Fault determination method and device, storage medium and electronic equipment

Abstract

The invention discloses a fault determination method and device, a storage medium and electronic equipment, and relates to the field of communication technology and signal processing. The method comprises the steps that a target image of a target cable is received, and the target image is used for representing the relation between the amplitude of an echo signal in the target cable and the propagation distance; inputting the target image into a target model, extracting target features of the target image according to priori knowledge learned by the target model in a model training stage, and predicting initial fault information corresponding to the target image based on the target features, the target features being used for representing distribution features of signals in the target image, and the initial fault information being used for predicting initial fault information corresponding to the target image; the initial fault information comprises a plurality of fault positions, and the target model is a model obtained by updating an initial model according to model parameters; and determining a target fault position in the target cable according to the initial fault information. The technical problem of low fault point positioning accuracy in a complex environment is solved.

Description

Technical Field

[0001] The present application relates to the fields of communication technology and signal processing, and in particular to a fault determination method, device, storage medium, and electronic device. Background Art

[0002] In the maintenance and monitoring of modern industrial facilities, accurate location of leaky cable faults is crucial for ensuring communication security and system stability. However, existing leaky cable fault detection technologies face the challenge of low fault location accuracy in complex environments, such as tunnels, subways, and large factories. These complex environments are often accompanied by high background noise, electromagnetic interference, multipath effects, and signal attenuation due to long-distance transmission, making it difficult to clearly identify the fault signal characteristics in leaky cables. In particular, when processing such signals, traditional time-domain reflectometers and frequency-domain reflectometers rely on specific signal characteristics, such as the height and shape of the reflection peak. These characteristics become blurred by signal distortion in complex environments, significantly compromising the accuracy and reliability of fault location.

[0003] Furthermore, manual analysis of fault distance spectra is not only inefficient but also subject to the operator's experience, making it difficult to ensure consistent and repeatable detection. This reliance on manual experience is no longer sufficient for large-scale, real-time monitoring, especially given the increasing demand for industrial monitoring.

[0004] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0005] The present application provides a fault determination method, device, storage medium and electronic device to at least solve the technical problem of low fault point location accuracy in complex environments.

[0006] According to one aspect of the present application, a fault determination method is provided, comprising: receiving a target image of a target cable, wherein the target image is used to characterize the relationship between the amplitude and propagation distance of an echo signal in the target cable; inputting the target image into a target model, extracting target features of the target image based on prior knowledge learned by the target model in a model training phase, and predicting initial fault information corresponding to the target image based on the target features, wherein the target features are used to characterize the distribution characteristics of the signal in the target image, the initial fault information includes multiple fault locations, and the target model is a model obtained by updating the initial model according to model parameters, wherein the model parameters are a combination of parameters that meet preset conditions, wherein the preset conditions are used to constrain the error value between the predicted position and the actual position of the target model to be less than a preset threshold; and determining the target fault location in the target cable according to the initial fault information.

[0007] Optionally, before receiving the target image of the target cable, the method also includes: receiving an echo signal in the target cable; mixing the echo signal with a preset signal to obtain a target signal, wherein the preset signal is used to characterize a periodically changing electrical signal; determining an amplitude spectrum of the target signal, and determining the target image based on the amplitude spectrum of the target signal.

[0008] Optionally, the target model is obtained by the following steps: obtaining N historical images, where N is an integer greater than 1, and each historical image includes the actual fault location marked; using a target method to expand the N historical images to M historical images, wherein the target method is used to simulate changes in the actual environment, and generate M historical images by performing target operations on the N historical images, wherein the target operation at least includes adding noise to the historical images; performing multiple iterative training and verification operations on the initial model based on the M historical images to obtain the target model, wherein the multiple iterative training is used to determine the model parameters of the model, and the verification operation is used to verify the performance parameters of the model.

[0009] Optionally, the initial model is iteratively trained and verified for multiple times based on M historical images to obtain a target model, including: initializing the initial model using preset model parameters to obtain a first model; dividing the M historical images into a model training set and a model verification set; inputting the model training set into the first model for iterative training until the number of model iterations is greater than a preset number or the training error of the model is less than a preset threshold, to obtain model parameters of the first model; updating the first model to a second model based on the model parameters; and verifying the second model based on the model verification set to obtain a target model.

[0010] Optionally, each iterative training includes the following steps: when there are T training sets in the model training set, input the T training sets into the first model, and determine the S fault locations predicted by the first model for the T training sets through forward propagation, wherein T is an integer greater than or equal to 1 and less than or equal to M, and S is an integer greater than or equal to T; determine the error value between the fault location corresponding to each training set and the actual location corresponding to each training set, and obtain S error values; based on the S error values, update the model parameters of the first model through the first algorithm and back propagation, wherein back propagation is used to determine the influence of the model parameters of the first model on the model error, and the first algorithm is used to adjust the model parameters of the first model based on the influence.

[0011] Optionally, determining a target fault location in a target cable based on the initial fault information includes: screening multiple fault locations in the initial fault information according to a second algorithm to obtain a target fault location in the target cable, wherein the second algorithm is used to compare the confidences between adjacent fault locations in the multiple fault locations, and filter out fault locations whose confidences are less than a preset value and any one of two fault locations with the same confidences.

[0012] Optionally, the method further includes: generating early warning information when it is detected that the initial fault information is empty, or when it is detected that the range of the fault location in the initial fault information does not conform to a preset range, wherein the preset range is used to constrain the distance of the fault location.

[0013] According to another aspect of the present application, a fault determination device is also provided, including: a receiving unit, configured to receive a target image of a target cable, wherein the target image is used to characterize the relationship between the amplitude and propagation distance of an echo signal in the target cable; a prediction unit, configured to input the target image into a target model, extract target features of the target image based on prior knowledge learned by the target model in a model training phase, and predict initial fault information corresponding to the target image based on the target features, wherein the target features are used to characterize the distribution features of the signal in the target image, the initial fault information includes multiple fault locations, and the target model is a model obtained by updating the initial model according to model parameters, wherein the model parameters are a combination of parameters that meet preset conditions, wherein the preset conditions are used to constrain the error value between the predicted position and the actual position of the target model to be less than a preset threshold; a determination unit, configured to determine the target fault location in the target cable according to the initial fault information.

