Conductive fault detection method for track socket

Abstract

The invention relates to the technical field of electrical equipment fault detection, in particular to a track socket conduction fault detection method, and aims to solve the problems that in the prior art, an electrical sensor such as a thermocouple is prone to electromagnetic interference, temperature data distortion is caused, a fault can be judged only through a fixed threshold value, and the detection accuracy is poor. And early faults with abnormal temperature change trends cannot be identified. According to the invention, accurate detection of conduction faults of the track socket in a strong electromagnetic environment is realized, especially timely identification of early-stage poor contact faults is realized. Through combination of anti-interference data acquisition and hierarchical feature extraction, misjudgment caused by data distortion or single threshold judgment in a traditional scheme is avoided. Meanwhile, different fault types can be distinguished through multi-dimensional feature analysis, and a reliable basis is provided for maintenance decision making.

Description

Technical Field

[0001] This application relates to the field of electrical equipment fault detection technology, and more specifically, to a method for detecting conductive faults in track sockets. Background Technology

[0002] Conductive faults are a prominent issue in the practical use of track sockets. Currently, conductive faults in track sockets mainly manifest as poor contact. Common causes include contact wear, foreign object intrusion into the track (such as oil fumes), and aging of the rubber strips. These problems lead to increased contact resistance, resulting in overheating, and in severe cases, even short circuits and fires. Traditional detection methods primarily rely on regular manual inspections, such as checking for burn marks on the exterior of the track socket or feeling for abnormal heat by touch. However, this method is inefficient and struggles to detect early potential faults. Some temperature-based monitoring methods utilize thermocouples and thermistors, but under high voltage conditions, these components are prone to signal drift, resulting in long response times, low measurement accuracy, and an inability to detect conductive faults promptly and accurately. Furthermore, existing technologies lack intelligent analysis capabilities for temperature change trends, failing to effectively distinguish between normal load fluctuations and early fault characteristics, leading to a high false alarm rate. Simultaneously, traditional detection methods struggle to accurately classify fault types and quantify their severity, hindering the implementation of preventative maintenance.

[0003] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0004] (a) Technical problems to be solved To address the aforementioned issues, this invention proposes a method for detecting conductive faults in track sockets. This method aims to solve the problem that existing electrical sensors, such as thermocouples, are susceptible to electromagnetic interference, leading to distorted temperature data. Consequently, they can only determine faults based on fixed thresholds and cannot identify early faults with abnormal temperature change trends.

[0005] (II) Technical Solution The present invention discloses a method for detecting conductive faults in a track socket, the technical solution of which is as follows: At least one fiber optic sensor is arranged at the conductive part of the track socket; the fiber optic sensor collects real-time dynamic temperature data of the conductive part under strong electromagnetic interference; the dynamic temperature data is transmitted to a fault detection model to obtain fault characteristics; the fault detection model extracts features from the dynamic temperature data based on a hierarchical progressive algorithm, identifying the temperature change trend and rate characteristics; if the fault characteristics match a conductive fault pattern in a preset fault feature library, the corresponding fault type and severity information are output, including early contact failure faults.

[0006] Furthermore, this application also proposes that the fiber optic sensor is a fiber Bragg grating sensor, which senses temperature changes by detecting the drift of the reflected light wavelength, and the drift is linearly related to the temperature change.

[0007] Furthermore, this application also proposes that the fault detection model extracts features from dynamic temperature data based on a hierarchical progressive algorithm, including: establishing a dynamic temperature model based on the temperature baseline data of the track socket under normal working conditions, the dynamic temperature model including temperature fluctuation thresholds under different load conditions; comparing the real-time collected dynamic temperature data with the dynamic temperature model in real time and calculating the deviation value; if the deviation value exceeds the fluctuation threshold and the duration meets the preset duration, it is determined that there is a conductive fault.

[0008] Furthermore, this application proposes that the preset fault feature library be constructed in the following manner: collecting dynamic temperature data of the track socket under conditions of poor contact, short circuit, foreign object intrusion, and aging of the rubber strip; performing cluster analysis on the data to extract the temperature rise rate threshold, abnormal temperature gradient pattern, and duration characteristics corresponding to each fault type; and storing the characteristics as the preset fault feature library.

[0009] Furthermore, this application also proposes that the fiber optic sensor is embedded in the insulating layer of the contact area between the conductive busbar and the conductive sheet, and the distance between the sensor and the conductive component is 0.5-2mm.

[0010] Furthermore, this application also proposes that the fiber optic sensor is encapsulated with a polyimide coating with a coating thickness of 50-100 μm, and that the sensor is fixed to the non-conductive support structure of the track socket by a metal clamp.

[0011] Furthermore, this application also proposes that the temperature dynamic data acquisition frequency is 100-500Hz, and each acquisition includes the temperature value and the acquisition timestamp.

[0012] Furthermore, this application proposes that the identification criteria for early contact failure are: the temperature continuously rises at a rate of 0.1-1℃ / min, and the rise exceeds the baseline value of the dynamic temperature model under the corresponding load by 5-10℃.

[0013] Furthermore, this application also proposes a computing device, including at least one processor and a memory, the memory storing instructions that, when executed by the processor, cause the processor to perform the above-described track socket conductive fault detection method.

[0014] Furthermore, this application also proposes a non-transitory machine-readable storage medium storing executable instructions that, when executed by a processor, implement the aforementioned method for detecting conductive faults in a track socket.

[0015] (III) Beneficial Effects Compared with the prior art, the beneficial effects of the present invention are as follows: In this invention, at least one fiber optic sensor is arranged on the conductive part of the track socket to collect dynamic temperature data in real time. Features are extracted through a hierarchical progressive algorithm and matched with preset fault modes. This enables accurate identification of faults such as poor early contact in strong electromagnetic interference environments. It has the advantages of improving detection accuracy, realizing early fault warning and reducing false alarm rate.

