A ship central cooler state evaluation method based on kernel fisher discriminant analysis

Abstract

The application discloses a ship central cooler state evaluation method based on kernel Fisher discriminant analysis, training sample sets are obtained by respectively randomly extracting data samples in normal monitoring data classes and abnormal monitoring data classes; and the data samples in the training sample sets are mapped to a high-dimensional feature space, so that a target function for obtaining an optimal projection vector is constructed according to obtained inter-class dispersion matrices and intra-class dispersion matrices; a generalized characteristic equation of a maximized target function is obtained based on a kernel Fisher strategy, and the current optimal projection vector is obtained according to the generalized characteristic equation, new ship central cooler nonlinear data are obtained, and ship central cooler state evaluation is realized according to the optimal projection vector, so that the problem that real-time state parameter data of the ship central cooler are various in type, the data amount is huge, and the data structure is nonlinear, which causes difficulty in effectively performing real-time state evaluation and poor evaluation precision, is solved, and more accurate and rapid state discriminant analysis is realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent ships and condition monitoring technology, and in particular to a method for assessing the condition of ship central coolers based on kernel Fisher discriminant analysis. Background Technology

[0002] With the rapid development of the global shipping industry, ships are playing an increasingly important role in the modern logistics chain. To ensure the efficient and safe operation of ships, real-time monitoring and accurate assessment of the operating status of various ship systems and equipment are crucial. Meanwhile, the intercooler, as an important component of the ship's engine, directly affects the overall performance and safety of the ship.

[0003] Currently, ship intercoolers face various harsh conditions such as high temperature, high pressure, and salt spray when operating in complex marine environments. Due to the special nature of ships, the collected operational monitoring data of ship intercoolers is complex, with many types of data, huge data volume, and the data fed back by ship intercoolers being mostly non-linear. This leads to difficulties in assessing the real-time status of ship intercoolers and poor assessment accuracy. Summary of the Invention

[0004] This invention provides a method for assessing the condition of a ship's central cooler based on nuclear Fisher discriminant analysis, in order to overcome the aforementioned technical problems.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] A method for assessing the condition of a ship's central cooler based on kernel Fisher discriminant analysis includes the following steps:

[0007] S1: Acquire historical nonlinear data of the ship's central cooler;

[0008] Furthermore, the historical nonlinear data includes normal monitoring data and abnormal monitoring data.

[0009] The monitored normal data category refers to the inlet and outlet temperatures and pressures of the high-temperature / low-temperature side seawater pipelines of the ship's central cooler under normal conditions.

[0010] The abnormal monitoring data category refers to the inlet and outlet temperatures and pressures of the high-temperature / low-temperature side seawater pipelines under abnormal conditions of the ship's central cooler.

[0011] S2: Randomly select data samples from the normal monitoring data class and the abnormal monitoring data class respectively to obtain the training sample set;

[0012] S3: Based on the nonlinear mapping method, the data samples in the training sample set are mapped to a high-dimensional feature space; and the inter-class discrete matrix and the intra-class discrete matrix of the same data class are obtained in the high-dimensional feature space.

[0013] Based on the inter-class discrete matrix and the intra-class discrete matrix, construct an objective function to obtain the optimal projection vector in the feature space;

[0014] S4: Obtain the generalized feature equation that maximizes the objective function based on the kernel Fisher strategy, and obtain the current best projection vector based on the generalized feature equation;

[0015] It also obtains data samples of normal monitoring data class and abnormal monitoring data class, the current mean of the current best projection vector, and compares the current mean with the preset mean threshold.

[0016] If the current mean satisfies the preset mean threshold, then the current best projection vector is the optimal projection vector; otherwise, the kernel parameters of the kernel Fisher policy are adaptively adjusted and steps S2 to S3 are repeated.

[0017] S5: Acquire new nonlinear data of the ship's central cooler and perform a state assessment of the ship's central cooler based on the optimal projection vector.