[0014] According to another aspect of the present application, a computer-readable storage medium is provided, which includes a stored executable program, wherein when the executable program runs, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned fault determination method.

[0015] According to another aspect of the present application, an electronic device is also provided, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the above-mentioned fault determination method.

[0016] In the present application, a target image of a target cable is first received, wherein the target image is used to characterize the relationship between the amplitude and propagation distance of the echo signal in the target cable. The target image is then input into a target model. Based on the prior knowledge learned by the target model during the model training phase, target features of the target image are extracted, and initial fault information corresponding to the target image is predicted based on the target features. The target features are used to characterize the distribution characteristics of the signal in the target image. The initial fault information includes multiple fault locations. The target model is a model obtained by updating the initial model according to model parameters, wherein the model parameters are parameter combinations that meet preset conditions, wherein the preset conditions are used to constrain the error value between the predicted position and the actual position of the target model to be less than a preset threshold. The target fault location in the target cable is then determined based on the initial fault information. That is, through the automatic extraction and intelligent analysis of the target image features by the model, the purpose of automated and high-precision fault point location is achieved, thereby achieving the technical effect of improving the accuracy and efficiency of fault detection and location in complex environments, thereby solving the technical problem of low fault point location accuracy in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 This is a flow chart of an optional fault determination method according to an embodiment of the present application. Figure 1 ;

[0019] Figure 2 is a schematic diagram of an optional fault determination method according to an embodiment of the present application;

[0020] Figure 3 This is a flow chart of an optional fault determination method according to an embodiment of the present application. Figure 2 ;

[0021] Figure 4 is a schematic diagram of an optional fault determination device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0024] It should be noted that the collected information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation portals for users to choose to authorize or refuse. For example, an interface is set up between this system and relevant users or institutions to provide users with corresponding operation portals for users to choose to agree or refuse the automated decision-making results; if the user chooses to refuse, the expert decision-making process will be entered.

[0025] According to an embodiment of the present application, a method embodiment of a fault determination method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0026] It should be noted that a fault detection system can be used as the execution subject of the fault determination method of the embodiment of the present application. It is understandable that the fault determination method provided in the embodiment of the present application can also be executed by other systems or devices, and the embodiment of the present application does not specifically limit this.

[0027] Figure 1 This is a flow chart of an optional fault determination method according to an embodiment of the present application. Figure 1 ,like Figure 1 As shown, the method includes the following steps:

[0028] Step S101: receiving a target image of a target cable.

[0029] In step S101 , the target image is used to characterize the relationship between the amplitude and propagation distance of the echo signal in the target cable.

[0030] Optionally, the target cable refers to a specific leaky cable that is undergoing fault detection, and may be any cable that requires such detection, such as a leaky coaxial cable used in application scenarios such as communication base stations and rail transit.

[0031] Optionally, the target image (fault distance image) is derived from FMCW (Frequency Modulated Continuous Wave) signal processing, including mixing and fast Fourier transform steps, which convert the signal into a visual fault distance image. This image characterizes how the echo signal amplitude changes with propagation distance. The horizontal axis of the target image represents distance, and the vertical axis represents the corresponding signal amplitude.

[0032] Optionally, the fault detection system is responsible for collecting the necessary signal data and converting it into a form that can be processed by the subsequent model (target model), namely the fault distance image. In this way, the analysis of electrical signals can be transformed into the problem of identifying image features, facilitating subsequent automatic feature extraction and fault location.

[0033] In step S102 , the target image is input into the target model, target features of the target image are extracted based on the prior knowledge learned by the target model during the model training phase, and initial fault information corresponding to the target image is predicted based on the target features.

[0034] In step S102, the target features are used to characterize the distribution features of the signals in the target image. The initial fault information includes multiple fault locations. The target model is a model obtained by updating the initial model according to the model parameters.

[0035] In step S102 , the model parameters are a combination of parameters that meet preset conditions.

[0036] In step S102 , the preset condition is used to constrain the error value between the predicted position and the actual position of the target model to be less than a preset threshold.

[0037] Optionally, the target model refers to a trained deep learning model that can predict fault information based on an input fault distance image. The model is obtained by training on a large number of fault distance images and has the ability to extract fault-related features from the images.

[0038] Alternatively, in the field of deep learning, prior knowledge refers to patterns or rules learned by the model during training. In this embodiment, prior knowledge is reflected in the model's ability to identify and locate fault peaks in the fault distance image.

[0039] Optionally, the target model uses automated feature extraction to identify key patterns in the image, known as fault peaks, and predicts initial fault information based on these patterns. This emphasizes the automated nature of deep learning models and their intelligent decision-making capabilities based on learning from historical data, eliminating the reliance on manual analysis or simple algorithmic rules for fault detection. By using the target model to automatically extract features from fault distance images, this solution effectively captures the characteristic relationships between signals in the distance and frequency dimensions.

[0040] Optionally, the initial fault information refers to the result of the target model prediction including multiple potential fault locations. This is based on a preliminary analysis of target image features and may include multiple location estimates because in reality, multiple fault points may exist in a cable.

[0041] Step S103: determining a target fault location in a target cable according to the initial fault information.

[0042] Optionally, the target fault location refers to the exact location of the actual fault point determined after further analysis or verification. This step may be a refinement of the initial prediction results to obtain more accurate positioning information.

[0043] Optionally, the fault detection system can confirm and optimize the initial prediction results by analyzing multiple predicted locations to eliminate false positives or identify the true fault point. Through comparison and verification, the fault location that truly requires attention is ultimately determined, guiding subsequent repair or replacement work and improving the effectiveness and accuracy of fault detection.

