Real-time evaluation method for reliability of marine autonomous surface ship mechanical equipment

Abstract

The invention discloses a real-time evaluation method for reliability of marine autonomous surface ship mechanical equipment, which comprises the following steps of: S1, constructing a synthetic data set of operation of the ship mechanical equipment according with an actual degradation rule, and solving the problem of imbalance between small samples and data; s2, evaluating a health index value of the ship mechanical equipment by adopting a comprehensive weighting method and a plurality of types of processed operation characteristic data; s3, constructing an unsupervised health index prediction model for predicting a health index value, and training the unsupervised health index prediction model to obtain a trained unsupervised health index prediction model; s4, performing Weibull distribution fitting on the health index value predicted by the trained unsupervised health index prediction model to obtain a shape parameter and a scale parameter of the ship mechanical equipment under Weibull distribution; and S5, real-time evaluation of the reliability of the ship mechanical equipment is realized based on the shape parameters and the scale parameters, early fault detection is supported, and the safety of the MASS is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of reliability assessment of ship machinery and equipment, and in particular to a real-time reliability assessment method for machinery and equipment of autonomous surface ships at sea. Background Art

[0002] With the rapid development of technologies related to Maritime Autonomous Surface Ships (MASS), in order to obtain regulatory approval, industry support and public recognition, MASS is required to achieve the same reliable operation time as traditional ships when completing the same tasks. In maritime transportation, the reliable operation of ships is usually achieved through reactive maintenance (RM) and preventive maintenance (PM). The former repairs the fault after it occurs, while the latter formulates regular maintenance plans based on preset time intervals, component life or working conditions. It also includes condition-based maintenance (CBM), which formulates maintenance plans by evaluating the degradation state of components. In recent years, predictive maintenance (PdM) has optimized maintenance strategies and laid the foundation for in-depth implementation and application.

[0003] Mechanical equipment, as the fundamental infrastructure that supports ship operations during navigation, ensuring its reliable operation can effectively reduce unplanned downtime and safety incidents caused by equipment failures. However, current implementations of MASS reliability focus primarily on hull structural integrity, collision avoidance algorithms, and remote control systems. Mechanical equipment reliability, a key element of navigation safety, has received insufficient attention. Filling this gap is crucial for MASS to effectively execute long-term autonomous missions. Mechanical equipment's heavy reliance on human intervention complicates the maintenance and repair of faulty components on MASS. Because MASS ultimately needs to operate without maintenance personnel, maintenance management (RM) cannot prevent failures, while maintenance maintenance (PM) relies on empirically defined fixed maintenance cycles that must be coordinated with scheduled port calls. While cost-effective CBM (Computer-Based Maintenance) (CBM) offers advantages, cumulative system degradation can reach a failure threshold, increasing the risk of failure. To ensure reliable MASS operation during navigation, PdM (Physical Device Design) (PDM) has become an effective method for ensuring mechanical equipment reliability. However, due to the slow accumulation of MASS mechanical equipment failure data and the tendency of purely algorithmic data to deviate from actual operating conditions, resulting in low model credibility, and the lack of fault labels in actual monitoring data, traditional methods have difficulty directly quantifying mechanical equipment reliability. Summary of the Invention

[0004] The present invention provides a real-time reliability evaluation method for mechanical equipment of autonomous surface ships at sea to overcome the above technical problems.

[0005] In order to achieve the above object, the technical solution of the present invention is:

[0006] A real-time reliability assessment method for mechanical equipment of an autonomous surface ship at sea, comprising the following steps:

[0007] S1: Constructing a synthetic data set of ship machinery and equipment operation that conforms to actual degradation laws, the synthetic data set includes several types of processed operation characteristic data;

[0008] S2: using a comprehensive weighting method and the several processed operating characteristic data to evaluate the health index value of the ship machinery and equipment to establish a health index data set;

[0009] S3: constructing an unsupervised health index prediction model for predicting a health index value, and training the unsupervised health index prediction model based on the health index dataset to obtain a trained unsupervised health index prediction model;

[0010] S4: performing Weibull distribution fitting on the health index value predicted by the trained unsupervised health index prediction model, and solving to obtain shape parameters and scale parameters of the ship machinery equipment under the Weibull distribution;

[0011] S5: Real-time evaluation of the reliability of ship machinery and equipment is achieved based on the shape parameters and scale parameters.

[0012] Furthermore, in S1, the specific steps of constructing a synthetic data set of ship machinery and equipment operation that conforms to the actual degradation law include:

[0013] S11: Acquiring ship machinery and equipment operating data, wherein the ship machinery and equipment operating data includes several types of operating characteristic data and degradation state coefficients;

[0014] S12: Filtering the ship machinery and equipment operation data to obtain a plurality of sub-data sets according to the required ship speed;

[0015] S13: removing one or more operating characteristic data that do not change over time from the multiple sub-data sets based on the change law of the Weibull distribution and the degradation state coefficient, to obtain an initial data set;

[0016] S14: Use the minimum-maximum normalization method to scale the initial data set to the interval [-1,1];

[0017] S15: Use Wasserstein generative adversarial network to perform data augmentation on the scaled dataset;

[0018] S16: Denormalize the data augmented dataset to convert it into its true form;

[0019] S17: setting a time period to be analyzed, and filtering out data within the time period from several sub-datasets in the denormalized data set to construct a synthetic data set including several types of processed operating characteristic data.