[0016] This application achieves accurate detection of conductive faults in track sockets under strong electromagnetic environments, particularly timely identification of early-stage poor contact faults. By combining anti-interference data acquisition with hierarchical feature extraction, it avoids misjudgments caused by data distortion or single-threshold judgments in traditional solutions. Simultaneously, multi-dimensional feature analysis can distinguish different fault types, providing a reliable basis for maintenance decisions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the logic structure for conductive fault detection. Detailed Implementation

[0019] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following drawings indicate similar items; therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0020] In existing technologies, the detection of conductive faults in rail sockets mainly relies on manual inspection or traditional electrical sensors. Manual inspection involves visual inspection or tactile assessment of heating conditions, but it cannot provide real-time monitoring and is difficult to detect early-stage problems. Electrical sensors, such as thermocouples, are susceptible to electromagnetic interference, leading to distorted temperature data. They can only identify faults based on fixed thresholds and cannot recognize early-stage faults with abnormal temperature change trends. For example, in industrial environments with strong electromagnetic interference, traditional sensors may falsely report or miss faults due to signal deviations, resulting in potential risks not being detected in a timely manner.

[0021] To address the aforementioned issues, the first step is to resolve the distortion in temperature data acquisition under strong electromagnetic environments. Considering the electromagnetic interference resistance of optical fiber materials, its application to temperature monitoring of conductive parts is explored. Secondly, traditional single-threshold judgments cannot distinguish between normal load fluctuations and abnormal heating; therefore, an algorithm capable of analyzing temperature change trends needs to be designed. By studying the temperature change patterns of different fault types, it was found that early contact defects manifest as a slow temperature rise, while short circuits result in a sudden temperature increase. Based on this, a hierarchical progressive algorithm combined with optical fiber sensing is proposed to achieve accurate fault identification through multi-dimensional feature extraction.

[0022] Example 1 Therefore, this application proposes a method for detecting conductive faults in a track socket, comprising the following steps: S100. At least one fiber optic sensor is arranged on the conductive part of the track socket. The fiber optic sensor collects the dynamic temperature data of the conductive part in real time under strong electromagnetic interference environment. S200. The temperature dynamic data is transmitted to the fault detection model to obtain fault characteristics. The fault detection model extracts features from the temperature dynamic data based on the hierarchical progressive algorithm and identifies the temperature change trend and rate characteristics. S300. If the fault characteristics match the conductive fault modes in the preset fault characteristic library, the corresponding fault type and severity information will be output. The fault types include early contact failure faults.

[0023] Among them, fiber optic sensors refer to devices that use optical fiber materials to sense temperature changes. Specifically, they can be implemented using fiber Bragg grating sensors. They reflect temperature changes by the amount of wavelength shift of reflected light, and their anti-electromagnetic interference characteristics can ensure the stability of data acquisition.

[0024] The hierarchical progressive algorithm refers to an analysis method that processes temperature data in stages. Specifically, it can adopt a three-layer structure of basic feature screening, key feature extraction, and feature verification to gradually eliminate interference and extract temperature change trends related to faults. In this embodiment, a machine learning algorithm can be used.

[0025] Specifically, the basic feature screening stage adopts a moving average filtering algorithm, with the filtering window size set to 5 sampling points. The window duration is 50ms when the sampling frequency is 100Hz and 10ms when the sampling frequency is 500Hz. By calculating the mean of the continuously collected temperature dynamic data, the interference caused by the ambient temperature fluctuation, such as instantaneous fluctuations within ±0.1℃, is eliminated.

[0026] In the key feature extraction stage, the least squares method is used to linearly fit the filtered temperature data to calculate the slope of the temperature change trend, i.e. the rate of change. At the same time, a 1-minute sliding time window is set to update the average rate within the window in real time to avoid misjudgment caused by a single abnormal data point. In the feature verification stage, dynamic threshold comparison logic is used to compare the real-time extracted temperature change rate and absolute temperature value with the rate threshold and amplitude threshold under the corresponding load output by the dynamic temperature model. If any parameter exceeds the threshold for 3 consecutive times with an interval of 100ms each time, the fault feature matching stage is entered.

[0027] Fault feature matching employs the Euclidean distance algorithm. Based on the three-dimensional features of temperature rise rate, temperature gradient, and duration of each fault type in the preset fault feature library, the Euclidean distance between the real-time fault feature and the benchmark feature is calculated. When the distance value is ≤0.5, it is considered a match. If there is partial matching of multiple features, such as rate matching but small gradient deviation, the load change trend is further combined. That is, the real-time current is collected by the ACS712 current sensor connected in series to determine the load change. The current fluctuation is ≤10% as the load is stable. If the load is stable, it is determined to be the corresponding fault type. If the load fluctuates, it re-enters the feature extraction stage.

[0028] The preset fault feature library refers to a database that stores the temperature characteristics of different fault types. Specifically, it can be constructed by clustering analysis of historical fault data and used to match abnormal patterns in real-time data.

[0029] Specifically, after the fiber optic sensor is placed on the conductive part, it collects real-time dynamic temperature data including timestamps. After the data is transmitted to the fault detection model, the hierarchical progressive algorithm first performs basic feature screening to remove interfering data such as ambient temperature fluctuations; then it extracts key features such as the rate of temperature change and duration; finally, it compares and verifies the features with thresholds under normal load. If the extracted features match the early contact failure patterns in the preset fault feature library, the corresponding fault type and severity are output.

