LCP cable manufacturing full-process data management and control system based on digital twinning

Abstract

The invention discloses a digital twinning-based LCP cable manufacturing full-process data management and control system, and particularly relates to the technical field of data management and control, and the system comprises the steps: a process coupling multimode sensing module synchronously collects macroscopic process time sequence data and microscopic response signals, and achieves the "one-object-one-code" tracing of a product unit through a readable and writable RFID tag; the dynamic causal graph construction and updating module fuses the data to construct a dynamic causal graph, and accurately identifies and quantifies a key causal path influencing the product quality; the collaborative optimization digital twinborn body module constructs a virtual simulation environment through a mixed model, and process decision prospective evaluation and causal model adaptive calibration are completed; the distributed decision control module generates a local optimization instruction and a global strategy and executes closed-loop control of the manufacturing process; according to the system, full-process data communication and intelligent management and control are achieved, the product percent of pass and production efficiency are remarkably improved, energy consumption is reduced, and technical support is provided for high-precision LCP cable manufacturing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data management and control, more specifically, the present application relates to an LCP cable manufacturing full-process data management and control system based on digital twinning. BACKGROUND

[0002] Liquid crystal polymer cable has been widely used in high-end fields such as 5G communication device interconnection and aerospace lightweight transmission system due to its excellent dielectric properties, mechanical strength and dimensional stability. Its manufacturing process covers three core sections of spinning, twisting and sheath extrusion, and the precise control of process parameters is extremely high. At present, the traditional manual inspection combined with single-point parameter monitoring mode is generally used in the LCP cable manufacturing industry. Through the deployment of basic sensors in each section to collect temperature, tension and other macro parameters, the process settings are adjusted by relying on the experience of the operating personnel, and the quality is checked through finished product detection. However, there are still some shortcomings in actual use, such as: on the one hand, the traditional control only focuses on the collection of macro process parameters, ignores the micro signals directly representing the intrinsic properties of the material such as molecular orientation degree and melt interface bonding state, and the data of each section is stored separately and lacks correlation. When there are quality defects such as insufficient breaking strength and sheath shedding, it is difficult to accurately locate the root cause, and manual disassembly analysis is required, which is extremely inefficient and prone to misjudgment.

[0003] On the other hand, there is a strong coupling effect between each section of LCP cable manufacturing, such as spinning single filament quality defects that can be transmitted to the twisting and extrusion sections. However, in the traditional mode, each station adjusts the parameters independently, lacks analysis of the coupling relationship across stations, and is prone to the problem of "local optimization but global inefficiency", resulting in low production efficiency and high energy consumption. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide an LCP cable manufacturing full-process data management and control system based on digital twinning, which solves the problems of weak quality traceability and poor cross-station collaboration in the background art through the following scheme.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: an LCP cable manufacturing full-process data management and control system based on digital twinning, comprising: A process coupling multi-mode perception module is used to synchronously collect macro process time series data and micro response signal sets of each product unit, and bind an intelligent identification code readable and writable for each product unit; A dynamic causal graph construction and updating module is used to fuse the macro and micro data sets of the same product unit based on the intelligent identification code, construct a manufacturing process dynamic causal graph and continuously update it, dynamically identify and quantify the key causal paths that affect the quality of the final product by analyzing real-time data flow; Cooperative optimization digital twin module: integrate the dynamic causal diagram as its inference kernel, including a virtual simulation environment, used for forward-looking performance evaluation of key process decisions based on the current causal diagram, and adaptive calibration of causal relationship models according to the difference between the evaluation results and physical feedback; Distributed decision control module: used to generate local optimization instructions for specific stations and global optimization strategies considering cross-station coupling effects according to the forward-looking performance evaluation results, and perform closed-loop control on the manufacturing process.

[0006] Preferably, the macro process timing data includes temperature, tension, pressure and speed collected from spinning, twisting to sheath extrusion section; The micro-response signal set includes signals directly representing the intrinsic physical and chemical state of the material obtained by non-contact sensing devices integrated in the equipment, including molecular orientation degree signals reflecting the arrangement state of LCP molecular chains, and melt rheological image signals reflecting the interface morphology of the sheath and core layer.

[0007] Preferably, the intelligent identification code is a read-write RFID tag; the intelligent identification code serves as a mobile data carrier, and the process execution summary and recommended downstream process parameters are written into it by the upstream station in the manufacturing process, which are read by the decision control module of the downstream station and used as the basis for decision-making.

