阅读说明：本技术 基于多维向量模型的船舶关键设备系统的监测方法 (Monitoring method of ship key equipment system based on multi-dimensional vector model ) 是由 陈冬梅 黄滔 徐在强 申志泽 周航 徐佩蓉 于 2018-09-03 设计创作，主要内容包括：本发明公开了基于多维向量模型的船舶关键设备系统的监测方法,该方法包括：S1：构建所述船舶关键设备系统的系统多维向量监测模型；S2：实时监测所述船舶关键设备系统的工况条件,按照当前工况,调用相应工况条件下的系统多维向量监测模型；S3：判断是否存在与子模块的特征参数所对应的多维向量未落入相应的正常态多维向量集,若是,则判断在所述当前工况下所述子模块处于故障状态；若否,则判断在所述当前工况下所述子模块处于正常状态。本发明的监测方法只需调用当前工况的各个子多维向量监测模型,即可实时监测船舶设备系统各部分的运行状态,且不需要通过异常数据样本来实现子多维向量监测模型的构建,使得适用性更加广泛。(The invention discloses a monitoring method of a ship key equipment system based on a multidimensional vector model, which comprises the following steps: s1: constructing a system multidimensional vector monitoring model of the ship key equipment system; s2: monitoring the working condition of the ship key equipment system in real time, and calling a system multidimensional vector monitoring model under the corresponding working condition according to the current working condition; s3: judging whether multidimensional vectors corresponding to the characteristic parameters of the sub-module do not fall into a corresponding normal-state multidimensional vector set or not, and if yes, judging that the sub-module is in a fault state under the current working condition; if not, the submodule is judged to be in a normal state under the current working condition. According to the monitoring method, the running states of all parts of the ship equipment system can be monitored in real time only by calling all the sub-multi-dimensional vector monitoring models under the current working condition, and the sub-multi-dimensional vector monitoring models are not required to be constructed through abnormal data samples, so that the applicability is wider.)

1. A monitoring method of a ship key equipment system based on a multi-dimensional vector model is characterized by comprising the following steps:

s1: constructing a system multidimensional vector monitoring model of the ship key equipment system;

s2: monitoring the working condition of the ship key equipment system in real time, and calling a system multidimensional vector monitoring model under the corresponding working condition according to the current working condition;

s3: judging whether multidimensional vectors corresponding to the characteristic parameters of the sub-module do not fall into a corresponding normal-state multidimensional vector set or not, and if yes, judging that the sub-module is in a fault state under the current working condition; if not, judging that the sub-module is in a normal state under the current working condition,

wherein the S1 includes:

s11: dividing the ship key equipment system into submodules;

s12: constructing a sub multi-dimensional vector monitoring model corresponding to each sub-module by utilizing characteristic parameters of each sub-module under various typical working conditions within a period of time of running of the sub-module in a healthy state based on a machine learning multi-dimensional vector model method, wherein the sub multi-dimensional vector monitoring model of each sub-module comprises a normal state multi-dimensional vector sample set;

s13: and constructing a system multi-dimensional vector monitoring model comprising the sub multi-dimensional vector monitoring models of the sub-modules by utilizing testability modeling software.

2. The monitoring method according to claim 1, wherein the S12 includes:

s121: selecting characteristic parameters of a sub multi-dimensional vector monitoring model corresponding to the sub-modules;

s122: extracting a training sample set and a testing sample set according to the selected characteristic parameters, wherein the training sample set and the testing sample set are composed of multi-dimensional vectors corresponding to the characteristic parameters;

s123: initializing model parameters of the sub multi-dimensional vector monitoring model;

s124: optimizing the model parameters to construct a sub-multidimensional vector monitoring model for each of the sub-modules:

and training the sub-multidimensional vector monitoring model by using the training sample set, and verifying the sub-multidimensional vector monitoring model by using the testing sample set so as to optimize the model parameters to obtain the optimized sub-multidimensional vector monitoring model.

