A cable joint arc multi-physics field simulation method and system

Abstract

The application relates to the technical field of data processing, and discloses a cable joint arc multi-physics field simulation method and system. The method comprises the following steps: a defect-containing geometric model is established; an adaptive time step is obtained by judging a temperature gradient change rate and a current rise rate; a cumulative damage factor is obtained by bidirectional coupling iteration of an arc temperature field and material dielectric parameters based on the time step; a virtual temperature detection point is arranged to convert a simulation temperature field into a virtual monitoring signal; and a fault mode fingerprint spectrum is obtained by projecting temperature characteristic parameters and ultrasonic spectral characteristic parameters into a two-dimensional characteristic space. The application realizes adaptive adjustment of the time step by establishing an arc evolution phase discrimination mechanism, and solves the problems of low calculation efficiency and insufficient transient feature capture caused by a fixed time step.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a multiphysics simulation method and system for electric arc of cable joints. Background Technology

[0002] Existing cable joint arc simulation technology mainly uses the finite element method with a fixed time step to numerically simulate the arc evolution process. After establishing a three-dimensional geometric model of the cable joint and setting defect parameters, the arc temperature field and electric field intensity distribution are solved based on Maxwell's equations and the energy conservation equation. The simulation results are output in the form of internal three-dimensional field quantity data, which are used to analyze the arc development law and insulation degradation trend of the cable joint under different operating conditions. This technology provides a theoretical analysis means for the study of cable joint fault mechanism and condition assessment.

[0003] However, existing simulation technologies have the following shortcomings: First, using a fixed time step for full-process simulation leads to unreasonable allocation of computational resources. During the initial breakdown stage of the arc, the temperature and current changes drastically, requiring a finer time resolution to capture transient characteristics. However, during the stable combustion stage, the field changes are gradual, yet the same time step is still used, resulting in a waste of computational resources. Second, existing methods only simulate the arc generation process under given defect conditions in a one-way manner, without considering that the high temperature and chemical corrosion generated by arc combustion will inversely exacerbate the deterioration of insulation materials. This results in the simulation failing to reflect the dynamic interaction between defect evolution and arc development, leading to insufficient accuracy in predicting the lifespan of long-term service processes. Third, the internal field data such as the three-dimensional temperature field and electric field intensity distribution output by the simulation cannot directly correspond to the external signals collected by the actual monitoring equipment. There is a lack of a mapping mechanism to convert the simulation results into measurable physical quantities, making it difficult to verify the simulation results and unable to guide sensor selection and placement optimization.

[0004] Based on the above shortcomings, it can be inferred that to improve the predictive ability of arc simulation for the early fault evolution process of cable joints, three progressive technical problems need to be solved: First, how to dynamically adjust the time discretization accuracy of the simulation according to the different phase characteristics of arc evolution, so as to reduce the consumption of computing resources while ensuring the accuracy of capturing transient processes. Second, since the thermal, electrical, and chemical damage generated by the arc will change the dielectric parameters and breakdown strength of the insulation material, thus affecting the electric field distribution and arc development in subsequent cycles, it is necessary to establish a two-way feedback mechanism between defect degradation and arc evolution to realize dynamic simulation of multi-cycle cumulative damage. Third, considering that the data gap between simulated field quantities and measured signals restricts the engineering application value of simulation technology, it is necessary to construct a conversion channel from the internal field quantities of the simulation to the external measurable signals, and establish a fault mode recognition system based on multi-dimensional feature fusion to transform the simulation data into a knowledge graph that can be directly used for fault diagnosis. Summary of the Invention

[0005] This application provides a multiphysics simulation method and system for electric arcs in cable joints, which uses an arc evolution phase discrimination mechanism to achieve adaptive adjustment of the time step, thus solving the problems of low computational efficiency and insufficient capture of transient features caused by a fixed time step.

[0006] This application solves the problems of simulation results failing to reflect the long-term cumulative damage evolution process and simulation data being disconnected from measured signals by constructing a two-way coupled iterative model of defect deterioration and arc evolution and establishing a virtual sensor network to realize field signal mapping, thereby improving the accuracy of cable joint fault prediction and the engineering application value of simulation technology.

[0007] Firstly, this application provides a multiphysics simulation method for cable joint arcs, the multiphysics simulation method for cable joint arcs comprising:

[0008] Step S1: Establish a three-dimensional geometric model of the cable joint, and embed defect parameterized feature vectors into the three-dimensional geometric model to obtain a geometric model containing defects;

[0009] Step S2: Perform discriminant analysis on the temperature gradient change rate and current rise rate in the defective geometric model to obtain the adaptive time step;

[0010] Step S3: Based on the adaptive time step, perform bidirectional coupling iteration of the arc temperature field and the material dielectric parameters to obtain the cumulative damage factor;

[0011] Step S4: Arrange virtual temperature detection points on the surface of the outer sheath of the defective geometric model, convert the simulated temperature field into infrared radiation intensity, and obtain a virtual monitoring signal;

[0012] Step S5: Extract temperature feature parameters and ultrasonic spectrum feature parameters from the virtual monitoring signal, and project the temperature feature parameters and ultrasonic spectrum feature parameters into a two-dimensional feature space to obtain a fault mode fingerprint spectrum.

[0013] Secondly, this application provides a multi-physics simulation system for cable joint arcs, the multi-physics simulation system for cable joint arcs comprising:

[0014] An embedding module is used to create a three-dimensional geometric model of a cable joint, and to embed defect parameterized feature vectors into the three-dimensional geometric model to obtain a geometric model containing defects.

[0015] The discrimination module is used to perform discrimination analysis on the temperature gradient change rate and current rise rate in the defective geometric model to obtain an adaptive time step.

[0016] The coupling module is used to perform bidirectional coupling iteration between the arc temperature field and the material dielectric parameters based on the adaptive time step to obtain the cumulative damage factor.

[0017] The conversion module is used to arrange virtual temperature detection points on the surface of the outer sheath of the defective geometric model, convert the simulated temperature field into infrared radiation intensity, and obtain a virtual monitoring signal.

[0018] The projection module is used to extract temperature feature parameters and ultrasonic spectrum feature parameters from the virtual monitoring signal, and project the temperature feature parameters and ultrasonic spectrum feature parameters onto a two-dimensional feature space to obtain a fault mode fingerprint spectrum.

[0019] Thirdly, a multiphysics simulation device for cable joint arc is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the multiphysics simulation device for cable joint arc to execute the above-described multiphysics simulation method for cable joint arc.

[0020] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the above-described multiphysics simulation method for cable joint arcs.

[0021] The technical solution provided in this application obtains a defect-containing geometric model by establishing a three-dimensional geometric model of the cable joint and embedding defect parameterized feature vectors. This overcomes the limitations of traditional idealized modeling. This technical feature uses four-dimensional feature vectors to parameterize the defects, including defect type number, geometric size parameters, spatial location parameters, and dielectric constant parameters. This allows different types of defects, such as air gaps, water trees, and metal spikes, to be expressed using a standardized data structure. By adjusting the feature vector parameters, different defect scenarios can be quickly generated without remodeling, significantly improving simulation modeling efficiency and laying the foundation for subsequent automated batch simulation. This defect-containing geometric model accurately embeds the geometric shape and physical properties of the defects into the five-layer structure of the cable joint, providing accurate boundary conditions and material property distributions for multi-physics coupling solutions. By adjusting the temperature gradient change rate and current rise rate in the defect-containing geometric model... Discriminant analysis yields an adaptive time step, resolving the imbalance between computational accuracy and efficiency in different stages of arc evolution caused by traditional fixed time step methods. This technique captures the dynamic characteristics of arc evolution by real-time monitoring of the second-order time derivative of the temperature field and the first-order time derivative of the current. When both parameters simultaneously exceed a preset breakdown threshold, the arc is determined to have entered the initial breakdown stage, and the time step is automatically reduced to one-tenth. This provides denser time sampling points during the rapid transient process of arc development, accurately recording the drastic changes in temperature and current. In the stable burning stage of the arc, a larger time step is restored to reduce unnecessary computation. This adaptive mechanism ensures high-resolution capture of key transient phenomena such as initial arc breakdown while avoiding the waste of computational resources caused by using a fine time step throughout the process. This allows the simulation to complete the long-span arc evolution process simulation at a reasonable computational cost.

[0022] Based on an adaptive time step, a cumulative damage factor is obtained through bidirectional coupling iteration of the arc temperature field and material dielectric parameters. This innovatively establishes a dynamic feedback mechanism between defect evolution and arc development. This technique overcomes the limitations of traditional unidirectional static simulation. By calculating the sum of thermal, electrical, and chemical damage components, the degree of insulation material degradation is quantified. When the cumulative damage factor exceeds a degradation threshold, the material dielectric parameters are updated to reflect decay. The degraded dielectric constant and breakdown field strength are used as initial conditions for the next cycle, enabling dynamic modeling of the impact of material performance degradation on the electric field distribution. This periodic iteration mechanism realistically reflects the vicious cycle of defects in cable joints during long-term service, leading to partial discharge (PD) aggravation and accelerated PD evolution. This significantly improves the accuracy of predicting joint insulation life. Virtual temperature detection points are arranged on the outer sheath surface containing the defect geometric model, converting the simulated temperature field into infrared radiation intensity to obtain a virtual monitoring signal. This solves the key problem of the disconnect between simulation results and actual monitoring data. This technique utilizes virtual detection points arranged at actual sensor installation locations. Based on the physical laws of thermal radiation, the internal temperature field is converted into an externally measurable infrared radiation intensity, ensuring that the data format of the simulation output is completely consistent with the signal acquired by the infrared thermal imager. This provides a direct comparison benchmark for experimental verification of the simulation results and also provides theoretical guidance for optimizing sensor placement. Temperature and ultrasonic spectral parameters are extracted from the virtual monitoring signal and projected onto a two-dimensional feature space to obtain a fault mode fingerprint spectrum. This constructs a complete transformation chain from simulation data to fault diagnosis knowledge. This technology extracts multidimensional feature parameters from the virtual monitoring signal through statistical analysis and frequency domain analysis. The t-SNE dimensionality reduction algorithm is used to map the high-dimensional feature space onto a two-dimensional plane while maintaining the local neighborhood relationship between samples. This allows different fault modes to form distinguishable clusters on the two-dimensional spectrum. The k-means clustering algorithm is used to divide the two-dimensional space into regions and label the fault type according to the sample labels. The resulting fault mode fingerprint spectrum can be directly used for pattern matching and fault type identification of actual monitoring data, transforming complex multiphysics simulation results into a diagnostic tool that is easy for engineers to understand and apply. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of an embodiment of the multiphysics simulation method for cable joint arc in this application.