[0020] Furthermore, in S13, the specific step of eliminating one or more operating characteristic data that do not change over time from the multiple sub-data sets in combination with the change law of the Weibull distribution and the degradation state coefficient includes:

[0021] S131: The initial shape parameters and scale parameters of the Weibull distribution are set through expert experience to ensure that the degradation curve conforms to the physical characteristics of mechanical wear. The degradation curve used to characterize the change of the degradation state coefficient over time is obtained by combining the change law of the Weibull distribution, which is expressed as:

[0022]

[0023] Where λ is the scale parameter, β is the shape parameter, t is the degradation time;

[0024] S132: adding time series to the plurality of operating characteristic data in the sub-dataset according to the degradation curve to obtain a sub-dataset with added time series;

[0025] S133: Eliminating one or more types of operation characteristic data in the sub-dataset to which the time series is added according to a set elimination rule;

[0026] The elimination rule is: determine whether each type of operation characteristic data has no change over time; if so, eliminate all data of this type of operation characteristic parameter; otherwise, retain this type of operation characteristic data.

[0027] Furthermore, in S2, the formula for evaluating the health index value of the ship machinery and equipment using the comprehensive weighted method and the several processed operating characteristic data is:

[0028]

[0029] Among them, h i (t) is the health index value of ship machinery equipment i based on multiple operating characteristic data; m is the total number of operating characteristic data; ρ j is the weight of the j-th running feature data;

[0030] h ij (t) is the final health index value of the j-th operating characteristic data of device i, and the steps of obtaining it include:

[0031] The several processed operating characteristic data are respectively normalized to obtain initial health index values ​​of the several operating characteristic data. The processing formula is:

[0032]

[0033] Where, is the initial health index value of the jth running characteristic data, ranging from [0,1]; X j is the data sample of the jth running feature data, X max and X min are the maximum and minimum values ​​of the j-th operating characteristic data respectively;

[0034] The initial health index values ​​of each operation characteristic data corresponding to the initial operation time of the ship are reset to the maximum health index value L, L = 1;

[0035] Calculate the difference between the initial health index value of each operation characteristic data corresponding to the initial time of ship operation and the maximum health index value, and add the absolute value of the difference to the initial health index value of each operation characteristic data corresponding to other ship operation times except the initial time of ship operation to obtain the changed health index value of each operation characteristic data;

[0036] Determine whether the health index value of each running characteristic data after the change is greater than or equal to 1. If it is greater than or equal to 1, transform the health index value greater than or equal to 1 according to the transformation formula, and use the transformed health index value as the final health index value of the running characteristic data. The transformation formula is:

[0037] h ij (t) = 2- h' ij (t) (4)

[0038] Where h' ij (t) is a health index value greater than or equal to 1.

[0039] Furthermore, in S3, the unsupervised health index prediction model for predicting the health index value is a model constructed based on the principal component analysis-long short-term memory network model PCA-LSTM architecture;

[0040] The specific steps of training the unsupervised health index prediction model based on the health index dataset include:

[0041] The health index dataset is normalized to eliminate the influence of the dimension on the calculation results, and the original matrix X is established based on the normalized health index dataset. * ;

[0042] Based on the original matrix X *, and determine the covariance matrix P according to formula (5), the formula is:

[0043]

[0044] Where S is the number of samples in the synthetic dataset;

[0045] Based on the covariance matrix P and according to formula (6), the eigenvalue λ is obtained i and eigenvector α i , expressed as:

[0046] Pα i =λ i α i (6)

[0047] Among them, λ i Indicates the variance in the direction of the principal component, α i Represents the direction of the principal component. The number of principal components α is calculated according to formula (7):

[0048]

[0049] Among them, δ is the threshold of cumulative contribution rate; is the cumulative contribution rate of the first k eigenvalues, k <n;

[0050] By combining the covariance matrix P and the number of principal components α, the original matrix X * Perform dimensionality reduction to obtain the original matrix after dimensionality reduction;

[0051] The original matrix after dimension reduction is input into a long short-term memory network model to train it, and a trained unsupervised health index prediction model is obtained.

[0052] Furthermore, in S4, the maximum likelihood estimation method (MLE) is used to obtain the shape parameters and scale parameters of the ship machinery and equipment.

[0053] Furthermore, in S5, the formula for real-time evaluation of the reliability of ship machinery and equipment based on the shape parameters and scale parameters is:

[0054]

[0055] Where R represents the reliability evaluation result, and t′ represents the navigation time in the synthetic dataset.