3. The monitoring method according to claim 2, wherein the S122 includes the steps of:

s1221: extracting and screening original data samples of the characteristic parameters corresponding to the sub-modules according to the time tags;

s1222: randomly sampling the original data sample extracted in the step S1221, visualizing the sampled data, and determining whether the original data sample is balanced, if so, going to the step S1223; if not, returning to the step S1221;

s1223: using the raw data samples extracted in the S1222 for training or testing of the sub-multidimensional vector monitoring model.

4. The monitoring method according to claim 2, wherein in the S123:

the model parameters include a penalty parameter nu and parameters of a kernel function, wherein the kernel function is obtained by searching and learning a corresponding training sample set.

5. The monitoring method according to claim 2, wherein the S124 includes the steps of:

s1241: the model parameters form a multi-dimensional parameter matrix;

s1242: searching a sub-multidimensional vector monitoring model trained by each group of model parameters in the multidimensional parameter matrix by using a training sample set corresponding to the sub-modules;

s1243: and verifying the accuracy of the same batch of training samples under different model parameters by using the test sample set corresponding to the sub-module, and selecting the optimal model parameter with the highest accuracy.

6. The method for monitoring set forth in claim 1, wherein in S2, each of the submodulesThe sub multi-dimensional vector monitoring model comprises a normal multi-dimensional vector sample set X1 ═ X₁₁，X₁₂，X₁₃...，X_1nH, and an abnormal-state multi-dimensional vector sample set X2 ═ X₂₁，X₂₂，X₂₃...，X_2n}；

After judging that the sub-modules of the ship key equipment system are in the fault state under the current working condition, the step of indicating specific fault information comprises the following steps:

s21: calculating the discrimination distance D between the normal state multi-dimensional vector sample set and the abnormal state multi-dimensional vector sample set of each characteristic parameter in the sub multi-dimensional vector monitoring model of the sub-module with the fault, wherein the calculation formula is as follows:

wherein m1 is the mean vector of the calculated feature parameters in the X1, m2 is the mean of the calculated feature parameters in the X2, s1 is the variance of the calculated feature parameters in the X1, and s2 is the variance of the calculated feature parameters in the X2;

s22: sorting the discrimination distances D of the characteristic parameters, wherein the larger the discrimination distance D is, the larger the deviation degree of the characteristic parameters from a normal multi-dimensional vector sample set is;

s23: and outputting specific fault characteristic parameters from large to small according to the judgment distance D.

Technical Field

The invention relates to the technical field of monitoring design of a ship key equipment system, in particular to a monitoring method of a ship key equipment system based on a multi-dimensional vector model.

Background

The state monitoring and maintenance of ship equipment are main means for guaranteeing the normal operation of the equipment, improving the working efficiency of the equipment and prolonging the service life of the equipment. The monitoring of the existing equipment mostly adopts threshold value alarm to monitor the running state, which can solve the problem of some conventional faults, but under the background that the working conditions of the ship equipment are changeable, the numerical value of the same parameter under different working conditions greatly fluctuates, sometimes, in order to avoid frequent alarm, a larger margin is reserved for the threshold value, so that a more serious fault often occurs during alarm, and the defect that the equipment state monitoring can not be flexibly carried out according to the real-time working conditions exists in the prior art; and for more complicated equipment, the parameters needing to be monitored are many, and when a fault occurs, the main and secondary factors causing the fault are difficult to be quickly positioned manually.

Therefore, there is a need to provide a monitoring method for a ship critical equipment system based on a multi-dimensional vector model to at least partially solve the above-mentioned problems.

Disclosure of Invention

In this summary, concepts in a simplified form are introduced that are further described in the detailed description section. This summary of the invention is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

