Optical cable fault intelligent diagnosis and positioning method and system based on multi-mode fusion

Abstract

The invention belongs to the technical field of fault intelligent diagnosis and positioning, and discloses an optical cable fault intelligent diagnosis and positioning method and system based on multi-mode fusion. The system comprises a multi-modal feature extraction module, a fault type intelligent classification module, a precise positioning integration module and a fault diagnosis positioning instruction output module, input preparation processing is carried out based on a multi-source processing data sequence, feature extraction is carried out according to a deep learning model, a fusion feature vector is obtained, and a fault diagnosis positioning instruction is output. Fault type classification is carried out based on the fusion feature vector, fault type confidence is calculated, a preliminary fault diagnosis report is obtained, a final fault positioning report is obtained, strategy instruction matching is carried out based on the final fault positioning report, and a matching result is output. The method has the remarkable advantages of high multi-modal data comprehensive degree, good system diagnosis intelligent effect and high operation and maintenance decision integration capability.

Description

Technical Field

[0001] This invention relates to the field of intelligent fault diagnosis and location technology, and more specifically, to a method and system for intelligent fault diagnosis and location of optical cables based on multimodal fusion. Background Technology

[0002] The intelligent diagnosis and location system for optical cable faults is an intelligent system that uses advanced sensing technology, data analysis, and artificial intelligence algorithms to monitor the health status of optical cable networks in real time or near real time, and automatically identify and locate faults. Its main purpose is to replace the traditional, inefficient fault diagnosis methods that rely on human experience, and to achieve automated, precise, and intelligent operation and maintenance. It is a key infrastructure to ensure the stability and reliability of the "nervous system" of modern communication networks.

[0003] However, traditional intelligent fault diagnosis and location systems for optical cables suffer from several shortcomings during use. First, these systems largely rely on single data provided by optical time-domain reflectometers. While capable of detecting faults such as breaks and bends, they are less effective at detecting minor vibrations, aging, and complex faults caused by multiple factors. This results in low accuracy in diagnosing complex faults and hinders real-time early warning. Second, traditional systems often rely on simple threshold judgments or preset rules, leading to high false alarm and false negative rates. Furthermore, they struggle to classify fault types precisely, requiring manual analysis and thus limiting their overall intelligence. Third, the diagnosis, location, and maintenance stages of traditional systems are largely independent. This necessitates manual reading, understanding, and execution of various data points after a fault is detected, resulting in inefficient decision-making and lengthy processing times. In summary, effectively addressing the issues of single data dimensions, low diagnostic intelligence, and isolated data across multiple stages in traditional systems is a crucial challenge that current intelligent fault diagnosis and location systems for optical cables must overcome. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution, including:

[0005] The multimodal feature extraction module is used to prepare input based on multi-source processed data sequences and extract features according to a deep learning model to obtain a fused feature vector.

[0006] The fault type intelligent classification module is used to classify fault types based on the fused feature vector, calculate the fault type confidence, and obtain a preliminary fault diagnosis report.

[0007] The precise positioning and integration module is used to calculate the coordinate position of the fault point based on the preliminary fault diagnosis report and the multi-source processing data sequence, and obtain the final fault location report;

[0008] The fault diagnosis and location instruction output module is used to perform policy instruction matching based on the final fault location report and output the matching results.

[0009] Optionally, the system further includes:

[0010] The multi-source data acquisition and verification module is used to acquire raw data in real time based on sensor networks and perform preliminary verification to obtain the raw processed dataset.

[0011] The data preprocessing and fusion module is used to preprocess the original processing dataset and fuse it based on the preprocessing results to obtain a multi-source processing data sequence.

[0012] The visual interaction module provides staff with a graphical interface to display system status and fault information in real time, and supports receiving manual interaction commands.

[0013] Optionally, the multi-source data acquisition and verification module includes:

[0014] The acquisition unit is used to acquire raw data in real time based on the sensor network to obtain the raw dataset, which includes the data items acquired by the sensor network and the acquisition timestamp of the data item acquisition time.

[0015] The verification unit is used to perform integrity verification on all data items in the original dataset based on the database data retrieval protocol to obtain the original verification dataset. The verification includes traversing all data items in the original dataset, checking whether the length of the data item is within the preset range and whether the necessary fields of the data item exist. When the check result is negative, a data acquisition failure report is generated, the acquisition failure report is output to the staff receiving end, and the data is returned to the acquisition unit to re-acquire the original data.

[0016] The dataset processing unit is used to convert the collection timestamps in the original verification dataset into the system's internal preset standard time to obtain the first processed dataset.