[0044] Alternatively, users can simply import a distance-to-fault image to obtain fault distance predictions, without requiring in-depth knowledge of Fourier transforms or deep learning models. This ease of use significantly reduces operator skill requirements and enhances the system's practicality and ease of use in industrial settings.

[0045] Optionally, a deep learning model can be used to automatically extract features from fault distance images and predict fault locations, fully exploiting the implicit information in long-distance weak signals. High-precision spectrum analysis technology can also be used to improve the resolution of echo signals, significantly enhancing long-distance fault detection capabilities. When processing fault signals, spectrum analysis is first used to convert the signal into an image. This image not only contains the signal's amplitude information but also preserves the distribution characteristics of distance and frequency. By automatically extracting these image features through a deep learning model, the system can effectively distinguish between valid information and noise interference in the signal, achieving high-precision fault location even in long-distance transmission and complex noise backgrounds.

[0046] As can be seen from steps S101 to S103, in the present application, a target image of a target cable is first received, wherein the target image is used to characterize the relationship between the amplitude and propagation distance of the echo signal in the target cable. The target image is then input into a target model. Target features of the target image are extracted based on prior knowledge learned by the target model during the model training phase, and initial fault information corresponding to the target image is predicted based on the target features. The target features are used to characterize the distribution characteristics of the signal in the target image. The initial fault information includes multiple fault locations. The target model is a model obtained by updating the initial model based on model parameters, wherein the model parameters are parameter combinations that meet preset conditions, wherein the preset conditions are used to constrain the error value between the predicted position and the actual position of the target model to be less than a preset threshold. The target fault location in the target cable is then determined based on the initial fault information. That is, through the automatic extraction and intelligent analysis of the target image features by the model, the purpose of automated and high-precision fault point location is achieved, thereby achieving the technical effect of improving the accuracy and efficiency of fault detection and location in complex environments, thereby solving the technical problem of low fault point location accuracy in complex environments.

[0047] In an optional embodiment, before receiving the target image of the target cable, the fault detection system first receives the echo signal in the target cable, and then mixes the echo signal with a preset signal to obtain a target signal, wherein the preset signal is used to characterize the periodically changing electrical signal, and then determines the amplitude spectrum of the target signal, and determines the target image based on the amplitude spectrum of the target signal.

[0048] Optionally, the fault detection system first captures the echo signal reflected from the target cable using a sensor or SDR device (wireless communication device). These signals contain electrical information about the internal state of the leaky cable and are the basis for fault detection. The system then mixes the received echo signal with a preset signal to generate a target signal. The preset signal is a periodically varying electrical signal, such as an FMCW signal, whose frequency varies linearly with time. This signal is used to mix with the echo signal, converting the frequency differences in the echo signals into amplitude differences, facilitating subsequent signal analysis. After mixing, the system performs a fast Fourier transform on the target signal to obtain its amplitude spectrum. The amplitude spectrum shows the amplitude distribution of the signal at different frequencies and is crucial for identifying fault characteristics in the echo signal. Finally, the system generates a target image based on the amplitude spectrum of the target signal. The horizontal axis of the target image represents distance, and the vertical axis represents signal amplitude. This intuitively illustrates the signal variation along the distance axis, providing a clear fault distance image input for the deep learning model.

[0049] As can be seen from the above, the fault detection system can effectively improve the accuracy and reliability of fault detection by implementing the above steps. First, the mixing process converts the frequency information in the echo signal into amplitude information, facilitating subsequent signal analysis and fault location. Second, the acquisition of the amplitude spectrum ensures that key information in the signal is fully preserved, laying the foundation for generating high-quality fault distance images. Finally, the generation of the target image enables the deep learning model to intuitively identify the fault point from the image, avoiding the potential errors and misjudgments in traditional signal processing.

[0050] In an optional embodiment, the target model is obtained by the following steps: first, the fault detection system obtains N historical images, where N is an integer greater than 1, and each historical image includes a marked actual fault location; secondly, a target method is used to expand the N historical images to M historical images, wherein the target method is used to simulate changes in the actual environment, and M historical images are generated by performing target operations on the N historical images, wherein the target operation at least includes adding noise to the historical images; and then the initial model is subjected to multiple iterative training and verification operations based on the M historical images to obtain the target model, wherein the multiple iterative training is used to determine the model parameters of the model, and the verification operation is used to verify the performance parameters of the model.

[0051] Optionally, the target model may be a deep learning model based on CNN (Convolutional Neural Network).

[0052] Alternatively, annotation refers to the specific location of the fault distance peak in the historical image pre-determined by professionals or algorithms. This process ensures the accuracy and reliability of the model training data and is the basis for the model to learn fault location rules.

[0053] Optionally, the fault detection system batch-imports N historical images from the database, where N is an integer greater than 1, representing the initial dataset size for model training. Each historical image is annotated with the actual fault location, meaning the dataset itself has been pre-processed, with each image correctly labeled with the fault location, providing a clear learning target for subsequent deep learning training.

[0054] Optionally, the fault detection system uses a targeted approach (data augmentation) to amplify N historical images to M historical images, aiming to simulate variations in the actual detection environment and improve the model's generalization capabilities. This targeted approach involves, at a minimum, adding noise to the historical images to simulate fault signal characteristics under varying signal-to-noise ratios. This series of simulations and processing allows the model to be exposed to a more diverse range of input data during training, ensuring high accuracy and robustness in the face of various noise interferences encountered in real-world applications.

[0055] Optionally, the target method includes but is not limited to adding random noise to the image, performing geometric transformations (such as rotation and translation), adjusting image brightness and contrast, etc., with the aim of simulating various interference situations that may be encountered in the actual environment, enriching the diversity of training data, and ensuring the stability and adaptability of the model in complex scenarios.