[0008] Preferably, the dynamic causal diagram of the manufacturing process is a topological network constructed with process parameters and material state as nodes and causal influence relationship between nodes as directed edges; by fusing process mechanism model and causal discovery algorithm based on conditional independence test, the existence and weight of causal edges between nodes are automatically learned and quantified from the fused macro and micro data.

[0009] Preferably, the key causal path is a path in the dynamic causal diagram, starting from the final product quality defect or performance index, traversing all incoming edges in reverse, and calculating the causal contribution degree of each potential path. The path composed of multiple causal edges with the sum of contribution degrees exceeding the preset threshold is determined as the key causal path.

[0010] Preferably, the virtual simulation environment is a hybrid model trained based on physical mechanism and historical data, which can receive real-time data from the dynamic causal relationship engine, and on this basis, simulate virtual disturbance of key process parameters beyond the current actual setting range to explore a better process window.

[0011] Preferably, the adaptive calibration of the causal relationship model is triggered by comparing the prospective evaluation results of the virtual simulation environment with the actual feedback results of the physical production line, and when the deviation continues to exceed the preset tolerance, the incremental learning and adjustment of the weight and structure of the corresponding causal edge in the dynamic causal graph are triggered.

[0012] Preferably, the local optimization instruction is generated by the sub-controller deployed at each station, and the decision objective focuses on optimizing the core quality index of the station, and the local causal chain directly related to the station in the dynamic causal graph is responded quickly.

[0013] Preferably, the global optimization strategy is generated by the central optimizer, and the decision objective considers the overall performance and production efficiency of the whole process, and by analyzing the complex coupling relationship across stations in the dynamic causal graph, the Pareto optimal solution of the set point of each station is solved to coordinate the local optimization instruction.

[0014] Technical effects and advantages of the present application: 1. The present application synchronously collects macro-process data and micro-response signals through process coupling multi-mode perception modules, realizes "one product one code" traceability of the whole life cycle of the product unit by combining the readable and writable RFID tags, fills the blank of traditional control which only focuses on macro parameters; the dynamic causal graph module further quantifies the causal relationship between parameters and quality, can quickly locate the core root cause of quality defects, shortens the traceability time from traditional hours to minutes, effectively reduces the misjudgment rate, realizes accurate rectification of defects combined with closed-loop control, and significantly improves the qualified rate of LCP cable products; 2. The present application solves the problem of "local optimization and global inefficiency" caused by coupling effect of each section, analyzes the complex correlation across stations through the dynamic causal graph module, optimizes the process parameters virtually by the digital twin module, and generates a strategy considering local quality and global benefit by the distributed decision control module; by solving the Pareto optimal solution to coordinate the parameters of each station, the production efficiency is greatly improved under the premise of ensuring product performance, and the energy consumption is reduced, breaking through the coordination bottleneck of traditional control; 3. The present application fuses physical mechanism and data-driven to build a hybrid simulation model, ensures high synchronization between virtual and physical production lines through adaptive calibration, explores the optimal process window without physical trial and error, reduces the trial and error cost of research and production; the whole process data is connected and the closed-loop control system realizes the transformation from "experience-driven" to "data-driven", and the model can be continuously optimized with the accumulation of production data, adapting to the manufacturing needs of different specifications of LCP cable, providing stable and reliable technical support for high-precision cable production in high-end fields. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 It is a schematic diagram of the overall structure of the present application; Figure 2A dynamic causal map construction and update flowchart of the present application; Figure 3 A collaborative optimization digital twin module flowchart of the present application; Figure 4 A distributed decision control and closed-loop control flowchart of the present application. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0017] As shown in the LCP cable manufacturing whole-process data management and control system based on digital twin, Figures 1-4 The LCP cable manufacturing whole-process data management and control system based on digital twin includes: A process coupling multi-mode perception module: used for synchronously collecting macro process timing data and micro response signal set of each product unit, and binding an intelligent identification code readable and writable for each product unit; It needs to be specifically pointed out that the macro process timing data is configured to collect schemes according to the characteristics of key process parameters for the three sections of spinning, stranding and sheath extrusion; in the spinning section, PT100 platinum resistance sensors are deployed at 3 positions of the screw extrusion section to collect the melt temperature, the collection frequency is 1 Hz, and the transmission is through 4-20 mA analog signal conversion to Ethernet; HBMS9M tension sensors are deployed at the godet of the spinning reel to collect the filament tension, the collection frequency is 1 Hz, and the transmission mode is the same as above; 1024-line incremental encoders are deployed at the screw drive motor of the spinning machine to collect the extrusion speed, the collection frequency is 1 Hz, and the transmission is through pulse signal conversion to Ethernet.

