Intelligent communication optical cable distribution box and data processing method

Abstract

The application discloses an intelligent communication optical cable junction box and a data processing method, relates to the technical field of communication optical cable monitoring, and comprises the following steps: collecting the optical power value of the optical fiber connector port in the box, obtaining the joint working temperature value of the corresponding connector end face area, calculating the port optical power deviation value, combining the deviation value and the joint temperature to construct the port working state vector, inputting the trained optical fiber link evaluation model, and outputting the aging early warning level and the fault probability score of each port. The port reaching the preset threshold generates a maintenance task instruction containing the port identification and the fault type code. The method realizes the coupling representation of the port optical power and the end face temperature, quantitatively analyzes the port aging and the fault probability through the model, weakens the single parameter determination error, improves the fault identification and positioning accuracy, realizes the precision and classification of the operation and maintenance instruction, and has the advantages of high accuracy, high reliability and the like.

Description

Technical Field

[0001] This invention belongs to the field of communication optical cable monitoring technology, specifically an intelligent communication optical cable junction box and data processing method. Background Technology

[0002] Traditional optical cable junction box maintenance and monitoring only collects optical power data from the optical fiber connector ports. The link transmission status is determined by comparing it with the standard calibrated optical power. The temperature of the box is collected only for the overall environment and does not measure the temperature of the optical fiber connector end face area independently. Optical power and temperature data are collected and analyzed independently, and no port-level correlation is established between the two.

[0003] Optical power attenuation is influenced by multiple factors, including connector end-face temperature, contact condition, and device aging. A single optical power parameter cannot distinguish the specific causes of attenuation, and the overall ambient temperature cannot reflect the true local operating thermal state of the port connector. Parameters are limited in scope and the data are independent, resulting in low accuracy in link status determination. Conventional threshold comparisons can only provide simple anomaly alarms, failing to quantify port aging and the probability of failure. Alarm information only indicates link anomalies, lacking precise port identification and fault type differentiation capabilities. Fault location and repair lack specificity, leading to significant false alarms and missed alarms, and low efficiency in operation and maintenance.

[0004] To address the issues of the inability to correlate port optical power with end-face temperature, reliance on a single parameter for status determination, inability to quantify aging and failure probabilities, and inability to accurately identify and classify faulty ports, it is necessary to establish a port-level multi-parameter fusion characterization method. This method should be used to achieve quantitative evaluation through a model and generate maintenance instructions with port identification and fault type codes. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art;

[0006] Therefore, this invention proposes an intelligent communication optical cable junction box and a data processing method, including:

[0007] The optical power values ​​of the ports of multiple fiber optic connectors installed in the intelligent communication optical cable junction box are collected. The multiple fiber optic connectors are connected to external optical cable lines to form multiple optical signal transmission channels.

[0008] The connector operating temperature value is obtained from the temperature sensor deployed corresponding to the plurality of fiber optic connectors. The connector operating temperature value is the measured temperature of the fiber optic connector end face area.

[0009] The optical power value of the port is compared with the preset reference optical power value to calculate the optical power deviation value of each port. The reference optical power value is the calibrated optical power of each port under standard operating conditions.

[0010] The optical power deviation value is combined with the connector operating temperature value to form a port operating state vector that reflects the state of the optical fiber link. The number of dimensions of the port operating state vector is the same as the number of ports.

[0011] The port operating state vectors of all ports are input into a pre-trained fiber optic link evaluation model, which outputs the aging warning level and fault probability score for each port.

[0012] For ports whose aging warning level reaches a preset threshold or whose fault probability score exceeds a preset threshold, a port maintenance task instruction is generated, which includes a port identifier and a fault type code.

[0013] Further, the optical power value of the port is compared with a preset reference optical power value to calculate the optical power deviation value of each port, including:

[0014] Read the standard operating parameters for each port from the junction box configuration database. The standard operating parameters include the reference optical power value and its allowable fluctuation range.

[0015] Extract the current port optical power value from the real-time acquired data sequence;

[0016] Calculate the difference between the port optical power value and the corresponding reference optical power value to generate a preliminary optical power deviation value;

[0017] Determine whether the preliminary optical power deviation value exceeds the allowable fluctuation range. If it does, mark the preliminary optical power deviation value as a valid deviation value.

[0018] The effective deviation value is divided by the reference optical power value to calculate the normalized relative optical power deviation value, which is then output as the optical power deviation value.

[0019] Furthermore, the optical power deviation value is combined with the connector operating temperature value to form a port operating state vector reflecting the fiber optic link status, including:

[0020] The optical power deviation value is linearly normalized to map its value to a standard value range.

[0021] The working temperature value of the joint is normalized by subtracting the standard temperature reference value and then dividing by the length of the standard temperature range.

[0022] The normalized optical power deviation value and the normalized connector operating temperature value are spliced ​​together in port number order.

[0023] Add a Boolean flag reflecting the current online status of the port to the spliced ​​data. The online status flag is determined by the presence or absence of the port's optical signal.

[0024] The concatenated data from all ports is aggregated to generate a multi-row, multi-column matrix. This matrix is ​​the working status vector of the ports, where each row corresponds to the status data of one port.

[0025] Furthermore, the port operating state vectors of all ports are input into a pre-trained fiber optic link evaluation model, which outputs an aging warning level and a fault probability score for each port, including:

[0026] The fiber optic link evaluation model includes a feature extraction sub-model, which processes the input port operating state vector to extract high-dimensional state features.

[0027] The high-dimensional state features are input into a classification decision sub-model. The classification decision sub-model classifies the current port state based on the classification boundary obtained by training on historical fault data, and divides the port state into three types: normal state, aging state, and fault state.

[0028] For ports classified as aging states, the aging warning level is calculated based on the position of their state characteristics in the aging state characteristic space. The aging warning level is a value between zero and one hundred.

[0029] For all ports, another regression prediction sub-model predicts the failure probability score based on state characteristics. The failure probability score is the probability value of a port failing within a specified time period in the future.

[0030] Output the status classification, aging warning level, and fault probability score for each port.

[0031] Furthermore, the high-dimensional state features are input into a classification decision sub-model. This sub-model classifies the current port state based on classification boundaries obtained through training on historical fault data, categorizing the port state into three types: normal state, aging state, and fault state.

[0032] The classification decision sub-model contains multiple trained decision rules, each decision rule corresponding to a combination of feature values ​​of different dimensions in the port working state vector;

[0033] Each component of the high-dimensional state feature is matched sequentially with the conditions in the decision rule;

[0034] When the state characteristics of a port meet all the conditions of a certain decision rule, the port is determined to be a state type corresponding to the decision rule, namely, one of the normal state, aging state, and fault state.

