Method for predicting degradation of underwater sealing performance of armored optical and electrical composite cable based on data analysis

Abstract

The present invention relates to the technical field of data analysis, and particularly to a method for predicting the degradation of the underwater sealing performance of an armored fiber-optic composite cable based on data analysis, comprising the following steps: S1: Collecting real-time operation data during the operation of the armored fiber-optic composite cable through sensors; S2: Preprocessing the operation data collected in S1 to generate a structured preprocessing data set; S3: Extracting feature parameters related to the degradation of the sealing performance; S4: Training a Gaussian process regression model; S5: Verifying the Gaussian process regression model generated in S4; S6: Adjusting the kernel function parameters and noise terms in the Gaussian process regression model; S7: Generating a degradation report. With the present invention, through real-time data collection, feature extraction, and accurate prediction of the Gaussian process regression model, the efficient prediction of the degradation time point, rate, and trend of the sealing performance of the armored fiber-optic composite cable is achieved, improving the scientific nature of equipment maintenance and the reliability of operation.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and particularly to a method for predicting the degradation of the underwater sealing performance of armored fiber optic composite cables based on data analysis. Background Art

[0002] Armored fiber optic composite cables are widely used in deep-sea exploration, marine communication, and underwater engineering. Their main functions are to achieve data transmission, power transmission, and mechanical protection. However, the fiber optic composite cables are in a complex marine environment for a long time, such as high pressure, low temperature, high humidity, and salt spray. These environmental factors will continuously erode and degrade the sealing performance of the cable body. The decline in sealing performance may lead to problems such as water infiltration, fiber optic signal attenuation, and circuit short circuit, which will further threaten the safety and reliability of the entire system. Traditional maintenance methods rely on regular physical inspections and empirical judgments. This method not only consumes time and effort but also may cause serious economic losses and safety hazards due to untimely inspections or inaccurate judgments.

[0003] The existing technology lacks an effective prediction method for the degradation trend of the sealing performance of armored fiber optic composite cables. Usually, it is impossible to identify potential degradation risks in advance and take preventive measures. On the one hand, the existing data processing methods are mostly limited to simple statistical analysis and fail to fully utilize the characteristic information of the operation data. On the other hand, the construction of the prediction model fails to effectively model the non-linear change characteristics of the sealing performance, resulting in insufficient prediction accuracy. Due to the lack of efficient and accurate prediction means, it is difficult to scientifically formulate maintenance and replacement plans, which is likely to cause waste of resources or delay in maintenance. Summary of the Invention

[0004] Based on the above objectives, the present invention provides a method for predicting the degradation of the underwater sealing performance of armored fiber optic composite cables based on data analysis.

[0005] The method for predicting the degradation of the underwater sealing performance of armored fiber optic composite cables based on data analysis includes the following steps:

[0006] S1: Collect real-time operation data during the operation of the armored fiber optic composite cable through sensors, including temperature, pressure, fiber optic signal attenuation value, and current.

[0007] S2: Preprocess the operation data collected in S1 and perform normalization operations on all data to generate a structured preprocessed data set.

[0008] S3: Based on the preprocessed data set in S2, extract characteristic parameters related to the degradation of the sealing performance, including the temperature change rate, pressure fluctuation amplitude, and fiber optic signal attenuation rate, and form a characteristic vector set.

[0009] S4: Using the set of feature vectors extracted in S3, train a Gaussian process regression model for predicting the trend of sealing performance degradation.

[0010] S5: Validate the Gaussian process regression model generated in S4 using an independent validation dataset. Input the feature vectors of the independent validation set into the model, output the prediction results, and compare them with the actual sealing performance annotation values to calculate the accuracy and error of the initial prediction model.

[0011] S6: Based on the accuracy and error results output in S5, adjust the kernel function parameters and noise terms in the Gaussian process regression model to optimize the prediction accuracy of the Gaussian process regression model.

[0012] S7: Input the data preprocessed in S2 into the Gaussian process regression model optimized in S6 to generate the prediction results of the sealing performance degradation of the armored fiber optic composite cable, and generate a degradation report based on the prediction results.

[0013] Optionally, S1 specifically includes:

[0014] S11: Install a digital temperature sensor at the connection between the outer wall of the sealing cavity of the armored fiber optic composite cable and the joint. The digital temperature sensor uses a platinum resistance element and outputs a linear voltage signal. The temperature data is collected after being processed by an analog-to-digital conversion circuit, and the sampling frequency is once per second.

[0015] S12: Install a piezoresistive pressure sensor inside the sealing cavity of the armored fiber optic composite cable. The piezoresistive pressure sensor specifically uses a silicon piezoresistive chip, with a measuring range of 0 - 10 MPa and an accuracy of ±0.1% FS. The pressure signal is converted into a standard voltage signal by a signal conditioning circuit and then collected, and the sampling frequency is once per second.

[0016] S13: Lay multimode optical fibers along the armored fiber optic composite cable, and use a distributed fiber optic sensing system and the Raman scattering principle to collect the attenuation values of the fiber optic signals. The collection points include 10 meters, 20 meters, and 50 meters from both ends of the fiber optic, and the sampling frequency is once per minute.

[0017] S14: Install a Hall effect current sensor on the power transmission conductor of the armored fiber optic composite cable. The Hall effect current sensor is located on the conductor section near the connection terminal and is used to monitor the real-time current flowing through the entire power transmission line; the sampling frequency is once per second.

