An OPGW optical cable production whole-process data acquisition and quality traceability method, device, equipment and medium based on an industrial Internet of Things

Abstract

The application discloses an OPGW optical cable production whole-process data acquisition and quality tracing method, device and equipment based on an industrial Internet of Things, and relates to the technical field of intelligent manufacturing. The method comprises the following steps: constructing a multi-physical field digital twin body covering the whole process of OPGW optical cable production; deploying a multi-modal sensing device in a key process in the whole process of production, and acquiring multi-modal sensing data at least comprising optical fiber excess length distribution data obtained based on the acoustic emission principle; based on the multi-physical field digital twin body and the multi-modal sensing data, performing quality tracing based on process interval comparative learning to identify the root cause of quality abnormalities. The application realizes accurate quality tracing and root cause diagnosis for the whole process of OPGW optical cable production, and improves the quality abnormality troubleshooting efficiency and process closed-loop control capability.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, and more specifically, to a method, apparatus, equipment, and medium for data acquisition and quality traceability throughout the entire production process of OPGW optical cables based on the Industrial Internet of Things. Background Technology

[0002] OPGW (Optical Fiber Composite Overhead Ground Wire) optical cable is a core infrastructure of power communication systems. Its production process involves multiple steps, including aluminum-clad steel wire drawing, stainless steel strip forming, fiber insertion into tubes, grease filling, laser welding, stranding into cables, sheath extrusion, and finished product testing. It is characterized by a long process chain, strong multi-physical field coupling, and complex quality-influencing factors. With the increasing demands for communication reliability in power systems, the quality traceability requirements for OPGW optical cables have evolved from simple pass / fail judgments to refined analysis of the entire process data and precise location of the root causes of anomalies. The rapid development of Industrial Internet of Things (IIoT) technology provides a technological foundation for deploying multimodal sensing devices on manufacturing sites and collecting massive amounts of process parameters in real time. However, how to effectively integrate dispersed, multi-source, heterogeneous data for quality traceability remains a core challenge for the industry.

[0003] Existing OPGW optical cable production quality traceability technologies suffer from the following prominent problems: First, data from different processes are fragmented, lacking a unified multi-physics digital twin model to characterize the causal mapping relationship between material properties, process history, testing characteristics, and finished product performance. Second, traditional testing methods are mostly offline sampling or destructive testing, unable to achieve online, full-length testing of key indicators such as fiber excess length distribution, residual stress in aluminum-clad steel wires, laser welding penetration depth, and interface bonding quality, resulting in a significant lag in the detection of quality anomalies. Third, the granularity of quality traceability is coarse, usually based on batch or reel number, unable to pinpoint specific length ranges, let alone accurately identify which process or parameter caused the non-conformity. Therefore, there is an urgent need for a quality traceability method that can integrate multi-modal sensing data throughout the entire process, construct a digital twin, and achieve intelligent root cause diagnosis at the process interval level. Summary of the Invention

[0004] To overcome the problems of data fragmentation, detection lag, coarse traceability granularity, and inability to accurately locate the root cause of abnormal processes and parameters in existing technologies, this invention discloses a method, device, equipment, and medium for data acquisition and quality traceability of the entire OPGW optical cable production process based on the Industrial Internet of Things, which can effectively solve the above-mentioned technical problems.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: A method for data acquisition and quality traceability throughout the entire OPGW optical cable production process based on the Industrial Internet of Things (IIoT), the method comprising: Construct a multi-physics digital twin covering the entire process of OPGW optical cable production; Multimodal sensing devices are deployed in each step of the entire production process to collect multimodal sensing data; the multimodal sensing data includes at least: fiber excess length distribution data of optical fibers in stainless steel tubes obtained based on the principle of acoustic emission; Based on the multiphysics digital twin and the multimodal sensing data, quality traceability based on process interval comparison learning is performed to identify the root cause of quality anomalies.

[0006] Furthermore, the fiber excess length distribution data is obtained through the following steps: After the fiber optic cable insertion process and before the grease filling, a distributed acoustic emission sensor array is coupled and installed at both ends of the stainless steel tube. A standard pulse signal is applied into the optical fiber to generate a vibration wave inside the tube; The distributed acoustic emission sensor array receives acoustic emission signals propagating along the optical fiber and the stainless steel pipe wall. The time delay difference and waveform characteristics of the acoustic emission signal are analyzed, and the time difference of arrival positioning algorithm is used to invert the distribution data of the excess fiber length per meter along the length direction of the optical fiber in the stainless steel tube.

[0007] Furthermore, the multimodal sensing data also includes: Data on the circumferential and axial residual stress distribution of aluminum-clad steel wire collected by a laser ultrasonic testing system; Estimated weld depth and measured weld width of stainless steel pipe laser welding process acquired by coaxial and lateral vision fusion detection device; The interface bonding quality assessment data of the stainless steel pipe and aluminum-clad steel wire interface is acquired by a ring phased array ultrasonic probe array. The interface bonding quality assessment data includes the interface reflection coefficient, backscattering intensity and interface spacing.

[0008] Furthermore, the multiphysics digital twin is constructed in the following manner: Establish a twin layer for material properties, a twin layer for process technology, a twin layer for quality inspection, a physical field coupling layer, and a twin layer for performance prediction. In the physical field coupling layer, a multi-physics coupling model including temperature field, stress field and deformation field is established based on the finite element method, and a reduced-order model with intrinsic orthogonal decomposition and radial basis function interpolation is used to accelerate the calculation. The multimodal sensing data is used to perform near real-time calibration, test verification calibration, and long-term learning updates on the multiphysics coupling model.

