Chemical fiber spinning production line virtual commissioning and process optimization system based on digital twinning

Abstract

The present application relates to a chemical fiber spinning production line virtual commissioning and process optimization system based on digital twinning, belonging to the technical field of chemical fiber spinning production. The system obtains equipment operation parameters through a data processing module and generates a set of operation parameters after preprocessing; a digital twinning module constructs a three-dimensional virtual model based on equipment three-dimensional feature data and process parameters, generates a virtual production environment, and establishes a two-way data synchronization mapping mechanism; a virtual commissioning module loads control programs based on the virtual production environment, simulates control instruction execution and the spinning production process, and generates simulation working condition data; a process optimization module generates a standardized data set based on historical production data and simulation working condition data, trains and constructs a mapping prediction model of process parameters and product performance indicators, performs global optimization based on the model and constraint conditions, and generates an optimal process parameter combination. The present application realizes the virtual-real linkage of the spinning production line through digital twinning technology, optimizing the commissioning and process optimization process.

Description

Technical Field

[0001] This invention belongs to the field of chemical fiber spinning production technology, specifically relating to a virtual debugging and process optimization system for chemical fiber spinning production lines based on digital twins. Background Technology

[0002] The chemical fiber spinning production process is characterized by its complexity, strong correlation of process parameters, and high requirements for equipment linkage. Its production quality and efficiency directly depend on the effectiveness of production line commissioning and the rationality of process parameters. Currently, the commissioning of chemical fiber spinning production lines mainly adopts on-site physical commissioning. This requires verifying the rationality of equipment linkage, control logic, and process parameters through shutdown and trial operation after the production line is installed. This method not only consumes a significant amount of production time and increases production costs, but also easily leads to equipment damage and product scrap due to improper parameters during commissioning.

[0003] In terms of process optimization, traditional methods mainly rely on the accumulated experience of technical personnel, adjusting process parameters through manual trial and error. This approach suffers from drawbacks such as low optimization efficiency, poor parameter matching, and the inability to achieve global optimization, making it difficult to adapt to the demands of large-scale and refined production. Furthermore, existing technologies lack an effective virtual-physical linkage mechanism, failing to map the operational status of the physical production line to the virtual space in real time, and also unable to directly guide process adjustments on the physical production line using virtual simulation results. This leads to a disconnect between virtual simulation and actual production, hindering the full realization of the advantages of simulation technology.

[0004] Digital twin technology, as an emerging technology integrating the Internet of Things, 3D modeling, and simulation analysis, enables real-time linkage between physical entities and virtual models, providing a new technical approach for production line debugging and optimization. However, there is currently no dedicated digital twin system for chemical fiber spinning production lines. Existing digital twin applications are mostly focused on single-device monitoring, failing to achieve full-process virtual debugging and intelligent process optimization, and thus failing to meet the actual needs of multi-device linkage and multi-parameter collaborative optimization in chemical fiber spinning production lines. Therefore, developing a virtual debugging and process optimization system for chemical fiber spinning production lines that can achieve virtual-physical mapping, simulation prediction, and closed-loop optimization is an urgent technical problem to be solved. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides a virtual debugging and process optimization system for a chemical fiber spinning production line based on digital twins. The virtual commissioning and process optimization system for chemical fiber spinning production lines based on digital twins includes a data processing module, a digital twin module, a virtual commissioning module, and a process optimization module. The data processing module acquires the equipment operating parameters of the chemical fiber spinning production line, and generates an operating parameter set after data preprocessing; The digital twin module constructs a three-dimensional virtual model that maps to the production line based on the equipment's three-dimensional feature data and process parameters, generating a virtual production environment; and establishes a two-way data synchronization mapping mechanism based on the virtual production environment and the set of operating parameters. The virtual debugging module, based on the virtual production environment, loads the production line control program, simulates the control instruction execution process, simulates the chemical fiber spinning production process, and generates simulated working condition data. The process optimization module generates a standardized dataset based on historical production data and the simulation operating condition data, and trains it to construct a mapping prediction model between spinning process parameters and product performance indicators. Using product performance indicators as constraints, it performs global optimization calculations based on the mapping prediction model to generate the optimal combination of process parameters.

[0006] Specifically, the process for setting the equipment operating parameters includes: IoT sensing elements are deployed at the monitoring points of the production equipment in the chemical fiber spinning production line to collect the operating parameters of the corresponding equipment in real time according to the preset collection frequency. The analog signals are converted into digital signals in real time. After each batch acquisition is completed, the acquired digital signals are initially sorted out, and the equipment number, acquisition point and acquisition timestamp corresponding to the operating parameters of each device are marked to generate the original parameter list.

[0007] Specifically, the data preprocessing process includes: The original parameter list is called up, the mean and standard deviation of each parameter are calculated, parameters that exceed the preset reasonable range are identified as outliers and removed, and the parameter type, collection time and corresponding device information of the outliers are recorded. All processed parameter data are mapped to preset parameter ranges, categorized by device type, and different parameters of the same device are sorted and organized according to parameter category to generate the operating parameter set.

[0008] Specifically, the construction process of the three-dimensional virtual model includes: Extract the equipment's 3D feature data from equipment design drawings and equipment operation and maintenance records, perform redundancy removal processing, and generate a 3D feature dataset of the equipment. Extract the process flow channel size requirements, equipment operating limit thresholds, and component motion range parameters from the process parameters as constraints. Using 3D modeling technology, based on the equipment's 3D feature dataset, independent 3D models of each core device are constructed sequentially. Based on the actual layout drawings of the physical production line, the installation position, spacing and connection method of each piece of equipment in the virtual space are determined, and each independent 3D model is virtually assembled to complete the construction of a 3D virtual model mapped to the production line.