[0030] Traditional electrical sensors are prone to signal drift in strong electromagnetic environments, leading to temperature measurement errors, while fiber optic sensors can eliminate such interference. Existing solutions rely on a single threshold, failing to identify early faults characterized by slow temperature increases. A hierarchical, progressive algorithm, by analyzing trends and rates of change, can distinguish between normal load fluctuations and abnormal heating. Furthermore, traditional methods can only qualitatively determine the presence or absence of a fault, while this solution utilizes a pre-defined fault feature library to achieve parallel identification of multiple fault types and quantitative assessment of their severity.

[0031] This application further proposes that the fiber optic sensor is a fiber Bragg grating sensor, which senses temperature changes by detecting the drift of the reflected light wavelength, and the drift is linearly related to the temperature change.

[0032] Fiber Bragg grating (FBG) sensors are optical sensors that utilize fiber optic gratings to reflect light of specific wavelengths. Specifically, this can be achieved by using ultraviolet lasers to form a periodic refractive index modulation structure in the fiber core. The reflected wavelength drifts with changes in external temperature, and temperature sensing is achieved by detecting this wavelength drift. The wavelength drift refers to the offset of the reflected center wavelength due to temperature changes, which can be monitored in real time using a spectrometer or demodulator. The drift directly reflects the temperature change. A linear relationship means that the drift and temperature change are directly proportional. This proportionality coefficient can be determined through calibration experiments, establishing a mathematical conversion model between wavelength and temperature changes to ensure the quantifiability and repeatability of the temperature detection results.

[0033] Specifically, fiber Bragg grating sensors sense temperature by utilizing the reflection characteristics of incident light at the grating structure; the reflected wavelength shifts linearly with temperature changes. Since the sensing signal is optical, it does not rely on electrical signal transmission, avoiding measurement errors caused by interference with electrical signals in strong electromagnetic environments. The linear relationship between the drift and temperature change allows temperature data to be directly obtained from the wavelength change, eliminating the need for complex data correction algorithms and simplifying the data processing flow.

[0034] Meanwhile, the rapid transmission characteristics of optical signals enable the sensor to capture dynamic temperature changes in real time, providing timely data support for early fault identification.

[0035] Traditional temperature sensors rely on electrical signal conversion and transmission, making them susceptible to interference from induced currents or electromagnetic radiation in strong electromagnetic environments, leading to signal deviation or distortion. For example, when thermocouples or thermistors operate near conductive parts of rail sockets, electromagnetic interference introduces additional noise, reducing measurement accuracy. In contrast, fiber Bragg grating sensors employ an all-optical signal sensing mechanism, fundamentally avoiding the effects of electromagnetic interference and ensuring stable temperature detection.

[0036] In addition, existing optical sensors, such as infrared temperature measurement devices, are susceptible to interference from environmental factors and cannot be used for contact measurement, while fiber Bragg grating sensors can be directly embedded near conductive components to achieve high-precision contact monitoring.

[0037] The linear correspondence calibration of the fiber Bragg grating (FBG) sensor is carried out as follows: The calibration environment is controlled at a temperature of -10℃ to 120℃, a humidity of 40% RH±5%, and an electromagnetic interference intensity of 100V / m. Temperature control is achieved using a constant temperature chamber, humidity control is achieved by adjusting the humidity controller built into the constant temperature chamber, and electromagnetic interference intensity is controlled by using an EMC electromagnetic interference generator to simulate the strong electromagnetic environment of the track socket. The calibration steps are as follows: Place the FBG sensor in a constant temperature chamber and gradually increase the temperature in 1℃ increments, holding each temperature for 5 minutes to ensure temperature stability. Collect the reflected light wavelength drift data corresponding to each temperature using a demodulator. Collect 10 sets of data for each temperature point and take the average. The demodulator parameters are: wavelength detection range 1520nm~1580nm, detection accuracy ±0.1pm. Finally, through linear regression analysis, the linear proportionality coefficient between wavelength drift and temperature change is obtained as 10pm / ℃, with an allowable error of ±0.5pm / ℃. This coefficient is then written into the temperature conversion module of the fault detection model to achieve direct conversion of wavelength drift to temperature value.

[0038] This application solves the problems of signal offset and response delay of traditional temperature sensors in strong electromagnetic environments, and achieves high-precision, anti-interference detection of temperature changes in conductive parts of track sockets. By combining optical signal transmission with linear correspondence, the reliability and real-time performance of temperature data are ensured, providing an effective data foundation for the early identification of conductive faults.

[0039] This application further proposes to establish a dynamic temperature model based on the temperature baseline data of the track socket under normal working conditions. The dynamic temperature model includes temperature fluctuation thresholds under different load conditions. The real-time collected dynamic temperature data is compared with the dynamic temperature model in real time to calculate the deviation value. If the deviation value exceeds the fluctuation threshold and the duration meets the preset duration, it is determined that there is a conductive fault.

[0040] The dynamic temperature model refers to a mathematical relationship model established by collecting historical temperature data of the track socket under various load conditions. Specifically, machine learning algorithms can be used to fit the temperature fluctuation range under normal operating conditions to reflect the dynamic correlation between different loads and temperature changes. The fluctuation threshold refers to the allowable range of temperature changes set according to different load levels. Specifically, statistical analysis methods can be used to calculate the standard deviation range of temperature fluctuations under each load to distinguish between normal temperature fluctuations and abnormal fault signals.

[0041] The deviation value refers to the difference between the real-time temperature data and the predicted value of the dynamic temperature model. It can be calculated using the mean square error algorithm to quantify the degree of temperature anomaly. The preset duration refers to the lower limit of the duration of the temperature anomaly that triggers fault judgment. It can be set using a sliding time window mechanism to eliminate misjudgments caused by transient interference.