[0018] Optionally, S2 specifically includes:

[0019] S21: Denoise the real-time operation data collected in S1.

[0020] S22: Detect and remove outliers from the denoised data using statistical methods.

[0021] S23: For the missing data remaining after outlier removal, interpolation method is used for completion;

[0022] S24: Normalize the completed data to convert data with different dimensions to the same scale range;

[0023] S25: Arrange the normalized data according to the time series to form a structured preprocessed data set, specifically in matrix form, where each row represents the observed data at a time point and each column represents a type of sensor data; and add the corresponding timestamp to each row of data to ensure the timeliness of the data.

[0024] Optionally, the S3 specifically includes:

[0025] S31: For the temperature data at each time point in the preprocessed data set, calculate the temperature change rate according to the following formula: , where and are the temperature values at adjacent time points respectively, is the time interval, is the temperature change rate; at the same time, summarize the temperature change rates calculated at all time points into time series data and record it as a column in the feature vector;

[0026] S32: Calculate the pressure fluctuation amplitude. For the pressure data in the preprocessed data set, take one hour as a fixed time window and calculate the maximum value and the minimum value of the pressure within the time window, and calculate the pressure fluctuation amplitude according to the following formula: , where is the pressure fluctuation amplitude of the time window; at the same time, summarize the pressure fluctuation amplitudes within all time windows into time series data and record it as a column in the feature vector;

[0027] S33: Calculate the optical fiber signal attenuation rate. For the optical fiber signal attenuation values in the preprocessed data set, take 10 minutes as a fixed time window and calculate the optical fiber signal attenuation rate according to the following formula: , where and are the optical fiber signal intensity values at the start time and the end time of the window respectively, is the optical fiber signal attenuation rate; at the same time, summarize the optical fiber signal attenuation rates within all time windows into time series data and record it as a column in the feature vector;

[0028] S34: Combine the calculated time series data of the temperature change rate, the pressure fluctuation amplitude, and the optical fiber signal attenuation rate to form a feature vector set; the format of the feature vector set adopts a matrix form, where each row represents the features of a time window, and the columns respectively correspond to the temperature change rate, the pressure fluctuation amplitude, and the optical fiber signal attenuation rate.

[0029] Optionally, the specific steps of S4 are as follows:

[0030] S41: Construct a training data set. From the feature vector set generated in S3, select the feature vectors that include the temperature change rate, the pressure fluctuation amplitude, and the optical fiber signal attenuation rate, and attach the corresponding sealing performance level labels to each feature vector. The specific level labels include normal, mild degradation, and severe degradation.

[0031] S42: According to the feature change pattern of the training data set, select the corresponding kernel function to construct a Gaussian process regression model; if the feature change is smooth, select the radial basis function kernel that can capture the smooth characteristics; if the feature change is irregular, select the Matern kernel that can adapt to the non-smooth characteristics.

[0032] S43: Set the initial kernel function parameters and noise term parameters, where the kernel function parameters are used to control the range of feature correlation, and the noise term parameters are used to describe the measurement noise in the data.

[0033] S44: Train the Gaussian process regression model by maximizing the marginal likelihood function of the training data. During the training process, calculate the correlation between feature vectors according to the feature matrix, and adjust the kernel function parameters and noise term parameters in combination with the sealing performance level labels.

[0034] S45: After the training is completed, save the Gaussian process regression model.

[0035] Optionally, the specific steps of S5 are as follows:

[0036] S51: Prepare an independent verification data set. Screen out the independent verification data set from the historical operation data that has not participated in the model training. The data set includes feature vectors with the same structure as the training data and the corresponding sealing performance level labels.

[0037] S52: Input each feature vector in the verification data set into the Gaussian process regression model generated in S4 one by one; output the prediction results, that is, the sealing performance level corresponding to each sample.

[0038] S53: Record and save the predicted sealing performance level of the model for each verification sample and the actual sealing performance level label in the verification data set to form a comparison data table of the prediction results and the actual labels.

[0039] S54: Traverse the prediction results and actual labels, and compare the predicted grade and the actual grade of each sample one by one; if the predicted grade is the same as the actual grade, it is recorded as a correct prediction; if the predicted grade is different from the actual grade, it is recorded as an incorrect prediction; record the number of samples with correct predictions and incorrect predictions.

[0040] S55: According to the comparison results in S54, count the proportion of the number of samples with correct predictions in the total number of samples, and calculate the accuracy rate of the model; the accuracy rate is used to measure the ability of the model to correctly identify the sealing performance grade.

[0041] S56: Calculate the model error by statistically analyzing the degree of difference between the prediction results and the actual grade labels and quantifying it as an error index.

[0042] Optionally, S55 specifically includes:

[0043] S551: Traverse the comparison data table of the prediction results and actual labels recorded in S54, and respectively count the number of samples with consistent prediction results and inconsistent prediction results. Let be the number of samples with correct predictions, and record the number of samples where the prediction result is equal to the actual label; let be the number of samples with incorrect predictions, and record the number of samples where the prediction result is not equal to the actual label; let be the total number of samples, and the calculation formula is: ;

[0044] S552: Calculate the accuracy rate of the model according to the number of samples counted in S551, and the formula is: , where A is the accuracy rate of the model, expressed as a percentage.

