling scheme for resource allocation of a task, including equipment R equip 、Human Resources human 、Raw materials material and its start time T start and end time T end , each particle represents a schedule, and its position and velocity are defined as:

[0069] Particle position x i (t): represents the current scheduling plan, including the resource allocation and time arrangement of each task:

[0070] x i (t) = [x i,1 (t),x i,2 (t),…,x i,n (t)]

[0071] Among them, x i,j (t) represents the allocation status of particle i on the jth task:

[0072] x i,j (t)=(R equip ,R human ,R material ,T start ,T end )

[0073] Particle velocity vi(t): defined as the search direction and step size of the particle in the solution space:

[0074] v i (t) = [v i,1 (t),v i,2 (t),…,v i,n (t)]

[0075] The initial population of PSO consists of M particles, and the initial position x of each particle is i (t) and speed v i (t) Random generation, the following conditions must be met:

[0076] 1) Feasibility: The scheduling scheme corresponding to the initial position must satisfy resource constraints and task dependency constraints;

[0077] 2) Diversity: The particles in the population are widely distributed to cover the solution space;

[0078] After initializing the population, the fitness value of each particle is calculated for subsequent iterative optimization;

[0079] The core of PSO lies in the speed and position update of each generation of particles. The speed update formula of particles is:

[0080] v i (t+1)=ωv i (t)+c1r1[p i -x i (t)]+c2r2[gx i (t)]

[0081] Among them, ω is the inertia weight, which controls the degree of inheritance of the particle to the current velocity and affects the balance between global and local searches; c1 and c2 are learning factors, which represent the weights of the particle learning from its own optimal position and the global optimal position respectively; r1 and r2 are random numbers in [0,1], which are used to introduce search randomness; p i is the individual optimal position of particle i in the historical iteration; g is the global optimal position, the optimal position of all particles;

[0082] The particle's position is updated according to the velocity:

[0083] x i (t+1)=x i (t)+v i (t+1)

[0084] In order to ensure the feasibility of resource allocation, the updated x i (t+1) needs to be constrained and corrected:

[0085] Resource constraint correction: Ensure that the total amount of allocated resources R used No more than the total available resources R avail :

[0086] R used ≤R avail

[0087] Time dependency correction: adjust the start time T of task j start,j Ensure the completion time of its dependent task k:

[0088] T start, j≥T end, k

[0089] Fitness function F(x i (t)) is used to evaluate the quality of the particle's current position. Its goal is similar to that of the genetic algorithm, taking into account the following goals:

[0090]

[0091] Among them, T total is the total processing time, the goal is to minimize; U resources is the resource utilization, the goal is to maximize; P violationsTo constrain the violation penalty term, the goal is to minimize it; the selection of weights w1, w2, and w3 is adjusted according to actual needs to balance the production goals;

[0092] The specific process of PSO is as follows:

[0093] 1) Randomly generate the initial positions and velocities of M particles to meet basic constraints.

[0094] 2) Calculate the initial fitness value of each particle and initialize the individual optimal position p i and the global optimal position g.

[0095] 3) Iterative optimization: For each generation t, perform the following steps:

[0096] 4) Speed ​​update: Calculate the speed v of each particle according to the speed update formula i (t+1);

[0097] 5) Position update: Calculate the new position x of each particle according to the position update formula i (t+1), and perform constraint correction;

[0098] 6) Fitness calculation: Evaluate the new fitness value F(x) of each particle i (t+1));

[0099] 7) Update individual and global optimal positions:

[0100] If F(x i (t+1))>F(p i ), update p i =x i (t+1);

[0101] If F(p i )>F(g), update g=p i ;

[0102] 8) Reach the predetermined maximum number of iterations T max Or the change of the global optimal fitness value is less than the threshold, and the global optimal position g is output as the optimal scheduling solution.

[0103] In one embodiment of the present invention, the inertia weight ω determines the exploration capability of the particle search and adopts a linear decreasing strategy:

[0104]

[0105] Initial value ω max Take 0.9, the final value ω min Take 0.4;

[0106] Learning factors c1, c2: c1 = c2 = 2 to balance individual learning and group collaboration;

[0107] Population size M: M is 20-100 and is adjusted according to the complexity of the problem.

[0108] The present invention also provides a multi-order scheduling device for a spinning enterprise based on a thinking tree structure, comprising:

[0109] Data collection and preprocessing module, which collects and cleans multi-source data including production workshops, order systems, and industry standards;

[0110] Construct a scheduling thinking tree module, with the spinning production demand of the current month as the root node, the first layer for different spinning types, the second layer for different production batches, the third layer for different processes, and the fourth layer for different production lines.

[0111] Resource allocation optimization module, which uses evolutionary computing methods to dynamically allocate and optimize resources including manpower, equipment, and materials to ensure maximum resource utilization;

[0112] Formulate a production scheduling module, comprehensively consider the delivery date, production capacity and resource constraints of each order, and generate the optimal production scheduling plan;

[0113] Production task and resource monitoring module, real-time tracking of the progress of production tasks and resource usage status;

[0114] The real-time monitoring and adjustment module, based on the production task and resource monitoring module, continuously monitors the actual situation of the production site and makes timely adjustments to cope with emergencies and changing production needs.

