A method for monitoring data acquisition and analysis

Abstract

This invention relates to the field of data acquisition and analysis technology, and discloses a monitoring data acquisition and analysis method. The method collects monitoring data from multiple data sources on a ship, including a longitude set A, a latitude set B, a heading set C, a speed set D, and a distance set E between the ship and its destination. The collected monitoring data is then preprocessed. This method achieves comprehensive monitoring of the ship's navigation status by collecting monitoring data from multiple data sources, including longitude, latitude, heading, speed, and distance between the ship and its destination. It comprehensively considers multiple dimensions of ship navigation, making the assessment of the ship's navigation status more comprehensive and accurate. Through accurate prediction of the ship's navigation trajectory, it can provide strong support for ship navigation safety, reduce potential risks during navigation, and, more importantly, help optimize the ship's navigation route and speed, improve navigation efficiency, and reduce operating costs.

Description

Technical Field

[0001] This invention relates to the field of data acquisition and analysis technology, specifically a method for monitoring data acquisition and analysis. Background Technology

[0002] Monitoring data refers to real-time or historical information about a specific environment, system, or object collected through various monitoring devices and technologies. This data is typically used in multiple fields such as security monitoring, system performance monitoring, and environmental monitoring, aiming to provide a comprehensive understanding of the target's status and help relevant personnel to promptly identify problems, make decisions, and take action.

[0003] With the continued growth of global trade and the increasing volume of maritime traffic, the prediction of ship trajectories has become particularly important. Accurate trajectories not only help improve the safety and efficiency of maritime traffic, but also provide strong support for ship scheduling, route planning, fuel consumption optimization, and emergency response. However, the prediction of ship trajectories faces many challenges, such as the processing of massive amounts of data from multiple monitoring data sources, complex and ever-changing navigation environments, and the nonlinear characteristics of ship dynamic behavior. At the same time, existing technologies lack mechanisms for monitoring data anomalies and alarms. When data anomalies occur, it is often impossible to detect the anomalies in a timely manner, and the troubleshooting work all requires manual work, resulting in a lag in the error correction function. Therefore, a monitoring data acquisition and analysis method is proposed to solve the above problems. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a monitoring data acquisition and analysis method that has advantages such as wide applicability, universal data source acquisition, and ease of anomaly analysis of monitoring data, thus solving the problems mentioned in the background section.

[0005] To achieve the aforementioned objectives of wide applicability, universal data source acquisition, and ease of anomaly analysis of monitoring data, this invention provides the following technical solution: A monitoring data acquisition and analysis method, comprising the following steps:

[0006] S1: Collect monitoring data from multiple data sources for the ship, including longitude set A, latitude set B, heading set C, speed set D, and distance set from the ship to its destination set E;

[0007] Configure a corresponding adapter for each data acquisition type. The data acquisition types include: longitude set A, latitude set B, heading set C, speed set D, and distance set from the ship to the destination E.

[0008] S2: Preprocess the collected monitoring data to obtain the longitude set A3, latitude set B3, heading set C3, speed set D3 and the distance set of the ship from the destination E3;

[0009] S3: Assign corresponding weights to the longitude set A3, latitude set B3, heading set C3, speed set D3, and distance set E3 from the ship to the destination, and calculate them as a comprehensive reference index S. The resulting sequence data S' of the comprehensive reference index S is then calculated, where S' = [S1, S2, ..., S...]. T ], T is the sequence length, S T S is the comprehensive reference index at time T;

[0010] S4: Construct a TBENet network learning model by inputting the sequence data of sequence data S' into the TBENet network learning model to improve the prediction ability of the TBENet network learning model;

[0011] S5: Test the predictive ability of the TBENet network learning model, optimize the parameters of the TBENet network learning model using the AdamW optimizer, and train the TBENet network learning model.

[0012] S6: The TBENet network learning model is hyperparameterized using the Bayesian optimization method to obtain the optimal parameters of the TBENet network learning model. The optimal parameters are then substituted into the TBENet network learning model to obtain the optimized TBENet network learning model, which improves the accuracy of ship trajectory prediction.

