Cable defect detection system

Abstract

The invention relates to the technical field of cable nondestructive testing, in particular to a cable defect detection system. The method mainly solves the problems that in the prior art, early-stage tiny defects in a cable are insufficient in recognition sensitivity, the omission ratio is high, and accurate classification cannot be achieved. According to the system, signals are synchronously collected through a multi-physics field composite sensing array, excitation parameters are optimized through a genetic algorithm, a three-dimensional defect probability graph is generated through fusion, a double-current Transform model is adopted to deeply analyze features and automatically recognize defect types, and finally, maintenance is guided through AR visualization, so that non-intrusive online accurate diagnosis and intelligent operation and maintenance of cable defects are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of cable nondestructive testing, in particular to a cable defect detection system. Background Art

[0002] In the field of cable nondestructive testing (NDT), existing mainstream testing methods have numerous inherent flaws. Time domain reflectometry (TDRS) technology is insensitive to changes in cable impedance, making it difficult to effectively identify early-stage, micron-level insulation defects. Furthermore, its limited spatial resolution can easily lead to misjudgments and missed detections of multiple defects in close proximity or faults at cable joints.

[0003] Ultrasonic testing relies heavily on the integrity and uniformity of the coupling agent. In actual field conditions or complex installation environments, coupling conditions are difficult to maintain, resulting in severe signal attenuation and poor stability. Furthermore, this method requires a high level of operator experience, resulting in highly subjective test results and difficulty in automated interpretation.

[0004] Infrared thermal imaging technology can only detect overheating faults on or near the cable surface. It is completely ineffective for detecting early-stage internal defects that have not yet caused a significant temperature rise. It is also highly susceptible to interference from ambient temperature, light, and radiation, resulting in a high false alarm rate. Partial discharge detection is extremely sensitive to environmental electromagnetic noise, making it extremely difficult to extract and isolate effective discharge signals in complex industrial environments.

[0005] Existing technologies often suffer from insufficient detection sensitivity, are subject to demanding testing conditions, or can only detect specific late-stage defects. These technologies fail to meet the urgent need for highly sensitive, reliable, and intelligent identification of early, minute defects within cables. Therefore, there is an urgent need to develop a cable defect detection system to overcome these technical limitations. Summary of the Invention

[0006] This invention provides a cable defect detection system designed to address the challenges of existing technologies for detecting early-stage defects within cables, including low sensitivity, poor anti-interference capabilities, and an inability to intelligently classify them. The system utilizes a multi-physics composite sensor array to simultaneously collect electromagnetic, acoustic, and thermal field signals. It employs a genetic algorithm to dynamically optimize excitation parameters to adapt to different cable defect characteristics. Multi-source data fusion technology is then used to generate a three-dimensional defect probability distribution map. Ultimately, a dual-stream Transformer deep learning model is employed to accurately identify and classify defects.

[0007] This system integrates a lightweight mobile platform to realize autonomous detection operations, and uses AR visualization technology to overlay diagnostic results and maintenance instructions on real scenes, thus building a complete set of services from precise perception, intelligent diagnosis to intuitive interaction. The solution significantly improves the reliability, automation level and operation and maintenance efficiency of cable defect detection.

[0008] The object of the present application can be achieved by the following technical solutions: A cable defect detection system, characterized in that the system comprises a multi-physical field composite sensing array module, an intelligent signal excitation module, a multi-source data fusion processing engine module, a deep defect diagnosis decision module, a lightweight mobile detection module, a collaborative positioning decision hub module and an AR visualization interaction module, wherein: The multi-physical field composite sensing array module is configured to perform multi-physical field coupling scanning on the surface of the cable by relying on a reconfigurable probe shell, thereby obtaining a multi-dimensional original signal set of the cable. The intelligent signal excitation module is configured to call the multi-dimensional original signal set to perform genetic algorithm optimization on excitation waveform parameters, thereby generating type-adaptive excitation instructions for the cable. The multi-source data fusion processing engine module is configured to perform cross-domain feature fusion on multi-source sensing data using the type-adaptive excitation instructions, thereby outputting a three-dimensional probability distribution map of the cable defects. The deep defect diagnosis decision module is configured to analyze the three-dimensional probability distribution map to perform double-flow Transformer analysis on defect features, thereby forming a diagnosis decision report for the type of cable defects. The lightweight mobile detection module is configured to control the path planning of a folding mechanical arm based on the diagnosis decision report, thereby completing the autonomous mobile detection trajectory of the cable. The AR visualization interaction module is configured to match the autonomous mobile detection trajectory to LiDAR-AR calibration matching of the repair space, thereby constructing a visual repair guide model for the cable defects.

[0009] Preferably, when performing multi-physical field coupling scanning on the surface of the cable by relying on a reconfigurable probe shell, the multi-physical field composite sensing array module comprises: The reconfigurable probe shell is configured with a ring-shaped slide rail mechanism to realize continuous circumferential scanning. The microwave antenna array is driven by a stepping motor to rotate and switch the polarization direction, thereby enhancing the response of defects at different angles. The liquid metal coupled ultrasonic transducer is built-in with a piezoelectric ceramic array to form a directional acoustic beam focus. The wide-spectrum infrared thermal imager is integrated with a narrow-band filter set to suppress environmental light interference. The optical axes of the microwave antenna array, ultrasonic transducer and infrared thermal imager are coincident, thereby ensuring spatial coordinate consistency.

[0010] Preferably, when performing coupling scanning, the liquid metal coupled ultrasonic transducer comprises: The gallium-indium alloy fills the cavity to connect a micro-pump, thereby realizing dynamic volume adjustment. The air pressure adjusting mechanism controls the liquid metal pressure in real time according to the cable curvature radius; The cavity bottom is provided with an elastic silica gel sealing film to prevent leakage of the liquid metal; The nanometer particles dispersed in the liquid metal enhance the acoustic impedance matching characteristics; The working frequency range of the ultrasonic transducer covers multiple harmonic frequency bands.

[0011] Preferably, the intelligent signal excitation module includes the following when performing genetic algorithm optimization of excitation waveform parameters based on a multi-dimensional original signal set: The fitness function calculates the energy density ratio of the defect echo amplitude to the background noise; The iterative optimization process constrains the microwave pulse width to the sub-microsecond to millisecond range; The ultrasonic sweep frequency range is dynamically adjusted according to the dielectric properties of the cable insulation material; The genetic algorithm uses a tournament selection strategy to preserve superior genes; The type adaptive excitation instructions are output to a programmable signal generator.

[0012] Preferably, the multi-source data fusion processing engine module includes the following when performing cross-domain feature fusion on multi-source sensing data using type adaptive excitation instructions: The microwave time-domain reflection coefficient is extracted through quadrature demodulation to obtain complex envelope features; The ultrasonic echo attenuation slope is calculated through piecewise linear regression; The infrared temperature rise gradient uses a gradient operator to strengthen the hot spot boundary; The unit size of the unified three-dimensional grid is associated and matched with the cable structure size; The three-dimensional probability distribution map labels the defect spatial distribution and morphological parameters.