In order to at least partially solve the above problems, the present invention provides a monitoring method for a ship key equipment system based on a multidimensional vector model, comprising the following steps:

s1: constructing a system multidimensional vector monitoring model of the ship key equipment system;

s2: monitoring the working condition of the ship key equipment system in real time, and calling a system multidimensional vector monitoring model under the corresponding working condition according to the current working condition;

s3: judging whether multidimensional vectors corresponding to the characteristic parameters of the sub-module do not fall into a corresponding normal-state multidimensional vector set or not, and if yes, judging that the sub-module is in a fault state under the current working condition; if not, judging that the sub-module is in a normal state under the current working condition,

wherein the S1 includes:

s11: dividing the ship key equipment system into submodules;

s12: constructing a sub multi-dimensional vector monitoring model corresponding to each sub-module by utilizing characteristic parameters of each sub-module under various typical working conditions within a period of time of running of the sub-module in a healthy state based on a machine learning multi-dimensional vector model method, wherein the sub multi-dimensional vector monitoring model of each sub-module comprises a normal state multi-dimensional vector sample set;

s13: and constructing a system multi-dimensional vector monitoring model comprising the sub multi-dimensional vector monitoring models of the sub-modules by utilizing testability modeling software.

According to the monitoring method of the ship key equipment system based on the multi-dimensional vector model, the ship equipment system is divided into sub-modules according to the composition structure, the corresponding sub-multi-dimensional vector monitoring models are constructed by adopting the characteristic parameters of the sub-modules in various typical working conditions within a period of time when the sub-modules operate in a healthy state, so that the multi-dimensional vector monitoring model of the system is formed, and in the monitoring process, the operation states of all parts of the ship equipment system can be monitored in real time only by calling the sub-multi-dimensional vector monitoring models in the current working conditions.

Preferably, the S12 includes:

s121: selecting characteristic parameters of a sub multi-dimensional vector monitoring model corresponding to the sub-modules;

s122: extracting a training sample set and a testing sample set according to the selected characteristic parameters, wherein the training sample set and the testing sample set are composed of multi-dimensional vectors corresponding to the characteristic parameters;

s123: initializing model parameters of the sub multi-dimensional vector monitoring model;

s124: optimizing the model parameters to construct a sub-multidimensional vector monitoring model for each of the sub-modules:

and training the sub-multidimensional vector monitoring model by using the training sample set, and verifying the sub-multidimensional vector monitoring model by using the testing sample set so as to optimize the model parameters to obtain the optimized sub-multidimensional vector monitoring model.

Therefore, the optimized sub-multidimensional vector monitoring model of the sub-module can be obtained.

Preferably, the S122 includes the steps of:

s1221: extracting and screening original data samples of the characteristic parameters corresponding to the sub-modules according to the time tags;

s1222: randomly sampling the original data sample extracted in the step S1221, visualizing the sampled data, and determining whether the original data sample is balanced, if so, going to the step S1223; if not, returning to the step S1221;

s1223: using the raw data samples extracted in the S1222 for training or testing of the sub-multidimensional vector monitoring model.

Thus, a suitable training sample set and test sample set can be obtained.

Preferably, in S123:

the model parameters include a penalty parameter nu and parameters of a kernel function, wherein the kernel function is obtained by searching and learning a corresponding training sample set.

Preferably, the S124 includes the steps of:

s1241: the model parameters form a multi-dimensional parameter matrix;

s1242: searching a sub-multidimensional vector monitoring model trained by each group of model parameters in the multidimensional parameter matrix by using a training sample set corresponding to the sub-modules;

s1243: and verifying the accuracy of the same batch of training samples under different model parameters by using the test sample set corresponding to the sub-module, and selecting the optimal model parameter with the highest accuracy.

Therefore, the optimal model parameters of the sub multi-dimensional vector monitoring model can be obtained.

Preferably, in said S2, the first step of said first step,

the sub-multidimensional vector monitoring model of each sub-module comprises a normal-state multidimensional vector sample set X1 ═ X₁₁，X₁₂，X₁₃…,X_1nH, and an abnormal-state multi-dimensional vector sample set X2 ═ X₂₁，X₂₂，X₂₃…,X_2n}；

After judging that the sub-modules of the ship key equipment system are in the fault state under the current working condition, the step of indicating specific fault information comprises the following steps:

s21: calculating the discrimination distance D between the normal state multi-dimensional vector sample set and the abnormal state multi-dimensional vector sample set of each characteristic parameter in the sub multi-dimensional vector monitoring model of the sub-module with the fault, wherein the calculation formula is as follows:

wherein m1 is the mean vector of the calculated feature parameters in the X1, m2 is the mean of the calculated feature parameters in the X2, s1 is the variance of the calculated feature parameters in the X1, and s2 is the variance of the calculated feature parameters in the X2;

s22: sorting the discrimination distances D of the characteristic parameters, wherein the larger the discrimination distance D is, the larger the deviation degree of the characteristic parameters from a normal multi-dimensional vector sample set is;

s23: and outputting specific fault characteristic parameters from large to small according to the judgment distance D.