[0115] Compared with the prior art, the present invention has the following beneficial effects: the present invention solves the complexity, uncertainty, resource optimization, efficiency improvement and cost control problems in the scheduling of spinning enterprises. The method comprises the following steps: data collection and preprocessing: collect relevant data and perform cleaning and preprocessing to ensure the accuracy and integrity of the data; construct a scheduling mind tree: adopt the mind tree method to construct a scheduling mind tree, convert production requirements into specific production tasks, and decompose and clarify the specific tasks and processes of each production line layer by layer. The mind tree method can decompose complex scheduling problems into multiple manageable sub-problems, and form an efficient scheduling plan through layer-by-layer analysis and decision-making; resource allocation optimization: optimize resource allocation through evolutionary computing to ensure the best use of equipment, manpower and raw materials; formulate a production scheduling plan: according to production tasks and resource allocation, determine the production tasks of each time period layer by layer, and formulate a preliminary production scheduling plan; real-time monitoring and adjustment: use the production management system to monitor the progress and resource usage of each process in real time, and adjust the task and resource allocation in time according to the monitoring data to ensure the smooth execution of the production plan. This method makes full use of the hierarchical decomposition and clear decision-making paths of the thinking tree to improve the transparency and operability of scheduling; the adaptive optimization capability of evolutionary computing effectively responds to dynamic changes and complex constraints in resource allocation, avoids the inefficiency and errors of manual decision-making, promotes intelligent manufacturing, improves production efficiency and reduces production costs, thereby achieving intelligent optimization of production plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0116] Figure 1 The figure is a flow chart of the method of the present invention.

[0117] Figure 2 The present invention is a flow chart of the steps of a batch process for producing pure white cotton yarn according to an example of the present invention.

[0118] Figure 3 This is a resource allocation scheme according to an example of the present invention. DETAILED DESCRIPTION

[0119] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.

[0120] The present invention provides a multi-order scheduling method for a spinning enterprise based on a thinking tree structure, comprising:

[0121] Data collection and preprocessing: Collect relevant data and perform cleaning and preprocessing;

[0122] Constructing a scheduling mind tree: Using the mind tree method to construct a scheduling mind tree, converting production demand into specific production tasks, and decomposing and clarifying the specific tasks and processes of each production line layer by layer;

[0123] Resource allocation optimization: Optimize resource allocation through evolutionary computing to ensure the best utilization of equipment, manpower and raw materials;

[0124] Formulate production scheduling plan: According to production tasks and resource allocation, determine the production tasks for each time period layer by layer, and formulate a preliminary production scheduling plan;

[0125] Real-time monitoring and adjustment: Monitor the progress and resource usage of each process in real time, and adjust tasks and resource allocation in a timely manner based on monitoring data to ensure the smooth implementation of the production plan.

[0126] The present invention also provides a multi-order scheduling device for a spinning enterprise based on a thinking tree structure, comprising:

[0127] Data collection and preprocessing module, which collects and cleans multi-source data including production workshops, order systems, and industry standards;

[0128] Construct a scheduling thinking tree module, with the spinning production demand of the current month as the root node, the first layer for different spinning types, the second layer for different production batches, the third layer for different processes, and the fourth layer for different production lines.

[0129] Resource allocation optimization module, which uses evolutionary computing methods to dynamically allocate and optimize resources including manpower, equipment, and materials to ensure maximum resource utilization;

[0130] Formulate a production scheduling module, comprehensively consider the delivery date, production capacity and resource constraints of each order, and generate the optimal production scheduling plan;

[0131] Production task and resource monitoring module, real-time tracking of the progress of production tasks and resource usage status;

[0132] The real-time monitoring and adjustment module, based on the production task and resource monitoring module, continuously monitors the actual situation of the production site and makes timely adjustments to cope with emergencies and changing production needs.

[0133] The following is a specific implementation process of the present invention by way of example.

[0134] like Figure 1As shown, this example provides a spinning scheduling method and device based on a textile industry big model, a mind tree and evolutionary computing, which aims to solve the complexity, uncertainty, resource optimization, efficiency improvement and cost control problems in the scheduling of spinning enterprises. The method includes the following steps: data collection and preprocessing: collect relevant data and perform cleaning and preprocessing to ensure the accuracy and completeness of the data; build a scheduling mind tree: use the mind tree method to build a scheduling mind tree, convert production requirements into specific production tasks, and decompose and clarify the specific tasks and processes of each production line layer by layer. The mind tree method can decompose complex scheduling problems into multiple manageable sub-problems, and form an efficient scheduling plan through layer-by-layer analysis and decision-making; resource allocation optimization: optimize resource allocation through evolutionary computing to ensure the best use of equipment, manpower and raw materials; formulate a production scheduling plan: according to production tasks and resource allocation, determine the production tasks of each time period layer by layer, and formulate a preliminary production scheduling plan; real-time monitoring and adjustment: use the production management system to monitor the progress and resource usage of each process in real time, and adjust the task and resource allocation in time according to the monitoring data to ensure the smooth implementation of the production plan. This method makes full use of the hierarchical decomposition and clear decision-making paths of the thinking tree to improve the transparency and operability of scheduling; the adaptive optimization capability of evolutionary computing effectively responds to dynamic changes and complex constraints in resource allocation, avoids the inefficiency and errors of manual decision-making, promotes intelligent manufacturing, improves production efficiency and reduces production costs, thereby achieving intelligent optimization of production plans.