[0009] Furthermore, the quality traceability based on process interval comparison learning specifically includes: The entire production process of the OPGW optical cable to be traced is divided into multiple continuous process intervals according to the length coordinate. Each process interval is associated with all process parameters and quality characteristic data within that interval. Obtain a pre-constructed normal process interval template corresponding to the model of the optical cable to be traced. The normal process interval template is generated based on the data of qualified optical cables through cluster analysis and statistical process control. The data of each process interval of the optical cable to be traced are compared with the normal process interval template to calculate the parameter value deviation, parameter fluctuation mode deviation and parameter correlation deviation. Process intervals where at least one deviation exceeds a preset threshold are marked as suspected root cause intervals; Based on the suspected root cause interval, and combined with the data from the preceding and subsequent processes, the root cause process and root cause parameters that caused the quality abnormality are inferred through a structural causal model.

[0010] Furthermore, the method also includes: In response to the deduction of the root cause process and root cause parameters, a process parameter correction instruction is generated and output to the process closed-loop control layer; The process closed-loop control layer automatically adjusts the process parameter settings for the corresponding process according to the correction instructions, forming a quality-driven adaptive process closed loop.

[0011] Furthermore, the entire production process includes at least: aluminum-clad steel wire drawing, stainless steel strip forming, optical fiber insertion into tubes, grease filling, laser welding, stranding into cables, sheath extrusion, and finished product testing; the key processes include the wire drawing process, the laser welding process, the stranding into cables process, and the optical fiber insertion into tubes process.

[0012] Furthermore, a data acquisition and quality traceability device for the entire OPGW optical cable production process based on the Industrial Internet of Things includes: The digital twin building block is used to construct a multi-physics digital twin covering the entire process of OPGW optical cable production; A multimodal data acquisition module is used to control multimodal sensing devices deployed in key processes and acquire multimodal sensing data; the multimodal sensing data includes at least fiber optic excess length distribution data; The traceability analysis module is used to perform quality traceability based on process interval comparison learning based on the multi-physics digital twin and the multimodal sensing data, so as to identify the root cause of quality anomalies.

[0013] Furthermore, a computer device includes: a memory and a processor, wherein the memory stores a computer program; When the computer program is executed by the processor, it implements the above-described method for data acquisition and quality traceability of the entire OPGW optical cable production process based on the Industrial Internet of Things.

[0014] Furthermore, a computer-readable storage medium stores a computer program that, when executed by a processor, implements the above-described method for data acquisition and quality traceability throughout the entire OPGW optical cable production process based on the Industrial Internet of Things.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention constructs a multi-physics digital twin covering the entire process, integrating five-layer models of materials, processes, detection, physical fields, and performance prediction, establishing a full-chain causal mapping, and using a reduced-order model to compress computation time from hours to minutes. Multi-modal sensing data enables multi-level model calibration, improving real-time performance and accuracy. Multi-modal sensing devices are deployed in key processes to acquire meter-level resolution fiber excess length distribution data based on acoustic emission principles. Combined with laser ultrasonic residual stress detection, visual weld depth and width detection, and phased array ultrasonic interface detection, full-length, non-destructive online detection is achieved. Quality traceability based on process interval comparison learning is implemented, dividing the entire production process into meter-level process intervals according to length coordinates. By comparing parameter values, fluctuation patterns, and correlation deviations with normal process interval templates, suspected root cause intervals are marked. A structural causal model is used to accurately infer the root cause process and parameters, increasing the traceability granularity from batch / reel number by several orders of magnitude. In response to the root cause inference results, process parameter correction instructions are generated and corresponding process settings are automatically adjusted, forming a quality-driven adaptive closed loop, achieving a leap from passive traceability to proactive prevention. In summary, this invention effectively solves the problems of data fragmentation, detection lag, and coarse traceability granularity, improving the efficiency of quality anomaly detection and the accuracy of root cause diagnosis. Attached Figure Description

[0016] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other embodiments can be derived from the provided drawings without creative effort.

[0017] Figure 1 A flowchart of a method for data acquisition and quality traceability of the entire OPGW optical cable production process based on the Industrial Internet of Things provided in this application embodiment; Figure 2 This is a structural diagram of a data acquisition and quality traceability device for the entire production process of OPGW optical cable based on the Industrial Internet of Things, provided in an embodiment of this application. Detailed Implementation

[0018] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent. To better illustrate this embodiment, some parts in the accompanying drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions; It will be understood by those skilled in the art that certain well-known structures and their descriptions may be omitted in the accompanying drawings.

[0019] The embodiments of this application will be further described in detail below with reference to the accompanying drawings and examples.

[0020] It is understood that the specific embodiments described herein are merely illustrative of the embodiments of this application and are not intended to limit the embodiments of this application. Furthermore, it should be noted that, for ease of description, the accompanying drawings only show the parts related to the embodiments of this application, not all structures. Those skilled in the art, after reading this specification, should be able to realize that any combination of technical features can constitute an optional implementation method, provided that the technical features do not contradict each other.

[0021] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship. In the description of this application, "multiple" means two or more, and "several" means one or more.

[0022] Please see Figure 1 This application provides a method for data acquisition and quality traceability throughout the entire OPGW optical cable production process based on the Industrial Internet of Things (IIoT). This method is applied to the manufacturing process of OPGW optical cables. It establishes a multi-physics digital twin of all production processes and deploys multi-modal sensing devices in key processes for data acquisition. Based on process interval comparative learning, it achieves accurate location of quality anomalies and root cause tracing.

[0023] The entire production process of OPGW optical cable includes at least the following steps: aluminum-clad steel wire drawing, stainless steel strip forming, fiber insertion, grease filling, laser welding, stranding into cable, sheath extrusion, and finished product testing. Specifically, the aluminum-clad steel wire drawing process involves drawing aluminum-clad steel wire into a specified diameter using multiple drawing dies; the stainless steel strip forming process involves forming a stainless steel tube with a specific cross-sectional shape using rollers; the fiber insertion process involves inserting a bundle of optical fibers into the stainless steel tube; the grease filling process involves injecting water-blocking grease into the stainless steel tube; the laser welding process involves longitudinally butt-jointing the formed stainless steel strips and forming a sealed stainless steel tube using laser welding; the stranding into cable process involves stranding multiple stainless steel tube optical units with aluminum-clad steel wire according to a specific stranding structure to form the cable core; the sheath extrusion process involves extruding polyethylene or low-smoke halogen-free sheath material onto the outside of the cable core; and the finished product testing process involves conducting mechanical and optical performance tests on the finished optical cable. Key processes include wire drawing, laser welding, stranding into cables, and fiber optic cable insertion. Multimodal sensing devices are deployed in these processes to collect multimodal sensing data that reflects the process status and quality characteristics.

