Textile equipment dispatching management and optimization system of textile factory

Abstract

The invention relates to the technical field of textile production scheduling and resource allocation planning, in particular to a textile equipment scheduling management and optimization system of a textile factory, which comprises a data perception and integration module, a scheduling optimization decision module, a plan execution and equipment control module and a closed-loop feedback and self-learning module. The data sensing module collects and fuses order data, equipment operation state data and production environment data in real time, and a unified real-time data view is generated through cleaning and alignment processing; the scheduling module runs a mixed integer programming dynamic model, and minimizes the comprehensive cost and synchronously optimizes the equipment utilization rate and energy consumption in combination with rolling horizon optimization under the condition of meeting the process constraints of order delivery time limit and process dependency matrix representation; the plan execution module analyzes the scheduling instruction into an equipment executable instruction, drives equipment operation and collects execution deviation; and the closed loop module triggers rescheduling when the deviation exceeds the limit or the order is plugged. The scheduling accuracy and adaptability are improved, the cost is reduced, and efficient and stable production is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of textile production scheduling and resource allocation planning technology, specifically a textile equipment scheduling management and optimization system for textile factories. Background Technology

[0002] Current textile factories are facing production trends such as multi-variety, small-batch orders, continuously shortening delivery cycles, and increasing energy consumption control requirements. This has significantly increased the demands on the accuracy, dynamic adaptability, and overall efficiency of equipment scheduling. However, existing textile equipment scheduling and management systems generally remain at a traditional semi-automated stage, lacking full-process digital integration and intelligent optimization capabilities. They exhibit significant shortcomings in data support, decision-making models, process control, execution coordination, and self-learning mechanisms, making it difficult to adapt to complex and ever-changing production needs. Specific technical problems are as follows: The scheduling decision model lacks adaptability. It mostly uses static planning models, which cannot dynamically respond to changes in equipment status and order adjustments during production. It has not constructed a comprehensive cost function that includes equipment idling energy consumption, production preparation energy consumption and order delay penalties. It has also not included equipment utilization rate and unit output energy consumption in the synchronous optimization objectives. In addition, it lacks a rolling time-domain optimization strategy for periodic re-solution, making it difficult to achieve efficient scheduling under the premise of meeting order delivery deadlines and production process path constraints. The lack of standardized control over production process paths means that the logical sequence of core processes such as dyeing, weaving, and finishing relies on manual experience. Without a unified process dependency matrix for standardized representation and mandatory constraints, the order of processes is easily reversed or the connections are chaotic, leading to product quality defects and production rework, and increasing production costs and time losses.

[0003] Therefore, a textile equipment scheduling management and optimization system for textile factories is proposed to address the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide a textile equipment scheduling management and optimization system for textile factories to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A textile equipment scheduling management and optimization system for a textile factory, comprising: The data sensing and integration module is used to acquire and integrate multi-source heterogeneous data from the production environment of the textile factory in real time. The multi-source heterogeneous data includes order data from the manufacturing execution system, textile equipment operating status data from the sensor network, and production environment data from the environmental monitoring unit. The textile equipment operating status data includes equipment start-up and shutdown, operating speed, instantaneous power, and fault codes. The scheduling optimization decision module, connected to the data perception and integration module, is used to run a dynamic scheduling model based on mixed integer programming and periodically re-solve the dynamic scheduling model using a rolling time-domain optimization strategy. The objective function of the dynamic scheduling model is constructed as a comprehensive cost function that minimizes the energy consumption of equipment idling, energy consumption of production preparation, and order delay penalties, while satisfying all order delivery deadlines and production process path constraints, and simultaneously optimizing the comprehensive utilization rate of equipment and energy consumption per unit output value. Production process path constraints are represented by a process dependency matrix to ensure the logical sequence of dyeing, weaving, and finishing processes in any production order. The planning execution and equipment control module is connected to the scheduling optimization decision module. It is used to parse the obtained scheduling instructions into executable instructions specific to each textile equipment controller, and drive the corresponding physical textile equipment to start, stop, adjust speed or switch working modes. At the same time, it collects the instruction execution status and plan execution deviation data in real time. The closed-loop feedback and self-learning module is connected to the planning execution and equipment control module and the scheduling optimization decision module, respectively. The closed-loop feedback and self-learning module is configured to immediately trigger the scheduling optimization decision module to perform online rescheduling when the planning execution deviation data exceeds a preset threshold or when new order insertion data is received. At the same time, based on the comparison dataset of historical scheduling schemes and actual execution results, the stochastic gradient descent method is used to iteratively optimize the weight parameters related to energy consumption and time estimation in the dynamic scheduling model with the goal of minimizing the long-term average comprehensive cost.

[0006] As a preferred solution, the data sensing and integration module acquires and integrates multi-source heterogeneous data from the textile factory production environment in real time, including: Through a distributed data acquisition interface, order data from the manufacturing execution system, textile equipment operating status data from the sensor network, and production environment data from the environmental monitoring unit are collected in parallel to form a multi-source heterogeneous data stream. Data cleaning and timestamp alignment are performed on multi-source heterogeneous data streams. Data cleaning includes removing abnormally changing data points from sensors and completing missing data segments caused by communication interruptions. Timestamp alignment maps various types of data with different acquisition periods to the factory's standard timeline. Multi-source data that has been timestamp aligned is associated and fused according to equipment units and order batches to generate a real-time data view of the production site with a consistent data structure. The data association process is based on the physical topology of equipment and the process route of orders to establish a mapping relationship between operating status data, environmental data and order production batches. The real-time data view of the production site is standardized and encapsulated, and output to the scheduling optimization decision module as the input data source for the dynamic scheduling model.

[0007] As a preferred solution, the scheduling optimization decision module is used to run a dynamic scheduling model based on mixed-integer programming, and periodically re-solves the dynamic scheduling model using a rolling time-domain optimization strategy. The objective function of the dynamic scheduling model is constructed as minimizing the comprehensive cost function consisting of equipment idling energy consumption, production preparation energy consumption, and order delay penalties, while satisfying all order delivery deadlines and production process path constraints. This simultaneously optimizes equipment utilization and energy consumption per unit of output, including: Receive real-time data views of the production site from the data perception and integration module as the input data source for the dynamic scheduling model; Based on real-time data views of the production site, a dynamic scheduling model is constructed with order processes as decision-making units. The decision variables of the dynamic scheduling model include the start-stop status, operating speed level, and order process allocation sequence of each textile equipment. The constraints include order delivery deadline constraints, equipment capacity constraints, and production process path constraints represented by the process dependency matrix. The rolling time-domain optimization strategy decomposes the solution process of the dynamic scheduling model into multiple consecutive time windows. Within each time window, the dynamic scheduling model is solved based on the real-time data view of the current production site, and a scheduling scheme within the current time window is generated. The scheduling scheme within the current time window is used as the initial solution. Combined with newly arrived order data and equipment status data, the dynamic scheduling model is periodically re-solved to adjust the scheduling scheme for subsequent time windows. The obtained scheduling scheme is output to the planning execution and equipment control module as a scheduling instruction to drive the textile equipment.

[0008] As a preferred approach, production process path constraints are represented by a process dependency matrix to ensure the logical sequence of dyeing, weaving, and finishing processes in any production order, including: Construct a process dependency matrix, where the rows and columns of the process dependency matrix correspond to the process types of the textile factory, and the matrix elements represent the sequential dependencies between process types. The process types include at least dyeing, weaving, and finishing. Based on the process dependency matrix, a process sequence constraint set is generated for each production order. The process sequence constraint set defines the logical order of the processes required for the order. The process sequence constraint set is integrated into the constraints of the dynamic scheduling model. By checking whether the process allocation sequence satisfies the process sequence constraint set, it is ensured that the order process strictly follows the production process path in the scheduling scheme. During the rolling time-domain optimization process, the process sequence constraint set is dynamically updated based on real-time production data to adapt to changes in order processes or equipment status, thereby maintaining the feasibility and optimality of the scheduling scheme.