[0135] 1. Data collection and preprocessing, including but not limited to the following data categories: equipment data, such as the number and type of spinning machines (ring spinning, rotor spinning, air-jet spinning, etc.), the number of spindles of each machine, the maximum operating speed and actual operating speed, equipment efficiency and utilization rate, and equipment maintenance and maintenance plan; raw material data, including raw material type (such as cotton, chemical fiber, etc.) and its proportion, inventory, quality parameters (fiber length, strength, fineness, etc.), supplier information and delivery cycle; product data, covering yarn varieties (pure cotton yarn, blended yarn, etc.), specifications (such as count, twist, etc.), quality requirements and finished product inventory; order data, including customer order information (quantity, specifications, delivery date, etc.), order priority and historical order data; production process data, such as the production capacity and output of each process, process Data on inter-sequence connection time, yarn change time and production line balance; human resource data, including the number of operators, skill levels, shift arrangements and personnel efficiency data; energy consumption data, such as energy consumption of each device and peak and valley electricity price information; quality control data, including online detection data, quality abnormality records, rework rate and scrap rate; logistics data, such as transportation time, storage capacity and location information of raw materials and finished products; cost data, covering raw material costs, labor costs and energy costs; environmental data, including workshop temperature and humidity and air quality parameters; production plan history data, such as past scheduling plans, plan execution and deviation analysis; enterprise management data, such as enterprise resource planning (ERP) system data and key performance indicators (KPI); compliance data, such as industry standards and regulatory requirements and environmental protection indicators. Pre-process the collected data to ensure the accuracy, completeness and consistency of the data, thereby improving the efficiency of data utilization.

[0136] 2. The above data is extracted through data mining, machine learning, natural language processing and other technologies, and the textile industry big model is fine-tuned in the form of JSON questions and answers to construct a company-level spinning-specific big model that understands the enterprise. By using the mind tree framework, it aims to guide the company-level spinning-specific big model of the enterprise to solve complex scheduling problems through multi-path exploration and heuristic evaluation. First, the root node is set to the production demand for spinning in the current month. The production demand is broken down into a series of smaller and more manageable sub-problems. This step is similar to breaking down a big problem into multiple small problems that are easy to solve.

[0137] 3. The first layer of the mind tree is to decompose the root node "production demand of spinning in the current month" into multiple sub-nodes according to different product categories. Each sub-node represents a specific product category. These sub-nodes include but are not limited to "cotton yarn production demand", "blended yarn production demand", "dyed yarn production demand" and "special yarn production demand". The "cotton yarn production demand" sub-node covers the production demand of all pure cotton yarns, which is one of the main products of spinning mills and is suitable for all kinds of clothing and home textiles. The "blended yarn production demand" sub-node involves the production demand of multi-fiber blended yarns. These yarns are usually blended from a variety of fibers such as cotton, polyester, wool, etc., and have the advantages of various fibers, such as improving the strength, elasticity and durability of the yarn. The "dyed yarn production demand" sub-node covers the demand for dyed spinning products. Dyed yarn is dyed during the spinning process and has a rich color selection. It is widely used in fashion clothing, home textiles and decorations. The "special yarn production demand" sub-node targets the demand for yarns with special functions or uses, such as fire-resistant yarn, antibacterial yarn, antistatic yarn and conductive yarn. These specialty yarns have important applications in high-performance and professional use areas, such as medical textiles, the electronics industry and safety protection equipment.

[0138] 4. Each product category sub-node is further decomposed into different production batches to enable more refined production management and scheduling. For example, under the "cotton yarn production demand" sub-node, it is decomposed into multiple specific production batches according to different specifications and yarn count requirements. These production batches may include "32-count cotton yarn production batch", "40-count cotton yarn production batch", "pure white cotton yarn production batch", etc. Each production batch represents a yarn production task with specific specifications and characteristics, ensuring that customer needs and special requirements of orders can be accurately met during the production process.

[0139] 5. For each production batch, continue to decompose it into specific process steps, with each process step as a child node. For example, "pure white cotton yarn production batch" can be decomposed into a series of detailed process steps, including: raw cotton / chemical fiber blending, cotton opening and cleaning, carding, combing / pre-combining, mixed sliver, roving, spun yarn and winding. By decomposing each production batch into specific process steps, every link in the production process can be carefully managed to ensure that each step is performed in accordance with standard operations. For specific process step diagrams, see Figure 2 .

[0140] 6. For each process step, assign it to each production line to ensure that the tasks of each production line are clear. Resource allocation includes three aspects: equipment resources, human resources and raw material resources. In terms of equipment resource allocation, production equipment is reasonably allocated to ensure that each production task has appropriate equipment support; in terms of human resource allocation, the best match is made according to the skills and experience of the workers to ensure that each process has appropriate workers to operate; in terms of raw material resource allocation, ensure that the raw materials required for the production tasks are sufficient to ensure the continuity and efficiency of production. Through refined resource allocation, the utilization rate and production efficiency of the production line can be improved, while ensuring product quality. Figure 3 To allocate resources.

