An on-line regulating device and method for eccentricity of cable insulation layer

Abstract

The application discloses a kind of on-line regulation and control device and method of cable insulation layer eccentricity, it is related to eccentricity on-line regulation and control field, by obtaining the full circumferential high frequency profile point cloud data in cable insulation layer extrusion process, the full circumferential high frequency profile point cloud data is executed based on the entity layering reconstruction preprocessing of fluid space exclusivity, obtain discrete cross section set, the discrete cross section set is carried out time-frequency domain joint analysis, identifies and extracts spiral noise component, based on the spiral noise component, reconstructs virtual equivalent conductor reference circle, calculates real eccentric displacement vector, by the real eccentric displacement vector through core mechanism transfer function, in combination extruder head melt rheological delay characteristics, generate dynamic compensation parameter, eliminate the spiral phantom phenomenon caused by tight pressure conductor edge rotation scanning, provide pure geometric reference for eccentric control.

Description

Technical Field

[0001] This invention relates to the field of online eccentricity control, specifically to an online control device and method for the eccentricity of cable insulation layers. Background Technology

[0002] In the insulation extrusion process of high-performance power cables, non-contact online profile measurement devices are widely used to monitor the eccentricity of the insulation layer. Existing eccentricity calculation algorithms are generally based on an ideal cylindrical coupling model, which assumes that both the inner conductor core and the outer insulation layer have smooth circular boundaries, and characterizes the eccentricity by fitting the Euclidean distance between the inner and outer centers. However, when measuring compacted stranded conductors, especially large-section sector or circular compacted cores, the microstructure formed by multiple filaments during the compaction process is not a perfect cylindrical surface, but rather exhibits a "polygonal-like" profile with several protruding edges and planar transitions. When the measurement beam or ray is projected onto the conductor surface that rotates at high speed with traction, the polygonal edges and flat surfaces of the conductor surface alternately sweep across the sensor's field of view, causing the geometric projection algorithm based on the ideal circle assumption to incorrectly interpret the radial undulations of the conductor's microstructure as abrupt changes in the macroscopic thickness of the insulation layer.

[0003] This measurement error, caused by the deception of the conductor surface morphology, manifests as a spiral phantom oscillation phenomenon on the online monitoring terminal. Specifically, the measured thinnest point of the insulation does not stably point to the actual die offset direction, but rather rotates rapidly in a spiral shape with the stranding pitch of the conductor, resulting in high-frequency, large-amplitude numerical fluctuations. Existing core-aligning control strategies cannot distinguish between the false geometric projection error caused by the conductor edges and the true physical eccentricity caused by the die position offset. This makes it difficult for operators or automatic closed-loop control systems to lock onto the true eccentricity vector, often leading to mis-adjustment oscillations caused by high-frequency noise, and even inducing continuous oscillations in the core-aligning actuator, severely affecting the control stability of the insulation layer concentricity.

[0004] To address the aforementioned shortcomings, a technical solution is provided. Summary of the Invention

[0005] To address the technical problems mentioned in the background section, this invention is proposed. This invention provides an online control device and method for the eccentricity of cable insulation layers.

[0006] The objective of this invention can be achieved through the following technical solutions: A method for online control of cable insulation eccentricity includes the following steps: Acquire full-circumferential high-frequency contour point cloud data during the cable insulation extrusion process, and perform solid layer reconstruction preprocessing based on fluid-structure space exclusivity on the full-circumferential high-frequency contour point cloud data to obtain a set of discrete sections; A joint time-frequency domain analysis was performed on the discrete cross-section set to identify and extract the helical noise component; Based on the spiral noise components, a virtual equivalent conductor reference circle is reconstructed, and the real eccentric displacement vector is calculated. The actual eccentric displacement vector is transferred through the core-aligning mechanism transfer function, and combined with the melt rheological delay characteristics of the extruder head, to generate dynamic compensation parameters.

[0007] Furthermore, the analysis steps for the entity hierarchical reconstruction preprocessing based on fluid-structure interaction space exclusivity are as follows: A three-dimensional cylindrical projection space is constructed using the mechanical centerline of the extruder die as the absolute physical reference. The high-frequency contour point cloud data of the entire circumference is mapped onto the cross section of the three-dimensional cylindrical projection space, and a virtual elastic contraction ring is initialized on the periphery of the cross section. Furthermore, the analysis steps for the entity hierarchical reconstruction preprocessing based on fluid-structure interaction space exclusivity also include: Based on the melt rheological tension characteristics of insulating materials, a dynamic shrinkage annular geometric constraint boundary is set in the virtual elastic shrinkage ring. The equivalent tension function is calculated in combination with the head pressure. The radial shrinkage dynamic equation of each node is constructed using the pre-calibrated equivalent elastic modulus and damping coefficient. Based on the radial contraction dynamic equations of each node, the contact reaction force is obtained according to the contact relationship between the node and the projected point cloud during the numerical integration iteration process, and the stable boundary radius is obtained. The projected spatial physics is divided into an external steady-state fluid domain and an internal transient interference domain using this stable boundary.

[0008] Furthermore, the analysis steps for the entity hierarchical reconstruction preprocessing based on fluid-structure interaction space exclusivity also include: Spatial density clustering is performed on discrete data points falling within the internal transient interference domain to identify and instantiate them as solid interference clusters. The geometric centroid coordinates and radial intrusion vector of the entity interference cluster are calculated respectively to generate spatial pose attribute parameters; The data points retained within the external steady-state fluid domain are reconstructed into a continuous fluid flow envelope; The continuous fluid flow envelope and the spatial pose attribute parameters are encapsulated together to form the discrete cross-section set.

[0009] Furthermore, the analysis steps for performing joint time-frequency domain analysis to identify and extract the spiral noise component are as follows: A fast Fourier transform is performed on the time series stream of the discrete cross-section set to lock the characteristic frequency that is linearly related to the conductor traction speed and stranding pitch, which is used as the fundamental frequency of the helical motion. A three-dimensional spiral phase space is constructed using the fundamental frequency of the spiral motion. The two-dimensional cross-sectional profile data is mapped along the time axis into a continuous surface descriptor, and the local Gaussian curvature distribution on the continuous surface descriptor is calculated. Based on the extreme value distribution of local Gaussian curvature, the region with abrupt curvature changes and the region with smooth curvature are segmented to obtain the spiral noise component.

[0010] Furthermore, the analytical steps for calculating the true eccentric displacement vector are as follows: The spiral noise component is used to generate a spatial mask, and the discrete cross-section set is subjected to inverse mask filtering to retain data points belonging to the planar region as the effective fitting support set. The effective fitted support set is fitted with cylindricity using the weighted least squares method to reconstruct an ideal circular contour that eliminates the interference of polygonal edges, thus obtaining the virtual equivalent conductor reference circle. Furthermore, the analytical step of calculating the true eccentric displacement vector also includes: The geometric center of the outer contour of the insulation layer is obtained by using the area moment integral method based on the continuous fluid flow envelope. Calculate the Euclidean distance and azimuth angle between the geometric center of the outer contour of the insulation layer and the center of the virtual equivalent conductor reference circle, and output the true eccentric displacement vector.

[0011] Furthermore, the analysis steps for generating dynamic compensation parameters are as follows: Based on the real eccentric displacement vector, frequency domain analysis and statistical calculations are performed on the eccentric modulus and azimuth sequence to calculate the dominant frequency, root mean square value of modulus, azimuth variation amplitude and azimuth standard deviation. The system comprehensively determines whether the system is in the spiral phantom oscillation range based on the dominant frequency, root mean square value of the modulus, azimuth angle variation amplitude, and azimuth angle standard deviation. If it is confirmed to be in the spiral phantom oscillation range, a zero command is output. Furthermore, the analysis step for generating dynamic compensation parameters also includes: If the true eccentric displacement vector is determined to be a low-frequency effective signal, the true eccentric displacement vector is decomposed into lateral and longitudinal displacement components in the mold coordinate system, and the transfer function of the core-aligning mechanism is constructed through the stroke-eccentric coupling matrix and the proportional coefficient matrix. Based on the transfer function of the self-aligning mechanism, the lateral and longitudinal displacement components are converted into the mechanical stroke increment of each self-aligning actuator to generate the target value of mechanical stroke. Based on the melt rheological time constant, the target value of mechanical stroke is smoothed by using a discrete-time differential form, and stroke amplitude and speed constraints are applied to generate dynamic compensation parameters.

[0012] An online control device for the eccentricity of cable insulation layer, comprising: The acquisition and reconstruction module is used to acquire full-circumferential high-frequency contour point cloud data during the extrusion process of cable insulation layer, and to perform solid layer reconstruction preprocessing based on fluid-structure space exclusivity on the full-circumferential high-frequency contour point cloud data to obtain a set of discrete sections. The noise extraction module is used to perform time-frequency domain joint analysis on the discrete cross-section set to identify and extract the helical noise component; The eccentricity calculation module is used to reconstruct a virtual equivalent conductor reference circle based on the spiral noise component and calculate the real eccentricity displacement vector. The compensation and control module is used to generate dynamic compensation parameters by transferring the actual eccentric displacement vector through the core-aligning mechanism transfer function and combining it with the melt rheological delay characteristics of the extruder head.

[0013] Compared with the prior art, the beneficial effects of the present invention are: This invention acquires circumferential high-frequency profile point cloud data during the extrusion process of cable insulation, performs entity layer-by-layer reconstruction preprocessing based on fluid-structure interaction space exclusivity on the circumferential high-frequency profile point cloud data to obtain a set of discrete cross-sections, performs time-frequency domain joint analysis on the set of discrete cross-sections to identify and extract helical noise components, reconstructs a virtual equivalent conductor reference circle based on the helical noise components, and calculates the true eccentric displacement vector; through the dynamic shrinkage evolution of the virtual elastic shrinkage ring, the interference of the polygonal edges of the compressed conductor and the macroscopic profile of the insulating melt are effectively separated. The virtual elastic shrinkage ring is initialized in a three-dimensional cylindrical projection space, and an equivalent shrinkage driving force calculated from the die head pressure is applied, causing the virtual ring to simulate the natural shrinkage behavior of the melt from the outside to the inside. Using the dynamic shrinkage ring geometric constraint boundary formed after the virtual ring stops due to obstruction, the measurement data is physically divided into an internal transient interference domain and an external steady-state fluid domain. By reconstructing a virtual equivalent conductor reference circle by extracting only data points from the external steady-state fluid domain, the equivalent center of the conductor after ignoring edge details can be physically restored. This allows for the calculation of the true eccentric displacement vector, which eliminates interference from the conductor's micro-morphology. It also eliminates the spiral phantom phenomenon caused by rotating and scanning the tightly pressed conductor edges, providing a pure geometric reference for eccentricity control.