[0056] Beneficial Effects: This invention constructs a synthetic dataset of ship machinery and equipment operation that conforms to actual degradation patterns, resolving the issues of small samples and data imbalance. A comprehensive weighted approach and several processed operational characteristic data are used to evaluate the health index of ship machinery and equipment, establishing a health index dataset. An unsupervised health index prediction model is constructed to predict health index values ​​in real time. The health index values ​​predicted by the trained unsupervised health index prediction model are fitted with a Weibull distribution, and the shape and scale parameters of the ship machinery and equipment under the Weibull distribution are obtained. This allows for dynamic, real-time health index-reliability assessment, supports early fault detection, and ultimately improves the safety of MASS. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0058] Figure 1 This is a first flow chart of a real-time reliability assessment method for mechanical equipment of an autonomous surface ship at sea according to the present invention;

[0059] Figure 2 This is a second flow chart of a real-time reliability assessment method for mechanical equipment of an autonomous surface ship at sea according to the present invention;

[0060] Figure 3 This is a structural diagram of the Generative Adversarial Network (GAN) mentioned in the present invention;

[0061] Figure 4 is a structural diagram of a turbine in an embodiment of the present invention;

[0062] Figure 5 is a comparison diagram of degradation curves of a compressor and a turbine in an embodiment of the present invention;

[0063] Figure 6 Graph showing changes in health indexes for three main ship speeds according to an embodiment of the present invention;

[0064] Figure 7 This is a diagram showing the calculation results of the weighted health index value in an embodiment of the present invention;

[0065] Figure 8 This is a PCA-LSTM health index value prediction result diagram in an embodiment of the present invention;

[0066] Figure 9 Calculation results of size parameters and shape parameters in an embodiment of the present invention;

[0067] Figure 10 This is a diagram of the reliability evaluation results in an embodiment of the present invention. DETAILED DESCRIPTION

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0069] This embodiment provides a real-time evaluation method for the reliability of mechanical equipment of autonomous surface ships at sea. Figure 1 and Figure 2 As shown, the specific steps include:

[0070] S1: Constructing a synthetic data set of ship machinery and equipment operation that conforms to actual degradation laws, the synthetic data set includes several types of processed operation characteristic data;

[0071] S2: using a comprehensive weighting method and the several processed operating characteristic data to evaluate the health index value of the ship machinery and equipment to establish a health index data set;

[0072] S3: constructing an unsupervised health index prediction model for predicting a health index value, and training the unsupervised health index prediction model based on the health index dataset to obtain a trained unsupervised health index prediction model;

[0073] S4: performing Weibull distribution fitting on the health index value predicted by the trained unsupervised health index prediction model, and solving to obtain shape parameters and scale parameters of the ship machinery equipment under the Weibull distribution;

[0074] S5: Real-time evaluation of the reliability of ship machinery and equipment is achieved based on the shape parameters and scale parameters.

[0075] In a specific embodiment, in S1, the specific steps of constructing a synthetic data set of ship machinery and equipment operation that conforms to actual degradation laws include:

[0076] S11: Acquire ship machinery and equipment operation data, wherein the ship machinery and equipment operation data includes several types of operation characteristic data and degradation state coefficients, as shown in Table 1;

[0077] S12: Since the obtained ship machinery and equipment operation data does not have a time series, the ship machinery and equipment operation data needs to be further processed. Specifically, in this embodiment, multiple sub-datasets are obtained from the ship machinery and equipment operation data according to the required ship speed;

[0078] S13: removing one or more operating characteristic data that do not change over time from the multiple sub-data sets based on a change pattern of Weibull distribution and a degradation state coefficient to obtain an initial data set;

[0079] Specifically, the Weibull distribution is a continuous probability distribution that is widely used in reliability analysis, survival analysis, and failure time modeling due to its flexibility. This distribution is often used to describe the failure time characteristics of mechanical equipment. This embodiment combines the Weibull distribution with the actual degradation law of the equipment, combines the degradation state coefficient, and uses shape parameters and scale parameters to characterize the changes in several operating characteristic data over time to facilitate subsequent analysis.

[0080] In a specific embodiment, in S13, the specific step of eliminating one or more operating characteristic data that do not change over time from the multiple sub-data sets by combining the change law of the Weibull distribution and the degradation state coefficient includes:

[0081] S131: The initial shape parameters and scale parameters of the Weibull distribution are set through expert experience to ensure that the degradation curve is more consistent with the physical characteristics of mechanical wear. The degradation curve used to characterize the change of the degradation state coefficient over time is obtained by combining the change law of the Weibull distribution, which is expressed as:

[0082]

[0083] Where λ is the scale parameter, β is the shape parameter, t is the degradation time;

[0084] S132: adding time series to the plurality of operating characteristic data in the sub-dataset according to the degradation curve to obtain a sub-dataset with added time series;

[0085] Specifically, according to the change of the degradation state coefficient, this embodiment selects data within a two-year range, and uses the degradation curve to fit the change trend of several types of operating characteristic data over time, so that the added time series is more consistent with the initial ship machinery and equipment operation data obtained.