[0009] Specifically, the specific processes of the virtual production environment include: Call the three-dimensional virtual model to import the complete process flow rules of the chemical fiber spinning production line, as well as the equipment operation logic and environmental parameters. The three-dimensional virtual model is rendered to set the background, lighting, and appearance features of materials in the virtual production space. The material flow process and equipment linkage process of the physical production line are simulated, and the virtual production environment is debugged to generate the virtual production environment.

[0010] Specifically, the process of the bidirectional data synchronization mapping mechanism includes: Based on the parameter information of the virtual production environment and the operating parameter set, a one-to-one correspondence between virtual model parameters and physical device parameters is defined, and a parameter mapping lookup table is generated. The system presets a data synchronization cycle, establishes data upload and command distribution channels, debugs the data synchronization process, simulates changes in physical device parameters, updates virtual model parameters synchronously, simulates adjustments to virtual model parameters, and enables physical devices to receive and execute commands.

[0011] Specifically, the execution process of the simulation control command includes: The production line control program is parsed and logically decomposed to extract all control instructions, instruction execution sequence, and corresponding device objects. According to the execution logic of the control program, the process of issuing control commands is simulated. Control commands are issued from the virtual control terminal, the transmission process of commands in the transmission channel is simulated, and the command transmission delay is recorded. After the commands are transmitted to the virtual device, the response process of the virtual device to the control commands is simulated, and the device response time, response action, and post-response operating parameters are recorded. For timing-based control instructions, we simulate a scenario of parallel execution of multiple instructions to verify the rationality of the instruction execution timing and troubleshoot instruction conflicts.

[0012] Specifically, the process of generating the simulation condition data includes: The virtual production environment is invoked, the virtual material model is loaded, and each piece of equipment in the virtual production environment is started sequentially according to the actual production steps of the physical production line. The entire process of simulating the changes in the operating status of each piece of equipment during the production process, the real-time adjustment of parameters, and the changes in the form and flow path of materials is simulated. During the simulation, the operating parameters of each device, the process deviations of each process step, the completion time of each process node, and the material status data are recorded according to the preset acquisition frequency. All recorded data are classified and organized, and structured and integrated according to process steps and equipment types to generate the simulation operating condition data.

[0013] Specifically, the process of generating the standardized dataset includes: Historical production data and simulation data from the chemical fiber spinning production line are called up and uniformly cleaned. The two types of data are converted into a unified data format, and all data are mapped to a preset range. According to a preset ratio, all processed data are randomly divided into training subsets and validation subsets. The training subsets and validation subsets are integrated to generate the standardized dataset.

[0014] Specifically, the specific process of the mapping prediction model includes: The training subset of the standardized dataset is called, the spinning process parameters in the training subset are used as the model input, the product performance indicators are used as the model output, the initial hyperparameters are set, the model training process is started, the prediction error of the mapping prediction model is calculated in real time during the training process, and the hyperparameters are continuously adjusted and the model structure is optimized by using the gradient descent method. After each training round, the validation subset of the standardized dataset is called, the process parameters of the validation subset are input, the predicted values ​​of the product performance indicators output by the mapping prediction model are obtained, and the prediction accuracy is calculated by comparing them with the actual product performance indicators in the validation subset. When the prediction accuracy reaches a preset threshold, training is stopped, the current model structure and hyperparameters are saved, and the construction of the mapping prediction model is completed.

[0015] Specifically, the process of setting the constraints includes: Based on industry standards and enterprise production standards for chemical fiber spinning products, the upper and lower limits of each product's performance indicators are used as core constraints and entered into the constraint parameter library. Obtain energy consumption data and production efficiency data of chemical fiber spinning production line, and set upper limit values ​​for energy consumption and lower limit values ​​for production efficiency as auxiliary constraints; The constraints are quantified by converting each constraint into a numerical range or mathematical expression, clarifying the priority of the constraints, and verifying the constraints by substituting historical process parameters and corresponding product performance, energy consumption, and production efficiency data.

[0016] Specifically, the process of generating the optimal combination of process parameters includes: The mapping prediction model is invoked to start global optimization calculation. Initial process parameter combinations are randomly generated according to a preset number of iterations. Each set of process parameter combinations is input into the mapping prediction model to obtain the corresponding product performance index prediction value. Based on the constraints, the process parameter combinations that satisfy all constraints are selected, and the comprehensive performance score of each qualified combination is calculated. After completing the preset number of iterations, the process parameter combination with the best performance is obtained. The optimal combination of process parameters is verified and input into the mapping prediction model to confirm that the predicted values ​​of product performance indicators meet the constraints and reach the optimal level, thus generating the optimal combination of process parameters.

[0017] The beneficial effects of this invention are as follows: This invention constructs a three-dimensional virtual model and virtual production environment through a digital twin module and establishes a two-way data synchronization mapping mechanism to achieve real-time linkage between the physical spinning production line and the virtual model. This enables the virtual space to accurately replicate the operating status of the physical production line, providing a high-fidelity simulation foundation for subsequent virtual debugging and process optimization.

[0018] The virtual debugging module enables the loading of control programs, simulation of control command execution, and simulation of the spinning production process in a virtual production environment. Production line debugging can be completed without stopping the machine, effectively reducing on-site debugging time and production costs, avoiding the risk of equipment damage and product scrapping during physical debugging, and improving the safety and efficiency of production line debugging.