[0035] If the state characteristics of a port fail to match any of the defined decision rules, the port's state type is marked as unknown and enters the manual verification queue.

[0036] The classification results are bound to the port identifiers to form a preliminary list of port classification results.

[0037] Furthermore, the fiber optic link evaluation model is constructed in the following manner:

[0038] Collect historical operating data, which includes port optical power values, connector operating temperature values, corresponding port status labels, and fault records of multiple fiber optic connectors at multiple time periods, to form the original training sample set;

[0039] The port optical power values ​​and connector operating temperature values ​​in the original training sample set are preprocessed to calculate the historical optical power deviation values, and after normalization, a historical port operating state vector is constructed.

[0040] Using the historical port working state vector as input features and the corresponding port state label as supervision signal, a deep neural network model is trained. The deep neural network model includes a feature extraction layer for extracting high-dimensional state features, a classification layer for port state classification, and a regression output layer for calculating fault probability.

[0041] The trained deep neural network model is validated and tuned using a labeled validation set until the model's classification accuracy and regression error on the validation set reach the preset performance indicators, thereby obtaining the trained fiber optic link evaluation model.

[0042] Furthermore, for ports where the aging warning level reaches a preset threshold or the fault probability score exceeds a preset threshold, a port maintenance task instruction is generated, including:

[0043] Set aging warning level thresholds and fault probability score thresholds;

[0044] Traverse all ports and filter out ports whose aging warning level is greater than or equal to the aging warning level threshold, and ports whose fault probability score is greater than or equal to the fault probability score threshold.

[0045] The identifier of the selected port, the aging warning level, the fault probability score, and the current online status of the port are encapsulated into a maintenance task data structure;

[0046] Based on the geographical location identifier of the port, match the maintenance task data structure with the maintenance personnel responsible for the area and the available spare parts inventory information;

[0047] The encapsulated maintenance task data structure, the matched maintenance personnel information, and the spare parts inventory information are integrated to generate a port maintenance task instruction containing complete dispatch elements, and stored in the task queue to be executed.

[0048] Furthermore, the method also includes a trend analysis step on historical port data:

[0049] The port optical power deviation value and the connector operating temperature value of each port are stored at a preset period to form a port historical data sequence;

[0050] For each port's historical data sequence, the average rate of change of the optical power deviation value over a long time window is calculated as the optical power degradation rate.

[0051] Calculate the average rate of change of the joint's operating temperature value over the same long time window, as the temperature change rate;

[0052] The optical power degradation rate is combined with the temperature change rate to form a long-term trend vector for the port;

[0053] The long-term trend vector is input into a trend analysis model, and the trend analysis model outputs the remaining useful life prediction value of the port.

[0054] The remaining service life prediction value is associated with and stored in conjunction with the current port's aging warning level and failure probability score to enrich the basis for port maintenance decisions;

[0055] The step of inputting the long-term trend vector into a trend analysis model, wherein the trend analysis model outputs a predicted remaining useful life value, includes:

[0056] The trend analysis model incorporates multiple typical fiber optic port failure mode curves. Each failure mode curve describes the evolution of parameters such as optical power and temperature over time under a specific failure mode.

[0057] The similarity is calculated between the long-term trend vector of the current port and the built-in curves of each of the typical fiber optic port failure modes.

[0058] Select the failure mode curve with the highest similarity as the matching failure mode for the current port;

[0059] Read the standard time length from the current state to complete failure corresponding to the matching failure mode curve;

[0060] The standard time length is scaled and corrected based on the change magnitude of the long-term trend vector to calculate the specific predicted value of the remaining useful life.

[0061] Furthermore, the method also includes a data model update step based on port maintenance results:

[0062] After the maintenance personnel complete the on-site maintenance according to the port maintenance task instruction, they receive maintenance result feedback data. The maintenance result feedback data includes a description of the maintenance measures, the identification of the replaced spare parts, and the optical power value and temperature value of the port after maintenance.

[0063] Based on the description of the maintenance measures, determine the classification of the root causes of the faults in the port being maintained;

[0064] Extract the status data of the port within a certain period after maintenance from the real-time monitoring data stream of the junction box, and calculate its status stability index.

[0065] The maintenance result feedback data, the fault root cause classification, the post-maintenance port status stability index, and the pre-maintenance port historical status data are combined to form a model training sample.

[0066] The fiber optic link evaluation model is incrementally trained using a new dataset containing the training samples of the model, and the internal parameters of the fiber optic link evaluation model are updated to improve the model's accuracy in identifying similar faults in the future.

[0067] The step of incrementally training the fiber optic link evaluation model using a new dataset containing the model's training samples and updating the model's internal parameters includes:

[0068] Load the current version of the fiber optic link evaluation model and its parameters from the model storage area;

[0069] The collected training samples of the model are divided into an incremental training set and a validation set according to a preset ratio;

[0070] The data from the incremental training set is input into the fiber optic link evaluation model, and the backpropagation algorithm is used to calculate the loss function value between the model's predicted output and the true label.

[0071] The gradient of the model parameters is calculated based on the loss function value, and the optimizer is used to make minor adjustments to the model's internal parameters.

[0072] The performance of the model after parameter adjustment is evaluated on the validation set. If the performance meets the preset improvement standard, the adjusted model parameters are saved as the model parameters of the new version.

[0073] The updated fiber optic link evaluation model is redeployed into the data processing flow of the smart communication fiber optic cable junction box.

[0074] Furthermore, the present invention also includes an intelligent communication optical cable junction box, the junction box comprising a box body, a main control module, an optical power monitoring array, a temperature sensor array, a data storage module, and a communication module;

[0075] The enclosure is equipped with a port panel for installing multiple fiber optic connectors, which are used to connect with external optical cable lines to form multiple optical signal transmission channels.

[0076] The optical power monitoring array is installed inside the enclosure and electrically connected to the main control module, and is used to collect the port optical power value of each of the plurality of optical fiber connectors;

[0077] The temperature sensor array is arranged corresponding to the plurality of fiber optic connectors and electrically connected to the main control module, and is used to collect the joint working temperature value of the end face area of ​​each fiber optic connector.

[0078] The data storage module is electrically connected to the main control module and is used to store preset reference optical power values, junction box configuration database, pre-trained fiber optic link evaluation model, and preset thresholds.