[0017] The data items of different formats in the first processing dataset are uniformly converted into the system's preset matrix form to obtain the second processing dataset;

[0018] Add a unique data identifier and a corresponding sensor ID to all data items in the second processing dataset to obtain the original processing dataset;

[0019] The preprocessing fusion unit is used to output the original processed dataset to the data preprocessing fusion module.

[0020] Optionally, the data preprocessing and fusion module includes:

[0021] The denoising unit is used to remove noise from all data items in the sensor synchronization dataset based on the original processed dataset using a moving average filter, so as to obtain a denoised dataset.

[0022] The data cell is used to remove outliers from the denoised dataset using the 3σ criterion, resulting in a denoising dataset.

[0023] The linear interpolation unit is used to resample the data sequence in the removed dataset using linear interpolation to obtain an aligned dataset.

[0024] The multi-source processing data unit is used to concatenate all data items in the aligned dataset into a multi-dimensional feature vector based on sensor type and the same time axis, thus obtaining a multi-source processing data sequence.

[0025] The modal feature extraction unit is used to output multi-source processed data sequences to the multimodal feature extraction module.

[0026] Optionally, the multimodal feature extraction module includes:

[0027] The normalization processing unit is used to perform normalization processing on the multi-source processed data sequence, and convert the normalized dataset into a tensor format required by the deep learning model to obtain the model requirement dataset.

[0028] The pre-training unit is used to retrieve a pre-trained deep learning model based on the database, input the model requirement dataset into the deep learning model, and output waveform feature vectors representing waveform features and sequence feature vectors representing sequence features.

[0029] The fusion feature vector unit is used to concatenate the waveform feature vector and the sequence feature vector to obtain the fusion feature vector;

[0030] The classification unit is used to output the fused feature vector to the fault type intelligent classification module.

[0031] Optionally, the fault type intelligent classification module includes:

[0032] The fully connected layer computation unit is used to calculate the activation value vector by inputting the fused feature vector into the fully connected layer of the fault intelligent classification model based on the fused feature vector and retrieving the pre-trained fault intelligent classification model from the database. The specific formula set for the calculation is as follows:

[0033] ;

[0034] The first one was obtained respectively The weighted input vector of the layer and activation value vector ,in, For the first The weight matrix of the layer, For the first The activation vector of the layer, For the first Layer bias vector, It is the ReLU activation function;

[0035] The Softmax function unit is used to calculate the activation vector of the last fully connected layer using the Softmax function to obtain a preliminary fault diagnosis report. The specific formula for the calculation is as follows:

[0036] ;

[0037] Obtain the weighted input vector of the last layer Fault type under the condition Predicted failure probability value ,in, It is a natural constant. For the corresponding fault type The last layer of weighted input vector , This represents the total number of fault types. For the corresponding fault type The last layer of weighted input vector ;

[0038] The item with the largest value among the fault probability values ​​is used as the predicted fault type, and the corresponding fault probability value is used as the fault type confidence level.

[0039] By packaging the predicted fault type and the confidence level of the fault type, a preliminary fault diagnosis report is obtained;

[0040] The positioning integration unit is used to output the preliminary fault diagnosis report to the precise positioning integration module.

[0041] Optionally, the precise positioning integration module includes:

[0042] The diagnostic report reading unit is used to select the localization algorithm based on the predicted fault type in the preliminary fault diagnosis report;

[0043] The location result unit is used to calculate the location result based on the location algorithm of the diagnostic report reading unit and the multi-source processing data sequence corresponding to the preliminary fault diagnosis report.

[0044] The fault location reporting unit is used to integrate the location results into the preliminary fault diagnosis report to obtain the final fault location report.

[0045] The instruction output unit is used to output the final fault location report to the fault diagnosis and location instruction output module.

[0046] Optionally, the fault diagnosis and location command output module includes:

[0047] The report extraction unit is used to extract the final fault location report to obtain the strategy requirement report;

[0048] The strategy requirements report includes predicted fault types, fault type confidence levels, location results, and data collection timestamps.

[0049] The strategy matching instruction set unit is used to retrieve a preset strategy instruction matching rule table from the database, match it with the strategy requirement report, and output the strategy matching instruction set.

[0050] The policy matching instruction unit is used to call via API and output the policy matching instruction set to the external system according to the content of the policy matching instruction set.

[0051] The system also includes a display module, which provides a graphical interface for staff to view system status and fault information in real time, and supports receiving manual interaction commands. The display module includes:

[0052] The location command output unit is used to continuously monitor the fault diagnosis location command output module and receive the latest final fault location report and strategy matching command set in real time to obtain a visual report.

