An automatic control system for ultra-thin rubber tube production

Abstract

The present application relates to the field of industrial automation and control technology, and more particularly to an automatic control system for ultra-thin rubber tube production, which comprises a real-time measurement subsystem, a data processing module, a twin updating module, a predictive control engine, an adaptive control module, and a detection and early warning module. The present application significantly improves the precision, stability, yield, and automation level of ultra-thin rubber tube production, and reduces the dependence on human experience.

Description

Technical Field

[0001] This invention relates to the field of industrial automation and control technology, and in particular to an automatic control system for the production of ultra-thin rubber hoses. Background Technology

[0002] Ultrathin rubber tubing, as a critical industrial consumable, plays an indispensable role in many high-tech fields such as medical devices, automotive manufacturing, electronic components, and aerospace. Its unique flexibility, sealing properties, insulation, and extremely low wall thickness requirements give it significant advantages in applications with strict limitations on space, weight, and performance. Currently, the production of ultrathin rubber tubing typically relies on a relatively mature automated control system, which usually covers core processes such as material extrusion, molding, cooling, traction, and cutting.

[0003] However, with the continuous development of technology and the increasingly stringent requirements for the performance indicators of ultra-thin rubber tubes in application scenarios, the existing technology is gradually showing its limitations. The reason is that the definition of high precision in the production of ultra-thin rubber tubes has far exceeded the traditional standards. Its requirements for wall thickness uniformity, inner diameter concentricity, dimensional stability, etc., have reached the micron or even sub-micron level. Against this background, the deep contradictions faced by the existing automated control system are gradually becoming prominent. Summary of the Invention

[0004] The purpose of this invention is to provide an automatic control system for the production of ultra-thin rubber tubes, in order to solve the technical problems of insufficient production accuracy and poor stability caused by the high sensitivity of materials, strong coupling of multiple parameters, and limitations of traditional reactive control strategies when manufacturing micron- or submicron-level ultra-thin rubber tubes.

[0005] This invention provides an automated control system for the production of ultra-thin rubber hoses, comprising: a real-time measurement subsystem for data exchange and command transmission via a high-speed industrial Ethernet protocol, a data processing module, a twin update module, a predictive control engine, an adaptive control module, and a detection and early warning module. All data collected by the real-time measurement subsystem is transmitted to the data processing module.

[0006] The real-time measurement subsystem is set at key process nodes in the ultra-thin rubber tube production line to acquire real-time data on the geometric dimensions, internal structure, and process parameters of the ultra-thin rubber tube.

[0007] The data processing module is configured to receive the high-speed raw data stream generated by the real-time measurement subsystem, preprocess the data, and generate a real-time updated process state feature vector.

[0008] The twin update module is used to construct a virtual simulation model of the entire process of ultra-thin rubber tube extrusion, cooling, and traction. The virtual simulation model reflects and predicts the physical behavior in the actual production process in real time, and updates the key material parameters in the virtual simulation model online in real time based on the process state feature vector.

[0009] The predictive control engine, based on the virtual simulation model provided by the twin update module, performs forward optimization of multiple coupled process parameters to calculate the optimal control action sequence.

[0010] The adaptive control module receives control commands from the predictive control engine and converts them into physical actions to achieve micron-level and millisecond-level dynamic adjustment.

[0011] The detection and early warning module enables predictive early warning and self-correction guidance for potential defects and system anomalies through in-depth analysis of real-time data streams.

[0012] In some embodiments, the real-time measurement subsystem includes at least: a laser triangulation sensor array, uniformly distributed radially along the ultrathin rubber tube, and installed approximately 50 mm to 150 mm downstream of the extrusion die outlet;

[0013] An optical coherence tomography (OCT) module, positioned at the beginning of the cooling section, performs a penetrating scan of the pipe using the principle of low coherence interference without contacting the pipe. A high-frequency ultrasonic sensor, installed at the entrance of the traction section, assists in detecting internal defects or local density inconsistencies in the pipe. A traction tension sensor, integrated on the roller of the traction device, monitors the axial tension experienced by the ultra-thin rubber tube during traction in real time. A melt pressure sensor, installed at the die entrance of the extruder head, monitors the rubber melt pressure inside the extrusion die in real time.

[0014] In some embodiments, the data processing module processing flow includes at least:

[0015] The raw displacement data of the laser triangulation sensor array is denoised, and geometric characteristic parameters such as real-time outer diameter, outer diameter fluctuation rate, and ellipticity are calculated.

[0016] Image enhancement, edge detection, and region segmentation are performed on the raw two-dimensional or three-dimensional image data output by the optical coherence tomography module to accurately identify and extract the inner diameter, wall thickness distribution, and potential internal defect features.

[0017] Time-domain or frequency-domain analysis is performed on the echo signal received by the high-frequency ultrasonic sensor to identify anomalies in signal attenuation or propagation speed, thereby extracting the type, location, and size characteristics of internal defects.

[0018] The analog signals output by the traction tension sensor and the melt pressure sensor are digitally converted and filtered, and the tension fluctuation amplitude and pressure fluctuation amplitude are calculated.

[0019] In some embodiments, the twin update module includes at least:

[0020] A coupled rheology-thermodynamics model, constructed based on the finite element method, is used to simulate the non-Newtonian rheological behavior of rubber materials within an extrusion die, the heat exchange process in the cooling section, and the viscoelastic deformation in the traction section. A parameter adaptive identification unit receives real-time process state feature vectors from the data processing module and compares them with the predicted output of the coupled rheology-thermodynamics model. Based on the comparison results, an adaptive Kalman filter algorithm is used to update the key material parameters in the coupled rheology-thermodynamics model online in real time. The correction period of the parameter adaptive identification unit is no more than 100 milliseconds.