[0024] The construction process of a multiphysics digital twin includes establishing a material performance field twin layer, a process field twin layer, a quality inspection field twin layer, a physical field coupling layer, and a performance prediction field twin layer. The material performance field twin layer stores and manages a database of physical performance parameters for all materials involved in OPGW optical cable production. The database uses the material batch number as the primary key, and the fields include elastic modulus, Poisson's ratio, yield strength, tensile strength, coefficient of thermal expansion, thermal conductivity, specific heat capacity, density, and resistivity. Specifically, the material parameters for aluminum-clad steel wire include the individual parameters of the steel wire matrix and the aluminum cladding, as well as composite equivalent parameters; the material parameters for stainless steel strip include the elastic modulus, Poisson's ratio, coefficient of thermal expansion, and high-temperature yield strength of austenitic stainless steel; the material parameters for optical fiber include the elastic modulus, coefficient of thermal expansion, and density of the quartz glass core and coating; the material parameters for grease include dynamic viscosity, density, and water-blocking performance indicators; and the material parameters for sheathing materials include the elastic modulus, coefficient of thermal expansion, melt flow index, and density of polyethylene or low-smoke halogen-free materials.

[0025] The process field twin layer maps real-time process parameters for each step in the entire production process, including: drawing die diameter, drawing speed, drawing temperature, and lubricant flow rate for aluminum-clad steel wire drawing; rolling pressure, forming roll gap, forming speed, and strip tension for stainless steel strip forming; fiber tension, fiber pitch, and stainless steel tube straightness for fiber optic tube insertion; grease filling pressure, grease temperature, and filling speed for grease filling; laser welding power, welding speed, shielding gas flow rate, and defocusing amount for laser welding; stranding pitch, stranding tension, stranding direction, and cabling die diameter for stranding cabling; and extrusion temperature, extrusion speed, cooling water temperature, and traction speed for sheath extrusion. Each process parameter is collected in real-time at a preset sampling frequency by sensors and programmable logic controllers in the Industrial Internet of Things (IIoT), and stored in the process field twin layer with timestamps and length coordinate tags.

[0026] The twin layer of the quality inspection field receives and stores multimodal sensing data collected by the multimodal sensing device. The data storage structure adopts a time-series database format, and each data record includes a collection timestamp, length coordinates, process identifier, data type identifier, and sensing data value. The multimodal sensing data includes at least fiber excess length distribution data, as well as circumferential and axial residual stress distribution data of aluminum-clad steel wire, estimated weld depth and measured weld width values ​​of the laser welding process of stainless steel pipe, interface bonding quality assessment data of the interface between stainless steel pipe and aluminum-clad steel wire, and tensile strength test values, Young's modulus test values, attenuation coefficient test values, and dispersion coefficient test values ​​of the finished product testing process.

[0027] The physics coupling layer establishes a multiphysics coupling model encompassing temperature, stress, and deformation fields based on the finite element method. The temperature field governing equations employ unsteady-state heat conduction equations, the stress field governing equations employ elastoplastic equilibrium equations, and the deformation field governing equations employ geometrically nonlinear equations. These three fields are solved simultaneously using thermo-mechanical coupling boundary conditions. Finite element discretization utilizes eight-node hexahedral isoparametric elements, with local mesh refinement strategies applied to critical regions. To reduce online computational complexity, a reduced-order model employing eigenorthogonal decomposition and radial basis function interpolation is used for computational acceleration. The process of constructing the reduced-order model is as follows: First, representative sample points are selected in the process parameter design space. A high-fidelity finite element model is run at each sample point to obtain the snapshot matrix of the system response. The snapshot matrix is ​​then subjected to intrinsic orthogonal decomposition to calculate the covariance matrix and solve the eigenvalue problem. The eigenvectors corresponding to the top k dominant eigenvalues ​​whose cumulative energy percentage reaches a preset threshold are selected as POD basis functions. The original high-dimensional finite element system is projected onto a low-dimensional subspace spanned by the POD basis functions using Galerkin projection, resulting in a reduced-order ordinary differential equation system. For new process parameter points, the corresponding POD modal coefficients are predicted using radial basis function interpolation: the modal coefficients calculated at the sample points are used as training data to construct a radial basis function interpolation matrix. The kernel function adopts a multivariate quadratic function form. The interpolation weights are obtained by solving a system of linear equations, thereby achieving response prediction for new process parameter points.

[0028] The multiphysics digital twin utilizes multimodal sensing data to perform near-real-time calibration, test verification calibration, and long-term learning updates on the physics coupling layer. Near-real-time calibration is triggered when multimodal sensing data for key processes is available. A recursive least squares algorithm is used to correct the modal coefficients of the reduced-order model online. The correction cycle is synchronized with the production cycle. Each correction uses multimodal sensing data from the current process interval as observations to update the estimated values ​​of the modal coefficients. Test verification calibration is performed after the finished product testing process. The measured mechanical and optical performance test values ​​from the finished product testing process are compared with the predicted values ​​from the performance prediction twin layer. Residual calculations are performed. If the absolute value of the residual exceeds the preset allowable error range, a reverse correction of the material performance parameters and boundary condition parameters in the physics coupling layer is triggered. Sensitivity analysis is used to determine the parameters to be corrected and their correction amounts. Long-term learning updates accumulate historical production data and periodically execute batch learning algorithms to update the spatial distribution density of the radial basis function interpolation nodes and the shape parameters of the kernel function to adapt to material batch changes and process drift.