[0013] Preferably, the deep defect diagnosis decision module includes the following when performing bi-flow Transformer analysis of defect features based on the three-dimensional probability distribution map: The time-frequency domain signal sequence of the first flow is input into a multi-head self-attention encoding layer; The spatial topology features of the second flow are extracted through a graph neural network to obtain neighborhood associations; The cross-attention layer establishes a time-space feature association mapping matrix; The fully connected layer outputs a probability distribution vector of the cable defect type; The diagnosis decision report includes defect material degradation mechanism analysis.

[0014] Preferably, the lightweight mobile detection module includes the following when performing path planning control of the folding mechanical arm based on the diagnosis decision report: The cable CAD model import parameterized surface geometry; Real-time point cloud is collected by a depth vision sensor and registered to a world coordinate system; The heuristic search algorithm generates an optimized detection path that avoids joints / scaffolds; The mechanical arm joint motion trajectory is smoothly transitioned using a spline curve; The anti-collision monitoring unit interrupts abnormal motion trajectories in real time.

[0015] As a preferred, the lightweight mobile detection module further comprises: The annular coil of the electromagnetic induction power unit coaxially surrounds the cable conductor; The alternating magnetic field frequency is adaptively matched to the fundamental component of the power supply network; The rectifier and voltage stabilizing circuit converts the induced current into a stable DC voltage; The distributed energy storage module buffers the power supply current; The dynamic power distribution unit adjusts the power output according to system load.

[0016] As a preferred, the AR visualization interaction module, when performing LiDAR-AR calibration matching of the repair space according to the matched autonomous mobile detection trajectory, comprises: The defect three-dimensional coordinates are projected to the display coordinate system through the rigid body transformation matrix; The AR coordinate system takes the head-mounted device inertial navigation system as the spatial reference; The spatial anchor points are bound to physical locations through feature point cloud registration; The visual rendering parameters of the virtual identifier are associated with the physical properties of the defect; The LiDAR point cloud data stream is synchronously coupled with the AR display frame rate.

[0017] As a preferred, the AR visualization interaction module, when constructing a visual repair guide model, comprises: The defect type database is associated with a failure mode effects analysis knowledge base; The tool list is intelligently filtered and recommended according to electrical insulation level; The operation steps are decomposed into spatially registered augmented reality animations; The gesture recognition engine verifies the compliance of repair actions in real time; The augmented reality overlay guide integrates a natural language interaction interface.

[0018] The application discloses a cable defect detection system, which is characterized in that a multi-physical field composite sensing array is adopted to synchronously collect electromagnetic wave, sound wave and thermal field signals of a cable, and a genetic algorithm is used to dynamically optimize excitation parameters to accurately adapt to different defect characteristics.

[0019] The application effectively solves the technical problems of low sensitivity, poor anti-interference ability and inability to intelligently identify early internal defects of a cable in the prior art, realizes the integration of the whole process from signal collection, intelligent diagnosis to maintenance guidance, and significantly improves the reliability, automation level and operation and maintenance efficiency of cable defect detection.

[0020] Compared with the prior art, the application has the following beneficial effects: 1. The application significantly improves the detection ability of early internal small defects of a cable through the synergistic effect of multi-physical field composite sensing and intelligent optimization algorithm, can accurately identify hidden dangers such as insulation deterioration, microscopic cracks and partial discharge that cannot be found by traditional single detection means, greatly reduces the risk of cable failure caused by defect missed detection, and provides key technical support for the safe and stable operation of a power system.

[0021] 2. The system realizes automatic analysis and classification of defect characteristics relying on a deep learning model, completely changes the traditional mode of relying on manual experience for signal interpretation, greatly improves the objectivity and consistency of the detection result, reduces the technical dependence of professional technical personnel on the detection operation, and makes the diagnosis work of complex cable defects more efficient and reliable.

[0022] 3. The application deeply integrates advanced detection technology with a mobile and visual platform, intuitively presents the defect position and maintenance scheme through an augmented reality interface, realizes intelligent operation of the whole process from detection, diagnosis to maintenance, effectively solves the actual problems of difficult positioning, slow decision-making and low efficiency in field operation, and comprehensively improves the quality and efficiency of cable operation and maintenance work. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 FIG. 1 is a structural schematic diagram of the cable defect detection system of the application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0025] Embodiment: Figure 1 A structural schematic diagram of a cable defect detection system is given. The cable defect detection system comprises the following modules: A multi-physical field composite sensing array module: multi-physical field coupling scanning is performed on the cable surface relying on a reconfigurable probe shell to obtain a multi-dimensional original signal set of the cable; An intelligent signal excitation module: genetic algorithm optimization is performed on excitation waveform parameters by calling the multi-dimensional original signal set to generate a type adaptive excitation instruction of the cable; A multi-source data fusion processing engine module: cross-domain feature fusion is performed on multi-source sensing data by using the type adaptive excitation instruction to output a three-dimensional probability distribution map of the cable defect; A deep defect diagnosis decision module: double-flow Transformer analysis is performed on the defect features by analyzing the three-dimensional probability distribution map to form a diagnosis decision report of the cable defect type; A lightweight mobile detection module: path planning control is performed on the folding mechanical arm according to the diagnosis decision report to complete the autonomous mobile detection trajectory of the cable; An AR visualization interaction module: LiDAR-AR calibration matching is performed on the maintenance space according to the autonomous mobile detection trajectory to construct a visual maintenance guide model of the cable defect.

[0026] In the embodiment, when the multi-physical field composite sensing array module performs multi-physical field coupling scanning on the cable surface relying on the reconfigurable probe shell, the following is included: The specific implementation of the multi-physical field composite sensing array module involves a highly integrated precision mechanical structure and a sensing system. The core carrier of the module is a reconfigurable probe shell carefully manufactured by using a lightweight ceramic composite material. The selection of this material is based on its excellent comprehensive performance: it not only has the advantage of light weight to reduce the load of the entire mobile platform, but also has the characteristics of high strength, high hardness, high temperature resistance, corrosion resistance and transparency to electromagnetic waves, perfectly adapting to harsh conditions such as high temperature, oil stains and electromagnetic interference that may exist in complex industrial sites. The internal structure of the shell is precisely designed and divided into three independent mounting positions and wiring channels for fixing and mounting three core sensor devices of different principles respectively, ensuring the accurate fixing of the relative positions of the sensors in space, thereby providing a stable spatial registration basis for subsequent data fusion.

[0027] The first component to be installed inside the housing is a microwave antenna array operating in Ka band. The array is composed of multiple microstrip patch antenna elements arranged in a specific pattern to generate directional beams. The operating frequency band is extremely wide, allowing the transmission and reception of linear frequency modulated pulse signals with extremely large bandwidth. The significant advantage of such signals is that they can achieve extremely high range resolution at the receiving end through pulse compression techniques, thus sensitively capturing the small impedance changes caused by defects inside the cable. Adjacent to the microwave array is a 64-element ultrasonic phased array probe with a center frequency of 5 MHz. The probe is composed of 64 independent piezoelectric ceramic chips arranged in a matrix form. Each element can be independently controlled in terms of its transmission and reception delay and phase, allowing the deflection and focusing of the acoustic beam through electronic control without the need for mechanical movement. This greatly improves the flexibility and efficiency of the detection. The third key sensor is a long-wave uncooled infrared focal plane detector. The size of its pixels determines the clarity of the temperature image, and it has extremely high detection sensitivity for the weak temperature rise caused by partial discharge or increased dielectric loss due to internal defects on the cable surface.