[0021] In some embodiments, the coupled rheology-thermodynamics model strongly couples the viscoelastic constitutive relation of rubber with the temperature field, stress field, velocity field, and pressure field for calculation; the input parameters of the model include the shear viscosity, tensile viscosity, thermal conductivity, specific heat capacity, density, elastic modulus, Poisson's ratio of the rubber material, as well as the geometry of the extrusion die, the temperature of each section, and the traction speed.

[0022] In some embodiments, the predictive control engine includes at least a model predictive control algorithm kernel, a feedforward compensation unit, and a control sequence output unit.

[0023] In some embodiments, the adaptive control module includes at least:

[0024] The mold gap adjustment unit has an adjustment accuracy of 0.1 micrometers and a response time of less than or equal to 20 milliseconds.

[0025] The propulsion speed control unit is equipped with a traction roller group driven by an AC servo motor;

[0026] The extrusion pressure control unit is used to combine the real-time data from the melt pressure sensor and apply an adaptive PID control algorithm to control the melt pressure fluctuation at the extrusion die inlet within ±0.05 MPa.

[0027] The segmented dynamic cooling rate control unit consists of multiple independently temperature-controlled cooling zones.

[0028] In some embodiments, the detection and early warning module includes at least:

[0029] The multidimensional data fusion and identification unit receives real-time process state feature vectors from the data processing module and uses machine learning algorithms to perform real-time analysis on the fused multidimensional time series data.

[0030] The hierarchical predictive early warning unit identifies early signs of system failure based on the analysis results of the multi-dimensional data fusion and identification unit, and issues hierarchical early warnings.

[0031] When a slight or moderate abnormal pattern is detected, the self-calibrating guidance unit feeds back the identified abnormal pattern and its possible causes to the predictive control engine.

[0032] In some embodiments, a human-computer interaction and visualization interface module is also included, which is used by operators to intuitively monitor the production process.

[0033] Compared with existing technologies, this invention has the following advantages: the real-time measurement subsystem provides comprehensive, high-resolution real-time data; the twin update module builds a precise understanding and adaptive prediction capability for complex processes; the predictive control engine achieves coordinated and forward-looking optimization control of multiple coupled parameters, overcoming the limitations of traditional independent PID control; the adaptive control module ensures precise execution at the micron and millisecond levels; the detection and early warning module transforms post-event response into pre-event prediction and self-correction; this invention fundamentally improves the precision, stability, yield, and automation level of ultra-thin rubber tube production. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is an overall structural diagram of the system of the present invention;

[0036] Figure 2 This is the control logic diagram of the system of the present invention. Detailed Implementation

[0037] The following will refer to the appendices in the embodiments of the present invention. Figures 1-2 The technical solutions in the embodiments of the present invention will be clearly and completely described together. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0038] To better understand this invention, it should be noted that the production of ultra-thin rubber tubes demands a level of precision far exceeding traditional standards. Specifically, due to their extremely low wall thickness, ultra-thin rubber tubes are exceptionally sensitive to minute fluctuations in temperature, pressure, and tension during extrusion, traction, and cooling processes. The viscoelastic characteristics of rubber materials mean that their deformation after being subjected to external forces often exhibits hysteresis, and a nonlinear relationship exists between deformation and stress. Existing distance measurement systems typically employ discrete or periodic measurements, resulting in relatively long feedback cycles. Combined with the viscoelasticity of rubber materials, when the system detects dimensional deviations and makes adjustments (such as adjusting the die gap or changing the feed speed), its response is often delayed. Furthermore, the adjustment action itself may introduce new fluctuations into other parts of the tube, making it difficult to achieve continuous, uniform micron-level high-precision control. This easily leads to phenomena such as localized excessive thinness, out-of-round inner diameter, or uneven wall thickness.

[0039] In the production of ultra-thin rubber hoses, the various control variables are not independent but highly coupled. For example, simply adjusting the height of the extrusion die (i.e., the die gap) to control the wall thickness directly affects the pressure at the extrusion head and the material flow rate. Changing the feed speed to maintain a stable outer diameter directly affects the residence time of the hose in the cooling zone and the load (tension) it bears, thus in turn affecting the uniformity of the wall thickness and inner diameter. Existing systems often use relatively independent PID (proportional-integral-derivative) control loops to manage these parameters separately. This decentralized control strategy is inadequate when dealing with the complex, nonlinear, and dynamically changing strong coupling relationships in the ultra-thin rubber hose production process. Optimizing a single parameter may trigger fluctuations in other related parameters, causing the system to oscillate between multiple performance indicators, making it difficult to achieve optimal performance simultaneously. This can even amplify errors, rendering pressure and load warning systems only as post-event remedies rather than proactive preventative mechanisms. This inherent limitation makes it difficult for the system to simultaneously guarantee high precision while maintaining production efficiency and product yield, increasing reliance on human experience and the complexity of debugging.

[0040] To meet the demands of ultra-thin rubber tubes for extreme uniformity and zero defect rate, existing control strategies often rely on reactive adjustments based on the current state, rather than predictive control based on process models and real-time data. Due to the complex rheological behavior of rubber materials and various uncertainties in the production process, the lack of forward-looking and adaptive control logic makes it difficult for the system to effectively compensate for the effects of external disturbances or internal parameter drift. For example, traditional ranging systems combined with simple controllers struggle to provide refined, feedforward compensation for the feed rate to eliminate uneven tube stretching caused by minute fluctuations in extrusion volume. This hysteresis directly results in the difficulty of continuously and stably manufacturing ultra-thin rubber tubes that meet micron-level precision requirements on the production line.