[0056] Optionally, the fault detection system uses the expanded M historical images to train the initial model, continuously adjusting the model parameters through multiple iterations until the model meets the preset performance indicators. This process includes forward propagation, loss calculation, backpropagation, and weight updates until the model converges to the optimal state. Verification operations are performed regularly or after training, using independent datasets that were not involved in training to evaluate the model's performance parameters, such as accuracy and recall. This ensures that the model not only performs well on the training set but can also be reliably applied to unknown data to avoid overfitting.

[0057] Optionally, multiple iterations of training involve gradually optimizing model parameters through repeated learning and feedback until the model can accurately predict fault locations. Validation, using an independent validation dataset, tests the model's generalization capabilities, ensuring that the model's predictive performance on new data is comparable to that of the training dataset. This is a key step in evaluating the model's overall performance.

[0058] Alternatively, by combining the fault distance spectrum input with the feature recognition capabilities of a deep learning model, along with diverse training data and noise enhancement strategies (such as random noise addition and perturbation simulation), the fault detection system's noise tolerance in complex industrial environments is significantly improved. This solution significantly enhances the stability of fault point identification in low signal-to-noise ratio conditions, overcoming the significant fluctuations in results associated with traditional methods under noise interference.

[0059] As can be seen from the above, the fault detection system has built a highly adaptable and accurate target model. Based on learning from a large number of real-world fault cases and supplemented by data augmentation techniques that simulate actual environmental variations, the model is able to accurately identify and locate leaky cable faults in complex and changing industrial inspection environments. This not only significantly improves the efficiency and accuracy of fault detection and reduces maintenance costs, but also ensures the model's robustness under varying signal strengths and complex noise environments. More importantly, through continuous training and validation, the model can continuously evolve and improve to adapt to various new or extreme fault scenarios that may arise in the future.

[0060] In an optional embodiment, the fault detection system initializes the initial model using preset model parameters to obtain a first model, then divides M historical images into a model training set and a model verification set, inputs the model training set into the first model for iterative training until the number of model iterations is greater than a preset number or the training error of the model is less than a preset threshold, and obtains the model parameters of the first model. Thereafter, the first model is updated to a second model according to the model parameters, and finally, the second model is verified according to the model verification set to obtain a target model.

[0061] Optionally, preset model parameters refer to a set of weights and biases that are pre-set before model training begins, based on the model architecture and design experience. These parameters ensure that the model starts from a relatively reasonable state and avoids falling into local optimal solutions during the initial training phase.

[0062] Optionally, the model training set is a data subset used for direct model learning and parameter adjustment. It contains a large number of images with labeled fault points and serves as direct input to the model training process. The preset number of iterations and the preset threshold are the stopping criteria for model training. The preset number of iterations limits the maximum number of training iterations, preventing endless computation; the preset threshold sets an acceptable lower limit for training error, ensuring sufficient model accuracy.

[0063] Optionally, the fault detection system first initializes the initial model using preset model parameters to generate a first model. These preset model parameters are typically based on the default settings of the model architecture or borrowed from previous similar models, providing a reasonable starting point for model training. Initialization is the first step in deep learning model training. It determines the model's initial weights and biases, which influence the convergence speed and ultimate performance of subsequent training.

[0064] Optionally, the fault detection system then divides the amplified M historical images into a model training set and a model validation set, ensuring sufficient data for model training while retaining a portion for objective evaluation of model performance. The system then inputs the model training set into the first model for iterative training, aiming to gradually optimize the model parameters until the number of model iterations exceeds a preset number or the model training error falls below a preset threshold, thereby obtaining initially optimized model parameters. This process is a critical stage in model learning and self-correction. Through multiple iterations, the model gradually learns to identify fault distance peak patterns in the fault distance image, improving prediction accuracy.

[0065] Optionally, once the first model reaches the training target, the fault detection system applies the model parameters to the first model, updating it to the second model. The second model is then tested on a validation set. This validation process evaluates the model's performance on unseen data and ensures its generalization ability. If the second model's performance on the validation set meets or exceeds the expected standard, it becomes the final target model. This validation step is crucial to preventing model overfitting and ensuring the model's reliability in real-world applications.

[0066] Optionally, the second model is an optimized model based on the first model after parameter updates, with the first model performing better in the task of identifying fault points. Validation is the process of objectively evaluating model performance using a validation set. The data in the validation set is not used during model training and can provide an independent assessment of the model's generalization capabilities.

[0067] As can be seen from the above, by executing the above steps, the fault detection system is able to build a highly adaptable and accurate target model, which demonstrates excellent performance in fault distance image analysis. Model initialization ensures a reasonable training starting point. Dataset partitioning and iterative training enable the model to learn and optimize from historical fault cases. Model updating and verification further enhance the model's robustness and generalization capabilities, enabling it to maintain stable performance when faced with new data. Overall, this process not only improves the accuracy and speed of fault point prediction, but also enhances the system's intelligence level, reduces reliance on manual experience, promotes the automation and standardization of leaky cable fault detection technology, and provides strong technical support for the industrial inspection field.

[0068] In an optional embodiment, each iterative training includes the following steps: when there are T training sets in the model training set, the fault detection system inputs the T training sets into the first model, and determines the S fault locations predicted by the first model for the T training sets through forward propagation, wherein T is an integer greater than or equal to 1 and less than or equal to M, and S is an integer greater than or equal to T, and then determines the error value between the fault location corresponding to each training set and the actual location corresponding to each training set to obtain S error values, and then updates the model parameters of the first model based on the S error values through the first algorithm and back propagation, wherein back propagation is used to determine the influence of the model parameters of the first model on the model error, and the first algorithm is used to adjust the model parameters of the first model based on the influence.

[0069] Optionally, forward propagation is the process by which a deep learning model calculates a prediction result. The model processes the input data according to the current parameters and generates a prediction of the fault location. Backward propagation is one of the core mechanisms for optimizing deep learning model parameters. By calculating the gradient of the loss function with respect to each weight, the model parameters are adjusted to minimize the prediction error. The first algorithm can be a gradient descent method, an Adam algorithm (Adaptive Moment Estimation), an Adagrad algorithm (Adaptive Gradient Algorithm), etc., which is used to update the model parameters based on the gradient obtained by backpropagation to achieve automatic optimization of the model.