[0009] As a preferred embodiment, the planning execution and equipment control module parses the solved scheduling instructions into executable instructions specific to each textile equipment controller, and drives the corresponding physical textile equipment to start / stop, adjust speed, or switch working modes. Simultaneously, it collects real-time data on instruction execution status and plan execution deviations, including: Receive scheduling instructions from the scheduling optimization decision module. The scheduling instructions include the order process allocation sequence, equipment start and stop time points, and running speed level information. The scheduling instructions are parsed and translated to generate an executable instruction sequence that matches the communication protocol of each textile equipment controller. The parsing and translation process includes converting abstract scheduling logic into specific control commands. The executable instruction sequence is sent to the corresponding textile equipment controller to drive the physical textile equipment to perform start-stop, speed adjustment or working mode switching operations. Real-time monitoring of feedback signals from textile equipment controllers, collection of command execution status data, including actual equipment start-up and shutdown times, operating speed values, and current working mode; The execution status data of the instruction is compared and analyzed with the expected status of the scheduling instruction to calculate the planned execution deviation data, which includes time deviation, speed deviation and mode deviation. The plan execution deviation data is sent to the closed-loop feedback and self-learning module to trigger online rescheduling or model parameter optimization.

[0010] As a preferred solution, when the planned execution deviation data exceeds a preset threshold or new order insertion data is received, the scheduling optimization decision module is immediately triggered to perform online rescheduling, including: It receives real-time data on plan execution deviations and newly arrived insert data from the plan execution and equipment control module; The plan execution deviation data is analyzed and processed to extract the time deviation, speed deviation and mode deviation components; The time deviation, velocity deviation, and mode deviation components are compared with preset time deviation thresholds, velocity deviation thresholds, and mode deviation thresholds, respectively. If any deviation component exceeds its corresponding preset threshold, a rescheduling request is generated. When a new order insertion is received, a rescheduling request is generated directly. The generated rescheduling requests are evaluated for priority, and rescheduling priorities are assigned based on the degree of deviation exceeding the limit or the urgency of the inserted data. An online rescheduling trigger instruction is generated based on the rescheduling priority, and then sent to the scheduling optimization decision module. In response to the online rescheduling trigger command, the scheduling optimization decision module interrupts the current scheduling calculation process and restarts the solution of the dynamic scheduling model based on the latest real-time data view of the production site and the inserted order data to generate an updated scheduling scheme.

[0011] As a preferred approach, the closed-loop feedback and self-learning module, based on a comparison dataset of historical scheduling schemes and actual execution results, aims to minimize the long-term average comprehensive cost. It iteratively optimizes the weight parameters related to energy consumption and time estimation in the dynamic scheduling model using stochastic gradient descent, including: Collect historical scheduling schemes and corresponding actual execution results data to form a historical scheduling-execution comparison dataset; Based on the historical scheduling-execution comparison dataset, the difference between the actual comprehensive cost and the predicted comprehensive cost of each scheduling scheme is calculated to generate a cost difference dataset; Based on the cost variance dataset, with the goal of minimizing the long-term average comprehensive cost, the gradient direction of the weight parameters in the dynamic scheduling model is calculated using the stochastic gradient descent method. The weight parameters in the dynamic scheduling model are adjusted based on the gradient direction to generate an updated set of weight parameters. The updated set of weight parameters is applied to the dynamic scheduling model to optimize the accuracy and energy efficiency of subsequent scheduling decisions.

[0012] As can be seen from the technical solution provided by the present invention above, the textile equipment scheduling management and optimization system for textile factories provided by the present invention has the following beneficial effects: The data sensing and integration module collects order data from the manufacturing execution system, equipment operation status data from the sensor network, and production environment data from the environmental monitoring unit in parallel. After cleaning and removing abnormal and abrupt data and filling in missing segments, the data is then aligned and uniformly mapped to the factory's standard timeline using timestamps. Finally, the data is fused together according to equipment and order batch to generate a real-time data view with a consistent structure, providing an accurate, comprehensive, and time-synchronized data foundation for scheduling decisions and avoiding data deviations from interfering with decision-making. The scheduling optimization decision module constructs a mixed integer programming dynamic scheduling model with the goal of minimizing equipment idle energy consumption, production preparation energy consumption, and order delay penalties. It combines a rolling time-domain optimization strategy to periodically resolve the problem, while strictly ensuring the process sequence of dyeing, weaving, and finishing through the process dependency matrix. This model can not only meet order delivery and process constraints, but also reduce equipment idle waste and ineffective energy consumption, and avoid order delay losses, thus simultaneously improving the overall utilization rate of equipment and reducing energy consumption per unit output. The planning execution and equipment control module parses the scheduling instructions into executable instructions that are compatible with the protocols of each equipment controller, driving the equipment to accurately execute start-stop speed regulation mode switching operations; at the same time, it collects the actual start-stop time, running speed and working mode of the equipment in real time, compares them with the expected state to calculate the deviation of time, speed and mode, promptly detects and feeds back execution deviations, avoids production disruptions due to instruction incompatibility or execution deviations, and ensures the smooth implementation of scheduling plans from decision-making to execution. The closed-loop feedback and self-learning module receives execution deviations and new order insertion data in real time. When the deviation exceeds the threshold or a new order is inserted, a rescheduling request is immediately generated and the scheduling optimization decision module is triggered to resolve the problem according to priority. This allows for rapid response to production emergencies and avoids order delays or production stoppages. At the same time, based on the comparison data between historical scheduling schemes and actual execution results, the model uses stochastic gradient descent to iteratively optimize the weight parameters related to energy consumption and time estimation in the model. This allows the model to continuously adapt to actual production, and the accuracy and energy efficiency of subsequent scheduling decisions are gradually improved. The system operates with full automation, automatically processing multi-source data through data perception and integration, automatically generating solutions for scheduling optimization decisions, automatically driving equipment and monitoring its status through planned execution control, and automatically optimizing model parameters through closed-loop self-learning. This significantly reduces the workload of manually collecting data and manually scheduling and adjusting parameters. At the same time, it reduces problems such as energy waste and order delays caused by human decision-making errors, and helps factories reduce labor costs and production loss costs in the long term. The historical scheduling scheme execution results deviation data accumulated during system operation form production data assets, which can be used to analyze equipment energy consumption patterns, order execution efficiency, process bottlenecks, etc.; provide data support for factory equipment upgrades, process improvements, and production process optimization, and promote the transformation of textile factories from traditional management models to digital and intelligent management models. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the overall structure of a textile equipment scheduling management and optimization system for a textile factory according to the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0015] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific embodiments.

[0016] like Figure 1 As shown, this embodiment of the invention provides a textile equipment scheduling management and optimization system for a textile factory, comprising: The data sensing and integration module is used to acquire and integrate multi-source heterogeneous data from the production environment of the textile factory in real time. The multi-source heterogeneous data includes order data from the manufacturing execution system, textile equipment operating status data from the sensor network, and production environment data from the environmental monitoring unit. The textile equipment operating status data includes equipment start-up and shutdown, operating speed, instantaneous power, and fault codes. The scheduling optimization decision module, connected to the data perception and integration module, is used to run a dynamic scheduling model based on mixed integer programming and periodically re-solve the dynamic scheduling model using a rolling time-domain optimization strategy. The objective function of the dynamic scheduling model is constructed as a comprehensive cost function that minimizes the energy consumption of equipment idling, energy consumption of production preparation, and order delay penalties, while satisfying all order delivery deadlines and production process path constraints, and simultaneously optimizing the comprehensive utilization rate of equipment and energy consumption per unit output value. Production process path constraints are represented by a process dependency matrix to ensure the logical sequence of dyeing, weaving, and finishing processes in any production order. The planning execution and equipment control module is connected to the scheduling optimization decision module. It is used to parse the obtained scheduling instructions into executable instructions specific to each textile equipment controller, and drive the corresponding physical textile equipment to start, stop, adjust speed or switch working modes. At the same time, it collects the instruction execution status and plan execution deviation data in real time. The closed-loop feedback and self-learning module is connected to the planning execution and equipment control module and the scheduling optimization decision module, respectively. The closed-loop feedback and self-learning module is configured to immediately trigger the scheduling optimization decision module to perform online rescheduling when the planning execution deviation data exceeds a preset threshold or when new order insertion data is received. At the same time, based on the comparison dataset of historical scheduling schemes and actual execution results, the stochastic gradient descent method is used to iteratively optimize the weight parameters related to energy consumption and time estimation in the dynamic scheduling model with the goal of minimizing the long-term average comprehensive cost.