[0041] Based on the above, the present invention provides an automatic control system for the production of ultra-thin rubber tubes, which aims to solve a series of problems in the traditional production process, such as insufficient product precision, poor production stability and low efficiency caused by the high sensitivity of materials, strong coupling of multiple parameters and limitations of control strategies, especially for the precision manufacturing needs of micron-level and even submicron-level ultra-thin rubber tubes.

[0042] Example

[0043] The automatic control system for ultra-thin rubber tube production provided in this embodiment includes a real-time measurement subsystem, a data processing module, a twin update module, a predictive control engine, an adaptive control module, and a detection and early warning module. The modules are interconnected via a high-speed industrial Ethernet protocol, such as PROFINET IRT or EtherCAT, or a dedicated real-time bus, such as PCIe or Fibre Channel, to achieve efficient data exchange and command transmission. This ensures synchronous acquisition, parallel processing, in-depth analysis of data from all aspects of the system, as well as the coordinated issuance of control commands, thereby constructing a low-latency, highly reliable real-time control closed loop.

[0044] first,

[0045] The real-time measurement subsystem is deployed at multiple key process nodes in the ultra-thin rubber hose production line to achieve comprehensive, accurate, and real-time monitoring of the hose's geometry, internal structure, material stress state, and process parameters. The real-time measurement subsystem includes:

[0046] A laser triangulation sensor array is meticulously installed at a specific location approximately 75 mm to 125 mm downstream of the extrusion die exit. This array consists of at least six sets of high-performance non-contact laser displacement sensors, evenly distributed radially along the ultra-thin rubber tube and spaced at 60-degree angles around the outer circumference of the tube. Each sensor set employs digital signal processing technology based on a CMOS image sensor, possessing a synchronous sampling frequency of at least 15 kHz and an ultra-high measurement accuracy of ±0.3 micrometers. Simultaneously, the laser triangulation sensor array can continuously acquire real-time data on the external contour dimensions of the rubber tube at a rate of 300 frames per second, including but not limited to the absolute value of the instantaneous outer diameter, the amplitude of circumferential fluctuation, roundness deviation, and ellipticity, among other geometric characteristic parameters. By fusing data from multiple sensor sets, the cross-sectional shape of the tube can be accurately reconstructed.

[0047] Optical coherence tomography (OCT) modules, as non-contact optical imaging devices, are positioned at the beginning or middle of the cooling section, such as in the drying zone after the pipe leaves the initial cooling water bath. The OCT module employs swept-frequency OCT (SS-OCT) technology, utilizing the principle of low-coherence interference. It performs a through-scan scan of the pipe by emitting near-infrared light and receiving the reflected echo. Equipped with a high-speed two-dimensional scanning galvanometer, the OCT module can acquire real-time cross-sectional images of the pipe at a scanning frequency of at least 4 kHz. These image data are analyzed by a built-in dedicated image processing unit, allowing for precise measurement of the pipe's inner diameter, wall thickness, and circumferential uniformity. It can also identify micron-level defects within the pipe, such as tiny bubbles, impurities, or delamination. The OCT module boasts an axial resolution of up to 1 micrometer and a lateral resolution of up to 15 micrometers, ensuring accurate and non-destructive testing of ultra-thin wall structures (e.g., less than 200 micrometers thick).

[0048] In some preferred embodiments, a high-frequency ultrasonic sensor is installed at the entrance of the traction section, immediately adjacent to the end of the cooling section. The high-frequency ultrasonic sensor is made of piezoelectric composite material, with an operating frequency set between 10MHz and 18MHz and a pulse repetition frequency of 3kHz to 4kHz. By emitting high-frequency ultrasonic pulses and analyzing their propagation speed, attenuation characteristics, and echo time-domain waveform within the pipe, this module can assist in detecting internal defects in the pipe, such as micro-voids, uneven density, or minute cracks. For example, by calculating changes in the ultrasonic velocity, small differences in density deviations or material elastic modulus in localized areas can be inferred, thereby determining the internal structural integrity of the pipe. This complements the OCT module.

[0049] Furthermore, to achieve precise control of the pipe tension, a traction tension sensor is integrated into the guide wheel of the traction device or the bearing seat of the main traction roller. The traction tension sensor adopts the advanced bridge strain gauge principle, combined with a high-resolution differential amplifier and a high-speed analog-to-digital converter, to monitor the axial tension borne by the ultra-thin rubber tube during traction in real time. Its measurement range has been optimized and set between 0.05 Newtons and 75 Newtons, providing a measurement accuracy of ±0.03 Newtons, and has a fast response frequency of not less than 2kHz.

[0050] Furthermore, the melt pressure sensor is installed at the die inlet of the extruder die head, employing high-melting-point piezoresistive or capacitive sensor technology to monitor the instantaneous pressure of the rubber melt inside the extrusion die in real time. Its measurement range is 0 MPa to 25 MPa, with an accuracy of ±0.03 MPa, and a response frequency of at least 1 kHz. The real-time pressure data provided by the melt pressure sensor is crucial for evaluating extrusion stability, predicting die filling status, and controlling extrusion volume, especially for products requiring constant extrusion pressure to achieve uniform tube wall thickness.

[0051] It should be noted that all raw data collected by the real-time measurement subsystem, including high spatial resolution laser displacement data, high frame rate OCT image data, high sampling rate ultrasonic echo data, rapidly changing tension data, and instantaneous pressure data, are transmitted to the data processing module in the form of a continuous stream through high-bandwidth data interfaces, such as industrial gigabit Ethernet or PCIe bus, to ensure the integrity and real-time performance of data transmission.

[0052] Secondly

[0053] The data processing module serves as the core of the entire control system's data processing workflow, and its detailed process is as follows:

[0054] For the raw displacement data generated by the laser triangulation sensor array, multi-stage denoising processing is performed, including the use of an adaptive moving average filtering algorithm with a window size that can be dynamically adjusted according to the production speed, and a threshold denoising algorithm based on wavelet transform to effectively filter out high-frequency random noise and occasional spikes.