[0045] Optionally, S56 specifically includes:

[0046] S561: For each sample, calculate the difference between the predicted sealing performance grade and the actual sealing performance grade. The calculation formula for the difference value is: , where is the predicted sealing performance grade of the i-th sample; is the actual sealing performance grade of the i-th sample; is the difference value of the i-th sample;

[0047] S562: Use the difference values of all samples to calculate the mean square error MSE of the model, and the formula is: , where is the total number of samples in the validation dataset; is the difference value of the i-th sample; MSE is the mean square error used to quantify the overall deviation between the prediction results and the actual labels;

[0048] S563: To analyze the error in depth, calculate the Mean Absolute Error (MAE). The formula is: , where MAE is the Mean Absolute Error, representing the average deviation magnitude of the model prediction results;

[0049] S564: Compare the absolute error with the average value of the actual grade label range to calculate the error rate. The formula is: , where E is the error rate, expressed in percentage; R is the value range of the sealing performance grade.

[0050] Optionally, the specific steps of S6 include:

[0051] S61: Based on the accuracy rate and error calculated in S5, judge the prediction performance of the model. If the accuracy rate is lower than 85%, it is considered that the model needs to be adjusted; if the mean square error is higher than 0.1 or the mean absolute error is higher than 0.1, it is considered that the model needs to be adjusted;

[0052] S62: Use an optimization algorithm to adjust the kernel function parameters in the Gaussian process regression model;

[0053] S63: Adjust the noise term of the model according to the error analysis to balance the signal and noise in the training data; the optimization of the noise term is carried out by minimizing the sum of squared residuals or maximizing the likelihood function;

[0054] S64: After adjusting the kernel function and the noise term, use the validation data set to re-evaluate the accuracy rate and error of the model: compare the accuracy rate and error before and after optimization; if the accuracy rate of the optimized model improves and the error decreases, determine the final optimization result.

[0055] Optionally, the specific steps of S7 include:

[0056] S71: Input the preprocessed data set in S2 into the Gaussian process regression model optimized in step S6;

[0057] S72: The model predicts the degradation trend of the sealing performance of the armored fiber optic composite cable according to the input feature parameters and outputs the prediction results, including the degradation time point, degradation rate, and degradation trend;

[0058] S73: Generate a degradation report according to the prediction results of the model. The degradation report includes the current sealing performance status, predicted degradation time point, degradation rate analysis, degradation trend chart, and maintenance suggestions.

[0059] Advantages of the present invention:

[0060] In the present invention, by collecting multi-dimensional operation data in real time, denoising, removing outliers and interpolating and complementing the data, a high-quality preprocessed data set is generated, and by extracting feature parameters related to the degradation of the sealing performance and combining the powerful non-linear modeling ability of the Gaussian process regression model, the degradation time point, degradation rate and trend of the sealing performance are accurately predicted, providing a scientific basis for the preventive maintenance of equipment.

[0061] In the present invention, through optimized feature extraction and modeling, the prediction accuracy and stability are improved; through the verification data set for model evaluation and parameter optimization, the applicability of the model under different operating conditions is ensured; by generating an intuitive degradation trend report, it is convenient for users to quickly understand the operating state of the equipment and formulate targeted maintenance plans; the limitations relying on manual inspection and empirical judgment are greatly reduced. Brief Description of the Drawings

[0062] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0063] Figure 1 Schematic diagram of the method for predicting the degradation of the underwater sealing performance of the composite cable according to the embodiment of the present invention;

[0064] Figure 2 Schematic diagram of the verification process of the model according to the embodiment of the present invention. Detailed Embodiments

[0065] The following will describe the present invention in detail with reference to the drawings and specific embodiments. At the same time, it should be noted here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments. For some well-known technologies, those skilled in the art can also adopt other alternative methods for implementation; moreover, the drawing part is only for more specifically describing the embodiments, and is not intended to specifically limit the present invention.

[0066] As Figure 1 - Figure 2 shown, the method for predicting the degradation of the underwater sealing performance of the armored optoelectronic composite cable based on data analysis includes the following steps:

[0067] S1: Collect real-time operation data during the operation of the armored optoelectronic composite cable through sensors, including temperature, pressure, optical fiber signal attenuation value and current;

[0068] S2: Preprocess the operation data collected in S1, including removing noise, removing outliers, interpolating and complementing missing data, and normalizing all data to generate a structured preprocessed data set;

[0069] S3: Based on the preprocessed dataset in S2, extract the characteristic parameters related to the degradation of the sealing performance, including the temperature change rate, the pressure fluctuation amplitude, and the optical fiber signal attenuation rate, and form a set of characteristic vectors;

[0070] S4: Use the set of characteristic vectors extracted in S3 to train a Gaussian Process Regression (GPR) model. The Gaussian Process Regression model models the non-linear characteristic relationships by selecting appropriate kernel functions (such as Gaussian kernel, Matern kernel, etc.). The Gaussian Process Regression model is used to predict the trend of the degradation of the sealing performance;

[0071] S5: Use the independent validation dataset to validate the Gaussian Process Regression model generated in S4. Input the characteristic vectors of the independent validation set into the model, output the prediction results, and compare them with the actual sealing performance annotation values. Calculate the accuracy and error of the initial prediction model, and evaluate the prediction ability of the model;

[0072] S6: Based on the accuracy and error results output in S5, adjust the kernel function parameters and noise terms in the Gaussian Process Regression model to optimize the prediction accuracy of the Gaussian Process Regression model and improve the prediction accuracy and stability of the degradation of the sealing performance;

[0073] S7: Input the data preprocessed in S2 into the Gaussian Process Regression model optimized in S6 to generate the prediction results of the degradation of the sealing performance of the armored fiber-optic composite cable, and generate a degradation report according to the prediction results.