[0141] 7. In order to optimize these resource allocations and production line scheduling, evolutionary computing methods are used. For complex multi-order scheduling problems, evolutionary computing is an efficient and flexible optimization method that can handle multiple variables and constraints. By simulating the natural evolution process, it continuously optimizes the scheduling plan and ultimately finds the optimal or near-optimal solution that meets various constraints. For example, for genetic algorithms, each production task on each production line can be represented as a gene, and a complete scheduling plan is represented as a chromosome. In this representation, the genetic algorithm optimizes and iterates the scheduling plan through genetic operations such as selection, crossover, and mutation. The fitness function is used to evaluate the quality of each chromosome, taking into account factors such as production efficiency, equipment utilization, human resource matching, and raw material supply. Specifically, the production task allocation and scheduling plan are represented by chromosomes, which are composed of a series of genes. The specific gene encoding method is as follows:

[0142] Each gene i Represents the resource allocation of a task, including equipment R equip 、Human Resources human 、Raw materials material and its start time T start and end time T end :

[0143] g i =(R equip ,R human ,R material ,T start ,T end )

[0144] A chromosome C contains the resource allocation and scheduling information of all tasks:

[0145] C=[g1,g2,…,g n ]

[0146] Where n is the total number of tasks. The encoding method needs to ensure that the chromosome can meet resource constraints and task dependencies.

[0147] The initial population is the starting point of the genetic algorithm optimization, which consists of a set of randomly generated chromosomes. Each chromosome represents a feasible scheduling solution. The initialization process must ensure that: a) all tasks meet the production order constraints (such as task dependencies). b) the total amount of allocated resources does not exceed the available resources. The population size N is usually 50-200, adjusted according to the complexity of the problem.

[0148] The fitness function F(C) is used to evaluate the quality of each chromosome (scheduling scheme). It is defined based on production goals and constraints in the form of:

[0149]

[0150] Where T total : Total processing time, defined as the latest end time of all tasks:

[0151] T total =max(T end,1 ,T end,2 ,…,T end,,n )

[0152] The goal is to minimize.

[0153] U resources : Resource utilization, calculated as the ratio of actual resource usage to available resources, taking into account equipment, manpower, and raw materials:

[0154]

[0155] The goal is to maximize.

[0156] P violations : Constraint violation penalty term, used to penalize solutions that do not meet the constraints:

[0157] P violations =λ1P resource +λ2P time

[0158] in:

[0159] ·P resource : Total amount of resource overallocation.

[0160] ·P time : Number of violations of task dependency order.

[0161] ·λ1,λ2: penalty coefficients.

[0162] The weights w1, w2, and w3 are adjusted according to actual needs to balance the importance of different objectives.

[0163] The selection operation selects chromosomes to enter the next generation based on the fitness value, and the selection probability of excellent chromosomes is higher. Common selection methods include:

[0164] Roulette method: The selection probability is allocated according to the fitness value ratio.

[0165] Tournament method: Randomly select multiple chromosomes to compete and choose the one with higher fitness.

[0166] The chromosomes selected and retained constitute the intermediate population P t .

[0167] The crossover operation simulates the recombination of genetic information by combining some genes of two chromosomes to generate new offspring chromosomes.

[0168] Common methods include:

[0169] Single-point crossover: A crossover point is randomly selected and the genes of the two chromosomes behind the crossover point are exchanged.

[0170] Multi-point crossover: Select multiple crossover points and alternately exchange chromosome segments.

[0171] For example, for two chromosomes:

[0172] C1=[g1,g2,g3,g4],C2=[g5,g6,g7,g8]

[0173] After crossover at the second gene, it generates:

[0174] C′1=[g1,g2,g7,g8],C′2=[g5,g6,g3,g4]

[0175] Crossover probability P c Usually 0.6-0.9 is taken.

[0176] The mutation operation randomly changes the gene values ​​in the chromosome to increase the diversity of the population and prevent the algorithm from falling into the local optimum. Common methods include:

[0177] Randomly adjust the task start time T start .

[0178] ·Change the resource allocation of tasks (such as changing R equip or R human .

[0179] For example, chromosome C = [g1, g2, ..., g n ] before the gene g2 mutates:

[0180] g2=(R equip =1,R human =2,R material =3,Tstart =5,T end =10)

[0181] After mutation:

[0182] g′2=(R equip =2,R human =1,R material =4,T start =6,T end =11)

[0183] Mutation probability R m Usually smaller, ranging from 0.01 to 0.1.

[0184] The genetic algorithm terminates when any of the following conditions are met:

[0185] ·Achieve the required algebra G max .

[0186] The population fitness converges, that is, the maximum fitness change is less than the threshold ∈.

[0187] After termination, the chromosome with the highest fitness value is selected from the population as the optimal scheduling solution C*, and its corresponding task allocation and scheduling information is output.

[0188] In addition to genetic algorithms, particle swarm algorithms are suitable for complex resource allocation and production line scheduling problems. For particle swarm algorithms, each particle represents a resource allocation and production line scheduling plan. The position of the particle represents the current scheduling plan, including equipment resource allocation, human resource allocation, and raw material resource allocation. By adjusting the speed and position of the particle, the PSO algorithm searches for the optimal solution in the solution space. The fitness function is also used to evaluate the pros and cons of each solution to ensure that each iteration moves in a better direction. By using these evolutionary computing methods, resource allocation and production line scheduling can be continuously optimized in a complex multi-order environment to maximize production efficiency and resource utilization while meeting customer needs and quality requirements. Specifically, in PSO, each particle represents a resource allocation and production scheduling plan, and its position and speed are defined as:

[0189] Particle position x i (t): represents the current scheduling plan, including resource allocation (equipment, manpower, raw materials) and time arrangement for each task:

[0190] x i (t) = [x i,1 (t),x i,2 (t),…,x i,n (t)]

[0191] Among them, x i,j(t) represents the allocation status of particle i on the jth task, for example:

[0192] x i,j (t)=(R equip ,R human ,R material ,T start ,T end )

[0193] Particle velocity vi(t): defined as the search direction and step size of the particle in the solution space:

[0194] v i (t) = [v i,1 (t),v i,2 (t),…,v i,n (t)]

[0195] The initial population of PSO consists of M particles, and the initial position x of each particle is i (t) and speed v i (t) Random generation, the following conditions must be met:

[0196] Feasibility: The scheduling scheme corresponding to the initial position must satisfy resource constraints and task dependency constraints.