[0055] The denoised data is used to calculate a series of important geometric characteristic parameters of the ultrathin rubber tube, such as the real-time outer diameter, outer diameter fluctuation rate (e.g., by calculating the root mean square deviation between the instantaneous outer diameter and the target outer diameter), ellipticity (by fitting a least square ellipse), and instantaneous roundness deviation. The accuracy of these parameters directly affects the accuracy of downstream control.

[0056] For the raw 2D or 3D image data stream output by the OCT module, advanced image processing algorithms are applied, including image enhancement techniques based on nonlocal mean filtering, to improve the image signal-to-noise ratio and contrast. An improved Canny operator or a deep learning-based U-Net semantic segmentation network is used for precise edge detection to identify the inner and outer wall boundaries of the pipe. Subsequently, through region segmentation and geometric morphology analysis, the inner diameter and wall thickness distribution of the pipe (including the minimum, maximum, and average wall thickness and their non-uniformity along the circumference) are accurately extracted. Furthermore, a feature point matching algorithm is used to identify potential internal defect features, such as the number, size, location, and shape of bubbles.

[0057] Before entering the analysis stage, the echo signal received by the high-frequency ultrasonic sensor undergoes bandpass filtering and signal enhancement processing, followed by time-domain analysis, such as calculating the signal transit time, peak amplitude, and pulse broadening, to identify anomalies in signal attenuation or propagation speed; simultaneously, frequency-domain analysis, such as Fast Fourier Transform (FFT), is performed to detect changes in specific frequency components. The aforementioned analysis aims to extract the type (e.g., bubbles or inclusions), location, and size characteristics of internal defects, and achieve preliminary defect classification by matching them with a pre-set defect feature library.

[0058] The analog signals output by the traction tension sensor and the melt pressure sensor are digitized by a high-precision analog-to-digital converter, and then smoothed by Kalman filtering or Wiener filtering to eliminate the measurement noise of the sensors themselves and the mechanical vibration interference in the production process. After that, the tension fluctuation amplitude, pressure fluctuation amplitude, and the mean, variance and other time-series characteristics of these parameters are calculated. These characteristics reflect the dynamic stability of the production process and the uniformity of material rheology.

[0059] More importantly, since different sensors are located at different positions on the production line and have different sampling frequencies, the data processing module uses high-precision GPS synchronization or PTP (Precise Time Protocol) timestamp alignment, and uses interpolation algorithms based on the current traction speed of the pipe and the physical position of each sensor or mapping algorithms based on the physical position to ensure that all parameters are correlated in the same production position and the same time reference, thereby solving the problems of data heterogeneity and spatiotemporal deviation.

[0060] then,

[0061] The twin update module establishes a highly realistic virtual simulation model of the entire process of ultra-thin rubber tube extrusion, cooling, and traction, enabling real-time reflection, accurate prediction, and adaptive learning of the actual production process. The twin update module includes:

[0062] The coupled rheology-thermodynamics model is developed based on advanced finite element method (FEM) or computational fluid dynamics (CFD) principles using professional simulation software platforms (such as ANSYS Polyflow or COMSOL Multiphysics). It simulates in detail the non-Newtonian rheological behavior of rubber materials in the extrusion die (e.g., shear thinning effect, inlet pressure loss), the complex heat exchange process in the multi-stage cooling section (including convective heat transfer, radiative heat transfer, and evaporative cooling), and the viscoelastic deformation and stress relaxation process experienced in the traction section.

[0063] The coupled rheology-thermodynamics model strongly couples the viscoelastic constitutive relations of rubber materials—for example, using the Ogden model, which is more suitable for large deformations, or the polynomial form of the Mooney-Rivlin model—with real-time temperature, stress, velocity, and pressure fields. The input parameters of the coupled rheology-thermodynamics model include at least accurately measured rubber material properties, such as shear viscosity curves at different temperatures, tensile viscosity curves, dynamic shear modulus, Poisson's ratio, thermal conductivity, specific heat capacity, density, coefficient of thermal expansion, crystallization kinetics parameters, the precise geometry of the extrusion die, the temperature distribution and flow rate of each cooling section, the rotational speed of the traction roller, the screw speed, and the melt inlet temperature, among other parameters.

[0064] The coupled rheology-thermodynamics model can accurately predict a series of key performance indicators of ultrathin rubber tubes, such as wall thickness distribution, inner diameter, outer diameter, concentricity, residual stress distribution, and surface quality (e.g., roughness, presence of defects caused by mold expansion), given the above combination of input parameters. Under steady-state conditions, the geometric dimension deviation of the coupled rheology-thermodynamics model can be controlled within ±0.5 micrometers, and the stress prediction deviation can be controlled within ±5%.

[0065] Furthermore, the system includes a parameter adaptive identification unit integrated within the twin update module. This unit receives real-time, high-dimensional process state feature vectors from the data processing module and precisely compares them with the current prediction output of the coupled rheology-thermodynamics model. Based on the comparison results, the parameter adaptive identification unit employs advanced adaptive Kalman filtering algorithms (e.g., extended Kalman filtering or unscented Kalman filtering), recursive least squares algorithms, or gradient descent-based deep neural network optimization algorithms to perform online real-time parameter identification and model correction. The aim is to update or correct key material parameters (e.g., viscoelastic parameters of rubber, shear dilution index, thermal conductivity, and cooling surface heat transfer coefficient) or environmental disturbance parameters (e.g., small fluctuations in ambient temperature, humidity, cooling water temperature, and friction coefficient changes caused by equipment wear) in the process twin update module. The correction cycle of the parameter adaptive identification unit is no more than 50 milliseconds, which means that in high-speed production processes, it can essentially achieve real-time response and adapt to external changes.