[0019] By constructing a mapping prediction model based on historical production data and virtual simulation data through the process optimization module, and combining it with intelligent optimization algorithms, global optimization of process parameters is achieved. This eliminates the reliance on human experience and can accurately find the optimal combination of process parameters that meets product performance indicators. It solves the defects of traditional process optimization, such as strong blindness and low efficiency, and improves the stability and consistency of spinning product quality. Attached Figure Description

[0020] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0021] Figure 1 This is a system architecture diagram of the virtual debugging and process optimization system for chemical fiber spinning production lines based on digital twins, as presented in this invention. Figure 2 This is a flowchart illustrating the construction process of the three-dimensional virtual model in this invention. Figure 3 This is a flowchart of the virtual debugging process in this invention. Detailed Implementation

[0022] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0023] Please see Figures 1-3 A virtual debugging and process optimization system for chemical fiber spinning production lines based on digital twins includes a data processing module, a digital twin module, a virtual debugging module, and a process optimization module. The data processing module acquires the equipment operating parameters of the chemical fiber spinning production line, and generates an operating parameter set after data preprocessing; The digital twin module constructs a three-dimensional virtual model that maps to the production line based on the equipment's three-dimensional feature data and process parameters, generating a virtual production environment; and establishes a two-way data synchronization mapping mechanism based on the virtual production environment and the set of operating parameters. The virtual debugging module, based on the virtual production environment, loads the production line control program, simulates the control instruction execution process, simulates the chemical fiber spinning production process, and generates simulated working condition data. The process optimization module generates a standardized dataset based on historical production data and the simulation operating condition data, and trains it to construct a mapping prediction model between spinning process parameters and product performance indicators. Using product performance indicators as constraints, it performs global optimization calculations based on the mapping prediction model to generate the optimal combination of process parameters.

[0024] Specifically, the process for setting the equipment operating parameters includes: IoT sensing elements are deployed at key monitoring points of each core production equipment in the chemical fiber spinning production line. These IoT sensing elements include temperature sensors, pressure sensors, speed sensors, and tension sensors. Each sensor collects the operating parameters of the corresponding equipment in real time according to a preset collection frequency. The collected parameters specifically include equipment operating temperature, internal operating pressure, core component speed, and material transmission tension. During the collection process, the sensors convert analog signals into digital signals in real time. After each batch collection is completed, the collected digital signals are initially processed, and the equipment number, collection point, and collection timestamp corresponding to each parameter are labeled. A raw parameter collection list containing parameter type, parameter value, and collection information is generated, thus completing the acquisition of operating parameters. The raw parameter collection list is synchronously transmitted to the temporary storage area of ​​the data processing module for subsequent preprocessing steps.

[0025] Specifically, the data preprocessing process includes: The system retrieves the original parameter collection list from the temporary storage area and uses an outlier removal algorithm to remove outliers from the original operating parameters. It calculates the mean and standard deviation of each parameter, identifies parameters exceeding the reasonable range as outliers, and removes them. Simultaneously, it records the parameter type, collection time, and corresponding device information of each outlier. For missing values ​​in the dataset after outlier removal, it uses linear interpolation to calculate the estimated parameter value at the missing position based on the adjacent valid parameter values, thus completing the missing value completion. A normalization algorithm is then used to map all processed parameter data to a preset range, eliminating dimensional differences between different parameters and completing data standardization. All standardized data is categorized by device type, and different parameters from the same device are sorted and organized by parameter category to generate a structured operating parameter set. This operating parameter set includes parameter name, collection time, parameter value, device identification information, and preprocessing identifier. The operating parameter set is stored in a designated database for use by the digital twin module.

[0026] Specifically, the construction process of the three-dimensional virtual model includes: Three-dimensional feature data of the equipment is extracted from equipment design drawings and equipment operation and maintenance records. Specifically, this includes the structural dimensions, contour features, assembly tolerances, and connection relationships between the components. The extracted data undergoes redundancy removal, eliminating duplicate and invalid feature data to generate a three-dimensional feature dataset. Simultaneously, key parameters such as process flow channel size requirements, equipment operating limit thresholds, and component movement ranges are extracted from the process parameters and used as constraints for model construction. Subsequently, using three-dimensional modeling technology, independent three-dimensional models of each core piece of equipment are constructed sequentially based on the equipment's three-dimensional feature dataset. During the modeling process, the structural dimensions and assembly tolerances of the components are strictly followed to ensure that the shape and size of each component are consistent with the physical equipment. After the independent models are constructed, the installation position, spacing, and connection method of each piece of equipment in virtual space are determined according to the actual layout drawings of the physical production line. The independent three-dimensional models are then virtually assembled. During the assembly process, the assembly gaps between the equipment and between equipment components are detected, and the assembly position is adjusted until it meets the actual layout requirements of the physical production line. This completes the construction of a three-dimensional virtual model mapped to the production line, and the completed three-dimensional virtual model is synchronously stored in the model library of the digital twin module.

[0027] Specifically, the specific processes of the virtual production environment include: The system imports a completed 3D virtual model from the model library into the virtual simulation platform. Then, it imports the complete process flow rules of the chemical fiber spinning production line, including the material feeding sequence, the start-up and shutdown sequence of each piece of equipment, the connection logic of process links, and the material flow path. Simultaneously, it imports the equipment operation logic, including the equipment's start-up conditions, stop conditions, parameter adjustment logic, and fault response logic. The imported environmental parameters include temperature, humidity, and air pressure in the production space, consistent with the physical production line. Next, it renders the virtual model, setting the background, lighting, and material appearance characteristics of the virtual production space to ensure visual consistency between the virtual and physical production scenes. Then, it configures the virtual model's parameters, assigning values ​​to equipment operating thresholds and process flow parameters to achieve parameter matching between the virtual model and the physical equipment. Finally, it simulates the material flow and equipment linkage processes of the physical production line, debugging the virtual production environment to ensure that the operating logic and process flow of each piece of equipment in the virtual environment are completely consistent with the physical production line. This generates a virtual production environment that supports real-time parameter updates and dynamic simulation. After the virtual production environment is generated, it establishes a data interaction channel with the data processing module.