[0029] In the fiber optic tube insertion process, multimodal sensing data includes at least fiber excess length distribution data. This data is acquired through the following steps: After the fiber optic tube insertion process and before the grease filling process, a distributed acoustic emission sensor array is coupled and installed at both ends of the stainless steel tube. The distributed acoustic emission sensor array consists of piezoelectric acoustic emission sensors. The number of sensors is determined based on the stainless steel tube diameter and the required circumferential coverage area. These sensors are evenly arranged circumferentially on the end faces or near-end outer walls of the stainless steel tube, with equal central angles between adjacent sensors. Acoustic impedance matching is achieved between the sensors and the stainless steel tube wall using an acoustic coupling agent, or by fixing them with a magnetic adsorption device and applying constant contact pressure. Each sensor is connected to a multi-channel data acquisition card via a preamplifier and a bandpass filter. The sampling rate of the data acquisition card is set to an equivalent sampling frequency corresponding to at least ten sampling points per meter of length, with a bandwidth covering the main frequency range of Lamb waves in the stainless steel tube and guided waves in the optical fiber.

[0030] A standard pulse signal, generated by a pulse generator, is applied into the optical fiber and injected into the fiber end via an optical fiber coupler. The pulse width of the standard pulse signal is set according to the fiber length and the required resolution. The pulse amplitude is below the fiber nonlinearity threshold, and the repetition frequency is matched to the moving speed of the stainless steel tube to ensure continuous coverage along the length direction. As the standard pulse signal propagates in the optical fiber, it couples with the micro-contact and oil film between the fiber and the inner wall of the stainless steel tube, generating vibration waves inside the tube. These vibration waves propagate along the fiber axis and the stainless steel tube wall, forming an acoustic emission signal containing both longitudinal and transverse wave components.

[0031] Acoustic emission signals propagating along optical fibers and stainless steel pipe walls are received using a distributed acoustic emission sensor array. Each channel synchronously records the waveform data of the acoustic emission signal, with the recording duration covering the complete propagation time of a standard pulse signal from the transmitter to the receiver. The time delay difference and waveform characteristics of the acoustic emission signal are analyzed: the time delay difference is obtained by calculating the cross-correlation function of waveform data from adjacent sensor channels and locating the delay time corresponding to the peak value of the cross-correlation function; waveform characteristics are extracted through time-domain analysis (peak amplitude, root mean square value, rise time, and duration) and through frequency-domain analysis (power spectral density, center frequency, and bandwidth).

[0032] The time-of-arrival (TOA) positioning algorithm is used to invert the fiber optic excess length distribution data. A one-dimensional positioning model is established with the axis of the stainless steel pipe as the coordinate axis. The length of the stainless steel pipe is L, and the axial coordinates of the sensor arrays at both ends are x=0 and x=L, respectively. The circumferential position of each sensor within the array does not affect the axial positioning. For the vibration wave source distributed axially inside the stainless steel pipe, its axial coordinate is... The standard pulse signal is transmitted at the following time: The equivalent propagation speed of the vibration wave in the fiber-stainless steel tube composite structure is v. For the i-th sensor, the arrival time of its received signal is... Satisfy the equation ,in Let be the axial coordinate of the i-th sensor. An overdetermined system of equations formed by multiple sensors is solved using the least squares method. and This minimizes the sum of squared residuals. Combining the transmission time of the standard pulse signal and the theoretical length of the stainless steel tube, the fiber excess length distribution data with a resolution of one meter along the length direction within the stainless steel tube is obtained through inversion. Specifically, the entire length of the stainless steel tube is divided into sub-intervals of one meter in length. Within each sub-interval, a time difference of arrival (TDOA) positioning algorithm is applied using a moving window method to obtain the average source position deviation within that sub-interval. This deviation is compared with the theoretical straight-line position of the fiber to calculate the fiber excess length value within that meter-length interval. This value is recorded in millimeters per meter, forming the fiber excess length distribution data along the entire length of the stainless steel tube.

[0033] In the wire drawing process, a laser ultrasonic testing system is used to collect data on the circumferential and axial residual stress distribution of the aluminum-clad steel wire. The laser ultrasonic testing system includes a pulsed laser excitation source and an interferometric laser detector receiver. The pulsed laser excitation source uses a Q-switched pulsed laser or a nanosecond pulsed laser with a wavelength in the near-infrared band and pulse energy on the order of millijoules. A micrometer-scale spot is formed on the surface of the aluminum-clad steel wire through a focusing lens, and ultrasonic waves are excited on the surface of the aluminum-clad steel wire using the thermoelastic effect. The interferometric laser detector receiver uses a Michelson interferometer or a Fabry-Perot interferometer structure. The wavelength of the detection laser is different from that of the excitation laser to avoid interference. By measuring the Doppler frequency shift or optical path difference caused by vibration on the surface of the aluminum-clad steel wire, the arrival time of the ultrasonic waves at a specific location on the surface of the aluminum-clad steel wire is obtained. In the circumferential direction of the aluminum-clad steel wire, multiple detector receivers arranged in a ring acquire ultrasonic signals at preset angular intervals at different azimuth angles. In the axial direction of the aluminum-clad steel wire, ultrasonic signals at different axial positions are acquired through a linear scanning mechanism or an array detector at preset intervals. By analyzing the changes in the propagation speed of ultrasound in aluminum-clad steel wire and the effect of acoustic birefringence, and combining the elastic constant tensor and thermoelastic coefficient of aluminum-clad steel wire, a quantitative relationship between stress and sound velocity change is established based on the acoustoelastic theory. The circumferential residual stress value and axial residual stress value on the surface of aluminum-clad steel wire are calculated, and residual stress distribution data are formed along the length of aluminum-clad steel wire at preset intervals.