[0013] Preferably, the specific steps of step S2 are as follows:

[0014] S2.1: Use Lagrange interpolation to supplement the monitoring data sets A (longitude), B (latitude), C (heading), D (speed), and E (distance from the ship to the destination), resulting in sets A1 (longitude), B1 (latitude), C1 (heading), D1 (speed), and E1 (distance from the ship to the destination). Specifically:

[0015] S2.1.1: For n+1 known data points (X0, Y0), (X1, Y1), ..., (X... n Y n Lagrange interpolation polynomial L Y The expression for (X) is:

[0016]

[0017] Where X is the independent variable, Y i For X i The value of L, Y∈{A, B, C, D, E} i (X) is the basic polynomial;

[0018] S2.1.2: Fundamental Polynomial L i (X) Specifically:

[0019]

[0020] Among them, X i With X j Let L be the independent variable of the known data points, satisfying L i (X) i ) = 1 and satisfy L i (X) i ) = 0;

[0021] S2.1.3: The calculated Lagrange interpolation polynomial expression L A (X), L B (X), L C (X), L D (X) and L E (X) The longitude set A1, latitude set B1, heading set C1, speed set D1, and distance set E1 from the ship to the destination are calculated, and the expression is:

[0022]

[0023]

[0024] Where Y1∈{A1, B1, C1, D1, E1}, Y new For interpolation, Y new ∈{A new B new C new D new E new};

[0025] S2.2: Perform Min-Max normalization on the longitude set A1, latitude set B1, heading set C1, speed set D1, and distance set E1 from the ship to the destination, scaling the data to [0,1], to obtain the longitude set A2, latitude set B2, heading set C2, speed set D2, and distance set E2 from the ship to the destination, expressed as:

[0026]

[0027] Where Y2∈{A2、B2、C2、D2、E2};

[0028] S2.3: Calculate the standard deviations of the longitude set A2, latitude set B2, heading set C2, speed set D2, and distance set E2 of the ship from the destination. Then, standardize these sets to obtain the standardized longitude set A3, latitude set B3, heading set C3, speed set D3, and distance set E3 of the ship from the destination. The expression is:

[0029]

[0030]

[0031] Among them, Y` σ Y represents the standard deviation. i For the i-th data in each set, Y3∈{A3、B3、C3、D3、E3}, The average values ​​are the values ​​of A3, B3, C3, D3, and E3.

[0032] Preferably, step S3 specifically comprises:

[0033]

[0034] in, W is the average value in the longitude set A3. A The relative weights of the average values ​​in the longitude set A3. W is the average value in the latitude set B3. B The relative weights of the average values ​​in the latitude set B3. W is the average value in the heading set C3. C The relative weights of the average values ​​in the heading set C3. W is the average value of the speed set D3. D The relative weight of the average value in the speed set D3. W is the average value of the set E3 of distances from the ship to its destination. E The weight relative to the average value in the set E3 of distances between ships and their destinations.

[0035] Preferably, the specific steps of step S4 are as follows:

[0036] S4.1: After sequence data S' enters the first Bi-GRU layer, it is processed to obtain sequence data H1. The ReLU activation function is used to process sequence data H1 into sequence data H'1, and sequence data H'1 is then transmitted to the first fully connected layer and the first Dropout layer. The expression is:

[0037]

[0038]

[0039] S4.2: After receiving sequence data H'1, the first fully connected layer processes sequence data H'1 into sequence data H2, and the first Dropout layer processes sequence data H2 into sequence data H'2. Then, sequence data H'2 is transmitted to the second Bi-GRU layer. The expression is as follows:

[0040]

[0041]

[0042] S4.3: After receiving sequence data H'2, the second Bi-GRU layer processes it to obtain sequence data H3, uses the ReLU activation function to process sequence data H3 into sequence data H'3, and then transmits sequence data H'3 to the second fully connected layer and the second Dropout layer. The expression is:

[0043]

[0044]

[0045] S4.4: After receiving sequence data H'3, the second fully connected layer processes sequence data H'3 into sequence data H4, and the second Dropout layer processes sequence data H4 into H'4. The expression for transmitting sequence data H'4 to the third Bi-GRU layer is as follows:

[0046]

[0047]