[0014] This invention generates dynamic compensation parameters by using the transfer function of the core-aligning mechanism through the actual eccentric displacement vector and combining it with the melt rheological delay characteristics of the extruder head. By combining the spiral phantom dead-zone logic with the melt rheological time constant constraint, smooth and precise control of the core-aligning action is achieved. Through frequency domain analysis and statistical calculation of the actual eccentric displacement vector, a judgment logic for the spiral phantom oscillation interval is constructed, which can actively identify and block invalid high-frequency oscillation signals and output a zero command. When generating the target value of the mechanical stroke, the melt rheological time constant, which describes the response hysteresis characteristics of the extruder head-melt-cooling system, is introduced. The core-aligning command is smoothed using a discrete-time differential form, and dual constraints on the amplitude and speed of the action are applied. The dynamic compensation parameters generated in this way ensure that the core-aligning mechanism can follow the actual low-frequency eccentricity changes while avoiding system resonance caused by sudden command changes or over-adjustment, ensuring the stability of online control of the insulation layer eccentricity under complex working conditions. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. The following drawings are not drawn to scale according to the actual size, but are intended to show the main idea of ​​the present invention.

[0016] Figure 1 A flowchart illustrating an online method for controlling the eccentricity of cable insulation; Figure 2 This is a system block diagram of an online control device for the eccentricity of cable insulation. Figure 3 This is a schematic diagram illustrating the principle of hierarchical reconstruction of entities based on the exclusivity of fluid-structure interaction. Figure 4 This is a schematic diagram of feature extraction and reconstruction of a discrete cross-section set. Detailed Implementation

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are also within the scope of protection of the present invention.

[0018] Example 1 like Figure 1 As shown, an online method for controlling the eccentricity of cable insulation includes the following steps: Step S100: Obtain the full circumferential high-frequency contour point cloud data during the cable insulation extrusion process, and perform solid layer reconstruction preprocessing based on fluid-structure space exclusivity on the full circumferential high-frequency contour point cloud data to obtain a set of discrete sections. A non-contact multi-axis high-speed laser scanning measurement unit is installed in front of the cooling section or at the inlet of the warm water tank in the cable insulation extrusion production line. A three-dimensional cylindrical measurement coordinate system is constructed with the mechanical center axis of the extruder die as the Z-axis (i.e., the cable traction direction) and the section perpendicular to the Z-axis as the XY plane. The specific process for acquiring full-circumferential high-frequency contour point cloud data is as follows: The multi-axis high-speed laser scanning measurement unit is activated. This unit has points evenly distributed along the circumferential direction... One (e.g., eight or more) high-precision laser displacement sensors, or a single-beam laser scanning head rotating at extremely high speeds, can be used to achieve full circumferential measurement of the cable cross-section from 0 to 360° without blind spots. For example, 8 or more. Set a high-frequency sampling strategy, sampling frequency... The sampling frequency must satisfy the Nyquist sampling theorem and oversampling must be performed on the polygonal edge features of the compacted conductor. The calculation formula is .in, This refers to the real-time traction speed during cable production. For compacting the strand pitch of the conductor, This refers to the number of single-wire edges or polygon vertices on the cross-section of a conductor. An oversampling safety factor, typically greater than 10, is set to ensure waveform fidelity. This frequency setting ensures the measurement system can clearly capture the minute undulations in the insulation layer caused by the rapidly rotating conductor edges passing through the measurement cross-section. During measurement, the sensor continuously acquires the radial distance of the cable surface relative to the mechanical center axis. Combined with the current real-time traction line speed and sampling timestamp By spatially stacking two-dimensional cross-sectional measurements from a continuous time series along the Z-axis, a model containing spatial coordinates is constructed. The massive set of discrete data points is defined as high-frequency contour point cloud data in the full circumference direction, where The circumferential azimuth angle of the measurement point on the cable cross-section is specifically defined by a fixed radial direction (such as the positive rightward X-axis) set in the XY plane perpendicular to the machine's central axis (Z-axis) as the zero-degree reference line. This represents the angle from the reference line, along a predetermined rotation direction (e.g., counterclockwise), to the current measurement point. If the device uses... If there are sensors evenly distributed along the circumference, then The fixed installation orientation corresponds to each sensor; if a rotating scanning head is used, then... The instantaneous angular position corresponding to the scanning beam. (Through...) This uniquely determines the geometric position of the measurement point in the three-dimensional cylindrical coordinate system. The point cloud data not only contains the macroscopic geometric contour information of the insulating layer (low-frequency part, corresponding to the subsequent external steady-state fluid domain), but also completely preserves the microscopic texture features formed by the internal pressure of the conductor edges against the insulating layer, distributed spirally along the axial direction (high-frequency part, corresponding to the subsequent internal transient interference domain). This provides a data foundation containing high-precision surface morphology features for subsequent entity layer-by-layer reconstruction based on fluid-structure space exclusivity. The entity layer-by-layer reconstruction based on fluid-structure space exclusivity refers to the process in steps S101-S104 where, in the three-dimensional cylindrical projection space, the dynamic convergence boundary of the virtual elastic contraction ring driven by melt tension is used to exclusively divide the circumferential point cloud data in physical space into an external steady-state fluid domain representing the macroscopic contour of the insulating melt and an internal transient interference domain representing the interference from the conductor edges. Then, continuous fluid envelope and entity interference cluster features are extracted from the data in both domains to achieve layered independent reconstruction of the fluid morphology of the insulating layer and the microscopic features of the conductor solid, resulting in a set of discrete sections.

[0019] Step S100 involves performing entity hierarchical reconstruction preprocessing based on fluid-structure interaction space exclusivity, specifically including the following steps: like Figure 3 The diagram shown is a schematic of the principle of entity hierarchical reconstruction based on the exclusivity of fluid-structure space.

[0020] Reference Figure 3 , Figure 3 This demonstrates the physical modeling process in a coordinate system with the mechanical center (Z-axis) of the extruder die as the origin. The outermost dashed circle represents a virtual elastic contraction ring (its initial radius is...). During the simulation, the ring is subjected to an equivalent contraction driving force calculated from the nose pressure. (As shown by the orange downward arrow in the figure) simulates the natural contraction behavior of the melt, moving from the outside to the inside.

[0021] The blue closed solid line in the middle of the diagram represents the stable boundary formed after contraction reaches quasi-static equilibrium. This boundary physically divides the entire projection space into two parts: the inner transient interference domain, as shown in the pink highlighted circular area in the figure, corresponding to the disturbed area pressed against the conductor edge in step S102, which belongs to the high-frequency noise that needs to be isolated; and the outer steady-state fluid domain, corresponding to the smooth part on the stable boundary, representing the macroscopic flow profile of the insulating melt, used for subsequent effective fitting. The entire process demonstrates a modeling mechanism that automatically segments the physical measurement point cloud into conductor edge interference and effective fluid profile using the dynamic evolution of the virtual ring.

[0022] Step S101: Using the mechanical centerline of the extruder die as the absolute physical reference, construct a three-dimensional cylindrical projection space, map the full circumferential high-frequency contour point cloud data onto the cross section of the three-dimensional cylindrical projection space, and initialize a virtual elastic contraction ring around the periphery of the cross section. Reference Figure 3 To establish an absolute physical reference, the geometric center axis of the extruder die exit is calibrated using a high-precision laser alignment instrument and set as the Z-axis in the three-dimensional cylindrical projection space. (Refer to...) Figure 3 A cylindrical virtual space with a radius larger than the maximum theoretical outer diameter of the cable is constructed with the Z-axis as the center; this is the three-dimensional cylindrical projection space. The maximum theoretical outer diameter of the cable mentioned here is the sum of the nominal outer diameter value determined by the process specifications of the current cable model or relevant national standards and the maximum allowable positive tolerance. Relevant national standards include GB / T12706. For example, if the nominal outer diameter of the cable is... The maximum allowable upper deviation of the process is Then the maximum theoretical outer diameter of the cable is This setting is to ensure that the constructed virtual space is physically large enough to accommodate cable entities under any operating conditions, preventing data overflow. It will include timestamps. and spatial coordinates The full-circumferential high-frequency contour point cloud data is radially projected onto a two-dimensional XY plane perpendicular to the Z-axis, according to its corresponding Z-axis coordinates. This two-dimensional XY plane represents the cross-section. The original three-dimensional spiral point cloud data is compressed and mapped into a discrete set of points on a two-dimensional plane. .in, This represents the total number of valid discrete data points contained within the current sampling period or unit projected length. Its value depends on the ratio of the sampling frequency of the laser scanning unit to the cable traction speed. This is the sequence index number of a single data point in the discrete point set, with values ​​increasing from 1 to... ; For the first The radial distance from each point to the Z-axis, For the first The circumferential angle at each point. On this cross-section, with the Z-axis as the center, initialize a circle with a radius of... Virtual elastic contraction ring, such as Figure 3 The outermost dashed circle is shown. Set to be greater than all projected points A value 1.1 to 1.5 times the maximum value ensures that the virtual ring completely surrounds and floats outside all measured point clouds in the initial state.