[0079] The main control module is configured to execute the data processing method for an intelligent communication optical cable junction box as described above.

[0080] Compared with the prior art, the beneficial effects of the present invention are:

[0081] By matching and combining the port optical power deviation value with the corresponding fiber optic connector end-face area's connector operating temperature, a port operating state vector with dimensions consistent with the number of ports is constructed, achieving coupled characterization of optical power and temperature parameters at the single-port level. Local port temperature data directly reflects the contact heating and heat loss of the connector end-face, forming a corresponding correlation with the optical power deviation. This eliminates the misalignment between the overall enclosure temperature and port data, and avoids the data fragmentation caused by independent acquisition and analysis. Multi-parameter fusion can fully reflect the intrinsic relationship between port transmission attenuation and end-face thermal state, distinguishing whether transmission anomalies are caused by temperature changes or device degradation, avoiding the one-sidedness of judging by a single optical power parameter, and improving the detail and accuracy of port operating state characterization.

[0082] Using port operating status vectors as model input, and relying on a trained fiber optic link evaluation model, the model outputs independent aging warning levels and fault probability scores for each port. Abnormal ports are filtered through preset thresholds, and maintenance task instructions containing port identifiers and fault type codes are generated. The model can quantitatively analyze port aging and fault risk, with output results possessing both level and score quantitative attributes, replacing the traditional simple threshold comparison method and reducing the probability of false alarms and missed alarms. It can distinguish and identify different fault causes, solidify the correspondence between fault types and corresponding codes, and allow maintenance instructions to directly target the port and specific fault type, achieving accurate fault location and classification, reducing manual judgment, and ensuring a precise correspondence between maintenance targets and faulty ports, thus improving the targeting and effectiveness of maintenance operations. Attached Figure Description

[0083] Figure 1 This is a flowchart of a data processing method for an intelligent communication optical cable junction box according to the present invention;

[0084] Figure 2 A flowchart for calculating the optical power deviation values ​​at each port;

[0085] Figure 3 A flowchart for forming the port operating state vector. Detailed Implementation

[0086] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0087] See Figure 1 The data processing method for an intelligent communication optical cable junction box provided by the present invention has the following overall implementation scheme:

[0088] During the operation of the intelligent communication optical cable junction box, the port optical power values ​​of multiple optical fiber connectors installed inside the box are collected. These optical fiber connectors are connected to external optical cable lines, forming multiple optical signal transmission channels. Simultaneously, the operating temperature values ​​of the connectors are acquired from temperature sensors corresponding to each optical fiber connector; these temperatures are the measured temperatures of the connector end face area. The collected port optical power values ​​are compared with preset reference optical power values, which are the calibrated optical power of each port under standard operating conditions. The comparison generates optical power deviation values ​​for each port. The calculated optical power deviation values ​​are combined with the corresponding connector operating temperature values ​​to form a port operating status vector reflecting the health status of the optical fiber link. The dimension of this vector is consistent with the number of ports. The port operating status vectors of all ports are input into a pre-trained optical fiber link evaluation model. This model calculates and outputs an aging warning level and a fault probability score for each port. For ports whose aging warning level reaches a preset threshold or whose fault probability score exceeds a preset threshold, a port maintenance task instruction containing a port identifier and fault type code is automatically generated to guide maintenance work.

[0089] In one embodiment of the present invention, when calculating the optical power deviation value of each port, refer to... Figure 2 The system reads the standard operating parameters corresponding to each port from the junction box configuration database. These standard operating parameters include the reference optical power value and its allowable fluctuation range. The current port optical power value is extracted from the real-time acquired data sequence. The difference between this port optical power value and the corresponding reference optical power value is calculated to generate a preliminary optical power deviation value. It is determined whether this preliminary optical power deviation value exceeds the allowable fluctuation range; if it does, the preliminary optical power deviation value is marked as a valid deviation value. The valid deviation value is divided by the reference optical power value to calculate the normalized relative optical power deviation value, which is then output as the final optical power deviation value.

[0090] In practical implementation, assume that a smart communication optical cable junction box contains 24 optical fiber connectors, numbered P1 to P24. Each optical fiber connector corresponds to a preset reference optical power value and its allowable fluctuation range. These standard operating parameters are stored in the junction box configuration database. In practice, when the system initiates the deviation value calculation process, it reads the standard operating parameters of port P1 from the junction box configuration database. For example, the reference optical power value of port P1 is -2.5dBm, and the allowable fluctuation range is ±0.2dBm. Simultaneously, the system extracts the current port optical power value of port P1 from the real-time acquired and cached data sequence; for example, the obtained real-time port optical power value is -2.85dBm.

[0091] In some embodiments, the difference between the port optical power value of port P1 and the corresponding reference optical power value is calculated to generate a preliminary optical power deviation value. Based on the example data above, the calculation process is as follows: Preliminary optical power deviation value = (-2.85) - (-2.5) = -0.35dBm. The system determines whether the absolute value of this preliminary optical power deviation value exceeds the allowable fluctuation range. The allowable fluctuation range for port P1 is ±0.2dBm, and the calculated absolute deviation value is 0.35dBm, exceeding the 0.2dBm threshold. Therefore, this preliminary optical power deviation value is marked as a valid deviation value. It can be understood that if the preliminary deviation value of a port does not exceed its allowable fluctuation range, the deviation calculation process for that port terminates at this step, and no subsequent relative optical power deviation value output is generated.

[0092] In practice, data marked as valid deviation values ​​need to be normalized to calculate the relative optical power deviation value. The calculation method is to divide the valid deviation value by the corresponding reference optical power value. Optionally, this calculation process can be expressed by the following formula:

[0093]

[0094] Where: the character ΔP in the formula relative This represents the calculated relative deviation value of optical power, which is used as the final output optical power deviation value; the character P effective_deviation This indicates the initial optical power deviation value marked as an effective deviation value; character P reference This represents the reference optical power value read from the database. Substituting the example data for port P1, the relative optical power deviation value for port P1 is calculated as: (-0.35) / (-2.5) = 0.14. In some embodiments, this calculated result of 0.14 is output as the optical power deviation value for port P1, used to subsequently form the port operating state vector. It can be understood that this normalization process makes the deviations between ports with different reference power levels comparable. The system sequentially executes the above complete process—from reading parameters, extracting real-time values, calculating preliminary deviations, judging validity, to calculating relative deviations—for all ports in the intelligent communication optical cable junction box, such as P1 to P24, generating an optical power deviation value for each eligible port.