[0018] In the stranding section, 1024-line incremental encoders are deployed at the main shaft of the stranding machine to collect the stranding speed, the collection frequency is 1 Hz; HBMS9M tension sensors are deployed at the take-up end of the stranding machine to collect the cable core tension, the collection frequency is 1 Hz; Keyence IL-600 laser displacement sensors are deployed at the wire arrangement mechanism of the stranding machine to collect the stranding pitch, the collection frequency is 0.5 Hz, and the transmission is through Ethernet.

[0019] In the sheath extrusion section, PT100 platinum resistance sensors are deployed at the extruder barrel and die to collect extrusion temperature at a frequency of 1Hz, and transmitted via Ethernet through a 4-20mA analog signal; a Kistler 9021 piezoelectric pressure sensor is deployed at the extruder die inlet to collect extrusion pressure at a frequency of 1Hz, and transmitted via Ethernet; a 1024-line incremental encoder is deployed at the sheath traction machine to collect traction speed at a frequency of 1Hz, and transmitted via Ethernet through a pulse signal.

[0020] It should be further explained that the collected macroscopic data needs to be preprocessed through edge computing nodes, and a moving average filter with a window size of 5 sampling points is used to remove sensor noise; based on the timestamp of the product unit entering the workstation, the asynchronous data of different sensors are aligned to the same time axis, and the time accuracy is controlled within ±100ms to ensure the consistency of the timing of macroscopic data of a single cable.

[0021] It should be specifically noted that the microscopic response signal directly characterizes the intrinsic physicochemical state of the LCP material. A non-contact sensing device is used to avoid interference with the production process. Before winding in the spinning section, an Axometrics AxoScan laser polarimeter is deployed to calculate the molecular orientation degree by measuring the birefringence effect of polarized light on the LCP monofilament. The molecular orientation degree is defined as the degree of order in which the molecular chains are arranged along the axis of the monofilament, with a value range of 0 to 1. The higher the value, the more regular the arrangement. The acquisition frequency is 0.2 Hz, once every 5 seconds. The signal is output digitally through an RS485 interface. The qualified molecular orientation degree of LCP cable must be ≥0.75.

[0022] At the die exit of the sheath extrusion section, a Photron FASTCAM Mini UX100 high-speed camera was deployed, along with a ring LED light source, to capture dynamic images of the LCP melt wrapping the cable core. The image resolution was 1280×720, and the frame rate was 50fps. The acquired images were processed using the Python OpenCV library to extract the melt flow boundary contour and calculate the melt thickness uniformity and interface bonding clarity. The melt thickness uniformity was defined as the ratio of the maximum to the minimum melt thickness in the image, with a acceptable range of 1.0~1.2. The interface bonding clarity was calculated by using an edge detection algorithm to determine the grayscale difference between the sheath and the core layer. A grayscale difference ≥30 was considered a good bond.

[0023] It should be specifically noted that the intelligent identification code adopts an ultra-high frequency readable and writable RFID tag, model ImpinjMonza R6. This tag has the characteristics of high temperature resistance, re-reading and rewriting capability, ≥100,000 erase and write cycles, and a reading distance of 1~3m. The specific application process is as follows: before the LCP resin raw material enters the spinning section, the RFID tag is fixed to the side of the pay-off reel that carries a single cable. The initial code is written through the Impinj R2000 RFID card writer. The code format is production date + batch number + workstation number + serial number, such as 20250320-B01-S01-0001, realizing the "one item, one code" binding for each product unit.

[0024] It should be further explained that after the product unit completes the spinning process, the edge computing node writes the spinning process execution summary and recommended downstream process parameters into the RFID tag. When the product unit enters the stranding section, the RFID reader deployed at the entrance of the stranding machine automatically reads the data stored in the tag, which serves as the decision basis for the stranding section sub-controller. If the reading shows that the spinning molecular orientation degree is only 0.72, which is lower than the qualified threshold of 0.75, the stranding section sub-controller will automatically reduce the stranding tension by 10N to avoid the problem of yarn breakage.