[0023] Step S101, the cable insulation eccentricity, is essentially the cross-sectional geometric offset relative to the mechanical centerline of the extruder die. This is a typical cross-sectional geometry problem. Therefore, the data must first be normalized and projected within a unified cylindrical coordinate system centered on the die centerline, ensuring that all subsequent "radial distances" and "eccentric displacements" have a unified physical meaning. Simultaneously, the original contour data obtained from laser scanning during extrusion is spirally distributed along the axial direction. Direct processing in three-dimensional space is not conducive to extracting cross-sectional features and defining control quantities. Converting and projecting this data according to axial position onto a cross-section perpendicular to the centerline simplifies the problem dimension and preserves the most crucial cross-sectional information for eccentricity analysis. Pre-setting a virtual annular boundary that completely encloses all measurement points within this unified cross-sectional space provides stable initial geometric conditions for subsequent "outside-in" searching of the actual fluid boundary using numerical evolution methods, resulting in good convergence and repeatability of the region partitioning process.

[0024] Step S102: Based on the melt rheological tension characteristics of the insulating material, a dynamic shrinkage annular geometric constraint boundary is set in the virtual elastic shrinkage ring. The equivalent tension function is calculated in combination with the head pressure. The radial shrinkage dynamic equation of each node is constructed using the pre-calibrated equivalent elastic modulus and damping coefficient. Based on the radial contraction dynamic equations of each node, the contact reaction force is obtained according to the contact relationship between the node and the projected point cloud during the numerical integration iteration process, and the stable boundary radius is obtained. The projected spatial physics is divided into an external steady-state fluid domain and an internal transient interference domain using this stable boundary.

[0025] In the constructed three-dimensional cylindrical projection space and its cross-section, with the mechanical centerline of the extruder die as the Z-axis and the section perpendicular to the Z-axis as the XY plane, the aforementioned virtual elastic contraction ring is discretized and physically modeled within this XY plane: the initial radius is set as... The virtual elastic contraction ring is uniformly discretized along the circumferential direction as follows: The virtual mass point, i.e., the node, is the first virtual elastic contraction ring. Each node at time... The instantaneous radius is defined as ,in The node number is obtained by discretizing the virtual elastic contraction ring along the circumference, and the value ranges from 1 to 1. , This represents the total number of discrete nodes in the virtual elastic contraction ring. Indicates the azimuth angle The radial distance from this node to the Z-axis. As a time variable synchronized with the cable extrusion process, it characterizes the time step of the virtual elastic shrinkage ring during its shrinkage evolution. Based on the rheological tension characteristics of the insulating material melt under the current extrusion conditions, an equivalent tension function is defined to drive the ring to shrink inward. Specifically, a melt pressure sensor is installed at the extruder die outlet, and the melt pressure sensor measures the current time in real time at a preset sampling period. Extruder head pressure value During the equipment commissioning phase, the user or equipment manufacturer pre-selects a constant proportional coefficient based on the type of insulation material used, the extruder specifications, and typical production conditions. This proportionality coefficient is then fixed in the control program or parameter table; during actual operation, the equivalent tension function value is calculated at each sampling moment according to the following simple linear relationship: ; in, For the current moment The equivalent tension scalar used to drive the virtual elastic contractile ring to contract inward. For the current moment The measured pressure value of the extruder head. To convert the dimensions of the head pressure into a proportional constant of the equivalent contraction tension dimension, due to The contraction response was selected based on experience under typical operating conditions during the commissioning phase and remains unchanged during operation. To give the equivalent elastic modulus of the virtual elastic contraction ring material and damping coefficient In this embodiment, the virtual elastic shrinkage ring is regarded as a numerical analogy to the macroscopic deformation behavior of the real insulating melt on the cross-section near the mold exit, and is determined in the following way. and Under laboratory conditions, typical cable specifications were selected, and the extrusion temperature, traction speed, and screw speed were kept stable at different operating points. The radial shrinkage of the insulation outer diameter under traction tension near the die exit was recorded using an online profile measurement system. Response curves as a function of time or linear velocity, with corresponding extruder head pressures also recorded. The least squares fitting method is used to find a set of... and This causes the virtual ring radius calculated based on the mass-spring-damping model to change. In terms of time response characteristics, it should approximate the aforementioned actual contraction curve as closely as possible. When the fitting error is lower than a preset threshold, the corresponding... and As equivalent parameters for this type of cable structure and material system, these are written into the control system, and the time response characteristics include contraction speed, overshoot, and stability value; [the text abruptly ends here, likely due to an incomplete sentence or missing information.] and Then, the virtual elastic contraction ring is discretized into Nodes The coordinates of each node in the polar coordinate system are , The azimuth angle of the node is usually divided into equal intervals as follows: In the initial state season , The initial radius set in step S101, The distance should be 1.1 to 1.5 times greater than the maximum radial distance of all measurement points at the current cross-section to ensure that the entire ring is outside the cloud of measurement points; during the dynamic shrinkage process, for each node... Apply a contraction driving force radially towards the Z-axis. ,like Figure 3 As indicated by the orange arrow, this contraction driving force is based on the equivalent melt rheological tension. Calculation, for example, can be represented as ,in, To map the real melt stress to the virtual ring shrinkage equivalent stress, For nodes The corresponding minute arc length along the circumferential tangential direction. Combined with the equivalent elastic modulus of the virtual elastic shrinkage ring material and damping coefficient The following radial contraction dynamic equation is established for the radial motion of each node: ; It should be noted that all terms in the equations represent line loads per unit axial depth, expressed in Newtons per meter (N / m), such as elastic restoring force and viscous damping force. The term represents the elastic restoring force generated by the virtual ring when it is compressed. The term represents the viscous damping force generated during the radial movement of the node. This represents the contact reaction force exerted by the rigid data points on the virtual flexible ring when the node comes into contact with the projected point cloud; this is for calculating... At each time step, the current time is... Next node radius Compared to the previous time step radius at time Comparison, among which This represents the radial distance from the node to the Z-axis at the end of the previous time step: in azimuth angle Retrieve the projected point cloud within its neighborhood. All measurement points When a certain measurement point satisfies and At that time, it is considered that the node The first contact occurs at the measurement point within the current time step; this measurement point is recorded as the contact point, and the contact depth is calculated. ,when At that time, the contact reaction force is calculated based on Hooke's law or a simplified linear elastic contact model. For example, it is advisable ,in The contact geometry correlation coefficient is used to reflect whether the contact is closer to a line contact or a surface contact, etc., in terms of force distribution. For the node The maximum contact depth among all contact measurement points within the azimuth neighborhood reflects the strongest rigid barrier effect in that azimuth; if within a certain time step, at a certain node If no contact points meeting the above conditions are detected nearby, then let This indicates that the virtual ring contraction motion in this angular direction is in a free contraction state, unhindered by the point cloud entities, and will continue to move inward according to the contraction law driven by melt rheological tension. It should be noted that the fundamental basis for establishing this radial contraction dynamic equation is the force balance principle of Newtonian mechanics. The left side of the equation is strictly constructed based on the viscoelastic constitutive equation of polymer materials. Hooke's law is used to convert the elastic modulus of the material into the elastic restoring force resisting deformation, and the law of viscous fluids is used to convert the damping coefficient of the material into the viscous drag hindering motion. Combined with the geometric physical quantity of the tiny arc length of the node, the actual force on the melt element represented by the node at a unit axial depth is accurately calculated. The right side of the equation is calculated based on the measured extruder head pressure to obtain the equivalent contraction thrust. The combination of the two and the introduction of fluid-solid space contact reaction force realizes the physical quantitative description from the microscopic properties of the material to the macroscopic contraction motion. Based on the above dynamic equation, conventional numerical integration methods such as the Euler method, the improved Euler method, or the fourth-order Runge-Kutta method are used to calculate the force at each time step. The change in the value is solved iteratively to obtain the radius of each node. From initial value It gradually decreases during the evolution of time; the corresponding movement inward according to the contraction law driven by melt rheological tension specifically means: in The free contraction region, Only subject to , and The three factors work together, and their evolution is entirely obtained by numerical integration of the above dynamic equations. That is, the same set of dynamic equations is solved repeatedly at each new time step to achieve the gradual inward contraction of the loop in that orientation. During the shrinking iteration, when all nodes Change over several consecutive time steps All are below the preset convergence threshold At this point, it is considered that the virtual elastic contraction ring has reached a quasi-static equilibrium state, and at this time, each node... coordinates A closed curve obtained by interpolation or spline fitting, i.e. Figure 3 The stable boundary shown by the blue closed solid line in the middle represents the final stable form of the dynamically shrinking annular geometric constraint boundary. This closed curve remains variable in shape and position throughout the entire shrinkage evolution process, but the stable boundary curve obtained after convergence is denoted as... Based on this, for any projection point Through On-axis search and The corresponding boundary azimuth angle interval is determined, and the stable boundary radius at that azimuth angle is obtained using linear interpolation or spline interpolation. ,Right now In the angle The node set The radius value of the fitted stable boundary curve; using this stable boundary radius For all measurement points in the projection space Perform physical region division: divide all regions that meet the requirements The set of measurement points is labeled as being within the internal transient interferometric domain, such as... Figure 3 The highlighted pink area represents the region where solid-fluid coupling is disturbed during the actual extrusion process due to factors such as the pressure from the conductor edge and local melt flow disturbances; all areas satisfying the requirements... and The set of measurement points is labeled as being located in an external steady-state fluid domain, such as... Figure 3 As shown, this region corresponds to the macroscopically smooth, approximately axisymmetric stable flow region formed by the insulating melt near the die exit during the actual extrusion process. Through the above numerical evolution and region division process with the dynamic shrinking annular geometric constraint boundary as the core, a one-to-one correspondence is established between each discrete point cloud data point and its physical domain type, forming a projected point cloud with physical semantic annotations of "steady-state fluid domain / transient interference domain". This provides directly usable and physically meaningful input data for the subsequent step S103, which only performs spatial density clustering, identifies entity interference clusters, and calculates their spatial pose attribute parameters for data points within the internal transient interference domain.