[0028] Specifically, the process of the bidirectional data synchronization mapping mechanism includes: The process involves analyzing the 3D virtual model structure of the virtual production environment, clarifying the parameter types and identifiers for each device and component within the virtual model, and simultaneously analyzing the operating parameter set generated by the data processing module to identify the parameter types, identifiers, and ranges of the physical devices. Based on this parameter information, a one-to-one correspondence between virtual model parameters and physical device parameters is defined, forming a parameter mapping table to ensure that every parameter in the virtual model corresponds to a specific operating parameter on the physical device, and vice versa. A data synchronization cycle is set, and the synchronization priority of different parameters is adjusted according to their importance, with core parameters set to the highest priority and non-core parameters to a normal priority. A data upload channel and a command delivery channel are then established. The data upload channel transmits real-time operating parameters of the physical devices from the data processing module to the virtual production environment, while the command delivery channel transmits parameter adjustment commands from the virtual production environment to the physical device control terminal. The data synchronization process is debugged by simulating changes in physical device parameters to verify whether the virtual model parameters can be updated synchronously, and by simulating virtual model parameter adjustments to verify whether the physical devices can receive and execute commands. After confirming accurate synchronization and error-free command execution, the bidirectional data synchronization mapping mechanism is established.

[0029] Specifically, the execution process of the simulation control command includes: The actual control program is exported from the physical control terminal of the production line and its format is converted to ensure compatibility with the simulation platform of the virtual production environment. The converted control program is then imported into the virtual production environment, where the simulation platform performs syntax parsing and logical decomposition, extracting all control instructions, instruction execution timing, and corresponding device objects. Following the execution logic of the control program, the process of issuing control instructions is simulated, sending instructions from the virtual control terminal and simulating their transmission in the transmission channel, recording the instruction transmission delay. After the instructions are transmitted to the virtual device, the response process of the virtual device to the control instructions is simulated, recording the device response time, response action, and post-response operating parameters. Simultaneously, the preset execution results of the control instructions are compared with the actual response results of the virtual device to verify the accuracy of the control instruction execution. For timing-based control instructions, a scenario of multiple instructions executing in parallel is simulated to verify the rationality of the instruction execution timing and identify instruction conflicts. After the simulation is completed, a control instruction execution simulation report is generated, including information such as instruction execution accuracy, timing rationality, and device response parameters, completing the simulation of the control instruction execution process.

[0030] Specifically, the process of generating the simulation condition data includes: The virtual production environment generated by the digital twin module is invoked, and a virtual material model is loaded. The physical characteristics of the virtual material model are consistent with those of the materials used in physical production. Following the actual production steps of the physical production line, each piece of equipment in the virtual production environment is started sequentially. The material feeding process is simulated, and the virtual material is introduced into the virtual screw extruder at a preset feeding speed and quantity. Next, the material processing process is simulated, including the heating and extrusion processes of the screw extruder, the spinning process in the virtual spinning box, and subsequent cooling, drawing, and winding processes. The entire process simulates the changes in the operating status of each piece of equipment during production, real-time parameter adjustments, changes in the morphology of the material, and its flow path. During the simulation, the operating parameters of each piece of equipment, the process deviations of each process step, the completion time of each process node, and the material status data are recorded at a preset acquisition frequency. After the simulation, all recorded data is classified and organized, invalid simulation data is removed, and the data is structured and integrated according to process steps and equipment types to generate complete simulation condition data. This simulation condition data is consistent with the format of the operating parameter set generated by the data processing module, including parameter name, acquisition time, parameter value, equipment identifier, and process step identifier, facilitating its use by the process optimization module.

[0031] Specifically, the process of generating the standardized dataset includes: Historical production data from the production line is accessed via a database interface. This historical data includes equipment operating parameters, process parameters, product performance testing data, and production environment data from a past period. Simulation operating condition data generated by the virtual debugging module is also accessed. Both types of data undergo unified cleaning, employing a combination of manual review and algorithmic filtering to remove invalid and duplicate data. Outliers are removed using the same outlier removal algorithm as the data processing module. Next, the cleaned data is converted to a unified data format, clearly defining the meaning and data type of each field to ensure format consistency. Then, a normalization algorithm is used to map all data to a preset range, eliminating the influence of different data types on their dimensions. Finally, according to a preset ratio, all processed data is randomly divided into a training subset and a validation subset. The training subset is used for model training, and the validation subset is used for model validation. The training and validation subsets are integrated to generate a standardized dataset, which is stored in the dataset storage area of ​​the process optimization module for subsequent model training.

[0032] Specifically, the specific process of the mapping prediction model includes: Determine the type of machine learning algorithm and select either a neural network or a random forest algorithm as the model training algorithm; call the training subset of the standardized dataset, use the spinning process parameters in the training subset as the model input, and the product performance indicators as the model output, and import them into the selected machine learning algorithm; set the initial hyperparameters of the algorithm, start the model training process, calculate the prediction error of the model in real time during the training process, and continuously adjust the hyperparameters and optimize the model structure using the gradient descent method.

[0033] After each training round, the validation subset of the standardized dataset is called, and the process parameters of the validation subset are input into the trained model. The predicted values ​​of the product performance indicators output by the model are obtained and compared with the actual product performance indicators in the validation subset to calculate the model prediction accuracy. If the prediction accuracy does not reach the preset threshold, the hyperparameters are adjusted and the training process is repeated. If the prediction accuracy reaches the preset threshold, the training is stopped, the current model structure and hyperparameters are saved, and the mapping prediction model between spinning process parameters and product performance indicators is completed. The completed prediction model is synchronously stored in the model library of the process optimization module.