[0025] Dynamic causal graph construction and update module: Based on the intelligent identification code, it integrates macro and micro datasets of the same product unit to construct and continuously update a dynamic causal graph of the manufacturing process. By analyzing real-time data streams, it dynamically identifies and quantifies key causal paths that affect the quality of the final product. It should be specifically noted that the nodes of the dynamic cause-effect graph are classified into three categories: process parameter nodes P, microscopic state nodes M, and product performance nodes Q. Among them, the process parameter nodes include spinning temperature P1, spinning tension P2, spinning speed P3, stranding speed P4, stranding tension P5, extrusion temperature P6, extrusion pressure P7, and traction speed P8; the microscopic state nodes include molecular orientation degree M1, melt uniformity after spinning, and after extrusion M2; the product performance nodes include product tensile strength Q1, finished product tensile test, dielectric constant Q2, finished product electrical test, sheath adhesion Q3, and finished product peel test.

[0026] It should be further explained that the edges of the causal graph are directed edges, which are used to represent the causal relationship between nodes. The direction of the edge is from the cause node to the result node. For example, P1→M1 indicates that the spinning temperature has a causal effect on the molecular orientation degree, and M1→Q1 indicates that the molecular orientation degree has a causal effect on the product's breaking strength.

[0027] It should be specifically noted that the construction of the dynamic cause-effect graph of the manufacturing process adopts a method that integrates the process mechanism model and the cause-effect discovery algorithm based on the conditional independence test. First, the initial edge structure is defined according to the manufacturing mechanism of LCP material. For example, in the LCP melt spinning process, the temperature rise will reduce the melt viscosity and promote the alignment of molecular chains along the stretching direction, so P1→M1 is defined; excessive stranding tension will cause the single filament to break and reduce the overall strength of the cable core, so P5→Q1 is defined; too low extrusion temperature will cause poor melt fluidity and affect the bonding quality between the sheath and the core layer, so P6→M2 and M2→Q3 are defined.

[0028] Subsequently, the PC algorithm was used, with historical production data from the past three months, including complete data from 5000 product units, as input. The initial edge structure was verified and corrected through conditional independence tests (using chi-square test, significance level α=0.05). For example, the initial hypothesis was P3→M1, but the test found that under the condition of controlling P1, the conditional mutual information between P3 and M1 approached 0, indicating that the influence of spinning speed on molecular orientation was negligible. Therefore, the edge P3→M1 was deleted. During the test, it was also found that under the condition of controlling M1, the conditional mutual information between P2 and Q1 was still significantly >0.3, indicating that the spinning tension not only indirectly affects Q1 through M1, but also has a direct influence. Therefore, the edge P2→Q1 was added.

[0029] Edge weights are quantized using standardized mutual information, and the calculation formula is as follows: Wherein, I(X,Y) is the mutual information between variables X and Y, used to measure the degree of correlation between variables; H(X) is the entropy of variable X, used to measure the uncertainty of the variable; H(Y) is the entropy of variable Y; the standardized mutual information ranges from 0 to 1, and the larger the value, the stronger the causal influence. For example, the NMI of P1→M1 is 0.82, indicating that the spinning temperature has a very strong influence on the molecular orientation degree; the NMI of P4→Q1 is 0.15, indicating that the twisting speed has a weak influence on the breaking strength of the product.

[0030] It should be further explained that the continuous update cycle of the cause-effect graph is set to trigger an incremental update every 100 product units produced. The update process is as follows: First, extract the macroscopic data, microscopic data, and product performance data of the latest 100 product units; use the incremental PC algorithm to verify the edge structure of the existing cause-effect graph. If the conditional independence test of an edge in the new data fails, the edge is deleted; if a new significant correlation is found, an edge is added; recalculate the standardized mutual information weights of all edges and replace the original weights; store the updated cause-effect graph on the central server for use by other modules.