[0197] Diversity: The particles in the population are widely distributed so as to cover the solution space.

[0198] After initializing the population, the fitness value of each particle is calculated for subsequent iterative optimization.

[0199] The core of PSO lies in the speed and position update of each generation of particles. The speed update formula of particles is:

[0200] v i (t+1)=ωv i (t)+c1r1[p i -x i (t)]+c2r2[gx i (t)]

[0201] ω: Inertia weight, which controls the degree to which the particle inherits the current velocity and affects the balance between global and local searches.

[0202] c1, c2: learning factors, representing the weights of the particle learning from its own optimal position and the global optimal position respectively.

[0203] r1, r2: Random numbers in the range [0, 1] used to introduce search randomness.

[0204] ·p i : The individual optimal position of particle i in the historical iteration.

[0205] g: global optimal position, the best position among all particles.

[0206] The particle's position is updated according to the velocity:

[0207] x i (t+1)=x i (t)+v i (t+1)

[0208] In order to ensure the feasibility of resource allocation, the updated x i (t+1) needs to be constrained and corrected:

[0209] Resource constraint correction: Ensure that the total amount of allocated resources R used No more than the total available resources R avail :

[0210] R used ≤R avail

[0211] Time dependency correction: adjust the start time T of task j start,j Ensure the completion time of its dependent task k:

[0212] T start,j ≥T end,k

[0213] Fitness function F(x i (t)) is used to evaluate the quality of the particle's current position. Its goal is similar to that of the genetic algorithm, taking into account the following goals:

[0214]

[0215] ·T total : Total processing time, the goal is to minimize it.

[0216] ·U resources : Resource utilization, the goal is to maximize.

[0217] ·P violations : Constraint violation penalty term, the goal is to minimize it.

[0218] The selection of weights w1, w2, and w3 can be adjusted according to actual needs to balance production goals.

[0219] The selection of algorithm parameters affects the performance of the algorithm to a certain extent. The inertia weight ω determines the exploration ability of the particle search, and a linear decreasing strategy is often used:

[0220]

[0221] Initial value ω max Usually 0.9 is taken, and the final value ωmin Usually 0.4 is taken.

[0222] Learning factors c1, c2: usually c1 = c2 = 2 to balance individual learning and group collaboration.

[0223] Population size M: M is usually 20-100, adjusted according to the complexity of the problem.

[0224] The process of PSO is as follows:

[0225] Randomly generate the initial positions and velocities of M particles to satisfy basic constraints.

[0226] Calculate the initial fitness value of each particle and initialize the individual optimal position p i and the global optimal position g.

[0227] Iterative optimization: For each generation t, perform the following steps:

[0228] Speed ​​update: Calculate the speed v of each particle according to the speed update formula i (t+1).

[0229] Position update: Calculate the new position x of each particle according to the position update formula i (t+1), and perform constraint correction.

[0230] Fitness calculation: Evaluate the new fitness value F(x i (t+1)).

[0231] Update individual and global optimal positions:

[0232] If F(x i (t+1))>F(p i ), update p i =x i (t+1).

[0233] If F(p i )>F(g), update g=p i .

[0234] · Reach the predetermined maximum number of iterations T max Or the global optimal fitness value changes less than the threshold.

[0235] Output the global optimal position g as the optimal scheduling solution.

[0236] 8. In general, for different evolutionary computing methods, fitness functions are needed to evaluate the pros and cons of each individual. The fitness function is an important tool to measure the pros and cons of scheduling schemes. Its evaluation indicators include total processing time, delivery time, resource utilization, etc. When formulating the fitness function, it is necessary to consider the production constraints at the same time. These constraints include the capacity of the production line, the skills of the workers, the availability of equipment, etc. When optimizing task allocation, it is necessary to ensure that these conditions are met. By executing different evolutionary algorithms, optimization can be performed based on the production tasks and resource allocation. Specifically, different evolutionary algorithms will generate multiple production scheduling schemes, which ensure the balance of production line load and maximize production efficiency under the evaluation of the fitness function. Finally, the optimal scheduling scheme is selected to formulate a preliminary production scheduling plan. In this process, the preliminary production scheduling scheme will be input into the enterprise-level spinning-specific large model. The enterprise-level spinning-specific large model is based on the mind tree framework, which can deeply learn and analyze the details of the production scheduling scheme. Through this process, the enterprise-level spinning-specific large model can become familiar with and understand the details of the production scheduling plan under the thinking tree framework, thereby improving its ability to handle future production scheduling tasks.

[0237] 9. Based on the mind tree framework, the enterprise-level spinning-specific big model can ask questions about each node to deeply analyze and optimize the production scheduling plan. In this way, the big model can refine each production step, discover potential problems and provide corresponding solutions. The specific process of this node questioning is as follows:

[0238] 1) Production resource assessment

[0239] Question: Please describe the current available production resources, including raw material inventory, equipment status, and staffing.