[0086] S133: Eliminating one or more types of operation characteristic data in the sub-dataset to which the time series is added according to a set elimination rule;

[0087] The elimination rule is: determine whether each type of operating characteristic data has no change over time. If so, eliminate all data of this type of operating characteristic parameter; otherwise, retain this type of operating characteristic data.

[0088] S14: Use the minimum-maximum normalization method to scale the initial data set to the [-1,1] interval. The formula is:

[0089]

[0090] In the formula, target_data is the initial data set, data_min is the minimum value in the initial data set, data_max is the maximum value in the initial data set, and normalized_data is the normalized data.

[0091] S15: Use Wasserstein generative adversarial network (WGAN) to perform data augmentation on the scaled dataset;

[0092] Specifically, Generative Adversarial Network (GAN), as a widely used data augmentation technology, can effectively alleviate the problem of data imbalance by generating synthetic samples. This adversarial architecture consists of two neural network components that perform minimax optimization: the generator (G) generates artificial samples by capturing the latent feature distribution, and the discriminator (D) distinguishes between real data and generated data through a learnable decision boundary, and seeks the Nash equilibrium between the two through adversarial training and iterative optimization. The optimization of GAN fundamentally depends on the strategic construction of the adversarial objective function. According to the original GAN ​​framework setting, the adversarial training process constitutes a two-player game theory scenario, and its minimax optimization objective can be mathematically expressed as:

[0093]

[0094] Where, P data (x) represents the real data distribution, P z (z) is the prior distribution of the latent variable. The discriminator dynamically adjusts the decision boundary through the parameter function D(x). When the output value exceeds 0.5, it indicates a higher confidence level in the authenticity of the sample. G(z) is the sample data generated by the generator network using random noise.

[0095] During GAN model training, problems such as gradient vanishing and model collapse may occur. Furthermore, coordinating the training parameters of the generator and discriminator networks often takes a significant amount of time. Therefore, this embodiment employs WGAN, which can alleviate the vanishing gradient problem caused by the inconsistency between the distribution of real samples and generated data in low-dimensional manifolds. WGAN is based on improvements to the GAN architecture, including the following:

[0096] 1) Removing the Sigmoid activation function at the end of the discriminator in the Generative Adversarial Network (GAN) to achieve unbounded gradient propagation, enhance the network's sensitivity to distribution differences, and enforce the Lipschitz continuity constraint through backpropagation weight clipping to prevent parameter divergence and improve the stability of generated data;

[0097] 2) Using Wasserstein distance to reconstruct the adversarial target loss function to replace the logarithmic loss function in the generative adversarial network (GAN), it can more effectively capture the relationship between distributions;

[0098] 3) During training, the stochastic gradient descent (SGD) and root mean square propagation (RMSProp) optimizer are used to replace the momentum-based Adam algorithm in the generative adversarial network (GAN). This reduces the random fluctuations in parameter updates and makes the confrontation between the discriminator and the generator more stable.

[0099] Specifically, the definition of Wasserstein distance in WGAN is expressed as:

[0100]

[0101] Where p r is the real data distribution, p g Represents the generated data distribution, the joint probability distribution set Ώ(p r ,p g ) with p r and p g is a marginal distribution, γ is one of the joint probability distributions, W(p r ,p g ) represents the lower bound of the expectation for γ, that is, the real data distribution p is required r The sample x and the generated data distribution p g The distance of sample y ||xy|| needs to be adapted to p r and p g .

[0102] In a specific embodiment, the adversarial target loss function reconstructed using Wasserstein distance is:

[0103]

[0104] Where, f w Refers to the final output layer of the discriminator and maintains linear characteristics, w is the bias top of the corresponding neuron; Represents the expected value of the discriminator's output for the real sample; Represents the expected value of the discriminator's output for the generated sample.

[0105] S16: Denormalize the data augmented dataset to convert it into its true form;

[0106] S17: setting a time period to be analyzed, and filtering out data within the time period from several sub-datasets in the denormalized data set to construct a synthetic data set including several types of processed operating characteristic data.

[0107] Specifically, in the process of constructing the synthetic dataset, it is necessary to combine physical knowledge, match it according to the actual ship speed during navigation, and reasonably divide the time proportion occupied by different datasets.

[0108] Specifically, MASS mechanical equipment failure data accumulates slowly, and purely algorithmic data generation (such as traditional GANs) easily deviates from actual operating conditions, resulting in low model credibility. This example uses WGAN for data augmentation and incorporates physical knowledge (such as Weibull distribution parameter priors) to generate a MASS mechanical equipment operation dataset for a single voyage, ensuring that it conforms to actual degradation patterns. This provides a dataset synthesis method for research on mechanical equipment reliability, addressing the issues of small sample sizes and data imbalance.