[0070] Optionally, at the beginning of each training iteration, the fault detection system extracts T training sets from the model training set, where T is an integer greater than or equal to 1 and less than or equal to the total number of training sets M, representing the number of training instances input into the model at a single time. The system then inputs these T training sets into the first model and, through forward propagation, calculates the S fault locations predicted by the first model for each training set. It is worth noting that the number of S fault locations may be greater than T because a leaky cable may contain multiple fault points, and the model needs to predict all potential fault locations.

[0071] Optionally, the fault detection system then calculates the error between the S fault locations predicted by the first model and the actual locations of each image in the training set, generating a set of S error values. These error values reflect the degree of deviation between the model's predictions and the actual fault locations. The system then applies the first algorithm and, through backpropagation, updates the model parameters of the first model based on the gradient information derived from the S error values. This process ensures that the model progresses towards reducing prediction error with each iteration.

[0072] Optionally, the above steps are repeated in each training iteration until a preset stopping condition is met, the model iteration count reaches a preset number, or the model training error falls below a preset threshold. This iterative process is a key step in gradually approaching the optimal solution for the model parameters, ensuring that the model fully learns the characteristic patterns in the fault distance image.

[0073] Optionally, the fault detection system utilizes an end-to-end deep learning framework to rapidly process fault distance images, enabling accurate prediction of real-time fault distances. The automated analysis capabilities of the deep learning model avoid subjective errors associated with manual observation or empirical reliance, improving the efficiency, accuracy, and intelligence of leaky cable detection, making it suitable for dynamic and ever-changing industrial inspection scenarios.

[0074] Optionally, the fault detection system compares the model's predictions with actual fault data during operation, continuously collects new data to expand the training sample set, and regularly retrains the deep learning model to ensure its adaptability to new data and changing scenarios.

[0075] Optionally, the model training process not only locates the fault point but also involves identifying and learning features in the fault distance image, which includes a mechanism for compensating for errors introduced by hardware delays. By incorporating a large amount of historical data into the deep learning model, it can automatically identify and adjust for signal propagation delays caused by device hardware. This adjustment capability is equivalent to building an error correction model that automatically compensates for deviations caused by these hardware delays. In the optimization and iteration module, through continuous data collection and model retraining to adapt to different fault detection needs, the model can rapidly expand its capabilities by learning signal data from new environments or new devices without changing the existing hardware or underlying algorithm architecture. This ensures the model's adaptive learning ability and the ability to automatically identify and compensate for nonlinear signal deviations in complex environments. Furthermore, because the model can be continuously optimized based on historical detection data, it demonstrates higher positioning accuracy in high-precision and long-distance detection scenarios.

[0076] Alternatively, the fault detection system uses standardized fault distance images as input, combined with a deep learning model, allowing rapid adaptation to different fault detection requirements through retraining. Simply by collecting and annotating signal data from new environments or devices, the model's capabilities can be expanded with minimal training, without changing existing hardware or the underlying algorithm architecture. This design reduces system maintenance costs while enhancing its portability and scalability, enabling its flexible application to fault monitoring tasks in various fields, including power and communications.

[0077] Optionally, by combining the automatic feature extraction and modeling capabilities of deep learning networks, the fault detection system significantly improves the efficiency and accuracy of quickly locating fault distance peaks in fault distance images. Compared to manual observation or traditional peak-finding functions, this system can automatically process complex data in large-scale real-time monitoring scenarios, avoiding errors caused by manual misjudgment or improper algorithm parameter settings. This significantly reduces equipment downtime and troubleshooting costs, and improves the efficiency and accuracy of industrial monitoring.

[0078] Optionally, the fault detection system performs a Fourier transform on the intermediate frequency signal and generates a fault distance image based on the amplitude spectrum. Using a deep learning network to directly analyze the fault distance image data, this significantly improves the signal-to-noise ratio for fault distance peak location compared to traditional peak-finding methods. By learning and optimizing the fault distance image using a deep learning model, the system effectively suppresses the effects of noise and accurately identifies fault peak locations even in high-noise or complex environments, thereby improving the system's detection reliability and stability.

[0079] Optionally, the following is part of the target model code:

[0080] import torch

[0081] import torch.nn as nn

[0082] import torch.optim as optim

[0083] class PeakDetection2DCNN(nn.Module):

[0084] def__init__(self):

[0085] super(PeakDetection2DCNN,self).__init__()

[0086] #Convolutional layer extracts two-dimensional image features;

[0087] self.conv1=nn.Conv2d(1,16,kernel_size=3,stride=1,padding=1)

[0088] self.conv2=nn.Conv2d(16,32,kernel_size=3,stride=1,padding=1)

[0089] self.conv3=nn.Conv2d(32,64,kernel_size=3,stride=1,padding=1)

[0090] self.pool=nn.MaxPool2d(kernel_size=2, stride=2)

[0091] #Fully connected layer predicts peak position;

[0092] self.fc1 = nn.Linear(64*64*64,256) #Assume the input image is 128x128;

[0093] self.fc2 = nn.Linear(256,2) #output two-dimensional coordinates (x,y);

[0094] def forward(self,x):

[0095] x=self.pool(torch.relu(self.conv1(x)))

[0096] x=self.pool(torch.relu(self.conv2(x)))

[0097] x=self.pool(torch.relu(self.conv3(x)))

[0098] x=x.view(x.size(0),-1)#Flatten the feature map;

[0099] x = torch.relu(self.fc1(x))

[0100] x = self.fc2(x)

[0101] return x

[0102] #Create model instance;

[0103] model = PeakDetection2DCNN()

[0104] #Define loss function and optimizer;

[0105] criterion = nn.MSELoss()

