A real-time tension fluctuation detection method and device for cable processing speed optimization

Abstract

This invention provides a real-time tension fluctuation detection method and device for optimizing cable processing speed, relating to the field of automatic control technology for power cable extrusion production lines. The method includes: deploying multiple non-contact detection points along the production line, located at the extrusion nozzle of the extruder, the middle section of the vulcanizing tube, and before the traction machine; synchronously acquiring laser Doppler vibration signals and eddy current deformation signals at each detection point to obtain multimodal raw synchronous data; inputting the multimodal raw synchronous data into a preset cross-point causal comparison dynamic mode decoupling network, which employs a transfer entropy causal constraint and bimodal comparison learning mechanism to separate the true tension fluctuation component from the multimodal raw synchronous data. This invention can effectively decouple and extract the true tension fluctuation component in real time at high linear speeds, improving cable processing efficiency and quality.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for power cable extrusion production lines, and in particular to a real-time tension fluctuation detection method and device for optimizing cable processing speed. Background Technology

[0002] In a three-layer co-extrusion insulation production line for high-voltage power cables, copper conductors are drawn, stranded, and continuously passed through an extruder, simultaneously extruding three layers of material: inner shielding, insulation, and outer shielding. After cross-linking and curing in a vulcanizing tube, they are pulled and wound up by a traction machine. To increase production capacity, the production line typically needs to operate continuously at high linear speeds. Tension fluctuations are the core factor restricting the increase in linear speed and insulation quality. Existing production lines generally install single-point contact tension sensors between the extruder outlet and the traction machine. The tension value is obtained by deforming the sensor wheel by pressing the cable, which is then converted into an electrical signal.

[0003] However, the aforementioned contact-based single-point measurement methods mostly suffer from the following drawbacks at high linear speeds: difficulty in real-time decoupling and extraction of the true tension fluctuation component. Specifically, the sensor wheel continuously experiences sliding friction and mechanical vibration with the high-speed moving cable surface. The resulting vibration signal is superimposed on the true tension fluctuation signal in the sensor output. The contact sensor can only output a mixed signal containing mechanical artifacts and lacks the ability to identify and remove false disturbances from multimodal and multi-positional information. For example, the tension feedback signal received by the control system is lagging and contains false fluctuations, making it difficult to accurately compensate for the speed and torque of the traction machine and take-up reel within a millisecond time window. The tension fluctuation amplitude increases sharply with increasing linear speed. Moreover, the misadjustment caused by false fluctuations causes the insulation layer to be subjected to non-uniform stretching during extrusion, resulting in insulation layer eccentricity or local thinning, increasing the risk of withstand voltage breakdown. Summary of the Invention

[0004] This invention provides a real-time tension fluctuation detection method and device for optimizing cable processing speed, which can effectively decouple and extract the real tension fluctuation component at high line speed, thereby improving cable processing efficiency and quality.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: In a first aspect, a real-time tension fluctuation detection method for optimizing cable processing speed is provided, the method comprising: Step 1: Install multiple non-contact detection points along the production line, located at the extrusion port of the extruder, the middle section of the vulcanizing pipe, and before the traction machine; synchronously collect laser Doppler vibration signals and eddy current deformation signals at each detection point to obtain multimodal raw synchronous data; Step 2: Input the multimodal raw synchronization data into a preset cross-point causal comparison dynamic mode decoupling network. The preset cross-point causal comparison dynamic mode decoupling network adopts the working mechanism of transfer entropy causal constraint and bimodal comparison learning to separate the real tension fluctuation component from the multimodal raw synchronization data. By sparsely reconstructing the real tension fluctuation component, real-time real tension fluctuation data is formed. Step 3: Based on real-time tension fluctuation data, calculate the stress distribution along the entire line from the extrusion section to the traction section; extract the implicit viscoelastic stress fluctuation that causes the insulation layer to be eccentric from the stress distribution to obtain the stress distribution and implicit viscoelastic stress fluctuation information along the entire line. Step 4: Based on the stress distribution and implicit viscoelastic stress fluctuation information along the entire line, control commands are generated for synchronously adjusting the traction machine speed and the take-up reel torque, thus obtaining the synchronous control command; Step 5: Drive the traction machine and take-up reel according to the synchronous control command to keep the tension fluctuation of the cable within the preset accuracy range during high-speed operation.

[0006] Furthermore, multiple non-contact detection points are deployed along the production line, located at the extrusion nozzle of the extruder, the middle section of the vulcanizing tube, and before the traction machine; laser Doppler vibration signals and eddy current deformation signals are simultaneously collected at each detection point to obtain multimodal raw synchronous data, including: Non-contact sensing nodes are established at the extrusion port, the middle section of the vulcanizing pipe, and before the traction machine. Each non-contact sensing node synchronously senses the vibration response and deformation response of the cable surface at the corresponding measuring point, forming a multi-physics information acquisition system. Based on the multi-physics information acquisition system, each non-contact sensing node is driven by a preset global synchronous triggering benchmark to synchronously acquire vibration and deformation signals of each measuring point, thereby obtaining multi-measuring point synchronous vibration signal and multi-measuring point synchronous deformation signal. The synchronous vibration signals and synchronous deformation signals from multiple measurement points are spatiotemporally aligned and fused according to the measurement point location identifiers and timestamps to form multimodal raw synchronous data.

[0007] Furthermore, the original multimodal synchronous data is input into a pre-defined cross-point causal comparison dynamic mode decoupling network. This network employs a working mechanism of transfer entropy causal constraints and bimodal contrastive learning to separate the true tension fluctuation component from the original multimodal synchronous data, including: The multimodal raw synchronous data is received by a preset cross-point causal comparison dynamic mode decoupling network, and the transfer entropy value between adjacent measurement point pairs is calculated along the production line direction. The transfer entropy value is used to construct a cross-measurement point causal correlation map that characterizes the temporal relationship of tension fluctuation propagation between each measurement point. Using the cross-measurement point causal correlation spectrum as a constraint, the vibration signal and deformation signal of each measurement point are subjected to dual-modal comparative learning. In the hidden space inside the preset cross-point causal comparison dynamic mode decoupling network, the features that have cross-modal consistent response to tension fluctuations are separated to obtain the dual-modal intrinsic feature primitive set. The preset cross-point causal comparison dynamic mode decoupling network extracts the time-series components that satisfy cross-measurement point causal consistency from the dual-modal intrinsic feature primitives, which are then used as the true tension fluctuation components.

[0008] Furthermore, by sparsely reconstructing the actual tension fluctuation components, real-time actual tension fluctuation data is formed, including: By using a pre-defined cross-point causal comparison dynamic pattern decoupling network, the real tension fluctuation component is projected into a sparse representation space spanned by a pre-defined dynamic pattern atom library; and the sparse activation coefficient set corresponding to each dynamic pattern atom for the real tension fluctuation component is solved in the sparse representation space. Based on the sparse activation coefficient set and the pre-set dynamic pattern atom library, the tension fluctuation waveform after denoising and artifact suppression is reconstructed in the time domain to obtain real-time tension fluctuation data.

[0009] Furthermore, based on real-time tension fluctuation data, the stress distribution along the entire line from the extrusion section to the traction section is calculated; the implicit viscoelastic stress fluctuations causing the insulation layer eccentricity are extracted from the stress distribution to obtain the overall stress distribution and implicit viscoelastic stress fluctuation information, including: Based on the real-time tension fluctuation data, a coaxial cylindrical differential manifold model of the cable is constructed along the production line direction. On the coaxial cylindrical differential manifold model, an isomorphic mapping relationship between the tension fluctuation at each measuring point and the surface curvature of the cable is established to obtain the curvature-tension isomorphic mapping function. The real-time tension fluctuation data is mapped to the spatial position between each measuring point based on the instantaneous linear velocity to support stress calculation across the entire line. Based on the curvature-tension isomorphic mapping function, the corresponding time series values ​​of the real-time tension fluctuation data at each measuring point are transformed into a curvature distribution sequence along the entire line from the extrusion section to the traction section. Stress inversion is performed on the curvature distribution sequence to obtain the stress distribution along the entire line. The stress distribution along the entire line is projected onto the cross section of the coaxial cylindrical differential manifold model, and the circumferential stress asymmetry component caused by the non-equilibrium of the material curing path is extracted as the implicit viscoelastic stress fluctuation that causes the insulation layer to be eccentric, thus obtaining the information on the stress distribution along the entire line and the implicit viscoelastic stress fluctuation.

[0010] Furthermore, based on the stress distribution and implicit viscoelastic stress fluctuation information along the entire line, control commands are generated for synchronously adjusting the traction machine speed and the take-up reel torque, resulting in synchronization control commands, including: The tension deviation characteristics within the traction section are extracted from the stress distribution along the entire line. The characteristics representing the eccentricity tendency of the insulation layer are extracted from the implicit viscoelastic stress fluctuation information. The tension deviation characteristics are combined with the characteristics representing the eccentricity tendency of the insulation layer to obtain the tension-eccentricity joint deviation characteristics. Based on the combined tension-eccentricity deviation characteristics, the traction machine speed compensation amount matching the tension deviation characteristics and the take-up reel torque compensation amount matching the eccentricity tendency characteristics are calculated to obtain the synchronization control command.