[0017] In this embodiment, the data perception and integration module is the core data support for the textile equipment scheduling management and optimization system of the textile factory. It constructs a unified and standardized real-time data view of the production site by comprehensively collecting, accurately processing, deeply associating and standardizing various scattered data in the production environment, providing a high-quality and reliable input data source for the scheduling optimization decision module, and ensuring the accuracy and timeliness of subsequent scheduling model solving and decision instruction generation. The data perception and integration module is primarily responsible for the parallel collection of order data from the manufacturing execution system, textile equipment operating status data from the sensor network, and production environment data from the environmental monitoring unit. It cleans, completes, and timestamps the collected multi-source heterogeneous data streams, eliminating data noise and format differences. Then, it establishes data associations based on equipment units and order batches to achieve data fusion, ultimately generating a consistent real-time production site data view and completing standardized encapsulation. This continuously provides usable data to the scheduling optimization decision-making module, laying the data foundation for the efficient operation of the entire scheduling system. The data perception and integration module includes: Multi-source data parallel acquisition unit: The data collection scope covers key data dimensions of the entire production process of textile factories; order data includes core information such as order delivery deadlines, product specifications, and production quantities; textile equipment operation status data covers real-time parameters such as equipment start-up and shutdown status, operating speed, instantaneous power, and fault codes; and production environment data involves environmental indicators that affect the production process, such as temperature, humidity, and air pressure. A distributed data acquisition interface architecture is adopted, simultaneously connecting to three major data sources: the manufacturing execution system sensor network and the environmental monitoring unit. A parallel acquisition mechanism synchronously acquires various types of data, forming a multi-source heterogeneous data stream. This avoids data delays or loss caused by a single acquisition channel, ensuring the comprehensiveness and real-time nature of data acquisition. The distributed parallel acquisition is based on a multi-node collaborative acquisition architecture, distributing acquisition tasks to multiple distributed acquisition nodes. Each node is specifically responsible for connecting to one type of data source and independently completing data acquisition. Nodes maintain a consistent acquisition rhythm through a synchronization mechanism, avoiding data acquisition conflicts. This technology overcomes the performance limitations of a single acquisition node, while improving the efficiency and reliability of data acquisition, ensuring the synchronous acquisition of multi-source heterogeneous data and meeting the system's real-time data requirements. During the data acquisition process, the connection status and data transmission of each data source are monitored in real time. When a connection to a data source is interrupted, a retry mechanism is automatically initiated to attempt to re-establish the connection. At the same time, the interruption time and fault information are recorded to provide a reference for subsequent maintenance. Data cleaning and timestamp alignment unit: Data cleaning focuses on improving data quality. For anomalous data jumps during sensor network transmission, it identifies and removes data by determining whether the data change amplitude exceeds a reasonable range, preventing abnormal data from interfering with subsequent analysis. For data gaps caused by communication interruptions, it uses interpolation based on adjacent valid data to complete the data, ensuring the continuity of the data sequence. Data cleaning identifies anomalous data by setting reasonable threshold ranges; data values ​​exceeding the threshold range are identified as anomalous jumps and removed. Data completion technology uses interpolation to calculate reasonable data for the missing position based on valid data points before and after the missing data segment, ensuring the continuity of the data sequence. The combined use of these two techniques effectively eliminates noise and missing data, improves data quality, and provides a reliable foundation for subsequent data processing and fusion. Timestamp alignment addresses the issue of inconsistent data collection cycles from multiple sources. Due to differences in collection frequencies across data sources, the timestamps for various data types differ. By mapping all data uniformly to the factory's standard timeline, different types of data are synchronized in the time dimension, ensuring time consistency during subsequent data association. Timestamp alignment establishes a mapping relationship between the collection time of each data source and the standard time, using the factory's standard timeline as a benchmark. By calculating the deviation between the collection cycle of different data sources and the standard time, the timestamp of each data point is adjusted, ensuring uniformity across all data in the time dimension. This technology solves the problem of asynchronous time among multiple data sources, ensuring that data can be matched in the same time dimension during association and fusion, guaranteeing the accuracy of data association. After the cleaning and alignment operations are completed, the processed data is initially verified to check its integrity and rationality. If any data that does not meet the requirements is found, the cleaning or completion operations will be performed again until the data meets the subsequent processing standards. Data association and fusion unit: Data association establishes a mapping relationship based on equipment physical topology and order process routes. According to the actual layout and connection relationships of factory equipment, as well as the preset production process path for each order, equipment operating status data is bound to the corresponding equipment unit, and production environment data is associated with the equipment operating in that environment and the batches of orders received, achieving a precise correspondence between data and equipment orders. Data association fusion, based on preset mapping rules, establishes connections between different types of data through the two core association dimensions of equipment physical topology and order process routes. It aggregates and integrates relevant data scattered across various data sources, eliminating data redundancy and format differences to form a comprehensive data view reflecting the production site status. This technology can uncover the inherent relationships between data, maximizing data value and providing comprehensive data support for scheduling optimization decisions. Data fusion integrates multi-source data that has undergone alignment processing into a unified data structure; it summarizes and integrates the operating status data, environmental data, and order data of the same equipment unit or the same order batch, eliminates data redundancy, and forms a comprehensive data set that includes equipment status, order requirements, and environmental conditions, fully reflecting the real-time situation of the production site. During the fusion process, data priority rules are followed. When there are differences in the same type of data from different sources, the real-time data directly collected by the sensors is taken as the standard. This is combined with the fixed parameters of the order data and the dynamic changes of the environmental data to ensure the authenticity and validity of the fused data. Data standardization encapsulation unit: Establish unified data encapsulation specifications, clarify data field definitions, data type format requirements, and transmission protocols; perform structured processing on the integrated real-time production site data view, organize data fields according to specifications, unify data formats, and ensure that the data has a consistent structure and interface; The standardized data is packaged and encapsulated to form a data format that meets the input requirements of the scheduling optimization decision module. During the encapsulation process, the integrity of the data is ensured, key information is not omitted, and data transmission efficiency is optimized to facilitate the scheduling optimization decision module to receive and parse the data quickly. After encapsulation, a real-time monitoring mechanism for the data transmission channel is established to ensure that the data can be stably and continuously output to the scheduling optimization decision module. When data transmission is delayed or abnormal, it is promptly fed back to the system for processing. Standardized encapsulation technology, by establishing unified data specifications, performs structured processing and format conversion on the merged data; clarifies the meaning of data fields, data types, and transmission formats, giving the data a unified interface and structure, ensuring that the scheduling optimization decision module can quickly parse and use the data; and simultaneously optimizes data transmission efficiency during the encapsulation process, reducing data transmission losses and delays, and ensuring the stability and timeliness of data transmission.