[0240] Purpose of analysis: Ensure that all production resources are sufficient and in optimal condition to prevent bottlenecks or resource shortages in the production process.

[0241] 2) Production capacity analysis

[0242] Question content: Please evaluate the maximum and average capacity of the current production line and analyze the production rhythm of each process.

[0243] Purpose of analysis: To understand the actual capacity of the production line, ensure that the production plan is in line with the load capacity of the production line, and avoid overload or idle resources.

[0244] 3) Worker Skill Matching

[0245] Question: Please list the skills of workers and explain how to best match them to production tasks.

[0246] Purpose of analysis: Improve production efficiency and product quality by rationally allocating worker tasks and avoid production delays or quality problems caused by skill mismatch.

[0247] 4) Equipment maintenance plan

[0248] Question: Please provide an equipment maintenance plan, including scheduled maintenance and emergency repair strategies.

[0249] Purpose of analysis: Ensure that the equipment operates in the best condition, reduce downtime caused by equipment failure, and improve overall production efficiency.

[0250] 5) Scheduling plan evaluation

[0251] Question content: Please describe a specific scheduling scheme and evaluate it based on the fitness function, pointing out its advantages and disadvantages.

[0252] Purpose of analysis: Evaluate the scheduling plan, identify potential improvement points, and make optimization suggestions to achieve higher production efficiency and resource utilization.

[0253] 6) Solution optimization and adjustment

[0254] Question content: Based on the above evaluation results, please propose an optimized scheduling plan and explain its optimization strategy and expected results.

[0255] Purpose of analysis: To optimize the scheduling plan to ensure that the plan is more in line with actual production needs and to improve the feasibility and efficiency of the production plan.

[0256] 7) Risk assessment and management

[0257] Question content: Please evaluate the possible risks in the current production plan and propose corresponding risk management strategies.

[0258] Purpose of analysis: To ensure smooth implementation of production plans and reduce production interruptions or delays caused by unexpected situations by identifying and managing potential risks.

[0259] By asking questions and analyzing each node, the enterprise-level spinning-specific large model can fully understand all aspects of the production scheduling plan, thereby formulating a more scientific and reasonable production plan and improving the efficiency and benefits of the entire production process.

[0260] A sample of the evaluation input text input into the enterprise-level textile macro model is shown in Table 1 below.

[0261] Table 1

[0262]

[0263]

[0264]

[0265] The present invention constructs a scheduling thinking tree: innovatively constructs a scheduling thinking tree structure with "production demand for the current month" as the root node, the first layer is different types of spinning, the second layer is different production batches, the third layer is different processes, and the fourth layer is different production lines. This hierarchical structure effectively organizes and manages complex production information, and improves the operability and flexibility of scheduling. Evolutionary computing optimizes resource allocation: using evolutionary computing methods, dynamically allocates and optimizes production resources such as manpower, equipment, and materials, maximizes resource utilization, avoids resource waste and production bottlenecks, and significantly improves production efficiency. Intelligent production scheduling plan formulation: comprehensively considers the delivery date, production capacity and resource constraints of each order, generates the optimal production scheduling plan, ensures the rationality and executability of the production plan, and effectively solves the problem of low efficiency of traditional scheduling methods. Real-time monitoring and adjustment mechanism: through real-time monitoring of the actual situation of the production site, provides a timely adjustment mechanism to respond to emergencies and changing production needs, ensures that the production plan is synchronized with actual production, and improves the flexibility and adaptability of production. The present invention greatly improves the scheduling efficiency and production management level of spinning enterprises through intelligent and automated means, ensures on-time delivery of orders, optimizes resource allocation, and enhances the overall competitiveness of enterprises.

[0266] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions do not exceed the scope of the technical solution of the present invention, belong to the protection scope of the present invention.

Claims

1. A multi-order scheduling method for a spinning enterprise based on a mind tree structure, characterized in that: include: Data collection and preprocessing: Collect relevant data and perform cleaning and preprocessing; Constructing a scheduling mind tree: Using the mind tree method to construct a scheduling mind tree, converting production demand into specific production tasks, and decomposing and clarifying the specific tasks and processes of each production line layer by layer; Resource allocation optimization: Optimize resource allocation through evolutionary computing to ensure the best utilization of equipment, manpower and raw materials; Formulate production scheduling plan: According to production tasks and resource allocation, determine the production tasks for each time period layer by layer, and formulate a preliminary production scheduling plan; Real-time monitoring and adjustment: Monitor the progress and resource usage of each process in real time, and adjust tasks and resource allocation in a timely manner based on monitoring data to ensure the smooth implementation of the production plan.

2. A multi-order scheduling method for a spinning enterprise based on a mind tree structure according to claim 1, characterized in that: The data collection and preprocessing are specifically implemented as follows: The collection includes equipment data, raw material data, including raw material type, product data, order data, production process data, human resources data, energy consumption data, quality control data, logistics data, cost data, environmental data, production plan history data, enterprise management data, and compliance data; Clean and preprocess the collected data.