[0028] To ensure that these sensors can detect the cable surface without missing any areas, the housing is mounted on a precision motion mechanism. The core of the mechanism is a ring-shaped servo slide rail at the bottom of the housing and a high-precision linear motor. During the detection process, the ring-shaped servo slide rail is first started to drive the entire probe housing to rotate smoothly and uniformly around the axis of the cable, achieving a complete scanning of the cable. At the same time, the linear motor drives the entire rotating mechanism to move a small distance along the axial direction of the cable according to the preset step distance after each rotation cycle is completed. The combination of rotational and linear motion allows the sensors at the end of the probe to form a complete spiral scanning trajectory relative to the cable surface, ensuring full coverage of the detection area without any dead zones.

[0029] During the scanning process, the three sensors perform their respective data acquisition tasks according to their physical characteristics. The microwave antenna array, under the control of the FPGA, periodically emits nanosecond-level broadband linear frequency modulation pulse signals that penetrate the cable insulation layer and propagate. When encountering impedance discontinuities such as defects, water droplets, bubbles, or conductor deformations within the insulation layer, a portion of the backscattered signal is generated. This signal is captured and recorded by the sensitive receiving channel. Almost at the same time, the ultrasonic phased array, under the control of another FPGA channel, excites each array element to emit ultrasonic waves according to the pre-calculated delay rule. The sound waves propagate in the cable medium and also encounter acoustic impedance difference interfaces, generating echoes. The echo signals are received by each array element and undergo amplification and digitization processing. At the same time, the infrared focal plane detector continuously captures infrared radiation images of the cable surface at a very high frame rate, thereby recording the subtle spatiotemporal changes in the cable surface temperature field during the scanning process. The three data streams are synchronized and labeled in time by a high-stability atomic clock and spatially associated by the encoder data of the motion mechanism.

[0030] All collected raw data, including microwave I / Q raw data, ultrasonic RF waveform data, infrared thermal image data, and corresponding spatial position coordinates and timestamps, are packaged, compressed, and indexed in real time by a data management unit to form a structured multi-dimensional raw signal set. This data package contains synchronized perception information of the cable's state at the same position from the electromagnetic, acoustic, and thermal dimensions, laying a solid data foundation for subsequent intelligent signal processing and fusion analysis.

[0031] In this embodiment, the liquid metal coupled ultrasonic transducer includes, when performing coupling scanning: The core of this module is to completely solve the most critical coupling problem in ultrasonic detection, i.e., how to establish an efficient, stable, and non-destructive low-impedance sound wave transmission channel between cables of different diameters and surface states and transducers. Traditionally, gel or grease solid-state coupling agents are commonly used, but they have many limitations that are difficult to overcome in actual field or complex industrial environments: they are volatile and dry, leading to a sharp decline in coupling effectiveness; they easily pick up dust and oil, forming new acoustic barriers; it is difficult to ensure uniform thickness when applying them to curved surfaces, and they are prone to trapping air bubbles, introducing significant acoustic energy loss and signal interference; in addition, their maintenance and cleaning work is tedious, severely affecting detection efficiency. This design aims to solve these inherent problems once and for all through a new active liquid metal coupling mechanism.

[0032] The key component of its implementation is a closed flexible cavity directly machined in front of the sound emission surface of the ultrasonic transducer. The design of this cavity is critical, its shape is a flat pouch structure, the bottom is a flexible pressure-bearing surface in contact with the cable, made of biomedical-grade high-strength silicone film, its thickness is precisely calculated, requiring good flexibility while minimizing its own attenuation of specific frequency sound waves. The four sides and top of the cavity are supported and sealed by a stiffer engineering plastic frame, forming a closed container. The container is filled with a sufficient amount of gallium, indium, and tin ternary eutectic alloy liquid metal with a specific weight percentage ratio. The choice of this ratio is verified by a large number of experiments, aiming to ensure that the alloy remains liquid at the expected field working temperature range, while having low viscosity and high surface tension, and its acoustic impedance characteristics are designed to be between the piezoelectric ceramic wafer and the typical cable insulation material, acting as an ideal acoustic transition layer.

[0033] In order to make this cavity from a passive container into an intelligent mechanism that can actively adapt to the shape of the cable, a high-precision, low-noise micro-step air pump is connected to the closed cavity through a flexible Teflon tube. The air pump has a built-in precision pressure sensor, forming a closed-loop feedback control system. Its working process is as follows: when the transducer is driven by the robot arm to be in contact with the cable surface, the air pump first starts, injecting or extracting gas into the cavity, thereby dynamically and accurately adjusting the air pressure in the cavity. This pressure change directly acts on the liquid metal in the cavity, pushing it to produce a conformable deformation. The flexible bottom membrane will therefore bulge outward or contract inward under the hydraulic transmission of the liquid metal, like a flexible "palm", actively wrapping and fitting the current cable curve, whether its diameter is thick or thin, whether the surface has concave and convex factory marks or armor traces. This process continues to fine-tune until the system determines through pressure readings that the liquid metal has reached the preset optimal contact pressure state with the cable surface, thereby physically completely eliminating any air gap, creating almost perfect conditions for ultrasonic transmission.

[0034] In addition, in order to further improve the transmission efficiency of acoustic energy, that is, to achieve better acoustic impedance matching and enhance the conduction of ultrasonic waves, the basic liquid metal medium is functionally enhanced at the nanoscale. In the gallium-indium-tin alloy liquid metal, specific mass fraction of nano-sized zinc oxide particles are uniformly dispersed through specific surfactant treatment, mechanical stirring, ultrasonic dispersion and other processes. The introduction of these nanoparticles is not a simple physical mixture, and their size, morphology and concentration are strictly designed and controlled. These high-hardness nanoparticles form a stable suspension system in the liquid metal, which effectively changes the overall density and sound speed of the composite medium, and then finely regulates its acoustic impedance value, making it more perfect to bridge the acoustic property difference between the transducer wafer and the cable insulation material. More importantly, these dispersed nanoparticles can effectively reduce the scattering loss of sound waves during propagation under the action of the acoustic field, so that the sound wave energy can be more concentrated and efficient injected into the measured cable, thereby significantly enhancing the signal-to-noise ratio of the echo signal and the detection ability of the detection system to small defects.

[0035] The module, through the synergistic effect of the "flexible cavity structure", "functional liquid metal medium", "active air pressure control system" and "nanoparticle enhancement technology", forms an intelligent, self-adaptive and high-performance ultrasonic coupling solution. It not only physically ensures the absolute consistency of the coupling interface, eliminates the uncertainty of manual operation and the inherent defects of traditional coupling agents, but also optimizes the energy transmission path from the acoustic physics level, thereby laying a solid foundation for subsequent acquisition of high-quality and reliable ultrasonic detection data. It is an indispensable key link in the entire high-precision detection system.

