CN111260893B - Fault early warning method and device for ocean platform propeller - Google Patents

Fault early warning method and device for ocean platform propeller Download PDF

Info

Publication number
CN111260893B
CN111260893B CN202010029817.7A CN202010029817A CN111260893B CN 111260893 B CN111260893 B CN 111260893B CN 202010029817 A CN202010029817 A CN 202010029817A CN 111260893 B CN111260893 B CN 111260893B
Authority
CN
China
Prior art keywords
propeller
characteristic parameter
vector
early warning
sample
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010029817.7A
Other languages
Chinese (zh)
Other versions
CN111260893A (en
Inventor
蒋爱国
王金江
谷明
于昊天
秦建安
李文锦
冼敏元
张来斌
黄玉泉
赵吉振
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China University of Petroleum Beijing
China Oilfield Services Ltd
China National Offshore Oil Corp CNOOC
Original Assignee
China University of Petroleum Beijing
China Oilfield Services Ltd
China National Offshore Oil Corp CNOOC
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China University of Petroleum Beijing, China Oilfield Services Ltd, China National Offshore Oil Corp CNOOC filed Critical China University of Petroleum Beijing
Priority to CN202010029817.7A priority Critical patent/CN111260893B/en
Publication of CN111260893A publication Critical patent/CN111260893A/en
Application granted granted Critical
Publication of CN111260893B publication Critical patent/CN111260893B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/18—Status alarms
    • G08B21/182—Level alarms, e.g. alarms responsive to variables exceeding a threshold
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/211—Selection of the most significant subset of features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B31/00—Predictive alarm systems characterised by extrapolation or other computation using updated historic data

Abstract

A failure early warning method for an ocean platform propeller comprises the following steps: acquiring a plurality of groups of characteristic parameter vector sets of an ocean platform propeller as a plurality of groups of sample data sets; performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrixes according to a preset relevance rule; selecting a sample matrix corresponding to the maximum value of the working condition space index in the reconstructed sample matrix as a process memory matrix; acquiring a plurality of characteristic parameter vectors of a propeller, and determining an estimation vector corresponding to each characteristic parameter vector according to a process memory matrix; calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector; determining an early warning threshold value of the early warning of the faults of the propeller according to the distance similarity; and acquiring the current characteristic parameters of the propeller in real time, and judging the fault state of the propeller according to the early warning threshold value. According to the scheme provided by the invention, the fault early warning accuracy of the ocean platform propeller is improved.

Description

Fault early warning method and device for ocean platform propeller
Technical Field
The present invention relates to oil drilling technology, and is especially failure warning method and device for marine platform propeller.
Background
The semi-submersible drilling platform has the advantages of large structure, high manufacturing cost, complex equipment, severe operation environment due to complex load effects of wind, sea waves, ocean currents and the like in complex and variable environments with high salt and high humidity. The thruster is used as key equipment of the semi-submersible drilling platform and is an important part for ensuring the normal operation of the platform. In the fields of fault diagnosis, machine learning and data mining, in order to reflect fault type information as comprehensively as possible, various sensors are generally required to collect a large amount of equipment information to characterize the real-time operation state of the equipment, however, most of the propeller diagnosis and early warning methods up to now usually rely on the processing of vibration signals, while the information of oil parameters, temperature and the like which are closely related to the operation state of the propeller cannot be effectively utilized, once the vibration signals are accidentally interfered, the fault early warning which depends on the vibration signals cannot be normally carried out, and thus serious consequences are caused.