Therefore, specific fault characteristic parameters can be further pointed out by calculating the discrimination distance D of each characteristic parameter and comparing the sizes, and clear guide information is provided for subsequent fault diagnosis and assistant decision.

Drawings

The following drawings of embodiments of the invention are included as part of the present invention for an understanding of the invention. The drawings illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention. In the drawings, there is shown in the drawings,

FIG. 1 is a flow diagram of a preferred embodiment of a method for monitoring a critical equipment system of a ship based on a multidimensional vector model according to the present invention;

FIG. 2 is a detailed flowchart of step S12 in FIG. 1;

FIG. 3 is a detailed flowchart of step S122 in FIG. 2;

FIG. 4 is a detailed flowchart of step S124 in FIG. 2;

FIG. 5 is a detailed flowchart of step 2 in FIG. 1;

fig. 6 is a schematic diagram of a ship cabin main power equipment system divided into sub-modules according to the monitoring method of the ship key equipment system based on the multidimensional vector model.

Detailed Description

In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. It will be apparent, however, to one skilled in the art, that embodiments of the invention may be practiced without one or more of these specific details. In other instances, well-known features have not been described in detail so as not to obscure the embodiments of the invention.

In the following description, a detailed structure will be presented for a thorough understanding of embodiments of the invention. It is apparent that the implementation of the embodiments of the present invention is not limited to the specific details familiar to those skilled in the art. The following detailed description of the preferred embodiments of the invention, however, the invention is capable of other embodiments in addition to those detailed.

Hereinafter, preferred embodiments according to the present invention will be described in detail.

Fig. 1 shows a flow chart of a monitoring method of a ship critical equipment system based on a multidimensional vector model according to the invention. Specifically, the monitoring method comprises the following steps:

s1: constructing a system multidimensional vector monitoring model of a ship key equipment system;

s2: monitoring working conditions of a key equipment system of a ship in real time, and calling a system multidimensional vector monitoring model under the corresponding working conditions according to the current working conditions;

s3: judging whether multidimensional vectors corresponding to the characteristic parameters of the sub-modules do not fall into a corresponding normal-state multidimensional vector set or not, and if so, judging that the sub-modules are in a fault state under the current working condition; if not, the submodule is judged to be in a normal state under the current working condition,

wherein, step S1 includes:

s11: dividing a ship key equipment system into submodules;

s12: the method comprises the steps that a multi-dimensional vector model method based on machine learning is adopted, characteristic parameters of each sub-module in a period of healthy state operation and under various typical working conditions are utilized to construct sub-multi-dimensional vector monitoring models corresponding to the sub-modules, and the sub-multi-dimensional vector monitoring models of each sub-module comprise normal state multi-dimensional vector sample sets;

s13: and constructing a system multi-dimensional vector monitoring model comprising the sub multi-dimensional vector monitoring models of the sub-modules by utilizing testability modeling software.

The key equipment of the ship key equipment system may be an engine room, a ship power station, electrical equipment, and the like. Referring to fig. 6, a schematic diagram of a marine nacelle primary power plant system divided into submodules is shown, with each plant being the smallest unit of the submodules.

In step S12, the health status of the sub-module means that the sub-module is in good operation status at the initial stage of commissioning, and the characteristic parameters corresponding to the sub-module at this time are considered to be healthy data samples, so that the sub-multidimensional vector monitoring model is constructed.

In step S3, after the sub-modules are determined to be in the normal state under the current working condition, the method returns to step S2 to continue the real-time monitoring of the operating states of the sub-modules.