[0031] It should be specifically explained that the identification of the critical causal path starts from the product performance defect node, traverses all incoming edges in reverse, and calculates the path contribution. The specific steps are as follows: First, define the causal contribution: the causal contribution of a path is equal to the product of the standardized mutual information weights of all edges in that path. For example, the contribution of the path P1→M1→Q1 = the weight of P1→M1 × the weight of M1→Q1. Then, starting from the product performance defect node (e.g., Q1<500MPa, i.e., the product's fracture strength is unqualified), traverse all incoming edges in reverse and trace the upstream node of each edge to form... All potential causal paths, such as P1→M1→Q1, P2→Q1, P5→Q1, P6→M2→Q3→Q1, etc., are considered. Finally, a total contribution threshold of 0.8 is preset. The contribution of each potential path is calculated and summed. Paths with a total contribution ≥ 0.8 are identified as critical causal paths. For example, the contribution of P1→M1→Q1 is 0.615, and the contribution of P2→Q1 is 0.22. The sum of the two is 0.835 ≥ 0.8. Therefore, these two paths are identified as critical causal paths, meaning that spinning temperature and spinning tension are the core factors causing the product's breaking strength to fail the test.

[0032] Collaborative optimization digital twin module: integrates the dynamic causal graph as its inference kernel, includes a virtual simulation environment, is used to perform forward-looking performance evaluation of key process decisions based on the current causal graph, and adaptively calibrate the causal relationship model according to the difference between the evaluation results and physical feedback; It should be specifically noted that the virtual simulation environment is constructed using a hybrid model based on physical mechanisms and historical data training: First, a 1:1 scale 3D model of the LCP cable manufacturing production line is constructed using Unity 2022. The model includes detailed structures of equipment such as spinning machines, stranding machines, and extruders, such as the thread structure of the spinning machine screw and the flow channel shape of the extruder die. A mapping relationship between physical parameters and virtual parameters is established. For example, the temperature sensor data of the physical spinning machine is directly mapped to the temperature field parameters of the virtual spinning machine screw, and the melt image captured by the physical high-speed camera is mapped to the melt flow simulation parameters at the virtual extruder die outlet.

[0033] It should be further explained that the hybrid model comprises two parts: a mechanistic model and a data-driven model. The mechanistic model uses ANSYS Polyflow to establish an LCP melt flow model based on the Navier-Stokes equations to calculate the velocity and pressure fields of the melt during extrusion. The Doi-Edwards molecular chain model is used to calculate the molecular orientation degree M1 during spinning. The calculation formula is as follows: Where θ is the angle between the molecular chain and the spinning axis. <cos 2θ> is the average value of the square of the cosine of the included angle, calculated by the spinning temperature P1 and spinning tension P2 using the Doi-Edwards model. The data-driven model uses an LSTM neural network, with the input being the values ​​of process parameter nodes P1-P8 and microstate nodes M1-M2, and the output being the predicted values ​​of product performance nodes Q1-Q3. The training set is 10,000 product units using historical qualified data from the past 6 months, and the Adam optimizer is used for training until the validation set error is <5%, which is used to quickly predict product performance.

[0034] When the hybrid model is fused, the values ​​of the micro-state nodes (M1, M2) are first calculated through the mechanism model. Then, the values ​​of M1, M2 and process parameter nodes are input into the data to drive the model and output the predicted values ​​of the product performance nodes. This balances the accuracy of the mechanism and the speed of prediction, and the time taken for a single simulation is controlled within 10 seconds.

[0035] It should be further explained that the forward-looking performance assessment explores a better process window by conducting virtual disturbance simulations on key process parameters. The specific process is as follows: First, based on the process safety boundary and historical optimization experience, the disturbance range of key process parameters is set, such as spinning temperature P1±10℃, spinning tension P2±15N, and extrusion pressure P7±2MPa, to ensure that the disturbance range does not exceed the equipment's load-bearing limit, such as P1 not exceeding 320℃, to prevent LCP resin degradation.

[0036] The evaluation process is as follows: Input the process parameters of the current product unit and the critical path in the dynamic cause-effect graph; perform multiple perturbations on the critical process parameters while keeping other parameters unchanged, and calculate the values ​​of the microstate node M1 and the product performance node Q1 corresponding to each perturbation through a hybrid model; with the goal of maximizing Q1 (≥600MPa) and M1≥0.8, select the optimal parameter combination. For example, when P1=300℃, M1=0.83 and Q1=620MPa, this parameter combination is the optimal solution, and a parameter-performance evaluation report is generated and output to the distributed decision control module.