[0053] The visualization analysis unit is used to analyze visualization reports and display the analysis results through a graphical interface, showing the content of the report:

[0054] The location results of the fault point are displayed using an electronic map, and the location results are rendered as a bright red display icon.

[0055] The link segment where the fault point is located is highlighted using the optical fiber network topology map;

[0056] The original dataset corresponding to the fault point is dynamically displayed using curves and charts.

[0057] The final fault location report is displayed in real time through the report interface;

[0058] The real-time monitoring unit is used to monitor the operations of staff on the graphical interface in real time and convert the operation results into human interaction instructions.

[0059] The fault diagnosis instruction output unit is used to output manual interaction instructions to the fault diagnosis location instruction output module.

[0060] Secondly, a method for intelligent diagnosis and localization of optical cable faults based on multimodal fusion is provided, including:

[0061] Input preparation is performed based on multi-source processed data sequences, and feature extraction is performed based on a deep learning model to obtain a fused feature vector;

[0062] Based on the fused feature vector, the fault type is classified and the fault type confidence is calculated to obtain a preliminary fault diagnosis report;

[0063] Based on the preliminary fault diagnosis report and the multi-source processing data sequence, the coordinates of the fault point are calculated to obtain the final fault location report;

[0064] Used to match policy instructions based on the final fault location report and output the matching results.

[0065] Optionally, the method further includes:

[0066] Raw data is collected in real time using a sensor network and preliminarily verified to obtain the raw processed dataset.

[0067] The original processing dataset is preprocessed, and the results are fused to obtain a multi-source processing data sequence.

[0068] It provides staff with a graphical interface that displays system status and fault information in real time, and supports receiving manual interaction commands.

[0069] The technical effects and advantages of the present invention, which is based on a multimodal fusion-based intelligent diagnosis and location method and system for optical cable faults, are as follows:

[0070] This invention acquires raw data in real time using a sensor network, performs preliminary verification to obtain a raw processing dataset, preprocesses the raw dataset, and fuses the preprocessed data to obtain a multi-source processing data sequence. Input preparation processing is performed based on this multi-source processing data sequence, and feature extraction is performed using a deep learning model to obtain a fused feature vector. Fault type classification is performed based on the fused feature vector, and fault type confidence is calculated to obtain a preliminary fault diagnosis report. The coordinates of the fault point are calculated based on the preliminary fault diagnosis report and the multi-source processing data sequence to obtain a final fault location report. Policy command matching is performed based on the final fault location report, and the matching results are output. A graphical interface is provided for staff to display system status and fault information in real time and supports receiving manual interaction commands. This allows the system to fundamentally solve the problem of a single data source in traditional systems through the collaborative operation of the multi-source data acquisition and verification module and the data preprocessing and fusion module. This invention provides a solid data foundation for subsequent system analysis, effectively reducing the high false alarm and false negative rates in traditional systems. Furthermore, by establishing a multimodal feature extraction module and a fault type intelligent classification module, the system can learn and distinguish various complex and subtle fault types, perform refined fault type classification, and provide the confidence level of each fault type. This maximizes the accuracy and intelligence of diagnosis, significantly reducing the false alarm and false negative rates of traditional systems and effectively reducing reliance on manual intervention. Finally, through the collaborative operation of the precise positioning integration module, the fault diagnosis and positioning command output module, and the visualization interaction module, the invention effectively solves the data isolation problem present in traditional systems, minimizing the time required for staff to locate, make decisions, and execute tasks, thereby improving operational efficiency. Overall, this invention has significant advantages in terms of comprehensive multimodal data, intelligent system diagnosis, and strong integrated operational decision-making capabilities. Attached Figure Description

[0071] Figure 1 This is a schematic diagram of the intelligent optical cable fault diagnosis and location system based on multimodal fusion of the present invention;

[0072] Figure 2 This is a schematic diagram of the intelligent diagnosis and location method for optical cable faults based on multimodal fusion according to the present invention. Detailed Implementation

[0073] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0074] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” used in the embodiments of this invention are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0075] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0076] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0077] In practice, the server-side equipment deployed in the multimodal fusion-based intelligent optical cable fault diagnosis and location system may consist of one or more devices. This system can be implemented as a service instance, a virtual machine, or hardware devices. For example, it can be implemented as a service instance deployed on one or more devices in a cloud node. Simply put, it can be understood as software deployed on a cloud node to provide multimodal fusion-based intelligent optical cable fault diagnosis and location services to various user terminals. Alternatively, it can be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage various user terminals. Or, it can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more devices configured to provide multimodal fusion-based intelligent optical cable fault diagnosis and location services to various user terminals.