[0042] Specifically, under normal operating conditions, temperature data of the track socket is collected at different load levels. Cluster analysis is used to extract temperature fluctuation characteristics corresponding to each load, constructing a dynamic model that includes the load-temperature mapping relationship. During real-time monitoring, the corresponding temperature fluctuation threshold is applied based on the current load level, and the real-time temperature data is compared with the model's predicted values ​​to calculate the temperature deviation.

[0043] When the deviation exceeds the fluctuation threshold corresponding to the current load, a timer is started to record the duration of the abnormality. If the duration reaches the preset duration, a fault determination is triggered.

[0044] Traditional detection methods, which use fixed temperature thresholds or single load models, cannot distinguish between normal temperature fluctuations caused by load changes and true fault signals, resulting in a high false alarm rate in dynamic load scenarios. This solution establishes a load-adaptive dynamic temperature model and combines deviation value and duration as dual criteria to effectively identify abnormal temperature change patterns during load fluctuations.

[0045] The load division adopts a current-power conversion method. Real-time current is collected by a current sensor, and the power is calculated in combination with the rated voltage of 220V of the track socket. It is divided into three levels: light load, medium load, and heavy load. Light load is less than 300W, medium load is 300W~800W, and heavy load is greater than 800W. The fluctuation threshold is calculated based on the normal temperature baseline data using the 3σ principle. First, the normal temperature data of the track socket is collected continuously for 24 hours under each load level, and the sampling frequency is 200Hz.

[0046] The mean temperature μ and standard deviation σ of each load level are calculated. The fluctuation threshold for the light load level is set to μ±2℃, corresponding to 3σ≈2℃; for the medium load level, it is set to μ±3℃, corresponding to 3σ≈3℃; and for the heavy load level, it is set to μ±5℃, corresponding to 3σ≈5℃. The preset duration is dynamically adjusted according to the load level. The light load level is set to 2 seconds due to the gradual temperature fluctuation, the medium load level is set to 1.5 seconds, and the heavy load level is set to 1 second due to the large temperature fluctuation. All of these settings have been verified through 100 sets of normal load fluctuation experiments to ensure that the duration can effectively eliminate the misjudgment caused by instantaneous fluctuations, such as the brief temperature rise when the load suddenly increases from 200W to 280W.

[0047] This application solves the problem of misjudgment caused by instantaneous temperature fluctuations under dynamic load conditions, and achieves accurate identification of conductive faults. By dynamically adjusting the temperature fluctuation threshold, false alarms are avoided in scenarios with sudden load changes; by introducing a duration determination condition, the influence of instantaneous interference signals is eliminated, thereby maintaining detection accuracy and reliability under complex operating conditions.

[0048] This application further proposes to construct a preset fault feature library in the following way: collecting dynamic temperature data of the track socket under conditions of poor contact, short circuit, foreign object intrusion, and aging of the rubber strip; performing cluster analysis on the data to extract the temperature rise rate threshold, abnormal temperature gradient pattern, and duration characteristics corresponding to each fault type; and storing the features as a preset fault feature library.

[0049] Among them, dynamic temperature data refers to temperature sequence information that changes continuously over time, which can be acquired using fiber optic sensors at a frequency of 100-500Hz. This data can reflect the complete trajectory of temperature changes during the development of a fault. Cluster analysis refers to classifying multidimensional temperature data into patterns using unsupervised learning algorithms, specifically the K-means algorithm, which automatically distinguishes the temperature change patterns of different fault types.

[0050] The temperature rise rate threshold refers to the critical value of temperature change per unit time. It can be determined by calculating the derivative distribution of each fault sample in the dataset and is used to quantify the gradual temperature rise characteristics caused by poor contact. The temperature gradient anomaly mode refers to the temperature difference characteristics between different areas of a conductive component. It can be calculated using a spatial temperature distribution matrix and is used to detect local hot spots caused by foreign object intrusion.

[0051] Duration characteristic refers to the duration of an abnormal temperature state, which can be achieved through time series segmented statistics and is used to identify continuous temperature rise caused by rubber strip aging.

[0052] Specifically, under poor contact conditions, the conductive parts experience a slow temperature rise due to increased contact resistance, with the temperature dynamics showing a linear upward trend of 0.1-1℃ / min; under short-circuit conditions, the sudden increase in current causes the temperature to rise rapidly to over 80℃ within 0.1 seconds; when foreign objects intrude, the contact point between the conductive sheet and the foreign object forms a localized high temperature, while the temperature of the surrounding area remains normal, creating a significant gradient difference; aging of the adhesive strip leads to a decline in insulation performance, with the temperature accumulating continuously at a rate of 0.2℃ per month.

[0053] Furthermore, during the sample collection phase, for poor contact faults, two gradients were set for contact wear depths of 0.1mm and 0.2mm, and two levels of contact pressure of 5N and 10N, with 10 sets of samples collected for each combination; for foreign object intrusion faults, two types of foreign objects were selected: metal debris and oil fumes, which intruded into the gap between conductive sheets and the busbar interface, respectively. The intrusion amount of metal debris was 0.1g and 0.5g, and the intrusion time of oil fumes was 1 hour and 2 hours, with 800 sets of samples collected for each condition; for short circuit faults, a short circuit circuit was artificially constructed to simulate the short circuit, and complete data from the moment of short circuit to temperature stabilization was collected, with a total of 2000 sets collected.

[0054] The aging failure of the rubber strip was determined through accelerated aging tests. The test parameters were 100 hours, 200 hours, and 300 hours at 85℃ and 85% RH, corresponding to actual aging degrees of 10%, 20%, and 30%, respectively. Fifteen samples were collected for each aging degree. All samples were continuously collected for 10 minutes at a sampling frequency of 200Hz, including temperature values ​​and timestamps.