[0018] Furthermore, the expression for the inter-class discrete matrix of different data classes in the high-dimensional feature space obtained in S3 is as follows:

[0019]

[0020] In the formula: S b denoted as the inter-class discrete matrix; c represents the class of normal monitoring data or abnormal monitoring data; m represents the mean of all data samples in the training sample set; m c The mean of the c-th class of data samples is represented by n; n represents the number of data samples in the training sample set; n c This represents the number of data samples in class c; T represents the transpose. This represents the i-th data sample in the feature space;

[0021] The expression for the intra-class discrete matrix of the same data class is:

[0022]

[0023] In the formula: S w Represents a discrete matrix within a class; m c The mean of the c-th class of data samples is represented by n; n represents the number of data samples in the training sample set; n c This indicates the number of data samples in class c; This represents the i-th data sample of class c in the feature space;

[0024] The expression for the objective function of the projection vector is:

[0025]

[0026] In the formula: J(w) represents the objective function of the projection vector; w represents the projection vector.

[0027] Furthermore, the step S4, which describes obtaining the generalized feature equation that maximizes the objective function based on the kernel Fisher strategy and obtaining the current optimal projection vector based on the generalized feature equation, specifically includes the following steps:

[0028] S41: Obtain the generalized characteristic equation that maximizes the objective function based on the objective function of the projection vector. Its expression is as follows:

[0029] S b w=λS w w

[0030] In the formula: λ represents the eigenvalue;

[0031] S42: Rewrite the generalized characteristic equation based on the kernel Fisher strategy;

[0032] The rewriting process includes the following steps:

[0033] S421: Use the Gaussian kernel function as the kernel function of the kernel Fisher strategy, and define the kernel matrix according to the kernel function, the expression of which is:

[0034]

[0035] K ij =K(x) i ,x j )

[0036] In the formula: K represents the kernel matrix of the kernel Fisher policy, and is K ij The abbreviation of ; σ represents the kernel parameter; K(x,y) represents the Gaussian kernel function; x,y represent different data points, i.e., x i ,x j ;

[0037] S422: From the inter-class discrete matrix S b With the intraclass discrete matrix S w able to obtain

[0038] S w =φHWHφ T

[0039] S b =φHBHφT

[0040] In the formula: H represents the centered matrix and I represents an n*n identity matrix; e represents an n-dimensional vector of all ones; B and W both represent diagonal matrices; φ represents... The matrix; φ T Indicates the transpose of φ;

[0041] S423: Based on steps S421 and S422, convert the inter-class discrete matrix S b With the intraclass discrete matrix S w Rewritten as

[0042] S w =KHWHK

[0043] S b =KHBHK

[0044] S424: The generalized characteristic equation is rewritten from step S423 as follows:

[0045] KHBHKv=λKHWHKv

[0046] In the formula: K represents the kernel matrix of the kernel Fisher policy, and is K ij The abbreviation of ; v represents the feature vector;

[0047] S425: By solving the generalized characteristic equation rewritten in step S424, the current optimal projection vector w is obtained. * The expression is

[0048] w * =φv.

[0049] Furthermore, S5 specifically includes the following steps:

[0050] S51: Acquire new nonlinear data for the ship's central cooler x n And obtain its projection value on the optimal projection vector, the expression of which is:

[0051]

[0052] In the formula: g(x) n ) represents nonlinear data x n The projection value on the optimal projection vector; w * Represents the optimal projection vector; Represents x in the feature space n High-dimensional data;

[0053] S52: Obtain data samples of normal monitoring data class and abnormal monitoring data class respectively, and obtain the first threshold range and the second threshold range by the mean of the optimal projection vector, so as to realize the status assessment of the ship's central cooler.

[0054] Furthermore, the judgment rule for assessing the condition of the ship's central cooler is as follows:

[0055] If the nonlinear data x is determined n The projection value g(x) n If the data meets the first threshold range, then the nonlinear data x is confirmed. n The category is normal monitoring data:

[0056] If the nonlinear data x is determined n The projection value g(x) n If the second threshold range is met, then the nonlinear data x is confirmed. n The category is anomaly monitoring data:

[0057] Otherwise, confirm the nonlinear data x n This refers to the process data from normal monitoring data to abnormal monitoring data.