[0074] S1 specifically includes:

[0075] S11: Install a digital temperature sensor at the connection between the outer wall of the sealing cavity of the armored fiber-optic composite cable and the joint. The digital temperature sensor uses a platinum resistance (Pt100) element to output a linear voltage signal. The temperature data is collected after being processed by an analog-to-digital conversion circuit, and the sampling frequency is once per second;

[0076] S12: Install a piezoresistive pressure sensor inside the sealing cavity of the armored fiber-optic composite cable. The piezoresistive pressure sensor specifically uses a silicon piezoresistive chip, with a measurement range of 0 - 10 MPa and an accuracy of ±0.1% FS. The pressure signal is collected after being converted into a standard voltage signal by a signal conditioning circuit, and the sampling frequency is once per second;

[0077] S13: Lay multimode optical fibers along the armored fiber-optic composite cable, and use a distributed optical fiber sensing system (DTS) and the Raman scattering principle to collect the attenuation values of the optical fiber signals. The collection points include 10 meters, 20 meters, and 50 meters from both ends of the optical fiber respectively, and the sampling frequency is once per minute;

[0078] S14: Install a Hall effect current sensor on the power transmission conductor of the armored optoelectronic composite cable. The Hall effect current sensor is located in the conductor segment close to the connection terminal and is used to monitor the real-time current flowing through the entire power transmission line. The sampling frequency is once per second. Through the above sub-steps, ensure the accurate collection of real-time operating data such as temperature, pressure, optical fiber signal attenuation value and current during the operation of the armored optoelectronic composite cable, and provide a reliable data basis for subsequent data processing and analysis.

[0079] S2 specifically includes:

[0080] S21: De-noise the real-time operation data collected by S1. Specifically, a sliding average filter is applied to the temperature and pressure data, and the window size is set to 5 sampling points to smooth short-term fluctuations and retain long-term trends. The wavelet denoising method is used for the fiber optic signal attenuation value, and the Daubechies wavelet basis function is selected. The number of decomposition layers is 3, and the threshold uses the soft threshold method to remove high-frequency noise. A low-pass filter is used for the current data, and the cutoff frequency is set to 50Hz to filter out high-frequency interference signals.

[0081] S22: For the denoised data, statistical methods are used to detect and remove outliers. Specifically, for temperature and pressure data, the mean and standard deviation of each group of data are calculated, and the threshold is set to mean ± 3 times the standard deviation. Values ​​outside this range are determined as outliers and removed; for the fiber signal attenuation value, the box plot method is used to calculate the interquartile range (IQR), and the upper limit is set to the upper quartile (Q3) plus 1.5 times the IQR, and the lower limit is set to the lower quartile (Q1) minus 1.5 times the IQR. Values ​​outside this range are determined as outliers and removed; for current data, a density-based clustering algorithm (such as DBSCAN) is used to identify outliers, which are determined as outliers and removed;

[0082] S23: For missing data after removing outliers, interpolation method is used to complete them; linear interpolation method is used for temperature and pressure data, and the value of the missing point is linearly estimated according to the value of the adjacent known data point; spline interpolation method is used for fiber signal attenuation value, and cubic spline function is used to fit the known data points to estimate the value of the missing point; polynomial interpolation method is used for current data, and polynomial function of appropriate order is selected to fit the known data points to estimate the value of the missing point;

[0083] S24: normalize the completed data to convert data of different dimensions to the same scale range. Specifically, a minimum-maximum normalization method is used to linearly map the data to the interval [0,1].

[0084] S25: Arrange the normalized data in time series to form a structured pre - processed data set. Specifically, use a matrix form where each row represents the observed data at a time point and each column represents a type of sensor data (temperature, pressure, optical fiber signal attenuation value, current); add the corresponding timestamp to each row of data to ensure the time sequence of the data; finally, store the structured pre - processed data set as a CSV file for subsequent analysis and model training. Through the above sub - steps, comprehensively pre - process the operation data collected in S1 to generate a high - quality structured data set, providing a reliable data basis for subsequent feature extraction and model construction.

[0085] S3 specifically includes:

[0086] S31: For the temperature data at each time point in the pre - processed data set, calculate the temperature change rate according to the following formula: , where and are the temperature values at adjacent time points respectively, is the time interval, is the temperature change rate; at the same time, summarize the temperature change rates calculated at all time points as time - series data and record it as a column in the feature vector.

[0087] S32: Calculate the pressure fluctuation amplitude. For the pressure data in the pre - processed data set, take one hour as a fixed time window and calculate the maximum value and the minimum value of the pressure within the time window, and calculate the pressure fluctuation amplitude according to the following formula: , where is the pressure fluctuation amplitude of the time window; at the same time, summarize the pressure fluctuation amplitudes within all time windows as time - series data and record it as a column in the feature vector.

[0088] S33: Calculate the optical fiber signal attenuation rate. For the optical fiber signal attenuation values in the pre - processed data set, take 10 minutes as a fixed time window and calculate the optical fiber signal attenuation rate according to the following formula: , where and are the optical fiber signal intensity values at the start time and the end time of the window respectively, is the optical fiber signal attenuation rate; at the same time, summarize the optical fiber signal attenuation rates within all time windows as time - series data and record it as a column in the feature vector.