[0066] Secondly

[0067] The predictive control engine utilizes the virtual simulation model provided by the twin update module to achieve forward-looking and optimized control of multiple strongly coupled process parameters, including at least:

[0068] Model Predictive Control (MPC) algorithm, within each control cycle, utilizes a coupled rheological-thermodynamic model from a twin update module to predict the system behavior of the ultrathin rubber tube production process at multiple future time steps (e.g., a prediction time domain of 5 to 45 seconds and a control time domain of 1 to 5 seconds). The MPC algorithm determines the optimal control strategy by iteratively solving a rolling optimization problem. The objective of the rolling optimization problem is to minimize a carefully constructed multi-objective function that comprehensively considers multiple key performance indicators, including at least:

[0069] The geometric dimension target is to minimize the weighted square error of the wall thickness deviation, inner diameter deviation, outer diameter deviation, and concentricity deviation of the ultra-thin rubber tube relative to the preset target value. The weights can be dynamically adjusted according to the product's sensitivity to each dimensional parameter.

[0070] The mechanical stress target focuses on the weighted square error of traction tension fluctuation and extrusion pressure fluctuation relative to the target set value. The purpose is to maintain the mechanical stability of the production process and reduce the accumulation of internal stress in the pipe.

[0071] Energy efficiency targets are used to measure the smoothness of actuator movements, switching frequency, and total energy consumption. By introducing penalty terms, excessively frequent or drastic control actions are avoided, thereby extending equipment life and reducing operating costs.

[0072] During the optimization process, the MPC algorithm comprehensively considers multiple controlled variables, including but not limited to the die clearance of the extrusion die (multi-point local adjustment), the rotational speed of the traction roller, the temperature of each area of ​​the cooling section (e.g., the temperature and flow rate of cooling water or airflow), the rotational speed of the extruder screw, or the rotational speed of the melt gear pump. At the same time, the MPC algorithm strictly incorporates the physical constraints of the controlled variables and their rates of change (e.g., the upper and lower limits of the stroke and speed of the die adjustment mechanism, the upper and lower limits of the traction speed, the maximum rotational speed and torque of the extruder screw, and the range and rate of change of the cooling temperature), and calculates a series of optimal future control action sequences under the premise of satisfying all equipment operating constraints.

[0073] The feedforward compensation unit is used to address known or measurable disturbances upstream of the production line, thereby improving the system's disturbance immunity. The feedforward compensation unit receives disturbance measurement data from upstream of the production line, such as core parameters of the rubber compounding batch (e.g., Mooney viscosity, Shore hardness, elastic modulus), minor fluctuations in the extruder feed rate, changes in ambient temperature or humidity, and instantaneous fluctuations in cooling water temperature. Based on a coupled rheological-thermodynamic model, the feedforward compensation unit simulates the potential impact of disturbances on the entire system in real time, calculating pre-compensation amounts for key control variables (e.g., extruder screw speed, die clearance, traction speed). These pre-compensation amounts are applied to the system before the MPC algorithm calculates the optimal control sequence. By actively canceling out disturbance effects before they reach the measurement point, the system's hysteresis response to disturbances is significantly reduced, improving the real-time performance and robustness of the control.

[0074] The control sequence output unit is responsible for converting the first control action of the optimal control sequence calculated by the MPC algorithm in the current control cycle into a specific, formatted actuator instruction, and sending it to the adaptive control module via a low-latency real-time bus (such as EtherCAT or PROFINET IRT). In subsequent control cycles, the MPC algorithm will recalculate and predict the future system behavior based on the latest measurement data from the data processing module and the updated coupled rheology-thermodynamics model, thereby achieving rolling, real-time closed-loop control.

[0075] Then,

[0076] The adaptive control module receives fine-grained control commands from the predictive control engine and translates them into precise, rapid actions on the physical actuators, achieving micron-level geometric adjustment and millisecond-level dynamic response of process parameters, including at least:

[0077] The die gap adjustment unit includes multiple distributed, high-resolution piezoelectric ceramic actuators or high-thrust voice coil motors. The actuators are directly integrated into the die area of ​​the extrusion die and arranged in a ring array. Each actuator is coupled to a precision mechanical amplification mechanism (e.g., a lever-type fine-tuning mechanism or a wedge-type fine-tuning mechanism with a flexible hinge design) to achieve local, independent or collaborative fine-tuning of the die gap.

[0078] Piezoelectric ceramic actuators or voice coil motors, with positioning accuracy down to 5 nanometers, combined with mechanical amplification mechanisms, can achieve radial adjustment accuracy of 0.05 micrometers for the die clearance, with a fast response time of no more than 15 milliseconds. The die clearance adjustment unit can perform multi-point synchronous or asynchronous fine-tuning of the radial clearance of the die, such as through differential adjustment, to accurately compensate for wall thickness deviations caused by uneven material flow or die temperature gradients, thereby precisely controlling the wall thickness distribution and roundness of the ultra-thin rubber tube.

[0079] The propulsion speed control unit employs a precision traction roller assembly driven by a high-performance AC servo motor. The servo motor is equipped with a high-resolution absolute encoder, such as one with a resolution of up to 28 bits, ensuring that the rotational speed control accuracy of the traction rollers reaches 0.0005 rpm, thereby achieving millisecond-level precise control of the pipe's linear speed. Furthermore, the propulsion speed control unit integrates a high-bandwidth torque control loop. By providing real-time feedback data from the traction tension sensor, it achieves closed-loop precise control of the traction tension, with a torque response time of no more than 3 milliseconds. This means the system can react to tension fluctuations in an extremely short time, thereby actively eliminating pipe tension fluctuations while maintaining a stable traction speed, preventing excessive pipe stretching or localized buckling.