[0036] In the embodiment, the intelligent signal excitation module includes the following steps when performing genetic algorithm optimization on the excitation waveform parameters by calling the multi-dimensional original signal set: The core task of the intelligent signal excitation module is to realize the adaptive optimization of detection parameters, and its specific implementation is a closed-loop feedback and decision-making process based on genetic algorithm. The process starts with deep mining of initial scanning data. The built-in high-speed digital signal processor in the module performs real-time spectrum analysis and feature extraction from the multi-dimensional original signal set. It does not process all the massive raw data, but calculates the width and amplitude envelope of the echo pulse from the microwave reflection signal, extracts the energy integral in a specific time window from the ultrasonic echo sequence, and calculates the temperature difference standard deviation of a specific area from the infrared image sequence. These preliminary refined time domain, frequency domain and spatial domain features together form the initial feature set for evaluating the effectiveness of the defect response under the current excitation parameters, providing a quantitative evaluation basis for the subsequent optimization algorithm.

[0037] On this basis, the system builds an optimization goal with signal-to-noise ratio as the core. Specifically, the algorithm establishes the ratio of the energy of the defect echo signal to the background noise energy as the only and most important fitness function of the genetic algorithm. The calculation of this function is extremely critical: the defect echo energy is obtained by integrating the specific window of the characteristic signal in the time domain or frequency domain, representing the strength of the useful information; and the noise energy is taken from the "quiet zone" in the signal far from the defect echo and without other obvious interference, representing the level of system background noise and random interference. Maximizing the ratio of the two directly means maximizing the detection sensitivity and reliability. This highly quantitative objective function, like a fair judge, provides an objective and unified standard for evaluating the pros and cons of each set of parameter schemes.

[0038] After establishing the goal, the genetic algorithm starts its iterative evolution process. In the initialization, the algorithm randomly generates a certain number of individuals in a pre-defined, broad but reasonable parameter space, forming an initial population. Each individual here is not a single numerical value, but a complete, multi-dimensional parameter combination scheme, which uniquely encodes all key adjustable excitation parameters in this detection task. These parameters usually include but are not limited to: the carrier frequency and instantaneous bandwidth of the microwave pulse, which determine the depth and resolution of microwave detection; the center frequency, pulse repetition frequency, and amplitude of the transmitted voltage of the ultrasonic probe, which together determine the penetration ability, detection speed, and signal-to-noise ratio level of the sound wave. Each individual represents a unique, "tentative" detection strategy.

[0039] Subsequently, the algorithm enters the cycle of "selection - crossover - mutation". In the selection stage, the algorithm sorts the individuals according to the calculated fitness function values of their parameter combinations, adopts strategies such as tournament selection, and decisively eliminates inferior individuals with low performance and poor signal-to-noise ratio, while retaining and copying outstanding high-fitness individuals to ensure that good "genes" are passed on. In the crossover stage, the retained excellent individuals exchange some of their parameter codes with a certain probability, for example, exchanging the microwave bandwidth parameter encoded in one individual with the ultrasonic frequency parameter encoded in another individual, so as to have the opportunity to produce new generation parameter combinations with the advantages of both parents. In the mutation stage, some parameters of individual individuals in the population are randomly adjusted with a small probability, for example, slightly changing the pulse repetition frequency. Although this step occurs with a low probability, it is crucial as it injects new genetic diversity into the population, preventing the algorithm from converging prematurely to a local optimal solution rather than a global optimal solution. This cycle continues for tens or even hundreds of generations.

[0040] After the above-mentioned sufficient iterative evolution, the quality of individuals in the population is continuously improved, and finally converges to one or a few excellent performance optimal solutions. The algorithm selects the individual with the highest fitness from the optimal solutions, and decodes the parameter combination coded by the individual into a complete and executable control command. The final output of the "type adaptive excitation instruction" is the optimal detection scheme customized for the current cable being detected. The instruction set is immediately issued to the core hardware responsible for signal generation, field programmable gate array, through the high-speed communication bus. The FPGA accurately controls the direct digital frequency synthesis module and high-voltage pulse generation circuit inside it according to the optimal instruction received, generates microwave and ultrasonic driving signals with perfectly optimized amplitude, frequency and timing, and synchronously outputs them to the front-end sensor, thereby starting a new round of detection scanning with higher performance and higher precision, and finally realizing the most sensitive perception and capture of cable defects.

[0041] In the embodiment, the multi-source data fusion processing engine module includes the following when performing cross-domain feature fusion on multi-source sensor data using the type adaptive excitation instruction: The specific implementation of the multi-source data fusion processing engine module is a key process of converting heterogeneous data into unified cognition. The core task of the module is to receive the instructions issued by the upper computer based on the genetic algorithm optimization, and drive the sensor system to perform a new scanning with optimized parameters, so as to obtain a set of high-quality, synchronously collected multi-source sensor data. This set of data carries the mapping information of the internal state of the cable from three different physical dimensions of electromagnetism, acoustics and heat, but the original forms are different, the dimensions are different, and the reflected physical phenomena are completely different, which cannot be directly compared and comprehensively judged. Therefore, the first task of the fusion engine is to independently preprocess and extract features from the three data streams with unified goals, so as to prepare standardized and structured feature fields for subsequent deep fusion.

[0042] The preprocessing is performed on the data of the microwave channel. The originally collected microwave signal is a time-domain waveform containing amplitude and phase. In order to more clearly reveal the impedance discontinuity points inside the cable caused by defects, the module uses the classical signal processing tool Hilbert transform to analyze it. The transform can accurately strip and extract the instantaneous envelope and instantaneous phase information of the signal by constructing the analytic signal of the original signal. The peak position of the extracted envelope directly corresponds to the axial distance of the defect, and the amplitude and phase changes of the envelope contain rich information about the dielectric constant change of the medium at this place. Through inversion calculation of this information, the system can preliminarily reconstruct the dielectric constant distribution profile of the cable axis, thereby obtaining a feature field reflecting the change of electromagnetic properties of the insulating material.

[0043] Almost in parallel, the data from the ultrasonic channel also undergoes a series of complex corrections and transformations. Since the ultrasonic wave inevitably experiences dispersion and amplitude attenuation when propagating in a cable-like viscoelastic medium, it will cause distortion of the echo signal and degradation of resolution. Therefore, the module will call the dispersion curve and attenuation coefficient of this type of cable, which is calibrated by experiment in advance, to perform dispersion compensation and attenuation correction on the received raw radio frequency ultrasonic signal. Dispersion compensation aims to synchronize the arrival of sound wave components of different frequencies, thereby sharpening the echo waveform; attenuation correction aims to compensate for the loss of sound energy with increasing propagation depth, so that echo amplitudes of defects of the same size but different depths are comparable. After these precise compensations, the algorithm calculates the decay rate of the echo signal energy with propagation time to finally generate a decay slope field that can directly reflect the acoustic loss characteristics of the material. This field is extremely sensitive to defects such as micro-cracks and pores.