[0034] In the laser welding process, a coaxial and lateral vision fusion inspection device is used to acquire estimated weld depth and measured weld width values ​​for stainless steel pipe laser welding. The coaxial vision inspection device uses a beam splitter on the optical axis of the laser welding head to align the optical axis of the industrial camera with that of the laser welding head. The resolution of the industrial camera is determined based on the weld pool size and required measurement accuracy, and the frame rate is matched to the welding speed to capture dynamic changes in the weld pool. The beam splitter has a higher reflectivity for laser wavelengths than for visible light, allowing the industrial camera to acquire an image of the weld pool on the front side of the weld through the beam splitter, while avoiding damage to the camera sensor from the welding laser. The lateral vision inspection device uses a high-speed camera positioned to the side of the laser welding head to acquire an image of the weld pool contour on the side at an angle. The optical axis of the high-speed camera forms a preset angle with the welding direction. The intrinsic and extrinsic parameter matrices of the high-speed camera are pre-calibrated using a calibration plate, establishing a mapping relationship between the image coordinate system and the world coordinate system.

[0035] Spatial registration of coaxial and lateral molten pool images is achieved using an image fusion algorithm. First, the coaxial and lateral molten pool images are preprocessed separately, including grayscale conversion, noise filtering, and contrast enhancement. Then, molten pool edge feature points are extracted from both images, and feature descriptors are calculated using a scale-invariant feature transformation algorithm or an accelerated robust feature transformation algorithm. The homography matrix or affine transformation matrix between the two images is calculated using nearest neighbor matching and random sample consensus algorithms to achieve pixel-level spatial registration. In the fused image, molten pool geometric feature parameters are extracted, including molten pool length, width, area, trailing angle, and center position. Based on a pre-established mapping model between molten pool geometric feature parameters and molten depth and width (obtained by fitting calibrated experimental data, in the form of a polynomial function or neural network mapping), the extracted molten pool geometric feature parameters are input into the mapping model to calculate the estimated molten depth and measured molten width.

[0036] In the stranding process, a ring phased array ultrasonic probe array is used to collect interfacial bonding quality assessment data between the stainless steel tube and the aluminum-clad steel wire. The ring phased array ultrasonic probe array consists of multiple ultrasonic transducers arranged circumferentially at the exit of the stranding mold. The transducer frequency is selected based on the stainless steel tube wall thickness and the aluminum-clad steel wire diameter, while the crystal size is determined based on the required beamwidth and near-field length. Each transducer is connected to an ultrasonic transmitter and receiver via a multiplexer. The transmitter pulse voltage, pulse width, and damping coefficient of the ultrasonic transmitter and receiver are set according to the material acoustic impedance matching conditions.

[0037] By controlling the transmission and reception delays of each transducer using a phased array focusing method, dynamic focusing of the ultrasonic beam is achieved in the interface region between the stainless steel pipe and the aluminum-clad steel wire. The transmission focusing delay is calculated based on the Fresnel approximation, ensuring that the ultrasonic waves emitted by each transducer are superimposed in phase at the target focal point. The reception focusing delay is calculated based on the acoustic path difference from the target focal point to each transducer, ensuring that the echo signals received by each transducer are superimposed in phase during beamforming. Electronic scanning allows the focused acoustic beam to move circumferentially and axially along the interface, covering the entire interface detection area.

[0038] After receiving the interface reflected echo signal, the interface reflection coefficient is extracted through spectral analysis. The interface reflection coefficient is defined as the ratio of the amplitude of the interface reflected echo signal to the amplitude of the echo signal from a total internal reflection reference under the same conditions. The backscattered signals received by each transducer are processed using a total focusing imaging algorithm to extract the backscattering intensity of the interface region. The backscattering intensity reflects the scattering characteristics of the interface's micro-roughness and local debonding defects. The interface spacing is calculated by measuring the flight time between the transmitted pulse and the interface reflected echo signal, combined with the sound velocity of ultrasound in stainless steel pipes and aluminum-clad steel wires. The interface bonding quality assessment data includes the interface reflection coefficient, backscattering intensity, and interface spacing.

[0039] Based on multi-physics digital twins and multimodal sensing data, quality traceability based on process interval comparative learning is implemented. The entire production process of the OPGW optical cable to be traced is divided into multiple continuous process intervals according to the length coordinate. The origin of the length coordinate is set at the entrance end of the aluminum-clad steel wire drawing process, increasing along the production flow direction. The length of each process interval is set according to the granularity requirements of quality traceability, and each process interval is associated with all process parameters and quality characteristic data within that interval. The process parameters include the wire drawing die aperture, wire drawing speed, wire drawing temperature, and lubricant flow rate of the wire drawing process; the rolling pressure, forming roller gap, forming speed, and strip tension of the stainless steel strip forming process; the fiber laying tension, fiber laying pitch, and stainless steel tube straightness of the fiber insertion process; the grease filling pressure and grease temperature of the grease filling process; the laser welding power, welding speed, shielding gas flow rate, and defocusing amount of the laser welding process; the stranding pitch, stranding tension, stranding direction, and cabling die aperture of the stranding process; and the extrusion temperature, extrusion speed, and cooling water temperature of the sheath extrusion process. The quality characteristic data include the circumferential residual stress distribution data and axial residual stress distribution data of aluminum-clad steel wire collected during the wire drawing process within this length range, the estimated melt depth and measured melt width collected during the laser welding process, the fiber excess length distribution data collected during the fiber insertion process, and the interface reflection coefficient, backscattering intensity, and interface spacing collected during the stranding process.