[0011] Furthermore, the traction machine and take-up reel are driven according to synchronous control commands to maintain the tension fluctuation of the cable throughout the high-speed operation within a preset accuracy range, including: By decoupling the traction machine speed compensation amount and the take-up reel torque compensation amount in the synchronous control command, the traction machine speed adjustment command and the take-up reel torque adjustment command are obtained. The current speed setting value of the traction machine is updated with the traction machine speed adjustment command, and the current torque setting value of the take-up reel is updated with the take-up reel torque adjustment command, so that the traction machine and the take-up reel perform adjustment actions synchronously. After the traction machine and take-up reel perform adjustment actions, the real-time tension fluctuation data of the next sampling cycle is used to verify whether the tension fluctuation of the entire line has returned to the preset accuracy range, thus completing the closed-loop adjustment.

[0012] Secondly, a real-time tension fluctuation detection device for optimizing cable processing speed includes: The multimodal synchronous acquisition module is used to deploy multiple non-contact detection points along the production line, located at the extrusion port of the extruder, the middle section of the vulcanizing pipe, and before the traction machine; it synchronously acquires laser Doppler vibration signals and eddy current deformation signals at each measurement point to obtain multimodal raw synchronous data; The tension fluctuation decoupling and reconstruction module is used to input multimodal raw synchronous data into a preset cross-point causal comparison dynamic mode decoupling network. The preset cross-point causal comparison dynamic mode decoupling network adopts the working mechanism of transfer entropy causal constraint and bimodal comparison learning to separate the real tension fluctuation component from the multimodal raw synchronous data. By sparsely reconstructing the real tension fluctuation component, real-time real tension fluctuation data is formed. The stress calculation and latent fluctuation extraction module is used to calculate the stress distribution along the entire line from the extrusion section to the traction section based on real-time tension fluctuation data; and to extract the latent viscoelastic stress fluctuations that cause the insulation layer to be eccentric from the stress distribution, thereby obtaining the stress distribution and latent viscoelastic stress fluctuation information along the entire line. The synchronization control command generation module is used to generate control commands for synchronously adjusting the traction machine speed and the take-up reel torque based on the stress distribution and implicit viscoelastic stress fluctuation information of the entire line, thus obtaining the synchronization control command. The closed-loop control module is used to drive the traction machine and take-up reel according to the synchronous control command, so that the tension fluctuation of the cable is maintained within the preset accuracy range during high-speed operation.

[0013] Thirdly, a computing device, comprising: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0014] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0015] The above-described solution of the present invention has at least the following beneficial effects: By employing non-contact multimodal multi-point synchronous acquisition combined with a cross-point causal comparison dynamic mode decoupling network based on transfer entropy causal constraints and bimodal contrastive learning, the true tension fluctuation data is separated and sparsely reconstructed from the original multimodal data in real time. Then, combined with the calculation of the stress distribution along the entire line and the extraction of implicit viscoelastic stress fluctuations, synchronous control commands for the traction machine and take-up reel are generated and closed-loop regulation is executed. Therefore, it overcomes the engineering problems of existing contact single-point measurement, such as the inability to decouple true and false tension fluctuations in real time, feedback signal lag, and the presence of mechanical artifacts. As a result, it achieves the technical effect of maintaining the tension fluctuation of the entire line within the preset accuracy range under high-speed operation conditions, effectively suppressing insulation layer eccentricity and local thinning, and effectively improving production line speed and cable insulation quality. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a real-time tension fluctuation detection method for optimizing cable processing speed, provided by an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of a real-time tension fluctuation detection device for optimizing cable processing speed, provided by an embodiment of the present invention. Detailed Implementation

[0018] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0019] like Figure 1As shown, an embodiment of the present invention proposes a real-time tension fluctuation detection method for optimizing cable processing speed. The method includes the following steps: Step 1: Install multiple non-contact detection points along the production line, located at the extrusion port of the extruder, the middle section of the vulcanizing pipe, and before the traction machine; synchronously collect laser Doppler vibration signals and eddy current deformation signals at each detection point to obtain multimodal raw synchronous data; Step 2: Input the multimodal raw synchronization data into a preset cross-point causal comparison dynamic mode decoupling network. The preset cross-point causal comparison dynamic mode decoupling network adopts the working mechanism of transfer entropy causal constraint and bimodal comparison learning to separate the real tension fluctuation component from the multimodal raw synchronization data. By sparsely reconstructing the real tension fluctuation component, real-time real tension fluctuation data is formed. Step 3: Based on real-time tension fluctuation data, calculate the stress distribution along the entire line from the extrusion section to the traction section; extract the implicit viscoelastic stress fluctuation that causes the insulation layer to be eccentric from the stress distribution to obtain the stress distribution and implicit viscoelastic stress fluctuation information along the entire line. Step 4: Based on the stress distribution and implicit viscoelastic stress fluctuation information along the entire line, control commands are generated for synchronously adjusting the traction machine speed and the take-up reel torque, thus obtaining the synchronous control command; Step 5: Drive the traction machine and take-up reel according to the synchronous control command to keep the tension fluctuation of the cable within the preset accuracy range during high-speed operation.

[0020] In this embodiment of the invention, multiple non-contact detection points are deployed at the extrusion port, the middle section of the vulcanizing pipe, and before the traction machine to simultaneously acquire multimodal vibration and deformation information, avoiding artifact interference introduced by contact sensors due to sliding friction and mechanical vibration. A preset cross-point causal comparison dynamic mode decoupling network is used to separate the real tension fluctuation component from the multi-point data through transfer entropy constraints and dual-modal comparison learning, and the denoised tension fluctuation waveform is recovered through sparse reconstruction. On this basis, combined with the calculation of the stress distribution of the entire line and the extraction of implicit viscoelastic stress fluctuation, the non-uniform stress causing the insulation layer eccentricity is accurately identified, and synchronous control commands for the traction machine and take-up reel are generated and closed-loop adjustment is executed. The whole method eliminates the hidden dangers of feedback signal lag and misadjustment in traditional schemes, so that the tension fluctuation of the entire line under high line speed can be stabilized within the preset accuracy range, effectively improving the uniformity of insulation thickness and the operating speed of the production line.

[0021] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1: Establish non-contact sensing nodes at the extrusion port, the middle section of the vulcanizing tube, and before the traction machine. Each non-contact sensing node synchronously senses the vibration and deformation responses of the cable surface at its corresponding measuring point, forming a multi-physics information acquisition system. Specifically, this includes: at the extrusion port of the extruder, deploying non-contact sensing nodes in the transition section between the extruder die outlet and the vulcanizing tube inlet. This location is adjacent to the initial forming area of ​​the insulation layer after the three-layer co-extrusion, enabling direct sensing of the instantaneous vibration and deformation responses of the cable surface before it enters the cross-linking and curing state after extrusion; at the middle section of the vulcanizing tube, deploying non-contact sensing nodes in the middle section of the vulcanizing tube... At the midpoint of the axial length of the vulcanizing tube, corresponding to the intermediate stage of the cross-linking reaction of the insulation layer, stress fluctuations caused by the viscoelastic transformation of the material when the cable is subjected to high temperature and high pressure inside the vulcanizing tube can be sensed. Before the traction machine, non-contact sensing nodes are deployed on a smooth straight section between the vulcanizing tube outlet and the contact point of the traction machine track. At this position, the cable has completed cross-linking and curing and has not yet been directly subjected to the mechanical clamping force of the traction machine, which can sense the comprehensive tension response of the entire line before traction. The above three measuring points form a spatially distributed sampling layout along the production line direction, covering the three process sections: the extrusion molding section, the vulcanization cross-linking section, and the traction transition section.

[0022] Each non-contact sensing node includes a laser Doppler sensing unit and an eddy current sensing unit. The laser Doppler sensing unit is a non-contact optical sensor based on the laser Doppler vibration measurement principle, which can obtain the instantaneous velocity of the measured surface by detecting the frequency change of the reflected laser. The eddy current sensing unit is a non-contact displacement sensor based on the eddy current effect, which can obtain the instantaneous distance between the measured conductor surface and the sensing unit by detecting the change of coil impedance. During installation, the two sensing units are simultaneously aligned with the same circumferential position on the cable surface to ensure that they synchronously sense the vibration response and deformation response of the same physical point.

[0023] Specifically, a laser Doppler sensing unit emits a frequency-stable, narrow-linewidth laser beam onto the cable surface. Upon impact with the rapidly moving cable surface, the laser beam scatters, generating a Doppler frequency shift. This reflected light is received by a laser interference module within the sensing unit, interferes with a reference beam, and the Doppler frequency shift is extracted. This shift is then converted into a voltage signal proportional to the instantaneous vibration velocity of the cable surface, thus obtaining the instantaneous vibration velocity response of the cable surface. Simultaneously, a high-frequency alternating current is passed through the excitation coil within the eddy current sensing unit, generating a high-frequency alternating magnetic field that acts on the metal of the cable. The shielding layer induces eddy currents within it, and the secondary magnetic field generated by these eddy currents reacts to the excitation coil, changing its equivalent impedance. Simultaneously, the change in equivalent impedance is detected in real time, and the calibration curve maps this impedance change to the instantaneous deformation displacement response between the cable surface and the sensing node. Based on this, the non-contact sensing nodes at three measuring points form a multi-physics information acquisition system along the production line. This system simultaneously covers the three process sections: extrusion, vulcanization, and traction, allowing the synchronous capture of the propagation of the same tension fluctuation event along the line.