[0095] In one embodiment of the present invention, when forming a port operating state vector reflecting the state of the optical fiber link, refer to... Figure 3The optical power deviation value is linearly normalized, mapping its value to a preset standard value range. The connector operating temperature value is normalized by subtracting a standard temperature reference value from the connector operating temperature value and then dividing by the length of a standard temperature range. The normalized optical power deviation value and the normalized connector operating temperature value are then concatenated in port number order. A Boolean flag reflecting the current online status of the port is added to the concatenated data; this online status flag is determined by the presence of an optical signal at the port. The concatenated data from all ports in the intelligent communication optical cable junction box are summarized to generate a multi-row, multi-column matrix. This matrix is ​​the port operating status vector, where each row of the matrix corresponds to the status data of one port.

[0096] In the specific implementation, assuming that the optical power deviation value of port P1 in the intelligent communication optical cable junction box is calculated to be 0.14, and the operating temperature of the port P1 connector collected by the temperature sensor is 41.2 degrees Celsius, the optical power deviation value is linearly normalized. For example, a standard value range is set as [-1, 1], and the optical power deviation value of 0.14 is mapped to this range. The mapping process uses a linear transformation. If the theoretical maximum positive offset of the optical power deviation value is known to be +0.5 and the theoretical maximum negative offset is -0.5, then the normalization calculation is: subtract the minimum value (-0.5) from the original deviation value and divide by the full range (1.0). After this calculation, the optical power deviation value of 0.14 for port P1 is obtained as a normalized value within the range [-1, 1].

[0097] In some embodiments, the connector operating temperature value is normalized by subtracting the standard temperature reference value and then dividing by the standard temperature range length. For example, the standard temperature reference value is set to 20 degrees Celsius, and the standard temperature range length is 30 degrees Celsius. For the connector operating temperature value of 41.2 degrees Celsius at port P1, the normalization calculation process is as follows: Normalized temperature = (41.2 - 20) / 30. Optionally, this calculation can be uniformly expressed by a single formula:

[0098]

[0099] Where: the character T in the formula normalized This indicates the normalized operating temperature value of the connector; the character T measured This indicates the operating temperature value of the connector collected by the temperature sensor; the character T base Indicates the preset standard temperature reference value; the character L range This indicates the length of the preset standard temperature range. Based on the example data, the normalized temperature value of port P1 is (41.2-20) / 30=0.7067.

[0100] In practice, the normalized optical power deviation value and the normalized connector operating temperature value are concatenated in port number order. The concatenated data for port P1 is a sequence containing two numerical elements, such as [normalized value of 0.14, 0.7067]. A Boolean flag reflecting the current online status of the port needs to be added to the concatenated data for port P1. The online status flag is determined by the presence or absence of an optical signal at the port. It can be understood that if port P1 detects a valid optical signal, the online status flag is set to logical "true" or the value 1; if there is no optical signal, it is set to logical "false" or the value 0. Assuming port P1 is online, its final single-port status data becomes a sequence containing three elements, such as [normalized optical power deviation, normalized temperature, 1].

[0101] In some embodiments, the above operations are performed on all ports within the intelligent communication optical cable junction box. For example, if the normalized value of the optical power deviation of port P2 is -0.3, the normalized value of the connector operating temperature is 0.5, and the online status is "true", then the status data of port P2 is [-0.3, 0.5, 1]. The spliced ​​data of all ports are summarized. Assuming there are 24 ports in the junction box, the status data of ports P1, P2, up to P24 are arranged in rows to generate a 24-row, 3-column numerical matrix. This matrix is ​​the port operating status vector. Each row of the port operating status vector strictly corresponds to the status data of one port, and its dimension is consistent with the number of ports.

[0102] In one embodiment of the present invention, the port operating state vectors of all ports are input into a pre-trained fiber optic link evaluation model, which outputs an aging warning level and a fault probability score for each port. The fiber optic link evaluation model includes a feature extraction sub-model, which processes the input port operating state vectors to extract high-dimensional state features. These high-dimensional state features are input into a classification decision sub-model, which classifies the current port state based on classification boundaries trained from historical fault data, categorizing the port state into three types: normal state, aging state, and fault state. The classification decision sub-model contains multiple pre-trained decision rules, each corresponding to a combination of different dimensional feature values ​​in the port operating state vector. Each component of the high-dimensional state features is sequentially matched with the conditions in the decision rules. When a port's state features satisfy all conditions of a certain decision rule, the port is determined to be in the state type corresponding to that decision rule. If a port's state features fail to match any defined decision rule, the port's state type is marked as unknown and enters a manual verification queue. The classification results are bound to port identifiers to form a preliminary port classification result list. For ports classified as aging, an aging warning level is calculated based on the position of their state characteristics in the aging state feature space, with a value between zero and one hundred. For all ports, a regression prediction sub-model predicts a failure probability score based on the state characteristics; this score represents the probability of a port failing within a specified future time period. The model ultimately outputs the state classification, aging warning level, and failure probability score for each port.

[0103] For ports whose aging warning level reaches a preset threshold or whose fault probability score exceeds a preset threshold, a port maintenance task instruction is generated. The aging warning level threshold and fault probability score threshold are set. All ports are traversed, and ports with an aging warning level greater than or equal to the aging warning level threshold, and ports with a fault probability score greater than or equal to the fault probability score threshold are selected. The identifier, aging warning level, fault probability score, and current online status of the selected ports are encapsulated into a maintenance task data structure. Based on the port's geographical location identifier, the maintenance task data structure is matched with the information of maintenance personnel responsible for that area and available spare parts inventory information. The encapsulated maintenance task data structure, the matched maintenance personnel information, and the spare parts inventory information are integrated to generate a port maintenance task instruction containing complete dispatch elements, and stored in the pending task queue.

[0104] In practical implementation, assuming the port operating state vector of the intelligent communication optical cable junction box is a 24x3 matrix, the feature extraction sub-model of the optical fiber link evaluation model first processes this matrix to extract the high-dimensional state features corresponding to each port. In practice, these high-dimensional state features are input into the classification decision sub-model, which contains multiple trained decision rules. Each decision rule corresponds to a combination of different dimensional feature values ​​in the port operating state vector. For example, one decision rule might have the condition "normalized optical power deviation > 0.7 and normalized temperature > 0.8," while another rule might have the condition "normalized optical power deviation < -0.5 and online status is false." The system sequentially matches each component of the high-dimensional state features of port P1 with the conditions of all decision rules in the classification decision sub-model. When the state features of port P1 satisfy all the conditions of a certain decision rule, port P1 is determined to be the state type corresponding to that decision rule. Assuming the state characteristics of port P1 match the rule with the conditions "normalized optical power deviation > 0.1 and normalized temperature > 0.6", and this rule is preset to correspond to "aging state", then port P1 is classified as aging state. If the state characteristics of a port fail to match any of the defined decision rules, the state type of that port is marked as unknown and enters the manual verification queue. The classification results of all ports are bound to the port identifier to form a preliminary port classification result list.