[0025] Figure 2This is a schematic diagram illustrating the relationship between the temperature field and the adaptive time step adjustment during the arc evolution process in an embodiment of this application.

[0026] Figure 3 This is a schematic diagram of one embodiment of the multiphysics simulation system for cable joint arcs in this application.

[0027] Figure 4 This is a schematic block diagram of the structure of the multiphysics field simulation device for cable joint arc in an embodiment of the present invention. Detailed Implementation

[0028] This application provides a method and system for simulating the electric arc of a cable joint using a multiphysics field. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data used can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0029] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the multiphysics simulation method for cable joint arcs in this application includes:

[0030] Step S1: Establish a three-dimensional geometric model of the cable joint, embed the defect parameterized feature vector into the three-dimensional geometric model, and obtain a geometric model containing defects;

[0031] Step S2: Perform discriminant analysis on the rate of change of temperature gradient and the rate of rise of current in the defect-containing geometric model to obtain the adaptive time step;

[0032] Step S3: Based on the adaptive time step, the arc temperature field and the material dielectric parameters are coupled and iterated in two directions to obtain the cumulative damage factor;

[0033] Step S4: Arrange virtual temperature detection points on the surface of the outer sheath containing the defective geometric model, convert the simulated temperature field into infrared radiation intensity, and obtain the virtual monitoring signal;

[0034] Step S5: Extract temperature feature parameters and ultrasonic spectrum feature parameters from the virtual monitoring signal, and project the temperature feature parameters and ultrasonic spectrum feature parameters onto the two-dimensional feature space to obtain the fault mode fingerprint spectrum.

[0035] It is understood that the executing entity of this application can be a multiphysics simulation system for cable joint arcs, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiment uses a server as an example for illustration.

[0036] Specifically, defect modeling is achieved by establishing a precise three-dimensional geometric structure for the cable joint. First, a five-layer concentric cylindrical structure consisting of a conductor connecting pipe, a semi-conductive stress control layer, a cross-linked polyethylene main insulation layer, an outer semi-conductive shielding layer, and a silicone rubber outer sheath is constructed as the basic geometric model. Then, a flattened ellipsoid geometric parameter is embedded in the main insulation layer to characterize the air gap defect. The flattened ellipsoid's geometric shape and position are defined by three parameters: major axis radius, minor axis radius, and spatial position coordinates. Next, a four-dimensional feature vector is assigned to the air gap defect model, including a defect type number to distinguish different defect types such as air gaps, water trees, and metal spikes; geometric dimension parameters to record the major and minor axis dimensions of the defect; and spatial position parameters to record... The radial distance, circumferential angle, and axial distance of the defect in the cylindrical coordinate system are recorded. The dielectric constant parameter sets the dielectric properties inside the defect. By spatially registering these parameters with the basic geometric model, the geometric model containing the defect is constructed. For example, the main insulation layer of a cable joint is 8.5 mm thick. An ellipsoidal air gap with a major axis of 2.5 mm and a minor axis of 0.8 mm is embedded at a distance of 5 mm from the conductor surface, at a circumferential angle of 45 degrees, and at an axial distance of 90 mm from the center of the joint. The dielectric constant of the air gap is set to 1.0, which is consistent with the dielectric constant of air, while the dielectric constant of the surrounding cross-linked polyethylene is 2.3. Through this parametric modeling method, the defect is accurately located in three-dimensional space and assigned physical properties.

[0037] The time step is adaptively adjusted by monitoring the dynamic characteristics of arc evolution. First, the time derivative of the arc temperature field in the defect-containing geometric model is obtained to obtain the rate of temperature change. Then, the second time derivative of this rate of temperature change is obtained to obtain the rate of temperature gradient change. At the same time, the time derivative of the current in the conductor is obtained to obtain the current rise rate. These two dynamic parameters are compared with preset breakdown thresholds. When both the rate of temperature gradient change and the rate of current rise exceed the threshold, the system is determined to have entered the initial breakdown stage of the arc. At this time, the reference time step is dynamically reduced according to the discrimination results, reducing the time step from the reference value to one-tenth to improve the time resolution. For example, the reference time step is initially set to 0 in a certain simulation. In 1 millisecond, when the temperature of the arc temperature field at a certain time t is detected to be 800 Kelvin, after 0.1 milliseconds the temperature rises to 1200 Kelvin, the rate of temperature change is 4000 Kelvin per second. Differentiating this rate of change, we find that it has increased from 2000 Kelvin per second in the previous moment to 4000 Kelvin per second, so the rate of temperature gradient change is 20000 Kelvin per second squared. At the same time, the conductor current increases from 500 Amperes to 650 Amperes, with a current increase rate of 1500000 Amperes per second. When both of these parameters exceed their respective preset breakdown thresholds, the system determines that it has entered the initial breakdown stage of the arc and automatically reduces the time step from 0.1 milliseconds to 0.01 milliseconds for more refined time discretization.

[0038] A dynamic interactive modeling of material degradation and arc evolution is achieved through a two-way coupling mechanism. Finite element numerical solutions are performed on the defective geometric model based on an adaptive time step to obtain the arc temperature field distribution and electric field intensity distribution for the Nth power frequency cycle. The thermal damage component is calculated from the temperature field, obtained by integrating the time the temperature exceeds the reference temperature and considering the activation energy effect. The electrical damage component is calculated from the electric field intensity distribution, obtained by integrating the electric field intensity to the higher power of the ratio of the initial breakdown field intensity. Simultaneously, the ozone concentration is calculated from the arc current density, and the ultraviolet radiation intensity is calculated from the arc temperature to obtain the chemical damage component. These three damage components are summed to obtain the cumulative damage factor. When this damage factor exceeds the degradation threshold, material parameter updates are triggered. The dielectric constant of the cross-linked polyethylene main insulation layer is updated by decreasing according to the decay rule related to the damage factor, and the breakdown field intensity is also updated accordingly. The updated material dielectric parameters are used as the initial conditions for the N+1th cycle for a new finite element solution. This process is iterated until the cumulative damage factor exceeds the failure threshold. For example, a cable joint in the N+1th cycle... During the 50-cycle simulation, the peak arc temperature field reached 3500 Kelvin, which lasted for 0.02 seconds, causing the local temperature of the insulation layer to rise from the initial 300 Kelvin to 380 Kelvin. The thermal damage component, calculated by temperature integration, was 0.15. Simultaneously, the electric field strength at the defect tip reached 45 kV / mm, exceeding 1.5 times the initial breakdown field strength. The electrical damage component, calculated by the 9th power integral of the electric field strength, was 0.08. The arc current density of 50,000 A / m² generated an ozone concentration of 15 ppm, combined with strong ultraviolet radiation. The chemical damage component calculated was 0.05. The cumulative damage factor obtained by summing the three components was 0.28, which did not exceed the degradation threshold of 0.3. Therefore, the material parameters remained unchanged. When the iteration continued to the 65th cycle, the cumulative damage factor reached 0.32, which exceeded the degradation threshold. At this time, the dielectric constant was updated from the initial value of 2.3 to 2.23, and the breakdown field strength was updated from the initial value of 30 kV / mm to 29.52 kV / mm. The simulation continued for the 66th cycle using these updated parameters until the cumulative damage factor reached the failure threshold of 0.95 and the iteration was terminated.

[0039] Virtual sensor technology was used to map and convert simulated field quantities to measured signals. On the surface of a silicone rubber outer sheath containing a defective geometric model, virtual temperature detection points were uniformly distributed at three circumferential angles (0°, 120°, and 240°). At each angle, virtual temperature detection points were also arranged at three axial positions (-50 mm, 0 mm, and +50 mm), for a total of nine detection points, and their coordinates were recorded. Surface temperature values ​​at these nine coordinates were extracted from the arc temperature field corresponding to the final cumulative damage factor distribution. These temperature values ​​were arranged in chronological order to form a surface temperature time series. Based on the emissivity and atmospheric transmittance of the silicone rubber material, the surface temperature was converted into infrared radiation intensity. The conversion relationship was: infrared radiation intensity equals material emissivity multiplied by the Stefan-Boltzmann constant multiplied by atmospheric transmittance multiplied by the fourth power of surface temperature. The calculated infrared radiation intensity time series was then converted to a value from 0 to 255 using a linear quantization relationship. The grayscale digital signal within the range is used to obtain the virtual monitoring signal. For example, if the surface temperature of a certain detection point at time t is 355 Kelvin (82 degrees Celsius), the emissivity of the silicone rubber material is 0.92, the Stefan-Boltzmann constant is 5.67 x 10^-8 watts per square meter per Kelvin, and the atmospheric transmittance is 0.85, then the infrared radiation intensity is calculated as 0.92 x 5.67 x 10^-8 x 0.85 x 355^4 equals 697 watts per square meter. The minimum radiation intensity is set to 418 watts per square meter, corresponding to 20 degrees Celsius, and the maximum radiation intensity is set to 1475 watts per square meter, corresponding to 150 degrees Celsius. The grayscale value corresponding to this radiation intensity is 697 minus 418 divided by 1475 minus 418 multiplied by 255, which equals 67. The grayscale values ​​of all nine detection points at each time are combined into a virtual monitoring signal data stream.

[0040] Automatic fault mode identification is achieved through multidimensional feature extraction and dimensionality reduction projection. Statistical analysis is performed on the surface temperature time series in the virtual monitoring signal to extract the maximum temperature value (peak value), the temperature rise rate (slope obtained through linear regression of the temperature time series), the temperature non-uniformity (standard deviation of temperature values ​​at nine detection points), and the temperature fluctuation coefficient (ratio of standard deviation to average temperature). These four parameters constitute the temperature feature parameters. Simultaneously, a Fast Fourier Transform (FFT) is performed on the ultrasonic sound pressure signal in the virtual monitoring signal to convert the time-domain signal into a frequency-domain power spectrum. From the power spectrum, the dominant frequency (frequency corresponding to the maximum power value), the spectral centroid (weighted average of each frequency point multiplied by its power value), the spectral bandwidth (frequency range containing 90% of the energy), and the spectral kurtosis (sharpness of the power spectrum distribution) are extracted. These four parameters constitute the ultrasonic spectral feature parameters. The eight parameters—temperature feature parameters and ultrasonic spectral feature parameters—are combined to form a feature parameter vector, creating a multidimensional feature space. The t-SNE dimensionality reduction algorithm is used to perform nonlinear dimensionality reduction on this multidimensional feature space. This algorithm constructs… The probability distributions in high-dimensional and low-dimensional spaces are constructed and the difference between the two distributions is minimized. The multidimensional feature parameter vector is mapped to a two-dimensional feature space, and then the k-means clustering algorithm is applied to the two-dimensional projected coordinates. This algorithm updates the cluster centers iteratively to minimize the sum of the squared distances from each sample point to its respective cluster center. The location of each cluster center is calculated to obtain the distribution of cluster regions. Based on the fault state labels of the sample data, each cluster region is labeled with a fault type to obtain a fault mode fingerprint spectrum. For example, the temperature feature parameters of a simulation case are a maximum temperature of 85 degrees Celsius, a temperature rise rate of 5 degrees Celsius per hour, a temperature non-uniformity of 8 degrees Celsius, and a temperature fluctuation coefficient of 0.11. The ultrasonic spectrum feature parameters are a main frequency of 40 kHz, a spectral centroid of 42 kHz, a spectral bandwidth of 20 kHz, and a spectral kurtosis of 3.5. After the feature vector composed of these eight parameters is dimensionality reduced by the t-SNE algorithm, the projected coordinates on the two-dimensional plane are abscissa 4.2 and ordinate 2.8. This coordinate falls within cluster region 3. The fault type label corresponding to this region is air gap partial discharge state, thus completing the automatic identification of fault modes.