[0024] Step 1.2: Based on the multi-physics information acquisition system, each non-contact sensing node is driven by a preset global synchronous triggering benchmark to synchronously acquire vibration and deformation signals at each measuring point, resulting in multi-point synchronous vibration and deformation signals. Specifically, this includes: based on the established multi-physics information acquisition system, each non-contact sensing node is driven by a preset global synchronous triggering benchmark for synchronous sampling. Specifically, the preset global synchronous triggering benchmark is generated by a synchronous triggering controller. This synchronous triggering controller has a built-in high-stability crystal oscillator clock source, and the reference frequency of the clock source is preset according to the Nyquist sampling frequency required for the production line's highest line speed and tension fluctuation detection. The synchronous triggering controller is connected to the sensing node at the extrusion port, the sensing node in the middle section of the vulcanizing pipe, and the sensing node before the traction machine via transmission cables of equal length, ensuring that the transmission delay of the trigger signal from the controller to each sensing node is consistent.

[0025] Furthermore, at the beginning of each sampling cycle, the synchronous trigger controller simultaneously sends sampling clock pulses to the laser Doppler sensing unit and eddy current sensing unit of each sensing node. The rising edge of the sampling clock pulse arrives at all six sensing units of the three measuring points at the same time. The analog-to-digital converter inside each sensing unit starts sampling under the trigger of the same clock edge, so that the sampling time of the extrusion outlet vibration signal, extrusion outlet deformation signal, vulcanization section vibration signal, vulcanization section deformation signal, pre-traction vibration signal and pre-traction deformation signal are strictly aligned on the time axis with sub-microsecond precision.

[0026] Within each sampling period, the laser Doppler sensing unit at the extrusion port sensor node outputs the extrusion port vibration signal, and the eddy current sensing unit outputs the extrusion port deformation signal; the laser Doppler sensing unit at the mid-section sensor node of the vulcanizing pipe outputs the vulcanizing section vibration signal, and the eddy current sensing unit outputs the vulcanizing section deformation signal; the laser Doppler sensing unit at the sensor node before the traction machine outputs the pre-traction vibration signal, and the eddy current sensing unit outputs the pre-traction deformation signal; the vibration and deformation signals at each measuring point are accompanied by a sampling sequence number marker injected by the synchronous trigger controller when output, which is used for frame alignment and fusion processing in subsequent steps.

[0027] Based on this, all vibration signals output from the three measuring points are framed according to the sampling sequence number to form a multi-measuring-point synchronous vibration signal; all deformation signals output from the three measuring points are framed according to the sampling sequence number to form a multi-measuring-point synchronous deformation signal; the multi-measuring-point synchronous vibration signal and the multi-measuring-point synchronous deformation signal correspond one-to-one in time sequence, providing a synchronous data foundation with sampling-level precision for subsequent spatiotemporal alignment and fusion.

[0028] Step 1.3 involves spatiotemporally aligning and fusing the multi-point synchronous vibration signals and multi-point synchronous deformation signals according to the measurement point location identifiers and timestamps to form multimodal raw synchronous data. Specifically, this includes: attaching corresponding measurement point location identifiers and sampling timestamps to each frame of synchronous vibration and deformation signals, where the measurement point location identifiers are used to distinguish the physical location of the signal source, and the sampling timestamps are used to record the precise time of signal acquisition; based on this, the extrusion outlet vibration signals and extrusion outlet deformation signals with the same timestamps are modally aligned and combined within the data frame, and the same modal alignment and combination operation is performed on the signals of the vulcanization section and the pre-traction measurement points; further, the data frames of the three measurement points after modal alignment and combination are arranged sequentially according to the timestamp order to form continuous multimodal raw synchronous data along the time axis. Each time step in this multimodal raw synchronous data contains vibration and deformation signals from the three measurement points, providing a spatiotemporally consistent data foundation for cross-measurement point causal analysis and dual-modal comparative learning in subsequent steps.

[0029] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1 involves receiving the multimodal raw synchronous data via a pre-defined cross-point causal comparison dynamic mode decoupling network, calculating the transfer entropy value between adjacent measurement point pairs along the production line direction, and constructing a cross-measurement point causal correlation map representing the temporal relationship of tension fluctuation propagation between measurement points using the transfer entropy value. Specifically, the pre-defined cross-point causal comparison dynamic mode decoupling network is a pre-trained multi-task neural network model used to separate the real tension fluctuation components from the multimodal raw synchronous data of multiple measurement points in real time. This network consists of four functional layers: a multimodal data access and preprocessing layer, a cross-measurement point causal correlation modeling layer, a causal constraint bimodal comparison learning layer, and a causal consistency extraction and output layer. The functional layers are connected in series in the direction of data flow to result output. The output of the previous functional layer serves as the input of the next functional layer. At the same time, the causal correlation map generated by the cross-measurement point causal correlation modeling layer is also fed in parallel as a constraint condition to the causal constraint bimodal comparison learning layer and the causal consistency extraction and output layer to guide the subsequent feature separation and component extraction process.

[0030] Specifically, the multimodal data access and preprocessing layer receives the original multimodal synchronous data from step 1. After performing frame synchronization verification and timestamp alignment on the data, it organizes the vibration signal sequence and deformation signal sequence of each measuring point into a multi-channel matrix of vibration signal and a multi-channel matrix of deformation signal, respectively. The cross-measuring point causal correlation modeling layer receives the multi-channel matrix of vibration signal and calculates the transfer entropy value between adjacent measuring point pairs sequentially along the production line direction using the built-in transfer entropy calculation engine. It then constructs a cross-measuring point causal correlation graph using the transfer entropy value as the weight of the directed edge. The causal constraint bimodal contrastive learning layer receives the multi-channel matrix of vibration signal, the multi-channel matrix of deformation signal, and the cross-measuring point causal correlation graph. It performs causal weighted bimodal contrastive learning operations in the latent space within this layer to separate features that have a cross-modal consistent response to tension fluctuations. The causal consistency extraction and output layer receives the feature set output from the latent space and the cross-measuring point causal correlation graph. Based on the causal temporal relationship in the graph, it extracts feature components that satisfy cross-measuring point causal consistency and outputs the true tension fluctuation components.

[0031] The four functional layers mentioned above work together to enable the pre-defined cross-point causal comparison dynamic mode decoupling network to complete end-to-end decoupling extraction from the original synchronous data to the actual tension fluctuation components within a unified framework. Based on this, the pre-defined cross-point causal comparison dynamic mode decoupling network performs the following steps: The network first extracts the vibration signal time series of the extrusion port measuring point, the vibration signal time series of the vulcanizing pipe mid-section measuring point, and the vibration signal time series of the measuring point before the traction machine from the multimodal original synchronous data through the multimodal data access and preprocessing layer group, denoted as follows: , and ,in This indicates the sampling time; furthermore, the cross-measurement point causal relationship modeling layer calculates the transfer entropy value from the extruder to the middle section of the vulcanizing pipe sequentially along the production line direction. And the entropy value transferred from the middle section of the vulcanizing pipe to the traction machine. The formula for calculating the transfer entropy, taking the section from the extruder to the middle of the vulcanizing pipe as an example, is as follows:

[0032] in, This indicates the extrusion outlet vibration signal at time [time]. The value of , This indicates the vibration signal of the middle section of the vulcanizing pipe at time [time]. The value of , This indicates the vibration signal of the middle section of the vulcanizing pipe at the next moment. The possible values ​​of ; For joint probability distribution, and All are conditional probability distributions; the transfer entropy value from the middle section of the vulcanizing pipe to the traction machine. The same calculation method is used, only the corresponding signal sequence is replaced with and and through variables 、 、 The probability distributions are summed; the joint probability and conditional probability mentioned above are obtained by the cross-measurement point causal relationship modeling layer group based on sample statistics within the sliding time window. The cross-measurement point causal relationship modeling layer group uses the two calculated transfer entropy values ​​as the weights of the directed edges, connecting the three measurement points: the extrusion port, the middle section of the vulcanizing pipe, and the traction machine, to construct a cross-measurement point causal relationship graph that characterizes the temporal relationship of tension fluctuation propagation between each measurement point. This graph contains three directed edges: the directed edge from the extrusion port to the middle section of the vulcanizing pipe, the directed edge from the middle section of the vulcanizing pipe to the traction machine, and the directed edge from the extrusion port to the traction machine. The weight of each edge is the corresponding transfer entropy value. The graph completely describes the causal propagation path of tension fluctuation along the production line direction.

[0033] Step 2.2: Using the cross-measurement point causal correlation map as a constraint, perform bimodal comparative learning on the vibration and deformation signals of each measurement point. This aims to separate features with consistent cross-modal responses to tension fluctuations within the latent space of the preset cross-point causal comparative dynamic mode decoupling network, obtaining a bimodal intrinsic feature primitive set. Specifically, this includes: using the preset cross-point causal comparative dynamic mode decoupling network as a constraint on the cross-measurement point causal correlation map, performing bimodal comparative learning on the vibration and deformation signals of each measurement point. This operation is performed by the causal constraint bimodal comparative learning layer group within the network. Specifically, the causal constraint bimodal comparative learning layer group contains two parallel encoding branches: a vibration encoding branch and a deformation encoding branch. The vibration encoding branch is composed of multiple stacked temporal convolutions, used to extract temporal features layer by layer from the vibration signal. The deformation encoding branch is also composed of multiple stacked temporal convolutions, used to extract temporal features layer by layer from the deformation signal. The structures of the two encoding branches are symmetrical, but their weight parameters are independent.