[0040] Obtain a pre-constructed normal process interval template corresponding to the model of the optical cable to be traced. The normal process interval template is generated based on the historical production data of qualified optical cables through cluster analysis and statistical process control. Collect historical multimodal sensing data and historical process parameters of qualified optical cables of the same model within a preset historical time window, and divide them into standard-length process intervals according to the same length coordinate division rules as the optical cable to be traced. Perform feature standardization on the process parameters and quality characteristic data in each process interval, and use the zero-mean standardization method to convert each feature parameter into a standardized feature vector with a mean of zero and a standard deviation of one. Use a cluster analysis algorithm to cluster the standardized feature vectors. The clustering algorithm uses K-means clustering algorithm or Gaussian mixture model clustering algorithm. The number of clusters is determined by elbow rule or silhouette coefficient method, dividing the historical process interval into multiple typical process state clusters. For each typical process state cluster, the mean vector and covariance matrix of each process parameter and quality characteristic data are calculated. Control limits for each characteristic parameter are established based on statistical process control methods. These control limits include a center line, upper control limit, and lower control limit. The upper and lower control limits are determined based on the mean plus or minus three standard deviations or percentiles estimated using nonparametric kernel density. The normal process interval template consists of the mean vector, covariance matrix, control limits, and prior probabilities of each typical process state cluster.

[0041] The data for each process interval of the optical cable to be traced are compared with the template of the normal process interval to calculate the parameter value deviation, parameter fluctuation mode deviation, and parameter correlation deviation. The parameter value deviation is obtained by calculating the Mahalanobis distance between the process parameters and quality characteristic data of the process interval to be traced and the mean vector of the corresponding typical process state cluster. The formula for calculating the Mahalanobis distance is: Σ is the feature vector of the process interval to be traced, μ is the mean vector of the typical process state cluster, Σ is the covariance matrix of the typical process state cluster, and T represents the matrix transpose. The parameter fluctuation mode deviation is obtained by extracting the time series features of each parameter within the process interval to be traced. These time series features include standard deviation, skewness, kurtosis, and first-order autocorrelation coefficient. The extracted time series feature vector is compared with the fluctuation mode benchmark vector in the normal process interval template to calculate the mode difference. The mode difference is measured using Euclidean distance or cosine distance. The parameter correlation deviation is calculated by calculating the Pearson correlation coefficient matrix between the process parameters and quality feature data within the process interval to be traced. The obtained Pearson correlation coefficient matrix is ​​then compared with the correlation benchmark matrix in the normal process interval template using the Frobenius norm distance calculation formula. ,in Here is the Pearson correlation coefficient matrix for the process range to be traced. Let F be the correlation benchmark matrix in the normal process interval template, and ||·||F denote the Frobenius norm of the matrix.

[0042] Process intervals exhibiting at least one deviation exceeding a preset threshold are designated as suspected root cause intervals. The preset thresholds include parameter value deviation thresholds, parameter fluctuation pattern deviation thresholds, and inter-parameter correlation deviation thresholds. The parameter value deviation threshold is set based on the distribution characteristics of Mahalanobis distances across historical process intervals for this type of optical cable; the parameter fluctuation pattern deviation threshold is set based on the quantiles of historical fluctuation pattern differences; and the inter-parameter correlation deviation threshold is set based on the quantiles of historical Frobenius norm distances. When a process interval exhibits a parameter value deviation exceeding a parameter value deviation threshold, a parameter fluctuation pattern deviation exceeding a parameter fluctuation pattern deviation threshold, or an inter-parameter correlation deviation exceeding an inter-parameter correlation deviation threshold, that process interval is designated as a suspected root cause interval.

[0043] Based on the suspected root cause intervals, and combining data from preceding and subsequent processes, a structural causal model is used to infer the root cause processes and parameters that lead to quality anomalies. The construction process of the structural causal model is as follows: The process parameters and quality characteristic data of each process—aluminum-clad steel wire drawing, stainless steel strip forming, optical fiber insertion, grease filling, laser welding, stranding into cables, and sheath extrusion—are used as nodes. The node set includes the following parameters: drawing die aperture, drawing speed, drawing temperature, aluminum-clad steel wire diameter, aluminum-clad steel wire residual stress, rolling pressure, forming roller gap, stainless steel tube straightness, optical fiber tension, optical fiber excess length, grease filling pressure, laser welding power, welding speed, penetration depth, penetration width, stranding pitch, stranding tension, interface reflection coefficient, backscattering intensity, interface spacing, sheath extrusion temperature, sheath thickness, finished product tensile strength, and finished product attenuation coefficient. A directed acyclic graph (DAG) is constructed using the physical causal relationships and process transfer relationships between processes as edges. The direction of the edges points from the cause node to the effect node; for example, from wire drawing speed to aluminum-clad steel wire diameter, from laser welding power to weld penetration depth, and from weld penetration depth to interface reflection coefficient. Conditional probability distributions between nodes are estimated using historical production data. For continuous variable nodes, a multivariate Gaussian distribution or kernel density estimation method is used to fit the conditional probability density function; for discrete variable nodes, a conditional probability table is used. Structural equation modeling is used to quantify the strength of causal effects between nodes. The structural equations can be linear or nonlinear, and the equation parameters are learned from historical data using maximum likelihood estimation or Bayesian estimation methods.

[0044] Once the suspected root cause interval is marked, input parameters affecting the process state of that interval are extracted from the preceding process data within the corresponding length coordinate range. These parameters include the drawing speed, drawing temperature, rolling pressure, forming roll gap, and fiber tension of the preceding process. Quality characteristic data reflecting the process results of that interval are extracted from the subsequent process inspection data, including the interface reflection coefficient, backscattering intensity, finished product tensile strength test value, and finished product attenuation coefficient test value of the subsequent process. The preceding process data, process parameters and quality characteristic data within the suspected root cause interval, and subsequent process inspection data are input into the structural causal model. The average causal effect of each process node on the quality anomaly result is calculated using do-calculus intervention analysis. Specifically, the do operator is executed on each process node within the suspected root cause interval, fixing the value of that process node to its mean value in the normal process interval template, keeping other nodes unchanged, and observing the conditional expectation change of the quality anomaly result variable. The probability that the quality result is within the normal range if the process parameter takes a normal value is calculated using counterfactual reasoning. The formula for calculating the average causal effect is as follows: Where Y is the quality anomaly result variable, and X is the process node variable. and Here, represents the intervention value and the reference value, respectively. `do()` is the do-calculus intervention operator, indicating that an external intervention is performed on variable X and its value is fixed. `E[]` is the conditional expectation operator, and `|` is the conditional separator. The process with the largest average causal effect value is identified as the root cause process. Within this root cause process, the process parameter with the largest absolute value of the path coefficient that has a direct causal path to the quality anomaly is identified as the root cause parameter.