[0026] In step S102, the insulating layer is in a molten or highly elastic state near the mold exit. Macroscopically, it behaves as a viscoelastic fluid subjected to the combined effects of traction tension and die head pressure. Its overall contraction and local undulations on the cross-section can be mechanically analogized to a "viscoelastic ring driven by equivalent tension." By experimentally calibrating the equivalent elastic modulus and damping parameters, the contraction time response of the virtual ring is matched with the outer diameter contraction curve of the actual melt in terms of speed, overshoot, and stability. Numerically, the deformation behavior of the real melt cross-section can be approximated using a set of dynamic equations with discrete nodes. Furthermore, the actual measurement points are considered as rigid "obstacles" that do not move with the virtual ring. During the inward contraction of the virtual ring, a contact reaction force is introduced based on whether the node contacts the point cloud and the depth of contact. This is equivalent to realizing a fluid-structure interaction process in numerical simulation where a "flexible structure automatically stops or slows down its contraction when encountering a rigid entity." As the time integration progresses and the motion of each node gradually converges, a stable closed boundary is formed. This boundary can be understood as the effective dividing contour of the melt on the cross-section under the current operating conditions: its inner side is a region strongly affected by the pressure of the conductor edges and local flow field disturbances, while its outer side is a smooth region mainly controlled by overall flow and traction tension. Using this stable boundary to physically divide the projected space is a practice based on the viscoelastic mechanics model, simplified assumptions of contact mechanics, and the convergence of numerical integration. It can abstract and layer the complex three-dimensional fluid-structure interaction effects on the cross-section in a computable and realizable manner. This is the fundamental reason why step S102 is proposed separately and is the core physical modeling step.

[0027] like Figure 4 The figure shows a schematic diagram of feature extraction and reconstruction of discrete cross-section sets.

[0028] Reference Figure 4 The black dot at the center of the diagram represents the mechanical center (Z-axis) of the extruder die, serving as the absolute geometric reference for feature extraction. The area within the red dashed box in the diagram is labeled as a solid interference cluster. These clusters are independent physical objects identified based on the spatial density clustering of measurement points falling into the internal transient interferometric domain in step S103. Each entity interferometric cluster Physically, this corresponds precisely to the localized mechanical pressure region exerted on the insulation layer by a polygonal edge or vertex of a compacted conductor on the current cross-section. The solid red arrows emanating from within each solid interference cluster represent radial intrusion vectors. Its direction is directly towards the center of the circle (Z-axis) in the radial direction, and its modulus quantifies the penetration depth of the edge relative to the ideal fluid boundary, thus characterizing the intensity and spatial orientation of the microscopic disturbance of the conductor edge to the insulating layer in a vectorized form.

[0029] Step S103: Perform spatial density clustering on discrete data points falling within the internal transient interference domain, identify them, and instantiate them as solid interference clusters; The geometric centroid coordinates and radial intrusion vector of the entity interference cluster are calculated respectively to generate spatial pose attribute parameters; From the projection point cloud Measurement points that are selected and divided into internal transient interferometric domains are used to form an interferometric point set. For this set of interference points, a density-based spatial clustering algorithm, such as the DBSCAN algorithm, is used, with a defined neighborhood radius. and minimum points Cluster analysis is performed on these discretely distributed interference points. Through clustering, spatially adjacent interference points are grouped into independent clusters, and each cluster is instantiated as a physical interference cluster, denoted as . , , This represents the total number of clusters identified. Each entity interference cluster... Physically, this corresponds to the pressure region of an edge or vertex of a compacted conductor against the insulating layer. For each solid interference cluster... Calculate its geometric centroid coordinates The calculation formula is: , ,in, For clusters The number of measurement points included. Let be the projection point. and They represent the first The radial distance coordinates and circumferential azimuth coordinates of the geometric centroid of an interference cluster on the cross-section of a three-dimensional cylindrical projection space are calculated. atan2 is the arctangent function in the four quadrants. The azimuth coordinates of the geometric centroid are calculated using the circular mean method to correctly handle the case where the azimuth crosses the 0 / 2π boundary. The radial intrusion vector of this cluster is calculated. The magnitude of this vector is defined as the sum of the magnitudes of all points within the cluster relative to the stable boundary. The maximum depth of penetration, i.e. The direction is radial, pointing towards the center of the circle, and `max` represents the function that takes the maximum value. The geometric centroid coordinates of all identified solid interference clusters are... With radial intrusion vector Combining these parameters generates spatial pose attribute parameters that describe the spatiotemporal distribution characteristics of the conductor's edges, denoted as... Spatial pose attribute parameters quantify the intensity and location of the microscopic perturbation of the conductor edge to the insulating layer.

[0030] In step S103, each measurement point in the projected point cloud has been assigned a clear physical domain attribute label, namely, "internal transient interference domain" or "external steady-state fluid domain." The former mainly represents the pressure and local disturbance regions of the conductor polygon edges on the insulating layer. For these regions, their spatial distribution characteristics are not uniform and random, but rather originate from several edges that periodically sweep across the cross-section as the conductor rotates. On any cross-section, they appear as a finite number of locally high-density clusters of points. The density-based spatial clustering method is used to automatically group these points, taking advantage of their distribution characteristics of "small point spacing within clusters and large inter-cluster spacing." Moreover, it does not require pre-specifying the number of clusters and naturally adapts to the variation in the number of edges under different conductor structures. Compared with the traditional approach of fitting the overall contour to a single curve, this method can more finely distinguish the local influence of different edges on the insulating layer. Furthermore, the geometric centroid of each cluster is used to represent the comprehensive position of the corresponding edge on the insulating layer pressure region, and the "intrusion depth" or disturbance intensity of the edge is measured by the maximum radial intrusion relative to the stable boundary. This is equivalent to compressing the massive point cloud information into a small number of "edge pose parameters" with clear physical meaning. Therefore, performing spatial density clustering on the points in the internal transient interference domain and then calculating the centroid and intrusion vector is a feature extraction method based on the combination of statistical geometry and physical semantics. This allows the present invention to process the spatiotemporal distribution of conductor edges in vector form, rather than directly dealing with the original noise point cloud, thereby improving the robustness and feasibility of the algorithm. This is the technical logic behind setting up a separate step S103.

[0031] Step S104: Reconstruct the data points retained in the external steady-state fluid domain into a continuous fluid flow envelope; The continuous fluid flow envelope and the spatial pose attribute parameters are encapsulated together to form the discrete cross-section set.

[0032] The measurement points are extracted and divided into external steady-state fluid domains to form a fluid point set. These points represent the natural fluid state of the insulation layer when it is not subjected to pressure from conductor edges. The fluid point set is then fitted using B-spline curves or Fourier series methods. A smooth reconstruction is performed to generate a continuous closed curve, which is the envelope of the continuous fluid flow, denoted as . This envelope physically characterizes the macroscopic profile of the insulating layer melt at the mold exit, eliminating interference from high-frequency edge noise. The reconstructed continuous fluid flow envelope... Spatial pose attribute parameters Data encapsulation is performed. The encapsulated data structure is a set of discrete cross-sections. ,Right now This discrete cross-section set, as a type of structured data containing macroscopic fluid profiles and microscopic edge perturbation features (the macroscopic fluid profiles being low-frequency signals and the microscopic edge perturbation features being high-frequency signals), provides precise inputs with physical layering and semantic annotation for the subsequent analysis based on helical manifold features in step S200 and the eccentricity calculation in step S300, realizing the dimensionality upgrade from the original point cloud to the physical feature set.

[0033] In step S104, the measurement points within the external steady-state fluid domain primarily reflect the macroscopic contour determined by the overall melt flow and mold geometry. Their shape should be relatively smooth and approximately axisymmetric. However, the points in the internal transient interference domain are superimposed with high-frequency fluctuations caused by factors such as conductor edge geometry, twist pitch, and axial helical motion. If the entire point cloud is used indiscriminately during overall contour reconstruction, local protrusions caused by the edges will be mistakenly treated as part of the "true outer contour," leading to helical artifacts in subsequent eccentricity calculations. Therefore, this invention selects only the points identified as belonging to the external steady-state fluid domain to fit a continuous closed curve, essentially preserving only the "natural fluid morphology unaffected by conductor edge pressure." Residual high-frequency noise is further suppressed using smoothing reconstruction methods such as B-splines or Fourier fitting, resulting in a low-frequency contour line representing the macroscopic flow envelope of the melt. Furthermore, this continuous fluid flow envelope, together with the edge spatial pose parameters obtained in step S103, is encapsulated into a set of discrete cross-sections. Each cross-section carries both "macroscopic contour information" and "microscopic edge perturbation information," providing structured input data for subsequent helical manifold analysis and eccentricity calculation. This processing logic—"first filtering data based on physical partitioning, then constructing a macroscopic envelope using smooth fitting, and simultaneously encapsulating it with microscopic features"—ensures the physical purity of the eccentricity calculation benchmark and the completeness of the data description.

[0034] A method for online control of cable insulation eccentricity further includes the following steps: Step S200: Perform time-frequency domain joint analysis on the discrete cross-section set to identify and extract the helical noise component; The process of extracting the spiral noise component in step S200 specifically includes the following steps: Step S201: Perform a fast Fourier transform on the time series stream of the discrete cross-section set to lock the characteristic frequency that is linearly related to the conductor traction speed and stranding pitch, as the fundamental frequency of the helical motion; Using the timestamp t of each recorded measurement point, the set of static discrete cross-sections for each frame is arranged longitudinally according to the chronological order of occurrence, constructing a time series of discrete cross-section sets that evolves continuously over time. Based on the time series of discrete cross-section sets, spatial pose attribute parameters encapsulated in each frame of data are extracted. These parameters include the radial intrusion vectors of each entity interference cluster. To transform these spatial features scattered on the two-dimensional cross-sections into a time-series signal suitable for one-dimensional frequency domain analysis, a spiral texture pulsation signal characterizing the pressure intensity of the conductor edge on the insulating layer is constructed. This signal at each time step... The value is defined as the arithmetic mean of the radial intrusion vector magnitudes of all identified entity interference clusters at that moment. The arithmetic mean is chosen as the analysis object because, ideally, the geometric dimensions and pressure of each edge of the conductor should be uniform. The arithmetic mean can effectively smooth out random noise caused by single-point measurement errors or local impurities, thus more robustly reflecting the periodic excitation effect of the overall macroscopic profile characteristics of the conductor cross-section on the insulation layer. A fast Fourier transform is performed on the helical texture pulsating signal to obtain the frequency domain power spectrum, and several peak frequencies with significant energy are searched in the frequency domain spectrum. The real-time cable traction speed is used to... and the preset tight conductor twist pitch The theoretical reference frequency for conductor rotation is calculated based on the helical propulsion formula. The frequency domain spectrum is related to the theoretical reference frequency. The main energy peak with deviation within the preset allowable range is locked as the characteristic frequency. Within the preset allowable range, for example, ±5%, it is confirmed as the fundamental frequency of the spiral motion that characterizes the spiral rotation of the conductor's edge. Thus, the physical interference characteristics separated in the previous steps are accurately used to precisely anchor the conductor's own rotational properties from the complex random vibration noise of the production line.