[0080] The extrusion pressure control unit precisely controls the speed of the extruder screw through a high-precision vector frequency converter, or in some embodiments, directly controls the speed of a high-precision melt gear pump. The extrusion pressure control unit combines real-time data from the melt pressure sensor and uses an adaptive PID control algorithm or an advanced fuzzy adaptive control algorithm to control the melt pressure fluctuation at the extrusion die inlet to within a very small range of ±0.03 MPa. Even when faced with changes in raw material viscosity or fluctuations in extruder thermal load, it can ensure the stability and continuity of the extrusion volume.

[0081] The segmented dynamic cooling rate control unit consists of multiple independently temperature-controlled cooling zones. These zones can be multi-segmented cooling water tanks or cooling airflow boxes that precisely control airflow temperature, humidity, and flow rate. Each cooling zone is equipped with an independent platinum resistance temperature sensor (e.g., a Pt100 sensor with an accuracy of ±0.05 degrees Celsius), a flow sensor, and a high-precision proportional valve or solenoid valve. The segmented dynamic cooling rate control unit achieves precise temperature control of each cooling zone through a predictive control strategy based on a coupled rheological-thermodynamic model or an adaptive PID control algorithm; for example, temperature fluctuations can be controlled within ±0.05 degrees Celsius. Through the aforementioned methods, the system in this embodiment can dynamically adjust the cooling curve of the pipe along the production line direction, optimizing its crystallization process, curing rate, and internal microstructure, thereby further stabilizing pipe dimensions, minimizing internal stress, and effectively preventing uneven dimensional shrinkage or warping deformation that may occur during the cooling process.

[0082] at last,

[0083] The detection and early warning module should include at least the following components:

[0084] The multi-dimensional data fusion and identification unit receives real-time process status feature vectors from the data processing module. These vectors include high-precision geometric dimensional parameters (such as outer diameter, inner diameter, wall thickness, and their distribution), internal defect features (such as the number, size, and location of bubbles), real-time traction tension, extrusion pressure, temperatures of each cooling section, and fluctuation information of each parameter. The multi-dimensional data fusion and identification unit employs advanced machine learning algorithms, such as multi-classifiers based on Support Vector Machines (SVM), stacked long short-term memory (LSTM) neural networks (particularly suitable for handling complex dependencies in time-series data), or Gaussian mixture models (GMM), for real-time analysis. By training and optimizing on a large amount of historical normal production data and defect sample data, the machine learning algorithms can identify weak, complex, and multi-dimensional patterns or abnormal trends related to defects in ultra-thin rubber tubes (e.g., localized excessive wall thickness, out-of-round inner diameter, micro-scratches on the surface, internal bubble clusters, and even precursors to impending tube rupture).

[0085] When the hierarchical predictive early warning unit and the multi-dimensional data fusion identification unit identify early signs that foreshadow future defects or system failures, the hierarchical predictive early warning unit will issue hierarchical early warning information based on the severity, development speed, and potential impact of the abnormal pattern. For example, it can be set up as a three-level early warning mechanism.

[0086] Slight fluctuations (system parameters deviate slightly but continuously, but have not yet affected product quality), moderate anomalies (parameter deviations exceed thresholds, product quality may be affected, requiring manual attention or system intervention), and severe malfunctions (parameter deviations are severe, product quality has been affected or the system is about to shut down, requiring immediate intervention).

[0087] Tiered early warning information is displayed to operators in real time and intuitively through a human-machine interface, and can trigger audible and visual alarms, or send detailed tiered early warning information to remote monitoring terminals or upper-level MES / ERP systems through an Industrial Internet of Things (IIoT) interface, so that managers can make timely decisions.

[0088] When the self-calibrating guidance unit detects a slight or moderate abnormal pattern, it does not immediately trigger a shutdown. Instead, it feeds back the identified abnormal pattern and its possible causes, obtained through source analysis (e.g., slight drift in coupled rheological-thermodynamic model parameters, minor changes in local material properties, minor disturbances in the external environment, or a decrease in control accuracy due to slight wear of an actuator), to the predictive control engine. The predictive control engine uses this feedback information to dynamically update the weights in its optimization objective (e.g., increasing the penalty weight for deviations in specific geometric dimensions), adjust constraints (e.g., temporarily tightening the fluctuation range of a parameter), or correct the gain parameters in the control strategy.

[0089] The aforementioned process can guide the system to make fine adjustments without interrupting production, achieve online self-repair, actively correct deviations in process parameters, thereby preventing further development or deterioration of defects, stabilizing the production process in an optimal state, minimizing scrap rate, and improving yield.

[0090] To better realize this invention, the automatic control system for ultra-thin rubber hose production also includes a human-machine interaction and visualization interface module. This module provides operators with an intuitive, comprehensive, and real-time monitoring and management platform for the production process. In some possible implementations, the human-machine interaction and visualization interface module can display, in a graphical and dashboard-like format, real-time the raw data from all sensors (e.g., laser scanning point clouds, OCT images), key characteristic parameters processed by the data processing module (e.g., real-time wall thickness distribution map, inner and outer diameter curves, tension and pressure fluctuation trends), predicted trends from the coupled rheological-thermo-mechanical model (e.g., prediction of future pipe size changes, prediction of residual stress), control output commands from the MPC algorithm, and early warning information and self-correction guidance suggestions generated by the detection and early warning module.

[0091] Operators can use the human-machine interface and visualization interface to set and modify control parameters (e.g., target size, traction speed, cooling temperature curve), view historical data curves and trend charts, analyze production reports and quality statistics, and perform fault diagnosis and system maintenance.