[0105] In some embodiments, for ports classified as aging, the fiber optic link evaluation model calculates the aging warning level based on the position of its state characteristics in the aging state characteristic space. The aging warning level is a value between zero and one hundred, calculated based on the distance between the state characteristics and the aging state cluster center. Optionally, the calculation formula is as follows:

[0106]

[0107] Where: the character L in the formula a The character 'd' represents the calculated aging warning level; the character 'd' represents the Euclidean distance in the feature space from the high-dimensional state feature vector of the current port to the preset "fully aged" state reference point; the character 'd'... max The Euclidean distance from the "normal" state reference point to the "fully aged" state reference point is used as a normalized baseline. The aging warning level calculated for port P1 based on its state characteristics might be 65. For all ports, the regression prediction sub-model in the fiber optic link evaluation model predicts a fault probability score based on high-dimensional state characteristics. The fault probability score is the probability value, expressed as a percentage, of a port failing within a specified future time period. The model ultimately outputs the state classification, aging warning level, and fault probability score for each port. A simplified output is shown in Table 1.

[0108] Table 1: Port Status Evaluation Results

[0109]

[0110] In practical implementation, generating port maintenance task instructions based on the output of the fiber optic link evaluation model requires first setting aging warning level thresholds and fault probability score thresholds. Assume the aging warning level threshold is set to 60, and the fault probability score threshold is set to 15%. The system iterates through the evaluation results of all ports, filtering out ports with aging warning levels greater than or equal to the aging warning level thresholds, and ports with fault probability scores greater than or equal to the fault probability score thresholds. According to the example above, port P1's aging warning level of 65 is greater than the threshold of 60, and port P3's fault probability score of 85% is greater than the threshold of 15%, therefore ports P1 and P3 are filtered out. It can be understood that port P2 is not within the filtering range because neither of its two indicators exceeds the threshold.

[0111] In some embodiments, the identifier, aging warning level, fault probability score, and current online status of the selected ports are encapsulated into a maintenance task data structure. For example, the data structure generated for port P1 includes the fields: Port Identifier = P1, Aging Warning Level = 65, Fault Probability Score = 18%, Online Status = True. Subsequently, the system matches the maintenance task data structure with the information of the maintenance personnel responsible for that area and the available spare parts inventory information based on the port's geographical location identifier. For example, if port P1 is located in area A, the system matches the maintenance personnel responsible for area A, Zhang San, and his contact information, and queries the inventory of fiber optic connectors compatible with port P1. The encapsulated maintenance task data structure, the matched maintenance personnel information, and the spare parts inventory information are integrated to generate a port maintenance task instruction containing complete dispatch elements, and stored in the task queue to be executed. It can be understood that a complete port maintenance task instruction not only includes the identifier and type of the faulty port, but also includes the personnel and material information required to perform the maintenance.

[0112] In one embodiment of the present invention, the fiber optic link evaluation model is constructed as follows: Historical operational data is collected, including port optical power values, connector operating temperature values, corresponding port status labels, and fault records for multiple fiber optic connectors over multiple time periods, forming an original training sample set. The port optical power values ​​and connector operating temperature values ​​in the original training sample set are preprocessed to calculate historical optical power deviation values, which are then normalized to construct a historical port operating state vector. Using the historical port operating state vector as input features and the corresponding port status labels as supervision signals, a deep neural network model is trained. This deep neural network model includes a feature extraction layer for extracting high-dimensional state features, a classification layer for port status classification, and a regression output layer for calculating fault probabilities. The trained deep neural network model is validated and optimized using a labeled validation set until the model's classification accuracy and regression error on the validation set reach preset performance indicators, thereby obtaining the trained fiber optic link evaluation model.

[0113] The method also includes a data model update step based on port maintenance results. After maintenance personnel complete on-site maintenance according to the port maintenance task instructions, they receive maintenance result feedback data, which includes a description of maintenance measures, identification of replaced spare parts, and the optical power and temperature values ​​of the port after maintenance. Based on the maintenance measure description, the fault root cause classification of the repaired port is determined. The port's status data for a period of time after maintenance is extracted from the real-time monitoring data stream of the junction box, and its status stability index is calculated. The maintenance result feedback data, fault root cause classification, post-maintenance port status stability index, and historical port status data before maintenance are combined to form a model training sample. Using a new dataset containing this model training sample, the fiber optic link evaluation model is incrementally trained, updating the internal parameters of the fiber optic link evaluation model to improve the model's accuracy in identifying similar faults in the future. During incremental training, the current version of the fiber optic link evaluation model and its model parameters are loaded from the model storage area. The collected multiple model training samples are divided into incremental training sets and validation sets according to a preset ratio. The data from the incremental training set is input into the fiber optic link evaluation model, and the backpropagation algorithm is used to calculate the loss function value between the model's predicted output and the true label. The gradient of the model parameters is calculated based on the loss function value, and the internal parameters of the model are slightly adjusted using the optimizer. The performance of the model after parameter adjustment is evaluated on the validation set. If the performance meets the preset improvement criteria, the adjusted model parameters are saved as the new version of the model parameters. The updated fiber optic link evaluation model is then redeployed into the data processing flow of the intelligent communication fiber optic cable junction box.

[0114] In practical implementation, constructing a fiber optic link evaluation model first requires collecting historical operational data. This data includes port optical power values, connector operating temperatures, corresponding port status labels, and fault records for multiple fiber optic connectors over multiple time periods, collectively forming the original training sample set. For example, the port optical power and connector operating temperature values ​​recorded daily for port P5 over the past three months are collected, while maintenance records indicate that port P5 experienced a fault at the end of the third month. This data constitutes the original training sample sequence for port P5. In practical implementation, the port optical power and connector operating temperature values ​​in the original training sample set are preprocessed to calculate the historical optical power deviation value. After normalization, a historical port operating status vector is constructed. At a certain historical moment T1, the port optical power value of port P5 is -3.1 dBm, and its reference optical power value is -2.5 dBm, resulting in a calculated historical optical power deviation value of -0.6 dBm. After normalization, this yields a numerical feature. At the same moment T1, the connector operating temperature is 45 degrees Celsius, which, after normalization, yields another numerical feature. These two normalized features, together with the port online status identifier, constitute the historical port operating status vector of port P5 at time T1.