[0055] Cluster analysis uses the K-means algorithm, and the sum of squares within clusters corresponding to different K values ​​is calculated using the elbow rule. When K=4, the decreasing trend of the sum of squares slows down significantly, and K=4 is determined to correspond to four types of faults. The input feature dimensions are the temperature rise rate, temperature gradient (i.e., the temperature difference between the center and the edge of the conductive sheet), and duration. Euclidean distance is used as a similarity measure during the clustering process. Finally, the clustering accuracy is verified to be ≥95% through the confusion matrix.

[0056] The quantitative characteristics of each fault in the preset fault feature library are as follows: poor contact fault: temperature rise rate 0.1-1℃ / min, no obvious temperature gradient (i.e., temperature change ≤0.5℃), duration ≥5min; short circuit fault: temperature rise rate ≥10℃ / s, temperature gradient ≤1℃ (i.e., the trend is a sudden overall increase), duration ≤1s, and the temperature ≥80℃ after stabilization; foreign object intrusion fault: temperature rise rate 2-5℃ / min, local temperature gradient ≥5℃ (i.e., the temperature difference between the intrusion point and the surrounding area), duration ≥2min; rubber strip aging fault: temperature rise rate ≤0.1℃ / min (approximately 0.2℃ per month), temperature gradient ≤0.3℃, duration ≥30 days, and continuous data collection.

[0057] By collecting dynamic temperature data for the four types of faults mentioned above, a raw dataset containing time series, spatial distribution, and rate of change is formed. A clustering algorithm is used to group the multidimensional data and automatically identify the core characteristic parameters of each fault type: poor contact corresponds to low-rate, long-duration characteristics; short circuit corresponds to high-rate, short-duration characteristics; foreign object intrusion corresponds to high-gradient fluctuation characteristics; and aging of the rubber strip corresponds to low-rate accumulation characteristics. Finally, the extracted characteristic parameters are stored in a preset fault feature library as a benchmark template for real-time fault matching.

[0058] Traditional methods rely on manual experience to summarize fault characteristics, which suffers from subjective judgment errors and incomplete feature coverage. For example, operators may misjudge localized temperature rises caused by foreign object intrusion as poor contact, or fail to identify early, slow temperature rises due to rubber strip aging. Existing detection systems based on a single temperature threshold struggle to distinguish between similar temperature rise phenomena caused by short circuits and foreign object intrusion.

[0059] This solution automatically uncovers the essential differences in multidimensional temperature data through cluster analysis, establishes a standardized feature library, solves the subjectivity problem of human experience, and overcomes the limitations of single feature matching.

[0060] This application enables precise differentiation of four types of faults: poor contact, short circuit, foreign object intrusion, and aging of the rubber strip. When an abnormal temperature is detected, the system performs multi-dimensional matching between real-time data and rate thresholds, gradient patterns, and durations in the feature library to accurately determine the fault type. For example, a temperature rise rate of 0.5℃ / min matches the poor contact characteristic, and a local 5℃ gradient difference matches the foreign object intrusion characteristic.

[0061] This feature library supports dynamic expansion. When adding new fault types, only data collection and re-clustering are required; no modification to the detection algorithm is needed. Simultaneously, the severity parameters associated with the feature library can output maintenance priorities when identifying fault types. For example, a rate of 1℃ / min corresponds to moderate contact failure, triggering an early warning.

[0062] This application further proposes embedding an optical fiber sensor within the insulation layer of the contact area between the conductive busbar and the conductive sheet of the track socket, and maintaining a certain distance between the sensor and the conductive component.

[0063] Among them, embedding the fiber optic sensor in the insulating layer means placing the sensor inside the insulating material near the conductive contact surface. Specifically, this can be achieved by using a pre-embedding process to install the sensor simultaneously during the forming of the insulating layer, so that the sensor is directly close to the heat source.

[0064] Specifically, the insulation layer uses epoxy resin E-51 material and employs an injection molding pre-embedding process. A sensor positioning groove is pre-set within the insulation layer injection mold. The fiber optic sensor is placed within the positioning groove before injection molding, ensuring the sensor is embedded within the insulation layer. The 0.5-2mm distance between the sensor and the conductive component is controlled by the mold spacing between the positioning groove and the conductive component. In light-load scenarios, the distance is set to 1.5-2mm to reduce electromagnetic interference, while in heavy-load scenarios, the distance is set to 0.5-1mm to improve heat conduction efficiency. After installation, a metallographic microscope is used to measure the distance, ensuring an error ≤0.1mm. The distance between the sensor and the conductive component is limited to a specific range, which can be achieved by adjusting the installation position using a precision positioning device. This distance range is determined based on the electromagnetic interference attenuation law and the heat conduction efficiency.

[0065] Specifically, the contact area between the conductive busbar and the conductive sheet will generate a local temperature rise during a fault. By embedding the fiber optic sensor in the insulation layer of this area, the heat change of the contact surface can be directly captured, avoiding the signal delay caused by the excessively long heat conduction path of traditional external sensors.

[0066] Maintaining a specific distance between the sensor and conductive components ensures that the electromagnetic interference intensity is below the sensor's tolerance threshold while placing the sensor in the region with the highest thermal conductivity. The thermal resistance properties of the insulating layer material allow heat to be quickly transferred to the sensor, while simultaneously isolating the conductive components from the high-voltage environment, thus preventing signal distortion.

[0067] Traditional detection methods typically attach temperature sensors to the track housing or locations far from the contact area, requiring indirect temperature sensing through multi-layered structures. This results in signal lag and susceptibility to environmental interference. This solution eliminates intermediate conduction by embedding the sensor within the insulating layer of the contact area, enabling direct heat source monitoring. By limiting the installation distance, it simultaneously meets the requirements for interference resistance and detection sensitivity within confined spaces, resolving the measurement error problem caused by the limited installation location of traditional sensors.