[0044] For infrared thermal image data, the processing strategy focuses on capturing dynamic changes. The system receives a series of thermal image frames collected in time sequence. The module uses time difference processing technology to subtract the pixel-level gray value of each subsequent frame from a selected reference frame. This processing can greatly suppress static thermal noise interference from background, environmental radiation, and uneven surface emissivity of the cable itself, thereby highlighting small temperature rise areas newly generated or changing due to internal defect activity. Further calculation of these difference results can extract an abnormal temperature rise gradient field representing the spatial gradient of temperature change, which is direct evidence for locating active thermal defects.

[0045] After successfully extracting the three essentially different characteristic fields of "permittivity distribution", "acoustic attenuation slope" and "temperature rise gradient" from three independent data sources, the biggest challenge for fusion is how to make them talk on the same stage. This is achieved through a pre-set three-dimensional space registration matrix. This matrix is obtained through precise measurement and calculation during the system calibration stage, and it establishes a strict mathematical mapping relationship between the physical coordinates of each sensor, the encoder readings of the scanning motion mechanism, and a virtual three-dimensional grid space based on the accurate CAD model of the cable. Using this matrix, the algorithm can map the data of each feature point in the above three characteristic fields to the corresponding voxel of the virtual three-dimensional grid without deviation. At this point, the data originally in different coordinate systems and reflecting different physical quantities are perfectly unified in the same spatial reference system.

[0046] In this unified three-dimensional grid space, the D-S evidence theory, a high-level information fusion algorithm, is used to complete the final decision fusion. This theory is very suitable for dealing with multi-source information problems with uncertainty. The system regards the data from three different sensors in each voxel unit as three independent "evidence", each of which provides a certain support probability for the proposition "the voxel belongs to the defect" or "does not belong to the defect". The core combination rule of D-S theory is responsible for integrating the three evidences and calculating the comprehensive probability supported by their joint support. By traversing all voxels in the three-dimensional space, this fusion process finally generates a comprehensive three-dimensional probability distribution map. This map is no longer a reflection of a single physical quantity, but a comprehensive analysis based on electromagnetic, acoustic and thermal evidence. It directly reveals the most likely location, approximate spatial range and relative credibility of the defect in the cable in the form of probability values, providing the most direct and reliable basis for the final intelligent diagnosis.

[0047] In the embodiment, the deep defect diagnosis decision module includes the following steps when performing the double-flow Transformer analysis of the three-dimensional probability distribution map to analyze the defect characteristics: The specific implementation of the deep defect diagnosis decision module embodies an advanced artificial intelligence analysis paradigm that fuses spatial and temporal information. The core input of the module is the three-dimensional probability distribution map produced by the upstream fusion engine. Each voxel in this image contains a probability value fused by the D-S evidence theory, representing the comprehensive confidence of the existence of defects in that spatial location. However, this is still a raw probability field that needs to be further interpreted and understood. The core task of the module is to act as a highly specialized "image diagnosis doctor" to accurately interpret the type, range, location and other key information of the defect from this three-dimensional image. The core architecture to achieve this goal is a carefully designed double-flow neural network that analyzes data from both sequential and spatial patterns to simulate the thinking way of human experts in comprehensive analysis of problems.

[0048] The first to be launched is the temporal flow analysis pipeline, which draws its inspiration from the powerful modeling capabilities of natural language processing in handling sequential information. The system first cuts the three-dimensional probability volume data continuously like a stack of cross-sectional slices along the axial direction of the cable. These slice images are arranged in a strict temporal sequence according to their spatial order, and each "frame" in the sequence no longer represents a moment in time, but rather the cross-sectional probability distribution at a specific axial position on the cable. This carefully constructed sequence is fed into an encoder based on the Transformer architecture for in-depth analysis. The core of the Transformer model is its self-attention mechanism, which can dynamically calculate the correlation weight between any two slices in the sequence. Through this mechanism, the model can sensitively capture the distribution patterns, periodicity and long-range dependencies of defect probabilities along the length of the cable, such as identifying whether a defect presents a discontinuous chain distribution in the axial direction or whether an anomaly at one location is associated with an anomaly at another location, thereby deeply understanding the extension characteristics of defects in the axial dimension.

[0049] Working in parallel with the temporal flow is the spatial flow analysis pipeline, which aims to deeply mine the geometric and topological information of defects in three-dimensional space. In this pipeline, the entire three-dimensional probability distribution map is reinterpreted as a huge graph structure. Each node in the graph is a voxel, and the features carried by the node are its probability value and its three-dimensional coordinates. The connection between nodes is defined according to their spatial proximity, usually using the k-nearest neighbor or fixed radius neighborhood method to connect adjacent voxel nodes in physical space. This constructed graph is fed into a multi-layer graph convolutional network for processing. The core of the graph convolution operation is "message passing", that is, each node aggregates the feature information of its direct neighbor nodes to update its own feature representation. Through multi-layer stacking, each node can ultimately perceive the features of a larger region within a certain radius. This process enables the network to accurately extract complex spatial topological features of defect regions, such as distinguishing whether a defect is a spherical hole, a sheet-like crack or a tree-like fractal structure, and the smoothness and orientation of its boundary, which are crucial for distinguishing defect types.

[0050] After the two paths complete their respective feature extraction, two high-dimensional feature representations with different modalities and perspectives are generated: the temporal stream outputs a feature vector containing global sequence context information, and the spatial stream outputs a node feature set containing detailed three-dimensional morphological information. At this time, a cross-attention module is enabled, acting as a "translator" and "integrator" between the two information domains. The module takes the features of one stream as the Query vector, and the features of the other stream as the Key and Value. By calculating the similarity between the Query and the Key, a set of attention weights is obtained, and the values are weighted and summed using these weights. This process essentially allows the model to learn "when analyzing defects in a certain spatial location, which axial sequence context information should be referred to" and vice versa, thereby achieving the deep, dynamic, and nonlinear fusion of temporal and spatial features, generating a unified feature representation that combines macro distribution rules and micro morphological details.

[0051] This top-level feature representation after deep fusion is sent to an output network composed of a fully connected classification layer. This network is essentially a powerful pattern classifier, and its output neurons correspond to various defect types predefined by the system. The network not only outputs the most likely defect type label, but also gives the confidence probability of belonging to that category. In addition, through the regression head, the network can also accurately calculate the envelope size of the defect region in three-dimensional space and the precise three-dimensional coordinates of its centroid in the global coordinate system of the cable. All this information - defect type, confidence, three-dimensional size, and precise coordinates - is automatically assembled and formatted to generate a detailed, quantitative, and directly actionable final diagnostic decision report, thus completing the ultimate leap from raw sensor data to high-level cognitive decision-making.

[0052] In this embodiment, the lightweight mobile detection module includes the following when performing path planning control on the folding mechanical arm according to the diagnostic decision report: The specific implementation of the lightweight mobile detection module is the core executive mechanism for realizing automated precision scanning. Its hardware foundation is a high-performance six-degree-of-freedom collaborative robot. This robot is specially selected and customized, made of lightweight alloy materials to reduce overall weight, and has a repeat positioning accuracy that meets industrial standards, ensuring high pose certainty of the sensor probe in complex spatial motion. The robot is integrated on the mobile platform through a rigid and shockproof mounting base, and the sixth-axis flange at the end of the robot is specially designed to securely mount and connect the aforementioned multi-physical field composite sensor probe assembly, forming a terminal actuator that integrates sensing and motion. The entire robot system is centrally controlled by an embedded industrial computer with a high-performance multi-core processor that runs complex path planning algorithms and real-time kinematics calculations, serving as the "brain" of the mobile detection module.