[0089] S34: Combine the calculated time series data of the temperature change rate, pressure fluctuation amplitude, and optical fiber signal attenuation rate to form a feature vector set. The format of the feature vector set adopts a matrix form, where each row represents the features of a time window, and the columns correspond to the temperature change rate, pressure fluctuation amplitude, and optical fiber signal attenuation rate respectively. Through the above sub-steps, key feature parameters related to the degradation of the sealing performance are extracted from the S2 preprocessed dataset, and a complete feature vector set is formed, providing high-quality input data for the subsequent training of the Gaussian process regression model.

[0090] S4 specifically includes:

[0091] S41: Construct a training dataset. From the feature vector set generated in S3, select the feature vectors that include the temperature change rate, pressure fluctuation amplitude, and optical fiber signal attenuation rate, and attach the corresponding sealing performance level labels to each feature vector. The specific level labels include normal, mild degradation, and severe degradation. The training dataset includes a feature matrix and the corresponding sealing performance level labels, where the feature matrix contains parameters such as the temperature change rate, pressure fluctuation amplitude, and optical fiber signal attenuation rate.

[0092] S42: According to the feature change pattern of the training dataset, select the corresponding kernel function to construct a Gaussian process regression model. If the feature change is smooth, select the radial basis function kernel that can capture the smooth characteristics. If the feature change is irregular, select the Matern kernel that can adapt to the non-smooth characteristics. The selected kernel function is used to measure the correlation between different feature vectors.

[0093] S43: Set the initial kernel function parameters and noise term parameters. The kernel function parameters are used to control the range of feature correlation, and the noise term parameters are used to describe the measurement noise in the data.

[0094] S44: Train the Gaussian process regression model by maximizing the marginal likelihood function of the training data. During the training process, calculate the correlation between feature vectors according to the feature matrix, and adjust the kernel function parameters and noise term parameters in combination with the sealing performance level labels to optimally fit the training data.

[0095] S45: After the training is completed, save the Gaussian process regression model. The model contains the optimized kernel function parameters, noise term parameters, and the correlation information calculated during the training process, serving as the basis for subsequent prediction tasks.

[0096] The specific steps for training the Gaussian process regression model are as follows:

[0097] Prepare the training dataset: Select the feature vectors with clear sealing performance level labels from the feature vector set generated in S3 as the training data. The format of the training dataset is , where X is the feature matrix, each row contains feature parameters within a time window, and the columns represent the temperature change rate, pressure fluctuation amplitude, and optical fiber signal attenuation rate respectively; Y is the corresponding seal performance level label, in numerical form as 0 (normal), 1 (slight degradation), 2 (severe degradation);

[0098] Kernel function selection: According to the distribution characteristics of the training data, select an appropriate Gaussian process regression kernel function: When the features change smoothly over time, select the radial basis function (RBF kernel), and the formula is: , where, is the input feature vector; is the Euclidean distance between the feature vectors; is the width parameter of the kernel function, used to control the correlation range of the features; When the features show irregular changes, select the Matern kernel, and the formula is:

[0099] , where, is the input feature vector; is the Euclidean distance between the feature vectors; is the scale parameter, used to control the correlation range of the features; v is the smoothness parameter, used to control the smoothness characteristics of the kernel function; is the gamma function, defined as ; is the Bessel function, defined as the Bessel function of order v;

[0100] Set the initial model parameters: The kernel function parameter or , set the initial value according to the standard deviation of the feature data; The noise term parameter , estimate the initial value according to the measurement noise in the training data; Optimization method selection, use the maximum likelihood estimation (MLE) to optimize the kernel function parameters;

[0101] Model training: Train the Gaussian process regression model by maximizing the log marginal likelihood function, and the formula is: , where Y is the label vector; X is the feature matrix; is the covariance matrix, I is the identity matrix, is the noise term; n is the number of training samples.

[0102] S5 specifically includes:

[0103] S51: Prepare an independent validation dataset, screen out the independent validation dataset from the historical operation data that has not participated in model training. The dataset includes feature vectors with the same structure as the training data and the corresponding seal performance level labels; The validation dataset is used to evaluate the generalization ability of the model and ensure that the validation data distribution is consistent with the actual operation scenario;

[0104] S52: Input each feature vector in the validation dataset into the Gaussian process regression model generated in S4 one by one. The feature vectors include preprocessed features such as the temperature change rate, the pressure fluctuation amplitude, and the optical fiber signal attenuation rate, etc.; Output the prediction results, that is, the sealing performance level corresponding to each sample.

[0105] S53: Record and save the predicted sealing performance level of each validation sample by the model and the actual sealing performance level label in the validation dataset to form a comparison data table of the prediction results and the actual labels.

[0106] S54: Traverse the prediction results and the actual labels, and compare the predicted level and the actual level of each sample one by one; If the predicted level is the same as the actual level, it is recorded as a correct prediction; If the predicted level is different from the actual level, it is recorded as an incorrect prediction; Record the number of samples with correct predictions and incorrect predictions.

[0107] S55: According to the comparison results in S54, count the proportion of the number of samples with correct predictions in the total number of samples, and calculate the accuracy of the model; The accuracy is used to measure the ability of the model to correctly identify the sealing performance level.

[0108] S56: Calculate the model error by statistically analyzing the degree of difference between the prediction results and the actual level labels, and quantifying it as an error index, which is used to evaluate the effectiveness of the model's ability to distinguish between different levels; Through the above sub-steps, use the independent validation dataset to comprehensively verify the Gaussian process regression model generated in S4 to ensure the accuracy and reliability of the model in the actual operation scenario, and provide a basis for subsequent model optimization.