3. The multi-order scheduling method for a spinning enterprise based on a mind tree structure according to claim 1 is characterized in that: Equipment data includes the number and type of spinning machines, the number of spindles of each machine, the maximum operating speed and actual operating speed, equipment efficiency and utilization, and equipment maintenance and servicing plans; raw material data includes raw material types and their proportions, inventory, quality parameters, supplier information and delivery cycles; product data covers yarn varieties, specifications, quality requirements and finished product inventory; order data includes customer order information, order priority and historical order data; production process data includes the production capacity and output of each process, the connection time between processes, the yarn change time and production line balance data; human resource data includes the number of operators, skill levels, shift arrangements and personnel efficiency data; energy consumption data includes the energy consumption of each equipment and peak and valley electricity price information; quality control data includes online detection data, quality abnormality records, rework rate and scrap rate; logistics data includes the transportation time, storage capacity and storage location information of raw materials and finished products; cost data covers raw material costs, labor costs and energy costs; environmental data includes workshop temperature and humidity and air quality parameters; Historical production plan data, including past scheduling plans, plan execution status and deviation analysis; enterprise management data, including enterprise resource planning system data and key performance indicators; compliance data, including industry standards and regulatory requirements and environmental indicators.

4. The method for multi-order scheduling of a spinning enterprise based on a mind tree structure according to claim 1, characterized in that: The preprocessed data is extracted through data mining, machine learning, and natural language processing technologies, and the textile industry big model is fine-tuned in the form of question-answer pairs in JSON format to construct the company-level spinning-specific big model for the enterprise; by using the mind tree framework, a scheduling mind tree is built to guide the company-level spinning-specific big model for the enterprise to solve complex scheduling problems.

5. A multi-order scheduling method for a spinning enterprise based on a mind tree structure according to claim 1 or 4, characterized in that: The scheduling thinking tree is constructed with the spinning production demand of the current month as the root node. The first layer is different spinning types, the second layer is different production batches, the third layer is different processes, and the fourth layer is a hierarchical structure of different production lines. The specific construction method of the scheduling thinking tree is as follows: Decompose the spinning production demand of the month into multiple sub-nodes; each sub-node represents a specific product category; these sub-nodes include cotton yarn production demand, blended yarn production demand, dyed yarn production demand and special yarn production demand; Each product category sub-node is further broken down into different production batches; For each production batch, continue to decompose it into specific process steps, and each process step is used as a child node; Each process step is assigned to each production line to ensure that the task of each production line is clear.

6. The method for multi-order scheduling of a spinning enterprise based on a mind tree structure according to claim 1, characterized in that: Genetic algorithm is used to optimize resource allocation through evolutionary computing. The specific implementation is as follows: The scheduling scheme is represented by a chromosome, which is composed of a series of genes; the gene encoding method is as follows: Each gene i Represents the resource allocation of a task, including equipment R equip 、Human Resources human 、Raw materials material and its start time T start and end time T end : g i =(R equip ,R human ,R material ,T start ,T end ) A chromosome C contains the resource allocation and scheduling information of all tasks: C=[g1,g2,…,g n ] Where n is the total number of tasks; the encoding method needs to ensure that the chromosome can meet resource constraints and task dependencies; The initial population is the starting point of genetic algorithm optimization, which is composed of a group of randomly generated chromosomes. Each chromosome represents a feasible scheduling solution. The initialization process needs to ensure that: a) all tasks meet the production order constraints; b) the total amount of allocated resources does not exceed the available resources; the population size N is adjusted according to the complexity of the problem; The fitness function F(C) is used to evaluate the quality of each chromosome, i.e., the scheduling scheme. It is defined based on production objectives and constraints in the form of: Where T total : Total processing time, defined as the latest end time of all tasks: T total =max(T end,1 ,T end,2 ,…,T end,,n ) The goal is to minimize; U resources : Resource utilization, calculated as the ratio of actual resource usage to available resources, taking into account equipment, manpower, and raw materials: The goal is to maximize; P violations : Constraint violation penalty term, used to penalize solutions that do not meet the constraints: P violations =λ1P resource +λ2P time Where: P resource is the total amount of resource over-allocation; is the number of violations of the task dependency order; λ1 and λ2 are penalty coefficients; The weights w1, w2, and w3 are adjusted according to actual needs to balance the importance of different goals; The selection operation selects chromosomes to enter the next generation according to the fitness value, and the selection probability of excellent chromosomes is higher; the selection operation methods include: roulette method, tournament method, crossover operation, and mutation operation; The genetic algorithm terminates when any of the following conditions are met: 1) Reach the required algebra G max ; 2) The fitness of the population converges, that is, the maximum fitness change is less than the threshold ∈; After termination, the chromosome with the highest fitness value is selected from the population as the optimal scheduling solution C*, and its corresponding task allocation and scheduling information is output.

7. A method for scheduling multiple orders for a spinning enterprise based on a mind tree structure according to claim 6, characterized in that: The specific implementation of each selection operation method is as follows: 1) Roulette method: Allocate selection probability according to the fitness value ratio; 2) Tournament method: Randomly select multiple chromosomes to compete, select the one with higher fitness; select the retained chromosomes to form the intermediate population P t ; 3) The crossover operation generates new offspring chromosomes by combining some genes of two chromosomes, simulating the recombination of genetic information; including: 3.1) Single-point crossover: randomly select a crossover point and exchange the genes of the two chromosomes at the rear section of the crossover point; 3.2) Multi-point crossover: select multiple crossover points and alternately exchange chromosome segments; For two chromosomes: C1=[g1,g2,g3,g4],C2=[g5,g6,g7,g8] After crossover at the second gene, it generates: C′1=[g1,g2,g7,g8],C2=[g5,g6,g3,g4] Crossover probability P c Take 0.6-0.9; 4) The mutation operation randomly changes the gene values ​​in the chromosome to increase the diversity of the population and prevent the algorithm from falling into the local optimum; including: 4.1) Randomly adjust the task start time T start ; 4.2) Change the resource allocation of the task, including changing R equip or R human ; Chromosome C = [g1, g2, …, g n ] before the gene g2 mutates: g2=(R equip =1,R human =2,R material =3,T start =5,T end =10) After mutation: g′2=(R equip =2,R human =1,F material =4,T start =6,T end =11) Mutation probability R m Take 0.01-0.