[0078] In terms of implementation, the intelligent optical cable fault diagnosis and location system based on multimodal fusion and the user terminal are mutually adaptable. That is, if the intelligent optical cable fault diagnosis and location system based on multimodal fusion is implemented as an application installed on a cloud service platform, then the user terminal is implemented as a client that establishes a communication connection with the application; or if the intelligent optical cable fault diagnosis and location system based on multimodal fusion is implemented as a website, then the user terminal is implemented as a webpage; or if the intelligent optical cable fault diagnosis and location system based on multimodal fusion is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0079] like Figure 1 The figure shown is a system architecture diagram of an intelligent optical cable fault diagnosis and location system based on multimodal fusion provided in an embodiment of the present invention.

[0080] The intelligent optical cable fault diagnosis and location system based on multimodal fusion described in this invention can be set up in a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server or server cluster), or it can be developed as a website. Depending on the functions implemented, the intelligent optical cable fault diagnosis and location system based on multimodal fusion may include a multi-source data acquisition and verification module, a data preprocessing and fusion module, a multimodal feature extraction module, a fault type intelligent classification module, a precise location integration module, a fault diagnosis and location command output module, and a visualization interaction module. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0081] In this embodiment of the invention, in the intelligent optical cable fault diagnosis and location system based on multimodal fusion, each of the above modules can be implemented independently and can call other modules. Here, "calling" can be understood as a module connecting to multiple modules of another type and providing corresponding services to those connected modules. For example, the visualization interaction module can call the same information acquisition module to obtain information collected by that module. Based on the above characteristics, in the intelligent optical cable fault diagnosis and location system based on multimodal fusion provided in this embodiment of the invention, the applicable scope of the system architecture can be adjusted by adding modules and directly calling them without modifying the program code, achieving cluster-based horizontal expansion to quickly and flexibly expand the system. In practical applications, the above modules can be set in the same device or different devices, or in virtual devices, such as service instances in a cloud server.

[0082] Example 1

[0083] Please see Figure 1 As shown in this embodiment, the intelligent optical cable fault diagnosis and location system based on multimodal fusion includes:

[0084] The multi-source data acquisition and verification module is used to acquire raw data in real time based on the sensor network and perform preliminary verification to obtain the raw processed dataset.

[0085] Furthermore, the steps for real-time acquisition and preliminary verification of raw data based on sensor networks include:

[0086] The acquisition unit is used to acquire raw data in real time based on the sensor network to obtain the raw dataset, which includes the data items acquired by the sensor network and the acquisition timestamp of the data item acquisition time.

[0087] It should be explained that sensor networks include, but are not limited to, optical time-domain reflectometers, acoustic vibration sensors, and temperature sensors;

[0088] The verification unit is used to perform integrity verification on all data items in the original dataset based on the database data retrieval protocol to obtain the original verification dataset. The verification includes traversing all data items in the original dataset, checking whether the length of the data item is within the preset range and whether the necessary fields of the data item exist. When the check result is negative, a data acquisition failure report is generated, the acquisition failure report is output to the staff receiving end, and the process is returned to the step acquisition unit to re-acquire the original data.

[0089] It should be explained that the preset range and required fields are all provided by the data protocol;

[0090] The dataset processing unit is used to convert the collection timestamps in the original verification dataset into the system's internal preset standard time to obtain the first processed dataset.

[0091] The data items of different formats in the first processing dataset are uniformly converted into the system's preset matrix form to obtain the second processing dataset;

[0092] It should be explained that different formats refer to the fact that the data items are collected by different types of sensors, so the formats of the data items are also different, such as text or binary streams;

[0093] Add a unique data identifier and a corresponding sensor ID to all data items in the second processing dataset to obtain the original processing dataset;

[0094] The preprocessing fusion unit is used to output the raw processed dataset to the data preprocessing fusion module;

[0095] The data preprocessing and fusion module is used to preprocess the original processing dataset and fuse it based on the preprocessing results to obtain a multi-source processing data sequence.

[0096] Furthermore, the system also includes a data preprocessing and fusion module, which is used to preprocess the original dataset and fuse it based on the preprocessing results. The steps include:

[0097] The denoising unit is used to remove noise from all data items in the sensor synchronization dataset based on the original processing dataset using a moving average filter, resulting in a denoised dataset. The specific calculation formula for the moving average filter is as follows:

[0098] ;

[0099] Get the time point Filtered data values ,in, To adjust the sliding window size, For time points The data value of the data item;

[0100] The data cell is used to remove outliers from the denoised dataset using the 3σ criterion, resulting in a denoising dataset.