[0109] The specific calculation of the accuracy of the model in S55 includes:

[0110] S551: Traverse the comparison data table of the prediction results and the actual labels recorded in S54, and respectively count the number of samples with the same prediction results and the actual labels and the number of samples with different results. Let be the number of samples with correct predictions, and record the number of samples whose prediction results are equal to the actual labels; Let be the number of samples with incorrect predictions, and record the number of samples whose prediction results are not equal to the actual labels; Let be the total number of samples, and the calculation formula is: ;

[0111] S552: According to the sample numbers counted in S551, calculate the accuracy of the model. The formula is: , where A is the accuracy rate of the model, expressed as a percentage; this accuracy rate represents the proportion of samples correctly predicted by the model and is used to measure the ability of the model to distinguish the sealing performance levels on the validation dataset; if the accuracy rate is high, it indicates that the model can better identify different sealing performance levels; if the accuracy rate is low, it means that the model needs further optimization.

[0112] Calculating the model error in S56 specifically includes:

[0113] S561: For each sample, calculate the difference between the predicted sealing performance level and the actual sealing performance level. The formula for the difference value is: , where is the predicted sealing performance level of the i-th sample; is the actual sealing performance level of the i-th sample; is the difference value of the i-th sample, representing the deviation between the predicted level and the actual level;

[0114] S562: Using the difference values of all samples, calculate the mean squared error MSE of the model. The formula is: , where is the total number of samples in the validation dataset; is the difference value of the i-th sample; MSE is the mean squared error used to quantify the overall deviation between the predicted result and the actual label;

[0115] S563: To analyze the error in depth, calculate the mean absolute error MAE. The formula is: , where MAE is the mean absolute error, representing the average deviation magnitude of the model prediction results;

[0116] S564: Compare the absolute error with the average value of the actual grade label range and calculate the error rate. The formula is: , where E is the error rate, expressed as a percentage; R is the value range of the sealing performance level (such as from 0 to 2, R = 2); the mean squared error reflects the stability of the predicted result, and the smaller the value, the smaller the fluctuation of the model prediction result; the mean absolute error reflects the average deviation between the predicted result and the actual value, and the smaller the value, the smaller the model error; the E error rate reflects the percentage of the overall deviation of the model in the grade range, and the smaller the value, the higher the model prediction accuracy.

[0117] S6 specifically includes:

[0118] S61: Based on the accuracy rate and error calculated in S5, judge the prediction performance of the model. If the accuracy rate is lower than 85%, it is considered that the model needs to be adjusted; if the mean squared error is higher than 0.1 or the mean absolute error is higher than 0.1, it is considered that the model needs to be adjusted;

[0119] S62: Adjust the kernel function parameters (such as the length scale ) in the Gaussian process regression model using an optimization algorithm (such as grid search or Bayesian optimization). The optimization process selects the optimal kernel function parameters by maximizing the likelihood function; calculate the likelihood function : , where y is the observed value, X is the input data, K is the covariance matrix, are the kernel function parameters; by maximizing this likelihood function, obtain the optimized kernel function parameters to enhance the model's fitting ability;

[0120] S63: Adjust the noise term of the model according to the error analysis to balance the signal and noise in the training data; optimize the noise term by minimizing the sum of squared residuals or maximizing the likelihood function; the updated noise term is: , where is the predicted value of the i-th sample, is the actual value of the i-th sample, is the total number of samples;

[0121] S64: After adjusting the kernel function and the noise term, re-evaluate the accuracy and error of the model using the validation dataset: compare the accuracy and error before and after optimization; if the accuracy of the optimized model is improved and the error is reduced, determine the final optimization result; through the above steps, based on the accuracy and error results, adjust the kernel function parameters and the noise term of the Gaussian process regression model to optimize the prediction accuracy of the model.

[0122] S7 specifically includes:

[0123] S71: Input the preprocessed dataset in S2 into the Gaussian process regression model optimized in step S6. This dataset includes the characteristic parameters of the armored fiber-optic composite cable, such as the temperature change rate, the pressure fluctuation amplitude, and the fiber signal attenuation rate, etc.;

[0124] S72: The model predicts the degradation trend of the sealing performance of the armored fiber-optic composite cable based on the input characteristic parameters and outputs the prediction results, including the degradation time point, the degradation rate, and the degradation trend;

[0125] S73: Generate a degradation report based on the prediction results of the model. The degradation report includes the current sealing performance status, predicted degradation time point, degradation rate analysis, degradation trend chart, and maintenance suggestions. The current sealing performance status describes the current situation of the sealing performance based on the current characteristic parameters. The predicted degradation time point is used to specify in detail the time point when the sealing performance may degrade to a specific threshold. The degradation rate analysis is used to provide a quantitative analysis of the degradation rate of the sealing performance. The degradation trend chart shows the predicted degradation trend of the sealing performance in the form of a chart for intuitive understanding. The maintenance suggestions specifically propose corresponding maintenance or replacement suggestions based on the prediction results to extend the service life of the equipment or prevent failures from occurring.