1.

8. The method for multi-order scheduling of a spinning enterprise based on a mind tree structure according to claim 1, characterized in that: The particle swarm algorithm PSO is used to optimize resource allocation through evolutionary computing. The specific implementation is as follows: A scheduling scheme for resource allocation of a task, including equipment R equip 、Human Resources human 、Raw materials material and its start time T start and end time T end , each particle represents a schedule, and its position and velocity are defined as: Particle position x i (t): represents the current scheduling plan, including the resource allocation and time arrangement of each task: x i (t)=[x i,1 (t),x i,2 (t),…,x i,n (t)] Among them, x i,j (t) represents the allocation status of particle i on the jth task: x i,j (t)=(R equip ,R human ,R material ,T start ,T end ) Particle velocity vi(t): defined as the search direction and step size of the particle in the solution space: v i (t)=[v i,1 (t),v i,2 (t),…,v i,n (t)] The initial population of PSO consists of M particles, and the initial position x of each particle is i (t) and speed v i (t) Random generation, the following conditions must be met: 1) Feasibility: The scheduling scheme corresponding to the initial position must satisfy resource constraints and task dependency constraints; 2) Diversity: The particles in the population are widely distributed to cover the solution space; After initializing the population, the fitness value of each particle is calculated for subsequent iterative optimization; The core of PSO lies in the speed and position update of each generation of particles. The speed update formula of particles is: v i (t+1)=ωv i (t)+c1r1[p i -x i (t)]+c2r2[g-x i (t)] Among them, ω is the inertia weight, which controls the degree of inheritance of the particle to the current velocity and affects the balance between global and local searches; c1 and c2 are learning factors, which represent the weights of the particle learning from its own optimal position and the global optimal position respectively; r1 and r2 are random numbers in [0,1], which are used to introduce search randomness; p i is the individual optimal position of particle i in the historical iteration; g is the global optimal position, the optimal position of all particles; The particle's position is updated according to the velocity: x i (t+1)=x i (t)+v i (t+1) In order to ensure the feasibility of resource allocation, the updated x i (t+1) needs to be constrained and corrected: Resource constraint correction: Ensure that the total amount of allocated resources R used No more than the total available resources R avail : R used ≤R avail Time dependency correction: adjust the start time T of task j start,j Ensure the completion time of its dependent task k: T start,j ≥T end,k Fitness function F(x i (t)) is used to evaluate the quality of the particle's current position. Its goal is similar to that of the genetic algorithm, taking into account the following goals: Among them, T total is the total processing time, the goal is to minimize; U resources is the resource utilization, the goal is to maximize; P violations To constrain the violation penalty term, the goal is to minimize it; the selection of weights w1, w2, and w3 is adjusted according to actual needs to balance the production goals; The specific process of PSO is as follows: 1) Randomly generate the initial positions and velocities of M particles to meet basic constraints. 2) Calculate the initial fitness value of each particle and initialize the individual optimal position p i and the global optimal position g. 3) Iterative optimization: For each generation t, perform the following steps: 4) Speed ​​update: Calculate the speed v of each particle according to the speed update formula i (t+1); 5) Position update: Calculate the new position x of each particle according to the position update formula i (t+1), and perform constraint correction; 6) Fitness calculation: Evaluate the new fitness value F(x) of each particle i (t+1)); 7) Update individual and global optimal positions: If F(x i (t+1))>F(p i ), update p i =x i (t+1); If F(p i )>F(g), update g=p i ; 8) Reach the predetermined maximum number of iterations T max Or the change of the global optimal fitness value is less than the threshold, and the global optimal position g is output as the optimal scheduling solution.

9. A method for scheduling multiple orders for a spinning enterprise based on a mind tree structure according to claim 8, characterized in that: Inertia weight ω: determines the exploration ability of particle search, using a linear decreasing strategy: Initial value ω max Take 0.9, the final value ω min Take 0.4; Learning factors c1, c2: c1 = c2 = 2 to balance individual learning and group collaboration; Population size M: M is 20-100 and is adjusted according to the complexity of the problem.

10. A multi-order scheduling device for a spinning enterprise based on a mind tree structure, characterized in that: include: Data collection and preprocessing module, which collects and cleans multi-source data including production workshops, order systems, and industry standards; Construct a scheduling thinking tree module, with the spinning production demand of the current month as the root node, the first layer for different spinning types, the second layer for different production batches, the third layer for different processes, and the fourth layer for different production lines. Resource allocation optimization module, which uses evolutionary computing methods to dynamically allocate and optimize resources including manpower, equipment, and materials to ensure maximum resource utilization; Formulate a production scheduling module, comprehensively consider the delivery date, production capacity and resource constraints of each order, and generate the optimal production scheduling plan; Production task and resource monitoring module, real-time tracking of the progress of production tasks and resource usage status; The real-time monitoring and adjustment module is based on the production task and resource monitoring module. It continuously monitors the actual situation of the production site and makes timely adjustments to cope with emergencies and changing production needs.

Images