[0018] In this embodiment, the scheduling optimization decision module is the core decision-making core of the textile equipment scheduling management and optimization system in the textile factory. Based on the real-time production site data output by the data perception and integration module, it constructs and solves a dynamic scheduling model based on mixed integer programming, and combines a rolling time-domain optimization strategy to achieve periodic rescheduling. Under the premise of strictly meeting the order delivery deadline and production process path constraints, it minimizes the comprehensive cost consisting of equipment idling energy consumption, production preparation energy consumption and order delay penalties, while simultaneously optimizing the comprehensive utilization rate of equipment and energy consumption per unit output value, providing a scientific, accurate and feasible scheduling solution for the plan execution and equipment control modules. The core functions of the scheduling optimization decision module revolve around the entire process of "data input - model building - optimization solution - solution output - dynamic adjustment". First, it continuously receives real-time data views from the production site, extracting key information such as order delivery requirements, equipment operating status, and production process parameters. Then, based on this information, it constructs a dynamic scheduling model with the order process as the core decision unit, clarifying the decision variables to be optimized and the constraints that must be met. Next, it breaks down long-cycle scheduling into multiple continuous time windows through a rolling time-domain optimization strategy, solving the model within each window to generate a phased scheduling solution. Simultaneously, it periodically re-solves the model based on newly arrived order data and equipment status change data, adjusting subsequent solutions. Furthermore, when it receives a re-scheduling trigger signal from the closed-loop feedback and self-learning module, it immediately interrupts the current process and re-solves the model, ensuring that the scheduling solution always adapts to dynamic changes in production. Finally, it outputs the validated scheduling solution to the planning execution and equipment control module to guide the actual operation of the equipment. The scheduling optimization decision module includes: Data receiving and preprocessing subunit: Establish a highly reliable data link with the data perception and integration module to receive real-time data views from the production site. This view covers basic order information (delivery deadline, product specifications, production quantity), real-time equipment status (start / stop status, current speed, load status, fault repair status), production environment parameters (temperature, humidity, air pressure), and process progress data, ensuring that the data covers all dimensions of production. The received data undergoes secondary verification and filtering; invalid data caused by transmission fluctuations is identified and eliminated through preset rationality rules (such as equipment speed not exceeding the maximum rated value and order delivery deadline not being earlier than the current time); at the same time, core fields directly related to scheduling decisions are filtered out, such as order process list, current equipment idle status, and inter-process dependencies, to remove redundant information and form a concise model input dataset; The filtered dataset is format-converted and standardized; the equipment operating speed is converted into discrete gear values ​​that the model can recognize (such as low speed 1, medium speed 2, high speed 3), the order delivery deadline is converted into a timestamp format based on the factory standard time, and the equipment load data is converted into a percentage form to ensure that the format and unit of all input data are fully matched with the parameter requirements of the dynamic scheduling model, and to avoid model solution anomalies caused by data format incompatibility. Dynamic scheduling model construction sub-unit: Define a complete system of decision variables; clarify three core decision variables: First, equipment start / stop status variables, with values ​​of running or stopped, used to control the working state of a single piece of equipment; second, equipment operating speed gear variables, setting 3-5 discrete gears according to the technical characteristics of the equipment (e.g., setting three gears of 500 rpm, 800 rpm, and 1200 rpm for a weaving machine), used to balance production efficiency and energy consumption; third, order process allocation sequence variables, clarifying which piece of equipment each process (dyeing, weaving, finishing) of each order is assigned to, and the execution order on that equipment, to ensure smooth process connection; A multi-dimensional set of constraints is constructed. The first type is the order delivery deadline constraint, which requires that the completion time of all processes in all orders must not be later than the preset delivery deadline. If an order has multiple processes, it must be ensured that the completion time of the previous process does not affect the timely completion of the subsequent process. The second type is the equipment capacity constraint, which limits the maximum number of processes that a single piece of equipment can handle at the same time (e.g., a single dyeing machine can handle the dyeing processes of a maximum of 2 orders at the same time) to avoid equipment overload. The third type is the production process path constraint, which is implemented through a process dependency matrix to ensure that any order must first complete the dyeing process, then execute the weaving process, and finally perform the finishing process, strictly following the process logic of textile production and eliminating product quality problems caused by process reversal. Design a multi-objective-oriented comprehensive cost objective function; quantify the core objective of scheduling optimization as minimizing the sum of "equipment idling energy consumption + production preparation energy consumption + order delay penalty"; where equipment idling energy consumption is calculated based on equipment standby time and unit standby power, production preparation energy consumption is calculated based on equipment switching order / process debugging time and unit debugging power, and order delay penalty is calculated based on overdue time and unit overdue penalty cost; at the same time, embed the equipment comprehensive utilization rate coefficient and unit output value energy consumption coefficient into the objective function. When the two coefficients reach the preset threshold, the model will automatically adjust the decision variables to achieve multi-objective synergistic optimization of "cost reduction", "efficiency improvement" and "energy consumption reduction"; Rolling time-domain optimization solution sub-unit: Reasonably divide the scheduling time window; based on the average cycle of textile production processes (such as dyeing process 4 hours, weaving process 6 hours, finishing process 2 hours), break down the entire production scheduling cycle into multiple continuous and partially overlapping time windows (usually set at 4-8 hours / window) to ensure that the scheduling plan in each window can be implemented in a short time, while avoiding the plan from becoming disconnected from actual production due to excessively long windows; The execution window model solution process is as follows: Starting from the initial production status of the current time window (such as the current process of each piece of equipment and the progress of completed orders), the order data and equipment status data in the window are input into the dynamic scheduling model, and the mixed integer programming solution algorithm is called to perform iterative calculations. Finally, the scheduling scheme that satisfies all constraints and has the lowest overall cost in the window is obtained. The scheme includes the process allocation sequence, start and stop time points, and running speed gear switching nodes for each piece of equipment. Implement a periodic re-solution mechanism; use the solution result of the current time window as the initial solution of the next time window. When entering the next window period, combine the new data sent by the data perception and integration module (such as new orders, equipment fault repair, and process progress updates) to adjust the input parameters of the dynamic scheduling model, restart the solution process, generate a scheduling scheme adapted to the new production state, and ensure that the scheme is updated in real time with the production dynamics. The system responds quickly to rescheduling trigger signals. When it receives a rescheduling instruction from the closed-loop feedback and self-learning module (such as exceeding the plan execution deviation limit or the arrival of an emergency order), it immediately interrupts the solution process in the current time window, prioritizes receiving the latest real-time production data and order information, quickly adjusts the model's constraints (such as the delivery deadline constraints for the new order) and input parameters, starts the emergency solution process, shortens the algorithm iteration time, and ensures that an updated scheduling scheme is generated within 10-15 minutes to respond promptly to production anomalies or new demands. Scheduling scheme output and verification subunit: Feasibility verification of the execution plan; multi-dimensional constraint checks are performed on the obtained scheduling plan: first, check whether the execution order of all orders meets the requirements of the process dependency matrix; second, check whether the load of each piece of equipment is within the capacity constraint range; third, check whether the expected completion time of all orders is not later than the delivery deadline; fourth, check whether the start-up and shutdown time and speed switching of the equipment meet the equipment technical specifications (such as avoiding frequent start-up and shutdown to avoid equipment damage); if a constraint conflict is found in the plan, immediately return to the rolling time domain optimization solution sub-unit, adjust the model parameters (such as equipment allocation priority) and solve again until the plan fully satisfies the constraints; The scheme is structured; the verified scheduling scheme is decomposed and reorganized to generate an independent list of execution instructions for each piece of equipment. The list includes order number, process type, execution start time, execution end time, running speed level, preparation time before process switching, etc., to ensure that the instructions are clear and can be executed directly. Establish a stable output interface; establish a communication connection with the planning execution and equipment control module through industrial Ethernet, transmit the structured scheduling instruction list with a standardized protocol (such as Modbus-TCP), and record the output time of the scheme, the scheme version number, the number of orders involved and the number of equipment, and store them in the module's historical database to provide data support for subsequent closed-loop feedback and parameter optimization of the self-learning module; further: Mixed-integer programming is the mathematical core of dynamic scheduling models. It defines discrete variables in scheduling decisions (such as equipment start / stop status and order process allocation sequence) as integer variables (taking discrete values ​​such as 0 or 1, 1 or 2, etc.), and defines continuously changing variables (such as equipment operating speed and energy consumption) as continuous variables, thus constructing a mathematical model that simultaneously adapts to discrete decision-making and continuous optimization requirements. In the solution process, this technique uses iterative search algorithms such as branch and bound and cutting plane method to find the optimal solution that minimizes the comprehensive cost objective function in the solution set that satisfies all constraints. This ensures the feasibility of the scheduling scheme and achieves the optimal balance between cost and efficiency. Rolling time-domain optimization technology is designed to address the dynamic and uncertain nature of textile production. Its core logic is "short-cycle solution + rolling update." Due to uncertainties in textile production such as order insertion, equipment failure, and energy consumption fluctuations, long-term fixed scheduling schemes are prone to failure due to changes in actual production. This technology breaks down long-cycle (e.g., 24-hour) scheduling into multiple short-time windows (e.g., 4-hour windows). Each window only optimizes the production arrangement for the current stage, without needing to predict long-term unknown states. At the same time, the optimization result of the previous window serves as the initial solution for the next window, reducing the computational load in the solution process and improving the efficiency of scheme updates. This approach can quickly respond to sudden changes in production, avoid scheme failures caused by long-term prediction deviations, and significantly improve the flexibility and adaptability of scheduling decisions. The process dependency matrix is ​​a key tool for ensuring that production process path constraints are implemented. The rows and columns of the matrix correspond to the core process types in textile production (dyeing, weaving, finishing). Elements in the matrix take values ​​of 0 or 1. When an element is 1, it indicates that the process corresponding to the row must be executed before the process corresponding to the column (e.g., if the elements in the dyeing row and weaving column are 1, it means the dyeing process must precede the weaving process). When an element is 0, it indicates that the processes corresponding to the row and column have no direct sequential dependency (e.g., if the elements in the dyeing row and finishing column are 0, but the elements in the weaving row and finishing column are 1, the order of dyeing and finishing still needs to be indirectly constrained through the weaving process). During the model solving process, the system automatically checks whether the order process allocation sequence conforms to the dependency relationship specified by the matrix. If there is a case of reversed processes, the solution will be directly excluded, ensuring that the production process of all orders strictly follows the process requirements. Multi-objective integrated cost optimization technology achieves a balance of scheduling objectives through "cost coefficient quantification + multi-objective collaboration." First, cost coefficients are set for equipment idling energy consumption, production preparation energy consumption, and order delay penalties (e.g., idling energy consumption coefficient of 0.8 yuan / kWh and delay penalty coefficient of 50 yuan / hour), and these three types of costs are uniformly converted into monetary units to achieve quantitative calculation of integrated costs. Second, equipment comprehensive utilization rate thresholds (e.g., 85%) and unit output energy consumption thresholds (e.g., 0.3 kWh / yuan) are embedded in the objective function. When the equipment comprehensive utilization rate in the model solution is lower than the threshold, the weight of equipment load allocation is automatically increased to improve utilization. When unit output energy consumption is higher than the threshold, the speed level of high-energy-consuming equipment is automatically reduced or processes are prioritized for low-energy-consuming equipment. Ultimately, multi-objective collaborative optimization is achieved, balancing economic efficiency and sustainability, with the goals of "lowest integrated cost," "equipment utilization rate meeting standards," and "unit output energy consumption meeting standards." The workflow of the scheduling optimization decision module is as follows: After the module starts, it automatically loads the preset core parameter configuration, including the iterative accuracy of the mixed integer programming algorithm (e.g., 10). -4 The parameters include rolling time window duration (e.g., 6 hours), cost coefficients for various energy consumptions, order delay penalty coefficients, equipment capacity parameters, and initial values ​​of the process dependency matrix. These parameters can be adjusted by factory technicians based on actual production conditions. Establish communication connections with the data perception and integration module, the plan execution and equipment control module, and the closed-loop feedback and self-learning module. Test the stability of data transmission and the accuracy of command interaction through interface debugging tools to ensure smooth data flow between modules. Receive the initial real-time data view of the production site output by the data perception and integration module, perform the first data preprocessing, filter out the initial order data, equipment initial status data, and started process progress data, and generate the initial input dataset for the model. Based on the initial input dataset, the decision variable system, constraint set and comprehensive cost objective function of the dynamic scheduling model are constructed, and the value range of each decision variable (e.g., equipment speed level is 1-3) and the judgment criteria of each constraint (e.g., equipment capacity is 2 processes / time windows) are clarified. Import the preset process dependency matrix, embed the production process path constraints into the model's constraint set, and supplement the order-specific process sequence details based on the initial order's process requirements (e.g., if some orders require an additional pre-dyeing process, add a pre-dyeing row to the matrix and adjust the corresponding element values). The constructed model is checked for completeness to ensure that no decision variables are missing, no logical conflicts exist in the constraints, and the objective function calculation logic is correct. If any problems are found, the system will automatically prompt and pause the construction process until the technical personnel have investigated and fixed them before restarting. Divide the first scheduling time window, with the current factory standard time as the starting point of the window, and determine the end point of the window according to the preset duration (e.g., 6 hours). Extract the data such as the order process to be executed, the status of available equipment, and the progress of executed processes within the window from the initial input dataset, and use them as the solution input for the window. The mixed integer programming algorithm is invoked to solve the dynamic scheduling model. During the iterative calculation, the convergence of the algorithm is monitored in real time. When the iteration accuracy reaches the preset requirement (such as the overall cost change of less than 1% in 5 consecutive iterations), the iteration is stopped and the scheduling scheme of the window is output. Perform a feasibility check on the output scheduling scheme. If the scheme meets all constraints, proceed to the scheme output stage. If there are constraint conflicts (such as an equipment being overloaded), adjust the process allocation priority of the equipment, restart the solution, and continue until a feasible scheme is obtained. The feasible scheduling schemes are structured to generate an independent list of execution instructions for each piece of equipment, and a summary table of scheduling schemes is generated (including the process allocation, estimated completion time and overall cost of all orders). The structured list of execution instructions is sent to the planning execution and equipment control module through the communication interface. At the same time, the summary table of scheduling schemes is stored in the historical database and pushed to the factory production management system for managers to view. Record detailed information about the solution output, including output time, solution version, number of equipment involved, number of orders involved, estimated comprehensive cost, estimated equipment utilization rate, and estimated energy consumption per unit output, to form a complete solution output log; When the first time window reaches the preset update node (usually half of the window duration, such as 3 hours), it automatically receives the latest real-time production site data sent by the data perception and integration module, including the progress of completed processes, the current status of equipment (such as whether a fault has occurred), and newly arrived order data (such as new regular orders). Using the execution result of the scheduling scheme in the first window as the initial solution, the new data is integrated into the solution input for the next time window. The parameters of the dynamic scheduling model are adjusted (such as updating equipment state variables and constraints of new orders). The rolling solution process is restarted to generate the scheduling scheme for the next window. Repeat the "rolling solution-solution verification-solution output" process to achieve periodic updates of the scheduling scheme and ensure that the scheme always remains consistent with the actual state of the production site; The system monitors the closed-loop feedback and rescheduling trigger signals sent by the self-learning module in real time. When a signal is received, the current periodic update process is immediately paused, and the system prioritizes receiving the planned execution deviation data (such as time deviation, speed deviation) or emergency order insertion data sent by the module. Based on deviation data or emergency order insertion data, adjust the constraints (such as delivery deadline constraints for emergency orders) and input parameters (such as correcting the energy consumption coefficient corresponding to the actual operating speed of the equipment) of the dynamic scheduling model, activate the emergency solution mode, and shorten the number of algorithm iterations (such as relaxing the iteration accuracy to 10). -3 ), quickly generate updated scheduling schemes; Perform a rapid feasibility check on the updated plan (focusing on the delivery deadline and equipment load of emergency orders). Once the check is passed, immediately output it to the plan execution and equipment control module, and record the reason for rescheduling, rescheduling duration, and plan adjustment content, and add it to the historical database. When the factory receives a production end instruction (such as the completion of the day's production tasks), the module stops the rolling solution and rescheduling response process, automatically summarizes the execution data of all scheduling schemes for the day (such as actual comprehensive cost, actual equipment utilization rate, and actual order delivery rate), and generates a daily scheduling summary report; then it closes the communication connection with each module, saves the current model parameter configuration and historical data, releases the computing resources occupied by the solution algorithm, and completes the module's operation.