[0101] The linear interpolation unit is used to resample the data sequence in the removed dataset using linear interpolation to obtain an aligned dataset.

[0102] The multi-source processing data unit is used to concatenate all data items in the aligned dataset into a multi-dimensional feature vector based on sensor type and the same time axis, thus obtaining a multi-source processing data sequence.

[0103] It needs to be explained that sensor type refers to data obtained based on different types of sensors in a sensor network, such as acoustic wave vibration sensor data. Multi-dimensional feature vector refers to data obtained by unifying all sensor type data items on the same time axis and based on multiple time axes.

[0104] The modal feature extraction unit is used to output the multi-source processed data sequence to the multimodal feature extraction module;

[0105] The system also includes a multimodal feature extraction module, which is used to prepare input based on multi-source processed data sequences and extract features based on a deep learning model to obtain a fused feature vector.

[0106] The multimodal feature extraction module includes:

[0107] The normalization processing unit is used to perform normalization processing on the multi-source processed data sequence, and convert the normalized dataset into a tensor format required by the deep learning model to obtain the model requirement dataset.

[0108] It needs to be explained that the tensor format required by deep learning models refers to, for example, converting the acoustic wave data and temperature data collected by acoustic vibration sensors and temperature sensors into 1D sequence format, and converting the data items collected by optical time domain reflectometers into 2D image format.

[0109] The pre-training unit is used to retrieve a pre-trained deep learning model based on the database, input the model requirement dataset into the deep learning model, and output waveform feature vectors representing waveform features and sequence feature vectors representing sequence features.

[0110] The fusion feature vector unit is used to concatenate the waveform feature vector and the sequence feature vector to obtain the fusion feature vector;

[0111] The classification unit is used to output the fused feature vector to the fault type intelligent classification module;

[0112] The system also includes a fault type intelligent classification module, which is used to classify fault types based on fused feature vectors, calculate the fault type confidence, and obtain a preliminary fault diagnosis report.

[0113] The fault type intelligent classification module includes:

[0114] The fully connected layer computation unit is used to calculate the activation value vector by inputting the fused feature vector into the fully connected layer of the fault intelligent classification model based on the fused feature vector and retrieving the pre-trained fault intelligent classification model from the database. The specific formula set for the calculation is as follows:

[0115] ;

[0116] The first one was obtained respectively The weighted input vector of the layer and activation value vector ,in, For the first The weight matrix of the layer, For the first The activation vector of the layer, For the first Layer bias vector, It is the ReLU activation function;

[0117] The Softmax function unit is used to calculate the activation vector of the last fully connected layer using the Softmax function to obtain a preliminary fault diagnosis report. The specific formula for the calculation is as follows:

[0118] ;

[0119] Obtain the weighted input vector of the last layer Fault type under the condition Predicted failure probability value ,in, It is a natural constant. For the corresponding fault type The last layer of weighted input vector , This represents the total number of fault types. For the corresponding fault type The last layer of weighted input vector ;

[0120] It should be explained that the types of failures include, but are not limited to, normal, breakage, aging, and bending.

[0121] The item with the largest value among the fault probability values ​​is used as the predicted fault type, and the corresponding fault probability value is used as the fault type confidence level.

[0122] By packaging the predicted fault type and the confidence level of the fault type, a preliminary fault diagnosis report is obtained;

[0123] The positioning integration unit is used to output the preliminary fault diagnosis report to the precise positioning integration module;

[0124] The system also includes a precise positioning and integration module, which is used to calculate the coordinate position of the fault point based on the preliminary fault diagnosis report and the multi-source processing data sequence, and obtain the final fault location report.

[0125] The precise positioning integration module includes:

[0126] The diagnostic report reading unit is used to select the localization algorithm based on the predicted fault type in the preliminary fault diagnosis report;

[0127] It should be explained that selecting the localization algorithm based on the predicted fault type in the preliminary fault diagnosis report means, for example, when the predicted fault type is fracture, the OTDR distance analysis method is used, and when the predicted fault type is bending, the multi-sensor sound source triangulation method is used.

[0128] The location result unit is used to calculate the location result based on the location algorithm of the diagnostic report reading unit and the multi-source processing data sequence corresponding to the preliminary fault diagnosis report.

[0129] It needs to be explained that, taking the multi-sensor sound source triangulation method as an example, sound wave data is extracted based on multi-source processed data sequences. Assuming that sensors Q1, Q2, and Q3 detect vibrations, and the vibration coordinates are respectively... , and The time difference for the vibration to reach its peak is and And establish a system of hyperbolic equations, the specific expression of which is:

[0130] ;

[0131] in, The speed at which sound waves propagate within the sheath of an optical cable;

[0132] The location result was obtained by applying Newton's iterative method to the hyperbola equations. ;

[0133] The fault location reporting unit is used to integrate the location results into the preliminary fault diagnosis report to obtain the final fault location report.