[0034] During bimodal contrastive learning, the network sends the vibration signal sequence collected within the same sliding time window to the vibration coding branch and the deformation signal sequence collected within the same sliding time window to the deformation coding branch for each of the three measurement points: the extrusion port, the middle section of the vulcanizing pipe, and the point before the traction machine. The vibration coding branch performs multi-layer convolution and pooling processing on the vibration signal sequence to generate the vibration feature vector of the measurement point in the network's latent space. At the same time, the deformation coding branch performs multi-layer convolution and pooling processing on the deformation signal sequence to generate the deformation feature vector of the measurement point in the network's latent space. The aforementioned latent space is located at the output end of the causal constraint bimodal contrastive learning layer group and is a high-dimensional feature space automatically constructed by the network during the offline training phase.

[0035] Based on this, the constraint mechanism of bimodal contrastive learning is introduced into the cross-measurement point causal correlation graph obtained in step 2.1. Each directed edge in the cross-measurement point causal correlation graph corresponds to a transfer entropy value, which reflects the causal strength of tension fluctuation propagation between adjacent measurement points. This network incorporates the transfer entropy value as a weighting coefficient into the loss calculation of contrastive learning. For measurement point pairs with high transfer entropy values, the network increases the positive sample matching strength in contrastive learning, strengthens the mutual information consistency between the vibration feature vector and the deformation feature vector at that measurement point, and narrows the distance between the two modal features of vibration and deformation within the same measurement point in the latent space. For measurement points with low transfer entropy values, the network correspondingly widens the distance between different modal features in the latent space, suppressing false modal correlations caused by non-tension factors.

[0036] Furthermore, cross-modal alignment is performed on measurement points with high propagation entropy values. The cross-modal alignment operation makes the feature vectors output by the vibration coding branch and the deformation coding branch in the latent space statistically consistent, ensuring that the two sensing modes can produce coordinated and consistent feature representations for the same tension fluctuation event. After the above-mentioned contrastive learning and cross-modal alignment under causal constraints, the network separates a set of feature components in the latent space that have a consistent cross-modal response to tension fluctuations. These feature components are organized into a bimodal intrinsic feature primitive set. Each primitive in the bimodal intrinsic feature primitive set is a feature vector in the latent space, representing a tension fluctuation primitive mode that appears synchronously in the vibration signal and the deformation signal.

[0037] Step 2.3 involves the preset cross-point causal comparison dynamic mode decoupling network extracting time-series components that satisfy cross-measurement point causal consistency from the bimodal intrinsic feature primitives as the true tension fluctuation components. Specifically, this step is executed by the causal consistency extraction and output layer group within the preset cross-point causal comparison dynamic mode decoupling network. This layer group receives the output bimodal intrinsic feature primitives and the output cross-measurement point causal correlation graph, and performs causal consistency checks on each intrinsic feature primitive based on the causal propagation path in the graph. In detail... By extracting causal consistency and outputting the layer group along the production line direction, each primitive in the dual-modal intrinsic feature primitive group is mapped one by one to the three nodes of the cross-measurement point causal relationship graph. It is checked whether the feature expression of the primitive at the extrusion port measurement point, the middle section measurement point of the vulcanizing pipe and the measurement point before the traction machine satisfies the causal transmission relationship along the production line direction. The test standard is that the feature expression of the primitive between adjacent measurement points must show the unidirectional causal driving characteristic from the upstream measurement point to the downstream measurement point, and the causal direction is consistent with the propagation direction marked in the cross-measurement point causal relationship graph.

[0038] For primitives that exhibit stable causal propagation characteristics on two adjacent measurement point pairs, they are marked as causally consistent primitives by causal consistency extraction and output layer group and retained. For primitives that appear only at a single measurement point or whose causal direction is inconsistent with the cross-measurement point causal correlation map, they are marked as false features caused by local artifacts or environmental interference and are removed. On this basis, all retained causally consistent primitives are arranged in order according to their original time index, and the feature vectors in the latent space are restored to time-series signals. These signals are combined to obtain the time-series signal as the real tension fluctuation component. This time-series signal is the real tension fluctuation component finally separated from the multimodal original synchronization data and serves as the input for subsequent sparse reconstruction.

[0039] In a preferred embodiment of the present invention, step 2 above may further include: Step 2.4: The real tension fluctuation component is projected onto a sparse representation space spanned by a pre-defined cross-point causal comparison dynamic pattern decoupling network. In this sparse representation space, the sparse activation coefficient set corresponding to each dynamic pattern atom for the real tension fluctuation component is solved. Specifically, after obtaining the real tension fluctuation component, the pre-defined cross-point causal comparison dynamic pattern decoupling network further performs a sparse reconstruction operation on the component to remove any residual weak noise and artifacts. This operation is performed by a pre-defined sparse reconstruction processing unit within the network. Specifically, the sparse reconstruction processing unit pre-defined a dynamic pattern atom library. This dynamic pattern atom library is a set of primitive waveforms learned from a large number of real-world tension fluctuation samples during the network's offline training phase. The dynamic pattern atom library is stored in matrix form, denoted as D, with a size of [missing information]. M OK K Column, in which M This represents the number of sampling points for each atom. K This indicates the total number of atoms in the atomic library.

[0040] Each column of matrix D is an independent dynamic mode atom, denoted as ,in The atom is indexed, with values ​​ranging from 1 to... K Each dynamic mode atom Each is a length of M The time-series vector represents a typical tension fluctuation primitive waveform, such as step-type fluctuation primitive, damped oscillation-type fluctuation primitive, narrowband resonant fluctuation primitive, and broadband random fluctuation primitive. These atoms cover the tension fluctuation modes commonly found in high-voltage cable high-speed extrusion production lines, providing a complete set of basis functions for subsequent sparse representation.

[0041] Based on this, the sparse reconstruction processing unit truncates the actual tension fluctuation component output in step 2.3 into a length of [length missing] using a sliding window. M time vector The step size of the sliding window is consistent with the sampling period in step 1 to ensure that each time-series vector is seamlessly connected on the time axis; for each time-series vector The sparse representation space is projected onto the sparse representation space spanned by the dynamic mode atom library D through the sparse reconstruction processing unit. K The high-dimensional space spanned by linear combinations of dynamic mode atoms; within this sparse representation space, sparse reconstruction processing units process temporal vectors. Sparse decomposition is performed, and the following optimization problem is solved within a fixed number of iterations using a pre-defined sparse coding algorithm to obtain a sparse activation coefficient set. Its form is as follows:

[0042] In the formula, This indicates a search for the parameter that minimizes the expression within the following curly braces. This can be understood as searching in the parameter space to find the value that minimizes the objective function. ; Let be the sparse activation coefficient vector to be solved, with dimension . K Each element in the vector corresponds to the activation intensity of an atom in the dynamic mode atom library; This represents the reconstruction error term, which is calculated from the original signal vector. The signal reconstructed from the atomic library and coefficients The square of the Euclidean norm between the two signals measures how closely the reconstructed signal approximates the original signal; the smaller this term is, the more accurate the reconstruction is. Represents the coefficient vector The L1 norm, or sparse regularization term, is equal to the sum of the absolute values ​​of all elements in the vector; by minimizing this term, we can force... Most of the elements in the signal are 0 or close to 0, thus achieving sparsity and ensuring that only a few atoms are used to reconstruct the signal. The regularization parameter, whose value is determined during offline training using cross-validation, is used to balance the relationship between reconstruction accuracy and sparsity. During the solution process, the sparse reconstruction processing unit uses an iterative threshold shrinkage algorithm to successively approximate the optimal solution. The number of iterations is preset to a fixed value to meet real-time requirements, ultimately resulting in a sparse activation coefficient set. In the study, only a few atoms that match the current tension fluctuation pattern have non-zero activation coefficients, while the activation coefficients of the remaining atoms are 0 or very small values ​​close to 0.

[0043] Step 2.5: Based on the sparse activation coefficient set and the preset dynamic pattern atom library, the tension fluctuation waveform after denoising and artifact suppression is reconstructed in the time domain to obtain real-time tension fluctuation data. Specifically, this includes: after completing the sparse decomposition in step 2.4, the sparse reconstruction processing unit reconstructs the tension fluctuation waveform using the solved sparse activation coefficient set and the preset dynamic pattern atom library; specifically, the sparse reconstruction processing unit reads the sparse activation coefficient set... This coefficient set is a set of length . K A vector, where each element The corresponding dynamic mode atom library Atoms The activation intensity; simultaneously read the preset dynamic mode atom library matrix D, which contains K Each atom is in a dynamic mode. All are of length M The time-series vector.