[0035] Step S202: Construct a three-dimensional spiral phase space using the fundamental frequency of the spiral motion, map the two-dimensional cross-sectional profile data along the time axis into a continuous surface descriptor, and calculate the local Gaussian curvature distribution on the continuous surface descriptor; Read timestamps in a computer A time-series stream of discrete cross-sections arranged in sequence, wherein each frame of the discrete cross-section set encapsulates at least the continuous fluid flow envelope of the outer contour of that frame's cross-section. and the corresponding sampling time ,in Indicates the first Sampling time of frame segment, The value range is 1 to , This represents the total number of frames in the cross-section within the current analysis time window; the fundamental frequency of the helical motion is used. Define a spiral phase quantity for each frame section. This is used to characterize the relative phase position of the cross section with respect to the rotational motion of the conductor edge, and can be specifically expressed by the formula... and to according to Mold conversion to The interval is used to make it a periodic phase variable.

[0036] The outer contour of each frame section Using a fixed angular step size in polar coordinates Perform discrete sampling to obtain the result at that moment. Azimuth Envelope radius at point ,in For the first Azimuth angle sampling points The value range is 1 to , This represents the number of discrete sampling points along the circumferential direction on a frame cross-section. The horizontal axis is the independent variable, and the spiral phase is used. The vertical axis is the independent variable, and the envelope radius is the variable. The height value is in the three-dimensional coordinate system. All sampling points of all frames Perform regular grid interpolation or spline interpolation to construct a grid in... Dimensions and A scalar field that is continuous and smooth in all dimensions This scalar field geometrically corresponds to a continuous surface descriptor. This surface descriptor unfolds the original time-independent two-dimensional cross-sectional profile data along the helical phase axis into a two-dimensional manifold reflecting the correspondence between the helical motion of the conductor's edges and the profile undulations. To calculate the local Gaussian curvature distribution on this surface, a selection is made... any point on and in direction and For several adjacent grid nodes along a given direction, the two principal curvatures of the surface at that point in the principal direction are obtained using a quadratic polynomial local surface fitting method or a bicubic B-spline local fitting method. and And accordingly The local Gaussian curvature value at this point was calculated. , Point The local Gaussian curvature value; repeat the above operation for all mesh nodes to obtain the same... One-to-one correspondence of local Gaussian curvature distribution It is stored in the computer as a matrix for subsequent region segmentation based on curvature extremum distribution.

[0037] Step S203: Based on the extreme value distribution of local Gaussian curvature, segment the ridge region with abrupt curvature changes and the planar region with smooth curvature to obtain the spiral noise component.

[0038] To distinguish between the high-frequency fluctuation trajectory caused by the helical pressing of the conductor's polygonal edges and the smooth region formed by the macroscopic flow of the insulating melt, Preprocessing is performed, including absolute value normalization and spatial smoothing filtering. Specifically, one aspect is to normalize the absolute value of curvature of each grid node. Normalize by the global maximum value so that it falls within... The interval is used to set the threshold later; on the other hand, a two-dimensional Gaussian kernel or median filter kernel is used in... Plane Convolutional smoothing is performed to suppress the interference of isolated noise points on extremum determination. The statistical properties of the smoothed absolute curvature field are calculated, such as obtaining its global mean. with standard deviation Based on this, a threshold for determining curvature extrema is set. , can be adopted In the form of, This is an empirical coefficient, preferably ranging from 1 to 3, used to adjust the sensitivity of ridge region extraction. For continuous surfaces... Any discrete grid node on If the absolute value of the normalized curvature at that node satisfies If so, then mark the node as a high curvature node; if If so, then the node is marked as a low curvature node. On the plane, the set of all points marked as high curvature nodes approximately forms a narrow band along the spiral phase. The directional distribution of this high-curvature band physically corresponds to the curvature abrupt change trajectory formed when the polygonal edges of the conductor are swept sequentially across each measurement section along the stranding pitch. Therefore, the band-shaped region composed of high-curvature nodes is defined as the curvature abrupt change ridge region; the remaining region composed of low-curvature nodes is defined as the curvature smooth plane region, which physically corresponds to the macroscopic smooth contour of the insulating melt when it is far from the edge pressure interference. This is achieved through the above-mentioned local Gaussian curvature absolute value and single threshold... The classification process can be carried out in order to... Clear division between curvature abrupt ridge regions and curvature smooth plane regions can be obtained directly on the plane.

[0039] All mesh nodes belonging to the curvature abrupt ridge region By reversing the mapping relationship, the specific data point indices in the original discrete cross-section set time series stream are restored, that is, based on... Corresponding time Physical measurement points near the azimuth angle and time are located in the discrete cross-section set. These measurement points are uniformly labeled as helical noise components. Physical measurement points refer to individual data points in the contour point cloud data directly acquired by the multi-axis high-speed laser scanning measurement unit arranged on the extrusion production line, that is, those with polar coordinates defined in step S100. and corresponding timestamps The original point cloud samples are used, and these points are the most basic building blocks of the discrete cross-section set. Data points belonging to the curvature smooth plane region are marked as effective smooth region points without helical noise, serving as the effective fitting support set for constructing the heterogeneous contour topology decoupling operation architecture and reconstructing the virtual equivalent conductor reference circle in subsequent step S300. Through the above-mentioned ridge and plane region segmentation based on the local Gaussian curvature extremum distribution, this invention can accurately extract the helical noise component caused by the conductor edge geometry from the overall contour data while ensuring computational feasibility, providing a reliable data foundation for high-precision calculation of eccentricity.

[0040] In step S200, the "pressure marks" on the cross-section of the conductor's geometric edges are not static but sweep across each measurement cross-section sequentially with a near-fixed angular velocity over time, causing approximately periodic disturbances to the insulation layer profile. These disturbances possess distinct time-frequency characteristics and manifest as surface undulations propagating along a spiral trajectory in the cross-sectional space. Therefore, analyzing only in the time domain or on a single frame cross-section makes it difficult to cleanly separate this time-rotating "spiral noise" from other low-frequency eccentricities and random vibrations. Based on this, this invention constructs a one-dimensional time-series signal related to the conductor's edge pressure strength in the time domain and uses a fast Fourier transform to lock in characteristic frequencies linearly related to traction speed and stranding pitch, using these as the fundamental frequency of the spiral motion. Theoretical frequencies calculable from process parameters are used to screen and verify the spectral peaks, ensuring that the extracted frequency components indeed originate from the conductor's spiral motion rather than other electromechanical noise. Subsequently, using the helical phase variable, the cross-sectional profile data from multiple consecutive frames is unfolded into a continuous surface descriptor in a two-dimensional coordinate system of "azimuth-helical phase." Local Gaussian curvature is used to characterize the degree of deformation in different regions of this surface: the high-frequency fluctuations generated by the conductor edge sweeping are represented by narrow, high-curvature ridge regions, while the smooth, undisturbed melt profile is represented by low-curvature planar regions. By applying a statistical threshold to the curvature distribution, the ridges and planar regions are distinguished, and the corresponding physical measurement points are labeled as "helical noise components" or "smooth effective profile points," effectively performing a fine-grained physical semantic layering of the data in the time-frequency-phase joint space. This process comprehensively utilizes the theoretical helical frequency calculable in the extrusion process, the helical geometric motion laws, and the sensitivity of the surface differential geometry to local fluctuations. This allows the high-frequency helical artifacts introduced by the conductor polygon to be mathematically identified and explicitly labeled at the data level, providing a reliable "denoising" foundation for subsequent reconstruction of the equivalent conductor reference circle and eccentricity calculation.

[0041] A method for online control of cable insulation eccentricity further includes the following steps: Step S300: Based on the spiral noise components, reconstruct the virtual equivalent conductor reference circle and calculate the true eccentric displacement vector; Step S301: Use the spiral noise component to generate a spatial mask, perform reverse mask filtering on the discrete cross-section set, and retain the data points belonging to the planar region as the effective fitting support set; Based on the step The set of mesh nodes marked as helical noise components on the plane is used to construct a binary spatial mask matrix with the same dimension as the original discrete cross-section set time series flow. In this matrix, the elements corresponding to the spiral noise components (i.e., the curvature abrupt ridge regions) are set to 0, indicating occlusion / invalidation, while the elements corresponding to the curvature smooth plane regions are set to 1, indicating retention / validation. This spatial mask matrix... Reverse mapping back to each physical measurement point Inverse masking filtering, which performs logical AND operations: for any measurement point If the corresponding mask value is 0, the point is determined to be affected by the pressure of the conductor edge and is removed from subsequent fitting calculations; if the mask value is 1, the point is determined to belong to a smooth region of the insulation layer that is not disturbed and is retained. Finally, all the retained physical measurement points with high confidence are collected to form an effective fitting support set for reconstructing the virtual reference circle. This set fundamentally eliminates the interference of the spiral phantom on the calculation of the geometric center.