[0109] In a specific embodiment, in S2, the formula for evaluating the health index value of the ship machinery and equipment using the comprehensive weighted method and the several types of processed operating characteristic data is:

[0110]

[0111] Among them, h i (t) is the health index value of ship machinery equipment i based on multiple operating characteristic data; m is the total number of operating characteristic data; ρ j is the weight of the jth operating characteristic data, reflecting the impact of the operating characteristic data on the equipment status, 0≤ρ j ≤1, and

[0112] h ij (t) is the final health index value of the j-th operating characteristic data of device i, and the steps of obtaining it include:

[0113] The several processed operating characteristic data are respectively normalized to obtain initial health index values ​​of the several operating characteristic data. The processing formula is:

[0114]

[0115] Where, is the initial health index value of the jth running characteristic data, ranging from [0,1]; X j is the data sample of the jth running feature data, Xmax and X min are the maximum and minimum values ​​of the j-th operating characteristic data respectively;

[0116] The initial health index values ​​of each operation characteristic data corresponding to the initial operation time of the ship are reset to the maximum health index value L, L = 1;

[0117] Calculate the difference between the initial health index value of each operation characteristic data corresponding to the initial time of ship operation and the maximum health index value, and add the absolute value of the difference to the initial health index value of each operation characteristic data corresponding to other ship operation times except the initial time of ship operation to obtain the changed health index value of each operation characteristic data;

[0118] Determine whether the health index value of each running characteristic data after the change is greater than or equal to 1. If it is greater than or equal to 1, transform the health index value greater than or equal to 1 according to the transformation formula, and use the transformed health index value as the final health index value of the running characteristic data. The transformation formula is:

[0119] h ij (t) = 2- h' ij (t) (8)

[0120] Where h' ij (t) is a health index value greater than or equal to 1.

[0121] In a specific embodiment, in S3, the unsupervised health index prediction model for predicting the health index value is a model constructed based on the principal component analysis-long short-term memory network model PCA-LSTM architecture;

[0122] The specific steps of training the unsupervised health index prediction model based on the health index dataset include:

[0123] The health index dataset is normalized to eliminate the influence of the dimension on the calculation results, and the original matrix X is established based on the normalized health index dataset. * ;

[0124] Based on the original matrix X * , and determine the covariance matrix P according to formula (9), the formula is:

[0125]

[0126] Where S is the number of samples in the synthetic dataset;

[0127] Based on the covariance matrix P and according to formula (10), the eigenvalue λ is obtained i and eigenvector α i, expressed as:

[0128] Pα i =λ i α i (10)

[0129] Among them, λ i Indicates the variance in the direction of the principal component, α i Indicates the direction of the principal component. The number of principal components α is calculated according to formula (11) to retain the main variance of the data. The formula is:

[0130]

[0131] Among them, δ is the threshold of the cumulative contribution rate, which is usually set to 85%; is the cumulative contribution rate of the first k eigenvalues ​​(k <n);

[0132] By combining the covariance matrix P and the number of principal components α, the original matrix X * Perform dimensionality reduction to obtain the original matrix after dimensionality reduction;

[0133] The original matrix after dimension reduction is input into a long short-term memory network model to train it, and a trained unsupervised health index prediction model is obtained.

[0134] Specifically, PCA is a linear dimensionality reduction technique that can map high-dimensional data to a low-dimensional space through linear projection, that is, projecting the original features onto the dimension that captures the maximum variance in the data, ensuring that the low-dimensional representation retains the original information as much as possible. In this embodiment, PCA is used to reduce the original matrix X * Performing dimensionality reduction and using the original matrix to train the Long Short-Term Memory (LSTM) network model can improve LSTM training accuracy. Specifically, LSTM demonstrates significant advantages when dealing with time series prediction and nonlinear mapping problems. LSTM incorporates memory cells to retain past information and three gate structures—input, output, and forget gates—to control the transmission of historical information. This solves the vanishing and exploding gradient issues associated with traditional recurrent neural networks, caused by chained connections between network cells, effectively improving learning time.

[0135] Specifically, in this embodiment, the following evaluation indicators are used to compare the prediction performance of the PCA-LSTM model, including:

[0136] (1) Mean Squared Error (MSE) is the ratio of the sum of squared deviations between the observed value and the true value to the number of observations. It can sensitively reflect the size of the model prediction error, but is greatly affected by outliers. Since MSE squares the error, larger errors receive greater weight. It is often used as the loss function of the linear regression model to optimize the model parameters by minimizing MSE. The formula for MSE is:

[0137]

[0138] Among them, y i is the true value, is the predicted value, and n is the number of observations.

[0139] (2) Mean Absolute Error (MAE) is similar to MSE, but uses absolute error. It is more robust to outliers than MSE because it does not amplify these outliers by squaring the error. MAE can more intuitively reflect the prediction error in some practical applications. It is suitable for situations where a direct and interpretable error measurement is required. The formula for MAE is:

[0140]

[0141] (3)R 2 Measures the proportion of the true label variance that the model fails to capture. Its value ranges from 0 to 1, and the closer it is to 1, the better the model fit. It is widely used in statistical modeling, especially in evaluating the interpretability of the model. 2 The formula is:

[0142]

[0143] in, is the mean of the true values.