[0106] optimizer=optim.Adam(model.parameters(),lr=0.001)

[0107] As can be seen from the above content, the above implementation method builds an efficient and accurate model training process for the fault detection system. Through sophisticated data management and iterative training strategies, the system can ensure that the first model is moving towards the optimal state at every moment during the training process. The fault location predicted by the model on the training set is getting closer and closer to the actual fault location, which not only improves the accuracy of positioning, but also enhances the generalization ability of the model on unknown data and reduces the prediction error. In addition, through the combined use of backpropagation and the first algorithm, the update of model parameters becomes more intelligent, and it can automatically identify which parameters contribute the most to the prediction error and make targeted adjustments, which accelerates the convergence speed of the model and saves computing resources. Ultimately, the target model demonstrates excellent fault identification and positioning capabilities, provides advanced intelligent tools for the field of leaky cable fault detection, and promotes the advancement of industrial maintenance technology.

[0108] In an optional embodiment, the fault detection system screens multiple fault locations in the initial fault information according to a second algorithm to obtain a target fault location in the target cable, wherein the second algorithm is used to compare the confidence levels between adjacent fault locations in the multiple fault locations, and filter out fault locations with a confidence level less than a preset value and any one of two fault locations with the same confidence level.

[0109] Alternatively, the second algorithm can be a custom comparison logic or a confidence screening algorithm based on existing technologies, such as the non-maximum suppression (NMS) algorithm, which is used to filter out the points with the highest confidence from multiple predicted positions while removing duplicate and low-confidence prediction results. Confidence in the output of a deep learning model generally indicates the model's degree of certainty about the prediction result. A higher confidence level means the model has a higher degree of confidence in the prediction result.

[0110] Optionally, first, the fault detection system calculates the confidence of each fault location in the initial fault information through a second algorithm. Confidence is a numerical indicator that reflects the degree of certainty of the model's prediction of a specific fault location. It is based on the output probability or similarity score of the model's prediction. Then, the system compares the confidence between adjacent fault locations and performs a filtering operation to remove those fault locations with a confidence lower than a preset value. The preset value is a threshold used to distinguish between high-confidence predictions and low-confidence predictions. Filtering out low-confidence locations can effectively reduce the false alarm rate. Secondly, for two fault locations with the same confidence, the system will select one to retain. Because in actual fault detection, the same confidence of two locations may mean that they both point to the same fault point, or are caused by random fluctuations in the model prediction. Retaining one of the locations can avoid repeated reporting of the same fault point and simplify the final fault list.

[0111] From the above content, it can be seen that by implementing the above steps, the fault detection system can significantly improve the efficiency and accuracy of leaky cable fault detection. The introduction of the second algorithm not only filters out low-confidence prediction results and reduces false alarms, but also avoids repeated detection of the same fault point by processing predictions with the same confidence, further simplifying the maintenance workflow. Overall, this method optimizes the post-processing stage of fault point estimation, ensures that the list of target fault locations is more refined and reliable, and provides strong support for the rapid location and handling of faults. In addition, through the rational use of confidence, the system can provide maintenance personnel with a fault list with credibility identification, which facilitates the priority handling of high-confidence faults, improves the priority judgment ability of fault handling, and has a significant promoting effect on improving the efficiency of industrial maintenance and reducing maintenance costs.

[0112] In an optional embodiment, the fault detection system generates an early warning message when it detects that the initial fault information is empty, or detects that the range of the fault location in the initial fault information does not meet the preset range, wherein the preset range is used to constrain the distance of the fault location.

[0113] Optionally, the preset range is a distance interval set based on the actual length of the leaky cable and historical fault data. This is primarily used to filter out fault location information that is clearly unreasonable or beyond the detection capability. The preset range takes into account factors such as the installation range of the leaky cable and the reasonable distance from which common faults occur. It serves as an important basis for the fault detection system to determine the validity of fault information.

[0114] Optionally, the fault detection system performs a rigorous review of initial fault information during fault detection to ensure the reliability and validity of the detection results. If the system detects that the initial fault information is empty—that is, the model fails to identify any fault peaks or fault locations in the target image—or if the system detects initial fault information but analyzes the fault location and finds it falls within an abnormal distance that does not meet the preset range, the system immediately activates an early warning mechanism and generates an early warning message. The generation of early warning information is part of the system's self-monitoring and safety protection, helping to avoid incorrect decisions caused by model misjudgments or abnormal data.

[0115] Optionally, when test results exceed pre-set safety or reasonableness standards, the system generates an early warning message, prompting maintenance personnel to check the equipment status or re-test the fault. This information is recorded and an alert is sent to the relevant maintenance personnel or monitoring system, prompting them to take measures such as re-testing, checking hardware status, or adjusting model parameters to ensure the accuracy and safety of subsequent tests.

[0116] As can be seen from the above content, by implementing this step, the fault detection system can self-monitor the validity of its detection results, promptly eliminate empty information or abnormal locations in the model output, and ensure that subsequent location and maintenance actions are only carried out when the fault point detection results meet logical and preset standards. This has several significant technical benefits: by automatically filtering out invalid or abnormal detection information, the system avoids making decisions based on erroneous data, improving the accuracy and reliability of fault location; the generation of early warning information reduces unnecessary on-site inspections and maintenance work, avoids the waste of manpower and material resources caused by false alarms, and indirectly reduces maintenance costs and operational risks; for operators or maintenance teams using the fault detection system, this mechanism improves the transparency of fault identification, provides additional security, enhances their trust in system performance, and improves overall user experience and satisfaction; when frequent early warning messages are received, this indicates that there may be specific problems with the model or hardware, prompting maintenance personnel to promptly inspect and optimize the system, promoting long-term technological progress and iterative upgrades.