[0053] The initialization of the path planning process begins with the interpretation of the diagnostic conclusion. After receiving the structured diagnostic decision report from the upstream diagnostic module through the standard communication interface, the algorithm first parses the report and accurately extracts the three-dimensional spatial coordinates of each identified defect. These coordinates are usually defined in a "cable coordinate system" with the cable itself as the reference. In order to drive the motion of the robot arm, it is necessary to convert these target point coordinates into the "base coordinate system" that the robot arm can understand. This conversion process involves a key coordinate transformation, which is achieved through a homogeneous transformation matrix accurately obtained through a hand-eye calibration program. Hand-eye calibration is achieved by fixing a calibration object at the end of the robot arm and guiding the robot arm to move to multiple different poses, using the vision sensor on the platform to solve the relative position relationship between the camera, robot base and cable, and thus obtain this fixed transformation matrix. By applying this matrix, the coordinates of the defect points in the cable coordinate system are accurately mapped into the robot base coordinate system, providing accurate spatial targets for subsequent planning.

[0054] After obtaining the absolute coordinates of the target points, the path planning algorithm begins to find a safe and efficient motion path for them in the three-dimensional workspace they are in. This three-dimensional space is a virtual environment model that includes the robot arm itself, the cable to be tested, and the surrounding known obstacles. The module uses an improved Rapidly-exploring Random Tree (RRT) algorithm to complete this task. The RRT algorithm explores the feasible path by randomly sampling in the configuration space and expanding the tree structure. Its improvement lies in that it not only finds a feasible path, but also optimizes the initial path to minimize its cost. The algorithm considers the robot arm's links and probe as a whole, and performs collision detection under the support of its kinematic model, ensuring that every pose on the planned path will not interfere with the cable itself and any obstacles on it, thus achieving "collision-free" intelligent movement.

[0055] The output of the planning algorithm is usually a series of discrete waypoints defined in Cartesian space or joint space. If this sequence is sent directly to the controller, the motion of the robot arm will be stiff and jerky, with abrupt straight-line motion between points, which not only causes mechanical vibration and affects probe stability, but also may produce unnecessary impact. Therefore, the path must be smoothed. The module uses a quintic polynomial trajectory generator to complete this task. The generator will interpolate a motion curve that is continuous and smooth in time, position, velocity, acceleration and even jerk for each joint of the robot arm between two adjacent path points. The quintic polynomial ensures the continuity of acceleration, making the motion of the robot arm extremely smooth and compliant, minimizing impact and vibration, and providing a stable motion basis for high-precision detection.

[0056] The smoothed joint space trajectory parameterized by time is converted into real-time position command stream and sent to the underlying multi-axis servo controller of the robot arm through high-speed field bus. After receiving the commands, the controller closes the current loop, speed loop and position loop in its internal, and accurately drives the servo motor of each joint to move according to the predetermined and smoothed trajectory. Thus, the robot arm can autonomously and smoothly move the end-mounted sensing probe to each defect point coordinate that needs to be finely inspected, maintain the optimal coupling posture with the cable surface, and accurately complete the entire autonomous movement detection trajectory, and convert the intelligent diagnosis conclusion into precise physical execution action.

[0057] In the embodiment, the lightweight mobile detection module further comprises: To solve the long-term endurance problem of the mobile detection platform in the field or in the environment without external power supply, a high-efficiency and self-adaptive energy acquisition unit, i.e., an electromagnetic induction power taking unit, is designed and implemented. The core design concept of the unit is to acquire energy from the detected live cable itself, realizing the true "taking from the cable and using for detection", and the technical basis is Faraday's law of electromagnetic induction. The entire unit is not a simple coil, but a complex system integrating advanced soft magnetic materials, precise mechanical structures and high-efficiency power electronic conversion technology. It needs to solve a series of engineering challenges: how to efficiently extract energy from the power frequency magnetic field with varying intensity; how to adapt to cables with different thickness and current levels; and how to process the induced unstable alternating current into stable, clean and usable direct current power for electronic systems. The successful implementation of the unit is the key prerequisite for ensuring the long-term autonomous operation of the entire system.

[0058] The energy capture core of the unit is a specially designed current transformer. The magnetic core is wound with nanocrystalline alloy strip and annealed. This material has extremely high initial permeability and saturation magnetic induction strength, and extremely low coercive force, which means that even in the case of small cable load current and weak magnetic field, the magnetic core can be easily magnetized to generate sufficient magnetic flux, and it is not easy to be saturated when encountering large instantaneous current, ensuring the wide range adaptability and safety of energy taking. The magnetic core is shaped into a ring with an opening and packaged in a high-strength engineering plastic shell to form a classic clamp structure. One side of the shell contains a precise hinge and spring mechanism, so that the magnetic ring can easily open and close like a clamp, so it can be easily clamped around the live cable without any electrical connection and without damaging the original insulation of the cable, realizing truly non-invasive installation. The high-conductivity magnetic ring efficiently collects the surrounding magnetic field generated by the cable current and dispersed in space into its own magnetic circuit.

[0059] Wound on the nanocrystalline magnetic core is a secondary coil wound with high-strength enameled wire. The number of turns of the coil is carefully calculated and experimentally verified to achieve the best balance between induced voltage and output current to obtain maximum power output. When the clamp unit is clamped on the cable, the alternating magnetic field generated by the power frequency alternating current flowing in the cable is coupled to the secondary coil by the magnetic core, and according to Faraday's law of electromagnetic induction, an alternating voltage of the same frequency will be induced at both ends of the coil. The voltage amplitude is proportional to the current in the cable and the number of turns of the coil. However, the electric energy directly induced from the coil is rough and cannot be directly used, and its voltage and frequency fluctuate sharply with the cable load current, and must be carefully conditioned by subsequent power electronic circuits.

[0060] Therefore, the induced alternating current is then sent to a high-efficiency electric energy processing circuit module. The module first passes through a full-bridge rectifier circuit composed of fast recovery diodes to convert alternating current into pulsed direct current. Then, a multi-stage filter network composed of large-capacity electrolytic and ceramic capacitors smooths the rectified direct current, filters out the power frequency ripple and high frequency burr, and obtains a relatively stable direct current with a voltage value that still changes with the primary current. The most critical step is completed by a high-performance DC-DC buck-boost voltage regulator chip. The chip and its peripheral circuit form a closed-loop feedback system that can accept a wide range of input voltages and dynamically adjust the energy conversion process by controlling the internal switching MOSFET through high-frequency PWM, ultimately generating an extremely stable, low ripple coefficient direct current voltage at its output. This voltage fully meets the power supply needs of the mobile platform host computer, mechanical arm servo driver, sensors and all other subsystems.