[0126] The present invention covers any alternatives, modifications, equivalent methods, and solutions made within the spirit and scope of the present invention. To enable the public to have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, and those skilled in the art can fully understand the present invention without these detailed descriptions. Additionally, to avoid unnecessary confusion to the essence of the present invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0127] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for predicting underwater sealing performance degradation of armored optoelectronic composite cables based on data analysis, characterized in that: The following steps are involved: S1: Collect real-time operation data of armored photoelectric composite cables during operation through sensors, including temperature, pressure, optical fiber signal attenuation value and current; S2: preprocess the running data collected by S1 and normalize all the data to generate a structured preprocessed data set; S3: Based on the preprocessed data set of S2, characteristic parameters related to the degradation of sealing performance are extracted, including temperature change rate, pressure fluctuation amplitude and optical fiber signal attenuation rate, and a characteristic vector set is formed; The S3 specifically includes: S31: For the temperature data at each time point in the preprocessed data set, the temperature change rate is calculated according to the following formula: Among them, T i and T i+1 are the temperature values ​​at adjacent time points, Δt is the time interval, and ΔT is the temperature change rate. At the same time, the temperature change rates calculated at all time points are summarized as time series data and recorded as a column in the feature vector. S32: Calculate the pressure fluctuation amplitude. For the pressure data in the preprocessed data set, take each hour as a fixed time window and calculate the maximum value P of the pressure in the time window. max With the minimum value P min , and calculate the pressure fluctuation amplitude according to the following formula: A P =P max -P min , where A P is the pressure fluctuation amplitude in the time window; at the same time, the pressure fluctuation amplitudes in all time windows are summarized as time series data and recorded as a column in the feature vector; S33: Calculate the optical fiber signal attenuation rate. For the optical fiber signal attenuation value in the preprocessed data set, take 10 minutes as a fixed time window and calculate the optical fiber signal attenuation rate according to the following formula: in, and are the optical fiber signal strength values ​​at the start time t1 and end time t2 of the window, R atten is the fiber signal attenuation rate; at the same time, the fiber signal attenuation rates in all time windows are summarized as time series data and recorded as a column in the feature vector; S34: merging the calculated time series data of temperature change rate, pressure fluctuation amplitude and optical fiber signal attenuation rate to form a feature vector set; the format of the feature vector set is in matrix form, wherein each row represents a feature of a time window, and the columns correspond to the temperature change rate, the pressure fluctuation amplitude and the optical fiber signal attenuation rate respectively; S4: using the feature vector set extracted in S3, training a Gaussian process regression model, wherein the Gaussian process regression model is used to predict the trend of sealing performance degradation; S5: Use an independent validation data set to validate the Gaussian process regression model generated in S4, input the feature vector of the independent validation set into the model, output the prediction results, and compare them with the actual sealing performance annotation values ​​to calculate the accuracy and error of the initial prediction model; S6: Based on the accuracy and error results output by S5, adjust the kernel function parameters and noise terms in the Gaussian process regression model to optimize the prediction accuracy of the Gaussian process regression model; S7: Input the data preprocessed by S2 into the Gaussian process regression model optimized by S6 to generate the prediction results of the degradation of the sealing performance of the armored optoelectronic composite cable, and generate a degradation report according to the prediction results.

2. The method for predicting underwater sealing performance degradation of armored optoelectronic composite cables based on data analysis according to claim 1 is characterized in that: The S1 specifically includes: S11: A digital temperature sensor is installed at the connection between the outer wall of the sealed cavity of the armored optoelectronic composite cable and the joint. The digital temperature sensor uses a platinum resistance element and outputs a linear voltage signal. The temperature data is collected after being processed by an analog-to-digital conversion circuit. The sampling frequency is once per second. S12: A piezoresistive pressure sensor is installed inside the sealed cavity of the armored optoelectronic composite cable. The piezoresistive pressure sensor specifically uses a silicon piezoresistive chip with a measuring range of 0-10MPa and an accuracy of ±0.1%FS. The pressure signal is converted into a standard voltage signal by a signal conditioning circuit and then collected. The sampling frequency is once per second. S13: Multimode optical fiber is laid along the armored optoelectronic composite cable, and the attenuation value of the optical fiber signal is collected using a distributed optical fiber sensing system and the Raman scattering principle. The collection points include 10 meters, 20 meters, and 50 meters away from the two ends of the optical fiber, and the sampling frequency is once per minute; S14: Install a Hall effect current sensor on the power transmission conductor of the armored optoelectronic composite cable, wherein the Hall effect current sensor is located in the conductor segment close to the connection terminal and is used to monitor the real-time current flowing through the entire power transmission line; the sampling frequency is once per second.

3. The method for predicting underwater sealing performance degradation of armored optoelectronic composite cables based on data analysis according to claim 1 is characterized in that: The S2 specifically includes: S21: De-noising the real-time operation data collected by S1; S22: Statistical methods are used to detect and remove outliers from the denoised data; S23: For the missing data after removing outliers, interpolation method is used to complete them; S24: normalize the completed data to convert data of different dimensions into the same scale range; S25: Arrange the normalized data in time series to form a structured preprocessed data set in matrix form, where each row represents the observation data at a time point and each column represents a sensor data type; and add a corresponding timestamp to each row of data to ensure the time series of the data.