The Multiple State Estimation Technology (MSET) utilizes data covering a normal operation state to perform process similarity modeling, so that identification of an equipment operation state and fault early warning are realized. Firstly, relevant parameters capable of representing the running state of a propeller need to be obtained by adopting a multivariate state estimation technology, and if the selected propeller state parameters cannot describe the real-time running condition of the propeller to the maximum extent, a fault early warning method model is established on the basis of the parameters, so that no representativeness exists obviously. Secondly, when fault early warning is carried out by using the MSET, the estimation of the running state of the propeller is realized by selecting historical running data under normal working conditions to construct a process memory matrix, and if normal sample data is extracted from the historical running data of the propeller by using a simple sampling mode, the extracted data cannot be mixed with some abnormal working conditions, so that the precision of fault early warning is influenced. In addition, after the sample data under the normal working condition is screened out, if the selected historical data of the normal working condition generally has larger repeatability, the size of the distribution space of the working condition of the normal sample can be greatly reduced, and the prediction precision of the model is directly influenced. Finally, when performing early warning judgment, expert experience usually becomes an important reference for early warning threshold calculation, but depending on subjective components excessively, the final calculation result usually has a large deviation, thereby resulting in early warning errors. Therefore, how to realize the marine platform propeller fault early warning method based on multivariate state estimation is an urgent problem to be solved.
Disclosure of Invention
The application provides a fault early warning method for an ocean platform propeller, which comprises the steps of selecting characteristic parameters capable of describing the running state of the propeller as much as possible, screening out the characteristic parameters of the running state under normal working conditions to serve as sample data, constructing a process memory matrix, determining a proper early warning threshold value, and improving the fault early warning precision of the ocean platform propeller.
The application provides a fault early warning method for an ocean platform propeller, which comprises the following steps:
acquiring a plurality of groups of characteristic parameter vector sets of an ocean platform propeller as a plurality of groups of sample data sets;
performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrixes according to a preset relevance rule;
selecting a sample matrix corresponding to the maximum value of the working condition space index in the reconstructed sample matrix as a process memory matrix;
acquiring a plurality of characteristic parameter vectors of a propeller, and determining an estimation vector corresponding to each characteristic parameter vector according to the process memory matrix;
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector;
determining an early warning threshold value of early warning of the faults of the propeller according to the distance similarity;
acquiring the current characteristic parameters of the propeller in real time, and judging whether the propeller fails according to the early warning threshold;
when the distance similarity between the current characteristic parameter vector and the estimation vector corresponding to the current characteristic parameter vector is smaller than an early warning threshold value, determining that the propeller has a fault;
and when the distance similarity between the current characteristic parameter vector and the estimation vector of the current characteristic parameter vector is greater than an early warning threshold value, determining that the propeller is normal.
In an exemplary embodiment, before the obtaining a plurality of sets of feature parameter vectors of a propeller of an offshore platform as a plurality of sets of sample data sets, the method further includes:
determining an arrangement scheme of the propeller sensors by using a preset algorithm;
the propeller sensor is arranged by adopting the arrangement scheme, and the characteristic parameters of the running state of the propeller are collected in real time.
In an exemplary embodiment, after obtaining a plurality of sets of feature parameter vectors of a propeller of an offshore platform as a plurality of sets of sample data sets, the method further includes:
and classifying the characteristic parameter vector set of the running state of the propeller by adopting a preset algorithm, removing the characteristic parameter vector of the abnormal working condition, and obtaining the characteristic parameter vector set of the propeller under the normal working condition.
In an exemplary embodiment, the performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrices according to a predetermined relevance rule includes:
calculating the correlation of each two groups of characteristic parameter vectors in the sample data according to a Person correlation algorithm;
obtaining a reconstructed sample matrix according to the relevance rule;
the Person correlation algorithm comprises:
Figure BDA0002362703780000031
wherein: m represents the number of characteristic parameters; xiAnd YiRespectively representing the ith parameter in the characteristic parameter vector X and Y,
Figure BDA0002362703780000032
and
Figure BDA0002362703780000033
respectively, the parameter means of the characteristic parameter vectors X and Y, and r represents the correlation between the characteristic parameter vectors X, Y.
In an exemplary embodiment, the association rule includes:
and deleting the sample data corresponding to the strong correlation quantity in the sample data correlation table.