According to the monitoring method of the ship key equipment system based on the multi-dimensional vector model, the ship equipment system is divided into the sub-modules, the corresponding sub-multi-dimensional vector monitoring models are constructed by adopting the characteristic parameters of the sub-modules in various typical working conditions within a period of time when the sub-modules operate in a healthy state, so that the multi-dimensional vector monitoring model of the system is formed, and in the monitoring process, the operation states of all parts of the ship equipment system can be monitored in real time only by calling the sub-multi-dimensional vector monitoring models in the current working conditions.

Specifically, as shown in fig. 2, step S12 may include:

s121: selecting characteristic parameters of a sub multi-dimensional vector monitoring model corresponding to the sub-modules;

s122: extracting a training sample set and a test sample set according to the selected characteristic parameters, wherein the training sample set and the test sample set are composed of multidimensional vectors corresponding to the characteristic parameters;

s123: initializing model parameters of a sub-multidimensional vector monitoring model;

s124: optimizing the model parameters to construct a sub-multidimensional vector monitoring model for each sub-module:

and training the sub multi-dimensional vector monitoring model by using the training sample set, and verifying the sub multi-dimensional vector monitoring model by using the testing sample set so as to optimize the model parameters to obtain the optimized sub multi-dimensional vector monitoring model.

In step S121, continuing to refer to fig. 2, the supercharger in fig. 6 is used as a sub-module of the main power plant system of the ship cabin, and through mechanism analysis, the selected parameters include: the system comprises a supercharger, a supercharger vibration effective value, a supercharger lubricating oil inlet pressure, a supercharger lubricating oil outlet pressure, a supercharger inlet exhaust temperature and a supercharger outlet exhaust temperature, wherein the supercharger rotating speed is used for reflecting the working condition of the supercharger; the effective value of the vibration of the supercharger is used for reflecting the bearing failure of the supercharger and the dynamic balance change of a rotor of the supercharger; the pressure of a supercharger lubricating oil inlet and the pressure of a supercharger lubricating oil outlet are used for reflecting the lubricating condition of the supercharger; the supercharger inlet exhaust temperature and the supercharger outlet exhaust temperature are used for reflecting the work efficiency of the supercharger.

Specifically, as shown in fig. 3, step S122 may include:

s1221: extracting and screening original data samples of the characteristic parameters corresponding to the sub-modules according to the time tags;

s1222: randomly sampling the original data sample extracted in the step S1221, visualizing the sampled data, and determining whether the original data sample is balanced, if so, going to the step S1223; if not, returning to the step S1221;

s1223: the raw data samples extracted in step S1222 are used for training or testing of the sub-multidimensional vector monitoring model.

In this embodiment, that is, in steps S1221 to S1223, a training sample set and a test sample set of the corresponding sub-multidimensional vector monitoring model are extracted from each sub-module of the governor, the supercharger, the lube pump, the electric starter, the seawater cooler, the fuel delivery pump, the fuel duplex filter, and the like, and may be stored in the CSV format. The training sample set and the testing sample set of the sub-multidimensional vector monitoring model corresponding to the supercharger are formed by utilizing multidimensional vectors corresponding to the data of the 6 characteristic parameters in a period of time when the supercharger operates in a healthy state.

Specifically, in step S123:

the model parameters include a penalty parameter nu and parameters of a kernel function, wherein the kernel function is obtained by searching and learning a corresponding training sample set.

It should be noted that nu (penalty parameter) is an important parameter affecting the state monitoring of the sub-multidimensional vector monitoring model, i.e. tolerance to errors. The larger nu is, the less tolerable errors are, the easier overfitting is, and the smaller nu is, the easier underfitting is. nu is too large or too small, which leads to poor generalization ability of the model.

The kernel functions selected are different and their corresponding parameters are different. In this embodiment, the sub-multidimensional vector monitoring model of the supercharger may select an RBF (radial Basis function) kernel function, where the RBF kernel function has a gamma coefficient, and the gamma is a parameter of the kernel function after the RBF kernel function is selected. The distribution of the data after being mapped to a new feature space is determined implicitly, the larger the gamma is, the fewer the support vectors are, and the smaller the gamma value is, the more the support vectors are. The number of support vectors affects the speed of training and prediction.