[0019] In this embodiment, the plan execution and equipment control module is the execution center of the textile equipment scheduling management and optimization system in the textile factory. It takes over the scheduling scheme of the scheduling optimization decision module, transforms abstract decisions into specific operations that the equipment can execute, and captures the actual equipment operation status and plan deviations in real time, providing accurate data support for closed-loop feedback, ensuring that the scheduling scheme is implemented effectively and the production process is stable and controllable. The core of the planning execution and equipment control module is responsible for receiving scheduling instructions from the scheduling optimization decision module. It then parses and translates these instructions into specific commands adapted to the controllers of various textile equipment, driving the equipment to perform operations such as start-up, shutdown, speed adjustment, and switching of working modes. During equipment operation, the module monitors the execution status of instructions in real time, collects actual equipment operating data, compares and analyzes this data with the expected state of the scheduling instructions, accurately calculates deviations in dimensions such as time, speed, and mode, and finally synchronizes the deviation data to the closed-loop feedback and self-learning module. This provides a basis for online rescheduling or model parameter optimization, ensuring that the production process proceeds according to plan and has a dynamic adjustment capability. The planning execution and equipment control module includes: Dispatch instruction receiving unit: Establish a stable communication link with the scheduling optimization decision module to ensure the real-time performance and integrity of scheduling command transmission; It continuously receives scheduling instructions from the scheduling optimization decision module. The instructions include core information such as order process allocation sequence, equipment start and stop time points, and operating speed levels, covering the key parameters required for equipment operation. The received scheduling instructions are initially verified to check whether the instruction format is standardized and whether the information is complete. If any abnormality is found, feedback is sent to the scheduling optimization decision module to request resending or correction. Instruction parsing and translation unit: The communication protocol specifications of various textile equipment controllers were reviewed, and the format requirements of control commands and data interface standards for different equipment were clarified. The verified scheduling instructions are parsed in layers to extract key information such as order process allocation logic, equipment operation timing requirements, and speed gear parameters, and the abstract scheduling logic is broken down into specific equipment operation steps. According to the communication protocol of the corresponding equipment controller, the parsed scheduling information is translated into a sequence of directly executable instructions to ensure that the instructions can be seamlessly adapted to the equipment controller and avoid operation failure due to protocol incompatibility. Equipment drive control unit: A multi-channel command issuance mechanism was established, and dedicated communication channels were allocated for different types of textile equipment to ensure the targeted and efficient issuance of commands. The translated executable instruction sequence is accurately sent to the corresponding textile equipment controller to ensure that the instructions received by each piece of equipment are completely matched with the production tasks it undertakes. The system monitors the command issuance process in real time and confirms whether the equipment controller has successfully received the command. If the command transmission fails, the system automatically starts the retransmission mechanism until the command is successfully received. Based on the instructions, the physical textile equipment is driven to perform corresponding operations, including controlling the equipment to start or stop according to the start and stop time points, adjusting the equipment running rate according to the set running speed level, and switching the equipment working mode according to production needs, so as to ensure that the equipment operation is highly consistent with the scheduling instructions. Execution status acquisition unit: The feedback interface of the textile equipment controller is connected to capture the status signal after the equipment executes the command in real time, and establish the correspondence between equipment operation and status feedback. Collect actual equipment operation data, including actual start-up and stop time (the moment when the equipment actually starts or stops), actual operating speed value (the current operating rate of the equipment), and current working mode (the type of working mode that the equipment is executing), to ensure that the collected data can fully reflect the equipment's operating status. The collected status data is preprocessed in real time to remove invalid data generated during transmission, ensuring the authenticity and accuracy of the data and providing a reliable basis for subsequent deviation calculation; Deviation Calculation and Feedback Unit: Establish a comparison standard between the expected state of the dispatching instructions and the actual execution state of the equipment, and clarify the comparison rules for dimensions such as time, speed, and mode; The preprocessed actual execution status data is compared with the expected status in the scheduling instructions item by item, and the deviation data of each dimension is calculated respectively; the time deviation is the difference between the actual start and stop time of the equipment and the expected start and stop time; the speed deviation is the difference between the actual operating speed value of the equipment and the standard value corresponding to the expected operating speed level; and the mode deviation is the quantitative value of the degree of conformity between the current working mode of the equipment and the expected working mode. The calculated deviation data are summarized and organized to form a structured deviation dataset, ensuring that the data is clearly organized and easy for subsequent modules to use; Establish a real-time data transmission channel with the closed-loop feedback and self-learning module to send the deviation dataset to the module in a timely manner, providing data support for triggering online rescheduling or optimizing model parameters; further: The instruction parsing and adaptation technology is based on the communication protocols of various textile equipment controllers and constructs a standardized parsing and translation rule base. By identifying the core information in the scheduling instructions, it transforms the abstract scheduling logic into specific control commands that meet the requirements of the equipment controller according to the corresponding relationships in the rule base. This technology can automatically adapt to the protocol differences of different types of equipment and achieve accurate instruction conversion without manual intervention. It ensures that different models and functions of textile equipment can accurately understand and execute scheduling instructions, thus guaranteeing the universality and implementability of the scheduling scheme. The real-time status acquisition and synchronization technology adopts a high-frequency data acquisition mechanism, sets an acquisition cycle that matches the equipment's operating rhythm, and ensures that status changes during equipment operation can be captured in a timely manner. By establishing a dedicated communication link with the equipment controller's feedback interface, data transmission delay is reduced, and the acquired data is synchronized with the actual operating status of the equipment. At the same time, invalid information is filtered out through a data preprocessing mechanism to ensure the real-time performance and reliability of the acquired data, providing accurate data input for deviation calculation. The deviation quantification calculation technology establishes a multi-dimensional deviation calculation model based on preset comparison standards. For different dimensions such as time and speed modes, quantification calculation rules are set separately to transform the difference between the actual execution state and the expected state into specific numerical deviation data. This technology can comprehensively capture the deviation situation during equipment execution, quantify the degree of deviation, and provide clear numerical basis for subsequent judgment on whether to trigger rescheduling or optimize model parameters, ensuring the objectivity and practicality of the feedback data. The workflow of the plan execution and equipment control module is as follows: After the plan execution and equipment control module is started, it completes its own hardware and software self-checks, including communication interface detection, instruction parsing unit test, data acquisition module verification, etc., to ensure that each component is operating normally. Load the configuration information such as communication protocol parameters, instruction parsing rules, deviation calculation standards, etc. of each textile equipment controller, establish a communication connection with the closed-loop feedback and self-learning module of the scheduling optimization decision module, and complete the pre-operation preparation work. Initiate a command receiving and monitoring mechanism to continuously receive scheduling commands issued by the scheduling optimization decision module; The received scheduling instructions are format-verified and information integrity-checked to confirm that the instructions meet the module processing requirements; If the instruction verification passes, proceed to the next stage; if the verification fails, send an exception feedback to the scheduling optimization decision module and wait for the corrected instruction or the resent instruction. The verified scheduling instructions are parsed in layers to extract key information such as the order process allocation, equipment operation sequence, speed gear, etc. Based on the communication protocol of the corresponding equipment controller, the parsed information is translated into an executable instruction sequence; Perform compatibility testing on the translated instruction sequence to ensure that the instructions can be recognized and executed by the target equipment controller; The executable instruction sequence is sent to the corresponding textile equipment controller through a dedicated communication channel; After confirming that the equipment controller has successfully received the instruction, the equipment is triggered to perform the corresponding operation, including switching between start / stop speed regulation and working modes. Real-time monitoring of equipment operation process ensures that the equipment executes production tasks step by step according to instructions; Continuously capture feedback signals from the equipment controller and collect real-time status data such as actual start-up and shutdown time, operating speed, and current working mode of the equipment; The collected status data is preprocessed to remove invalid data and correct abnormal data to ensure data quality. The preprocessed actual state data is compared with the expected state in the scheduling instructions to calculate the time deviation, speed deviation, and mode deviation. Summarize the deviation data from each dimension to form a structured deviation dataset; The deviation dataset is sent to the closed-loop feedback and self-learning module to provide support for subsequent scheduling adjustments or model optimization. When the production task is completed or a stop instruction is received, the module stops receiving and parsing the drive control status acquisition and other operations; closes the communication connection with other modules, saves the instruction execution record status data deviation information, releases system resources, and completes the operation cleanup work.