[0058] Beneficial Effects: This invention provides a method for assessing the condition of a ship's central cooler based on kernel Fisher discriminant analysis. A training sample set is obtained by randomly sampling data from normal and abnormal monitoring data classes. The data samples in the training sample set are then mapped to a high-dimensional feature space. Based on the obtained inter-class and intra-class discrete matrices, an objective function is constructed to obtain the optimal projection vector. A generalized characteristic equation maximizing the objective function is obtained based on the kernel Fisher strategy, and the current optimal projection vector is obtained according to the generalized characteristic equation. By introducing a kernel function, the dimensionality of the original data can be increased, linearly segmenting nonlinearly inseparable data in a high-dimensional space, thus handling nonlinear data. New nonlinear data of the ship's central cooler is obtained, and the condition of the central cooler is assessed based on the optimal projection vector. That is, based on the changing trend of the projected values ​​of operating parameters, the development trend of the ship's central cooler condition can be determined, providing an effective means for early detection of repetitive failures in ship machinery. This solves the problem of difficulty in effectively assessing the real-time condition of ship central coolers due to the large variety, volume, and nonlinearity of the real-time condition parameters, resulting in poor assessment accuracy. This achieves more accurate and faster condition discriminant analysis. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart of the ship's central cooler condition assessment method based on kernel Fisher discriminant analysis according to the present invention;

[0061] Figure 2 This is the core block diagram of the ship's central cooler condition assessment method based on kernel Fisher discriminant analysis in this embodiment;

[0062] Figure 3 This is a simulation diagram of the projected values ​​of normal data, process data, and abnormal data in this embodiment. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0064] This embodiment provides a method for assessing the condition of a ship's central cooler based on kernel Fisher discriminant analysis, such as... Figures 1 to 2 As shown, it includes the following steps:

[0065] S1: Acquire historical nonlinear data of the ship's central cooler;

[0066] Furthermore, the historical nonlinear data includes normal monitoring data and abnormal monitoring data.

[0067] The monitored normal data category refers to the inlet and outlet temperatures and pressures of the high-temperature / low-temperature side seawater pipelines of the ship's central cooler under normal conditions.

[0068] The abnormal monitoring data category refers to the inlet and outlet temperatures and pressures of the high-temperature / low-temperature side seawater pipelines under abnormal conditions of the ship's central cooler.

[0069] Specifically, the cooling system of a ship's central cooler is simulated using a central cooling water system in the laboratory. Data from pressure and temperature sensors are collected during normal and stable operation, including at least the seawater inlet and outlet pressures and temperatures on both the low-temperature and high-temperature sides. That is, keeping the heat source constant, after the central cooling system has been running for a long time, there will be some scale buildup between the heat exchange plates of the central cooler. At this time, the inlet and outlet temperatures and pressures will be affected to some extent. The data recorded during this period are data from abnormal conditions.

[0070] S2: Randomly select data samples from the normal monitoring data class and the abnormal monitoring data class respectively to obtain the training sample set;

[0071] Specifically, the training sample set is represented as (xi, yi)i = 1,2 ..., n, where x i Let y represent the i-th sample. i This indicates the label category corresponding to each sample, namely, normal monitoring data class or abnormal monitoring data class. There are a total of n samples and two label categories: normal monitoring data class ω1 and abnormal monitoring data class ω2.

[0072] S3: Based on a nonlinear mapping method, the data samples in the training sample set are mapped to a high-dimensional feature space; wherein the high-dimensional feature space is the data samples in the training sample set mapped to a preset high-dimensional space with a higher dimension than their data dimensions.

[0073] And obtain the inter-class discrete matrix and the intra-class discrete matrix of different data classes in the high-dimensional feature space; based on the inter-class discrete matrix and the intra-class discrete matrix, construct an objective function to obtain the optimal projection vector in the feature space;

[0074] Specifically, the expression for the inter-class discrete matrix of different data classes in the obtained high-dimensional feature space is as follows:

[0075]

[0076] In the formula: S b denoted as the inter-class discrete matrix; c represents the class of normal monitoring data or abnormal monitoring data; m represents the mean of all data samples in the training sample set; m c The mean of the c-th class of data samples is represented by n; n represents the number of data samples in the training sample set; n c This represents the number of data samples in class c; T represents the transpose. This represents the i-th data sample in the feature space;

[0077] The expression for the intra-class discrete matrix of the same data class is:

[0078]

[0079] In the formula: S w Represents a discrete matrix within a class; m c The mean of the c-th class of data samples is represented by n; n represents the number of data samples in the training sample set; n c This indicates the number of data samples in class c; This represents the i-th data sample of class c in the feature space;

[0080] The expression for the objective function of the projection vector is:

[0081]

[0082] In the formula: J(w) represents the objective function of the projection vector; w represents the projection vector.