In an exemplary embodiment, the selecting a sample matrix corresponding to a maximum value of a spatial indicator of an operating condition in a reconstructed sample matrix as a process memory matrix includes:
calculating the average distance from each sample in each group of reconstructed sample matrixes to the adjacent sample of the sample;
summing the average distances between all samples and adjacent samples to obtain an average value, wherein the average value represents a working condition space index;
the reconstructed sample matrix with the largest average value is selected as the process memory matrix.
In an exemplary embodiment, the obtaining a plurality of eigenvectors of the propeller, and determining an estimated vector corresponding to each of the eigenvectors according to the process memory matrix includes:
obtaining a plurality of characteristic parameter vectors of a propeller, and determining similarity measure vectors between the plurality of characteristic vectors and a process memory matrix by using a Mahalanobis distance algorithm;
and multiplying the process memory matrix and the similarity measure vector to obtain an estimation vector corresponding to each characteristic parameter vector.
In an exemplary embodiment, the calculating a distance similarity between each feature parameter vector and the corresponding estimation vector includes:
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector by adopting a Markov distance algorithm;
wherein, the horse-type distance algorithm comprises:
Figure BDA0002362703780000041
in the formula: xmon，XestRespectively representing the characteristic parameter vector and the corresponding estimation vector; s is a covariance matrix and the covariance matrix,
Figure BDA0002362703780000042
m is the number of feature variables of the feature parameters and the corresponding estimation vectors,
Figure BDA0002362703780000043
and
Figure BDA0002362703780000044
respectively representing a characteristic parameter vector XmonAnd estimate vector XestIs measured.
In an exemplary embodiment, the determining an early warning threshold for early warning of a propeller fault according to the distance similarity includes:
calculating the average value of the N distance similarities;
and calculating the average value by adopting a 3 sigma principle to obtain an early warning threshold value of the propeller fault early warning, wherein the 3 sigma principle formula comprises:
Figure BDA0002362703780000045
in the above formula, σ is an early warning threshold value of the early warning of the propeller fault, and N is the number of distance similarities;
Figure BDA0002362703780000046
the average value of the distance similarity is obtained; MDiIs distance similarity;
and normalizing the early warning threshold value.
In order to solve the above problems, the present invention further provides a device for early warning of a failure of an offshore platform propeller, the device comprising: a memory and a processor; the method is characterized in that:
the memory is used for storing a program for early warning of faults of the marine platform propeller;
the processor is used for reading and executing the program for early warning the failure of the marine platform propeller and executing the following operations:
acquiring a plurality of groups of characteristic parameter vector sets of an ocean platform propeller as a plurality of groups of sample data sets;
performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrixes according to a preset relevance rule;
selecting a sample matrix corresponding to the maximum value of the working condition space index in the reconstructed sample matrix as a process memory matrix;
acquiring a plurality of characteristic parameter vectors of a propeller, and determining an estimation vector corresponding to each characteristic parameter vector according to the process memory matrix;
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector;
determining an early warning threshold value of early warning of the faults of the propeller according to the distance similarity;
acquiring a current characteristic parameter vector in real time, and judging whether the propeller fails according to the early warning threshold;
when the distance similarity between the current characteristic parameter vector and the estimation vector corresponding to the current characteristic parameter vector is smaller than an early warning threshold value, determining that the propeller has a fault;
and when the distance similarity between the current characteristic parameter vector and the estimation vector of the current characteristic parameter vector is greater than an early warning threshold value, determining that the propeller is normal.
In an exemplary embodiment, before the obtaining the sets of vectors of the characteristic parameters of the plurality of sets of marine platform propellers as the sets of sample data, the processor further performs the following operations:
determining an arrangement scheme of the propeller sensors by using a preset algorithm;
the propeller sensor is arranged by adopting the arrangement scheme, and the characteristic parameters of the running state of the propeller are collected in real time.