[0117] In an optional embodiment, Figure 2 is a schematic diagram of an optional fault determination method according to an embodiment of the present application, such as Figure 2 As shown, first, the fault distance image of the leaky cable is received, and data preprocessing is performed on the fault distance image, including normalization, data enhancement and other operations, to improve the generalization ability of the model for actual complex scenarios. Then, the preprocessed fault distance image is input into the convolutional neural network for feature extraction and output of the fault distance, including: the feature extraction module in the convolutional neural network captures the local pattern and overall features of the fault distance image, extracts information related to the fault peak position, and then integrates the output of the feature extraction module through the fully connected layer, further processes the information of the fault distance image, and gradually focuses on the fault peak position. Finally, based on the feature analysis results, the position of the fault peak in the image is predicted, and the corresponding fault distance value is output.

[0118] In an optional embodiment, Figure 3 This is a flow chart of an optional fault determination method according to an embodiment of the present application. Figure 2 ,like Figure 3As shown in the figure, first, the original FMCW signal is obtained from the SDR device, and the intermediate frequency signal is obtained by mixing. Then, the intermediate frequency signal is processed by FFT, and the amplitude spectrum is taken to draw the leaky cable fault distance image. Then, the fault distance image is normalized to eliminate the differences between different signal strengths and improve the adaptability of the model to the input data. At the same time, data enhancement operations are performed on the fault distance image, such as rotation, translation, adding noise, etc., to improve the generalization ability of the model to actual complex scenes. Then, the real fault distance peak position is marked by manual or traditional peak finding methods, and a large-scale training sample set is constructed (that is, a fault distance image dataset containing fault peaks, and each image is marked with the actual position of the fault peak). Then, a CNN-based network model is designed to automatically extract the characteristic pattern of the fault peak in the image, and the model is trained using a large amount of labeled fault image data. This includes: using partially labeled training data to preliminarily train the model so that it can learn the basic characteristic pattern of the fault peak position, and then using the full dataset to further train the network. , optimize the model parameters (use mean square error or cross entropy loss (for classification tasks) to measure the difference between the model predicted position and the actual peak position), and improve the robustness of the model in complex noise background by introducing data enhancement technology (such as adding noise simulation). After the model training is completed, an independent validation set is applied to the model to evaluate the performance of the model, and the model is adjusted according to the evaluation results to ensure its high accuracy and robustness. After the model is adjusted, the real-time generated fault map is input into the trained deep learning model for fault diagnosis, which includes: receiving the fault distance image as input in real time and directly providing it to the trained deep learning model. The model automatically extracts features from the input fault distance image and analyzes the peak position features. Then the model outputs the predicted position of the fault peak (which may include multiple fault winds, that is, multiple fault locations) and provides the corresponding distance value. Finally, for the multi-peak results output by the model, the system combines post-processing algorithms (such as non-maximum suppression) to screen the most significant peak position to ensure the reliability of the results.

[0119] The present application also provides a fault determination device. It should be noted that the fault determination device of the present application can be used to execute the fault determination method provided in the present application. The following describes the fault determination device provided in the present application.

[0120] According to an embodiment of the present application, a device for implementing the above-mentioned fault determination method is also provided. Figure 4 is a schematic diagram of an optional fault determination device according to an embodiment of the present application, such as Figure 4 As shown, the device includes: a receiving unit 401, a prediction unit 402 and a determination unit 403.

[0121] Optionally, a receiving unit 401 is used to receive a target image of a target cable, wherein the target image is used to characterize the relationship between the amplitude of the echo signal in the target cable and the propagation distance; a prediction unit 402 is used to input the target image into a target model, extract target features of the target image based on prior knowledge learned by the target model in the model training phase, and predict initial fault information corresponding to the target image based on the target features, wherein the target features are used to characterize the distribution characteristics of the signal in the target image, the initial fault information includes multiple fault locations, and the target model is a model obtained by updating the initial model according to the model parameters, wherein the model parameters are a parameter combination that meets preset conditions, wherein the preset conditions are used to constrain the error value between the predicted position and the actual position of the target model to be less than a preset threshold; a determination unit 403 is used to determine the target fault location in the target cable according to the initial fault information.

[0122] Optionally, the fault determination device further includes: a first receiving unit, a first processing unit, and a first determining unit. The first receiving unit is configured to receive an echo signal from a target cable; the first processing unit is configured to mix the echo signal with a preset signal to obtain a target signal, wherein the preset signal is configured to represent a periodically varying electrical signal; and the first determining unit is configured to determine an amplitude spectrum of the target signal and determine a target image based on the amplitude spectrum of the target signal.

[0123] Optionally, the prediction unit 402 includes: a first acquisition subunit, a second processing subunit, and a third processing subunit. The first acquisition subunit is configured to acquire N historical images, where N is an integer greater than 1, and each historical image includes a marked actual fault location; the second processing subunit is configured to expand the N historical images to M historical images using a target method, where the target method is used to simulate changes in an actual environment and generate M historical images by performing a target operation on the N historical images, where the target operation at least includes adding noise to the historical images; and the third processing subunit is configured to perform multiple iterative training and verification operations on the initial model based on the M historical images to obtain a target model, where the multiple iterative training is used to determine model parameters of the model, and the verification operation is used to verify the performance parameters of the model.

[0124] Optionally, the third processing subunit includes: a first processing module, a first partitioning module, a first training module, a first updating module, and a first verification module. The first processing module is configured to initialize the initial model using preset model parameters to obtain a first model; the first partitioning module is configured to partition the M historical images into a model training set and a model verification set; the first training module is configured to input the model training set into the first model for iterative training until the number of model iterations exceeds a preset number or the model training error is less than a preset threshold, thereby obtaining model parameters of the first model; the first updating module is configured to update the first model to a second model based on the model parameters; and the first verification module is configured to verify the second model based on the model verification set to obtain a target model.