[0048] S4.5: After receiving sequence data H'4, the third Bi-GRU layer processes it to obtain sequence data H5, uses the ReLU activation function to process sequence data H5 into sequence data H'5, and then transmits sequence data H'5 to the third fully connected layer and the third Dropout layer. The expression is:

[0049]

[0050]

[0051] S4.6: After receiving sequence data H'5, the second fully connected layer processes sequence data H'5 into sequence data H6, and the third Dropout layer processes sequence data H6 into H'6. Sequence data H'6 is then transmitted to the Bi-LSTM layer. The expression is:

[0052]

[0053]

[0054] S4.7: After receiving sequence data H'6, the Bi-LSTM layer processes sequence data H'6 into sequence data H7, and transmits sequence data H7 to the fourth fully connected layer. The expression is as follows:

[0055]

[0056] S4.8: After receiving the sequence data H7, the fourth fully connected layer processes the sequence data H7 using the Sigmoid function to obtain the sequence data M, with the following expression:

[0057]

[0058] Preferably, step S4 specifically comprises:

[0059] S4.1: The processed sequence data M is evaluated for quality using binary cross-entropy, expressed as:

[0060]

[0061] Where N is the total sample size, Q t The true label indicates whether the ship's trajectory has deviated at time t. t Let L be the predicted value and L be the loss amount;

[0062] S4.2: The AdamW optimizer is used to optimize the parameters of the TBENet network learning model. After multiple iterations of optimization, the optimal TBENet network learning model is obtained. The AdamW optimizer specifically optimizes as follows:

[0063]

[0064] Where, θ z Let η be the parameter for step z. z Let P be the learning rate. z For the first-order moment estimate of the gradient, U z For the second moment estimate of the gradient, ε is a constant, and λ is the weight decay function to prevent division by zero;

[0065] S4.3: The specific steps for training the TBENet network learning model are as follows:

[0066] S4.3.1: Initialize the TBENet network learning model parameters;

[0067] S4.3.2: For each epoch, the following steps are included:

[0068] Forward propagation: Calculate the predicted value M t And the amount of loss L;

[0069] Backpropagation: Calculates gradients and updates the TBENet network to learn model parameters;

[0070] Evaluate the performance of the TBENet network learning model and record the optimal TBENet network learning model parameters.

[0071] Preferably, step S6 specifically includes:

[0072] S6.1: Select initial parameter space;

[0073] S6.2: Train a surrogate model using parameters from the initial parameter space to approximate an unknown objective function;

[0074] S6.3: Based on the current proxy model, use the acquisition function to determine the next sampling point;

[0075] S6.4: Evaluate the true objective function at the selected sampling points, and add the newly obtained sampling points to the existing sampling point set to obtain an updated sampling point set. Retrain the surrogate model using the updated sampling point set.

[0076] S6.5: Repeat steps S6.3-S6.4 until the stopping condition is met to obtain the trained surrogate model. Use the parameters in the trained surrogate model as the optimal parameters for the TBENet network learning model.

[0077] Compared with existing technologies, the present invention provides a monitoring data acquisition and analysis method, which has the following beneficial effects:

[0078] This monitoring data acquisition and analysis method collects monitoring data from multiple sources on the ship, including longitude, latitude, heading, speed, and distance from the destination, to achieve comprehensive monitoring of the ship's navigation status. By comprehensively considering multiple dimensions of ship navigation, the assessment of the ship's navigation status is more comprehensive and accurate. Through precise prediction of the ship's navigation trajectory, it can provide strong support for the ship's navigation safety, reduce potential risks during navigation, and at the same time, the accurate prediction results can also help optimize the ship's navigation route and speed, improve navigation efficiency, and reduce operating costs. Attached Figure Description

[0079] Figure 1 This is a schematic diagram of a monitoring data acquisition and analysis method proposed in this invention;

[0080] Figure 2This is a schematic diagram of the internal process of the TBENet network learning model, a monitoring data acquisition and analysis method proposed in this invention. Detailed Implementation

[0081] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0082] Please see Figure 1-2 A method for monitoring data collection and analysis includes the following steps:

[0083] S1: Collect monitoring data from multiple data sources for the ship, including longitude set A, latitude set B, heading set C, speed set D, and distance set from the ship to its destination set E;

[0084] S2: Preprocess the collected monitoring data to obtain the longitude set A3, latitude set B3, heading set C3, speed set D3 and the distance set of the ship from the destination E3;

[0085] By collecting monitoring data from multiple sources on the ship, including longitude, latitude, heading, speed, and distance from the destination, comprehensive monitoring of the ship's navigation status was achieved. The data preprocessing steps further ensured the accuracy and reliability of the data, providing a solid foundation for subsequent analysis and prediction.