[0061] This highly integrated electromagnetic induction power unit converts the magnetic energy originally dispersed around the cable into a stable and reliable "micro power station" through multi-level conversion of physics, circuits and control. It completely eliminates the dependence on traditional battery endurance time or on-site socket, providing nearly unlimited continuous power supply for the entire mobile detection platform. This enables the system to perform long-term and large-scale automated inspection operations, greatly expanding its application scenarios and practical value, and is an indispensable cornerstone for the high degree of autonomy of the entire equipment. The wide input range design ensures that the system can stably and reliably obtain energy whether in a light load line during a low electricity consumption period or in a heavy load line during a high electricity consumption period, thereby truly achieving all-weather energy self-sufficiency.

[0062] In the embodiment, when the AR visualization interaction module performs LiDAR-AR calibration matching on the maintenance space according to the matched autonomous mobile detection track, it includes: In this embodiment, the specific implementation of the AR visualization interaction module is the key technical bridge for realizing the superposition of maintenance guidance from abstract data to concrete space. Its core lies in solving the problem of high-precision, real-time and stable spatial registration between virtual information and the physical world. The hardware carrier of this module is a high-performance mixed reality glasses. Unlike ordinary consumer-grade AR devices, it is highly integrated with two core sensors on the temples to meet the high-precision requirements of industrial inspection: a high-precision area array LiDAR scanning module and a high-resolution RGB color camera. The LiDAR module can actively and quickly acquire depth information of the surrounding environment by emitting invisible laser beams and measuring their return time. Its accuracy can reach the millimeter level and is not affected by changes in ambient light; the RGB camera passively captures the color texture information of the environment. The two complement each other and together provide a rich data source for spatial perception, forming the "eyes" of the system.

[0063] After the system boots up, it first executes the environmental perception and map-building process. The LiDAR module is activated, emitting laser pulses at its inherent high frequency and performing a high-speed rotational scan of the entire maintenance scene, including cables, brackets, the ground, and even walls within its current field of view. Millions of laser point ranging data are received and processed in real time. Each point has its own three-dimensional coordinates (X, Y, Z) relative to the LiDAR sensor itself. This massive collection of points forms a high-density three-dimensional point cloud map that describes the geometry of the environment. This process is rapid and automated, typically completing the initial scan of the local work area within seconds, establishing an initial, accurate digital twin framework of the physical environment. Simultaneously, the RGB camera continuously captures video streams at a high frame rate.

[0064] Next, the visual SLAM algorithm begins working in parallel, tasked with solving the core problem of determining the glasses' position and orientation in space. The algorithm performs real-time feature extraction on incoming consecutive video frames, identifying stable features such as corners and edges within the image and matching those features between consecutive frames. By calculating the pixel displacement of these feature points within the image and combining them with possible inertial measurement unit data, the algorithm can infer the AR glasses' six degrees of freedom (DOF) position in space in real time at extremely high frequency. These are three translational components (movement along the X, Y, and Z axes) and three rotational components. This real-time pose estimation ensures the system always knows where the user is and where they are looking, a prerequisite for correctly placing virtual objects.

[0065] With the precise point cloud of the physical environment and the real-time pose of the glasses, the system needs to achieve the most critical step: precisely align the virtual cable model with the real cable entity. This is done through a point cloud registration algorithm. The input of the algorithm has two: one is the point cloud just scanned by LiDAR, which describes the real cable and its surrounding environment; the second is the virtual cable model point cloud generated by the CAD model from the upstream autonomous movement detection trajectory, which is absolutely accurate. The core task of the algorithm is to calculate an optimal rigid transformation matrix, which, when applied to the "source point cloud", can make the virtual model point cloud and the real scene point cloud overlap in space to the highest degree. This calculation process is iterative optimization, and the final transformation matrix accurately defines the mathematical mapping relationship between the virtual coordinate system and the real world coordinate system.

[0066] With this calculated accurate transformation matrix, all virtual information that needs to be displayed-including three-dimensional highlight marks of defect positions, text annotations, arrow directions, and sectioned models-are first defined in the virtual cable model coordinate system. At the beginning of each frame's rendering period, the system converts the coordinates of these virtual elements to the current AR glasses display coordinate system according to the real-time acquired glasses pose and applies the above transformation matrix. The graphics rendering engine then superimposes these virtual contents onto the real world picture seen through the camera according to the final screen coordinates, with correct perspective and occlusion relationship. As a result, the maintenance personnel see through the AR glasses: a realistic cable virtual twin that is completely locked in space with the real cable, with defect points clearly marked at their real and accurate positions, with related diagnostic data and operation instructions suspended around, achieving millimeter-level precise, stable, and seamless fusion of virtual information and physical entities, and converting abstract detection reports into intuitive and operable on-site maintenance guidelines.

[0067] In the embodiment, when the AR visualization interaction module constructs the visualization maintenance guide model, it includes: The AR visualization interaction module is the key link to transform all the detection and diagnosis results into the final executable maintenance operation. Its core is an intelligent decision support system integrating expert knowledge, computer graphics and natural interaction. The cornerstone of the system is a deep structured defect type knowledge base, which is not a simple database table, but a professional knowledge graph based on ontology. The knowledge base is built by long-term accumulation of experimental data, field cases and industry standards. Each node represents a specific cable defect type, and the nodes are connected by attribute relationships. Each node encapsulates multi-dimensional attribute data of the defect type, including its typical feature pattern under multi-physical field scanning, its root physical and chemical cause, its risk level according to relevant standards, and most importantly, the verified standard maintenance process. This knowledge base is the intelligence source of the system's expert-level diagnosis and guidance capability.

[0068] When the diagnosis report generated by the upstream deep defect diagnosis decision module is transmitted to this module, the system immediately starts the intelligent scheme matching process. The diagnosis report contains the type identifier, confidence, size and coordinates of the defect. The core reasoning engine of the system first analyzes the report and extracts the most critical type identifier as the main search key to perform accurate query and matching in the vast knowledge graph. The matching process is not a simple key-value lookup, but a complex process that may involve uncertain reasoning. For example, when the system's recognition confidence for a defect is not 100%, the engine may search for several most likely defect type nodes in parallel and recommend a most conservative or comprehensive maintenance scheme based on the risk level and the difference of maintenance processes. After successful matching, the engine will extract the complete set of maintenance schemes customized for the defect from the corresponding knowledge node. These information are structured into a data object to be rendered.

[0069] Subsequently, the AR rendering engine is activated to start receiving the processed data objects and perform the multi-level, multi-modal information visualization rendering task. This is the core process of translating abstract data into intuitive visual experience. The rendering is performed in layers: at the bottom layer, which is the layer of direct interaction with the real world, the engine renders a semi-transparent and highlighted 3D warning model, such as a pulsating red sphere, around the defect point or a wireframe model that outlines the defect edge, at the corresponding real position in the AR field of view, according to the precise 3D coordinates of the defect and its physical size, in real time. The purpose is to guide the maintenance personnel to lock the target without ambiguity in the first time in a complex on-site environment. On the side or above the field of view, the engine generates a suspended information panel HUD in the form of dynamic graphics, which clearly lists and highlights the required list of special tools, the list of consumables, and the decomposed operation steps. For extremely complex or critical steps, the engine calls pre-made 3D animation assets to perform immersive demonstrations on the side of the panel or directly above the real object. The animation can show the motion path of the tool, the assembly sequence of the parts, and other microscopic processes that are difficult to understand visually.