[0042] Step S302: The effective fitted support set is fitted with cylindricity using the weighted least squares method to reconstruct an ideal circular profile that eliminates the interference of polygonal edges, thereby obtaining the virtual equivalent conductor reference circle; Assuming the outer contour of the insulation layer should conform to the equation of an ideal cylindrical surface in the undisturbed region. ,in Let the center of the virtual equivalent conductor reference circle be the target. and These correspond to the actual observed components of the measurement point in a Cartesian coordinate system, specifically manifested as follows: and , Let the radius of the reference circle be used. Construct a weighted least squares objective function. ,in The weighting coefficients assign higher weights to points in the smooth central region far from the edge of the spiral noise, and lower weights to points at the edge, in order to further suppress the influence of residual noise. It is to effectively fit the support points polar diameter The x-coordinate observations obtained by projecting them onto the x-axis The physical meaning is the polar radius of that point. The observed ordinate values ​​are projected onto the Y-axis. The objective function is then iteratively solved using the Gauss-Newton method or the Levenberg-Marquardt algorithm. Minimize the set of parameters The coordinates of the circle's center obtained by solving the problem. and radius The determined ideal circular profile is the virtual equivalent conductor reference circle. This reference circle physically represents the theoretical concentric reference position of the insulation layer if the conductor is a perfect cylinder without any edge compression. Since the fitting data has had helical noise components removed in step S301 and weighted least squares has been used to suppress residual local errors, the obtained virtual equivalent conductor reference circle can statistically characterize the equivalent circular boundary of the real compacted conductor cross-section after ignoring polygonal edge details, providing a reliable geometric reference for accurate calculation of eccentricity.

[0043] Step S303: Based on the continuous fluid flow envelope, obtain the geometric center of the outer contour of the insulation layer using the area moment integral method; Calculate the Euclidean distance and azimuth angle between the geometric center of the outer contour of the insulation layer and the center of the virtual equivalent conductor reference circle, and output the true eccentric displacement vector.

[0044] First, calculate the geometric center of the outer contour of the insulation layer. The center is based on a reconstructed continuous fluid flow envelope. The closed curve is obtained by the method of area moment integration. The enclosed region is considered as a thin, planar plate (planar layer) with uniform mass distribution. The centroid of this planar region is calculated using numerical integration. Specifically, the envelope region can be divided into several tiny sector-shaped area elements along the circumference. The area and local centroid coordinates of each tiny sector-shaped area element are calculated, and then a weighted average of all area elements is taken to obtain the geometric center coordinates of the entire insulation layer cross-section. . This is the macroscopic fluid profile of the insulating layer after physical layering reconstruction. It eliminates high-frequency fluctuation interference caused by conductor edge pressure. Therefore, the centroid obtained based on this envelope can most accurately represent the macroscopic physical center of gravity of the insulating layer as a fluid medium at the mold exit. The position of this centroid is only affected by the melt flow distribution and is independent of the microscopic shape of the internal conductor, thus ensuring the physical purity of the eccentricity calculation reference. Calculate this geometric center. Center of the virtual equivalent conductor reference circle The eccentricity difference vector between Based on this vector, calculate the Euclidean distance (i.e., the modulus of eccentricity) that characterizes the degree of eccentricity. and the azimuth angle representing the direction of eccentricity. , This is the arctangent function. Ultimately, it will contain the modulus. and direction The vector data is packaged and output as the true eccentric displacement vector. This vector eliminates the false high-frequency components caused by the conductor polygon effect and accurately reflects the macroscopic physical offset between the extruder die and the conductor core, providing an accurate control target value for the core-aligning mechanism.

[0045] In step S300, because the actual compressed conductor has a polygonal or polygonal approximate shape in cross-section, its outer boundary has a geometric difference from the ideal circular conductor. If the polygonal outline is directly used as the reference for eccentricity calculation, the spiral undulations caused by the polygonal edges will inevitably be superimposed on the eccentricity result. This makes the so-called eccentricity include both the macroscopic offset of the mold-conductor and the irregularity of the conductor's own geometry, losing its clear physical meaning. To obtain a control target that can truly represent the offset between the mold center and the equivalent geometric center of the conductor, this invention, based on the aforementioned spiral noise identification results, first uses a spatial mask to remove all measurement points judged as spiral noise, retaining only those effective support points that are classified as smooth regions in time-phase-space, used to fit the position where the insulation layer should be when the conductor is assumed to be an ideal cylinder. This process is equivalent to constructing a virtual equivalent conductor reference circle supported only by data from the undisturbed region: by using weighted least squares to fit the cylindricity of the smooth point set and assigning higher weights to points far from the noise boundary, an equivalent circular contour that statistically reconstructs a shape that neither contains the high-frequency fluctuations of polygonal edges nor fails to faithfully reflect the overall position of the conductor can be recovered. Meanwhile, the geometric center of the outer contour of the insulation layer is not a simple average radius, but should be considered as the centroid of the uniform density section from a fluid perspective. Therefore, a centroid calculation method based on area moment integrals is used to numerically integrate the reconstructed continuous fluid envelope, which can more accurately obtain the overall geometric center position of the melt section. The center of the equivalent conductor reference circle is defined as the geometric reference center for eccentricity calculation, and the geometric center of the insulation layer is taken as the actual centroid of the insulation layer. The Euclidean distance and azimuth angle between the two directly constitute the true eccentricity displacement vector, thus clearly distinguishing the geometric effects of the conductor edges and the macroscopic offset of the mold-conductor at the result level, retaining only the latter as the core-aligning control quantity. The starting point of the entire step S300 is to reconstruct an equivalent circular reference that only reflects the overall position of the conductor by eliminating the spurious spiral components caused by polygonal geometry, and then use this reference to establish a geometric relationship with the macroscopic contour of the insulation layer, so as to realize the transformation of the eccentricity from a mixed quantity to a purely controlled object.

[0046] A method for online control of cable insulation eccentricity further includes the following steps: Step S400: The actual eccentric displacement vector is transferred through the core-aligning mechanism transfer function and combined with the melt rheological delay characteristics of the extruder head to generate dynamic compensation parameters; The process of generating dynamic compensation parameters as described in step S400 specifically includes the following steps: Step S401: Based on the real eccentric displacement vector, perform frequency domain analysis and statistical calculation on the eccentric modulus and azimuth sequence to calculate the dominant frequency, root mean square value of modulus, azimuth variation amplitude and azimuth standard deviation. The system comprehensively determines whether the system is in the spiral phantom oscillation range based on the dominant frequency, root mean square value of the modulus, azimuth angle variation amplitude, and azimuth angle standard deviation. If it is confirmed to be in the spiral phantom oscillation range, a zero command is output. To construct the dead-zone threshold logic, this embodiment collects eccentricity characteristic data through a set of baseline tests under controllable operating conditions during the equipment commissioning phase. Specifically, under the conditions that the conductor and die are strictly concentric, the conductor stranding pitch and traction speed remain stable, and the core-aligning mechanism is closed or frozen, the eccentric die length sequence is continuously recorded over a period of time. and azimuth sequence Simultaneously, spectral analysis was used to determine the amplitude statistics and dominant frequency range of the typical spiral phantom oscillation generated by the superposition of factors such as the polygonal edges of the conductor and mechanical vibrations of the production line under this operating condition. Based on the above baseline data, the steady-state mean of the eccentric modulus under this ideal concentric operating condition was statistically obtained. and standard deviation and the minimum threshold Set it to a value slightly higher than the upper limit of background noise, for example, take... To ensure that when the eccentric mold length is lower than At this time, it can be assumed that only measurement noise exists and there is no uncontrollable eccentricity; maximum threshold By simulating extreme operating conditions on the same equipment, such as intentionally introducing known eccentricity or observing transient impacts in historical data, the maximum reasonable value of the eccentricity modulus in the non-resonance state can be obtained. Multiply by the safety factor ,get This is used to limit the upper bound of the allowable modulus of the helical phantom oscillation. If this upper bound is exceeded, the oscillation will no longer be considered merely a phantom, but will be treated as an abnormally large eccentricity and trigger the protection logic; it also limits the bandwidth threshold for the frequency range. The method of obtaining the data is as follows: from the above concentric baseline test data and several typical production batch data, the eccentric modulus sequence obtained for each time period is... Perform a Fast Fourier Transform to obtain its amplitude-frequency response curve. On this curve, first search for frequency points in descending order of amplitude, retaining frequencies whose amplitude is greater than the average amplitude of the entire frequency band by several times (e.g., 3 times), and denoted as the candidate peak frequency set. ,in This is the sequence index number of the candidate peak frequency, used to distinguish different candidate frequency points; here, energy concentration is specifically manifested in the fact that the spectral amplitude of a certain frequency point is significantly higher than the surrounding neighborhood and the average level of the entire frequency band; based on the theoretical frequency of conductor rotation... Centered on the center, a preset relative deviation is taken on both sides (e.g. Forming a reference frequency band Candidate peak frequencies falling within this reference frequency band are classified as peaks related to conductor stranding rotation; several major mechanical natural frequencies are identified based on equipment technical data or experiments. ,in This is a sequence index number for the mechanical natural frequency, used to distinguish various natural vibration frequencies from different sources present in the equipment, such as motor rotation frequency and gear meshing frequency. Similarly, a preset relative deviation is taken on both sides of each natural frequency (e.g., A local reference band is formed, and candidate peak frequencies falling within these bands are classified as peaks related to mechanical vibration. These two types of frequency points are then uniformly clustered: frequencies spaced less than a preset merging threshold on the frequency axis are clustered together. Candidate frequencies (e.g., 0.1 Hz) are considered to be in the same group. The minimum and maximum frequencies within this group are designated as the lower and upper boundaries of the group, respectively, thus obtaining several non-overlapping frequency sub-intervals. ;in This refers to the sequence index number of the frequency sub-intervals obtained after clustering and merging; and These represent the lower and upper boundary frequencies of the first frequency sub-interval, respectively. and These represent the lower and upper boundary frequencies of the second frequency sub-interval, respectively. and Each refers to the first The lower and upper boundary frequencies of each frequency sub-interval are defined as the typical spiral phantom frequency bands observed in actual data. The minimum value among the lower boundaries of all sub-intervals is defined as... , To find the minimum value function, the maximum value in the upper boundary of all subintervals is defined as... , To find the maximum value function, the following is obtained: That is, the total frequency band covering all identified spiral phantom-related peaks, and because it consists only of... The algorithm is derived from the narrowband peak near the device's natural frequency, naturally excluding frequency components in the low-frequency region that correspond to real, slow eccentric changes. This ensures that the determination of the phantom oscillation frequency band in the dead-zone logic has a clear and reproducible physical and mathematical basis. (Preset angle threshold) The method for obtaining this data is as follows: From the aforementioned concentric baseline and typical operating condition data, extract multiple time windows with a length of [missing information]. Azimuth sequence The standard deviation of the azimuth angle within each time window is calculated to obtain a set of standard deviation samples. Comparative analysis shows that under the condition that the conductor and mold are basically concentric and no core adjustment is performed, these standard deviations mainly reflect the slow drift of the true eccentricity and measurement noise, and their values ​​are generally small. However, under the condition of deliberately creating a spiral illusion, such as increasing the traction speed, selecting a conductor with obvious polygonal edges and keeping it concentric without adjustment, the azimuth angle changes drastically due to false oscillations, and the calculated standard deviation value is significantly larger. Based on the statistical results of the above two working conditions, the typical upper limit of the azimuth angle standard deviation under the true slowly varying eccentricity working condition is selected as... This causes the standard deviation of the azimuth angle calculated within a certain time window during online operation to be greater than [a certain value]. When the azimuth angle has oscillated significantly within the time window, it can be determined that the azimuth angle has oscillated significantly, thus providing an effective angular domain support criterion for the identification of spiral phantom oscillation. A typical upper limit value is, for example, taking the larger value in the measured sample or the mean of the sample plus several times the standard deviation.