[0086] S3: Assign corresponding weights to the longitude set A3, latitude set B3, heading set C3, speed set D3, and distance set E3 from the ship to the destination, and calculate them as a comprehensive reference index S. The resulting sequence data S' of the comprehensive reference index S is then calculated, where S' = [S1, S2, ..., S...]. T ], T is the sequence length, S T S is the comprehensive reference index at time T;

[0087] By setting corresponding weight thresholds for key parameters such as longitude, latitude, heading, speed, and distance to the destination, and calculating them as a comprehensive reference index S, this step comprehensively considers multiple dimensions of ship navigation, making the assessment of ship navigation status more comprehensive and accurate.

[0088] S4: Construct a TBENet network learning model by inputting the sequence data of sequence data S' into the TBENet network learning model to improve the prediction ability of the TBENet network learning model;

[0089] The TBENet model combines the advantages of Bidirectional Long Short-Term Memory (BiLSTM) and Bidirectional Gated Recurrent Unit (BiGRU) with a unique three-layer architecture, effectively capturing both long-term dependencies and short-term variations in time-series data. This combination not only ensures comprehensive data processing but also maintains information integrity over extended periods, thereby improving prediction accuracy and stability. Furthermore, the trajectory curves predicted by the TBENet method closely match reality, demonstrating superior prediction accuracy compared to other evaluation models.

[0090] S5: Test the predictive ability of the TBENet network learning model, optimize the parameters of the TBENet network learning model using the AdamW optimizer, and train the TBENet network learning model.

[0091] The AdamW optimizer was used to optimize the parameters of the TBENet network learning model, which not only accelerated the training process but also effectively prevented overfitting and further improved the model's generalization ability.

[0092] S6: The TBENet network learning model is hyperparameterized using the Bayesian optimization method to obtain the optimal parameters of the TBENet network learning model. The optimal parameters are then substituted into the TBENet network learning model to obtain the optimized TBENet network learning model, which improves the accuracy of ship trajectory prediction.

[0093] By fine-tuning the hyperparameters of the TBENet network learning model using Bayesian optimization, the optimal parameter configuration of the model was obtained. This step enabled the model to achieve higher prediction accuracy while maintaining high efficiency, providing strong support for the accurate prediction of ship trajectories.

[0094] The specific steps of step S2 are as follows:

[0095] S2.1: Use Lagrange interpolation to supplement the monitoring data sets A (longitude), B (latitude), C (heading), D (speed), and E (distance from the ship to the destination), resulting in sets A1 (longitude), B1 (latitude), C1 (heading), D1 (speed), and E1 (distance from the ship to the destination). Specifically:

[0096] Lagrange interpolation approximates known data points by constructing a polynomial function, thus calculating more accurate data values. During long-term data acquisition, errors may gradually accumulate, causing the data to deviate from the true values. Lagrange interpolation can reduce this error accumulation to some extent, making the data closer to reality. Lagrange interpolation ensures the accuracy and continuity of navigation data, thereby enabling better route planning and avoidance of potential risks.