[0034] Specifically, the process of setting the constraints includes: Industry standards and enterprise production standards for chemical fiber spinning products were consulted to clarify the acceptable range of product performance indicators. The upper and lower limits of each product performance indicator were used as core constraints and entered into the constraint parameter library of the optimization algorithm. Subsequently, energy consumption data and production efficiency data of the production line were collected to analyze the correlation between energy consumption and process parameters, and the correlation between production efficiency and process parameters. Upper limits of energy consumption and lower limits of production efficiency were set as auxiliary constraints, which took effect in parallel with the core constraints. Next, the constraints were quantified, converting each constraint into a numerical range or mathematical expression that the optimization algorithm could recognize, and clarifying the priority of the constraints, with core constraints having higher priority than auxiliary constraints. Finally, the constraints were verified by substituting historical process parameters and corresponding product performance, energy consumption, and production efficiency data to confirm that the constraints were set reasonably, without contradictions or omissions, forming a complete optimization constraint system for subsequent global optimization calculations.

[0035] Specifically, the process of generating the optimal combination of process parameters includes: Genetic algorithm or particle swarm optimization algorithm is selected as the intelligent optimization algorithm. The mapping prediction model in the process optimization module model library is called, and the constraints in the optimization constraint system are entered into the intelligent optimization algorithm. Then, the optimization range of process parameters is set. The optimization range is determined based on the reasonable range of historical process parameters and the effective range of virtual debugging simulation data, and the adjustable step size of each process parameter is defined. Global optimization calculation is started. The algorithm randomly generates multiple sets of initial process parameter combinations according to the preset number of iterations. Each set of process parameter combinations is input into the mapping prediction model to obtain the corresponding predicted value of product performance index.

[0036] Based on the optimization constraints, process parameter combinations that satisfy all constraints are selected, and the comprehensive product performance score of each qualified combination is calculated. During the iteration process, the process parameter combinations are continuously optimized, eliminating combinations with poor performance and retaining combinations with better performance, until the preset number of iterations is completed to obtain the process parameter combination with optimal performance. Finally, the selected optimal process parameter combination is verified by inputting it into the mapping prediction model to confirm that the predicted value of the product performance index meets the constraints and reaches the optimal level. At the same time, the feasibility of the parameter combination is verified by combining it with historical production data. After confirming that there are no abnormalities, the optimal process parameter combination is generated. The optimal process parameter combination includes the specific parameters of each process step, the adjustable range of the parameters, and the execution standard, and is synchronously stored in the parameter library.

[0037] This embodiment provides a virtual debugging and process optimization system for a chemical fiber spinning production line based on digital twins. The system includes a data processing module, a digital twin module, a virtual debugging module, and a process optimization module. These modules work together to achieve virtual debugging and process optimization of the chemical fiber spinning production line. The specific implementation process is as follows: 1. Implementation process of the data processing module The data processing module first uses IoT sensing elements deployed at monitoring points of each core production equipment in the chemical fiber spinning production line to collect the operating parameters of each equipment in real time according to a preset collection frequency. The equipment operating parameters include equipment operating temperature T, equipment internal operating pressure P, core component rotation speed N, and material transmission tension F. During the collection process, the IoT sensing elements convert the collected analog signals into digital signals in real time. After each batch collection is completed, the collected digital signals are initially sorted, and the equipment number, collection point, and collection timestamp corresponding to each parameter are labeled to generate an original parameter list.

[0038] The data processing module then preprocesses the original parameter list, calling up all parameter data from the original list, calculating the mean μ and standard deviation σ for each parameter, and identifying and removing parameters that exceed the preset reasonable range [μ-σ, μ+σ]. Simultaneously, it records the parameter type, acquisition time, and corresponding equipment information for each outlier. Next, a normalization algorithm is used to map all processed parameter data to the preset parameter range [0, 1], eliminating dimensional differences between different parameters. Finally, the processed parameters are categorized by equipment type, and different parameters of the same equipment are sorted and organized according to parameter category (temperature, pressure, speed, tension) to generate an operating parameter set S. The operating parameter set S is stored in a designated database for use by the digital twin module.

[0039] 2. Implementation process of the digital twin module The core of the digital twin module is to construct a 3D virtual model that is precisely mapped to the physical production line. The specific construction process is as follows: Data Extraction and Processing: Three-dimensional feature data of the equipment is extracted from the design drawings and maintenance records of the core equipment in the chemical fiber spinning production line. This data includes the structural dimensions L, contour features C, assembly tolerances Δ, and connection relationships R between components. Redundancy is removed from the extracted 3D feature data, eliminating duplicate and invalid data to generate a 3D feature dataset M. Simultaneously, the process flow channel size requirement D, equipment operating limit threshold Y, and component movement range S1 are extracted from the process parameters. These parameters are used as constraints for constructing the 3D virtual model, forming a constraint set G = {D, Y, S1}.

[0040] Independent Equipment Model Construction: Using 3D modeling technology, based on the equipment's 3D feature dataset M, independent 3D models of each core piece of equipment are constructed sequentially. During the modeling process, the structural dimensions L and assembly tolerances Δ of the components are strictly adhered to, ensuring that the shape and size of each component are consistent with the physical equipment. Each independent 3D model includes a corresponding parameter interface for receiving parameter data from the subsequent operating parameter set S. After completing the construction of the independent 3D models, an independent model set M1={M11 M 12 M 1n}, where n is the number of core devices, M 11 M 12 ... M 1n Each is an independent 3D model of a core device.