[0134] The instruction output unit is used to output the final fault location report to the fault diagnosis and location instruction output module.

[0135] The fault diagnosis and location instruction output module is used to perform strategy instruction matching based on the final fault location report and output the matching result.

[0136] The fault diagnosis and location command output module includes:

[0137] The report extraction unit is used to extract the final fault location report to obtain the strategy requirement report;

[0138] The strategy requirements report includes predicted fault types, fault type confidence levels, location results, and data collection timestamps.

[0139] The strategy matching instruction set unit is used to retrieve a preset strategy instruction matching rule table from the database, match it with the strategy requirement report, and output the strategy matching instruction set.

[0140] It should be explained that the preset policy instruction matching rule table means, for example, if the predicted fault type is fracture and the fault type confidence level is 0.92, then the policy matching instruction set is to send alarm instruction and emergency work order instruction.

[0141] The policy matching instruction unit is used to call via API and output the policy matching instruction set to the external system according to the content of the policy matching instruction set.

[0142] It should be explained that external systems include, but are not limited to, operation and maintenance work order systems, resource scheduling systems, and network management systems;

[0143] The system also includes a display module, which provides a graphical interface for staff to view system status and fault information in real time, and supports receiving manual interaction commands; the display module includes:

[0144] The location command output unit is used to continuously monitor the fault diagnosis location command output module and receive the latest final fault location report and strategy matching command set in real time to obtain a visual report.

[0145] The visualization analysis unit is used to analyze visualization reports and display the analysis results through a graphical interface, showing the content of the report:

[0146] The location results of the fault point are displayed using an electronic map, and the location results are rendered as a bright red display icon.

[0147] The link segment where the fault point is located is highlighted using the optical fiber network topology map;

[0148] The original dataset corresponding to the fault point is dynamically displayed using curves and charts.

[0149] The final fault location report is displayed in real time through the report interface;

[0150] The real-time monitoring unit is used to monitor the operations of staff on the graphical interface in real time and convert the operation results into human interaction instructions.

[0151] It should be explained that the operations performed by staff on the graphical interface include, but are not limited to, clicking on the location of the fault point on the electronic map, confirming alarm commands, and filling in processing opinions; manual interaction commands include, but are not limited to, confirming alarm commands, marking false alarms, and completing fault repair commands.

[0152] The fault diagnosis instruction output unit is used to output manual interaction instructions to the fault diagnosis location instruction output module;

[0153] This embodiment offers the following advantages: It acquires raw data in real-time using a sensor network, performs preliminary verification to obtain a raw processing dataset, preprocesses the dataset, and fuses the preprocessing results to obtain a multi-source processing data sequence. This multi-source processing data sequence is then used for input preparation processing, and feature extraction is performed using a deep learning model to obtain a fused feature vector. Based on this fused feature vector, fault types are classified, and fault type confidence is calculated to obtain a preliminary fault diagnosis report. Based on the preliminary fault diagnosis report and the multi-source processing data sequence, the coordinates of the fault point are calculated to obtain a final fault location report. Based on the final fault location report, strategy command matching is performed, and the matching results are output. A graphical interface is provided for staff to display system status and fault information in real time and supports receiving manual interaction commands. This allows the system to fundamentally solve the problem of a single data source in traditional systems through the collaborative operation of the multi-source data acquisition and verification module and the data preprocessing and fusion module. This provides a solid data foundation for subsequent system analysis, effectively reducing the high false alarm and false negative rates in traditional systems. Furthermore, the invention, through the establishment of a multimodal feature extraction module and a fault type intelligent classification module, enables the system to learn and distinguish various complex and subtle fault types, and to perform refined classification of fault types while providing confidence levels. This maximizes the accuracy and intelligence of diagnosis, significantly reducing the false alarm and false negative rates of traditional systems and effectively reducing reliance on manual intervention. Finally, through the collaborative operation of the precise positioning integration module, the fault diagnosis and positioning command output module, and the visualization interaction module, the data isolation problem in traditional systems is effectively solved, minimizing the time required for staff to locate, decide, and execute faults, thereby improving operational efficiency. Overall, this invention has significant advantages such as high comprehensiveness of multimodal data, excellent intelligent system diagnosis, and strong integrated operational decision-making capabilities.