[0097] S2.1.1: For n+1 known data points (X0, Y0), (X1, Y1), ..., (X... n Y n Lagrange interpolation polynomial L Y The expression for (X) is:

[0098]

[0099] Where X is the independent variable, Y i For X i The value of L, Y∈{A, B, C, D, E} i (X) is the basic polynomial;

[0100] S2.1.2: Fundamental Polynomial L i (X) Specifically:

[0101]

[0102] Among them, X i With X j Let L be the independent variable of the known data points, satisfying L i (X) i ) = 1 and satisfy L i (X) i ) = 0;

[0103] S2.1.3: The calculated Lagrange interpolation polynomial expression L A (X), L B (X), L C (X), L D (X) and L E (X) The longitude set A1, latitude set B1, heading set C1, speed set D1, and distance set E1 from the ship to the destination are calculated, and the expression is:

[0104]

[0105]

[0106] Where Y1∈{A1, B1, C1, D1, E1}, Y new For interpolation, Y new ∈{A new B new C new D new E new};

[0107] The Lagrange interpolation polynomial expression for longitude A is:

[0108]

[0109] Among them, A i For longitude in X i The value at that location, Li （A） (X) is the fundamental polynomial of longitude A;

[0110] The Lagrange interpolation polynomial expression for latitude B is:

[0111]

[0112] Among them, B i Latitude in X i The value at that location, Li （B） (X) is the fundamental polynomial of latitude B;

[0113] The Lagrange interpolation polynomial expression for the heading C is:

[0114]

[0115] Among them, C i For heading in X i The value at that location, Li （C） (X) is the fundamental polynomial of the heading C;

[0116] The Lagrange interpolation polynomial expression for the speed D is:

[0117]

[0118] Among them, D i For the speed at X i The value at that location, Li （D） (X) is the fundamental polynomial of the speed D;

[0119] The Lagrange interpolation polynomial expression for the distance E between the ship and its destination is:

[0120]

[0121] Among them, E i The distance of the ship from its destination in X i The value at that location, Li （E） (X) is the fundamental polynomial of the distance E between the ship and its destination;

[0122] S2.2: Perform Min-Max normalization on the longitude set A1, latitude set B1, heading set C1, speed set D1, and distance set E1 from the ship to the destination, scaling the data to [0,1], to obtain the longitude set A2, latitude set B2, heading set C2, speed set D2, and distance set E2 from the ship to the destination, expressed as:

[0123]

[0124] Where Y2∈{A2、B2、C2、D2、E2};

[0125] S2.3: Calculate the standard deviations of the longitude set A2, latitude set B2, heading set C2, speed set D2, and distance set E2 of the ship from the destination. Then, standardize these sets to obtain the standardized longitude set A3, latitude set B3, heading set C3, speed set D3, and distance set E3 of the ship from the destination. The expression is:

[0126]

[0127]

[0128] Among them, Y` σ Y represents the standard deviation. i For the i-th data in each set, Y3∈{A3、B3、C3、D3、E3}, The average values ​​are the values ​​of A3, B3, C3, D3, and E3.

[0129] Step S3 is as follows:

[0130]

[0131] in, W is the average value in the longitude set A3. A The relative weights of the average values ​​in the longitude set A3. W is the average value in the latitude set B3. B The relative weights of the average values ​​in the latitude set B3. W is the average value in the heading set C3. C The relative weights of the average values ​​in the heading set C3. W is the average value of the speed set D3. D The relative weight of the average value in the speed set D3. W is the average value of the set E3 of distances from the ship to its destination. E The weight relative to the average value in the set E3 of distances between ships and their destinations.

[0132] The TBENet model's structural design leverages the complementary advantages of different network units to enhance data fitting capabilities. Each network unit is followed by a fully connected layer to enhance nonlinear processing and adaptability. Furthermore, each component in the network can have a unique structure tailored to different functions, including enhancing long-term and short-term features. This design allows the TBENet model to better adapt to complex ship navigation trajectory data, improving prediction reliability. The specific steps in step S4 are as follows:

[0133] S4.1: After sequence data S' enters the first Bi-GRU layer, it is processed to obtain sequence data H1. The ReLU activation function is used to process sequence data H1 into sequence data H'1, and sequence data H'1 is then transmitted to the first fully connected layer and the first Dropout layer. The expression is:

[0134]

[0135]

[0136] S4.2: After receiving sequence data H'1, the first fully connected layer processes sequence data H'1 into sequence data H2, and the first Dropout layer processes sequence data H2 into sequence data H'2. Then, sequence data H'2 is transmitted to the second Bi-GRU layer. The expression is as follows:

[0137]

[0138]