[0020] In this embodiment, the closed-loop feedback and self-learning module is the "adaptive hub" of the textile equipment scheduling management and optimization system in the textile factory. On the one hand, it connects the planning execution and equipment control modules to capture real-time production deviations. On the other hand, it links with the scheduling optimization decision module to achieve dynamic adjustment. At the same time, it iteratively optimizes the scheduling model through historical data, so that the system can not only cope with sudden production changes, but also continuously improve scheduling accuracy and efficiency. The closed-loop feedback and self-learning module undertakes two main tasks: First, closed-loop feedback, which receives real-time data on plan execution deviations and new order insertions. When deviations exceed limits or new orders are inserted, it immediately triggers the scheduling optimization decision module for online rescheduling to ensure production does not deviate from the target. Second, self-learning optimization, based on comparison data between historical scheduling schemes and actual execution results, aims to minimize the long-term average comprehensive cost. Through algorithmic iteration, it adjusts the weight parameters related to energy consumption and time estimation in the scheduling model, allowing the model to continuously adapt to actual production and improve the accuracy and economy of subsequent scheduling decisions. The closed-loop feedback and self-learning module includes: Real-time feedback processing unit: The real-time feedback processing unit is the "response hub" for the module to deal with sudden changes in production. It focuses on capturing real-time data and triggering rescheduling, including: Data reception: Continuously receive plan execution deviation data from the plan execution and equipment control module, including time deviation, speed deviation, and mode deviation; at the same time, receive new arrival order data from the production management system, including order quantity, delivery deadline, and process requirements of the order. Data analysis: Decompose the plan execution deviation data to extract time deviation (the difference between the actual start-up and shutdown time of the equipment and the expected time), speed deviation (the difference between the actual operating speed of the equipment and the expected gear), and mode deviation (the quantified value of the consistency between the actual working mode of the equipment and the expected time); perform structured processing on the new order data to clarify the urgency of the order and the production requirements. Trigger judgment: Compare each deviation component with the preset time deviation threshold, speed deviation threshold, and mode deviation threshold. If any deviation component exceeds the corresponding threshold, a rescheduling request is immediately generated. When new order data is received, no threshold judgment is required, and a rescheduling request is generated directly. Priority assessment: Requests are assigned priority based on the degree of deviation or the urgency of the order. The more severe the deviation (e.g., a time deviation exceeding the threshold by 50% has a higher priority than one exceeding it by 10%), and the more urgent the delivery deadline (e.g., delivery within 24 hours has a higher priority than delivery within 72 hours), the higher the rescheduling priority. Command sending: Generate an online rescheduling trigger command based on priority, including the triggering reason (such as time deviation exceeding the limit or new order insertion) and priority level, and send it to the scheduling optimization decision module. At the same time, interrupt the current scheduling calculation process of the module and push it to re-solve based on the latest data. Historical data management unit: The historical data management unit is the "data foundation" for the module to achieve self-learning, and is responsible for collecting and organizing historical scheduling and execution data; Data collection: Periodically obtain historical scheduling plans from the scheduling optimization decision module, including order process allocation sequence, equipment start and stop time, operating speed level, and expected comprehensive cost; obtain the actual execution results of the corresponding plans from the plan execution and equipment control module, including actual comprehensive cost, actual execution time, and actual energy consumption data; Data association: Each historical scheduling plan is associated with its corresponding actual execution result by order number and production time period to ensure that every scheduling decision can find matching actual effect data, forming a "plan-result" correspondence; Data cleaning: Remove abnormal data caused by sudden equipment failure (not scheduling model issues) or human error, such as records of a dyeing machine experiencing a surge in energy consumption due to hardware damage, to avoid abnormal data interfering with subsequent optimization; Dataset Construction: The cleaned "Solution-Result" data is classified by order type (such as cotton spinning orders and chemical fiber orders) and equipment type (such as dyeing machines and weaving machines), and a historical scheduling-execution comparison dataset is constructed and stored in a dedicated database for easy retrieval and analysis later; Weight parameter optimization unit: The weight parameter optimization unit is the "core engine" of module self-learning, which adjusts the scheduling model parameters through algorithm iteration; Cost variance calculation: Extract the expected comprehensive cost and actual comprehensive cost of each scheme from the historical scheduling-execution comparison dataset, calculate the difference between the two, and generate a cost variance dataset; for example, the expected comprehensive cost of a certain scheduling scheme is 1,000 yuan, the actual cost is 1,200 yuan, and the cost variance is 200 yuan. Gradient direction calculation: With the goal of minimizing the long-term average comprehensive cost, the gradient direction of each weight parameter in the scheduling model is calculated using the stochastic gradient descent method. The gradient direction points to the parameter adjustment direction that can reduce cost differences. For example, when a certain energy consumption weight parameter is too low, causing the energy consumption estimate to be less than the actual value, the gradient direction will guide the parameter to increase. Parameter adjustment: Adjust the weight parameters related to energy consumption and time estimation in the scheduling model according to the gradient direction and the preset learning rate (control the adjustment range of parameters to avoid large fluctuations); for example, adjust the weight of energy consumption estimation for the dyeing process from 0.6 to 0.7 so that the model's subsequent estimation of dyeing energy consumption is closer to reality. Parameter validation: The adjusted set of weight parameters is validated to ensure that the parameter values ​​are within a reasonable range (e.g., the weight parameters are not negative and the sum does not exceed the preset upper limit). If the parameters exceed the range, they are corrected according to the reasonable boundary to ensure that the parameters can be used for the scheduling model. Optimization parameter application unit: The optimization parameter application unit is the "bridge" connecting the weight optimization and scheduling model, and is responsible for implementing the optimized parameters. Parameter synchronization: The updated set of weight parameters output by the weight parameter optimization unit is sent to the scheduling optimization decision module to request the update of the corresponding parameters in the dynamic scheduling model; Effect verification: After updating the scheduling model parameters, a test scheduling is triggered; the updated model is used to regenerate the scheduling plan for the currently incomplete orders, and the expected cost of the plan is compared with the actual cost of similar historical orders to verify the effect of parameter optimization; if the difference between the expected cost and the actual cost is reduced compared with before optimization, it indicates that the parameter application is effective. Secondary fine-tuning: If the test results show that the optimization effect does not meet expectations (e.g., the cost difference does not decrease), the test data will be fed back to the weight parameter optimization unit to restart the parameter adjustment process until the parameter optimization effect meets the preset standard (e.g., the cost difference rate is less than 5%). Log recording: Record the time of each parameter update, the parameter values ​​before and after the update, and the test results to form an optimization log, which is convenient for tracing the parameter adjustment process and analyzing optimization patterns. further: Online rescheduling is triggered based on a "real-time monitoring + priority grading" mechanism to ensure that the system responds to production changes in a timely and orderly manner. The module quickly identifies production anomalies by scanning plan execution deviation data and order insertion data in real time. Deviation exceeding the limit means that the current scheduling has deviated from the preset target, and new orders are new variables added to the production task. Both require urgent adjustment. At the same time, priority grading is introduced so that the system can prioritize the handling of situations with a greater impact on production. For example, urgent orders are given priority over minor speed deviations, avoiding the waste of resources on non-critical adjustments and ensuring production efficiency. Stochastic gradient descent optimization is the core of self-learning. Its principle is to find the parameter combination that minimizes the long-term average comprehensive cost through iteration. Based on historical cost differences, it adjusts the parameters along the gradient direction that reduces cost differences each time, and controls the adjustment range through the learning rate to avoid excessive parameter fluctuations. As the number of iterations increases, the weight parameters gradually approach the optimal value, making the scheduling model's estimation of energy consumption and time more and more accurate, thereby reducing the gap between "theoretical scheduling" and "actual production" and reducing the long-term comprehensive production cost. Historical data comparative analysis is the foundation of data-driven optimization. By comparing "solutions and results," it uncovers the shortcomings of the scheduling model. It links historical scheduling decisions with actual production results and discovers patterns of model deviation from the data. For example, if the estimated idling energy consumption of a certain type of equipment is consistently lower than the actual value, it indicates that the corresponding energy consumption weight parameter is too low. This analysis based on actual data makes parameter optimization no longer dependent on human experience, but has clear data support, ensuring that the self-learning process is scientific and reliable. The workflow of the closed-loop feedback and self-learning module includes: After the module starts, it loads preset configuration parameters, including the time threshold, speed threshold, and mode deviation threshold for planned execution deviation, the rescheduling priority evaluation criteria, the initial weight parameters of the dynamic scheduling model, and the historical data storage path. Establish communication connections with the planning execution and equipment control module, the scheduling optimization decision module, and the historical database; test the smoothness of data transmission; and ensure that data can be exchanged normally between the modules. The most recent set of weight parameters is retrieved from the historical database and synchronized to the weight parameter optimization unit as the initial optimization baseline; It receives real-time data on plan execution deviations from the plan execution and equipment control modules, as well as new order insertion data from the production management system. Analyze deviation data to extract time deviation, speed deviation, and pattern deviation; analyze new order data to determine the urgency of the order and production needs. Compare the deviation components with the preset threshold: if any deviation exceeds the limit, generate a rescheduling request; if a new order is inserted, generate a rescheduling request directly. The priority of the rescheduling request is evaluated, a trigger instruction is generated and sent to the scheduling optimization decision module to interrupt the current calculation and re-solve the scheduling scheme based on the latest data. Receive the new scheme generation results from the scheduling optimization decision module, record the trigger time, trigger reason, scheme adjustment content, and store them in the historical log; After the daily production is completed, all historical scheduling plans for the day are obtained from the scheduling optimization decision module, and the corresponding actual execution results are obtained from the plan execution and equipment control module. Link the "solution-result" data, calculate the cost difference for each solution, and clean up outlier data; The processed data is added to the historical scheduling-execution comparison dataset and stored in the database according to the classification rules. The historical schedule is retrieved from the database weekly to execute the comparison dataset, which is divided into multiple training batches according to the time series (e.g., one batch per day). For each training batch, calculate the average cost difference of that batch, and use stochastic gradient descent to calculate the gradient direction of the weight parameters; Adjust the weight parameters according to the gradient direction and learning rate to generate an updated parameter set, and complete one iteration. Repeat the iterative process until the long-term average cost difference stabilizes within the preset range, then stop the optimization and determine the final update parameter set. The final updated parameter set is sent to the scheduling optimization decision module to update the dynamic scheduling model parameters; Trigger test scheduling to compare the updated model-generated solution with historical data and verify the optimization effect; If the desired effect is achieved, record the parameter update information; if not, return to the weight parameter optimization stage for readjustment. When the system receives a production stop instruction (such as a monthly production stoppage for maintenance), the module stops real-time feedback processing and weight optimization operations, closes communication connections with other modules, backs up historical scheduling-execution comparison datasets and optimization logs, and releases system resources.