[0092] The present invention will now be described with reference to a specific practical embodiment. The automatic control system for the production of ultra-thin rubber tubes of the present invention is applied to the production of an ultra-thin catheter with an inner diameter of 2.000 mm, a wall thickness of 0.100 mm, and made of medical-grade silicone rubber. The silicone rubber has a Mooney viscosity ML(1+4) of 45 at 100°C and a Shore A hardness of 60. The extruder used is a single-screw extruder with a screw diameter of 30 mm and an L / D ratio of 28:1.

[0093] The real-time measurement subsystem is configured as follows:

[0094] The laser triangulation sensor array consists of eight SICK OD5000 series sensors, evenly distributed around the outer circumference of the pipe, with a sampling frequency of 20kHz and a measurement accuracy of ±0.2 micrometers.

[0095] The OCT module uses the Thorlabs Ganymede-II SS-OCT system with a scanning frequency of 5 kHz, an axial resolution of 1 micrometer, a lateral resolution of 12 micrometers, and an image processing delay of less than 10 milliseconds.

[0096] The high-frequency ultrasonic sensor uses the GE Inspection Technologies USM 36 series, with a probe frequency of 15MHz and a pulse repetition frequency of 4kHz.

[0097] The traction tension sensor is an HBM U9C series miniature force sensor with a measurement range of 0.1-50N, an accuracy of ±0.02N, and a response frequency of 2.5kHz.

[0098] The melt pressure sensor is a Dynisco PT4626 series, with a measurement range of 0-20MPa, an accuracy of ±0.02MPa, and a response frequency of 1.2kHz.

[0099] The data processing module is implemented using a hybrid architecture of FPGA and multi-core DSP. Laser data denoising uses a combination of 5-point moving average and Savitzky-Golay filtering. The image edge detection of the OCT module uses an improved Canny operator and a deep learning edge recognition model, achieving a data alignment accuracy of 0.5 milliseconds.

[0100] In the twin update module, the coupled rheological-thermodynamic model is built on the COMSOL Multiphysics platform, employing the Mooney-Rivlin constitutive model (C01=0.2MPa, C10=0.08MPa), combined with a non-Newtonian fluid (power-law exponent n=0.45) and a convective-radiative heat transfer model. The parameter adaptive identification unit uses the unscented Kalman filter (UKF) algorithm with an update cycle of 30 milliseconds, real-time correcting the viscosity coefficient and heat transfer coefficient of silicone rubber.

[0101] In the predictive control engine, the prediction time domain of the MPC algorithm is set to 20 seconds, and the control time domain to 3 seconds. In the optimization objective function, the weights for wall thickness deviation, inner diameter deviation, and tension fluctuation are set to 0.6, 0.3, and 0.1 respectively. The extrusion die gap adjustment range is ±0.2 mm, and the traction speed range is 0.5-5 m / min. The feedforward compensation unit monitors the batch viscosity changes of the compound and ambient temperature fluctuations in real time, and performs fine-tuning compensation for the extruder screw speed and traction speed in advance.

[0102] In the adaptive control module:

[0103] The mold gap adjustment unit uses eight Piezosystem Jena PSH 10 / 20 piezoelectric ceramic actuators to achieve a gap adjustment accuracy of 0.08 micrometers and a response time of 18 milliseconds through a lever mechanism.

[0104] The propulsion speed control unit uses a B&R ACOPOS P3 servo drive and synchronous motor, equipped with a 28-bit absolute encoder, with a traction roller speed control accuracy of 0.0003 rpm and a torque response time of 4 milliseconds.

[0105] The extrusion pressure control unit uses a Siemens S120 frequency converter to control the screw speed of the extruder, and combined with an adaptive PID algorithm, controls the melt pressure fluctuation within ±0.02MPa.

[0106] The segmented dynamic cooling rate control unit consists of three independent cooling water tanks, each equipped with a PT1000 sensor and a proportional valve controlled by Proportional-Integral-Derivative (PID), with a temperature control accuracy of ±0.03℃.

[0107] In the detection and early warning module, the multi-dimensional data fusion and identification unit employs an anomaly detection model based on an LSTM network. Input features include time-series data for 12 geometric dimensional parameters, 5 internal defect parameters, 2 tension / pressure parameters, and 3 temperature parameters. The model has been trained using 1000 hours of normal production data and 200 hours of known defect data. The hierarchical predictive early warning unit is set to a three-level warning system, with trigger thresholds set based on historical data and expert experience. The self-correcting guidance unit, upon detecting minor anomalies, feeds back to the predictive control engine, adjusting the target weights of wall thickness and concentricity by ±10%.

[0108] With the above configuration, the system of the present invention stably produces ultra-thin silicone rubber tubes at a traction speed of 3 meters per minute. After 24 hours of continuous production, a total of 4320 meters of tubes were produced.

[0109] To clearly demonstrate the superiority of the system provided by this invention, the inventors have provided a comparative embodiment, and the comparison results are shown in the table below:

[0110]

[0111] As can be seen from the table above, the automatic control system for ultra-thin rubber tubing production of the present invention exhibits significant advantages in all key performance indicators. Specifically, the system reduces the average outer diameter deviation, average inner diameter deviation, and average wall thickness deviation by 77.1%, 82.4%, and 85.7%, respectively, and reduces wall thickness non-uniformity and concentricity deviation by 82.3% and 79.4%, respectively, which is crucial for high-precision applications such as ultra-thin medical catheters. The internal bubble defect rate and surface scratch defect rate are reduced by 95.8% and 87.5%, respectively, which is directly attributed to the system's real-time detection capability and the predictive intervention of the self-correcting guidance mechanism for potential problems. The scrap rate is significantly reduced from 1.5% in the comparative example to 0.08% in the embodiment, which means a significant improvement in the yield rate. The traction tension fluctuation and extrusion pressure fluctuation are reduced by 80.0%, indicating that the system's dynamic stability control capability for the production process far exceeds that of traditional solutions. Due to the accurate process understanding provided by the twin update module and the optimization capability of the MPC algorithm, the product batch changeover time is shortened by 66.7%.