[0044] Based on this, the sparse activation coefficient set Each element in the process is thresholded one by one. For elements whose absolute activation intensity is greater than the set activation threshold, their corresponding atoms are marked as valid activated atoms. For elements whose absolute activation intensity is less than or equal to the set activation threshold, their activation coefficients are set to 0, and their corresponding atoms do not participate in subsequent reconstruction stacking. After thresholding, only a few valid activated atoms that directly match the current tension fluctuation pattern and their corresponding activation coefficients are retained. Further, the sparse reconstruction processing unit performs a weighted stacking operation item by item according to the atom index order: that is, for each atom marked as validly activated... Multiply each sample value in the atomic waveform by the activation coefficient corresponding to that atom. This yields a weighted atomic waveform. The weighted atomic waveform is scaled proportionally to the activation intensity in amplitude and remains consistent with the original atomic waveform in phase. Based on this, the weighted waveforms of all effective activated atoms are accumulated point-by-point along the same time axis according to the sampling point position. The accumulation process starts from the first sampling point, sequentially adding the weighted values ​​of each effective activated atom at that sampling point, until the [number missing]th sampling point. M Each sampling point is accumulated, and after accumulation, the reconstructed waveform vector is obtained. .

[0045] Reconstructed waveform vector This refers to the tension fluctuation signal after denoising and artifact suppression. In this reconstruction process, the activated atoms in the real tension fluctuation component separated in step 2.3 retain the basic components that are consistent with the physical propagation characteristics of tension. However, the residual environmental noise and mechanical artifacts that could not be completely eliminated in step 2.3 cannot form an effective match with any single atom or combination of atoms in the dynamic mode atom library. Therefore, they are automatically suppressed by the regularization term in the sparse decomposition step in step 2.4, and their corresponding activation coefficients are compressed to zero or below the preset activation threshold. Thus, they do not participate in the reconstruction of the final waveform.

[0046] The sparse reconstruction processing unit reconstructs each waveform vector obtained within the sliding window. The waveforms are sequentially stitched together according to their corresponding timestamps to reconstruct a continuous tension fluctuation waveform in the time domain. During the stitching process, the overlapping parts between adjacent windows are transitioned using a Hanning window weighted smoothing method. The waveform amplitude of the previous window gradually decays to 0 in the overlapping area, while the waveform amplitude of the next window gradually rises from 0 to full amplitude in the overlapping area. The two waveforms are then weighted and superimposed in the overlapping area to achieve a smooth connection, thereby eliminating possible signal discontinuities at the window boundaries. The final stitched continuous waveform is output as real-time tension fluctuation data. This data fully preserves the amplitude and timing characteristics of the real tension fluctuations while effectively removing the influence of non-real tension fluctuations such as contact interference, workshop vibration, and environmental electromagnetic noise.

[0047] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: Based on the real-time tension fluctuation data, construct a coaxial cylindrical differential manifold model of the cable along the production line direction. Establish an isomorphic mapping relationship between the tension fluctuation at each measuring point and the cable surface curvature on the coaxial cylindrical differential manifold model, obtaining the curvature-tension isomorphic mapping function. The real-time tension fluctuation data is mapped from instantaneous linear velocity to the spatial position between measuring points to support full-line stress calculation. Specifically, this step is executed by a stress calculation and implicit fluctuation extraction module. This module first constructs a coaxial cylindrical differential manifold model of the cable along the production line direction based on the real-time tension fluctuation data output in Step 2. Specifically, the stress calculation and implicit fluctuation extraction module uses the common perpendicular line of the extruder outlet center, the vulcanizing pipe axial centerline, and the traction machine track clamping centerline as the reference axis, and the nominal outer diameter of the cable as the reference radius, to generate an ideal reference cylindrical surface along the production line direction. This reference cylindrical surface is then spatially stepped along the axial direction. Discretize into a series of differential surface elements, each corresponding to an axial position. and azimuth ; Axial position From the extrusion port Starting from the beginning, increase along the production line direction to the position before traction. Based on this, the radial displacement of the actual cable surface at each measuring point from the ideal reference cylindrical surface is mapped onto the corresponding differential surface element to obtain the radial displacement fields along the axial and circumferential directions. .

[0048] Furthermore, after establishing the radial displacement field, based on the definition of surface curvature in differential geometry, the instantaneous principal curvature at each differential surface element is calculated, for the axial position... Location, azimuth The differential surface element, its axial curvature The circumferential curvature is given by the second-order partial derivative of the radial displacement with respect to the axial coordinate. The relationship between the second-order partial derivative of the radial displacement with respect to the azimuth angle and the instantaneous radius of curvature is given by the following formula:

[0049]

[0050] In the formula, Indicates radial displacement For axial coordinates The second-order partial derivative of the displacement characterizes the rate of change of the displacement along the axis of curvature and is the core of calculating the axial curvature. Indicates radial displacement opposite angle The second-order partial derivative of the displacement characterizes the rate of change of the convexity and concavity along the circumferential direction and is the core of calculating the circumferential curvature. Indicates the position of the partial derivative along the axis. The value at the specified point ensures that the calculation result of the continuous function is accurately mapped to the specific location of the measurement point. The nominal radius of the cable is consistent with the reference radius. The axial curvature reflects the degree of local bending of the cable in the traction direction, while the circumferential curvature reflects the non-uniformity of the insulation layer thickness along the circumferential direction. Together, they constitute the instantaneous principal curvature of this differential element.

[0051] Based on this, an isomorphic mapping relationship between tension fluctuations and cable surface curvature is established at each measuring point. For the three measuring points—the extrusion port measuring point, the vulcanizing pipe mid-section measuring point, and the pre-traction measuring point—the time-series values ​​of the real-time tension fluctuation data output in step 2 are extracted at each location, denoted as... , and ,in , , The axial position coordinates of the three measuring points are represented respectively; simultaneously, the instantaneous equivalent curvature of the cable surface at each measuring point is calculated on the coaxial cylindrical differential manifold model. The equivalent curvature is determined by the axial curvature. and circumferential curvature The components are synthesized using appropriate weights; the instantaneous tension value is correlated with the instantaneous equivalent curvature value to construct a curvature-tension isomorphic mapping function. This makes it possible for any measurement point location The mapping relationship is as follows:

[0052] In a preferred embodiment, the curvature-tension isomorphic mapping function Using a linear constitutive form, its formula is:

[0053] in, Indicates the location ,time The real-time tension value at the location; The effective elastic modulus of the insulating material varies with temperature and exhibits a gradient distribution along the production line direction. Different calibration values ​​are taken in the three process sections described in step 1.1. The moment of inertia of the cable cross-section about the neutral axis is determined by the nominal radius of the cable and the thickness of the insulation layer. This mapping function ensures a one-to-one isomorphic relationship between tension fluctuations and curvature changes within the elastic range, and maintains the structural mapping characteristics at all three measuring points.

[0054] Step 3.2: Based on the curvature-tension isomorphic mapping function, the corresponding time-series values ​​of the real-time tension fluctuation data at each measuring point are transformed into a curvature distribution sequence along the entire line from the extrusion section to the traction section. Stress inversion is then performed on the curvature distribution sequence to obtain the stress distribution along the entire line. Specifically, this includes: first dividing the entire line along the axial direction into... A series of consecutive spatial cells, each with an axial length of [missing value]. , No. The axial coordinates of each spatial cell are For each spatial cell, based on its axial coordinates With the positions of three known measuring points , , The spatial relationship is used to calculate the equivalent curvature value at the spatial cell using an axial interpolation method.

[0055] The interpolation process uses the curvature values ​​of three measurement points as reference points, where the curvature value of the extrusion outlet measurement point is... Curvature value of measuring point in the middle section of vulcanizing pipe curvature value of the measuring point before traction Interpolation weights are assigned based on the axial distance between the spatial cell and each measuring point, with higher weights for nearby measuring points and lower weights for points farther away. For spatial cells located in the extrusion section, their curvature values ​​are determined by interpolating the curvature values ​​of the extrusion port measuring point and the mid-section of the vulcanizing pipe measuring point. For spatial cells located in the vulcanizing section, their curvature values ​​are determined by the non-linear attenuation of the curvature values ​​of the mid-section of the vulcanizing pipe measuring point along the axial direction within the vulcanizing pipe. For spatial cells located in the traction section, their curvature values ​​are determined by interpolating the curvature values ​​of the mid-section of the vulcanizing pipe measuring point and the pre-traction measuring point measuring point. Through the above segmented interpolation, the curvature values ​​along the entire line from the extrusion section to the traction section are obtained. Curvature distribution sequence of spatial cells ,in from arrive .

[0056] Based on this, stress inversion is performed on the curvature distribution sequence to obtain the stress distribution along the entire line; the stress inversion is based on the viscoelastic constitutive relation of the cable, for the entire line... A spatial cell, its axial stress and circumferential stress The following relationships are given respectively:

[0057] in, The effective elastic modulus of the insulating material varies with temperature and exhibits a gradient distribution along the production line direction. Different calibration values ​​are taken in the three process sections described in step 1.1. This is the nominal radius of the cable. and The first The axial curvature and circumferential curvature of a spatial cell are given by the equivalent curvature of that cell. Obtained by azimuth decomposition; all The axial and circumferential stresses of each spatial cell are arranged in axial position order to form the stress distribution along the entire line.