[0125] Optionally, the first training module includes: a first determination submodule, a second determination submodule, and a first update submodule. The first determination submodule is used to input the T training sets into the first model when there are T training sets in the model training set, and determine the S fault locations predicted by the first model for the T training sets through forward propagation, wherein T is an integer greater than or equal to 1 and less than or equal to M, and S is an integer greater than or equal to T; the second determination submodule is used to determine the error value between the fault location corresponding to each training set and the actual location corresponding to each training set, to obtain S error values; the first update submodule is used to update the model parameters of the first model based on the S error values through the first algorithm and back propagation, wherein back propagation is used to determine the influence of the model parameters of the first model on the model error, and the first algorithm is used to adjust the model parameters of the first model based on the influence.

[0126] Optionally, the determination unit 403 includes: a first screening subunit, used to screen multiple fault locations in the initial fault information according to a second algorithm to obtain a target fault location in the target cable, wherein the second algorithm is used to compare the confidences between adjacent fault locations in the multiple fault locations, and filter out the fault location whose confidence is less than a preset value and any one of the two fault locations with the same confidence.

[0127] Optionally, the fault determination device further includes: a detection unit, which generates early warning information when it detects that the initial fault information is empty, or detects that the range of the fault location in the initial fault information does not meet the preset range, wherein the preset range is used to constrain the distance of the fault location.

[0128] According to another aspect of an embodiment of the present application, a computer-readable storage medium is further provided, which includes a stored executable program, wherein when the executable program runs, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned fault determination method.

[0129] According to another aspect of an embodiment of the present application, an electronic device is also provided, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the above-mentioned fault determination method.

[0130] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0131] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0132] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0133] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0134] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0135] If the integrated unit is implemented in the form of 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 application is essentially 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, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0136] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A fault determination method, characterized in that: include: receiving a target image of a target cable, wherein the target image is used to characterize a relationship between an amplitude and a propagation distance of an echo signal in the target cable; The target image is input into a target model, target features of the target image are extracted based on prior knowledge learned by the target model during a model training phase, and initial fault information corresponding to the target image is predicted based on the target features, wherein the target features are used to characterize distribution characteristics of signals in the target image, the initial fault information includes multiple fault locations, and the target model is a model obtained by updating the initial model according to model parameters, wherein the model parameters are a parameter combination that satisfies preset conditions, wherein the preset conditions are used to constrain an error value between the predicted position and the actual position of the target model to be less than a preset threshold value; A target fault location in the target cable is determined according to the initial fault information.

2. The fault determination method according to claim 1, characterized in that: Before receiving the target image of the target cable, the method further includes: receiving an echo signal from the target cable; Mixing the echo signal with a preset signal to obtain a target signal, wherein the preset signal is used to represent a periodically changing electrical signal; An amplitude spectrum of the target signal is determined, and the target image is determined according to the amplitude spectrum of the target signal.

3. The fault determination method according to claim 1, wherein: The target model is obtained by the following steps: Acquire N historical images, where N is an integer greater than 1, and each of the historical images includes a marked actual fault location; Expanding the N historical images to M historical images using a target approach, wherein the target approach is used to simulate changes in an actual environment, generating the M historical images by performing a target operation on the N historical images, wherein the target operation at least includes adding noise to the historical images; The initial model is subjected to multiple iterative training and verification operations according to the M historical images to obtain the target model, wherein the multiple iterative trainings are used to determine the model parameters of the model, and the verification operations are used to verify the performance parameters of the model.

4. The fault determination method according to claim 3, characterized in that: Performing multiple iterative training and validation operations on the initial model according to the M historical images to obtain the target model includes: Initializing the initial model using preset model parameters to obtain a first model; Dividing the M historical images into a model training set and a model validation set; Inputting the model training set into the first model for iterative training until the number of model iterations is greater than a preset number or the training error of the model is less than a preset threshold, thereby obtaining model parameters of the first model; Updating the first model to a second model according to the model parameters; The verification operation is performed on the second model according to the model verification set to obtain the target model.

5. The fault determination method according to claim 4, characterized in that: Each training iteration consists of the following steps: When there are T training sets in the model training set, input the T training sets into the first model, and determine S fault locations predicted by the first model for the T training sets through forward propagation, where T is an integer greater than or equal to 1 and less than or equal to M, and S is an integer greater than or equal to T; Determine an error value between a fault location corresponding to each of the training sets and an actual location corresponding to each of the training sets, to obtain S error values; Based on the S error values, the model parameters of the first model are updated through a first algorithm and back propagation, wherein the back propagation is used to determine the impact of the model parameters of the first model on the model error, and the first algorithm is used to adjust the model parameters of the first model based on the impact. The fault determination method according to claim 1 , wherein: Determining a target fault location in the target cable according to the initial fault information includes: The multiple fault locations in the initial fault information are screened according to a second algorithm to obtain the target fault location in the target cable, wherein the second algorithm is used to compare the confidences of adjacent fault locations among the multiple fault locations, and to filter out the fault location whose confidence is less than a preset value and any one of the two fault locations with the same confidence.

7. The fault determination method according to claim 1, characterized in that: The method further comprises: When it is detected that the initial fault information is empty, or when it is detected that the range of the fault location in the initial fault information does not conform to a preset range, a warning message is generated, wherein the preset range is used to constrain the distance of the fault location.

8. A fault determination device, characterized in that: include: a receiving unit, configured to receive a target image of a target cable, wherein the target image is used to characterize a relationship between an amplitude and a propagation distance of an echo signal in the target cable; a prediction unit, configured to input the target image into a target model, extract target features of the target image based on prior knowledge learned by the target model during a model training phase, and predict initial fault information corresponding to the target image based on the target features, wherein the target features are used to characterize distribution characteristics of signals in the target image, the initial fault information includes multiple fault locations, and the target model is a model obtained by updating the initial model according to model parameters, wherein the model parameters are a combination of parameters that meet preset conditions, wherein the preset conditions are used to constrain an error value between a predicted position and an actual position of the target model to be less than a preset threshold; A determining unit is configured to determine a target fault location in the target cable according to the initial fault information.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located executes the fault determination method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the fault determination method according to any one of claims 1 to 7.

Images