[0021] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A textile equipment scheduling management and optimization system for a textile factory, characterized in that: include: The data sensing and integration module is used to acquire and integrate multi-source heterogeneous data from the production environment of the textile factory in real time. The multi-source heterogeneous data includes order data from the manufacturing execution system, textile equipment operating status data from the sensor network, and production environment data from the environmental monitoring unit. The textile equipment operating status data includes equipment start-up and shutdown, operating speed, instantaneous power, and fault codes. The scheduling optimization decision module, connected to the data perception and integration module, is used to run a dynamic scheduling model based on mixed integer programming and to periodically resolve the dynamic scheduling model using a rolling time-domain optimization strategy. The objective function of the dynamic scheduling model is constructed as a comprehensive cost function that minimizes the energy consumption of equipment idling, energy consumption of production preparation, and penalty for order delays, while satisfying all order delivery deadlines and production process path constraints, and simultaneously optimizing the comprehensive utilization rate of equipment and energy consumption per unit output value. The production process path constraint is represented by a process dependency matrix to ensure the sequential order of dyeing, weaving, and finishing processes in the logical link for any production order. The planning execution and equipment control module is connected to the scheduling optimization decision module. It is used to parse the obtained scheduling instructions into executable instructions specific to each textile equipment controller, and drive the corresponding physical textile equipment to start, stop, adjust speed or switch working modes. At the same time, it collects the instruction execution status and plan execution deviation data in real time. The closed-loop feedback and self-learning module is connected to the planning execution and equipment control module and the scheduling optimization decision module, respectively. The closed-loop feedback and self-learning module is configured to immediately trigger the scheduling optimization decision module to perform online rescheduling when the planning execution deviation data exceeds a preset threshold or when new order insertion data is received. At the same time, based on the comparison dataset of historical scheduling schemes and actual execution results, the stochastic gradient descent method is used to iteratively optimize the weight parameters related to energy consumption and time estimation in the dynamic scheduling model with the goal of minimizing the long-term average comprehensive cost.