[0115] In some embodiments, a historical port operating state vector constructed from a large number of ports over a large number of historical time points is used as input features, and the port state labels recorded at corresponding time points are used as supervision signals to train a deep neural network model. For example, port state labels can be denoted as "0" for normal state, "1" for aging state, and "2" for fault state. Optionally, a simplified historical data sample is shown in Table 2.

[0116] Table 2: Historical Port Status and Tag Correspondence Table

[0117]

[0118] The deep neural network model comprises a feature extraction layer for extracting high-dimensional state features, a classification layer for port state classification, and a regression output layer for calculating failure probabilities. The goal of model training is to make the output of the classification layer as close as possible to the true port state labels, while simultaneously making the output of the regression output layer as close as possible to the true failure probabilities (if any). In practice, a labeled validation set is used to validate and fine-tune the trained deep neural network model. The validation set consists of another set of historical port operating state vectors and their true labels that were not involved in the training. Model performance is evaluated by calculating the classification accuracy and regression error on the validation set. For example, if the model achieves a 95% accuracy rate in port state classification on the validation set and the mean squared error of regression prediction is less than 0.01, it is considered to have met the preset performance indicators, thus obtaining a successfully trained fiber optic link evaluation model.

[0119] In practice, the data model update step based on port maintenance results is initiated after maintenance is completed. After maintenance personnel complete the on-site maintenance of port P5 according to the port maintenance task instructions, the system receives maintenance result feedback data. The maintenance result feedback data includes the maintenance measure description "replace fiber optic connector", the replacement spare part identification "FC-001A", and the optical power value of the port after maintenance (-2.6dBm) and temperature value (38 degrees Celsius). Based on the maintenance measure description "replace fiber optic connector", the root cause of the fault in the repaired port P5 is classified as "connector physical aging". In some embodiments, the status data of port P5 within 24 hours after maintenance is extracted from the real-time monitoring data stream of the junction box, and its status stability index is calculated. The status stability index can be the standard deviation of optical power and temperature during this period. For example, the standard deviation of optical power of port P5 after maintenance is calculated to be 0.05dBm, and the standard deviation of temperature is 0.5 degrees Celsius. The maintenance result feedback data, the fault root cause classification, the status stability index of the port after maintenance, and the historical status data of the port before maintenance are combined to form a model training sample. Understandably, this new sample contains information about the entire process from fault occurrence, diagnosis, repair to recovery.

[0120] In practice, the fiber optic link evaluation model is incrementally trained using a new dataset containing model training samples. The current version of the fiber optic link evaluation model and its parameters are loaded from the model storage area. Multiple collected model training samples are divided into incremental training and validation sets according to a preset ratio; for example, 80 of the 100 newly collected maintenance feedback samples are used for the incremental training set, and 20 for the validation set. The data from the incremental training set is input into the fiber optic link evaluation model, and the backpropagation algorithm is used to calculate the loss function value between the model's predicted output and the true label. Optionally, the calculation of the loss function value L may involve the following relationship:

[0121]

[0122] Where: L in the formula represents the loss function value, taking the mean squared error as an example; N represents the number of samples in the incremental training set; yi represents the true label or true value of the i-th sample; and so on. This represents the predicted output value of the fiber optic link evaluation model for the i-th sample. The gradient of the model parameters is calculated based on the loss function value, and the internal parameters of the model are slightly adjusted using an optimizer. The performance of the model after parameter adjustment is evaluated on the validation set. If the performance meets a preset improvement standard, such as an improvement of more than 0.5% in classification accuracy on the validation set compared to before adjustment, the adjusted model parameters are saved as the new version of the model parameters. The updated fiber optic link evaluation model is redeployed into the data processing flow of the intelligent communication fiber optic cable junction box for subsequent status evaluation.

[0123] In one embodiment of the present invention, the method further includes a trend analysis step for historical port data. The optical power deviation value and connector operating temperature value of each port are stored at a preset period to form a historical port data sequence. For each port's historical data sequence, the average rate of change of the optical power deviation value over a long time window is calculated as the optical power degradation rate. The average rate of change of the connector operating temperature value over the same long time window is calculated as the temperature change rate. The optical power degradation rate and the temperature change rate are combined to form a long-term trend vector for the port. The long-term trend vector is input into a trend analysis model, which outputs a predicted value for the remaining service life of the port. The predicted value for the remaining service life is associated with and stored in conjunction with the current port's aging warning level and failure probability score to enrich the basis for port maintenance decisions. The trend analysis model incorporates multiple typical fiber optic port failure mode curves, each describing the evolution of parameters such as optical power and temperature over time under a specific failure mode. The long-term trend vector of the current port is compared with each of the incorporated typical fiber optic port failure mode curves for similarity calculation. The failure mode curve with the highest similarity is selected as the matching failure mode for the current port. The standard time length from the current state to complete failure corresponding to the matching failure mode curve is read. The standard time length is scaled and corrected based on the magnitude of change in the long-term trend vector to calculate the specific remaining useful life prediction.

[0124] In specific implementation, the optical power deviation value and connector operating temperature value of each port are stored at a preset period to form a port historical data sequence. For example, for port P7 in the intelligent communication optical cable junction box, the system records the optical power deviation value and connector operating temperature value of port P7 once a day, and stores them continuously for 90 days, forming a port P7 historical data sequence containing 90 time points. In some embodiments, for the port historical data sequence of port P7, the average rate of change of the optical power deviation value over a long period of time is calculated as the optical power degradation rate. The long period of time is set to the most recent 60 days. The optical power deviation value sequence of port P7 within these 60 days is extracted, and the linear regression slope of the sequence is calculated. This slope is the optical power degradation rate. Assume that the calculated optical power degradation rate of port P7 is 0.008 per day. At the same time, the average rate of change of the connector operating temperature value of port P7 over the same 60-day long period of time is calculated as the temperature change rate. The calculation method is also to obtain the linear regression slope of the temperature value sequence. Assume that the calculated temperature change rate of port P7 is 0.02 degrees Celsius per day.