[0047] In actual operation, the controller operates at a fixed time step. Sample the eccentric vector, selecting a length of A sliding time window, for the eccentric modulus sequence within that time window. Perform a fast Fourier transform or bandpass filtering to estimate the dominant frequency. k is the discrete time series index; simultaneously, the root mean square value of the modulus within this time window is calculated. and azimuth standard deviation And define rapid back-and-forth oscillation as: the amplitude of azimuth change within the same time window. Exceeding the preset minimum swing amplitude, such as 30° or ,and After the above parameters are determined, the condition is met if and only if three conditions are met simultaneously: First, ;second, Third, within this time window and When the "rapid back-and-forth oscillation" judgment rule is met, the control system determines that the current real eccentric displacement vector is in the spiral phantom oscillation range. Within this control cycle, it outputs a zero displacement command to the self-aligning mechanism, that is, sets the target stroke increment of all self-aligning actuators to 0, and keeps the current position of the self-aligning mechanism unchanged.

[0048] Step S402: If the true eccentric displacement vector is determined to be a low-frequency effective signal, the true eccentric displacement vector is decomposed into lateral and longitudinal displacement components in the mold coordinate system, and the transfer function of the core-aligning mechanism is constructed through the stroke-eccentric coupling matrix and the proportional coefficient matrix. Based on the transfer function of the self-aligning mechanism, the lateral and longitudinal displacement components are converted into the mechanical stroke increment of each self-aligning actuator to generate the target value of mechanical stroke. Based on the determination that the current eccentric signal is not in the spiral phantom oscillation range and the eccentric modulus... Exceeding the minimum effective eccentricity threshold At that time, the true eccentric displacement vector This is considered a low-frequency effective signal, and the target mechanical stroke value of the self-aligning mechanism is calculated based on it. Among these, the minimum effective eccentricity threshold... The method for obtaining the data is as follows: During the equipment debugging phase, the conductor and the mold are made as concentric as possible without performing any core-aligning actions, and the eccentric mold length sequence is recorded over a period of time. Calculate the mean of the sequence. and standard deviation and will Set a value slightly higher than the upper limit for normal measurement noise and minor uncontrollable deviations, for example... ,in Choose 3 to 4, so that when running online... It can be assumed that the eccentricity is within a small disturbance range that cannot be effectively adjusted, and therefore does not trigger the core-aligning action.

[0049] To map the eccentric vector to the stroke of the self-aligning actuator, in the extrusion die coordinate system, the eccentric die length and azimuth angle are first decomposed into rectangular coordinate components: , ,in This represents the displacement of the center of the conductor's equivalent reference circle relative to the geometric center of the insulation layer along the X-axis at the mold exit section, with rightward displacement being positive. This represents the displacement along the Y-axis, with upward movement being positive. To establish a quantitative relationship between the actuator stroke change and the conductor center displacement, this embodiment obtains the stroke-eccentric coupling matrix through calibration experiments during the commissioning phase. : Sequentially drive each self-aligning actuator individually to apply several known axial displacement increments, for example , At each operating point, the displacement response of the conductor center in the X and Y directions is recorded by an online measurement system. The linear relationship between "actuator stroke and conductor displacement" is obtained by fitting using the least squares method. The displacement contribution coefficients of each actuator per unit stroke in the X and Y directions are arranged into a matrix, thus obtaining... . elements Indicates "the Each self-aligning actuator moves one unit stroke along its axial direction, at the center of the conductor. The displacement generated in each direction (such as the X or Y direction). The theoretical stroke increment of each actuator is calculated from the target eccentric displacement using the following formula: ; in, for The axial travel increment of each actuator For the first The axial travel increment of each actuator for The Moore-Penrose generalized inverse matrix. The axial travel increment vector is obtained. Subsequently, to avoid mechanical shock or overcompensation caused by excessively large single movements, a proportional coefficient matrix is ​​added to the transfer function. : The corrected mechanical stroke increment is obtained. . Selected as a diagonal matrix, each diagonal element During the equipment commissioning phase, step tests were conducted to fine-tune the settings: for the first... For each actuator, several candidate values ​​are selected sequentially, such as 1.0, 0.8, 0.6, etc. A fixed theoretical stroke increment is applied to each candidate value, and the eccentric modulus data before and after adjustment are continuously recorded. For each candidate value, the minimum value of the adjusted eccentric modulus is compared with the final stable value of the eccentric modulus at the end of the experiment. If the measured minimum value is less than 80% of the final stable value, i.e., the minimum value is lower than 80% of the final value, then there is a significant overshoot under that candidate value, and the candidate value is judged as too large. As the candidate values ​​gradually decrease, when a certain value appears, the minimum value of the corresponding eccentric modulus in the experiment is not lower than 80% of the final stable value, and the eccentric modulus does not show repeated alternating changes between being greater than and less than the same stable level throughout the recording time, then that value is determined as stable. The final set value. This is how it is obtained. This ensures sufficient correction amount in each adjustment while avoiding system oscillation caused by excessive movement. The corrected mechanical travel increment is then... By superimposing the actual stroke positions of each actuator at the current time, the target mechanical stroke value for the next control cycle is obtained. , represents the absolute position vector of the reference target during the stroke, and serves as the input for superimposing the melt rheological time constant constraint in the subsequent step S403.

[0050] Step S403: Based on the melt rheological time constant, the target value of the mechanical stroke is smoothed by using a discrete-time difference form, stroke amplitude and speed constraints are applied, and dynamic compensation parameters are generated.

[0051] During the trial operation phase of the production line, the equivalent response time constant of the extruder head-melt-cooling system to changes in mechanical stroke is obtained through step tests or disturbance tests. For example, under steady-state operating conditions, a fixed stroke step is applied to a certain self-aligning actuator, and the curve of the change in the eccentric die length of the outer contour over time is recorded. The time interval corresponding to the eccentricity change increasing from 0 to its stable change of approximately 63% is measured, and this time interval is recorded as the melt rheological time constant. Based on this, the aforementioned continuous-time characteristics are implemented using a discrete-time instruction smoothing method, assuming the control period is... Then define the smoothing coefficient. Within each control cycle, only travel commands are allowed to approach the target value. The proportion is used to limit the rate of change at the mechanical end. The control update law for each actuator's stroke command is calculated using the following differential form: ; in, Let w be the absolute position vector of the stroke command actually issued to each self-aligning actuator in the w-th control cycle, and let its i-th position be the absolute position vector of the stroke command actually issued to each self-aligning actuator in the w-th control cycle. Each component is denoted as , For the previous control cycle (the first) The absolute position vector of the travel command that has been issued (period), its first... Each component is denoted as , To the melt rheological time constant and control cycle The relevant smoothing coefficient, For the first Each control cycle aims to determine the absolute position vector of the travel reference target that each actuator should reach at the end of the cycle. Within each control cycle, motion commands are only allowed to gradually approach the same target value by a certain proportion, thus simulating the asymptotic characteristics of melt flow. Based on this, to prevent the actuators from exceeding their mechanical limits due to instantaneous large-amplitude commands, two levels of limits are set in the controller: motion amplitude limit and motion speed limit. Apply upper and lower boundary constraints to each component. This ensures that the travel command at any given time does not exceed the maximum travel range allowed by the actuator. and They represent the first The minimum and maximum permissible stroke positions of an actuator under mechanical structure and safety conditions; the speed limit refers to the amount of stroke variation between adjacent control cycles. Apply maximum speed threshold constraint ,in For the first The maximum safe movement speed of each actuator It is the change in travel. The absolute value; the stroke command sequence after time constant smoothing and satisfying both amplitude and rate constraints. The output is packaged as dynamic compensation parameters, which are sent to the drive unit of the self-aligning actuator in the form of a time series. Each cycle includes the target displacement increment to be executed by each actuator and the corresponding maximum allowable speed setting. This allows the self-aligning mechanism to gradually guide the equivalent reference circle of the conductor to the geometric center of the insulation layer under the premise of meeting the rheological characteristics of the melt and the mechanical safety boundary, so as to achieve stable and accurate online control of the eccentricity.