[0139] S4.3: After receiving sequence data H'2, the second Bi-GRU layer processes it to obtain sequence data H3, uses the ReLU activation function to process sequence data H3 into sequence data H'3, and then transmits sequence data H'3 to the second fully connected layer and the second Dropout layer. The expression is:

[0140]

[0141]

[0142] S4.4: After receiving sequence data H'3, the second fully connected layer processes sequence data H'3 into sequence data H4, and the second Dropout layer processes sequence data H4 into H'4. The expression for transmitting sequence data H'4 to the third Bi-GRU layer is as follows:

[0143]

[0144]

[0145] S4.5: After receiving sequence data H'4, the third Bi-GRU layer processes it to obtain sequence data H5, uses the ReLU activation function to process sequence data H5 into sequence data H'5, and then transmits sequence data H'5 to the third fully connected layer and the third Dropout layer. The expression is:

[0146]

[0147]

[0148] S4.6: After receiving sequence data H'5, the second fully connected layer processes sequence data H'5 into sequence data H6, and the third Dropout layer processes sequence data H6 into H'6. Sequence data H'6 is then transmitted to the Bi-LSTM layer. The expression is:

[0149]

[0150]

[0151] S4.7: After receiving sequence data H'6, the Bi-LSTM layer processes sequence data H'6 into sequence data H7, and transmits sequence data H7 to the fourth fully connected layer. The expression is as follows:

[0152]

[0153] S4.8: After receiving the sequence data H7, the fourth fully connected layer processes the sequence data H7 using the Sigmoid function to obtain the sequence data M, with the following expression:

[0154]

[0155] The Sigmoid function maps the output value to the (0,1) interval, representing the probability of the offset.

[0156] Step S4 is as follows:

[0157] S4.1: The processed sequence data M is evaluated for quality using binary cross-entropy, expressed as:

[0158]

[0159] Where N is the total sample size, Q t For true labels, the output is generally 1 or 0, indicating whether the ship's trajectory has deviated at time t. tLet L be the predicted value, representing the probability of the ship's trajectory deviating at time t, and L be the amount of loss.

[0160] S4.2: The AdamW optimizer is used to optimize the parameters of the TBENet network learning model. After multiple iterations of optimization, the optimal TBENet network learning model is obtained. The AdamW optimizer specifically optimizes as follows:

[0161]

[0162] Where, θ z Let η be the parameter for step z. z Let P be the learning rate. z For the first-order moment estimate of the gradient, U z For the second moment estimate of the gradient, ε is a constant, and λ is the weight decay function to prevent division by zero;

[0163] S4.3: The specific steps for training the TBENet network learning model are as follows:

[0164] S4.3.1: Initialize the TBENet network learning model parameters;

[0165] S4.3.2: For each epoch, the following steps are included:

[0166] Forward propagation: Calculate the predicted value M t And the amount of loss L;

[0167] Backpropagation: Calculates gradients and updates the TBENet network to learn model parameters;

[0168] Evaluate the performance of the TBENet network learning model and record the optimal TBENet network learning model parameters.

[0169] Step S6 specifically includes:

[0170] S6.1: Select initial parameter space;

[0171] S6.2: Train a surrogate model using parameters from the initial parameter space to approximate an unknown objective function;

[0172] S6.3: Based on the current proxy model, use the acquisition function to determine the next sampling point;

[0173] S6.4: Evaluate the true objective function at the selected sampling points, and add the newly obtained sampling points to the existing sampling point set to obtain an updated sampling point set. Retrain the surrogate model using the updated sampling point set.

[0174] S6.5: Repeat steps S6.3-S6.4 until the stopping condition is met to obtain the trained surrogate model. Use the parameters in the trained surrogate model as the optimal parameters for the TBENet network learning model.

[0175] In summary, this monitoring data acquisition and analysis method, by collecting monitoring data from multiple sources on the ship, including longitude, latitude, heading, speed, and distance from the destination, achieves comprehensive monitoring of the ship's navigation status. It comprehensively considers multiple dimensions of ship navigation, making the assessment of the ship's navigation status more comprehensive and accurate. Through precise prediction of the ship's navigation trajectory, it can provide strong support for navigation safety, reduce potential risks during navigation, and, moreover, help optimize the ship's route and speed, improve navigation efficiency, and reduce operating costs.