[0041] In one specific embodiment, step S1 includes:

[0042] A five-layer concentric cylindrical structure consisting of a conductor connecting tube, a semiconductive stress control layer, a cross-linked polyethylene main insulation layer, an outer semiconductive shielding layer, and a silicone rubber outer sheath was established to obtain the basic geometric model of the cable joint.

[0043] An ellipsoidal geometric parameter is embedded in the main insulation layer of the basic geometric model of the cable joint. The ellipsoidal geometric parameter includes the major axis radius, minor axis radius and spatial position coordinates to obtain the air gap defect model.

[0044] The air gap defect model is assigned a defect type number, geometric dimension parameters, spatial location parameters, and dielectric constant parameters to obtain the parameterized feature vector of the defect.

[0045] The defect parameterized feature vector is spatially registered with the basic geometric model of the cable joint to obtain the geometric model containing the defect.

[0046] Specifically, when establishing a five-layer concentric cylindrical structure, the geometric dimensions of each layer are first determined. The conductor connecting tube, as the innermost cylinder, has its outer diameter set to a fixed value. The semi-conductive stress control layer covers the outer surface of the conductor connecting tube, forming the second cylinder. The cross-linked polyethylene main insulation layer, as the third cylinder, has its thickness determining the cable's insulation strength. The outer semi-conductive shielding layer forms the fourth cylinder to homogenize the electric field distribution. The silicone rubber outer sheath, as the outermost cylinder, provides mechanical protection. These five layers are constructed sequentially in 3D modeling software from the inside out, maintaining the concentricity between each layer to obtain the basic geometric model of the cable joint. Subsequently, when embedding the geometric parameters of the flattened ellipsoid in the main insulation layer, it is necessary to define... Three key parameters are defined: the major axis radius determines the length of the oblate spheroid in the main direction, the minor axis radius determines the thickness of the oblate spheroid in the vertical direction, and the spatial position coordinates are represented using a cylindrical coordinate system, including radial distance (distance from the central axis of the cylinder to the center of the oblate spheroid), circumferential angle (position angle of the oblate spheroid in the circumferential direction), and axial distance (position of the oblate spheroid along the axial direction of the cylinder). These geometric parameters are input into the solid model of the main insulation layer, and a Boolean subtraction operation is used to excavate an oblate spheroid cavity inside the insulation layer to obtain an air gap defect model. Then, four types of parameters are assigned to this defect model to form a parametric description. The defect type is numbered using a numerical coding method to distinguish different defects, such as air gap numbered 01, water... The tree is numbered 02, and the metal spike is numbered 03. Geometric parameters record the specific values ​​of the major and minor axes of the oblate spheroid. Spatial position parameters record the coordinate values ​​of radial distance, circumferential angle, and axial distance. The dielectric constant parameter sets the dielectric properties inside the defect; for example, the dielectric constant of an air gap is set to 1.0, indicating that its dielectric properties are close to those of a vacuum. These four types of parameters are arranged in the order of type number - geometric dimension - spatial position - dielectric constant to form a four-dimensional vector, which is the defect parameterized feature vector. Finally, when spatially registering this feature vector with the cable joint's basic geometric model, the precise location of the defect in the basic model is determined by the spatial position parameters in the feature vector. The geometry of the defect is embedded at the location, and the material properties of the region are set according to the dielectric constant parameter. This completes the spatial alignment and property association between the defect and the base model, resulting in a geometric model containing the defect. For example, the outer diameter of the conductor connecting pipe of a 10 kV cable joint is 32 mm, the thickness of the semiconductive stress control layer is 2 mm, so its outer diameter is 36 mm, the thickness of the cross-linked polyethylene main insulation layer is 8.5 mm, so its outer diameter is 53 mm, the thickness of the outer semiconductive shielding layer is 1 mm, so its outer diameter is 55 mm, and the thickness of the silicone rubber outer sheath is 6 mm, so its outer diameter is 67 mm. When embedding the air gap defect in the main insulation layer, the major axis radius of the flattened ellipsoid is set to 2.5 mm and the minor axis radius is 0.The spatial coordinates are set as follows: radial distance 20 mm (center of the main insulation layer), circumferential angle 45 degrees, axial distance 90 mm (90 mm axial distance from the joint center). This air gap defect is assigned the defect type number 01 to indicate the air gap type. The geometric dimensions are recorded as major axis 2.5 mm and minor axis 0.8 mm, spatial position parameters are recorded as 20 mm - 45 degrees - 90 mm, and dielectric constant is set to 1.0. These parameters form a feature vector: 01-2.5-0.8-20-45-90-1.0. Based on the spatial position parameter 20 mm - 45 degrees - 90 mm in this vector, the point is located in the main insulation layer of the basic geometric model. At this point, a flattened ellipsoidal cavity is carved out with a major axis of 2.5 mm and a minor axis of 0.8 mm. The dielectric constant of this cavity is set to 1.0, while the dielectric constant of the surrounding main insulation layer remains at 2.3. After spatial registration, a complete cable joint geometric model including the air gap defect is obtained.

[0047] In one specific embodiment, step S2 includes:

[0048] The temperature gradient rate is obtained by taking the derivative of the arc temperature field in the defective geometric model twice in consecutive time steps.

[0049] The current rise rate is obtained by taking the time derivative of the conductor current in the defective geometric model.

[0050] The rate of change of temperature gradient and the rate of rise of current are compared with preset breakdown thresholds respectively. When both exceed the preset breakdown threshold, it is determined to be the initial breakdown stage of the arc, and the arc evolution phase discrimination result is obtained.

[0051] The reference time step is dynamically reduced based on the arc evolution phase discrimination result, and then reduced to one-tenth to obtain the adaptive time step.

[0052] Specifically, when performing two consecutive time derivatives on the arc temperature field, the temperature field distribution data at each moment is first extracted from the finite element solution results of the defect-containing geometric model. This temperature field data includes temperature values ​​corresponding to the three-dimensional spatial coordinates and time labels. During the first time derivative, the temperature values ​​at two adjacent moments are subtracted and then divided by the time interval to obtain the temperature change rate, which represents the rate of temperature change with time and is measured in Kelvin per second. During the second time derivative, the temperature change rate values ​​at two adjacent moments are subtracted and then divided by the time interval to obtain the temperature gradient rate of change, which represents the rate of change of the temperature itself and is measured in Kelvin per second. The ampere-squared value per second (amperes per second) captures the abrupt changes in temperature. When performing time derivative on the conductor current, the current values ​​flowing through the conductor at each moment are extracted from the simulation results. The current rise rate is obtained by subtracting the current values ​​from two adjacent moments and dividing by the time interval. The current rise rate represents the rate at which the current increases with time, and its unit is amperes per second. This parameter reflects the steep increase in current during arc initiation. When comparing the rate of change of the temperature gradient with a preset breakdown threshold, if the absolute value of the rate of change of the temperature gradient exceeds the threshold, it indicates that the temperature field is undergoing a drastic change. Simultaneously, when comparing the current rise rate with another preset breakdown threshold, if the current rise rate exceeds the threshold, it indicates that the current is increasing rapidly. When both the rate of change of the temperature gradient and the rate of increase of the current exceed their respective thresholds, it is determined that initial arc breakdown has occurred inside the defective geometric model. At this time, the arc channel has just formed, and both temperature and current change drastically in a short period of time. This determination result is recorded as the initial arc breakdown stage. If either of the two parameters does not exceed the threshold, it is determined as a non-breakdown stage, such as the normal operation stage or the stable combustion stage. When adjusting the time step based on the arc evolution phase discrimination result, if the discrimination result is the initial arc breakdown stage, the currently used reference time step value is reduced to one-tenth of the original value. After the time step is reduced, the number of solutions per unit time increases tenfold, thereby enabling more precise capture of the transient changes in the initial arc breakdown stage. In the process, if the determination result is a non-breakdown stage, the baseline time step is kept unchanged to save computational resources. The adjusted time step is used as the adaptive time step for subsequent finite element iterative calculations. For example, in a cable joint simulation, the temperature at the tip of the defect is 800 Kelvin at time t0. After a baseline time step of 0.1 milliseconds, the temperature at the tip of the defect rises to 1200 Kelvin at time t1. The rate of temperature change is equal to 1200 minus 800 divided by 0.1 milliseconds, which gives a change of 4000 Kelvin per second. Continuing at time t2, which is 0.1 milliseconds after t1, the temperature at the tip of the defect rises to 1800 Kelvin. At this time, the rate of temperature change is equal to 1800 minus 1200 divided by 0.One millisecond yields a temperature change of 6000 Kelvin per second. Subtracting the temperature change rate of 4000 Kelvin per second at time t1 from the temperature change rate of 6000 Kelvin per second at time t2 gives an increment of 2000 Kelvin per second. Dividing this increment by the time interval of 0.1 milliseconds gives a temperature gradient rate of change of 20000 Kelvin per squared per second. Simultaneously, the conductor current is 500 Amperes at time t0, rising to 650 Amperes at time t1. The current rise rate is equal to 650 minus 500, divided by 0.1 milliseconds, giving an increase of 1,500,000 Amperes per second. Comparing the temperature gradient rate of change of 20000 Kelvin per squared per second with the preset breakdown threshold of 10000 Kelvin per squared per second reveals that the former exceeds the latter. Furthermore, the electrical... A comparison of the current rise rate of 1,500,000 amperes per second with the preset breakdown threshold of 1,000,000 amperes per second revealed that the former exceeded the latter. Since both parameters exceeded their respective thresholds, it was determined that the arc breakdown stage began at time t2. Based on this determination, the baseline time step was reduced from 0.1 milliseconds to 0.01 milliseconds as the adaptive time step after time t2. In subsequent simulations at this time step, the time interval from t2 to t3 was shortened from 0.1 milliseconds to 0.01 milliseconds. Within this shorter time interval, the temperature rose from 1800 Kelvin to 1900 Kelvin, and the current rose from 650 amperes to 680 amperes, allowing for a more detailed recording of the temperature and current changes during the arc breakdown process.