[0125] In specific implementation, the optical power degradation rate and temperature change rate are combined to form the long-term trend vector of the port. The long-term trend vector of port P7 is a two-dimensional vector [0.008, 0.02]. It can be understood that the long-term trend vector describes the direction and magnitude of the port's key parameters changing over time. The long-term trend vector is input into a trend analysis model, which outputs a predicted value for the port's remaining service life. The predicted value for the remaining service life is associated with and stored in relation to the current port's aging warning level and failure probability score to enrich the basis for port maintenance decisions. In some embodiments, the trend analysis model incorporates several typical fiber optic port failure mode curves. Each failure mode curve describes the evolution of parameters such as optical power and temperature over time under a specific failure mode. For example, failure mode curve A describes the "gradual decrease in optical power due to slow contamination" mode, with an optical power degradation rate of approximately 0.005 per day and a temperature change rate of approximately 0.01 per day; failure mode curve B describes the "rapid degradation due to connector mechanical loosening" mode, with an optical power degradation rate of approximately 0.015 per day and a temperature change rate of approximately 0.05 per day.

[0126] In practice, the long-term trend vector of the current port is compared with the built-in curves for each typical fiber optic port failure mode. Optionally, the similarity calculation can use the cosine similarity method, with the following relationship:

[0127]

[0128] Where: the character S in the formula cos Represents the calculated cosine similarity value; characters Represents the long-term trend vector of port P7; character This represents the vector representation of a typical fiber optic port failure mode curve built into the trend analysis model. The similarity between the long-term trend vector [0.008, 0.02] of port P7 and the failure mode curve A vector [0.005, 0.01] is calculated, along with the similarity between it and the failure mode curve B vector [0.015, 0.05]. The failure mode curve with the highest similarity is selected as the matching failure mode for the current port. Assuming the similarity between the long-term trend vector of port P7 and failure mode curve A is 0.98, and the similarity with failure mode curve B is 0.85, failure mode curve A is selected as the matching failure mode for port P7. The standard time length from the current state to complete failure corresponding to the matching failure mode curve A is read, assuming the standard time length is 300 days. The standard time length is scaled and corrected according to the change amplitude of the long-term trend vector to calculate the specific remaining service life prediction value. It can be understood that the scaling correction factor can be the ratio of the standard rate of change amplitude of the matching failure mode curve to the amplitude of the long-term trend vector of the current port. Assuming the rate of change amplitude of the standard failure mode curve A is... The amplitude of the long-term trend vector of port P7 is The correction factor is approximately the ratio of the standard amplitude to the current amplitude, and the calculated predicted remaining service life of port P7 is likely to be 180 days. This predicted value will be stored along with the real-time aging warning level and failure probability score of port P7, providing a longer-term trend basis for maintenance decisions.

[0129] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A data processing method for an intelligent communication optical cable junction box, characterized in that, The method includes: The optical power values ​​of the ports of multiple fiber optic connectors installed in the intelligent communication optical cable junction box are collected. The multiple fiber optic connectors are connected to external optical cable lines to form multiple optical signal transmission channels. The connector operating temperature value is obtained from the temperature sensor deployed corresponding to the plurality of fiber optic connectors. The connector operating temperature value is the measured temperature of the fiber optic connector end face area. The optical power value of the port is compared with the preset reference optical power value to calculate the optical power deviation value of each port. The reference optical power value is the calibrated optical power of each port under standard operating conditions. The optical power deviation value is combined with the connector operating temperature value to form a port operating status vector that reflects the state of the optical fiber link. The number of port operating status vectors is the same as the number of ports. The port operating state vectors of all ports are input into a pre-trained fiber optic link evaluation model, which outputs the aging warning level and failure probability score for each port, including: The fiber optic link evaluation model includes a feature extraction sub-model, which processes the input port operating state vector to extract high-dimensional state features. The high-dimensional state features are input into a classification decision sub-model. The classification decision sub-model classifies the current port state based on the classification boundary obtained by training on historical fault data, and divides the port state into three types: normal state, aging state, and fault state. For ports classified as aging states, the aging warning level is calculated based on the position of their state characteristics in the aging state characteristic space. The aging warning level is a value between zero and one hundred. For all ports, another regression prediction sub-model predicts the failure probability score based on state characteristics. The failure probability score is the probability value of a port failing within a specified time period in the future. Output the status classification, aging warning level, and fault probability score for each port; The fiber optic link evaluation model is constructed in the following way: Collect historical operating data, which includes port optical power values, connector operating temperature values, corresponding port status labels, and fault records of multiple fiber optic connectors at multiple time periods, to form the original training sample set; The port optical power values ​​and connector operating temperature values ​​in the original training sample set are preprocessed to calculate the historical optical power deviation values, and after normalization, a historical port operating state vector is constructed. Using the historical port working state vector as input features and the corresponding port state label as supervision signal, a deep neural network model is trained. The deep neural network model includes a feature extraction layer for extracting high-dimensional state features, a classification layer for port state classification, and a regression output layer for calculating fault probability. The trained deep neural network model is validated and tuned using a labeled validation set until the model’s classification accuracy and regression error on the validation set reach the preset performance indicators, thereby obtaining the trained fiber optic link evaluation model. For ports whose aging warning level reaches a preset threshold or whose fault probability score exceeds a preset threshold, a port maintenance task instruction is generated, which includes a port identifier and a fault type code. The method also includes a trend analysis step on historical port data: The port optical power deviation value and the connector operating temperature value of each port are stored according to a preset period to form a port historical data sequence; For each port's historical data sequence, the average rate of change of the optical power deviation value over a long time window is calculated as the optical power degradation rate. Calculate the average rate of change of the joint's operating temperature value over the same long time window, as the temperature change rate; The optical power degradation rate is combined with the temperature change rate to form a long-term trend vector for the port; The long-term trend vector is input into a trend analysis model, and the trend analysis model outputs the remaining useful life prediction value of the port. The remaining service life prediction value is associated with and stored in conjunction with the current port's aging warning level and failure probability score to enrich the basis for port maintenance decisions; The step of inputting the long-term trend vector into a trend analysis model, wherein the trend analysis model outputs a predicted remaining useful life value, includes: The trend analysis model incorporates multiple typical fiber optic port failure mode curves. Each failure mode curve describes the evolution of optical power and temperature parameters over time under each failure mode. The similarity is calculated between the long-term trend vector of the current port and the built-in curves of each of the typical fiber optic port failure modes. Select the failure mode curve with the highest similarity as the matching failure mode for the current port; Read the standard time length from the current state to complete failure corresponding to the matching failure mode curve; The standard time length is scaled and corrected based on the change range of the long-term trend vector to calculate the specific predicted value of the remaining useful life. The intelligent communication optical cable junction box includes: An optical power monitoring array, housed inside the enclosure and electrically connected to the main control module, is used to collect the port optical power value of each of the multiple fiber optic connectors. A temperature sensor array is deployed corresponding to the multiple fiber optic connectors and electrically connected to the main control module to collect the working temperature value of the connector end face area of ​​each fiber optic connector.