[0068] This application can capture minute temperature fluctuations on conductive contact surfaces in real time, avoiding signal distortion caused by installation misalignment or conduction path loss, and significantly improving the accuracy of early contact failure identification. The sensor's embedded insulating layer design also avoids the impact of external dust contamination or mechanical vibration on detection stability, ensuring long-term monitoring reliability.

[0069] This application further proposes that the fiber optic sensor is encapsulated with a polyimide coating, and that the sensor is fixed to the non-conductive support structure of the track socket by a metal clamp.

[0070] Polyimide coating encapsulation refers to the surface coating of fiber optic sensors with polyimide material. Specifically, it can be achieved by using automated coating equipment to achieve uniform coverage. This material has a higher thermal conductivity than traditional encapsulation materials, which can accelerate temperature transfer and form an insulating barrier.

[0071] Metal clamp fixing refers to using a ring-shaped metal clamp to secure the sensor to the installation position. Specifically, a standardized snap-fit ​​structure can be used to achieve rapid assembly. This method can avoid the risk of falling off caused by adhesive fixing and is adaptable to vibration environments.

[0072] Specifically, the polyimide coating, with its low thermal resistance, rapidly transfers temperature changes from conductive parts to the fiber optic sensor, while its anti-aging properties ensure the encapsulation layer maintains structural integrity over long periods in high-temperature environments. The metal clamps, working in conjunction with the non-conductive support structure, eliminate electromagnetic interference affecting the sensor signal and maintain relative positional stability between the sensor and the detection area through a rigid connection. The coating thickness is determined by balancing temperature conduction efficiency and mechanical strength, ensuring lossless heat transfer while preventing cracking of the encapsulation layer. The standardized installation method of the metal clamps adapts to industrial production processes, improving assembly efficiency and ensuring consistency across batches of products.

[0073] The imide coating is performed using a dip-coating machine at a speed of 5 mm / s. After coating, it is baked in a constant temperature oven at 150℃ for 30 minutes to cure, ensuring a coating thickness of 50-100 μm. The metal clamp is made of H62 brass and is formed by stamping. It has an inner diameter of 0.3 mm, which is suitable for sensors with a diameter of 0.225-0.325 mm, a width of 2 mm, and a thickness of 0.1 mm after coating. The non-conductive support structure is made of alumina ceramic material. The metal clamp is fastened to the pre-set mounting holes of the support structure with M2 screws to ensure that the sensor is not loose.

[0074] Traditional solutions often use epoxy resin coatings and adhesive bonding, which suffer from problems such as high thermal resistance leading to delayed temperature detection and adhesive aging causing sensor detachment. This solution replaces epoxy resin with polyimide, significantly reducing thermal resistance and improving temperature resistance. Furthermore, the metal clamp fixing method offers higher mechanical stability compared to adhesive bonding, enabling it to withstand long-term vibration conditions in the track socket.

[0075] This application effectively solves the signal distortion problem of fiber optic sensors in strong electromagnetic environments, ensuring the real-time performance and accuracy of temperature data acquisition. The combined design of coating encapsulation and metal clamp fixation enables reliable installation of the sensor near conductive components, avoiding detection errors caused by electromagnetic interference or mechanical vibration. This solution also meets the requirements of industrial production for assembly efficiency and product consistency, making it suitable for large-scale applications in track sockets.

[0076] This application further proposes that the temperature dynamic data acquisition frequency is 100-500Hz, and each acquisition includes the temperature value and the acquisition timestamp.

[0077] The sampling frequency refers to the number of times the temperature of the conductive part is sampled per unit time. This can be achieved by using a timer-triggered analog-to-digital converter module. By setting the sampling period to 2-10 milliseconds, the data acquisition interval is matched to the response speed of the fiber optic sensor, ensuring that the transient characteristics of temperature changes are fully captured. The timestamp records the precise time information of each temperature acquisition action. This can be achieved by using a real-time clock chip that stores the temperature data synchronously. By appending the hour, minute, second, and millisecond timecodes of the acquisition moment to the temperature value, a time reference is provided for multi-sensor data alignment and load correlation analysis.

[0078] Specifically, when poor contact or short circuit occurs in the conductive parts of the track socket, the temperature change exhibits different rate characteristics. When the sampling frequency is set to 100Hz, temperature data is acquired every 10 milliseconds, capturing the complete curve of a 20°C temperature surge within 0.1 seconds during a short circuit fault. When the sampling frequency is increased to 500Hz, data is acquired every 2 milliseconds, identifying the minute temperature rise trend of 0.1°C per minute caused by poor contact.

[0079] The timestamp recording function ensures strict time alignment of data from multiple fiber optic sensors deployed in different conductive areas, preventing misjudgments of temperature change sequence due to sampling time discrepancies. Simultaneously, the timestamp information is correlated with the load status recording data of the track socket, distinguishing between temperature change patterns caused by normal load fluctuations and abnormal faults.

[0080] Traditional temperature detection solutions often employ fixed low-frequency acquisition or adaptive frequency adjustment mechanisms. Fixed low-frequency acquisition cannot cover the rapid temperature changes required for track socket faults. For example, using a 10Hz acquisition frequency, a temperature surge within 0.1 seconds can only record one data point, resulting in the loss of fault characteristics. Adaptive frequency solutions rely on preprocessing algorithms to determine the rate of temperature change, and are prone to misjudgments due to signal noise in environments with strong electromagnetic interference, causing frequency switching delays. This solution, by limiting a fixed frequency range of 100-500Hz, can cover the temperature acquisition requirements of different fault types without complex algorithms. Combined with accurate timestamp recording, it overcomes the technical obstacles of multi-sensor data synchronization and load correlation analysis.