[0176] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0177] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring data acquisition and analysis, characterized in that, Includes the following steps: S1: Collect monitoring data from multiple data sources for the ship, including longitude set A, latitude set B, heading set C, speed set D, and distance set from the ship to its destination set E; S2: Preprocess the collected monitoring data to obtain the longitude set A3, latitude set B3, heading set C3, speed set D3 and the distance set of the ship from the destination E3; S3: Assign corresponding weights to the longitude set A3, latitude set B3, heading set C3, speed set D3, and distance set E3 from the ship to the destination, and calculate them as a comprehensive reference index S. The resulting sequence data S' of the comprehensive reference index S is then calculated, where S' = [S1, S2, ..., S...]. T ], T is the sequence length, S T S is the comprehensive reference index at time T; S4: Construct a TBENet network learning model by inputting the sequence data of sequence data S' into the TBENet network learning model to improve the prediction ability of the TBENet network learning model; S5: Test the predictive ability of the TBENet network learning model, optimize the parameters of the TBENet network learning model using the AdamW optimizer, and train the TBENet network learning model. S6: The TBENet network learning model is hyperparameter adjusted using the Bayesian optimization method to obtain the optimal parameters of the TBENet network learning model. The optimal parameters are then substituted into the TBENet network learning model to obtain the optimized TBENet network learning model, which improves the accuracy of ship trajectory prediction. The specific steps of step S4 are as follows: S4.1: After sequence data S' enters the first Bi-GRU layer, it is processed to obtain sequence data H1. The ReLU activation function is used to process sequence data H1 into sequence data H'1, and sequence data H'1 is then transmitted to the first fully connected layer and the first Dropout layer. The expression is: ； ； S4.2: After receiving sequence data H'1, the first fully connected layer processes sequence data H'1 into sequence data H2, and the first Dropout layer processes sequence data H2 into sequence data H'2. Then, sequence data H'2 is transmitted to the second Bi-GRU layer. The expression is as follows: ； ； S4.3: After receiving sequence data H'2, the second Bi-GRU layer processes it to obtain sequence data H3, uses the ReLU activation function to process sequence data H3 into sequence data H'3, and then transmits sequence data H'3 to the second fully connected layer and the second Dropout layer. The expression is: ； ； S4.4: After receiving sequence data H'3, the second fully connected layer processes sequence data H'3 into sequence data H4, and the second Dropout layer processes sequence data H4 into H'4. The expression for transmitting sequence data H'4 to the third Bi-GRU layer is as follows: ； ； S4.5: After receiving sequence data H'4, the third Bi-GRU layer processes it to obtain sequence data H5, uses the ReLU activation function to process sequence data H5 into sequence data H'5, and then transmits sequence data H'5 to the third fully connected layer and the third Dropout layer. The expression is: ； ； S4.6: After receiving sequence data H'5, the second fully connected layer processes sequence data H'5 into sequence data H6, and the third Dropout layer processes sequence data H6 into H'6. Sequence data H'6 is then transmitted to the Bi-LSTM layer. The expression is: ； ； S4.7: After receiving sequence data H'6, the Bi-LSTM layer processes sequence data H'6 into sequence data H7, and transmits sequence data H7 to the fourth fully connected layer. The expression is as follows: ； S4.8: After receiving the sequence data H7, the fourth fully connected layer processes the sequence data H7 using the Sigmoid function to obtain the sequence data M, with the following expression: 。 2. The monitoring data acquisition and analysis method according to claim 1, characterized in that, The specific steps of step S2 are as follows: S2.1: Use Lagrange interpolation to supplement the monitoring data sets A (longitude), B (latitude), C (heading), D (speed), and E (distance from the ship to the destination), resulting in sets A1 (longitude), B1 (latitude), C1 (heading), D1 (speed), and E1 (distance from the ship to the destination). Specifically: S2.1.1: For n+1 known data points (X0, Y0), (X1, Y1), ..., (X... n Y n Lagrange interpolation polynomial L Y The expression for (X) is: ； Where X is the independent variable, Y i For X i The value of L, Y∈{A, B, C, D, E} i (X) is the basic polynomial; S2.1.2: Fundamental Polynomial L i (X) Specifically: ； Among them, X i With X j Let L be the independent variable of the known data points, satisfying L i (X) i ) = 1 and satisfy L i (X) i ) = 0; S2.1.3: The calculated Lagrange interpolation polynomial expression L A (X), L B (X), L C (X), L D (X) and L E (X) The longitude set A1, latitude set B1, heading set C1, speed set D1, and distance set E1 from the ship to the destination are calculated, and the expression is: ； ； Where Y1∈{A1, B1, C1, D1, E1}, Y new For interpolation, Y new ∈{A new B new C new D new E new }; S2.2: Perform Min-Max normalization on the longitude set A1, latitude set B1, heading set C1, speed set D1, and distance set E1 from the ship to the destination, scaling the data to [0,1], to obtain the longitude set A2, latitude set B2, heading set C2, speed set D2, and distance set E2 from the ship to the destination, expressed as: ； Where Y2∈{A2、B2、C2、D2、E2}; S2.3: Calculate the standard deviations of the longitude set A2, latitude set B2, heading set C2, speed set D2, and distance set E2 of the ship from the destination. Then, standardize these sets to obtain the standardized longitude set A3, latitude set B3, heading set C3, speed set D3, and distance set E3 of the ship from the destination. The expression is: ； ； Among them, Y` σ Y represents the standard deviation. i For the i-th data in each set, Y3∈{A3、B3、C3、D3、E3}, The average values ​​are the values ​​of A3, B3, C3, D3, and E3.