[0053] Figure 2 This is a schematic diagram illustrating the relationship between the temperature field and adaptive time step adjustment during the arc evolution process in an embodiment of this application. Figure 2 As shown, the horizontal axis represents the simulation time (in milliseconds), the left vertical axis represents the arc temperature field value (in Kelvin), and the right vertical axis represents the temperature gradient rate of change (in Kelvin squared per second). The solid curve represents the evolution trend of the arc temperature field over time, and the dashed curve represents the change in the temperature gradient rate of change. The horizontal dashed line indicates the preset breakdown threshold (10000 K / s²). The shaded area in the figure represents the initial breakdown stage of the arc (50-70 ms). During this stage, when the temperature gradient rate of change exceeds the breakdown threshold, an adaptive time step adjustment mechanism is triggered, reducing the baseline time step from 0.1 ms to 0.01 ms. This achieves precise capture of the rapid transient process of arc development. During the normal operation and stable combustion stages, when the temperature gradient rate of change is below the threshold, the baseline time step remains at 0.1 ms. This figure intuitively demonstrates the technical effect of the adaptive time step adjustment achieved by the arc evolution phase discrimination mechanism in this application, solving the technical problems of low computational efficiency and insufficient capture of transient features caused by a fixed time step.

[0054] In one specific embodiment, step S3 includes:

[0055] The finite element method was used to solve the defective geometric model based on the adaptive time step, and the arc temperature field and electric field intensity distribution of the Nth period were obtained.

[0056] The thermal damage component, electrical damage component, and chemical damage component are calculated based on the arc temperature field and electric field intensity distribution, and the cumulative damage factor is obtained by summing the three components.

[0057] When the cumulative damage factor exceeds the degradation threshold, the dielectric constant and breakdown field strength of the cross-linked polyethylene main insulation layer in the defective geometric model are updated by attenuation to obtain the updated material dielectric parameters.

[0058] The dielectric parameters of the material are used as the initial conditions for the N+1th period to perform a new finite element solution. The process is repeated until the cumulative damage factor exceeds the failure threshold, and the final cumulative damage factor distribution is obtained.

[0059] Specifically, when performing finite element analysis on a defective geometric model based on an adaptive time step, this time step is input into the finite element software as a time discretization parameter of the solver. The solver discretizes the time interval from the start of the current cycle to the end of the cycle according to this time step. At each discrete time point, it solves a coupled set of equations including electromagnetic field equations, fluid dynamics equations, and energy conservation equations. The electromagnetic field equations calculate the electric and magnetic field intensity distributions inside the defective geometric model, the fluid dynamics equations calculate the velocity and pressure distributions of the arc plasma, and the energy conservation equations calculate the temperature field distribution. These equations are coupled together and solved iteratively to obtain the arc temperature field data at the end of the Nth cycle, including three-dimensional space. The temperature values ​​and electric field intensity distribution data at various points include the electric field intensity vectors at each point in three-dimensional space. When calculating the thermal damage component based on the arc temperature field, all spatial grid cells in the temperature field are traversed. The temperature value of each cell is compared with the reference temperature value. If the cell temperature exceeds the reference temperature, the temperature excess is calculated. The temperature excess is divided by the reference temperature to obtain the normalized temperature difference. The normalized temperature difference is multiplied by the length of time the cell has experienced high temperature to obtain the thermal damage contribution value of that cell. The thermal damage contribution values ​​of all cells are summed to obtain the overall thermal damage component. When calculating the electric damage component based on the electric field intensity distribution, all spatial grid cells in the electric field intensity distribution are traversed. The electric field intensity value of each cell is compared with the initial breakdown field intensity value. The ratio of calculated electric field strength to breakdown field strength is compared and then subjected to a high-order power operation, such as a 9th-order power operation, to reflect the nonlinear effect of electric field strength on insulation damage. The result of the power operation is multiplied by the duration of the high electric field experienced by the unit to obtain the electrical damage contribution value of that unit. The electrical damage contribution values ​​of all units are summed to obtain the overall electrical damage component. When calculating the chemical damage component, the ozone concentration generated by the arc is calculated based on the arc temperature field and current density distribution. Ozone concentration has a power relationship with arc current density; the higher the current density, the more ozone is generated. Simultaneously, the ultraviolet radiation intensity is calculated based on the arc temperature; ultraviolet radiation intensity has a fourth-order power relationship with temperature; the higher the temperature, the stronger the radiation. The chemical damage component is calculated by weighted combination of ozone concentration and ultraviolet radiation intensity. The cumulative damage factor is obtained by directly adding the values ​​of thermal damage, electrical damage, and chemical damage components. This factor characterizes the overall damage level of the cross-linked polyethylene main insulation layer at the end of the Nth cycle. The cumulative damage factor is compared with a preset degradation threshold, which represents the critical value at which the material begins to deteriorate. If the cumulative damage factor exceeds the degradation threshold, the material parameter update process is triggered. When updating the dielectric constant of the cross-linked polyethylene main insulation layer, the original dielectric constant is multiplied by an attenuation coefficient less than 1. This attenuation coefficient is calculated based on the difference between the cumulative damage factor and the degradation threshold; the larger the difference, the smaller the attenuation coefficient. If the updated dielectric constant is less than the original dielectric constant, it indicates a decrease in the dielectric properties of the material.Simultaneously, when updating the breakdown field strength, the original breakdown field strength is multiplied by another attenuation coefficient less than 1. This attenuation coefficient is also calculated based on the difference between the cumulative damage factor and the degradation threshold. If the updated breakdown field strength is less than the original breakdown field strength, it indicates a decrease in the insulation strength of the material. The updated dielectric constant and breakdown field strength are assigned as new material properties of the cross-linked polyethylene main insulation layer to the corresponding region of the defective geometric model to obtain the updated material dielectric parameters. These material dielectric parameters are set as the initial conditions for the finite element solution in the (N+1)th cycle. During the solution process in the (N+1)th cycle, the defective geometric model uses these already degraded material parameters for calculation. After obtaining the arc temperature field and electric field intensity distribution in the (N+1)th cycle, the thermal damage component and the electrical damage component are calculated again. The chemical damage component is added to the cumulative damage factor of the previous cycle to obtain a new cumulative damage factor. This new cumulative damage factor is compared with a failure threshold, which represents the damage limit for complete material failure and is usually greater than the degradation threshold. If the cumulative damage factor does not exceed the failure threshold, the process continues into the (N+2)th cycle, repeating the above solution and update process. If the cumulative damage factor exceeds the failure threshold, the iteration cycle terminates. The resulting cumulative damage factor distribution is the final cumulative damage factor distribution, containing the damage level values ​​at each point in three-dimensional space. For example, in the 50th cycle of a cable joint simulation, the adaptive time step is set to 0.01 milliseconds. The finite element solver performs 2000 time step iterations within the 20-millisecond time length of this cycle. After the solution is completed... The peak arc temperature field at the defect tip reached 3500 Kelvin, while the temperature of the surrounding insulation layer was 380 Kelvin. With a reference temperature set at 300 Kelvin, the temperature excess at the defect tip was 3500 minus 300, equaling 3200 Kelvin. The normalized temperature difference was 3200 divided by 300, equaling 10.67. This region experienced high temperatures for 20 milliseconds, and the thermal damage contribution was 10.67 multiplied by 20 milliseconds. Summing the contributions from all high-temperature regions yielded a thermal damage component of 0.15. Simultaneously, the electric field strength distribution showed an electric field strength of 45 kV / mm at the defect tip and an initial breakdown field strength of 30 kV / mm. The ratio of electric field strength to breakdown field strength was 45 divided by 30, equaling 1.5. Powering 1.5 by the 9th power yielded 38. 44. The region experienced a high electric field for 20 milliseconds, and the electrical damage contribution was 38.44 multiplied by 20 milliseconds. Summing all high electric field regions yielded an electrical damage component of 0.08. The ozone concentration generated by an arc current density of 50,000 amperes per square meter was proportional to the 1.2 power of the current density, resulting in an ozone concentration of 15 ppm. The ultraviolet radiation intensity generated by an arc temperature of 3500 Kelvin was proportional to the fourth power of the temperature, resulting in a radiation intensity value. A weighted combination of ozone concentration and ultraviolet radiation intensity yielded a chemical damage component of 0.05. Adding 0.08 to 0.15 gives a cumulative damage factor of 0.28. The degradation threshold was set to 0.3; therefore, 0.28 is less than 0.3, and the material parameters remain unchanged as we enter the 51st cycle.Continuing the iteration to the 65th cycle, the cumulative damage factor reaches 0.32, exceeding the degradation threshold of 0.3, triggering the material parameter update process. The dielectric constant is calculated from the initial value of 2.3 using an attenuation coefficient. The attenuation coefficient is determined by the difference of 0.32 minus 0.3, resulting in a calculated attenuation coefficient of 0.968. The new dielectric constant is 2.3 multiplied by 0.968, equaling 2.226. The breakdown field strength is updated from the initial value of 30 kV / mm using another attenuation coefficient, calculated from the difference of 0.02, resulting in a new breakdown field strength of 0.984. The electric field strength is 30 multiplied by 0.984, which equals 29.52 kV per millimeter. The dielectric constant of 2.226 and the breakdown electric field strength of 29.52 are assigned to the defective geometric model in the 66th cycle. Using these degraded parameters during the 66th cycle solution causes a change in the electric field strength distribution. Iteration continues until the 150th cycle when the cumulative damage factor reaches 0.95, exceeding the failure threshold of 0.95, at which point the iteration terminates. The final cumulative damage factor distribution shows that the damage factor at the defect tip is 0.95, while the damage factor in areas farther from the defect is only 0.1.

[0060] In one specific embodiment, step S4 includes:

[0061] Virtual temperature detection points were arranged at 0°, 120°, and 240° circumferential positions on the surface of the silicone rubber outer sheath containing the defect geometric model, respectively, to obtain the coordinates of the nine detection points.

[0062] The surface temperature values ​​at the coordinates of nine detection points were extracted from the arc temperature field corresponding to the final cumulative damage factor distribution to obtain the surface temperature time series.

[0063] The surface temperature time series was converted into infrared radiation intensity based on the emissivity and atmospheric transmittance of silicone rubber material, thus obtaining the infrared radiation intensity time series.

[0064] The infrared radiation intensity time series is converted into a digital signal within the grayscale range according to a linear quantization relationship to obtain a virtual monitoring signal.