In an exemplary embodiment, after obtaining the sets of vectors of the characteristic parameters of the plurality of sets of marine platform propellers as sets of sample data, the processor further performs the following operations:
and classifying the characteristic parameter vector set of the running state of the propeller by adopting a preset algorithm, removing the characteristic parameter vector of the abnormal working condition, and obtaining the characteristic parameter vector set of the propeller under the normal working condition.
In an exemplary embodiment, the performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrices according to a predetermined relevance rule includes:
calculating the correlation of each two groups of characteristic parameter vectors in the sample data according to a Person correlation algorithm;
obtaining a reconstructed sample matrix according to the relevance rule;
the Person correlation algorithm comprises:
Figure BDA0002362703780000061
wherein: m represents the number of characteristic parameters; xiAnd YiRespectively representing the ith parameter quantity in the characteristic parameter vectors X and Y,
Figure BDA0002362703780000062
and
Figure BDA0002362703780000063
respectively, the parameter means of the characteristic parameter vectors X and Y, and r represents the correlation between the characteristic parameter vectors X, Y.
In an exemplary embodiment, the association rule includes:
and deleting the sample data corresponding to the strong correlation quantity in the sample data correlation table.
In an exemplary embodiment, the selecting a sample matrix corresponding to a maximum value of a spatial indicator of an operating condition in a reconstructed sample matrix as a process memory matrix includes:
calculating the average distance from each sample in each group of reconstructed sample matrixes to the adjacent sample of the sample;
summing the average distances between all samples and adjacent samples to obtain an average value, wherein the average value represents a working condition space index;
the reconstructed sample matrix with the largest average value is selected as the process memory matrix.
In an exemplary embodiment, the obtaining a plurality of eigenvectors of the propeller, and determining an estimated vector corresponding to each of the eigenvectors according to the process memory matrix includes:
obtaining a plurality of characteristic parameter vectors of a propeller, and determining similarity measure vectors between the plurality of characteristic vectors and a process memory matrix by using a Mahalanobis distance algorithm;
and multiplying the process memory matrix and the similarity measure vector to obtain an estimation vector corresponding to each characteristic parameter.
In an exemplary embodiment, the calculating a distance similarity between each feature parameter vector and the corresponding estimation vector includes:
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector by adopting a Markov distance algorithm;
wherein, the horse-type distance algorithm comprises:
Figure BDA0002362703780000071
in the formula: xmon，XestRespectively representing the characteristic parameter vector and the corresponding estimation vector; s is a covariance matrix,
Figure BDA0002362703780000072
m is the number of feature variables of the feature parameters and the corresponding estimation vectors,
Figure BDA0002362703780000073
and
Figure BDA0002362703780000074
respectively representing a characteristic parameter vector XmonAnd estimate vector XestIs measured.
In an exemplary embodiment, the determining an early warning threshold for early warning of a propeller fault according to the distance similarity includes:
calculating the average value of the N distance similarities;
and calculating the average value by adopting a 3 sigma principle to obtain an early warning threshold value of the propeller fault early warning, wherein the 3 sigma principle formula comprises:
Figure BDA0002362703780000075
in the above formula, σ is an early warning threshold value of the early warning of the propeller fault, and N is the number of distance similarities;
Figure BDA0002362703780000076
the average value of the distance similarity is obtained; MDiIs distance similarity;
and normalizing the early warning threshold value.