[0144] In a specific embodiment, in S4, the health index value predicted by the trained unsupervised health index prediction model is fitted with a Weibull distribution, and the maximum likelihood estimation method MLE is used to solve the shape parameters and scale parameters of the ship machinery equipment under the Weibull distribution. The formula includes:

[0145]

[0146] Among them, L represents the likelihood function, x j represents the jth health index value, λ is the scale parameter, β is the shape parameter, and N is the total number of health indices used for calculation;

[0147] The maximum likelihood condition is calculated as follows:

[0148]

[0149] To simplify formula (15), the maximum likelihood condition is expressed as:

[0150]

[0151] Specifically, this embodiment evaluates the reliability of mechanical equipment through Weibull distribution fitting, adopts maximum likelihood estimation method to solve the shape parameters and scale parameters under Weibull distribution in real time, and realizes real-time reliability evaluation of ship mechanical equipment based on the dynamic changes of shape parameters and scale parameters.

[0152] In a specific embodiment, in S5, the formula for real-time evaluation of the reliability of ship machinery and equipment based on the shape parameters and scale parameters is:

[0153]

[0154] Where R represents the reliability evaluation result, and t′ represents the navigation time in the synthetic dataset.

[0155] Specifically, actual monitoring data lacks fault labels, and traditional methods are difficult to directly quantify the reliability of mechanical equipment. This embodiment enhances real data by combining physical knowledge and the WGAN algorithm to construct a health index dataset, and reduces the dimension of the health index dataset through PCA. By extracting key features through dimensionality reduction and combining LSTM to accurately capture long-term dependencies, the health index value is predicted in real time, the health index prediction accuracy is improved, and the predicted health index value is fitted with a Weibull distribution. The shape parameters and scale parameters are obtained by combining MLE real-time solution, and a mathematical relationship is directly established between the health status and the reliability of ship mechanical equipment to more accurately characterize the equipment degradation process and reliability trend. In the absence of fault label data, the real-time calculation results of shape parameters and scale parameters can be combined to achieve real-time evaluation of the reliability of ship mechanical equipment. With the help of real-time reliability evaluation, early fault detection can be carried out in a timely manner, preventive measures can be taken to avoid accidents, improve system safety, and provide real-time decision support for equipment maintenance and management.

[0156] Specifically, this embodiment applies MLE to update the Weibull distribution parameters every hour, combined with real-time health index feedback, to achieve dynamic calibration of the reliability assessment model.

[0157] To effectively evaluate the real-time reliability of MASS mechanical equipment, this example proposes a real-time reliability assessment method for MASS mechanical equipment. Combining physical knowledge, the degradation coefficient is fitted to a Weibull distribution to reflect the actual operating conditions. Time series features are added to the status data to overcome the limitations of actual monitoring data. WGAN is used to enhance data richness. A 720-hour voyage status monitoring dataset is extracted for case verification.

[0158] In view of the future development of MASS, it is crucial to ensure high reliability, power redundancy, minimize human intervention, and enable remote control of the propulsion system. In addition, prioritizing low emissions and environmental sustainability is also key. For smart ships, battery-powered and hybrid propulsion systems are considered to be the best solutions. Therefore, in this embodiment, a data set is generated through an advanced gas turbine (GT) propulsion system simulator, which is installed on a naval frigate equipped with a combined diesel-electric and gas (CODLAG) propulsion system. The data comes from the repository of the University of California, Irvine (UCI). The degradation of the turbine and compressor is represented by the degradation state coefficient. The GT compressor degradation state coefficient and the GT turbine degradation state coefficient are used as the characteristics of this model. Other characteristics of the GT components are used as the target of state prediction. The structure of the GT is as follows: Figure 4 As shown, its specific operating characteristic data are shown in characteristics 1-16 in Table 1.

[0159] Table 1

[0160]

[0161]

[0162] Among these operating characteristic data, lp (feature 1) is linearly related to the ship speed, while T1, P1 and Pexh remain unchanged throughout the dataset. Therefore, these four operating characteristics are excluded from the analysis, and the dataset covers nine different ship speeds: 3, 6, 9, 12, 15, 18, 21, 24 and 27 knots. Since the status of other gas turbine components will change with different ship speeds, the status monitoring data of each speed are analyzed separately. This embodiment weights the results according to the duration ratio of each speed in actual operation. Specifically, after consulting with experts on the operating speed range of marine ships and analyzing the dataset, infrequently occurring speed data are excluded, and only the three most commonly used speeds are retained. The corresponding ratios are shown in Table 2.

[0163] Table 2 Ship speed time ratio

[0164] Ship speed (knots) 12 15 18 time(%) 20％ 60％ 20％

[0165] In this example, the simulation period of the data set is set to two years. CDSC and TDSC decrease from 1 to 0.95 and 0.975 respectively within two years. The other features correspond to the monitoring state at each time point and match the attenuation coefficient value. In this example, the two degradation state coefficients (features 17 and 18) are fitted using Weibull distribution to incorporate time series data. The shape parameter and scale parameter are calculated in combination with formula (1), and the following is plotted: Figure 5The degradation curves for the turbine and compressor are shown. Each ship speed dataset contains 1,326 data points. To alleviate data imbalance and improve the accuracy of training results, this example uses the WGAN algorithm for data augmentation, expanding the dataset to 175,200 data points.