[0037] It should be specifically noted that the adaptive calibration of the causal relationship model is achieved by comparing the virtual simulation results with the actual feedback results from the physical production line. The specific steps are as follows: First, calculate the relative deviation, using the formula: The preset deviation tolerance is 5%. If the relative deviation persists for 3 consecutive times (i.e., 3 consecutive product units exceed 5%), calibration is triggered. If a single deviation exceeds 10%, calibration is triggered immediately. The calibration process is as follows: Locate the deviation source and trace the node with the largest predicted deviation through a dynamic causal graph. If the predicted value of M1 deviates the most from the actual value, the deviation source is determined to be the causal edge "P1→M1". Perform incremental learning adjustment, extract data from 50 product units during the deviation period, recalculate the standardized mutual information weights of the deviation source edge, and verify the prediction deviation of the adjusted model. If the deviation drops to 3% < 5% after adjustment, the calibration is complete. Replace the original model with the adjusted causal graph to ensure the accuracy of subsequent simulation predictions.

[0038] Distributed decision control module: Based on the aforementioned forward-looking performance evaluation results, it generates local optimization instructions for specific workstations and global optimization strategies that consider cross-workstation coupling effects, and performs closed-loop control on the manufacturing process.

[0039] It should be further explained that one PLC-based edge controller, model Siemens S7-1200, is deployed in each of the spinning, stranding, and sheathing extrusion sections. The sub-controller is directly connected to the station sensors and actuators and is responsible for local decision-making at its station.

[0040] It should be specifically explained that the generation logic of the local optimization instruction is as follows: the sub-controller generates instructions based on the evaluation report of the local causal chain and twin in the dynamic causal graph using a PID control algorithm; for example, if the spinning section sub-controller receives the evaluation result of "optimal P1=300℃", and the current actual P1=295℃, it calculates the opening adjustment of the temperature control valve through the PID algorithm and generates the instruction "adjust the setting value of the spinning machine screw to 300℃, adjustment rate 2℃ / min".

[0041] The instruction execution and feedback process is as follows: the sub-controller sends the instruction to the actuator and collects the adjusted P1 and M1 data in real time. If M1 reaches 0.83 within 10 minutes and meets the target requirement, the local optimization is completed; if M1 does not meet the target, the instruction is recalculated until the target requirement is met.

[0042] It should be further explained that a global optimizer is deployed on the central server, and a multi-objective genetic algorithm is used to find the Pareto optimal solution with the objectives of "maximizing the overall production efficiency, maximizing the product qualification rate, and minimizing energy consumption".

[0043] It should be specifically explained that the generation logic of the global optimization strategy is as follows: input the local optimization requirements of each sub-controller and the cross-station coupling relationship in the dynamic cause-effect graph; set constraints, including the safety range of each process parameter and product quality constraints; iterate 100 times through the NSGA-II algorithm to obtain the Pareto optimal solution set; and the production management personnel determine the final global strategy according to the current priority. The cross-workstation collaboration process is as follows: the central optimizer decomposes the final global strategy into coordination instructions for each work section. After receiving the coordination instructions, the sub-controllers of each work section adjust the local optimization targets to ensure the coordination of parameters throughout the entire process.

[0044] It should be further explained that the closed-loop control cycle of the manufacturing process is set to 1 minute, and the process is as follows: First, data acquisition is performed. Sensors at each workstation collect data at a frequency of 1Hz, which is preprocessed by the edge computing node and then uploaded to the central server. Next, causal analysis is performed. The dynamic causal graph module analyzes the real-time data to identify whether there are any deviations from the acceptable range of parameters. If so, it is marked as an "abnormal warning". Then, a digital twin evaluation is performed. Both abnormal warning data and normal data are prospectively evaluated through a collaborative optimization digital twin to determine whether the current parameters need to be adjusted. Then, decision generation is performed. If adjustment is required, local optimization instructions and global optimization strategies are generated. If the parameters are normal, the current settings are maintained. Finally, execution feedback is performed. After the actuator executes the instructions, the sensors collect the adjusted parameter data and feed it back to the data acquisition stage to form closed-loop control.

[0045] For example, when an alarm is triggered that the extrusion pressure P7 = 14MPa > 15MPa, the closed-loop process is as follows: Data is collected (P7 = 14MPa, M2 = 1.25), exceeding the acceptable range by 1.2; a cause-effect graph analysis is performed (P7 → M2 → Q3), determining that an abnormal P7 will cause a decrease in Q3; a twin simulation is performed (P7 = 12MPa), where M2 = 1.15 and Q3 = 55N, meeting the acceptable requirements; a local command is generated to reduce the extrusion pressure to 12MPa and a global strategy is generated to reduce the traction speed P8 from 1.5m / min to 1.4m / min; after the actuator executes the command, P7 = 12MPa and M2 = 1.15 are collected and fed back to the data acquisition stage, completing the closed-loop control.