[0065] Specifically, when arranging virtual temperature detection points on the surface of the silicone rubber outer sheath, the cylindrical coordinate system containing the defect geometric model is first determined. A coordinate system is established with the central axis of the cable joint as the z-axis. Three uniformly distributed azimuth angles (0°, 120°, and 240°) are selected on the outer sheath surface. Each azimuth angle corresponds to a radial section. On each radial section, three different positions are selected along the axial direction: 50 mm from the joint center (axial coordinate z = -50 mm), 0 mm from the joint center (axial coordinate z = 0 mm), and 50 mm from the joint center (axial coordinate z = +50 mm). The three azimuth angles are multiplied by three... Nine detection points were obtained along the axial direction. The position coordinates of each detection point were represented by three-dimensional coordinates, including a radial distance r equal to the outer radius of the outer sheath, a circumferential angle θ of 0, 120, or 240 degrees, and an axial distance z of -50 mm, 0 mm, or +50 mm. The position coordinates of these nine detection points were recorded as a coordinate array. When extracting the surface temperature value from the arc temperature field corresponding to the final cumulative damage factor distribution, the temperature field data file was first read. This file contains the temperature values ​​of all grid nodes in three-dimensional space. Based on the position coordinates of the nine detection points, the grid node closest to these coordinates was found in the temperature field data. For each If the coordinates of a probe point perfectly match the coordinates of a grid node, the temperature value of that node is directly extracted. If they do not perfectly match, a three-dimensional spatial interpolation method is used to calculate the interpolated temperature at the probe point based on the temperature values ​​of multiple surrounding grid nodes. The temperature values ​​of the nine probe points at each time step are arranged in chronological order to form nine surface temperature time series. Each time series contains temperature data from the start to the end of the simulation. When converting infrared radiation intensity based on the emissivity and atmospheric transmittance of silicone rubber material, the Stefan-Boltzmann thermal radiation law is used. This law describes the thermal radiation intensity of an object's surface and its surface temperature. The fourth power is directly proportional to the fourth power. During the conversion calculation, each temperature value in the surface temperature time series is raised to the fourth power to obtain the fourth power value. This fourth power value is then multiplied by the emissivity parameter of the silicone rubber material. Emissivity represents the proportion of the material's surface radiation capacity relative to an ideal blackbody; the emissivity of silicone rubber is typically around 0.92. This is then multiplied by the Stefan-Boltzmann constant, which is 5.67 multiplied by 10 to the power of -8 (watts per square meter per Kelvin to the fourth power). Finally, it is multiplied by the atmospheric transmittance parameter. Atmospheric transmittance represents the attenuation ratio of thermal radiation propagating through the air; in the 8-14 micrometer infrared band, the transmittance is approximately 0 at a distance of 1 meter.85. Multiply these parameters sequentially to obtain the infrared radiation intensity value in watts per square meter. Perform the above conversion calculation on each temperature value in the surface temperature time series to obtain the corresponding infrared radiation intensity value. Arrange these radiation intensity values ​​in chronological order to form an infrared radiation intensity time series. When converting the infrared radiation intensity time series into a grayscale digital signal, first determine the quantization range of the infrared radiation intensity. Set the minimum radiation intensity value to correspond to the radiation intensity at an ambient temperature of 20 degrees Celsius (293 Kelvin), and the maximum radiation intensity value to correspond to the radiation intensity at the upper limit of the measurement temperature of 150 degrees Celsius (423 Kelvin). For each radiation intensity value in the infrared radiation intensity time series, subtract the minimum value... The relative radiation intensity is obtained by dividing the relative radiation intensity by the difference between the maximum and minimum radiation intensity. The normalized radiation intensity ranges from 0 to 1. Multiplying the normalized radiation intensity by 255 yields a value within the range of 0 to 255. This value is then rounded to the nearest integer to obtain an integer grayscale value. A grayscale value of 0 corresponds to the lowest radiation intensity, i.e., the lowest temperature, while a grayscale value of 255 corresponds to the highest radiation intensity, i.e., the highest temperature. Each radiation intensity value in the infrared radiation intensity time series is converted into a corresponding grayscale value to obtain a grayscale value time series. This grayscale value time series is the virtual monitoring signal that simulates the digital image signal actually acquired by the infrared thermal imager. For example, a detection point is located at... At the circumferential 0-degree and axial 0-mm position, directly in front of the joint center, the surface temperature at the end of the simulation was 355 Kelvin (82 degrees Celsius). Powering 355 to the fourth power yields approximately 1,589.8 billion Kelvin to the fourth power. Multiplying this by the emissivity of silicone rubber (0.92) gives approximately 1,462.6 billion Kelvin to the fourth power. Multiplying this by the Stefan-Boltzmann constant (5.67) by 10⁻⁸ watts per square meter per Kelvin to the fourth power gives approximately 829 watts per square meter. Multiplying this by the atmospheric transmittance (0.85) gives approximately 705 watts per square meter as the infrared radiation intensity at that point. The minimum radiation intensity corresponding to an ambient temperature of 20 degrees Celsius (293 Kelvin) is calculated. The intensity, calculated similarly, is approximately 418 watts per square meter. The maximum radiation intensity corresponding to the upper limit temperature of 150 degrees Celsius (423 Kelvin) is calculated to be approximately 1475 watts per square meter. Subtracting the minimum value of 418 from the radiation intensity of 705 at this point yields a relative radiation intensity of 287 watts per square meter. Dividing 287 by the maximum value of 1475 minus the minimum value of 418 (i.e., dividing by 1057) gives a normalized radiation intensity of approximately 0.271. Multiplying 0.271 by 255 yields approximately 69.1, which is rounded to the nearest integer, resulting in a grayscale value of 69. The virtual monitoring signal value for this detection point at that moment is 69. The grayscale values ​​for the other eight detection points are calculated using the same procedure to form complete virtual monitoring signal data.

[0066] In one specific embodiment, step S5 includes:

[0067] Statistical analysis was performed on the surface temperature time series in the virtual monitoring signal to extract the maximum temperature value, temperature rise rate, temperature non-uniformity and temperature fluctuation coefficient, and temperature characteristic parameters were obtained.

[0068] Fast Fourier Transform is performed on the ultrasonic sound pressure signal in the virtual monitoring signal to extract the main frequency, spectral centroid, spectral bandwidth and spectral kurtosis, and ultrasonic spectral characteristic parameters are obtained.

[0069] The temperature characteristic parameters and the ultrasonic spectrum characteristic parameters are combined into an eight-dimensional characteristic parameter vector to obtain a multi-dimensional feature space.

[0070] The t-SNE dimensionality reduction algorithm is used to perform dimensionality reduction mapping on the multidimensional feature space, projecting the eight-dimensional feature parameter vector onto the two-dimensional feature space. The k-means clustering algorithm is then used to divide the two-dimensional feature space into regions and label the fault types to obtain the fault mode fingerprint map.