[0045] In response to the inferred root cause process and root cause parameters, a process parameter correction instruction is generated and output to the process closed-loop control layer. The process parameter correction instruction includes the corrected process parameter setpoint, the effective length coordinate range of the correction, and the correction priority. The corrected process parameter setpoint is determined based on the mean or median vector of the corresponding typical process state cluster in the normal process interval template. The effective length coordinate range of the correction is limited to the length coordinate interval corresponding to the suspected root cause interval. The correction priority is determined according to the magnitude of the average causal effect value; the larger the average causal effect value, the higher the correction priority.

[0046] The process closed-loop control layer automatically adjusts the process parameter setpoints for the corresponding processes based on process parameter correction commands, forming a quality-driven adaptive process closed loop. The process closed-loop control layer includes a programmable logic controller (PLC) or an industrial computer, communicating with the production equipment controllers of each process via industrial Ethernet or fieldbus. The communication protocol adopts the OPC unified architecture, Modbus transmission control protocol, or message queue telemetry transmission protocol. Upon receiving a process parameter correction command, the process closed-loop control layer parses the target process identifier, target parameter name, and corrected process parameter setpoints from the command. It then locates the process parameter register address of the corresponding equipment controller through a process parameter mapping table and writes the corrected process parameter setpoints into the register. Based on the new setpoints, the equipment controller transitions the current process parameters from the original setpoints to the corrected process parameter setpoints using proportional-integral-derivative control or model predictive control. The transition rate is set according to the allowable rate of change of the process parameter to avoid process instability caused by sudden parameter changes. The corrected process parameter setpoints are effective within the correction effective length coordinate range. When the production length exceeds the correction effective length coordinate range, the process closed-loop control layer automatically restores the process parameters to the original process parameter setpoints.

[0047] Please see Figure 2This application also provides a data acquisition and quality traceability device for the entire OPGW optical cable production process based on the Industrial Internet of Things (IIoT). This device, used to execute the aforementioned method, includes a digital twin construction module, a multimodal data acquisition module, a traceability analysis module, and a closed-loop control module. The digital twin construction module is used to construct a multi-physics digital twin covering the entire OPGW optical cable production process. This includes establishing a material performance field twin layer, a process field twin layer, a quality inspection field twin layer, a physical field coupling layer, and a performance prediction field twin layer. In the physical field coupling layer, a multi-physics coupling model containing temperature, stress, and deformation fields is established based on the finite element method. A reduced-order model using intrinsic orthogonal decomposition and radial basis function interpolation is employed for computational acceleration. Multimodal sensing data is used to perform near-real-time calibration, test verification calibration, and long-term learning updates on the multi-physics coupling model. The multimodal data acquisition module is used to control multimodal sensing devices deployed in key processes to acquire multimodal sensing data, which includes at least fiber excess length distribution data. The multimodal data acquisition module controls a distributed acoustic emission sensor array to acquire acoustic emission signals and invert fiber excess length distribution data after fiber insertion and before grease filling. It controls a laser ultrasonic testing system to acquire circumferential and axial residual stress distribution data of aluminum-clad steel wire during the wire drawing process. It controls a coaxial and lateral vision fusion detection device to acquire estimated weld depth and measured weld width during laser welding. It controls a ring phased array ultrasonic probe array to acquire interface reflection coefficient, backscattering intensity, and interface spacing during the stranding process. The traceability analysis module is used to perform quality traceability based on multiphysics digital twins and multimodal sensing data, performing process interval comparison learning to identify the root cause of quality anomalies. The traceability analysis module divides the entire production process of the traceable optical cable into multiple continuous process intervals according to length coordinates, obtains normal process interval templates corresponding to the model of the traceable optical cable, calculates parameter numerical deviations, parameter fluctuation mode deviations, and parameter correlation deviations, marks suspected root cause intervals, and infers the root cause process and root cause parameters through a structural causal model. The closed-loop control module converts the root causes identified by the traceability analysis module into process parameter correction instructions and issues them to the corresponding production processes. The closed-loop control module generates a process parameter correction instruction that includes the corrected process parameter setpoints, the effective length coordinate range of the correction, and the correction priority. The process parameter correction instruction is then output to the process closed-loop control layer, which automatically adjusts the process parameter setpoints for the corresponding process.

[0048] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement the steps of the above-described method. The electronic device also includes a communication interface, including an Ethernet interface, a wireless LAN interface, or a fieldbus interface, for data communication with multimodal sensing devices deployed in critical processes and a process closed-loop control layer. The processor receives multimodal sensing data through the communication interface, executes a quality traceability algorithm, and sends process parameter correction instructions to the process closed-loop control layer through the communication interface. The memory includes non-volatile memory and random access memory. The non-volatile memory stores model data of a multi-physics digital twin, template data of normal process intervals, and structural causal model parameters. The random access memory is used to load temporary data and intermediate calculation results during runtime.