2. The textile equipment scheduling management and optimization system for a textile factory according to claim 1, characterized in that: The data sensing and integration module acquires and integrates multi-source heterogeneous data from the textile factory production environment in real time, including: Through a distributed data acquisition interface, order data from the manufacturing execution system, textile equipment operating status data from the sensor network, and production environment data from the environmental monitoring unit are collected in parallel to form a multi-source heterogeneous data stream. Data cleaning and timestamp alignment are performed on multi-source heterogeneous data streams. Data cleaning includes removing abnormally changing data points from sensors and completing missing data segments caused by communication interruptions. Timestamp alignment maps various types of data with different acquisition periods to the factory's standard timeline. Multi-source data that has been timestamp aligned is associated and fused according to equipment units and order batches to generate a real-time data view of the production site with a consistent data structure. The data association process is based on the physical topology of equipment and the process route of orders to establish a mapping relationship between operating status data, environmental data and order production batches. The real-time data view of the production site is standardized and encapsulated, and output to the scheduling optimization decision module as the input data source for the dynamic scheduling model.

3. The textile equipment scheduling management and optimization system for a textile factory according to claim 1, characterized in that: The scheduling optimization decision module is used to run a dynamic scheduling model based on mixed integer programming and to periodically re-solve the dynamic scheduling model using a rolling time-domain optimization strategy. The objective function of the dynamic scheduling model is constructed as minimizing the comprehensive cost function consisting of equipment idling energy consumption, production preparation energy consumption, and order delay penalties, while satisfying all order delivery deadlines and production process path constraints. This simultaneously optimizes equipment utilization and energy consumption per unit output, including: Receive real-time data views of the production site from the data perception and integration module as the input data source for the dynamic scheduling model; Based on real-time data views of the production site, a dynamic scheduling model is constructed with order processes as decision-making units. The decision variables of the dynamic scheduling model include the start-stop status, operating speed level, and order process allocation sequence of each textile equipment. The constraints include order delivery deadline constraints, equipment capacity constraints, and production process path constraints represented by the process dependency matrix. The rolling time-domain optimization strategy decomposes the solution process of the dynamic scheduling model into multiple consecutive time windows. Within each time window, the dynamic scheduling model is solved based on the real-time data view of the current production site, and a scheduling scheme within the current time window is generated. The scheduling scheme within the current time window is used as the initial solution. Combined with newly arrived order data and equipment status data, the dynamic scheduling model is periodically re-solved to adjust the scheduling scheme for subsequent time windows. The obtained scheduling scheme is output to the planning execution and equipment control module as a scheduling instruction to drive the textile equipment.

4. The textile equipment scheduling management and optimization system for a textile factory according to claim 1, characterized in that: The production process path constraints are represented by a process dependency matrix to ensure the logical sequence of dyeing, weaving, and finishing processes in any production order, including: Construct a process dependency matrix, where the rows and columns of the process dependency matrix correspond to the process types of the textile factory, and the matrix elements represent the sequential dependencies between process types. The process types include at least dyeing, weaving, and finishing. Based on the process dependency matrix, a process sequence constraint set is generated for each production order. The process sequence constraint set defines the logical order of the processes required for the order. The process sequence constraint set is integrated into the constraints of the dynamic scheduling model. By checking whether the process allocation sequence satisfies the process sequence constraint set, it is ensured that the order process strictly follows the production process path in the scheduling scheme. During the rolling time-domain optimization process, the process sequence constraint set is dynamically updated based on real-time production data to adapt to changes in order processes or equipment status, thereby maintaining the feasibility and optimality of the scheduling scheme.

5. The textile equipment scheduling management and optimization system for a textile factory according to claim 1, characterized in that: The plan execution and equipment control module is used to parse the obtained scheduling instructions into executable instructions specific to each textile equipment controller, and drive the corresponding physical textile equipment to start / stop, adjust speed, or switch working modes. Simultaneously, it collects real-time data on instruction execution status and plan execution deviation, including: Receive scheduling instructions from the scheduling optimization decision module, the scheduling instructions including the order process allocation sequence, equipment start and stop time points and operating speed gear information; The scheduling instructions are parsed and translated to generate an executable instruction sequence that matches the communication protocol of each textile equipment controller. The parsing and translation process includes converting abstract scheduling logic into specific control commands. The executable instruction sequence is sent to the corresponding textile equipment controller to drive the physical textile equipment to perform start-stop, speed adjustment or working mode switching operations. Real-time monitoring of feedback signals from the textile equipment controller, collection of command execution status data, including actual start-up and stop times, operating speed values, and current working mode; The execution status data of the instruction is compared and analyzed with the expected status of the scheduling instruction to calculate the planned execution deviation data, which includes time deviation, speed deviation and mode deviation. The plan execution deviation data is sent to the closed-loop feedback and self-learning module to trigger online rescheduling or model parameter optimization.

6. The textile equipment scheduling management and optimization system for a textile factory according to claim 1, characterized in that: When the planned execution deviation data exceeds the preset threshold or new order insertion data is received, the scheduling optimization decision module is immediately triggered to perform online rescheduling, including: It receives real-time data on plan execution deviations and newly arrived insert data from the plan execution and equipment control module; The plan execution deviation data is analyzed and processed to extract the time deviation, speed deviation and mode deviation components; The time deviation, velocity deviation, and mode deviation components are compared with preset time deviation thresholds, velocity deviation thresholds, and mode deviation thresholds, respectively. If any deviation component exceeds its corresponding preset threshold, a rescheduling request is generated. When a new order insertion is received, a rescheduling request is generated directly. The generated rescheduling requests are evaluated for priority, and rescheduling priorities are assigned based on the degree of deviation exceeding the limit or the urgency of the inserted data. An online rescheduling trigger instruction is generated based on the rescheduling priority, and then sent to the scheduling optimization decision module. In response to the online rescheduling trigger command, the scheduling optimization decision module interrupts the current scheduling calculation process and restarts the solution of the dynamic scheduling model based on the latest real-time data view of the production site and the inserted order data to generate an updated scheduling scheme.

7. The textile equipment scheduling management and optimization system for a textile factory according to claim 1, characterized in that: The closed-loop feedback and self-learning module, based on a comparison dataset of historical scheduling schemes and actual execution results, aims to minimize the long-term average comprehensive cost. It iteratively optimizes the weight parameters related to energy consumption and time estimation in the dynamic scheduling model using stochastic gradient descent, including: Collect historical scheduling schemes and corresponding actual execution results data to form a historical scheduling-execution comparison dataset; Based on the historical scheduling-execution comparison dataset, the difference between the actual comprehensive cost and the predicted comprehensive cost of each scheduling scheme is calculated to generate a cost difference dataset; Based on the cost variance dataset, with the goal of minimizing the long-term average comprehensive cost, the gradient direction of the weight parameters in the dynamic scheduling model is calculated using the stochastic gradient descent method. The weight parameters in the dynamic scheduling model are adjusted based on the gradient direction to generate an updated set of weight parameters. The updated set of weight parameters is applied to the dynamic scheduling model to optimize the accuracy and energy efficiency of subsequent scheduling decisions.

Images