[0071] Specifically, when performing statistical analysis on the surface temperature time series, temperature data from nine detection points are first read from the virtual monitoring signal. The maximum value is found by iterating through the time series of each detection point. The temperature rise rate is obtained by performing linear regression on the temperature time series. Linear regression fits a straight line with time as the independent variable and temperature as the dependent variable. The slope of this line is the temperature rise rate, representing the average increase in temperature per unit time. Temperature non-uniformity is obtained by calculating the standard deviation of the temperature values ​​at the nine detection points at the same time. The standard deviation is calculated by first averaging the nine temperature values, then squaring the difference between each temperature value and the average, summing all the squared differences, dividing by the number of detection points, and then taking the square root. The square root yields the standard deviation value. A larger standard deviation value indicates a greater temperature difference between different locations, i.e., a more uneven temperature distribution. The temperature fluctuation coefficient is obtained by dividing the temperature standard deviation by the temperature average value. This coefficient is a dimensionless parameter used to normalize and represent the degree of temperature fluctuation. The four parameters—maximum temperature value, temperature rise rate, temperature unevenness, and temperature fluctuation coefficient—are combined to form the temperature characteristic parameters. When performing a Fast Fourier Transform (FFT) on the ultrasonic sound pressure signal, the time-domain sound pressure waveform data collected by the ultrasonic sensor is first read from the virtual monitoring signal. This waveform data is a sequence of sound pressure amplitude changes over time. The FFT algorithm converts the time-domain signal into a frequency-domain signal. During the conversion process, each sampling point of the time-domain sequence is multiplied by a complex exponent. The function is used to accumulate the complex amplitudes of each frequency component. Taking the modulus of these complex amplitudes yields the power spectral density for each frequency. The frequency corresponding to the maximum value in the power spectral density is the dominant frequency, representing the frequency component where the sound pressure signal energy is most concentrated. The centroid of the spectrum is obtained by calculating the weighted average of each frequency multiplied by its corresponding power value. Specifically, the product of each frequency value and its power value is multiplied, and the sum of these products is divided by the sum of the power values ​​of all frequencies to obtain the weighted average frequency, which is the centroid of the spectrum. The centroid represents the center of gravity of the spectral energy. When calculating the bandwidth, the power spectral density is first sorted from largest to smallest power value to find the frequency corresponding to 90% of the total power. The frequency range is defined as the upper limit minus the lower limit, which is the bandwidth. Bandwidth represents the width of the frequency distribution of the sound pressure signal. Spectral kurtosis is obtained by dividing the fourth central moment of the power spectrum distribution by the square of the second central moment. The fourth central moment is calculated by subtracting the average power from the power value at each frequency point, then raising the result to the fourth power and averaging the results. The second central moment is the variance of the power spectrum. Kurtosis indicates the sharpness of the power spectrum distribution; a larger value indicates that the spectrum is more concentrated at certain specific frequencies. Combining the four parameters—dominant frequency, centroid, bandwidth, and kurtosis—forms the ultrasonic spectrum characteristic parameters. Arranging the four values ​​of the temperature characteristic parameters and the four values ​​of the ultrasonic spectrum characteristic parameters in sequence forms an array containing eight elements.The array contains eight elements: the first is the maximum temperature, the second is the temperature rise rate, the third is the temperature non-uniformity, the fourth is the temperature fluctuation coefficient, the fifth is the dominant frequency, the sixth is the spectral centroid, the seventh is the spectral bandwidth, and the eighth is the spectral kurtosis. These eight elements constitute an eight-dimensional feature parameter vector, which represents a point in the eight-dimensional space, i.e., a sample point in the multidimensional feature space. When using the t-SNE dimensionality reduction algorithm for dimensionality reduction mapping, the Euclidean distance between each sample point in the multidimensional feature space is first calculated. The Euclidean distance is calculated by subtracting the eight eigenvalues ​​corresponding to two sample points, squaring the results, summing them, and finally taking the square root. Based on the Euclidean distance, a conditional probability distribution in the high-dimensional space is constructed. The conditional probability is represented as... Given a sample point, the probability of another sample point being its neighbor is calculated, with closer proximity increasing the probability and farther distance decreasing the probability. Simultaneously, the projected coordinates of each sample point are randomly initialized in a low-dimensional two-dimensional space. A conditional probability distribution is also constructed in the two-dimensional space based on Euclidean distance. The t-SNE algorithm iteratively optimizes the probability distribution in the two-dimensional space to approximate the probability distribution in the high-dimensional space as closely as possible. The optimization objective is to minimize the KL divergence between the two probability distributions, which represents the degree of difference between them. The KL divergence is gradually reduced by continuously adjusting the coordinates of each sample point in the two-dimensional space using gradient descent. After multiple iterations, the distribution of sample points in the two-dimensional space can maintain the local neighborhood relationships of sample points in the high-dimensional space. In other words, similar sample points in a high-dimensional space will also cluster together in a two-dimensional space. After iteration, the projected coordinates of all sample points in the two-dimensional feature space are obtained. These two-dimensional projected coordinates are input into the k-means clustering algorithm for region partitioning. The k-means algorithm first randomly selects k initial cluster centers, then calculates the Euclidean distance from each sample point to each cluster center, and assigns the sample points to the nearest cluster center to form k clusters. Then, the average value of the coordinates of all sample points in each cluster is recalculated as the new cluster center. The above allocation and update process is repeated until the cluster centers no longer change or the change is less than a threshold. After clustering, k cluster regions are obtained. Each region contains a group of similar sample points. Based on the data carried by the sample data... Fault status labels are used to annotate the fault types in each cluster region. During annotation, the number of samples of each fault type in each cluster region is counted, and the fault type with the most samples is taken as the annotation type for that region. After annotation, a fault mode fingerprint map is formed. This map displays the regional distribution corresponding to different fault types on a two-dimensional plane. For example, in a cable joint simulation case, the virtual monitoring signal shows that the temperatures at nine detection points are 82 degrees Celsius, 85 degrees Celsius, 80 degrees Celsius, 83 degrees Celsius, 88 degrees Celsius, 79 degrees Celsius, 81 degrees Celsius, 84 degrees Celsius, and 86 degrees Celsius, with the maximum value of 88 degrees Celsius being the maximum temperature value. Linear regression fitting of the temperature time series shows that the temperature increases linearly with time at a slope of 5 degrees Celsius per hour, which is the temperature rise rate.The average of nine temperature values ​​is calculated as 82 + 85 + 80 + 83 + 88 + 79 + 81 + 84 + 86, then divided by 9, which equals approximately 83.1 degrees Celsius. The squared difference between each temperature and the average is calculated as follows: 1.21², 3.61², 8.41², 0.01², 24.01², 16.81², 4.41², 0.81², 8.41² (degrees Celsius squared). These are summed, divided by 9, and the square root is taken to obtain the standard deviation of approximately 3.2 degrees Celsius, which is the temperature non-uniformity. Dividing the standard deviation of 3.2 by the average value of 83.1 gives the temperature fluctuation coefficient of approximately 0.038. The time-domain waveform of the ultrasonic sound pressure signal is subjected to a fast Fourier transform to obtain the power spectrum. The power spectrum shows a maximum power peak at 40 kHz, therefore the dominant frequency is 40 kHz. When calculating the centroid of the spectrum, each frequency point in the range of 20 kHz to 80 kHz is multiplied by... The weighted average of the power values ​​yielded a centroid position of approximately 42 kHz. Statistical analysis of the power values ​​from largest to smallest showed that the top 90% of the power was concentrated in the 30-50 kHz range, resulting in a spectral bandwidth of 20 kHz. Calculating the ratio of the fourth-order central moment to the squared variance of the power spectrum yielded a spectral kurtosis of approximately 3.5. These eight characteristic parameters (88°C, 5°C, 3.2°C, 0.038, 40 kHz, 42 kHz, 20 kHz, 3.5) were used to construct a feature parameter vector. After dimensionality reduction using the t-SNE algorithm on the feature parameter vectors of multiple simulation cases, projected coordinates were obtained in a two-dimensional plane. The projected coordinates for this case were 4.2 x and 2.8 y. Applying k-means clustering, the two-dimensional plane was divided into multiple regions. This coordinate point fell within cluster region 3. The fault type for this region was identified as air gap partial discharge, thus completing fault mode identification.

[0072] In one specific embodiment, the t-SNE dimensionality reduction algorithm is used to perform dimensionality reduction mapping on the multidimensional feature space, projecting the eight-dimensional feature parameter vector onto the two-dimensional feature space. The k-means clustering algorithm is then used to divide the two-dimensional feature space into regions and label the fault types, resulting in a fault mode fingerprint map, including:

[0073] Multiple simulations with different parameter combinations were performed on various typical defects to obtain feature parameter vectors and corresponding fault state labels, resulting in a labeled sample dataset.

[0074] The t-SNE dimensionality reduction algorithm is used to perform nonlinear dimensionality reduction mapping on the feature parameter vectors in the labeled sample dataset, preserving the local neighborhood structure and mapping the multidimensional feature space to a two-dimensional feature space to obtain two-dimensional projected coordinates.

[0075] The k-means clustering algorithm is applied to the two-dimensional projected coordinates to perform fault mode clustering, and the location of each cluster center is calculated to obtain the distribution of cluster regions.

[0076] Based on the fault state labels in the labeled sample dataset, the cluster region distribution is labeled with fault type to obtain the fault mode fingerprint map.

[0077] Specifically, when performing multiple parameter combination simulations on various typical defects, the first step is to determine a list of defect types, including typical types such as air gap defects, water tree defects, metal spike defects, and interface delamination defects. Different parameter combinations are set for each defect type. For air gap defects, parameter combinations include different major axis radii (e.g., 1 mm, 2 mm, 3 mm), different minor axis radii (e.g., 0.5 mm, 1 mm, 1.5 mm), and different spatial positions (e.g., radial distance 10 mm, 15 mm, 20 mm, circumferential angle 0 degrees, 90 degrees, 180 degrees, axial distance 50 mm, 100 mm, 150 mm). These parameters are combined using a Cartesian product to generate multiple parameter combination schemes. For each parameter combination scheme, a corresponding geometric model containing the defect is established, and a complete multiphysics simulation is performed. The process includes finite element method (FEM) solution for damage factor calculation, virtual sensor signal extraction, and feature parameter extraction. After each simulation case is completed, an eight-element feature parameter vector is obtained, including maximum temperature, temperature rise rate, temperature non-uniformity, temperature fluctuation coefficient, dominant frequency spectrum, centroid spectrum, bandwidth, and kurtosis. Simultaneously, the feature parameter vector is labeled with a corresponding fault state label, such as normal state, alert state, abnormal state, or severe fault state. The feature parameter vectors and their corresponding labels from multiple simulation cases are stored in a data structure to form a labeled sample dataset. Each sample record in the dataset contains an eight-dimensional feature parameter vector and a fault state label field. When processing the labeled sample dataset using the t-SNE dimensionality reduction algorithm, the data is first read from the dataset... A high-dimensional data matrix is ​​constructed from the feature parameter vectors of the samples. Each row of the matrix represents a sample, and each column represents a feature dimension. The Euclidean distance between any two samples in the high-dimensional space is calculated. This calculation involves subtracting the corresponding elements of the two sample vectors, squaring the result, summing the squares over all eight dimensions, and finally taking the square root. A joint probability distribution is then constructed based on the Euclidean distance matrix. The probability between any two samples in this distribution is calculated using a Gaussian kernel function. The Gaussian kernel function takes the distance between the two samples as input; a smaller distance indicates a higher probability. The probability values ​​of all sample pairs are normalized so that the total probability sums to 1. The projected coordinates of all samples are randomly initialized in the two-dimensional feature space. Initialization typically uses... Two-dimensional coordinate values ​​are generated using a uniform or normal distribution. The Euclidean distance between any two sample projection points is calculated in the two-dimensional space. A joint probability distribution in the low-dimensional space is constructed based on this two-dimensional distance. The low-dimensional probability distribution is calculated using the t-distribution kernel function to enhance sensitivity to distance changes. The objective function is defined as the KL divergence between the high-dimensional and low-dimensional probability distributions. The KL divergence measures the degree of difference between the two probability distributions; a smaller value indicates greater similarity. The two-dimensional projected coordinates are iteratively adjusted using a gradient descent optimization algorithm to minimize the KL divergence. During gradient calculation, the partial derivative of the KL divergence is calculated for each sample's two-dimensional coordinates. The coordinate positions are updated according to the gradient direction; each update moves the coordinates a certain step along the negative gradient direction.During the iteration process, similar sample points in the high-dimensional space gradually cluster together in the two-dimensional space, while sample points that are far apart also maintain a greater distance in the two-dimensional space. After multiple iterations, the KL divergence converges to a local minimum, at which point the stable projected coordinates of all samples in the two-dimensional feature space are obtained. When applying the k-means clustering algorithm to the two-dimensional projected coordinates, first set the number of clusters k, such as setting it to 15, which means dividing the two-dimensional plane into 15 cluster regions. Randomly select k sample coordinates as the initial cluster centers, and calculate the Euclidean distance from each sample point to each cluster center. When calculating the Euclidean distance, subtract the x-coordinate of the sample point from the x-coordinate of the cluster center, square the result, add the square of the subtraction of the y-coordinate of the sample point from the y-coordinate of the cluster center, and then take the square root. Sample points are assigned to the nearest cluster centers to form k clusters. For each cluster, the arithmetic mean of the coordinates of all sample points within it is calculated as the new cluster center. The x-coordinate of the new cluster center is equal to the sum of the x-coordinates of all sample points within the cluster divided by the number of sample points. The y-coordinate is calculated using the same method. Sample points are reassigned and cluster centers are updated using the new cluster centers, iterating this process continuously. The algorithm terminates when the change in cluster center position is less than a preset threshold or the maximum number of iterations is reached after two consecutive iterations. Upon termination, the final coordinates of the k cluster centers and the cluster number to which each sample point belongs are obtained. Based on the cluster numbers, a two-dimensional planar clustering region distribution is formed, with each region corresponding to a cluster containing all sample points of that cluster. When labeling clustered regions based on fault state labels in the labeled sample dataset, each cluster is traversed to count the frequency of various fault state labels within that region. Fault state labels include categories such as normal, attentional, abnormal, and severe. The number of samples for each category label within the region is calculated, and the category with the most frequent labels is identified and used as the fault type label for the entire clustered region. For example, if a clustered region contains 50 sample points, with 35 samples labeled as air gap partial discharge, 10 samples labeled as normal, and 5 samples labeled as attentional, then since air gap partial discharge labels are the most numerous, this region is labeled as the air gap partial discharge type. After labeling all clustered regions, a fault mode fingerprint is obtained. This fingerprint is displayed in a two-dimensional plane. The image displays the boundaries of each cluster region and the corresponding fault type labels. Different fault type regions in the map are distinguished by different colors or symbols. For example, a cable joint fault diagnosis system cumulatively simulated various defect types such as air gap defects, water tree defects, and metal spike defects. For air gap defects, the major axis radii were set to 1 mm, 2 mm, and 3 mm, and the minor axis radii were set to 0.5 mm and 1 mm. Spatial position combinations included radial 10 mm, circumferential 0 degrees, and axial 50 mm, as well as radial 15 mm, circumferential 90 degrees, and axial 100 mm. Each parameter configuration was subjected to a complete simulation to obtain a feature parameter vector, which was labeled as the air gap partial discharge state. Similarly, for water tree defects, different degradation region radii and position parameters were set to perform multiple simulations and labeled as water tree degradation states.Simulations were performed on metal spike defects with different spike length and position parameters, labeled as metal discharge states. After accumulating multiple parameter combinations for various defect types, a labeled dataset containing a large number of samples was obtained. This dataset was then input into the t-SNE algorithm for dimensionality reduction. The t-SNE algorithm calculates the distance between samples in the high-dimensional space and constructs a probability distribution. The distance between a certain air gap defect sample and another air gap defect sample in the eight-dimensional feature space is small, so their probability values ​​are relatively high. However, the distance between the same air gap defect sample and a certain water tree defect sample in the feature space is large, so their probability values ​​are relatively low. After randomly initializing the coordinates of each sample in two-dimensional space, iterative optimization was performed to make the probability distribution in two-dimensional space approximate the probability distribution in high-dimensional space. After multiple iterations, air gap defect samples clustered in a certain region in the two-dimensional plane, and water tree defect samples clustered in a certain region. Samples of different defect types in another region form clearly separated clusters in a two-dimensional plane. A k-means clustering algorithm is applied to the two-dimensional projected coordinates, with a cluster size of 15. The algorithm randomly selects 15 initial cluster centers, calculates the distance from each sample to the cluster center, and assigns the samples. The cluster center positions are iteratively updated until convergence yields 15 cluster regions. One cluster region contains multiple air gap defect samples and a small number of water tree defect samples. The number of air gap partial discharge tags in this region is significantly greater than the number of water tree degradation tags; therefore, this region is labeled as an air gap partial discharge type. After labeling all regions, a fault mode fingerprint map is generated. The map shows that the air gap partial discharge region is located in one quadrant of the two-dimensional plane, the water tree degradation region is located in another quadrant, and the metal discharge region is located in the third quadrant. The spatial distribution of different fault modes is clearly distinguishable on the map.