[0049] This application also provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method. The computer program includes instructions for performing the following operations: constructing a multi-physics digital twin covering the entire production process of OPGW optical cables; establishing a material performance field twin layer, a process field twin layer, a quality inspection field twin layer, a physical field coupling layer, and a performance prediction field twin layer; establishing a multi-physics coupling model containing temperature, stress, and deformation fields based on the finite element method in the physical field coupling layer and using a reduced-order model with intrinsic orthogonal decomposition and radial basis function interpolation for computational acceleration; controlling multi-modal sensing devices deployed in key processes to collect multi-modal sensing data, which includes at least fiber excess length distribution data; performing quality traceability based on process interval comparison learning based on the multi-physics digital twin and multi-modal sensing data; dividing the entire production process of the optical cable to be traced into multiple continuous process intervals according to length coordinates; obtaining normal process interval templates; calculating parameter numerical deviations, parameter fluctuation mode deviations, and parameter correlation deviations; marking suspected root cause intervals; inferring the root cause process and root cause parameters through a structural causal model; and generating and outputting process parameter correction instructions to the process closed-loop control layer in response to the inferred root cause process and root cause parameters.

[0050] The same or similar labels correspond to the same or similar parts; The terms used to describe positional relationships in the accompanying drawings are for illustrative purposes only and should not be construed as limiting this patent. Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. For those skilled in the art, other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all implementation methods here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the claims of the present invention.

Claims

1. A method for data acquisition and quality traceability throughout the entire OPGW optical cable production process based on the Industrial Internet of Things, characterized in that, The method includes: Construct a multi-physics digital twin covering the entire process of OPGW optical cable production; Multimodal sensing devices are deployed in each step of the entire production process to collect multimodal sensing data; the multimodal sensing data includes at least: fiber excess length distribution data of optical fibers in stainless steel tubes obtained based on the principle of acoustic emission; Based on the multiphysics digital twin and the multimodal sensing data, quality traceability based on process interval comparison learning is performed to identify the root cause of quality anomalies.

2. The method according to claim 1, characterized in that, The fiber excess length distribution data is obtained through the following steps: After the fiber optic cable insertion process and before the grease filling, a distributed acoustic emission sensor array is coupled and installed at both ends of the stainless steel tube. A standard pulse signal is applied into the optical fiber to generate a vibration wave inside the tube; The distributed acoustic emission sensor array receives acoustic emission signals propagating along the optical fiber and the stainless steel pipe wall. The time delay difference and waveform characteristics of the acoustic emission signal are analyzed, and the time difference of arrival positioning algorithm is used to invert the distribution data of the excess fiber length per meter along the length direction of the optical fiber in the stainless steel tube.

3. The method according to claim 1, characterized in that, The multimodal sensing data also includes: Data on the circumferential and axial residual stress distribution of aluminum-clad steel wire collected by a laser ultrasonic testing system; Estimated weld depth and measured weld width of stainless steel pipe laser welding process acquired by coaxial and lateral vision fusion detection device; The interface bonding quality assessment data of the stainless steel pipe and aluminum-clad steel wire interface is acquired by a ring phased array ultrasonic probe array. The interface bonding quality assessment data includes the interface reflection coefficient, backscattering intensity and interface spacing.

4. The method according to claim 1, characterized in that, The multiphysics digital twin is constructed in the following manner: Establish a twin layer for material properties, a twin layer for process technology, a twin layer for quality inspection, a physical field coupling layer, and a twin layer for performance prediction. In the physical field coupling layer, a multi-physics coupling model including temperature field, stress field and deformation field is established based on the finite element method, and a reduced-order model with intrinsic orthogonal decomposition and radial basis function interpolation is used to accelerate the calculation. The multimodal sensing data is used to perform near real-time calibration, test verification calibration, and long-term learning updates on the multiphysics coupling model.

5. The method according to claim 1, characterized in that, The quality traceability based on process interval comparison learning specifically includes: The entire production process of the OPGW optical cable to be traced is divided into multiple continuous process intervals according to the length coordinate. Each process interval is associated with all process parameters and quality characteristic data within that interval. Obtain a pre-constructed normal process interval template corresponding to the model of the optical cable to be traced. The normal process interval template is generated based on the data of qualified optical cables through cluster analysis and statistical process control. The data of each process interval of the optical cable to be traced are compared with the normal process interval template to calculate the parameter value deviation, parameter fluctuation mode deviation and parameter correlation deviation. Process intervals where at least one deviation exceeds a preset threshold are marked as suspected root cause intervals; Based on the suspected root cause interval, and combined with the data from the preceding and subsequent processes, the root cause process and root cause parameters that caused the quality abnormality are inferred through a structural causal model.

6. The method according to claim 5, characterized in that, The method further includes: In response to the deduction of the root cause process and root cause parameters, a process parameter correction instruction is generated and output to the process closed-loop control layer; The process closed-loop control layer automatically adjusts the process parameter settings for the corresponding process according to the correction instructions, forming a quality-driven adaptive process closed loop.

7. The method according to claim 1, characterized in that, The entire production process includes at least: aluminum-clad steel wire drawing, stainless steel strip forming, optical fiber insertion into tubes, grease filling, laser welding, stranding into cables, sheath extrusion, and finished product testing; key processes include wire drawing, laser welding, stranding into cables, and optical fiber insertion into tubes.

8. A data acquisition and quality traceability device for the entire production process of OPGW optical cable based on the Industrial Internet of Things, characterized in that, include: The digital twin building block is used to construct a multi-physics digital twin covering the entire process of OPGW optical cable production; A multimodal data acquisition module is used to control multimodal sensing devices deployed in key processes and acquire multimodal sensing data; the multimodal sensing data includes at least fiber optic excess length distribution data; The traceability analysis module is used to perform quality traceability based on process interval comparison learning based on the multi-physics digital twin and the multimodal sensing data, so as to identify the root cause of quality anomalies.

9. A computer device, characterized in that, include: A memory and a processor, wherein the memory stores a computer program; When the computer program is executed by the processor, it implements the method for data acquisition and quality traceability of the entire production process of OPGW optical cable based on industrial Internet of Things as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method for data acquisition and quality traceability of the entire OPGW optical cable production process based on the Industrial Internet of Things as described in any one of claims 1-7.

Images