[0154] Example 2

[0155] Please see Figure 2 As shown, parts not described in detail in this embodiment are described in Embodiment 1. A method for intelligent diagnosis and localization of optical cable faults based on multimodal fusion is provided, including:

[0156] Input preparation is performed based on multi-source processed data sequences, and feature extraction is performed based on a deep learning model to obtain a fused feature vector;

[0157] Based on the fused feature vector, the fault type is classified and the fault type confidence is calculated to obtain a preliminary fault diagnosis report;

[0158] Based on the preliminary fault diagnosis report and the multi-source processing data sequence, the coordinates of the fault point are calculated to obtain the final fault location report;

[0159] Used to match policy instructions based on the final fault location report and output the matching results.

[0160] Optionally, the method further includes:

[0161] Raw data is collected in real time using a sensor network and preliminarily verified to obtain the raw processed dataset.

[0162] The original processing dataset is preprocessed, and the results are fused to obtain a multi-source processing data sequence.

[0163] It provides staff with a graphical interface that displays system status and fault information in real time, and supports receiving manual interaction commands.

[0164] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the present invention.

Claims

1. A multi-modal fusion-based intelligent diagnosis and location system for optical cable faults, characterized in that, include: The multimodal feature extraction module is used to prepare input based on multi-source processed data sequences and extract features according to a deep learning model to obtain a fused feature vector. The fault type intelligent classification module is used to classify fault types based on the fused feature vector, calculate the fault type confidence, and obtain a preliminary fault diagnosis report. The precise positioning and integration module is used to calculate the coordinate position of the fault point based on the preliminary fault diagnosis report and the multi-source processing data sequence, and obtain the final fault location report; The fault diagnosis and location instruction output module is used to perform policy instruction matching based on the final fault location report and output the matching results.

2. The intelligent optical cable fault diagnosis and location system based on multimodal fusion according to claim 1, characterized in that, The system also includes: The multi-source data acquisition and verification module is used to acquire raw data in real time based on sensor networks and perform preliminary verification to obtain the raw processed dataset. The data preprocessing and fusion module is used to preprocess the original processing dataset and fuse it based on the preprocessing results to obtain a multi-source processing data sequence. The visual interaction module provides staff with a graphical interface to display system status and fault information in real time, and supports receiving manual interaction commands.

3. The intelligent optical cable fault diagnosis and location system based on multimodal fusion according to claim 2, characterized in that, The multi-source data acquisition and verification module includes: The acquisition unit is used to acquire raw data in real time based on the sensor network to obtain the raw dataset, which includes the data items acquired by the sensor network and the acquisition timestamp of the data item acquisition time. The verification unit is used to perform integrity verification on all data items in the original dataset based on the database data retrieval protocol to obtain the original verification dataset. The verification includes traversing all data items in the original dataset, checking whether the length of the data item is within the preset range and whether the necessary fields of the data item exist. When the check result is negative, a data acquisition failure report is generated, the acquisition failure report is output to the staff receiving end, and the data is returned to the acquisition unit to re-acquire the original data. The dataset processing unit is used to convert the collection timestamps in the original verification dataset into the system's internal preset standard time to obtain the first processed dataset. The data items of different formats in the first processing dataset are uniformly converted into the system's preset matrix form to obtain the second processing dataset; Add a unique data identifier and a corresponding sensor ID to all data items in the second processing dataset to obtain the original processing dataset; The preprocessing fusion unit is used to output the original processed dataset to the data preprocessing fusion module.

4. The intelligent optical cable fault diagnosis and location system based on multimodal fusion according to claim 3, characterized in that, The data preprocessing and fusion module includes: The denoising unit is used to remove noise from all data items in the sensor synchronization dataset based on the original processed dataset using a moving average filter, so as to obtain a denoised dataset. The data cell is used to remove outliers from the denoised dataset using the 3σ criterion, resulting in a denoising dataset. The linear interpolation unit is used to resample the data sequence in the removed dataset using linear interpolation to obtain an aligned dataset. The multi-source processing data unit is used to concatenate all data items in the aligned dataset into a multi-dimensional feature vector based on sensor type and the same time axis, thus obtaining a multi-source processing data sequence. The modal feature extraction unit is used to output multi-source processed data sequences to the multimodal feature extraction module.

5. The intelligent optical cable fault diagnosis and location system based on multimodal fusion according to claim 4, characterized in that, The multimodal feature extraction module includes: The normalization processing unit is used to perform normalization processing on the multi-source processed data sequence, and convert the normalized dataset into a tensor format required by the deep learning model to obtain the model requirement dataset. The pre-training unit is used to retrieve a pre-trained deep learning model based on the database, input the model requirement dataset into the deep learning model, and output waveform feature vectors representing waveform features and sequence feature vectors representing sequence features. The fusion feature vector unit is used to concatenate the waveform feature vector and the sequence feature vector to obtain the fusion feature vector; The classification unit is used to output the fused feature vector to the fault type intelligent classification module.