[0078] The multiphysics simulation method for cable joint arc in the embodiments of this application has been described above. The multiphysics simulation system for cable joint arc in the embodiments of this application is described below. Please refer to [link / reference]. Figure 3 One embodiment of the multiphysics simulation system for cable joint arcs in this application includes:

[0079] An embedding module is used to establish a three-dimensional geometric model of a cable joint, and to embed defect parameterized feature vectors into the three-dimensional geometric model to obtain a geometric model containing defects.

[0080] The discrimination module is used to perform discriminant analysis on the rate of change of temperature gradient and the rate of rise of current in the defective geometric model to obtain an adaptive time step.

[0081] The coupling module is used to perform bidirectional coupling iteration between the arc temperature field and the material dielectric parameters based on the adaptive time step to obtain the cumulative damage factor.

[0082] The conversion module is used to arrange virtual temperature detection points on the surface of the outer sheath of the defective geometric model, convert the simulated temperature field into infrared radiation intensity, and obtain a virtual monitoring signal.

[0083] The projection module is used to extract temperature feature parameters and ultrasonic spectrum feature parameters from the virtual monitoring signal, and project the temperature feature parameters and ultrasonic spectrum feature parameters onto a two-dimensional feature space to obtain a fault mode fingerprint spectrum.

[0084] Specifically, the embedding module is used to establish a three-dimensional geometric model of the cable joint and embed defect parameterized feature vectors into the three-dimensional geometric model to obtain a defect-containing geometric model. This module first calls the three-dimensional modeling engine to construct a five-layer concentric cylindrical structure consisting of a conductor connecting pipe, a semi-conductive stress control layer, a cross-linked polyethylene main insulation layer, an outer semi-conductive shielding layer, and a silicone rubber outer sheath as the basic geometric model of the cable joint. Then, it embeds the geometric parameters of an oblate spheroid into the main insulation layer. The oblate spheroid's geometric shape and spatial position are defined by three parameters: its major axis radius, minor axis radius, and spatial position coordinates. Next, it assigns defect parameterized feature vectors to the oblate spheroid. The parameters include a defect type number to distinguish different defects such as air gaps, water trees, and metal spikes; geometric dimension parameters to record the major and minor axis dimensions of the defect; spatial position parameters to record the radial distance, circumferential angle, and axial distance of the defect in cylindrical coordinates; and dielectric constant parameters to set the dielectric properties inside the defect. The embedding module spatially registers these parameters with the basic geometric model of the cable joint, determines the precise location of the defect in the basic model based on the spatial position parameters, embeds the geometry of the defect into that location, and sets the material properties of that area based on the dielectric constant parameters. This completes the construction of the geometric model containing the defect and outputs it to the discrimination module through the data interface.

[0085] The discrimination module is used to perform discriminant analysis on the rate of change of temperature gradient and the rate of rise of current in the defective geometric model to obtain an adaptive time step. After receiving the defective geometric model output by the embedded module, the module calls the finite element solver to perform a preliminary solution on the model to obtain time series data of arc temperature field and conductor current. The arc temperature field data is first differentiated to calculate the rate of change of temperature, and then the rate of change of temperature is second differentiated to obtain the rate of change of temperature gradient. At the same time, the conductor current data is differentiated to obtain the rate of rise of current. The discrimination module has a built-in comparator unit that compares the rate of change of temperature gradient with a preset breakdown threshold and the rate of rise of current with another preset breakdown threshold. When both parameters exceed their respective thresholds, the comparator outputs a discrimination signal for the initial breakdown stage of the arc. The discrimination module triggers the time step adjustment unit based on this signal to reduce the currently used reference time step to one-tenth as the adaptive time step. If the discrimination result is a non-breakdown stage, the reference time step remains unchanged. The adjusted adaptive time step is transmitted to the coupling module through the data interface as its solution parameter.

[0086] The coupling module is used to obtain the cumulative damage factor by bidirectionally coupling and iterating the arc temperature field and the material dielectric parameters based on an adaptive time step. After receiving the adaptive time step output by the discrimination module, this module sets it as the time discretization parameter of the finite element solver and performs a finite element solution for the Nth period on the defect-containing geometric model. The solution process includes coupled calculations of the electromagnetic field equation, hydrodynamic equation, and energy conservation equation. After the solution is completed, the arc temperature field distribution and electric field intensity distribution data for the Nth period are obtained. The coupling module has a built-in damage calculation unit. This unit calculates the thermal damage component based on the temperature field data, which is obtained by integrating the time the temperature exceeds the reference temperature and considering the activation energy effect. It also calculates the electrical damage component based on the electric field intensity data, which is obtained by integrating the ratio of the electric field intensity to the initial breakdown field intensity, and calculating the electrical damage component based on the arc current density. The chemical damage component is obtained by calculating the ozone concentration and the ultraviolet radiation intensity in combination with the arc temperature. The damage calculation unit sums the three components to obtain the cumulative damage factor. The coupling module has a built-in parameter update unit. When the cumulative damage factor exceeds the degradation threshold, this unit updates the dielectric constant and breakdown field strength of the cross-linked polyethylene main insulation layer by attenuation. The update rule is to multiply the original parameter value by an attenuation coefficient less than 1. The attenuation coefficient is calculated based on the difference between the cumulative damage factor and the degradation threshold. The parameter update unit reassigns the updated material dielectric parameters to the defective geometric model as the initial condition for the finite element solution in the N+1th cycle. The coupling module executes the iterative process of solution-damage calculation-parameter update repeatedly until the cumulative damage factor exceeds the failure threshold and the iteration terminates. Finally, the data containing the final cumulative damage factor distribution and the corresponding arc temperature field is output to the conversion module.

[0087] The conversion module is used to arrange virtual temperature detection points on the surface of the outer sheath containing the defective geometric model, converting the simulated temperature field into infrared radiation intensity to obtain a virtual monitoring signal. After receiving the arc temperature field data output by the coupling module, the module arranges nine virtual temperature detection points on the surface of the silicone rubber outer sheath according to a preset detection point arrangement scheme at three circumferential positions (0°, 120°, 240°) and three axial positions (-50 mm, 0 mm, +50 mm) in each position. The conversion module extracts the surface temperature values ​​at the coordinates of the nine detection points from the temperature field data, arranges the temperature values ​​in chronological order to form a surface temperature time series. The conversion module has a built-in thermal radiation conversion unit, which is based on the emissivity of silicone rubber material and atmospheric permeability. The infrared radiation intensity value is obtained by power-four operations on each temperature value in the surface temperature time series and then multiplying it by the material emissivity Stefan-Boltzmann constant and atmospheric transmittance in turn. This forms an infrared radiation intensity time series. The conversion module has a built-in quantization unit, which subtracts the minimum radiation intensity value from each radiation intensity value in the infrared radiation intensity time series, divides it by the radiation intensity range, multiplies it by 255, and rounds it to obtain a grayscale value in the range of 0 to 255. The grayscale value time series is the virtual monitoring signal, which simulates the digital image signal format actually acquired by the infrared thermal imager. The conversion module transmits the virtual monitoring signal to the projection module through the data interface for feature extraction and pattern recognition.