3. The monitoring data acquisition and analysis method according to claim 1, characterized in that, Step S3 specifically involves: ； in, W is the average value in the longitude set A3. A The relative weights of the average values ​​in the longitude set A3. W is the average value in the latitude set B3. B The relative weights of the average values ​​in the latitude set B3. W is the average value in the heading set C3. C The relative weights of the average values ​​in the heading set C3. W is the average value of the speed set D3. D The relative weight of the average value in the speed set D3. W is the average value of the set E3 of distances from the ship to its destination. E The weight relative to the average value in the set E3 of distances between ships and their destinations.

4. The monitoring data acquisition and analysis method according to claim 1, characterized in that, Step S4 specifically involves: S4.1: The processed sequence data M is evaluated for quality using binary cross-entropy, expressed as: ； Where N is the total sample size, Q t The true label indicates whether the ship's trajectory has deviated at time t. t Let L be the predicted value and L be the loss amount; S4.2: The AdamW optimizer is used to optimize the parameters of the TBENet network learning model. After multiple iterations of optimization, the optimal TBENet network learning model is obtained. The AdamW optimizer specifically optimizes as follows: ； Where, θ z Let η be the parameter for step z. z Let P be the learning rate. z For the first-order moment estimate of the gradient, U z This is the second-order moment estimate of the gradient, where ε is a constant and λ is the weight decay function; S4.3: The specific steps for training the TBENet network learning model are as follows: S4.3.1: Initialize the TBENet network learning model parameters; S4.3.2: For each epoch, the following steps are included: Forward propagation: Calculate the predicted value M t And the amount of loss L; Backpropagation: Calculates gradients and updates the TBENet network to learn model parameters; Evaluate the performance of the TBENet network learning model and record the optimal TBENet network learning model parameters.

5. The monitoring data acquisition and analysis method according to claim 1, characterized in that, Specifically, step S6 is as follows: S6.1: Select an initial parameter space; S6.2: Train a surrogate model using the parameters in the initial parameter space to approximate the unknown objective function; S6.3: Based on the current surrogate model, use the acquired function to determine the next sampling point; S6.4: Evaluate the true objective function at the selected sampling point, and add the newly obtained sampling point to the existing sampling point set to obtain an updated sampling point set, and retrain the surrogate model using the updated sampling point set; S6.5: Repeat steps S6.3-S6.4 until the stopping condition is met to obtain the trained surrogate model, and use the parameters in the trained surrogate model as the optimal parameters for the TBENet network learning model.

Images