6. The intelligent optical cable fault diagnosis and location system based on multimodal fusion according to claim 5, characterized in that, The fault type intelligent classification module includes: The fully connected layer computation unit is used to calculate the activation value vector by inputting the fused feature vector into the fully connected layer of the fault intelligent classification model based on the fused feature vector and retrieving the pre-trained fault intelligent classification model from the database. The specific formula set for the calculation is as follows: ； The first one was obtained respectively The weighted input vector of the layer and activation value vector ,in, For the first The weight matrix of the layer, For the first The activation vector of the layer, For the first Layer bias vector, It is the ReLU activation function; The Softmax function unit is used to calculate the activation vector of the last fully connected layer using the Softmax function to obtain a preliminary fault diagnosis report. The specific formula for the calculation is as follows: ； Obtain the weighted input vector of the last layer Fault type under the condition Predicted failure probability value ,in, It is a natural constant. For the corresponding fault type The last layer of weighted input vector , This represents the total number of fault types. For the corresponding fault type The last layer of weighted input vector ; The item with the largest value among the fault probability values ​​is used as the predicted fault type, and the corresponding fault probability value is used as the fault type confidence level. By packaging the predicted fault type and the confidence level of the fault type, a preliminary fault diagnosis report is obtained; The positioning integration unit is used to output the preliminary fault diagnosis report to the precise positioning integration module.

7. The intelligent optical cable fault diagnosis and location system based on multimodal fusion according to claim 6, characterized in that, The precise positioning integration module includes: The diagnostic report reading unit is used to select the localization algorithm based on the predicted fault type in the preliminary fault diagnosis report; The location result unit is used to calculate the location result based on the location algorithm of the diagnostic report reading unit and the multi-source processing data sequence corresponding to the preliminary fault diagnosis report. The fault location reporting unit is used to integrate the location results into the preliminary fault diagnosis report to obtain the final fault location report. The instruction output unit is used to output the final fault location report to the fault diagnosis and location instruction output module.

8. The intelligent optical cable fault diagnosis and location system based on multimodal fusion according to claim 7, characterized in that, The fault diagnosis and location command output module includes: The report extraction unit is used to extract the final fault location report to obtain the strategy requirement report; The strategy requirements report includes predicted fault types, fault type confidence levels, location results, and data collection timestamps. The strategy matching instruction set unit is used to retrieve a preset strategy instruction matching rule table from the database, match it with the strategy requirement report, and output the strategy matching instruction set. The policy matching instruction unit is used to call via API and output the policy matching instruction set to the external system according to the content of the policy matching instruction set. The system also includes a display module, which provides a graphical interface for staff to view system status and fault information in real time, and supports receiving manual interaction commands. The display module includes: The location command output unit is used to continuously monitor the fault diagnosis location command output module and receive the latest final fault location report and strategy matching command set in real time to obtain a visual report. The visualization analysis unit is used to analyze visualization reports and display the analysis results through a graphical interface, showing the content of the report: The location results of the fault point are displayed using an electronic map, and the location results are rendered as a bright red display icon. The link segment where the fault point is located is highlighted using the optical fiber network topology map; The original dataset corresponding to the fault point is dynamically displayed using curves and charts. The final fault location report is displayed in real time through the report interface; The real-time monitoring unit is used to monitor the operations of staff on the graphical interface in real time and convert the operation results into human interaction instructions. The fault diagnosis instruction output unit is used to output manual interaction instructions to the fault diagnosis location instruction output module.

9. A method for intelligent diagnosis and localization of optical cable faults based on multimodal fusion, characterized in that, include: Input preparation is performed based on multi-source processed data sequences, and feature extraction is performed based on a deep learning model to obtain a fused feature vector; Based on the fused feature vector, the fault type is classified and the fault type confidence is calculated to obtain a preliminary fault diagnosis report; Based on the preliminary fault diagnosis report and the multi-source processing data sequence, the coordinates of the fault point are calculated to obtain the final fault location report; Used to match policy instructions based on the final fault location report and output the matching results.

10. The intelligent optical cable fault diagnosis and location system based on multimodal fusion according to claim 9, characterized in that, The method further includes: Raw data is collected in real time using a sensor network and preliminarily verified to obtain the raw processed dataset. The original processing dataset is preprocessed, and the results are fused to obtain a multi-source processing data sequence. It provides staff with a graphical interface that displays system status and fault information in real time, and supports receiving manual interaction commands.

Images