[0088] The projection module extracts temperature and ultrasonic spectral characteristic parameters from the virtual monitoring signal. It projects these parameters onto a two-dimensional feature space to obtain a fault mode fingerprint. After receiving the virtual monitoring signal from the conversion module, the module calls the statistical analysis unit to process the surface temperature time series. The statistical analysis unit iterates through the time series to find the maximum temperature value, performs linear regression fitting to obtain the temperature rise rate, calculates the standard deviation of the temperature at nine detection points to obtain the temperature non-uniformity, and divides the standard deviation by the average temperature to obtain the temperature fluctuation coefficient. These four parameters constitute the temperature characteristic parameters. The projection module calls the frequency domain analysis unit to perform a fast Fourier transform on the ultrasonic sound pressure signal. The frequency domain analysis unit converts the time-domain sound pressure waveform into a frequency-domain power spectrum, extracting the dominant frequency, centroid, bandwidth, and kurtosis from the power spectrum. These four parameters constitute the ultrasonic spectral characteristic parameters. The projection module combines the temperature and ultrasonic spectral characteristic parameters into an eight-dimensional feature parameter vector to form a multi-dimensional feature space. The projection module incorporates a dimensionality reduction mapping unit, which employs the t-SNE dimensionality reduction algorithm to perform nonlinear dimensionality reduction on the multidimensional feature space. During dimensionality reduction, the Euclidean distance between samples in the high-dimensional space is first calculated, and a probability distribution is constructed. After randomly initializing the sample projection coordinates in the two-dimensional space, iterative optimization is used to approximate the high-dimensional probability distribution within the two-dimensional probability distribution. After iteration, the two-dimensional projection coordinates of all samples are obtained. The projection module also incorporates a clustering labeling unit, which applies the k-means clustering algorithm to the two-dimensional projection coordinates. The algorithm iteratively updates the cluster centers, dividing the two-dimensional plane into multiple cluster regions. The clustering labeling unit counts the number of various fault state labels within each cluster region, using the most frequent label type as the fault type label for that region. After labeling all regions, a fault mode fingerprint map is formed. This map displays the boundaries of each cluster region and the corresponding fault type labels on the two-dimensional plane. The projection module outputs the fault mode fingerprint map as the final diagnostic result of the system, providing engineers with fault mode identification and status assessment.

[0089] above Figure 3 The multiphysics simulation system for cable joint arc in this embodiment of the invention is described in detail from the perspective of modular functional entities. The multiphysics simulation device for cable joint arc in this embodiment of the invention is described in detail from the perspective of hardware processing.

[0090] Reference Figure 4 This invention also provides a multiphysics simulation device for cable joint arcs, which can be a server, and its internal structure can be as follows: Figure 4As shown, the cable joint arc multiphysics simulation device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor in this computer design provides computational and control capabilities. The memory of the cable joint arc multiphysics simulation device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the cable joint arc multiphysics simulation device is used to store the data corresponding to this embodiment. The network interface of the cable joint arc multiphysics simulation device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0091] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the cable joint arc multiphysics simulation device to which the present invention is applied.

[0092] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the multiphysics simulation method for cable joint arc.

[0093] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0094] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a cable joint arc multiphysics simulation device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0095] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multiphysics simulation method for electric arc at cable joints, characterized in that, The method includes: Step S1: Establish a three-dimensional geometric model of the cable joint, and embed defect parameterized feature vectors into the three-dimensional geometric model to obtain a geometric model containing defects; Step S2: Perform discriminant analysis on the temperature gradient change rate and current rise rate in the defective geometric model to obtain the adaptive time step; Step S3: Based on the adaptive time step, perform bidirectional coupling iteration of the arc temperature field and the material dielectric parameters to obtain the cumulative damage factor; Step S4: Arrange virtual temperature detection points on the surface of the outer sheath of the defective geometric model, convert the simulated temperature field into infrared radiation intensity, and obtain a virtual monitoring signal; Step S5: Extract temperature feature parameters and ultrasonic spectrum feature parameters from the virtual monitoring signal, and project the temperature feature parameters and ultrasonic spectrum feature parameters onto a two-dimensional feature space to obtain a fault mode fingerprint spectrum. This includes: performing statistical analysis on the surface temperature time series in the virtual monitoring signal to extract the maximum temperature value, temperature rise rate, temperature non-uniformity, and temperature fluctuation coefficient to obtain the temperature feature parameters; performing a fast Fourier transform on the ultrasonic sound pressure signal in the virtual monitoring signal to extract the dominant frequency, spectral centroid, spectral bandwidth, and spectral kurtosis to obtain the ultrasonic spectrum feature parameters; combining the temperature feature parameters and the ultrasonic spectrum feature parameters into an eight-dimensional feature parameter vector to obtain a multi-dimensional feature space; using the t-SNE dimensionality reduction algorithm to perform dimensionality reduction mapping on the multi-dimensional feature space, projecting the eight-dimensional feature parameter vector onto the two-dimensional feature space, and using the k-means clustering algorithm to divide the two-dimensional feature space into regions and label the fault types to obtain the fault mode fingerprint spectrum.

2. The multiphysics simulation method for cable joint arc according to claim 1, characterized in that, Step S1 includes: A five-layer concentric cylindrical structure consisting of a conductor connecting tube, a semiconductive stress control layer, a cross-linked polyethylene main insulation layer, an outer semiconductive shielding layer, and a silicone rubber outer sheath was established to obtain the basic geometric model of the cable joint. An ellipsoidal geometric parameter is embedded in the main insulation layer of the basic geometric model of the cable joint. The ellipsoidal geometric parameter includes the major axis radius, minor axis radius and spatial position coordinates to obtain the air gap defect model. The air gap defect model is assigned a defect type number, geometric dimension parameters, spatial location parameters, and dielectric constant parameters to obtain the parameterized feature vector of the defect. The defect parameterized feature vector is spatially registered with the basic geometric model of the cable joint to obtain the defect-containing geometric model.

3. The multiphysics simulation method for cable joint arc according to claim 1, characterized in that, Step S2 includes: The temperature gradient rate of change is obtained by taking the derivative of the arc temperature field in the defective geometric model twice in consecutive time steps. The current rise rate is obtained by taking the time derivative of the conductor current in the defective geometric model. The temperature gradient change rate and the current rise rate are compared with a preset breakdown threshold, respectively. When both exceed the preset breakdown threshold, it is determined to be the initial breakdown stage of the arc, and the arc evolution phase discrimination result is obtained. The reference time step is dynamically reduced based on the arc evolution phase discrimination result, and the reference time step is reduced to one-tenth to obtain the adaptive time step.

4. The multiphysics simulation method for cable joint arc according to claim 1, characterized in that, Step S3 includes: The defective geometric model is solved by finite element method based on the adaptive time step to obtain the arc temperature field and electric field intensity distribution in the Nth period. The thermal damage component, electrical damage component, and chemical damage component are calculated based on the arc temperature field and the electric field intensity distribution, and the three are summed to obtain the cumulative damage factor. When the cumulative damage factor exceeds the degradation threshold, the dielectric constant and breakdown field strength of the cross-linked polyethylene main insulation layer in the defective geometric model are updated by attenuation to obtain the updated dielectric parameters of the material. The dielectric parameters of the material are used as the initial conditions for the N+1th period to perform a new finite element solution. This process is repeated until the cumulative damage factor exceeds the failure threshold, resulting in the final cumulative damage factor distribution.

5. The multiphysics simulation method for cable joint arc according to claim 4, characterized in that, Step S4 includes: The virtual temperature detection points are arranged at 0 degrees, 120 degrees and 240 degrees circumferentially on the surface of the silicone rubber outer sheath containing the defective geometric model, respectively, to obtain the position coordinates of nine detection points; The surface temperature values ​​at the coordinates of the nine detection points are extracted from the arc temperature field corresponding to the final cumulative damage factor distribution to obtain the surface temperature time series. The surface temperature time series is converted into infrared radiation intensity based on the emissivity and atmospheric transmittance of silicone rubber material, thus obtaining an infrared radiation intensity time series. The infrared radiation intensity time series is converted into a digital signal within the grayscale range according to a linear quantization relationship to obtain the virtual monitoring signal.

6. The multiphysics simulation method for cable joint arc according to claim 1, characterized in that, The t-SNE dimensionality reduction algorithm is used to perform dimensionality reduction mapping on the multidimensional feature space, projecting the eight-dimensional feature parameter vector onto the two-dimensional feature space. The k-means clustering algorithm is then used to divide the two-dimensional feature space into regions and label the fault types, resulting in the fault mode fingerprint map, including: Multiple simulations with different parameter combinations were performed on various typical defects to obtain the feature parameter vector and the corresponding fault state label, resulting in a labeled sample dataset. The t-SNE dimensionality reduction algorithm is used to perform nonlinear dimensionality reduction mapping on the feature parameter vector in the labeled sample dataset, preserving the local neighborhood structure to map the multidimensional feature space to the two-dimensional feature space, thereby obtaining two-dimensional projected coordinates; The k-means clustering algorithm is applied to the two-dimensional projected coordinates to perform fault mode clustering, and the location of each cluster center is calculated to obtain the distribution of cluster regions. The clustered region distribution is labeled with fault type based on the fault state labels in the labeled sample dataset to obtain the fault mode fingerprint map.

7. A multiphysics simulation system for electric arc at a cable joint, characterized in that, For implementing the multiphysics simulation method for cable joint arcs as described in any one of claims 1 to 6, the multiphysics simulation system for cable joint arcs includes: An embedding module is used to create a three-dimensional geometric model of a cable joint, and to embed defect parameterized feature vectors into the three-dimensional geometric model to obtain a geometric model containing defects. The discrimination module is used to perform discrimination analysis on the temperature gradient change rate and current rise rate in the defective geometric model to obtain an adaptive time step. The coupling module is used to perform bidirectional coupling iteration between the arc temperature field and the material dielectric parameters based on the adaptive time step to obtain the cumulative damage factor. The conversion module is used to arrange virtual temperature detection points on the surface of the outer sheath of the defective geometric model, convert the simulated temperature field into infrared radiation intensity, and obtain a virtual monitoring signal. The projection module is used to extract temperature feature parameters and ultrasonic spectrum feature parameters from the virtual monitoring signal, and project the temperature feature parameters and ultrasonic spectrum feature parameters onto a two-dimensional feature space to obtain a fault mode fingerprint spectrum. This includes: performing statistical analysis on the surface temperature time series in the virtual monitoring signal to extract the maximum temperature value, temperature rise rate, temperature non-uniformity, and temperature fluctuation coefficient to obtain the temperature feature parameters; performing a fast Fourier transform on the ultrasonic sound pressure signal in the virtual monitoring signal to extract the dominant frequency, spectral centroid, spectral bandwidth, and spectral kurtosis to obtain the ultrasonic spectrum feature parameters; combining the temperature feature parameters and the ultrasonic spectrum feature parameters into an eight-dimensional feature parameter vector to obtain a multi-dimensional feature space; using a t-SNE dimensionality reduction algorithm to perform dimensionality reduction mapping on the multi-dimensional feature space, projecting the eight-dimensional feature parameter vector onto the two-dimensional feature space, and using a k-means clustering algorithm to divide the two-dimensional feature space into regions and label the fault types to obtain the fault mode fingerprint spectrum.

8. A multiphysics simulation device for electric arc at a cable joint, characterized in that, It includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the multiphysics simulation method for cable joint arcs as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the multiphysics simulation method for cable joint arcs as described in any one of claims 1 to 6.

Images