[0058] Step 3.3: Project the overall stress distribution onto the cross-section of the coaxial cylindrical differential manifold model, extract the circumferential stress asymmetric component caused by the non-equilibrium of the material curing path, as the implicit viscoelastic stress fluctuation causing the insulation layer eccentricity, and obtain the overall stress distribution and implicit viscoelastic stress fluctuation information. Specifically, the projection process is performed independently for each spatial cell, for the axial position... The spatial cell at that location, the axial stress of that cell. and circumferential stress Azimuth angle along the cross section at that location Expanding along the direction, we construct the circumferential stress distribution function of the cross-section. The circumferential stress distribution function describes the variation of stress along the circumferential direction at this axial position; in a preferred embodiment, the circumferential stress distribution function... The specific expression is:

[0059] in, and These are the axial stress and circumferential stress obtained in step 3.2, respectively. This expression, representing the in-plane shear stress caused by the non-equilibrium of the material curing path, fully describes the stress distribution along the circumferential direction within the cross-section at any axial position. The asymmetric component of the circumferential stress caused by the non-equilibrium of the material curing path is extracted from the aforementioned circumferential stress distribution function. The non-equilibrium of the material curing path refers to the temperature gradient along the circumferential direction in the cross-linking curing reaction of the insulating material within the middle section of the vulcanized pipe, leading to inconsistent curing degrees at various angular locations of the insulating layer, thus resulting in an asymmetric spatial distribution of circumferential stress. This asymmetric component is extracted using the following method: first, the circumferential stress distribution function... Along azimuth Fourier series expansion was performed to separate axisymmetric components and several non-axisymmetric harmonic components. The axisymmetric components correspond to the uniform stress caused by homogeneous curing, while the non-axisymmetric harmonic components of each order correspond to different modes of non-uniform curing stress. All non-axisymmetric harmonic components except the axisymmetric components were synthesized to obtain the circumferential stress asymmetric components. This asymmetric component of circumferential stress directly causes the insulation layer to become eccentric at the corresponding axial position.

[0060] The circumferential stress asymmetric components of all spatial cells along the entire line are aggregated according to their axial positions to form the implicit viscoelastic stress fluctuations that cause the insulation layer to deviate. Finally, the stress distribution along the entire line obtained in step 3.2 is output together with the aforementioned implicit viscoelastic stress fluctuation information to obtain the stress distribution and implicit viscoelastic stress fluctuation information along the entire line. This information simultaneously includes the stress state of the insulation layer at each position along the entire line from the extrusion section to the traction section and the asymmetric stress components that cause the deviator, providing a complete mechanical basis for generating synchronous control commands in subsequent steps.

[0061] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Extract tension deviation features within the traction section from the overall stress distribution, and extract features representing the insulation layer's eccentricity tendency from the implicit viscoelastic stress fluctuation information. Combine the tension deviation features with the features representing the insulation layer's eccentricity tendency to obtain the tension-eccentricity joint deviation features. Specifically, this includes: sending the overall stress distribution and implicit viscoelastic stress fluctuation information into the processing flow for generating synchronous control commands. In this processing flow, first, a spatial interval corresponding to the traction section is extracted from the axial range covered by the overall stress distribution. The starting boundary of this interval is the axial position corresponding to the vulcanized pipe outlet, and the ending boundary is the axial position corresponding to the contact point between the traction machine track and the cable. After extraction, the axial stress value of each spatial cell is extracted sequentially from the axial stress value sequence arranged along the spatial axis within this interval. The arithmetic mean of all axial stress values ​​within the interval is calculated to obtain the average axial stress of the traction section. Simultaneously, the difference between the average axial stress and the pre-set target tension value is compared. The pre-set target tension value is a comprehensive expected tension value determined based on process parameters such as the nominal cross-sectional area of ​​the conductor, the design thickness of the insulation layer, and the line speed of the production line for the current production batch of cables. This target tension value is pre-calculated and stored by the process parameter configuration system before the start of each batch of cable production, and is directly read and called during the execution of step 4.1.

[0062] The target tension value represents the optimal tension level that simultaneously satisfies the requirements of uniform extrusion of the insulation layer and conductor elongation under the current linear speed and specifications. The difference between the average axial stress and the target tension value is the tension deviation. The absolute value of this deviation directly reflects the degree to which the tension deviates from the optimal setting value within the traction section, and the positive or negative sign of the deviation indicates whether the tension is over-tensioned or under-tensioned. Based on this, the difference in axial stress between adjacent spatial cells is calculated sequentially from front to back along the spatial axis of the traction section. The absolute values ​​of all adjacent differences are arranged along the spatial axis to form a distribution sequence of axial stress fluctuation amplitude along the traction section. The gradient of axial stress fluctuation amplitude with spatial position is extracted from this distribution sequence. This gradient serves as the dynamic change characteristic of the tension deviation, reflecting the spatial evolution trend of tension fluctuation within the traction section. The aforementioned tension deviation and dynamic change characteristic together constitute the tension deviation characteristics within the traction section.

[0063] Furthermore, within the same processing flow, the insulation layer eccentricity tendency characteristics are extracted from the implicit viscoelastic stress fluctuation information. The implicit viscoelastic stress fluctuation information includes the circumferential stress asymmetric components in the cross-section of each spatial cell along the entire line. From these circumferential stress asymmetric components, the amplitude of the circumferential stress asymmetric components of all spatial cells covered by the traction section is extracted cell by cell. The extracted amplitudes are arranged sequentially along the axial direction of the traction section from the vulcanizing pipe outlet to the traction machine contact point, and accumulated cell by cell to obtain the cumulative distribution sequence of amplitudes along the axial direction. In the cumulative distribution sequence, the maximum value of the cumulative amplitude is located, and this maximum value is used as the maximum eccentricity stress index. This index reflects the most severe degree of insulation layer eccentricity within the traction section.

[0064] Simultaneously, the spatial rate of change of the cumulative amplitude distribution sequence is calculated along the axial direction. The calculation process is as follows: the cumulative amplitude values ​​of adjacent spatial cells in the cumulative distribution sequence are compared one by one, and the cumulative amplitude value of the previous spatial cell is subtracted from the cumulative amplitude value of the next spatial cell to obtain the cumulative increment between adjacent cells; each cumulative increment is divided by the axial distance between adjacent spatial cells to obtain the local spatial rate of change between each pair of adjacent cells; all local spatial rates of change are arranged sequentially along the axial direction and the arithmetic mean is taken. The average value is the eccentricity development trend index, which reflects the degree of insulation layer eccentricity with spatial position within the traction section. The overall evolution direction of intensification or weakening is as follows: when the eccentricity development trend index is positive, it indicates that the degree of eccentricity continues to intensify along the traction direction; when the eccentricity development trend index is negative, it indicates that the degree of eccentricity gradually weakens along the traction direction. The above-mentioned maximum eccentricity stress index and eccentricity development trend index together constitute the insulation layer eccentricity tendency characteristics. After the tension deviation characteristics and insulation layer eccentricity tendency characteristics are extracted, the two sets of characteristics are aligned and combined under the same time mark, so that the tension deviation amount, tension deviation change gradient, maximum eccentricity stress index and eccentricity development trend index in each control cycle are organized into the same data structure, that is, tension-eccentricity joint deviation characteristics.

[0065] Step 4.2: Based on the tension-eccentricity combined deviation characteristics, calculate the traction machine speed compensation amount that matches the tension deviation characteristics and the take-up reel torque compensation amount that matches the eccentricity tendency characteristics to obtain synchronous control commands. Specifically, after obtaining the tension-eccentricity combined deviation characteristics, these combined deviation characteristics are synchronously sent to two parallel compensation calculation processes: the tension control loop and the eccentricity suppression loop. The two processes are started synchronously and completed simultaneously within the same control cycle.

[0066] In the tension control loop, the tension deviation and the gradient of the tension deviation change are used as inputs. The pre-set speed proportional gain coefficient is multiplied by the tension deviation to obtain the basic component of the speed compensation. The pre-set speed proportional gain coefficient is a proportional factor describing the intensity of the traction machine speed response to the tension deviation. Its physical meaning is the traction machine speed adjustment range corresponding to a unit tension deviation. The value of this coefficient is obtained through step response test calibration during the production line commissioning phase. During calibration, a speed step disturbance of known amplitude is applied to the traction machine under steady-state operating conditions, and the steady-state change of the tension response curve is recorded. The value of this coefficient is determined by the ratio of the speed step amplitude to the steady-state change of the tension. After calibration, this coefficient is stored in the control parameter table and is directly called each time step 4.2 is executed. The basic component is proportional to the magnitude of the tension deviation and undertakes the main tension correction function.

[0067] Simultaneously, the pre-set differential gain coefficient of rotational speed is multiplied by the tension deviation change gradient to obtain the damping component of the rotational speed compensation. The pre-set differential gain coefficient of rotational speed is a differential factor describing the intensity of the traction machine's rotational speed response to the tension change trend. Its physical meaning is the damping adjustment amplitude of the traction machine's rotational speed corresponding to a unit tension deviation change gradient. The value of this coefficient is determined through online gradient response identification. During identification, the peak value of the deviation change gradient and the corresponding optimal damping amount are recorded at the rising and falling edges of the tension fluctuation, respectively. The value of this coefficient is determined by the ratio of the optimal damping amount to the gradient peak value. The component is used to suppress dynamic overshoot and oscillation during tension changes, enabling the tension to converge quickly and smoothly to the target value during adjustment. The traction machine speed compensation is obtained by adding the base component and the damping component. The positive or negative attribute of the traction machine speed compensation is directly determined by the sign of the tension deviation. When the tension deviation is positive, i.e., the real-time tension is lower than the target tension, the speed compensation is positive, and the traction machine speed increases in the next control cycle to tighten the cable. When the tension deviation is negative, i.e., the real-time tension is higher than the target tension, the speed compensation is negative, and the traction machine speed decreases in the next control cycle to relax the cable.