[0070] In order to free the hands of the maintenance personnel and improve the operation efficiency, the system builds a complete natural interaction interface to support the control of the entire guide model. The entire interaction system builds a closed loop from intelligent information pushing, multi-modal visualization to natural interaction control, and finally forms a highly intuitive, information-rich and convenient operation immersive maintenance guide model, which greatly reduces the technical threshold and error probability of maintenance work, and directly extends the knowledge and experience of experts to the fingertips and eyes of front-line personnel.

[0071] It should be noted that, in this text, the relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0072] Although the embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A cable defect detection system, characterized in that: The system includes a multi-physics field composite sensing array module, an intelligent signal excitation module, a multi-source data fusion processing engine module, a deep defect diagnosis and decision module, a lightweight mobile detection module, and an AR visualization interaction module, wherein: The multi-physics field composite sensing array module is used to perform multi-physics field coupling scanning on the cable surface based on the reconfigurable probe housing to obtain a multi-dimensional original signal set of the cable; The intelligent signal excitation module is used to call the multi-dimensional original signal set to perform genetic algorithm optimization on the excitation waveform parameters to generate type-adaptive excitation instructions for the cable; The multi-source data fusion processing engine module is used to perform cross-domain feature fusion on the multi-source sensor data using the type of adaptive excitation instruction, and output a three-dimensional probability distribution map of the cable defects; The deep defect diagnosis and decision module is used to analyze the three-dimensional probability distribution map, perform dual-stream Transformer analysis on the defect characteristics, and form a diagnosis and decision report on the cable defect type; The lightweight movement detection module is used to perform path planning and control on the folding robot arm according to the diagnosis decision report, so as to complete the autonomous movement detection trajectory of the cable; The AR visualization interaction module is used to match the autonomous mobile detection trajectory to perform LiDAR-AR calibration matching on the maintenance space and build a visual maintenance guidance model for the cable defect.

2. The cable defect detection system according to claim 1, characterized in that: When the multi-physics field composite sensing array module performs multi-physics field coupling scanning on the cable surface based on the reconfigurable probe housing, the multi-physics field composite sensing array module includes: The reconfigurable probe housing is equipped with an annular slide mechanism to achieve circumferential continuous scanning; The microwave antenna array is driven by a stepper motor to rotate and switch the polarization direction, enhancing the response to defects at different angles; The liquid metal-coupled ultrasonic transducer has a built-in piezoelectric ceramic array to form a directional sound beam focus; The wide-spectrum infrared thermal imager integrates a narrow-band filter set to suppress ambient light interference; The optical axes of the microwave antenna array, ultrasonic transducer, and infrared thermal imager coincide to ensure consistency of spatial coordinates.

3. The cable defect detection system according to claim 2, characterized in that: When performing coupling scanning, the liquid metal coupled ultrasonic transducer includes: The gallium-indium alloy-filled cavity is connected to a micro air pump to achieve dynamic volume adjustment; The air pressure regulating mechanism controls the liquid metal pressure based on the cable curvature radius through real-time feedback; An elastic silicone sealing membrane is set at the bottom of the cavity to prevent liquid metal leakage; Enhanced acoustic impedance matching properties by dispersing nanoparticles in liquid metal; The operating frequency range of the ultrasonic transducer covers multiple resonant frequency bands.

4. The cable defect detection system according to claim 1, characterized in that: When the intelligent signal excitation module calls the multi-dimensional original signal set to perform genetic algorithm optimization on the excitation waveform parameters, it includes: The fitness function calculates the energy density ratio of the defect echo amplitude to the background noise; The iterative optimization process constrains the microwave pulse width to the sub-microsecond to millisecond range; The ultrasonic sweep frequency range is dynamically adjusted according to the dielectric properties of the cable insulation material; The genetic algorithm adopts a tournament selection strategy to retain dominant genes; Outputting the optimized adaptive excitation instruction of the type to the programmable signal generating device.

5. The cable defect detection system according to claim 1, characterized in that: When executing the cross-domain feature fusion of multi-source sensor data using the type adaptive excitation instruction, the multi-source data fusion processing engine module includes: The microwave time domain reflection coefficient is orthogonally demodulated to extract the complex envelope characteristics; Ultrasound echo attenuation slope was calculated by piecewise linear regression; The infrared temperature rise gradient uses a gradient operator to enhance the hot spot boundary; The unit size of the unified 3D grid is correlated and matched with the cable structure scale; The three-dimensional probability distribution diagram marks the defect spatial distribution and morphological parameters.

6. The cable defect detection system according to claim 1, characterized in that: The deep defect diagnosis decision module, when performing a two-stream Transformer analysis of defect features by parsing a three-dimensional probability distribution graph, includes: The first-stream time-frequency domain signal sequence is input into the multi-head self-attention encoding layer; The second stream’s spatial topological features are used to extract neighborhood associations through graph neural networks; The cross attention layer establishes a temporal-spatial feature correlation mapping matrix; The fully connected layer outputs a probability distribution vector of the cable defect type; The diagnostic decision report includes an analysis of the degradation mechanism of the defective material.

7. The cable defect detection system according to claim 1, characterized in that: When the lightweight movement detection module performs path planning control on the folding robot arm according to the diagnosis decision report, it includes: The cable CAD model is imported into a parametric surface geometry; Real-time point clouds are collected by depth vision sensors and registered to the world coordinate system; A heuristic search algorithm generates an optimized inspection path that avoids joints / brackets; The motion trajectory of the robot arm joint adopts a spline curve for smooth transition; The anti-collision monitoring unit interrupts abnormal motion trajectory in real time.

8. The cable defect detection system according to claim 7, characterized in that: The lightweight movement detection module also includes: The annular coil of the electromagnetic induction energy acquisition unit coaxially surrounds the cable conductor; The frequency of the alternating magnetic field is adaptively matched to the fundamental wave component of the power supply network; The rectifier and voltage stabilizing circuit converts the induced current into a stable DC voltage; The distributed energy storage module buffers the power supply current; The dynamic power distribution unit adjusts the power output according to the system load.

9. The cable defect detection system according to claim 1, characterized in that: When the AR visualization interaction module performs LiDAR-AR calibration matching on the maintenance space by matching the autonomous mobile detection trajectory, the AR visualization interaction module includes: The three-dimensional coordinates of the defect are projected to the display coordinate system via the rigid body transformation matrix; The AR coordinate system uses the head-mounted device inertial navigation system as the spatial reference; The spatial anchor is bound to the physical location through feature point cloud registration; The visual rendering parameters of the virtual logo are associated with the physical properties of the defect; The LiDAR point cloud data stream is synchronously coupled with the AR display frame rate.

10. The cable defect detection system according to claim 9, characterized in that: When the AR visualization interaction module constructs a visualization maintenance guidance model, it includes: The defect type database is associated with a failure mode effects analysis knowledge base; The tool list is intelligently filtered and recommended by electrical insulation level; Operation steps are broken down into spatially registered augmented reality animations; The gesture recognition engine verifies the compliance of maintenance actions in real time; The augmented reality overlay guides the integration of a natural language interaction interface.

Images