[0112] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0113] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. An automatic control system for the production of ultra-thin rubber hoses, characterized in that, include: The system comprises a real-time measurement subsystem, a data processing module, a twin update module, a predictive control engine, an adaptive control module, and a detection and early warning module, which exchange data and transmit commands via high-speed industrial Ethernet protocol. All data collected by the real-time measurement subsystem is transmitted to the data processing module. The real-time measurement subsystem is set at key process nodes in the ultra-thin rubber tube production line to acquire real-time data on the geometric dimensions, internal structure, and process parameters of the ultra-thin rubber tube. The data processing module is configured to receive the high-speed raw data stream generated by the real-time measurement subsystem, preprocess the data, and generate a real-time updated process state feature vector. The twin update module is used to construct a virtual simulation model of the entire process of ultra-thin rubber tube extrusion, cooling, and traction. The virtual simulation model reflects and predicts the physical behavior in the actual production process in real time, and updates the key material parameters in the virtual simulation model online in real time based on the process state feature vector. The predictive control engine, based on the virtual simulation model provided by the twin update module, performs forward optimization of multiple coupled process parameters to calculate the optimal control action sequence. The adaptive control module receives control commands from the predictive control engine and converts them into physical actions to achieve micron-level and millisecond-level dynamic adjustment. The detection and early warning module achieves predictive early warning and self-correction guidance for potential defects and system anomalies through in-depth analysis of real-time data streams; The twin update module includes at least: a coupled rheology-thermodynamics model, constructed based on the finite element method, used to simulate the non-Newtonian rheological behavior of rubber materials in the extrusion die, the heat exchange process in the cooling section, and the viscoelastic deformation in the traction section; and a parameter adaptive identification unit, which receives the real-time process state feature vector from the data processing module, compares it with the prediction output of the coupled rheology-thermodynamics model, and updates the key material parameters in the coupled rheology-thermodynamics model online in real time using an adaptive Kalman filter algorithm based on the comparison result; the correction period of the parameter adaptive identification unit is no more than 100 milliseconds. The coupled rheology-thermodynamics model strongly couples the viscoelastic constitutive relation of rubber with the temperature field, stress field, velocity field, and pressure field for calculation. The input parameters of the coupled rheology-thermodynamics model include the shear viscosity, tensile viscosity, thermal conductivity, specific heat capacity, density, elastic modulus, Poisson's ratio of the rubber material, as well as the geometry of the extrusion die, the temperature of each section, and the traction speed.

2. The system according to claim 1, characterized in that, The real-time measurement subsystem includes at least: a laser triangulation sensor array, which is uniformly distributed radially along the ultrathin rubber tube and installed 50 mm to 150 mm downstream of the extrusion die outlet; An optical coherence tomography (OCT) module, positioned at the beginning of the cooling section, performs a penetrating scan of the pipe using the principle of low coherence interference without contacting the pipe. A high-frequency ultrasonic sensor, installed at the entrance of the traction section, assists in detecting internal defects or local density inconsistencies in the pipe. A traction tension sensor, integrated on the roller of the traction device, monitors the axial tension experienced by the ultra-thin rubber tube during traction in real time. A melt pressure sensor, installed at the die entrance of the extruder head, monitors the rubber melt pressure inside the extrusion die in real time.

3. The system according to claim 2, characterized in that, The data processing module's processing flow includes at least the following: The raw displacement data of the laser triangulation sensor array is denoised, and the real-time outer diameter, outer diameter fluctuation rate, and ellipticity are calculated. Image enhancement, edge detection, and region segmentation are performed on the raw two-dimensional or three-dimensional image data output by the optical coherence tomography module to accurately identify and extract the inner diameter, wall thickness distribution, and potential internal defect features. Time-domain or frequency-domain analysis is performed on the echo signal received by the high-frequency ultrasonic sensor to identify anomalies in signal attenuation or propagation speed, thereby extracting the type, location, and size characteristics of internal defects. The analog signals output by the traction tension sensor and the melt pressure sensor are digitally converted and filtered, and the tension fluctuation amplitude and pressure fluctuation amplitude are calculated.

4. The system according to claim 1, characterized in that, The predictive control engine includes at least a model predictive control algorithm kernel, a feedforward compensation unit, and a control sequence output unit.

5. The system according to claim 2, characterized in that, The adaptive control module includes at least: The mold gap adjustment unit has an adjustment accuracy of 0.1 micrometers and a response time of less than or equal to 20 milliseconds. The propulsion speed control unit is equipped with a traction roller group driven by an AC servo motor; The extrusion pressure control unit is used to combine the real-time data from the melt pressure sensor and apply an adaptive PID control algorithm to control the melt pressure fluctuation at the extrusion die inlet within ±0.05 MPa. The segmented dynamic cooling rate control unit consists of multiple independently temperature-controlled cooling zones.

6. The system according to claim 1, characterized in that, The detection and early warning module includes at least: The multidimensional data fusion and identification unit receives real-time process state feature vectors from the data processing module and uses machine learning algorithms to perform real-time analysis on the fused multidimensional time series data. The hierarchical predictive early warning unit identifies early signs of system failure based on the analysis results of the multi-dimensional data fusion and identification unit, and issues hierarchical early warnings. When a slight or moderate abnormal pattern is detected, the self-calibrating guidance unit feeds back the identified abnormal pattern and its possible causes to the predictive control engine.

7. The system according to claim 1, characterized in that, It also includes a human-computer interaction and visualization interface module, which allows operators to intuitively monitor the production process.

Images