[0068] In the eccentricity suppression circuit, the maximum eccentricity stress index and the eccentricity development trend index are used as input quantities. A pre-set torque proportionality coefficient is multiplied by the maximum eccentricity stress index to obtain the basic amplitude adjustment component of the torque compensation. This pre-set torque proportionality coefficient is a proportional factor describing the intensity of the take-up reel torque response to eccentricity stress. Its physical meaning is the torque adjustment amplitude of the take-up reel corresponding to a unit maximum eccentricity stress. The value of this coefficient is determined through a static calibration test of eccentricity stress and torque. During calibration, a circumferential eccentricity stress of known amplitude is applied to the cable through an external loading device while the take-up reel is stationary. The steady-state torque value of the take-up reel required to adjust the insulation eccentricity to the allowable range under different eccentricity stress levels is recorded. The value of this coefficient is obtained by fitting the ratio of the torque value to the eccentricity stress. The basic amplitude adjustment component is proportional to the maximum eccentricity stress index and plays the main role in eccentricity suppression adjustment.

[0069] Simultaneously, the pre-set torque advance compensation coefficient is multiplied by the eccentricity development trend index to obtain the advance adjustment component of the torque compensation amount. The pre-set torque advance compensation coefficient is an advance adjustment factor describing the intensity of the take-up reel torque's advance response to the eccentricity development trend. Its physical meaning is the advance adjustment amplitude of the take-up reel torque corresponding to a unit eccentricity development trend. The value of this coefficient is determined through joint optimization of the eccentricity development trend and torque timing. During optimization, under the condition of gradual accumulation of eccentric stress along the axial direction, torque adjustment is applied along the traction direction with different time advances, and the suppression effect of eccentric stress at each time advance is recorded. The optimal suppression... The value of this coefficient is derived from the ratio of the time advance corresponding to the control effect to the current eccentricity development trend. The advance adjustment component applies adjustment in advance according to the eccentricity development trend, and begins to intervene in compensation before the eccentricity reaches the most serious stage, preventing the eccentricity from continuing to deteriorate in the next time step. The torque compensation amount of the take-up reel is obtained by adding the basic amplitude adjustment component and the advance adjustment component. The direction of the torque compensation amount of the take-up reel is opposite to the deflection direction of the circumferential stress asymmetry component. By applying compensation torque on the take-up reel to change the interlayer clamping force distribution when the cable is wound, the further development of insulation layer eccentricity is suppressed from the structural perspective.

[0070] After the two loops are solved, the traction machine speed compensation and the take-up reel torque compensation are assigned the same time stamp and encapsulated as a synchronous control command. This ensures that the traction machine and the take-up reel maintain strict timing coordination when performing adjustment actions, and avoids introducing new tension disturbances in the traction section due to asynchronous adjustment.

[0071] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1: Decouple the traction machine speed compensation amount and the take-up reel torque compensation amount in the synchronization control command to obtain the traction machine speed adjustment command and the take-up reel torque adjustment command. Specifically, this includes: sending the synchronization control command into the command decoupling processing flow. This synchronization control command internally encapsulates the two compensation parameters, traction machine speed compensation amount and take-up reel torque compensation amount, with the same time stamp. Specifically, first, read the time stamp in the synchronization control command and verify its validity. After confirming that the synchronization control command belongs to the current control cycle, continue to execute subsequent processing. After the verification is successful, extract the traction machine speed from the data structure of the synchronization control command. The compensation amount field and the take-up reel torque compensation amount field are separated into two independent compensation signals. The traction machine speed compensation amount is projected as a traction machine speed adjustment command, which represents the increase or decrease in the traction machine speed relative to the current speed in the next control cycle in the form of a target speed increment. The take-up reel torque compensation amount is projected as a take-up reel torque adjustment command, which represents the increase or decrease in the take-up reel torque value relative to the current torque in the next control cycle in the form of a target torque increment. Both adjustment commands inherit the time stamp of the original synchronous control command to ensure that the timing synchronization between the two is not lost in subsequent execution stages.

[0072] Step 5.2: Update the current speed setting of the traction machine with a traction machine speed adjustment command, and simultaneously update the current torque setting of the take-up reel with a take-up reel torque adjustment command, so that the traction machine and the take-up reel perform adjustment actions synchronously. Specifically, this includes: after command decoupling, the traction machine speed adjustment command and the take-up reel torque adjustment command are sent to the traction machine drive channel and the take-up reel drive channel respectively for synchronous execution; in the traction machine drive channel, the traction machine speed adjustment command is written into the speed setting value register of the traction machine driver, overwriting the current speed setting value stored in the register. The traction machine driver reads the updated speed setting value, compares it with the current actual speed feedback value of the traction machine, and adjusts the power supply frequency or current of the traction machine motor through the internal speed closed-loop controller. This allows the actual linear speed of the traction machine track to smoothly transition to the new set speed. In the take-up reel drive channel, the take-up reel torque adjustment command is written into the torque setting value register of the take-up reel driver, overwriting the current torque setting value stored in the register. The take-up reel driver reads the updated torque setting value and compares it with the current actual torque feedback value of the take-up reel. The output torque of the take-up reel motor is adjusted through the internal torque closed-loop controller, so that the actual winding torque of the take-up reel smoothly transitions to the new set torque. The two drive channels start the adjustment action simultaneously under the same time mark trigger. At the same time, the adjustment rate of the traction machine speed and the adjustment rate of the take-up reel torque are smoothly transitioned to ensure that the cable is not subjected to transient impact due to sudden changes in speed or torque during the adjustment process.

[0073] Step 5.3: After the traction machine and take-up reel perform adjustment actions, verify whether the tension fluctuation of the entire line has returned to the preset accuracy range from the real-time tension fluctuation data of the next sampling cycle, and complete the closed-loop adjustment. Specifically, after the traction machine and take-up reel perform a complete adjustment action, the production line enters the next sampling cycle. In the new sampling cycle, according to the method in step 1, the multimodal original synchronous data is collected again at three measuring points: the extrusion port, the middle section of the vulcanizing pipe, and before the traction machine. At the same time, in step 2, the real-time tension fluctuation data corresponding to the next sampling cycle is re-separated and reconstructed based on the updated multimodal original synchronous data.

[0074] The system receives real-time tension fluctuation data from the next sampling cycle, extracts the instantaneous amplitude of the overall tension fluctuation corresponding to the traction section, and compares this instantaneous amplitude with the upper and lower limits of the preset accuracy range point by point. The preset accuracy range is the allowable range of tension fluctuations converted from the insulation thickness tolerance grade and conductor elongation index of the current production batch of cables. Its upper and lower limits are preset and stored during the production line debugging stage. When the instantaneous amplitude corresponding to all sampling points in the traction section falls between the upper and lower limits of the preset accuracy range, it is determined that the overall tension fluctuation has returned to the preset accuracy range, and the closed-loop adjustment task in the current control cycle is completed. When the instantaneous amplitude corresponding to any sampling point in the traction section exceeds the preset accuracy range, it is determined that the overall tension fluctuation has not yet returned to the preset accuracy range. At this time, the next control cycle continues to execute the closed-loop adjustment process of steps 3 to 5 until the overall tension fluctuation meets the accuracy requirements.

[0075] like Figure 2 As shown, embodiments of the present invention also provide a real-time tension fluctuation detection device for optimizing cable processing speed, comprising: The multimodal synchronous acquisition module is used to deploy multiple non-contact detection points along the production line, located at the extrusion port of the extruder, the middle section of the vulcanizing pipe, and before the traction machine; it synchronously acquires laser Doppler vibration signals and eddy current deformation signals at each measurement point to obtain multimodal raw synchronous data; The tension fluctuation decoupling and reconstruction module is used to input multimodal raw synchronous data into a preset cross-point causal comparison dynamic mode decoupling network. The preset cross-point causal comparison dynamic mode decoupling network adopts the working mechanism of transfer entropy causal constraint and bimodal comparison learning to separate the real tension fluctuation component from the multimodal raw synchronous data. By sparsely reconstructing the real tension fluctuation component, real-time real tension fluctuation data is formed. The stress calculation and latent fluctuation extraction module is used to calculate the stress distribution along the entire line from the extrusion section to the traction section based on real-time tension fluctuation data; and to extract the latent viscoelastic stress fluctuations that cause the insulation layer to be eccentric from the stress distribution, thereby obtaining the stress distribution and latent viscoelastic stress fluctuation information along the entire line. The synchronization control command generation module is used to generate control commands for synchronously adjusting the traction machine speed and the take-up reel torque based on the stress distribution and implicit viscoelastic stress fluctuation information of the entire line, thus obtaining the synchronization control command. The closed-loop control module is used to drive the traction machine and take-up reel according to the synchronous control command, so that the tension fluctuation of the cable is maintained within the preset accuracy range during high-speed operation.