[0083] S4: Obtain the generalized feature equation that maximizes the objective function based on the kernel Fisher strategy, and obtain the current best projection vector based on the generalized feature equation;

[0084] Specifically, the kernel Fisher strategy is a method to obtain a generalized characteristic equation that maximizes the objective function by combining the objective function of the projection vector obtained by the Fisher algorithm with a kernel function. The process of obtaining the generalized characteristic equation that maximizes the objective function based on the kernel Fisher strategy and obtaining the current optimal projection vector based on the generalized characteristic equation includes the following steps:

[0085] S41: To achieve data classification, the inter-class divergence should be maximized while the intra-class divergence should be minimized, i.e., the objective function should be maximized. The optimal projection direction should be found in the feature space. Solving the maximization problem can be achieved by solving the following generalized eigenvalue problem.

[0086] The generalized characteristic equation for maximizing the objective function is obtained based on the objective function of the projection vector, and its expression is:

[0087] S b w=λS w w

[0088] In the formula: λ represents the eigenvalue;

[0089] S42: However, since S is calculated directly in the high-dimensional feature space... w ,S b It is not feasible to use a kernel function to convert it into the input space for computation. Here, the Gaussian kernel function, which has the advantages of strong nonlinear mapping ability, fast computation speed and strong adjustability, is selected as the kernel function used.

[0090] The generalized characteristic equation is then rewritten based on the kernel Fisher strategy;

[0091] The rewriting process includes the following steps:

[0092] S421: Use the Gaussian kernel function as the kernel function of the kernel Fisher strategy, and define the kernel matrix according to the kernel function, the expression of which is:

[0093]

[0094] K ij =K(x) i ,x j )

[0095] In the formula: K represents the kernel matrix of the kernel Fisher policy, and is K ij The abbreviation of ; σ represents the kernel parameter; K(x,y) represents the Gaussian kernel function; x,y represent different data points, i.e., x i ,x j ;

[0096] S422: From the inter-class discrete matrix S b With the intraclass discrete matrix S w able to obtain

[0097] S w =φHWHφ T

[0098] S b =φHBHφ T

[0099] In the formula: H represents the centered matrix and I represents an n*n identity matrix; e represents an n-dimensional vector of all 1s; B and W both represent diagonal matrices, where if x i If it belongs to class c, then W ii =1, otherwise 0, if x i If it belongs to category c, then B ii =n c Otherwise, it is 0; φ represents The matrix; φ T Indicates the transpose of φ;

[0100] S423: Based on steps S421 and S422, convert the inter-class discrete matrix S b With the intraclass discrete matrix S w Rewritten as

[0101] S w =KHWHK

[0102] S b =KHBHK

[0103] S424: The generalized characteristic equation is rewritten from step S423 as follows:

[0104] KHBHKv=λKHWHKv

[0105] In the formula: K represents the kernel matrix of the kernel Fisher policy, and is K ij The abbreviation of ; v represents the feature vector;

[0106] S425: By solving the generalized characteristic equation rewritten in step S424, the current optimal projection vector w is obtained. * The expression is

[0107] w * =φv.

[0108] Since the parameters of the cooling water system are mostly nonlinear, a kernel function is introduced to address the nonlinearity problem. Kernel Fisher discriminant analysis, combining kernel functions and the Fisher algorithm, is used to model the data. A Gaussian kernel function, known for its strong nonlinear mapping capability, fast computation speed, and high adjustability, is selected based on the characteristics of each kernel function. The kernel parameter σ is determined using a trial-and-error method. Specifically, kernel parameters σ = 1, 5, 10, 20, 30, and 40 are used for experiments. Two sets of random data (x, y) without physical meaning are classified using the kernel function and the Fisher algorithm. The results show that a smaller kernel parameter σ results in stricter classification, but misclassification between the two classes is more likely. A larger kernel parameter σ results in looser classification, but there is a possibility of incorrectly classifying other data into the first or second class. To visually demonstrate the classification effect of kernel Fisher discriminant analysis and ensure its effectiveness in engineering applications, this paper selects a kernel parameter σ of 20.