2. The data processing method for an intelligent communication optical cable junction box according to claim 1, characterized in that, The optical power value of each port is compared with a preset reference optical power value to calculate the optical power deviation value of each port, including: Read the standard operating parameters for each port from the junction box configuration database. The standard operating parameters include the reference optical power value and its allowable fluctuation range. Extract the current port optical power value from the real-time acquired data sequence; Calculate the difference between the port optical power value and the corresponding reference optical power value to generate a preliminary optical power deviation value; Determine whether the preliminary optical power deviation value exceeds the allowable fluctuation range. If it does, mark the preliminary optical power deviation value as a valid deviation value. The effective deviation value is divided by the reference optical power value to calculate the normalized relative optical power deviation value, which is then output as the optical power deviation value.

3. The data processing method for an intelligent communication optical cable junction box according to claim 1, characterized in that, The optical power deviation value is combined with the connector operating temperature value to form a port operating state vector reflecting the fiber optic link status, including: The optical power deviation value is linearly normalized to map its value to a standard value range. The working temperature value of the joint is normalized by subtracting the standard temperature reference value and then dividing by the length of the standard temperature range. The normalized optical power deviation value and the normalized connector operating temperature value are spliced ​​together in port number order. Add a Boolean flag reflecting the current online status of the port to the spliced ​​data. The online status flag is determined by the presence or absence of the port's optical signal. The concatenated data from all ports is aggregated to generate a multi-row, multi-column matrix. This matrix is ​​the working status vector of the ports, where each row corresponds to the status data of one port.

4. The data processing method for an intelligent communication optical cable junction box according to claim 3, characterized in that, The high-dimensional state features are input into a classification decision sub-model. This sub-model classifies the current port state based on classification boundaries obtained from training with historical fault data, categorizing the port state into three types: normal state, aging state, and fault state. The classification decision sub-model contains multiple trained decision rules, each decision rule corresponding to a combination of feature values ​​of different dimensions in the port working state vector; Each component of the high-dimensional state feature is matched sequentially with the conditions in the decision rule; When the state characteristics of a port meet all the conditions of a certain decision rule, the port is determined to be a state type corresponding to the decision rule. The state type is one of the normal state, aging state, and fault state. If the state characteristics of a port fail to match any of the defined decision rules, the port's state type is marked as unknown and enters the manual verification queue. The classification results are bound to the port identifiers to form a preliminary list of port classification results.

5. The data processing method for an intelligent communication optical cable junction box according to claim 4, characterized in that, For ports whose aging warning level reaches a preset threshold or whose fault probability score exceeds a preset threshold, a port maintenance task instruction is generated, including: Set aging warning level thresholds and fault probability score thresholds; Traverse all ports and filter out ports whose aging warning level is greater than or equal to the aging warning level threshold, and ports whose fault probability score is greater than or equal to the fault probability score threshold. The identifier of the selected port, the aging warning level, the fault probability score, and the current online status of the port are encapsulated into a maintenance task data structure; Based on the geographical location identifier of the port, match the maintenance task data structure with the maintenance personnel responsible for the area and the available spare parts inventory information; The encapsulated maintenance task data structure, the matched maintenance personnel information, and the spare parts inventory information are integrated to generate a port maintenance task instruction containing complete dispatch elements, and stored in the task queue to be executed.

6. The data processing method for an intelligent communication optical cable junction box according to claim 5, characterized in that, The method also includes a data model update step based on port maintenance results: After the maintenance personnel complete the on-site maintenance according to the port maintenance task instruction, they receive maintenance result feedback data. The maintenance result feedback data includes a description of the maintenance measures, the identification of the replaced spare parts, and the optical power value and temperature value of the port after maintenance. Based on the description of the maintenance measures, determine the classification of the root causes of the faults in the port being maintained; Extract the status data of the port within a certain period after maintenance from the real-time monitoring data stream of the junction box, and calculate its status stability index. The maintenance result feedback data, the fault root cause classification, the post-maintenance port status stability index, and the pre-maintenance port historical status data are combined to form a model training sample. The fiber optic link evaluation model is incrementally trained using a new dataset containing the training samples of the model, and the internal parameters of the fiber optic link evaluation model are updated to improve the model's accuracy in identifying similar faults in the future. The step of incrementally training the fiber optic link evaluation model using a new dataset containing the model's training samples and updating the model's internal parameters includes: Load the current version of the fiber optic link evaluation model and its parameters from the model storage area; The collected training samples of the model are divided into an incremental training set and a validation set according to a preset ratio; The data from the incremental training set is input into the fiber optic link evaluation model, and the backpropagation algorithm is used to calculate the loss function value between the model's predicted output and the true label. The gradient of the model parameters is calculated based on the loss function value, and the optimizer is used to make minor adjustments to the model's internal parameters. The performance of the model after parameter adjustment is evaluated on the validation set. If the performance meets the preset improvement standard, the adjusted model parameters are saved as the model parameters of the new version. The updated fiber optic link evaluation model is redeployed into the data processing flow of the smart communication fiber optic cable junction box.

7. A smart communication optical cable junction box, characterized in that, It includes a housing, a main control module, an optical power monitoring array, a temperature sensor array, a data storage module, and a communication module; The enclosure is equipped with a port panel for installing multiple fiber optic connectors, which are used to connect with external optical cable lines to form multiple optical signal transmission channels. The optical power monitoring array is installed inside the enclosure and electrically connected to the main control module, and is used to collect the port optical power value of each of the plurality of optical fiber connectors; The temperature sensor array is arranged corresponding to the plurality of fiber optic connectors and electrically connected to the main control module, and is used to collect the joint working temperature value of the end face area of ​​each fiber optic connector. The data storage module is electrically connected to the main control module and is used to store preset reference optical power values, junction box configuration database, pre-trained fiber optic link evaluation model, and preset thresholds. The main control module is configured to execute a data processing method for an intelligent communication optical cable junction box as described in any one of claims 1 to 6.

Images