[0041] Virtual Assembly and Model Calibration: Based on the actual layout drawings of the physical production line, determine the installation position coordinates (X1, Y1, Z1), equipment spacing L1, and connection method R1 of each independent 3D model in the virtual space. Virtually assemble the independent 3D models in the independent model set M1. During the assembly process, detect the assembly gap Δ1 between each piece of equipment and between equipment parts. If Δ1 exceeds the assembly tolerance requirements in the constraint set G, adjust the installation position of the corresponding model until the assembly gap meets the constraint conditions. After the assembly is completed, call some parameters from the running parameter set S and input them into the 3D virtual model to verify the parameter response performance of the model, ensuring that the model can accurately receive and feedback parameter changes, and complete the construction of the 3D virtual model M2 mapped to the physical production line.

[0042] The virtual production environment generation and bidirectional data synchronization mapping mechanism is established by calling the completed 3D virtual model M2, importing the complete process flow rules K, equipment operation logic L2 and environmental parameters E of the chemical fiber spinning production line, rendering the scene of the 3D virtual model M2, setting the background, lighting and material appearance characteristics of the virtual production space; simulating the material flow process and equipment linkage process of the physical production line, debugging the virtual production environment, and generating the virtual production environment H.

[0043] Based on the parameter information of the virtual production environment H and the operating parameter set S, a one-to-one correspondence between virtual model parameters and physical equipment parameters is defined. That is, the temperature parameter T1 in the virtual model corresponds to the temperature parameter T of the physical equipment, the pressure parameter P1 in the virtual model corresponds to the pressure parameter P of the physical equipment, and so on, generating a parameter mapping table K1. A preset data synchronization period T2 is established, and a data upload channel T3 and an instruction distribution channel T4 are established. The data upload channel T3 is used to transmit the real-time operating parameters of the physical equipment from the data processing module to the virtual production environment H, and the instruction distribution channel T4 is used to transmit the parameter adjustment instructions of the virtual production environment H to the physical equipment control terminal. The data synchronization process is debugged by simulating changes in the physical equipment parameter T to verify whether the virtual model parameter T1 is updated synchronously, and simulating adjustments to the virtual model parameter P1 to verify whether the physical equipment can receive and execute the corresponding parameter adjustment instructions, ensuring that the two-way data synchronization is accurate and error-free, and completing the establishment of the two-way data synchronization mapping mechanism.

[0044] 3. Implementation process of the virtual debugging module The virtual debugging module performs virtual debugging based on the virtual production environment H generated by the digital twin module. The specific process is as follows: Load the production line control program, perform syntax parsing and logic decomposition on the control program, extract all control instructions I, instruction execution timing T5, and the corresponding device objects O, and generate an instruction set I1={I 11 I 12 , ..., I 1m}, where m is the number of control commands.

[0045] According to the execution logic of the control program, the process of issuing control commands is simulated. Control commands Iᵢ (i=1, 2, ..., m) are issued from the virtual control terminal, and the transmission process of the commands in the transmission channel is simulated, recording the command transmission delay t1. After the command is transmitted to the corresponding virtual device Oᵢ, the response process of the virtual device to the control command Iᵢ is simulated, recording the device response time t2, response action A, and the running parameters S1 after the response. For timing-based control commands, a scenario of parallel execution of multiple commands is simulated to verify the timing rationality of command execution, and to investigate command conflict issues. If command conflicts exist, the conflicting command I2 and the conflict type are recorded, and a command conflict report is generated.

[0046] For timing-based control instructions, simulate a scenario of parallel execution of multiple instructions to verify the timing rationality of instruction execution, investigate instruction conflict issues, and if instruction conflicts exist, record the conflicting instruction I and the conflict type, and generate an instruction conflict report.

[0047] The virtual production environment H is invoked, and the virtual material model W (the physical properties of the virtual material model are consistent with those of the materials used in physical production) is loaded. Following the actual production steps of the physical production line, each piece of equipment in the virtual production environment is started sequentially. The entire process of simulating the changes in the operating status of each piece of equipment during the production process, the real-time adjustment of parameters, the changes in the form of materials, and the flow path are simulated. During the simulation, the operating parameters of each piece of equipment, the process deviation Δ2 of each process link, the completion time t3 of each process node, and the material status data W1 are recorded according to the preset acquisition frequency. All recorded data are classified and organized, and structured and integrated according to process link and equipment type to generate simulation condition data D1. Simulation condition data D1 is stored in the database for use by the process optimization module.

[0048] 4. Implementation process of the process optimization module The process optimization module, based on historical production data D2 and simulation condition data D1 generated by the virtual debugging module, completes the generation of standardized datasets, construction of mapping prediction models, and generation of optimal process parameter combinations. The specific process is as follows: Standardized dataset generation: Historical production data D2 and simulated operating condition data D1 from the chemical fiber spinning production line are called. A combination of manual review and algorithm screening is used to perform unified cleaning processing to remove invalid and duplicate data. The two types of cleaned data are converted into a unified data format. A normalization algorithm is used to map all data to a preset interval [0, 1] to eliminate the influence of different dimensions between different types of data. According to a preset ratio, all processed data are randomly divided into a training subset D3 and a validation subset D4. The training subset D3 and the validation subset D4 are integrated to generate a standardized dataset D5.

[0049] Mapping prediction model construction and calculation process: The neural network algorithm is selected as the model training algorithm. The training subset D3 of the standardized dataset is called. The spinning process parameters (spinning temperature T6, cooling air parameter W2, draw ratio K1) in the training subset D3 are used as the model input X1={T6, W2, K1}, and the product performance indicators (fineness F1, strength Q1, elongation E1) are used as the model output Y1={F1, Q1, E1} and imported into the neural network algorithm.

[0050] Set the initial hyperparameters of the neural network algorithm, including the number of hidden layers h, the learning rate η, and the number of iterations k1, and start the model training process. During training, the prediction error E2 of the mapping prediction model is calculated in real time using the following formula: E2=(1 / m)×Σ(Y 11 -Y 12 ) 2 Where m is the number of samples in the training subset D3, Y 11 Y represents the predicted product performance index output by the mapping prediction model. 12 The actual values ​​of product performance metrics in the training subset D3.