[0166] Specifically, in order to obtain accurate prediction results, this embodiment uses the PCA-LSTM model to predict the status of various GT components based on the degradation of the turbine and compressor. However, due to the lack of fault labels in the dataset and in order to accurately reflect the actual degradation and real-time status of the equipment, this embodiment divides the calculation of the health index value into two parts: the first part is to calculate the real-time health index value through formula (6), and the relevant results are as follows: Figure 6 and Figure 7 The second part uses the PCA-LSTM model to predict the health index value of the device. The relevant results are shown in Figure 8 As shown. In addition, in order to improve the accuracy of the prediction, the model parameters are adjusted in time according to the prediction error. In order to be consistent with the operating mode of the ship, this embodiment merges the data of different ship speeds according to their respective time proportions during the operation period. As shown in Table 2, the synthetic data are reorganized according to the time proportion of each ship speed in the actual voyage (20% for 12 knots, 60% for 15 knots, and 20% for 18 knots) to ensure that the generated data reflects the actual operating mode. In addition, in order to simulate the actual operating conditions encountered in actual voyages, the new dataset allows the selection of time data for any given time period. In order to simplify the construction of the dataset, a continuous 720-hour time period was selected. Since different ship speeds mean different operating conditions, the differences between some features are more obvious. To address this problem, the dataset is normalized using the min-max method before merging the dataset. Figure 6 and Figure 7 The health index values ​​for three different ship speeds are shown, as well as the health index values ​​for the synthetic dataset. Figure 6 As shown in Figure 2, the health index shows a downward trend over time and drops below 0.8 after 17,520 hours (two years) without maintenance. Figure 7 As shown, the newly synthesized dataset shows an overall decline around hour 144. This can be attributed to the more stable state of the equipment operating at 12 knots compared to the data at 15 and 18 knots. After normalization, the health index for the 12 knot model is more favorable at hour 3000, which introduces unavoidable differences when using the combined dataset for prediction. If calculated using actual operating data, the health index results would likely be higher.

[0167] In order to achieve real-time and accurate prediction of the health index, the PCA-LSTM method was selected, combining the dimensionality reduction capability of PCA with the time series prediction capability of the LSTM network. The five most important features were retained by PCA, and the LSTM part used these reduced dimensionality features as input to predict the health index at each time step. The predicted health index was compared with the actual value to evaluate the performance of the model. The predicted health index is as follows: Figure 8 As shown. During the training process, MSE, MAE and R 2 To evaluate the prediction accuracy of various components. The results are shown in Table 3, where R 2 A value of 0.98 indicates that the model explains 98% of the variance in the health index, showing high accuracy of prediction.

[0168] Table 3 Average values ​​of measurement parameters

[0169] parameter MSE MAE <![CDATA[R 2 ]]> value 6.74E-05 6.50E-3 9.80E-1

[0170] In order to further evaluate the real-time reliability of the propulsion system, the predicted health index value is combined with the Weibull distribution to estimate the system reliability in real time through MLE. Based on the prediction results, the dual-parameter results are calculated using formulas (15) to (17) as follows: Figure 9 The final evaluation results of the reliability of ship machinery and equipment are shown in Figure 10 As shown, at the 720th hour, the system reliability dropped to 0.959. Based on the reliability definition, this propulsion system has a 95.9% probability of operating reliably without human intervention during the voyage (30 days). This validates the effectiveness of this method in the field of MASS mechanical equipment and provides support for addressing research gaps in existing technologies.

[0171] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A real-time evaluation method for the reliability of mechanical equipment of autonomous surface ships at sea, characterized in that: The specific steps include: S1: Constructing a synthetic data set of ship machinery and equipment operation that conforms to actual degradation laws, the synthetic data set includes several types of processed operation characteristic data; S2: using a comprehensive weighting method and the several processed operating characteristic data to evaluate the health index value of the ship machinery and equipment to establish a health index data set; S3: constructing an unsupervised health index prediction model for predicting a health index value, and training the unsupervised health index prediction model based on the health index dataset to obtain a trained unsupervised health index prediction model; S4: performing Weibull distribution fitting on the health index value predicted by the trained unsupervised health index prediction model, and solving to obtain shape parameters and scale parameters of the ship machinery equipment under the Weibull distribution; S5: Real-time evaluation of the reliability of ship machinery and equipment is achieved based on the shape parameters and scale parameters.

2. The real-time reliability assessment method for autonomous surface ship machinery and equipment according to claim 1 is characterized in that: In S1, the specific steps of constructing a synthetic dataset of ship machinery and equipment operation that conforms to the actual degradation law include: S11: Acquiring ship machinery and equipment operating data, wherein the ship machinery and equipment operating data includes several types of operating characteristic data and degradation state coefficients; S12: Filtering the ship machinery and equipment operation data to obtain a plurality of sub-data sets according to the required ship speed; S13: removing one or more operating characteristic data that do not change over time from the multiple sub-data sets based on the change law of the Weibull distribution and the degradation state coefficient, to obtain an initial data set; S14: Use the minimum-maximum normalization method to scale the initial data set to the interval [-1,1]; S15: Use Wasserstein generative adversarial network to perform data augmentation on the scaled dataset; S16: Denormalize the data augmented dataset to convert it into its true form; S17: setting a time period to be analyzed, and filtering out data within the time period from several sub-datasets in the denormalized data set to construct a synthetic data set including several types of processed operating characteristic data.