[0046] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A digital-twin-based LCP cable manufacturing whole-process data management and control system, characterized in that, Comprise: Process coupling multi-modal perception module: for synchronously collecting macroscopic process timing data and microscopic response signal set of each product unit, and binding an intelligent identification code readable and writable for each product unit; Dynamic causal graph construction and updating module: for fusing macroscopic and microscopic data sets of the same product unit based on the intelligent identification code, constructing and continuously updating the dynamic causal graph of the manufacturing process, dynamically identifying and quantifying the key causal path affecting the final product quality by analyzing real-time data flow; Collaborative optimization digital twin module: integrating the dynamic causal graph as its reasoning kernel, containing a virtual simulation environment for forward-looking performance evaluation of key process decisions based on the current causal graph, and self-adaptive calibration of the causal relationship model according to the difference between the evaluation results and the physical feedback; Distributed decision control module: for generating local optimization instructions for specific workstations and global optimization strategies considering cross-station coupling effects according to the forward-looking performance evaluation results, and performing closed-loop control on the manufacturing process.

2. The digital-twin-based LCP cable manufacturing whole-process data management and control system according to claim 1, characterized in that: The macroscopic process timing data includes temperature, tension, pressure and rotating speed collected from spinning, twisting to sheath extrusion section; The microscopic response signal set includes signals directly representing the intrinsic physical and chemical state of the material obtained by non-contact sensing devices integrated in the equipment, including molecular orientation degree signals reflecting the arrangement state of LCP molecular chains, and melt rheological image signals reflecting the interface morphology of the sheath and core layer.

3. The LCP cable manufacturing whole-process data management and control system based on digital twinning according to claim 1, characterized in that: The intelligent identification code is a readable and writable RFID tag; as a mobile data carrier, it writes process execution abstract and recommended downstream process parameters in the manufacturing process, which are read by the decision control module of the downstream workstation as the basis for decision-making.

4. The digital-twin-based LCP cable manufacturing whole-process data management and control system according to claim 1, characterized in that: The construction of the dynamic causal graph of the manufacturing process refers to constructing a topological network with process parameters and material state as nodes and causal influence relationship between nodes as directed edges; By fusing process mechanism model and causal discovery algorithm based on conditional independence test, the existence and weight of causal edges between nodes are automatically learned and quantified from the fused macroscopic and microscopic data.

5. The digital-twin-based LCP cable manufacturing whole-process data management and control system according to claim 1, characterized in that: The key causal path is the path that starts from the final product quality defect or performance index, traverses all incoming edges in reverse, and calculates the causal contribution degree of each potential path. The path that is composed of multiple causal edges and whose contribution degree exceeds the preset threshold is determined as the key causal path.

6. The digital-twin-based LCP cable manufacturing whole-process data management and control system according to claim 1, characterized in that: The virtual simulation environment is a hybrid model based on physical mechanism and historical data training, which can receive real-time data from the dynamic causal relationship engine and perform virtual perturbation simulation of key process parameters beyond the current actual setting range to explore a better process window.

7. The digital-twin-based LCP cable manufacturing whole-process data management and control system according to claim 1, characterized in that: The self-adaptive calibration of the causal relationship model compares the forward-looking evaluation results of the virtual simulation environment with the actual feedback results of the physical production line, and when the deviation continuously exceeds the preset tolerance, triggers the incremental learning and adjustment of the weight and structure of the corresponding causal edge in the dynamic causal graph.

8. The digital-twin-based LCP cable manufacturing whole-process data management and control system according to claim 1, characterized in that: The local optimization instruction is generated by the sub-controller deployed at each station, which focuses on optimizing the core quality indicators of the station and responds quickly based on the local causal chain directly related to the station in the dynamic causal diagram.

9. The digital-twin-based LCP cable manufacturing whole-process data management and control system according to claim 1, characterized in that: The global optimization strategy is generated by the central optimizer, which aims to coordinate the local optimization instructions by analyzing the complex coupling relationship across stations in the dynamic causal diagram and solving the Pareto optimal solution of the set points of each station to optimize the overall performance and production efficiency of the whole process.

Images