[0081] This application achieves real-time and accurate capture of temperature changes caused by early conductive faults, solving the problem of missed fault detection due to insufficient response speed of traditional detection elements. The timestamp recording function provides a benchmark for multi-sensor data alignment, avoiding misjudgments of faults caused by acquisition time deviations. It also supports correlation analysis between temperature changes and load status, effectively distinguishing between normal operating conditions and abnormal faults. The setting of the acquisition frequency range balances data integrity and system resource consumption, ensuring that temperature change characteristics at different rates are fully recorded, providing a reliable data foundation for fault mode identification.

[0082] This application further proposes that the identification criteria for early contact failure is that the temperature rises continuously at a specific rate and the rise exceeds the range of the baseline value under the corresponding load of the dynamic temperature model.

[0083] The term "temperature rising at a specific rate" refers to a unidirectional, stable increase in temperature throughout the load cycle, without significant decline. This characteristic is achieved by real-time monitoring of temperature gradient changes, used to capture the initial characteristic of gradually increasing contact resistance. The term "temperature rise exceeding the baseline value corresponding to the load in the dynamic temperature model" refers to setting a differential amplitude threshold based on the reference temperature value corresponding to the current load. This characteristic is achieved by dynamically matching the temperature baseline under load conditions, used to eliminate normal temperature fluctuations under different operating conditions.

[0084] Specifically, when the temperature of the conductive part shows a stable unidirectional upward trend without a significant drop during the load cycle, the system identifies this as a potential characteristic of early contact failure. At this point, the temperature rise is further compared to the dynamic baseline value corresponding to the current load. If it exceeds a preset range, a fault determination is triggered. This dual determination mechanism effectively distinguishes between normal load fluctuations and early fault characteristics through the synergistic effect of trend stability verification and load adaptation amplitude verification. For example, under light load conditions, the temperature baseline value is low, and a small amplitude deviation can trigger an early warning; under heavy load conditions, the temperature baseline value is high, requiring a larger amplitude deviation to determine a fault, thus avoiding misjudgments caused by normal fluctuations under high load.

[0085] Traditional detection methods rely on fixed temperature thresholds or single-rate standards, which cannot adapt to temperature variations under different load conditions and are prone to false positives or false negatives due to load fluctuations or environmental interference. This solution introduces a load-adaptive dynamic amplitude threshold and trend stability verification to construct a multi-dimensional judgment logic, significantly improving the accuracy of early contact failure identification.

[0086] This application can accurately identify the initial temperature rise characteristics of poor contact in track sockets, avoiding misjudgments caused by load fluctuations or changes in ambient temperature. At the same time, it enables differentiated detection of early faults under different load conditions, effectively reducing the risk of missed detection. Example

[0087] This application further proposes a computing device, including at least one processor and a memory, the memory storing instructions that, when executed by the processor, cause the processor to perform a method for detecting conductive faults in a rail socket.

[0088] The processor, or processor, is the core hardware unit that performs computation and control functions. It can be implemented using a multi-core CPU or an embedded microcontroller, and is used to process high-frequency dynamic temperature data collected by fiber optic sensors in real time, solving the problem of delayed response in traditional manual inspections. The memory, or memory, is the physical medium that stores program instructions and data. It can be implemented using flash memory chips or solid-state drives, and is used to solidify the hierarchical progressive algorithm of the fault detection model and the dynamic temperature model comparison logic, ensuring standardized execution of the detection process and avoiding human error.

[0089] Specifically, the dynamic temperature data collected by the fiber optic sensor is transmitted to the processor via a communication interface. The processor executes data parsing, feature extraction, and fault matching operations according to the instruction sequence stored in its memory. A hierarchical progressive algorithm processes the temperature data through multi-level filtering and trend analysis modules. The dynamic temperature model comparison module calculates the deviation between the real-time data and preset thresholds. When an abnormal temperature change rate or gradient is detected, the fault type identification logic is triggered. The processor outputs the identification results to a display module or a remote monitoring system, completing the real-time detection and early warning of conductive faults.

[0090] In some specific implementations, the processor may integrate a digital signal processing unit to accelerate the Fourier transform operation of temperature data, and the memory may be divided into a read-only memory area and a random access memory area, which are used to store the solidified algorithm code and the temporary cache of real-time data, respectively.

[0091] Traditional manual inspections rely on periodic visual checks, which cannot detect millisecond-level temperature anomalies. Furthermore, thermocouple-based detection equipment is susceptible to signal interference in strong electromagnetic environments. This solution achieves physical isolation between data processing and algorithm execution through the hardware architecture of the computing device. Leveraging the electromagnetic insulation properties of fiber optic sensors and the real-time computing power of the processor, it maintains detection accuracy while achieving millisecond-level response, effectively identifying early contact failures.

[0092] This application realizes the automated operation of conductive fault detection in track sockets. Through the deep integration of hardware equipment and detection algorithms, it solves the problems of temperature data acquisition distortion in strong electromagnetic environments and low efficiency of manual inspection. It can accurately identify abnormal temperature trends in the early stages of contact wear or rubber strip aging, thus avoiding safety accidents.

[0093] Example 2

[0094] This application further proposes a non-transitory machine-readable storage medium storing executable instructions. When executed by a processor, the instructions implement a method for detecting conductive faults in a track socket, comprising: arranging at least one fiber optic sensor on the conductive part of the track socket; the fiber optic sensor acquiring real-time dynamic temperature data of the conductive part under strong electromagnetic interference; transmitting the dynamic temperature data to a fault detection model to obtain fault features; the fault detection model extracting features from the dynamic temperature data based on a hierarchical progressive algorithm to identify temperature change trends and rate characteristics; and if the fault features match conductive fault patterns in a preset fault feature library, outputting corresponding fault type and severity information, wherein the fault type includes early contact failure faults.