[0076] It should be noted that this device is a device corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0077] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0078] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0079] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A real-time tension fluctuation detection method for optimizing cable processing speed, characterized in that, The method includes: Step 1: Install multiple non-contact detection points along the production line, located at the extrusion port of the extruder, the middle section of the vulcanizing pipe, and before the traction machine; synchronously collect laser Doppler vibration signals and eddy current deformation signals at each detection point to obtain multimodal raw synchronous data; Step 2: Input the multimodal raw synchronization data into a preset cross-point causal comparison dynamic mode decoupling network. The preset cross-point causal comparison dynamic mode decoupling network adopts the working mechanism of transfer entropy causal constraint and bimodal comparison learning to separate the real tension fluctuation component from the multimodal raw synchronization data. By sparsely reconstructing the real tension fluctuation component, real-time real tension fluctuation data is formed. Step 3: Based on real-time tension fluctuation data, calculate the stress distribution along the entire line from the extrusion section to the traction section; extract the implicit viscoelastic stress fluctuation that causes the insulation layer to be eccentric from the stress distribution to obtain the stress distribution and implicit viscoelastic stress fluctuation information along the entire line. Step 4: Based on the stress distribution and implicit viscoelastic stress fluctuation information along the entire line, control commands are generated for synchronously adjusting the traction machine speed and the take-up reel torque, thus obtaining the synchronous control command; Step 5: Drive the traction machine and take-up reel according to the synchronous control command to keep the tension fluctuation of the cable within the preset accuracy range during high-speed operation.

2. The real-time tension fluctuation detection method for optimizing cable processing speed according to claim 1, characterized in that, Multiple non-contact detection points are set up along the production line, located at the extrusion port of the extruder, the middle section of the vulcanizing pipe, and before the traction machine. Simultaneously acquire laser Doppler vibration signals and eddy current deformation signals at each measuring point to obtain multimodal raw synchronous data, including: Non-contact sensing nodes are established at the extrusion port, the middle section of the vulcanizing pipe, and before the traction machine. Each non-contact sensing node synchronously senses the vibration response and deformation response of the cable surface at the corresponding measuring point, forming a multi-physics information acquisition system. Based on the multi-physics information acquisition system, each non-contact sensing node is driven by a preset global synchronous triggering benchmark to synchronously acquire vibration and deformation signals of each measuring point, thereby obtaining multi-measuring point synchronous vibration signal and multi-measuring point synchronous deformation signal. The synchronous vibration signals and synchronous deformation signals from multiple measurement points are spatiotemporally aligned and fused according to the measurement point location identifiers and timestamps to form multimodal raw synchronous data.

3. The real-time tension fluctuation detection method for optimizing cable processing speed according to claim 2, characterized in that, The multimodal raw synchronous data is input into a pre-defined cross-point causal comparison dynamic mode decoupling network. This network employs a working mechanism of transfer entropy causal constraints and bimodal contrastive learning to separate the true tension fluctuation components from the multimodal raw synchronous data, including: The multimodal raw synchronous data is received by a preset cross-point causal comparison dynamic mode decoupling network, and the transfer entropy value between adjacent measurement point pairs is calculated along the production line direction. The transfer entropy value is used to construct a cross-measurement point causal correlation map that characterizes the temporal relationship of tension fluctuation propagation between each measurement point. Using the cross-measurement point causal correlation spectrum as a constraint, the vibration signal and deformation signal of each measurement point are subjected to dual-modal comparative learning. In the hidden space inside the preset cross-point causal comparison dynamic mode decoupling network, the features that have cross-modal consistent response to tension fluctuations are separated to obtain the dual-modal intrinsic feature primitive set. The preset cross-point causal comparison dynamic mode decoupling network extracts the time-series components that satisfy cross-measurement point causal consistency from the dual-modal intrinsic feature primitives, which are then used as the true tension fluctuation components.

4. The real-time tension fluctuation detection method for optimizing cable processing speed according to claim 3, characterized in that, By sparsely reconstructing the actual tension fluctuation components, real-time actual tension fluctuation data is formed, including: By using a pre-defined cross-point causal comparison dynamic pattern decoupling network, the real tension fluctuation component is projected into a sparse representation space spanned by a pre-defined dynamic pattern atom library; and the sparse activation coefficient set corresponding to each dynamic pattern atom for the real tension fluctuation component is solved in the sparse representation space. Based on the sparse activation coefficient set and the pre-set dynamic pattern atom library, the tension fluctuation waveform after denoising and artifact suppression is reconstructed in the time domain to obtain real-time tension fluctuation data.

5. The real-time tension fluctuation detection method for optimizing cable processing speed according to claim 4, characterized in that, Based on real-time tension fluctuation data, the stress distribution along the entire line from the extrusion section to the traction section is calculated. The implicit viscoelastic stress fluctuations causing the insulation layer eccentricity are extracted from the stress distribution to obtain information on the overall stress distribution and implicit viscoelastic stress fluctuations, including: Based on the real-time tension fluctuation data, a coaxial cylindrical differential manifold model of the cable is constructed along the production line direction. On the coaxial cylindrical differential manifold model, an isomorphic mapping relationship between the tension fluctuation at each measuring point and the surface curvature of the cable is established to obtain the curvature-tension isomorphic mapping function. The real-time tension fluctuation data is mapped to the spatial position between each measuring point based on the instantaneous linear velocity to support stress calculation across the entire line. Based on the curvature-tension isomorphic mapping function, the corresponding time series values ​​of the real-time tension fluctuation data at each measuring point are transformed into a curvature distribution sequence along the entire line from the extrusion section to the traction section. Stress inversion is performed on the curvature distribution sequence to obtain the stress distribution along the entire line. The stress distribution along the entire line is projected onto the cross section of the coaxial cylindrical differential manifold model, and the circumferential stress asymmetry component caused by the non-equilibrium of the material curing path is extracted as the implicit viscoelastic stress fluctuation that causes the insulation layer to be eccentric, thus obtaining the information on the stress distribution along the entire line and the implicit viscoelastic stress fluctuation.

6. The real-time tension fluctuation detection method for optimizing cable processing speed according to claim 5, characterized in that, Based on the stress distribution and implicit viscoelastic stress fluctuation information along the entire line, control commands are generated for synchronously adjusting the traction machine speed and the take-up reel torque, resulting in synchronization control commands, including: The tension deviation characteristics within the traction section are extracted from the stress distribution along the entire line. The characteristics representing the eccentricity tendency of the insulation layer are extracted from the implicit viscoelastic stress fluctuation information. The tension deviation characteristics are combined with the characteristics representing the eccentricity tendency of the insulation layer to obtain the tension-eccentricity joint deviation characteristics. Based on the combined tension-eccentricity deviation characteristics, the traction machine speed compensation amount matching the tension deviation characteristics and the take-up reel torque compensation amount matching the eccentricity tendency characteristics are calculated to obtain the synchronization control command.

7. The real-time tension fluctuation detection method for optimizing cable processing speed according to claim 6, characterized in that, Driven by synchronous control commands, the traction machine and take-up reel are maintained within a preset accuracy range for overall cable tension fluctuations during high-speed operation, including: By decoupling the traction machine speed compensation amount and the take-up reel torque compensation amount in the synchronous control command, the traction machine speed adjustment command and the take-up reel torque adjustment command are obtained. The current speed setting value of the traction machine is updated with the traction machine speed adjustment command, and the current torque setting value of the take-up reel is updated with the take-up reel torque adjustment command, so that the traction machine and the take-up reel perform adjustment actions synchronously. After the traction machine and take-up reel perform adjustment actions, the real-time tension fluctuation data of the next sampling cycle is used to verify whether the tension fluctuation of the entire line has returned to the preset accuracy range, thus completing the closed-loop adjustment.

8. A real-time tension fluctuation detection device for optimizing cable processing speed, the device implementing the method as described in any one of claims 1 to 7, characterized in that, include: The multimodal synchronous acquisition module is used to deploy multiple non-contact detection points along the production line, located at the extrusion port of the extruder, the middle section of the vulcanizing pipe, and before the traction machine. Simultaneously acquire laser Doppler vibration signals and eddy current deformation signals at each measuring point to obtain multimodal raw synchronous data; The tension fluctuation decoupling and reconstruction module is used to input multimodal raw synchronous data into a preset cross-point causal comparison dynamic mode decoupling network. The preset cross-point causal comparison dynamic mode decoupling network adopts the working mechanism of transfer entropy causal constraint and bimodal comparison learning to separate the real tension fluctuation component from the multimodal raw synchronous data. Real-time real tension fluctuation data is formed by sparsely reconstructing the real tension fluctuation components. The stress calculation and latent fluctuation extraction module is used to calculate the stress distribution along the entire line from the extrusion section to the traction section based on real-time tension fluctuation data; and to extract the latent viscoelastic stress fluctuations that cause the insulation layer to be eccentric from the stress distribution, thereby obtaining the stress distribution and latent viscoelastic stress fluctuation information along the entire line. The synchronization control command generation module is used to generate control commands for synchronously adjusting the traction machine speed and the take-up reel torque based on the stress distribution and implicit viscoelastic stress fluctuation information of the entire line, thus obtaining the synchronization control command. The closed-loop control module is used to drive the traction machine and take-up reel according to the synchronous control command, so that the tension fluctuation of the cable is maintained within the preset accuracy range during high-speed operation.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images