[0051] The gradient descent method is used to continuously adjust the hyperparameters (number of hidden layers h, learning rate η) to optimize the model structure. The parameter update formula for gradient descent is: η2=η1-α×∂E2 / ∂η Where α is the learning step size, and ∂E2 / ∂η is the partial derivative of the prediction error E2 with respect to the learning rate η.

[0052] After each training round, the validation subset D4 of the standardized dataset is called, and the process parameters of the validation subset D4 are input into the trained mapping prediction model to obtain the predicted value Y of the product performance index output by the model. 21 The actual product performance index Y in the validation subset D4 is compared with that of the product. 22 For comparison, the prediction accuracy A1 is calculated using the following formula: A1=1-(|Y 21 -Y 22| / Y 22 ) If the prediction accuracy A1 does not reach the preset threshold A1, continue to adjust the hyperparameters and repeat the training process; if the prediction accuracy A1 reaches the preset threshold A1, stop training, save the current model structure and hyperparameters, and complete the construction of the mapping prediction model M3 between spinning process parameters and product performance indicators.

[0053] Constraint Setting: Based on industry standards and enterprise production standards for chemical fiber spinning products, the upper limit value Y1 and lower limit value Y2 of each product performance index are used as core constraints and entered into the constraint parameter library; energy consumption data N1 and production efficiency data E2 of the chemical fiber spinning production line are obtained, and the upper limit value N2 of energy consumption and the lower limit value E3 of production efficiency are set as auxiliary constraints; the constraints are quantified by converting each constraint into a numerical range or mathematical expression, clarifying the priority of the constraints, with core constraints having higher priority than auxiliary constraints, forming a constraint set G1={Y1, Y2, N2, E3}; the constraints are verified by substituting historical process parameters and corresponding product performance, energy consumption, and production efficiency data to confirm that the constraint settings are reasonable, without contradictions or omissions.

[0054] The process of generating and calculating the optimal combination of process parameters is as follows: Genetic algorithm is selected as the intelligent optimization algorithm, the mapping prediction model M3 is called, the constraints in the constraint set G1 are entered into the genetic algorithm, the optimization range X2 of the process parameters is set, and the adjustable step size Δ3 of each process parameter is defined.

[0055] The global optimization calculation is initiated, and the algorithm randomly generates multiple sets of initial process parameter combinations X1={X according to the preset number of iterations k2. 11 X 12 , ..., X 1n Each initial process parameter combination X1ᵢ (i=1, 2, ..., n) is input into the mapping prediction model M3, and the corresponding predicted product performance index Y is obtained through model calculation. 31 .

[0056] Based on the constraint set G1, process parameter combinations X2 that satisfy all constraints are selected. A weighted summation method is used to calculate the comprehensive product performance score S1 for each qualified combination. The scoring formula is as follows: S1=ω1×F 11 +ω2×Q 11 +ω3×E 11 Wherein, ω1, ω2, and ω3 are the weighting coefficients of fineness, strength, and elongation, respectively, and ω1+ω2+ω3=1.

[0057] During the iteration process, the combination of process parameters is continuously optimized by following the selection, crossover, and mutation operations of the genetic algorithm. Combinations with lower overall performance scores S1 are eliminated, while combinations with higher scores S1 are retained, until the preset number of iterations k2 is completed, and the optimal combination of process parameters X3 is obtained.

[0058] The optimal process parameter combination X3 was selected and validated. It was then input into the mapping prediction model M3, and the predicted value of the product performance index Y was calculated again. 41 Confirm Y 41 If the set of constraints G1 is satisfied and the comprehensive score S2 is the highest, and the feasibility of the parameter combination is verified by combining historical production data D2, and after confirming that there are no anomalies, the optimal process parameter combination X3 is generated for the physical production line to call and execute.

[0059] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A virtual commissioning and process optimization system for a chemical fiber spinning production line based on digital twinning, characterized in that, It includes a data processing module, a digital twin module, a virtual debugging module, and a process optimization module; The data processing module acquires the equipment operating parameters of the chemical fiber spinning production line, and generates an operating parameter set after data preprocessing; The digital twin module constructs a three-dimensional virtual model that maps to the production line based on the equipment's three-dimensional feature data and process parameters, generating a virtual production environment; and establishes a two-way data synchronization mapping mechanism based on the virtual production environment and the set of operating parameters. The virtual debugging module loads the production line control program based on the virtual production environment, simulates the control instruction execution process, simulates the chemical fiber spinning production process, and generates simulated working condition data. The process optimization module generates a standardized dataset based on historical production data and the simulation working condition data, and trains it to construct a mapping prediction model between spinning process parameters and product performance indicators. Using product performance indicators as constraints, a global optimization calculation is performed based on the mapping prediction model to generate the optimal combination of process parameters.

2. The system of claim 1, wherein, The specific process for setting the equipment operating parameters includes: IoT sensing elements are deployed at the monitoring points of the production equipment in the chemical fiber spinning production line to collect the operating parameters of the corresponding equipment in real time according to the preset collection frequency. The analog signals are converted into digital signals in real time. After each batch acquisition is completed, the acquired digital signals are initially sorted out, and the equipment number, acquisition point and acquisition timestamp corresponding to the operating parameters of each device are marked to generate the original parameter list.

3. The system of claim 1, wherein, The specific process of data preprocessing includes: The original parameter list is called up, the mean and standard deviation of each parameter are calculated, parameters that exceed the preset reasonable range are identified as outliers and removed, and the parameter type, collection time and corresponding device information of the outliers are recorded. All processed parameter data are mapped to preset parameter ranges, categorized by device type, and different parameters of the same device are sorted and organized according to parameter category to generate the operating parameter set.