Specifically, as shown in fig. 4, step S124 may include:

s1241: the model parameters form a multi-dimensional parameter matrix;

s1242: searching a sub-multidimensional vector monitoring model trained by each group of model parameters in the multidimensional parameter matrix by using a training sample set corresponding to the sub-modules;

s1243: and verifying the accuracy of the same batch of training samples under different model parameters by using the test sample set corresponding to the sub-module, and selecting the optimal model parameter with the highest accuracy.

In this embodiment, a supercharger is taken as an example to describe in detail how to determine a sub-multidimensional vector monitoring model corresponding to the supercharger, and how to determine the sub-multidimensional vector monitoring model by other devices will not be described again:

model parameters of the sub-multidimensional vector monitoring model of the supercharger comprise nu (penalty parameter) and gamma coefficients of the RBF kernel function. The sub-multidimensional vector monitoring model is optimized by searching a model trained by each group of parameters in a two-dimensional parameter matrix consisting of nu and gamma, then, verifying by using a test sample set, and counting the accuracy of the same batch of training samples under different groups of parameters. Preferably, in order to ensure the accuracy of the result, a plurality of test samples can be used for verification, and the nu and the gamma with the highest accuracy are selected as the optimal ones. After the optimal model parameters nu and gamma are determined, the sub-multidimensional vector monitoring model corresponding to the supercharger can be trained by using the optimal model parameters.

The sub-multidimensional vector monitoring model of each sub-module after training has

Normal state multidimensionalVector sample set X1 ═ X₁₁，X₁₂，X₁₃…,X_1nH, and an abnormal-state multi-dimensional vector sample set X2 ═ X₂₁，X₂₂，X₂₃…,X_2n}，

Specifically, as shown in fig. 5, after the sub-module of the ship key equipment system under the current working condition is judged to be in the fault state, specific fault information can be further indicated, and the step includes:

s21: calculating the discrimination distance D between the normal state multi-dimensional vector sample set and the abnormal state multi-dimensional vector sample set of each characteristic parameter in the sub multi-dimensional vector monitoring model of the sub-module with the fault, wherein the calculation formula is as follows:

where m1 is the mean vector of the calculated feature parameters in X1, m2 is the mean of the calculated feature parameters in X2, s1 is the variance of the calculated feature parameters in X1, and s2 is the variance of the calculated feature parameters in X2;

s22: sorting the discrimination distances D of the characteristic parameters, wherein the larger the discrimination distance D is, the larger the deviation degree of the characteristic parameters from the normal multi-dimensional vector sample set is;

s23: and outputting specific fault characteristic parameters from large to small according to the judgment distance D.

In the present embodiment, assuming that the supercharger has a failure, the discrimination distance D between the normal-state multidimensional vector sample set and the abnormal-state multidimensional vector sample set of the 6 feature parameters is calculated, and the obtained 6 discrimination distances are sorted, and the greater the discrimination distance D, the greater the degree of deviation of the feature parameter from the normal-state multidimensional vector sample set, the more likely it is to be the main failure parameter.

According to the monitoring method of the ship key equipment system based on the multi-dimensional vector model, the ship equipment system is divided into sub-modules, characteristic parameters of the sub-modules in various typical working conditions are adopted within a period of time when the sub-modules operate in a healthy state, corresponding sub-multi-dimensional vector monitoring models are constructed, and then the multi-dimensional vector monitoring model of the system is formed; in addition, the method does not need to realize the construction of the model through abnormal data samples, so that the applicability is wider.

Unless defined otherwise, technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. Terms such as "component" and the like, when used herein, can refer to either a single part or a combination of parts. Terms such as "mounted," "disposed," and the like, as used herein, may refer to one component as being directly attached to another component or one component as being attached to another component through intervening components. Features described herein in one embodiment may be applied to another embodiment, either alone or in combination with other features, unless the feature is otherwise inapplicable or otherwise stated in the other embodiment.

The present invention has been described in terms of the above embodiments, but it should be understood that the above embodiments are for purposes of illustration and description only and are not intended to limit the invention to the scope of the described embodiments. It will be appreciated by those skilled in the art that many variations and modifications may be made to the teachings of the invention, which fall within the scope of the invention as claimed.