[0095] Among them, non-transitory machine-readable storage media refers to physical storage carriers that can still retain data after power failure. This can be implemented using solid-state drives (SSDs) or flash memory chips. The embedded detection algorithm ensures that the complete detection process can be restored after a system restart. Executable instructions are program code that can be parsed and executed by a computer processor. This can be stored as a compiled binary file and used to drive fiber optic sensors to collect data and call fault detection models to complete analysis. The hierarchical progressive algorithm is a feature extraction method that processes data in stages. Specifically, it can employ multi-layered computational logic, first extracting temperature change trends and then calculating rate features, to achieve progressive analysis from macroscopic to microscopic levels. The preset fault feature library is a data set storing known fault modes. This can be implemented using a relational database or key-value pair storage structure, and real-time data is compared with historical fault features using pattern matching rules.

[0096] Specifically, after fiber optic sensors are deployed on the conductive parts of the track socket, dynamic temperature data is transmitted to the processor via a photoelectric conversion module. Executable instructions from the storage medium are loaded into memory, triggering a data acquisition thread to acquire temperature values ​​and timestamps at a fixed frequency. Upon receiving the real-time data stream, the fault detection model first calculates the deviation between the current temperature and the dynamic temperature model using a baseline comparison module, and then extracts the temperature rise rate feature using a trend analysis module. When continuous fluctuations exceeding the threshold are detected, the feature matching module retrieves the temperature gradient parameters of poor contact patterns from the fault feature library, and generates a fault diagnosis result by combining this with the duration judgment condition. All processing is completed through a pre-set instruction sequence in the storage medium, requiring no manual intervention.

[0097] Traditional detection methods rely on manual inspection or electronic sensors, which suffer from weak anti-interference capabilities and slow response speeds. This solution, however, embeds the detection logic into a non-volatile storage medium, allowing fiber optic sensor data to be directly acquired and analyzed under program control, eliminating the efficiency bottleneck caused by manual operation. Simultaneously, the hierarchical progressive algorithm achieves multi-level feature extraction through software instructions, enabling earlier identification of abnormal temperature trends compared to a single threshold judgment mechanism, avoiding misjudgments caused by signal drift in traditional sensors.

[0098] This application automates the entire process of detecting conductive faults in rail sockets, ensuring stable temperature data acquisition even in strong electromagnetic environments. By programmatically executing a hierarchical, progressive algorithm, it effectively identifies subtle temperature change characteristics of early-stage contact failures. Utilizing non-volatile storage media to save the detection program avoids algorithm interruptions caused by system power outages, thus improving the standardization and repeatability of the detection process.

[0099] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for detecting conductive faults in a track socket, characterized in that, include: At least one fiber optic sensor is arranged on the conductive part of the track socket, and the fiber optic sensor collects the dynamic temperature data of the conductive part in real time under strong electromagnetic interference environment. The dynamic temperature data is transmitted to the fault detection model to obtain fault characteristics; The fault detection model extracts features from the dynamic temperature data based on a hierarchical progressive algorithm to identify temperature change trends and rate characteristics. If the fault feature matches a conductive fault pattern in a preset fault feature library, the corresponding fault type and severity information are output. The fault type includes early contact failure.

2. The method for detecting conductive faults in a track socket according to claim 1, characterized in that, The fiber optic sensor is a fiber Bragg grating sensor, which senses temperature changes by detecting the drift of the reflected light wavelength, and the drift is linearly related to the temperature change.

3. The method for detecting conductive faults in a track socket according to claim 1, characterized in that, The fault detection model extracts features from the dynamic temperature data based on a hierarchical progressive algorithm, including: establishing a dynamic temperature model based on the temperature baseline data of the track socket under normal working conditions, wherein the dynamic temperature model includes temperature fluctuation thresholds under different load conditions. The real-time collected dynamic temperature data is compared with the dynamic temperature model in real time, and the deviation value is calculated. If the deviation value exceeds the fluctuation threshold and the duration meets the preset duration, a conductive fault is determined to exist.

4. The method for detecting conductive faults in a track socket according to claim 3, characterized in that, The preset fault feature library is constructed in the following way: Collect dynamic temperature data of the track socket under conditions of poor contact, short circuit, foreign object intrusion, and aging of the rubber strip; Cluster analysis was performed on the data to extract the temperature rise rate threshold, temperature gradient anomaly pattern, and duration characteristics corresponding to each fault type. The features are stored in a preset fault feature library.

5. The method for detecting conductive faults in a track socket according to claim 1, characterized in that, The fiber optic sensor is embedded in the insulating layer of the contact area between the conductive busbar and the conductive sheet, and the distance between it and the conductive component is 0.5-2mm.

6. The method for detecting conductive faults in a track socket according to claim 1, characterized in that, The fiber optic sensor is encapsulated with a polyimide coating with a thickness of 50-100 μm, and the sensor is fixed to the non-conductive support structure of the track socket by a metal clamp.

7. The method for detecting conductive faults in a track socket according to claim 1, characterized in that, The temperature dynamic data is collected at a frequency of 100-500Hz, and each collection includes the temperature value and the collection timestamp.

8. The method for detecting conductive faults in a track socket according to claim 1, characterized in that, The criteria for identifying early contact failures are: the temperature rises continuously at a rate of 0.1-1℃ / min, and the increase exceeds the baseline value of the dynamic temperature model under the corresponding load by 5-10℃.

9. The method for detecting conductive faults in a track socket according to claim 1, characterized in that, A computing device includes at least one processor and a memory, the memory storing instructions that, when executed by the processor, cause the processor to perform the track socket conductive fault detection method as described in any one of claims 1-8.

10. A non-transitory machine-readable storage medium having executable instructions stored thereon, which, when executed by a processor, implement the track socket conductive fault detection method as described in any one of claims 1-8.

Images