4. The system according to claim 1, characterized in that, The construction process of the three-dimensional virtual model includes: Extract the equipment's 3D feature data from equipment design drawings and equipment operation and maintenance records, perform redundancy removal processing, and generate a 3D feature dataset of the equipment. Extract the process flow channel size requirements, equipment operating limit thresholds, and component motion range parameters from the process parameters as constraints. Using 3D modeling technology, based on the equipment's 3D feature dataset, independent 3D models of each core device are constructed sequentially. Based on the actual layout drawings of the physical production line, the installation position, spacing and connection method of each piece of equipment in the virtual space are determined, and each independent 3D model is virtually assembled to complete the construction of a 3D virtual model mapped to the production line.

5. The system according to claim 1, characterized in that, The specific process of the virtual production environment includes: Call the three-dimensional virtual model to import the complete process flow rules of the chemical fiber spinning production line, as well as the equipment operation logic and environmental parameters. The three-dimensional virtual model is rendered to set the background, lighting, and appearance features of materials in the virtual production space. The material flow process and equipment linkage process of the physical production line are simulated, and the virtual production environment is debugged to generate the virtual production environment.

6. The system according to claim 1, characterized in that, The specific process of the bidirectional data synchronization mapping mechanism includes: Based on the parameter information of the virtual production environment and the operating parameter set, a one-to-one correspondence between virtual model parameters and physical device parameters is defined, and a parameter mapping lookup table is generated. The system presets a data synchronization cycle, establishes data upload and command distribution channels, debugs the data synchronization process, simulates changes in physical device parameters, updates virtual model parameters synchronously, simulates adjustments to virtual model parameters, and enables physical devices to receive and execute commands.

7. The system according to claim 1, characterized in that, The specific process of the simulation control command execution flow includes: The production line control program is parsed and logically decomposed to extract all control instructions, instruction execution sequence, and corresponding device objects. According to the execution logic of the control program, the process of issuing control commands is simulated. Control commands are issued from the virtual control terminal, the transmission process of commands in the transmission channel is simulated, and the command transmission delay is recorded. After the commands are transmitted to the virtual device, the response process of the virtual device to the control commands is simulated, and the device response time, response action, and post-response operating parameters are recorded. For timing-based control instructions, we simulate a scenario of parallel execution of multiple instructions to verify the rationality of the instruction execution timing and troubleshoot instruction conflicts.

8. The system according to claim 1, characterized in that, The process of generating the simulation condition data includes: The virtual production environment is invoked, the virtual material model is loaded, and each piece of equipment in the virtual production environment is started sequentially according to the actual production steps of the physical production line. The entire process of simulating the changes in the operating status of each piece of equipment during the production process, the real-time adjustment of parameters, and the changes in the form and flow path of materials is simulated. During the simulation, the operating parameters of each device, the process deviations of each process step, the completion time of each process node, and the material status data are recorded according to the preset acquisition frequency. All recorded data are classified and organized, and structured and integrated according to process steps and equipment types to generate the simulation operating condition data.

9. The system according to claim 1, characterized in that, The process of generating the standardized dataset includes: Historical production data and simulation data from the chemical fiber spinning production line are called up and uniformly cleaned. The two types of data are converted into a unified data format, and all data are mapped to a preset range. According to a preset ratio, all processed data are randomly divided into training subsets and validation subsets. The training subsets and validation subsets are integrated to generate the standardized dataset.

10. The system according to claim 1, characterized in that, The specific process of the mapping prediction model includes: The training subset of the standardized dataset is called, the spinning process parameters in the training subset are used as the model input, the product performance indicators are used as the model output, the initial hyperparameters are set, the model training process is started, the prediction error of the mapping prediction model is calculated in real time during the training process, and the hyperparameters are continuously adjusted and the model structure is optimized by using the gradient descent method. After each training round, the validation subset of the standardized dataset is called, the process parameters of the validation subset are input, the predicted values ​​of the product performance indicators output by the mapping prediction model are obtained, and the prediction accuracy is calculated by comparing them with the actual product performance indicators in the validation subset. When the prediction accuracy reaches a preset threshold, training is stopped, the current model structure and hyperparameters are saved, and the construction of the mapping prediction model is completed.

11. The system according to claim 1, characterized in that, The specific process of setting the constraints includes: Based on industry standards and enterprise production standards for chemical fiber spinning products, the upper and lower limits of each product's performance indicators are used as core constraints and entered into the constraint parameter library. Obtain energy consumption data and production efficiency data of chemical fiber spinning production line, and set upper limit values ​​for energy consumption and lower limit values ​​for production efficiency as auxiliary constraints; The constraints are quantified by converting each constraint into a numerical range or mathematical expression, clarifying the priority of the constraints, and verifying the constraints by substituting historical process parameters and corresponding product performance, energy consumption, and production efficiency data.

12. The system according to claim 1, characterized in that, The process of generating the optimal combination of process parameters includes: The mapping prediction model is invoked to start global optimization calculation. Initial process parameter combinations are randomly generated according to a preset number of iterations. Each set of process parameter combinations is input into the mapping prediction model to obtain the corresponding product performance index prediction value. Based on the constraints, the process parameter combinations that satisfy all constraints are selected, and the comprehensive performance score of each qualified combination is calculated. After completing the preset number of iterations, the process parameter combination with the best performance is obtained. The optimal combination of process parameters is verified and input into the mapping prediction model to confirm that the predicted values ​​of product performance indicators meet the constraints and reach the optimal level, thus generating the optimal combination of process parameters.

Images