[0052] Step S400 involves the online measurement of the true eccentric displacement vector, which contains multiple time-scale components. These include high-frequency helical phantom oscillations caused by the superposition of conductor polygons and mechanical vibrations, as well as low-frequency true eccentricity changes caused by conductor trace offsets, mold thermal drift, and other factors. If these two types of components are not distinguished, and the full-band eccentricity is directly used as the control input for the core-aligning mechanism, the actuator will be frequently driven by the high-frequency components. This not only fails to effectively improve the true eccentricity but may also induce system oscillations due to frequent reverse actions, increasing mechanical wear. Based on this understanding, in step S401, this invention utilizes frequency domain analysis and statistical features to extract the dominant frequency, root mean square value, and angular fluctuation statistics from the eccentric modulus and azimuth sequence within a time window. These online operating data are then compared with threshold intervals pre-calibrated during the commissioning phase using concentric baseline conditions and typical production conditions. This allows the identification of operating conditions falling within a specific frequency band, amplitude range, and azimuth angle rapid reciprocating oscillation mode as "helical phantom oscillation intervals." Within this range, the control strategy selects to issue a zero-displacement command to the core-aligning mechanism, i.e., temporarily refrain from mechanical adjustment to avoid erroneous responses to false eccentricities purely caused by edge geometry and inherent vibrations. For low-frequency components that are not identified as phantom oscillations and whose eccentricity length exceeds the minimum effective eccentricity threshold, in step S402, the eccentricity vector is decomposed into the rectangular components of the die coordinate system. The theoretical stroke increment of each actuator is then derived from the stroke-eccentricity coupling matrix obtained during the debugging phase, and a proportional coefficient matrix tuned according to the step test is superimposed to ensure that the single adjustment amount has sufficient correction effect while avoiding overshoot. Considering that the extruder head-melt-cooling system has a significant time constant in response to stroke changes, and that the actuator itself has stroke and speed safety limits, step S403 further introduces these mechanical stroke target values ​​into a discrete-time difference form of command smoothing update law. Through the correspondence between the smoothing coefficient and the response time constant, the actual command gradually approaches the target value only by a certain proportion in each control cycle, while simultaneously applying dual constraints on stroke amplitude and speed, forming a set of dynamic compensation parameters that evolve over time. The purpose of this is twofold: firstly, to actively shield high-frequency phantom components using frequency domain and statistical criteria; and secondly, to introduce a time scale and safety boundary that matches the rheological and mechanical properties of the melt when correcting effective low-frequency eccentricity. This ensures that the movement of the core-aligning mechanism keeps up with eccentricity changes without inducing new dynamic instability, thereby achieving an effective mapping from "the physical eccentricity of the measurement space" to "the dynamic stroke command acceptable to the actuator".

[0053] Example 2 An online control device for the eccentricity of cable insulation layer, comprising the following: The acquisition and reconstruction module is used to acquire full-circumferential high-frequency contour point cloud data during the extrusion process of cable insulation layer, and to perform solid layer reconstruction preprocessing based on fluid-structure space exclusivity on the full-circumferential high-frequency contour point cloud data to obtain a set of discrete sections. The noise extraction module is used to perform time-frequency domain joint analysis on the discrete cross-section set to identify and extract the helical noise component; The eccentricity calculation module is used to reconstruct a virtual equivalent conductor reference circle based on the spiral noise component and calculate the real eccentricity displacement vector. The compensation and control module is used to generate dynamic compensation parameters by transferring the actual eccentric displacement vector through the core-aligning mechanism transfer function and combining it with the melt rheological delay characteristics of the extruder head.

[0054] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0055] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0056] The foregoing description is illustrative of the invention and should not be construed as limiting it. Although several exemplary embodiments of the invention have been described, those skilled in the art will readily understand that many modifications can be made to the exemplary embodiments without departing from the novel teachings and advantages of the invention. Therefore, all such modifications are intended to be included within the scope of the invention as defined in the claims. It should be understood that the foregoing description is illustrative of the invention and should not be construed as limiting it to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The invention is defined by the claims and their equivalents.

Claims

1. A method for online control of the eccentricity of cable insulation layer, characterized in that, Includes the following steps: Acquire full-circumferential high-frequency contour point cloud data during the cable insulation extrusion process, and perform solid layer reconstruction preprocessing based on fluid-structure space exclusivity on the full-circumferential high-frequency contour point cloud data to obtain a set of discrete sections; A joint time-frequency domain analysis was performed on the discrete cross-section set to identify and extract the helical noise component; Based on the spiral noise components, a virtual equivalent conductor reference circle is reconstructed, and the real eccentric displacement vector is calculated. The actual eccentric displacement vector is transferred through the core-aligning mechanism transfer function and combined with the melt rheological delay characteristics of the extruder head to generate dynamic compensation parameters; The analysis steps for the entity hierarchical reconstruction preprocessing based on fluid-structure space exclusivity are as follows: A three-dimensional cylindrical projection space is constructed using the mechanical centerline of the extruder die as the absolute physical reference. The high-frequency contour point cloud data of the entire circumference is mapped onto the cross section of the three-dimensional cylindrical projection space, and a virtual elastic contraction ring is initialized on the periphery of the cross section. Based on the melt rheological tension characteristics of insulating materials, a dynamic shrinkage annular geometric constraint boundary is set in the virtual elastic shrinkage ring. The equivalent tension function is calculated in combination with the head pressure. The radial shrinkage dynamic equation of each node is constructed using the pre-calibrated equivalent elastic modulus and damping coefficient. Based on the radial contraction dynamic equations of each node, the contact reaction force is obtained according to the contact relationship between the node and the projected point cloud during the numerical integration iteration process, and the stable boundary radius is obtained. The projected space physics is divided into an external steady-state fluid domain and an internal transient interference domain using this stable boundary. Spatial density clustering is performed on discrete data points falling within the internal transient interference domain to identify and instantiate them as solid interference clusters. The geometric centroid coordinates and radial intrusion vector of the entity interference cluster are calculated respectively to generate spatial pose attribute parameters; The data points retained within the external steady-state fluid domain are reconstructed into a continuous fluid flow envelope; The continuous fluid flow envelope and the spatial pose attribute parameters are encapsulated together to form the discrete cross-section set.

2. The method for online control of cable insulation layer eccentricity according to claim 1, characterized in that, The analysis steps for performing joint time-frequency domain analysis to identify and extract the spiral noise component are as follows: A fast Fourier transform is performed on the time series stream of the discrete cross-section set to lock the characteristic frequency that is linearly related to the conductor traction speed and stranding pitch, which is used as the fundamental frequency of the helical motion. A three-dimensional spiral phase space is constructed using the fundamental frequency of the spiral motion. The two-dimensional cross-sectional profile data is mapped along the time axis into a continuous surface descriptor, and the local Gaussian curvature distribution on the continuous surface descriptor is calculated. Based on the extreme value distribution of local Gaussian curvature, the region with abrupt curvature changes and the region with smooth curvature are segmented to obtain the spiral noise component.

3. The method for online control of cable insulation layer eccentricity according to claim 1, characterized in that, The analytical steps for calculating the true eccentric displacement vector are as follows: The spiral noise component is used to generate a spatial mask, and the discrete cross-section set is subjected to inverse mask filtering to retain data points belonging to the planar region as the effective fitting support set. The effective fitted support set is fitted with cylindricity using the weighted least squares method to reconstruct an ideal circular profile that eliminates the interference of polygonal edges, thus obtaining the virtual equivalent conductor reference circle.

4. The method for online control of cable insulation layer eccentricity according to claim 3, characterized in that, The analytical steps for calculating the true eccentric displacement vector also include: The geometric center of the outer contour of the insulation layer is obtained by using the area moment integral method based on the continuous fluid flow envelope. Calculate the Euclidean distance and azimuth angle between the geometric center of the outer contour of the insulation layer and the center of the virtual equivalent conductor reference circle, and output the true eccentric displacement vector.

5. The method for online control of cable insulation layer eccentricity according to claim 1, characterized in that, The analysis steps for generating dynamic compensation parameters are as follows: Based on the real eccentric displacement vector, frequency domain analysis and statistical calculations are performed on the eccentric modulus and azimuth sequence to calculate the dominant frequency, root mean square value of modulus, azimuth variation amplitude and azimuth standard deviation. The system comprehensively determines whether the oscillation is within the spiral phantom oscillation range based on the dominant frequency, root mean square value of the modulus, azimuth variation amplitude, and azimuth standard deviation. If it is confirmed to be within the spiral phantom oscillation range, a zero command is output.

6. The method for online control of cable insulation layer eccentricity according to claim 5, characterized in that, The analysis step for generating dynamic compensation parameters also includes: If the true eccentric displacement vector is determined to be a low-frequency effective signal, the true eccentric displacement vector is decomposed into lateral and longitudinal displacement components in the mold coordinate system, and the transfer function of the core-aligning mechanism is constructed through the stroke-eccentric coupling matrix and the proportional coefficient matrix. Based on the transfer function of the self-aligning mechanism, the lateral and longitudinal displacement components are converted into the mechanical stroke increment of each self-aligning actuator to generate the target value of mechanical stroke. Based on the melt rheological time constant, the target value of mechanical stroke is smoothed by using a discrete-time differential form, and stroke amplitude and speed constraints are applied to generate dynamic compensation parameters.

7. An online control device for the eccentricity of cable insulation layer, characterized in that, An online method for controlling the eccentricity of a cable insulation layer according to any one of claims 1-6 includes: The acquisition and reconstruction module is used to acquire full-circumferential high-frequency contour point cloud data during the extrusion process of cable insulation layer, and to perform solid layer reconstruction preprocessing based on fluid-structure space exclusivity on the full-circumferential high-frequency contour point cloud data to obtain a set of discrete sections. The noise extraction module is used to perform joint time-frequency domain analysis on the discrete cross-section set to identify and extract the helical noise component; The eccentricity calculation module is used to reconstruct a virtual equivalent conductor reference circle based on the spiral noise component and calculate the real eccentricity displacement vector. The compensation and control module is used to generate dynamic compensation parameters by transferring the actual eccentric displacement vector through the core-aligning mechanism transfer function and combining it with the melt rheological delay characteristics of the extruder head.

Images