4. The method for predicting underwater sealing performance degradation of armored optoelectronic composite cables based on data analysis according to claim 1 is characterized in that: The S4 specifically includes: S41: construct a training data set, select feature vectors including temperature change rate, pressure fluctuation amplitude and optical fiber signal attenuation rate from the feature vector set generated in S3, and attach corresponding sealing performance level labels to each feature vector, and the specific level labels include normal, slightly degraded and severely degraded; S42: According to the feature change pattern of the training data set, select the corresponding kernel function to build a Gaussian process regression model; if the feature changes smoothly, select the radial basis function kernel that can capture the smooth characteristics; if the feature changes irregularly, select the Matern kernel that can adapt to the non-smooth characteristics; S43: setting initial kernel function parameters and noise term parameters, wherein the kernel function parameters are used to control the range of feature correlation, and the noise term parameters are used to describe measurement noise in the data; S44: training the Gaussian process regression model by maximizing the marginal likelihood function of the training data, during which the correlation between the feature vectors is calculated according to the feature matrix, and the kernel function parameters and the noise term parameters are adjusted in combination with the sealing performance level label; S45: After the training is completed, the Gaussian process regression model is saved.

5. The method for predicting underwater sealing performance degradation of armored optoelectronic composite cables based on data analysis according to claim 1 is characterized in that: The S5 specifically includes: S51: preparing an independent verification data set, screening out the independent verification data set from historical operation data that has not participated in model training, wherein the data set includes feature vectors with the same structure as the training data and corresponding sealing performance level labels; S52: Input all feature vectors in the validation data set into the Gaussian process regression model generated by S4 one by one; output the prediction result, i.e., the sealing performance level corresponding to each sample; S53: Record and save the predicted sealing performance level of each verification sample by the model and the actual sealing performance level label in the verification data set to form a comparison data table of the predicted results and the actual labels; S54: traverse the prediction results and the actual labels, and compare the predicted level and the actual level of each sample one by one; if the predicted level is consistent with the actual level, it is recorded as a correct prediction; if the predicted level is inconsistent with the actual level, it is recorded as an incorrect prediction; record the number of samples with correct prediction and incorrect prediction; S55: According to the comparison result in S54, the ratio of the number of samples with correct prediction to the total number of samples is counted to calculate the accuracy of the model; the accuracy is used to measure the ability of the model to correctly identify the sealing performance level; S56: Calculate the model error by counting the difference between the predicted result and the actual grade label and quantify it as an error indicator.

6. The method for predicting underwater sealing performance degradation of armored optoelectronic composite cables based on data analysis according to claim 5 is characterized in that: The S55 specifically includes: S551: Traverse the comparison data table of the predicted results and actual labels recorded in S54, and count the number of samples whose predicted results are consistent with the actual labels and the number of samples whose labels are inconsistent. Let N c To predict the correct number of samples, record the predicted result equal to the number of samples with actual labels; let N e is the number of samples with incorrect predictions, and records the number of samples whose prediction results are not equal to the actual labels; let N t is the total sample size, and the calculation formula is: N t =N c +N e ; S552: Calculate the accuracy of the model based on the number of samples counted in S551. The formula is: Among them, A is the accuracy of the model, expressed as a percentage.

7. The method for predicting underwater sealing performance degradation of armored optoelectronic composite cables based on data analysis according to claim 6 is characterized in that: The S56 specifically includes: S561: For each sample, calculate the difference between the predicted sealing performance level and the actual sealing performance level. The calculation formula for the difference value is: D i =P i -A i , where P i is the predicted sealing performance level of the i-th sample; A i is the actual sealing performance level of the i-th sample; D i is the difference value of the i-th sample; S562: Using the difference values ​​of all samples, calculate the mean square error (MSE) of the model. The formula is: Among them, N t is the total number of samples in the validation dataset; D i is the difference value of the i-th sample; MSE is the mean square error, which is used to quantify the overall deviation between the predicted result and the actual label; S563: To further analyze the error, calculate the mean absolute error MAE, the formula is: Among them, MAE is the mean absolute error, which indicates the average deviation of the model prediction results; S564: Compare the absolute error with the average value of the actual grade label range and calculate the error rate. The formula is: Wherein, E is the error rate, expressed as a percentage; R is the value range of the sealing performance level.

8. The method for predicting underwater sealing performance degradation of armored optoelectronic composite cables based on data analysis according to claim 5 is characterized in that: The S6 specifically includes: S61: According to the accuracy and error calculated in S5, the prediction performance of the model is judged. If the accuracy is lower than 85%, it is considered that the model needs to be adjusted; if the mean square error is higher than 0.1 or the mean absolute error is higher than 0.1, it is considered that the model needs to be adjusted; S62: Use optimization algorithms to adjust kernel function parameters in Gaussian process regression models; S63: adjusting the noise term of the model according to the error analysis to balance the signal and noise in the training data; optimizing the noise term is performed by minimizing the residual sum of squares or maximizing the likelihood function; S64: After adjusting the kernel function and noise term, use the validation data set to re-evaluate the accuracy and error of the model: compare the accuracy and error before and after optimization; if the accuracy of the model is improved and the error is reduced after optimization, determine the final optimization result.

9. The method for predicting underwater sealing performance degradation of armored optoelectronic composite cables based on data analysis according to claim 1 is characterized in that: The S7 specifically includes: S71: input the data set preprocessed in S2 into the Gaussian process regression model optimized in step S6; S72: The model predicts the degradation trend of the sealing performance of the armored optoelectronic composite cable according to the input characteristic parameters, and outputs the prediction results, including the degradation time point, degradation rate and degradation trend; S73: Generate a degradation report based on the prediction results of the model, which includes the current sealing performance status, predicted degradation time point, degradation rate analysis, degradation trend chart and maintenance suggestions.

Images