[0109] It also obtains data samples of normal monitoring data class and abnormal monitoring data class, the current mean of the current best projection vector, and compares the current mean with the preset mean threshold.

[0110] If the current mean satisfies the preset mean threshold, then the current best projection vector is the optimal projection vector; otherwise, the kernel parameters of the kernel function in the kernel Fisher strategy are adaptively adjusted through the backpropagation algorithm and steps S2 to S3 are repeated.

[0111] S5: Acquire new nonlinear data of the ship's central cooler and perform a state assessment of the ship's central cooler based on the optimal projection vector. This includes the following steps:

[0112] S51: Acquire new nonlinear data for the ship's central cooler x n And obtain its projection value on the optimal projection vector, the expression of which is:

[0113]

[0114] In the formula: g(x) n ) represents nonlinear data x n The projection value on the optimal projection vector; w * Represents the optimal projection vector; Represents x in the feature space n High-dimensional data;

[0115] S52: Obtain data samples of normal monitoring data class and abnormal monitoring data class respectively, and obtain the first threshold range and the second threshold range by the mean of the optimal projection vector, so as to realize the status assessment of the ship's central cooler.

[0116] Furthermore, the judgment rule for assessing the condition of the ship's central cooler is as follows:

[0117] If the nonlinear data x is determined n The projection value g(x) n If the data meets the first threshold range, then the nonlinear data x is confirmed. n The category is normal monitoring data:

[0118] If the nonlinear data x is determined n The projection value g(x) n If the second threshold range is met, then the nonlinear data x is confirmed. n The category is anomaly monitoring data:

[0119] Otherwise, confirm the nonlinear data x n This refers to process data that transitions from normal monitoring data to abnormal monitoring data.

[0120] Specifically, the process data refers to the inlet and outlet temperatures and pressures of the high-temperature and low-temperature seawater pipes during the long-term operation of the central cooling system, recorded as process data. In this embodiment, the projection of the process data onto the optimal projection direction is obtained, and then... Figure 3 The distribution of projected values ​​for normal, process, and abnormal data can be observed. Projected values ​​are higher in normal states and lower in abnormal states. Process data shows a gradual change, converging from normal to abnormal data. The results are consistent with actual conditions, further validating the effectiveness of kernel Fisher discriminant analysis in assessing the condition of ship machinery and equipment in practical engineering.

[0121] In summary, the beneficial effects of this embodiment are as follows:

[0122] First, this application can effectively identify the normal and abnormal operating conditions of the central cooler, and can accurately describe the changes in process conditions through projected values. In the early stages of fault development, the trend of changes in the projected values ​​of operating parameters can be used to determine the development trend of the ship's system equipment status, providing an effective means for the early detection of recurring faults in ship machinery and equipment, and has significant practical engineering application value.

[0123] Secondly, this application combines the Fisher algorithm and kernel functions into a kernel Fisher discriminant analysis method. The introduced kernel function can increase the dimensionality of the original data, linearly partitioning nonlinearly inseparable data in a high-dimensional space. It can handle nonlinear data and has the advantages of high efficiency, flexibility, and strong universality.

[0124] Third, this application not only considers the dispersion of sample data of the same class, but also the dispersion between sample data of different classes. It can obtain a more accurate recognition model with fewer dimensions.

[0125] Fourth, this application can greatly reduce the computational complexity after kernel mapping and avoid the "curse of dimensionality".