3. The real-time reliability assessment method for autonomous surface ship machinery and equipment according to claim 2 is characterized in that: In S13, the specific steps of eliminating one or more types of operating characteristic data that do not change over time from the multiple sub-data sets by combining the change law of the Weibull distribution and the degradation state coefficient include: S131: The initial shape parameters and scale parameters of the Weibull distribution are set through expert experience to ensure that the degradation curve conforms to the physical characteristics of mechanical wear. The degradation curve used to characterize the change of the degradation state coefficient over time is obtained by combining the change law of the Weibull distribution, which is expressed as: Where λ is the scale parameter, β is the shape parameter, t is the degradation time; S132: adding time series to the plurality of operating characteristic data in the sub-dataset according to the degradation curve to obtain a sub-dataset with added time series; S133: Eliminating one or more types of operation characteristic data in the sub-dataset to which the time series is added according to a set elimination rule; The elimination rule is: determine whether each type of operation characteristic data has no change over time; if so, eliminate all data of this type of operation characteristic parameter; otherwise, retain this type of operation characteristic data.

4. The real-time reliability assessment method for autonomous surface ship machinery and equipment according to claim 3 is characterized in that: In S2, the formula for evaluating the health index value of ship machinery and equipment using the comprehensive weighted method and the several types of processed operating characteristic data is: Among them, h i (t) is the health index value of ship machinery equipment i based on multiple operating characteristic data; m is the total number of operating characteristic data; ρ j is the weight of the j-th running feature data; h ij (t) is the final health index value of the j-th operating characteristic data of device i, and the steps of obtaining it include: Normalization processing is performed on the several types of processed operating characteristic data to obtain initial health index values ​​of the several types of operating characteristic data. The processing formula is: Where, is the initial health index value of the jth operating characteristic data, ranging from [0,1]; X j is the data sample of the jth running feature data, X max and X min are the maximum and minimum values ​​of the j-th operating characteristic data respectively; The initial health index values ​​of each operation characteristic data corresponding to the initial operation time of the ship are reset to the maximum health index value L, L = 1; Calculate the difference between the initial health index value of each operation characteristic data corresponding to the initial time of ship operation and the maximum health index value, and add the absolute value of the difference to the initial health index value of each operation characteristic data corresponding to other ship operation times except the initial time of ship operation to obtain the changed health index value of each operation characteristic data; Determine whether the health index value of each running characteristic data after the change is greater than or equal to 1. If it is greater than or equal to 1, transform the health index value greater than or equal to 1 according to the transformation formula, and use the transformed health index value as the final health index value of the running characteristic data. The transformation formula is: h ij (t) =2- h’ ij (t) (4) Where h' ij (t) is a health index value greater than or equal to 1.

5. The real-time reliability assessment method for autonomous surface ship machinery and equipment according to claim 4 is characterized in that: In S3, the unsupervised health index prediction model for predicting the health index value is a model constructed based on the principal component analysis-long short-term memory network model PCA-LSTM architecture; The specific steps of training the unsupervised health index prediction model based on the health index dataset include: The health index dataset is normalized to eliminate the influence of the dimension on the calculation results, and the original matrix X is established based on the normalized health index dataset. * ; Based on the original matrix X * , and determine the covariance matrix P according to formula (5), the formula is: Where S is the number of samples in the synthetic dataset; Based on the covariance matrix P and according to formula (6), the eigenvalue λ is obtained i and eigenvector α i , expressed as: Ra i =λ i a i (6) Among them, λ i Indicates the variance in the direction of the principal component, α i Represents the direction of the principal component. The number of principal components α is calculated according to formula (7): Among them, δ is the threshold of cumulative contribution rate; is the cumulative contribution rate of the first k eigenvalues, k <n; By combining the covariance matrix P and the number of principal components α, the original matrix X * Perform dimensionality reduction to obtain the original matrix after dimensionality reduction; The original matrix after dimension reduction is input into a long short-term memory network model to train it, and a trained unsupervised health index prediction model is obtained.

6. The real-time reliability assessment method for autonomous surface ship machinery and equipment according to claim 5 is characterized in that: In S4, the maximum likelihood estimation method (MLE) is used to obtain the shape parameters and scale parameters of ship machinery and equipment.

7. The real-time reliability assessment method for autonomous surface ship machinery and equipment according to claim 6 is characterized in that: In S5, the formula for real-time evaluation of the reliability of ship machinery and equipment based on the shape parameters and scale parameters is: Where R represents the reliability evaluation result, and t′ represents the navigation time in the synthetic dataset.

Images