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions 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 method for assessing the condition of a ship's central cooler based on kernel Fisher discriminant analysis, characterized in that, Includes the following steps: S1: Acquire historical nonlinear data of the ship's central cooler; Furthermore, the historical nonlinear data includes normal monitoring data and abnormal monitoring data. The normal monitoring data category refers to the inlet and outlet temperatures and pressures of the high-temperature or low-temperature side seawater pipelines of the ship's central cooler under normal conditions. The abnormal monitoring data category refers to the inlet and outlet temperatures and pressures of the high-temperature or low-temperature side seawater pipelines of the ship's central cooler under abnormal conditions. S2: Randomly select data samples from the normal monitoring data class and the abnormal monitoring data class respectively to obtain the training sample set; S3: Based on a nonlinear mapping method, the data samples in the training sample set are mapped to a high-dimensional feature space; And obtain the inter-class discrete matrix and the intra-class discrete matrix of different data classes in the high-dimensional feature space; Based on the inter-class discrete matrix and the intra-class discrete matrix, construct an objective function to obtain the optimal projection vector in the feature space; S4: Obtain the generalized feature equation that maximizes the objective function based on the kernel Fisher strategy, and obtain the current best projection vector based on the generalized feature equation. This includes the following steps: S41: Obtain the generalized characteristic equation that maximizes the objective function based on the objective function of the projection vector. Its expression is as follows: In the formula: Represents eigenvalues; Represents the projection vector; S42: Rewrite the generalized characteristic equation based on the kernel Fisher strategy; The rewriting process includes the following steps: S421: Use the Gaussian kernel function as the kernel function of the kernel Fisher strategy, and define the kernel matrix according to the kernel function, the expression of which is: In the formula: K Denotes the kernel matrix of the kernel Fisher policy, and is The abbreviated form; Indicates kernel parameters; Represents the Gaussian kernel function; Different data points are represented as ; S422: From the discrete matrix between classes and intraclass discrete matrix able to obtain In the formula: Represents a centered matrix and ; I n n represents the identity matrix; e represents an n-dimensional vector of all 1s; and W Both represent diagonal matrices; express Matrix; express transpose; S423: Based on steps S421 and S422, the inter-class discrete matrix is... and intraclass discrete matrix Rewritten as S424: The generalized characteristic equation is rewritten from step S423 as follows: In the formula: K Denotes the kernel matrix of the kernel Fisher policy, and is The abbreviated form; Represents the eigenvector; S425: By solving the generalized characteristic equation rewritten in step S424, the current optimal projection vector is obtained. The expression is ； It also obtains data samples of normal monitoring data class and abnormal monitoring data class, the current mean of the current best projection vector, and compares the current mean with the preset mean threshold. If the current mean satisfies the preset mean threshold, then the current best projection vector is the optimal projection vector; otherwise, the parameter weights of the kernel parameters in the kernel Fisher policy are adaptively adjusted and steps S2 to S3 are repeated. S5: Acquire new nonlinear data of the ship's central cooler and perform a state assessment of the ship's central cooler based on the optimal projection vector. This includes the following steps: S51: Acquire new nonlinear data for the ship's central cooler And obtain its projection value on the optimal projection vector, the expression of which is: In the formula: Representing nonlinear data The projection value on the optimal projection vector; Represents the optimal projection vector; In the characteristic space High-dimensional data; S52: Obtain data samples of normal monitoring data class and abnormal monitoring data class respectively, and obtain the first threshold range and the second threshold range by the mean of the optimal projection vector, so as to realize the status assessment of the ship's central cooler. Furthermore, the judgment rule for assessing the condition of the ship's central cooler is as follows: If the nonlinear data is determined projection value If the first threshold range is met, then the nonlinear data is confirmed. The category is normal monitoring data: If the nonlinear data is determined projection value If the second threshold range is met, then the nonlinear data is confirmed. The category is anomaly monitoring data: Otherwise, confirm the nonlinear data. This refers to the process data from normal monitoring data to abnormal monitoring data.

2. The method for assessing the condition of a ship's central cooler based on kernel Fisher discriminant analysis according to claim 1, characterized in that, The expression for the inter-class discrete matrix of different data classes in the high-dimensional feature space obtained in S3 is: , In the formula: Represents the discrete matrix between classes; c Indicates the category of monitoring data and ,Right now This indicates normal monitoring data. This represents an anomaly monitoring data class; This represents the mean of all data samples in the training sample set; Indicates the first c The mean of a sample of class data; This indicates the number of data samples in the training sample set; Indicates the first c The number of data samples in each class; Indicates transpose; Represents the first in the feature space i One data sample; The expression for the intra-class discrete matrix of the same data class is: , In the formula: Represents a discrete matrix within a class; Indicates the first c The mean of a sample of class data; This indicates the number of data samples in the training sample set; Indicates the first c The number of data samples in each class; Represents the first in the feature space c The class of i One data sample; The expression for the objective function of the projection vector is: In the formula: The objective function representing the projection vector; This represents the projection vector.

Images