Compared with the related art, the method comprises the following steps: acquiring a plurality of groups of characteristic parameter vector sets of an ocean platform propeller as a plurality of groups of sample data sets; performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrixes according to a preset relevance rule; selecting a sample matrix corresponding to the maximum value of the working condition space index in the reconstructed sample matrix as a process memory matrix; acquiring a plurality of characteristic parameter vectors of a propeller, and determining an estimation vector corresponding to each characteristic parameter vector according to the process memory matrix; calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector; determining an early warning threshold value of early warning of the faults of the propeller according to the distance similarity; acquiring a current characteristic parameter vector in real time, and judging whether the propeller fails according to the early warning threshold; when the distance similarity between the current characteristic parameter vector and the estimation vector corresponding to the current characteristic parameter vector is smaller than an early warning threshold value, determining that the propeller has a fault; and when the distance similarity between the current characteristic parameter vector and the estimated vector of the current characteristic parameter vector is greater than an early warning threshold value, determining that the propeller is normal. By the scheme of the invention, the characteristic parameters capable of describing the operation state of the propeller as much as possible are selected, the characteristic parameters of the operation state under the normal working condition are screened out to be used as sample data, a process memory matrix is constructed, and finally, a reasonable early warning threshold value is determined, so that the precision of the early warning of the faults of the propeller of the ocean platform is improved.
Additional features and advantages of the application will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the application. Other advantages of the present application can be realized and attained by the instrumentalities and combinations particularly pointed out in the specification and the drawings.
Drawings
The accompanying drawings are included to provide an understanding of the present disclosure and are incorporated in and constitute a part of this specification, illustrate embodiments of the disclosure and together with the examples serve to explain the principles of the disclosure and not to limit the disclosure.
FIG. 1 is a flow chart of a method for early warning of a failure of an offshore platform propeller according to an embodiment of the present application;
FIG. 2 is a table of sample data correlations according to an embodiment of the present application;
fig. 3 is a schematic diagram of a fault early warning device for an offshore platform propeller according to an embodiment of the present application.
Detailed Description
The present application describes embodiments, but the description is illustrative rather than limiting and it will be apparent to those of ordinary skill in the art that many more embodiments and implementations are possible within the scope of the embodiments described herein. Although many possible combinations of features are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are possible. Any feature or element of any embodiment may be used in combination with or instead of any other feature or element in any other embodiment, unless expressly limited otherwise.
The present application includes and contemplates combinations of features and elements known to those of ordinary skill in the art. The embodiments, features and elements disclosed in this application may also be combined with any conventional features or elements to form a unique inventive concept as defined by the claims. Any feature or element of any embodiment may also be combined with features or elements from other inventive aspects to form yet another unique inventive aspect, as defined by the claims. Thus, it should be understood that any of the features shown and/or discussed in this application may be implemented alone or in any suitable combination. Accordingly, the embodiments are not limited except as by the appended claims and their equivalents. Furthermore, various modifications and changes may be made within the scope of the appended claims.
Further, in describing representative embodiments, the specification may have presented the method and/or process as a particular sequence of steps. However, to the extent that the method or process does not rely on the particular order of steps set forth herein, the method or process should not be limited to the particular sequence of steps described. Other orders of steps are possible as will be understood by those of ordinary skill in the art. Therefore, the particular order of the steps set forth in the specification should not be construed as limitations on the claims. Further, the claims directed to the method and/or process should not be limited to the performance of their steps in the order written, and one skilled in the art can readily appreciate that the sequences may be varied and still remain within the spirit and scope of the embodiments of the present application.
Example one
Fig. 1 is a flowchart of a method for early warning a failure of an offshore platform propeller according to an embodiment of the present invention.
And 100, acquiring a plurality of groups of characteristic parameter vector sets of the marine platform propeller as a plurality of groups of sample data sets.
In this embodiment, each sensor may be arranged by using a sensor optimal arrangement scheme of the propeller, and the sensors arranged by using the optimal arrangement scheme collect characteristic parameter combinations reflecting real-time operating states of the propeller on line as characteristic parameter vectors, where the characteristic parameter vectors may include characteristic parameters of m operating states such as bearing vibration speed, acceleration signals, temperature values of components such as a propeller bearing and a motor, and parameters of related lubricating oil and oil; and forming a characteristic parameter vector set by the collected multiple groups of characteristic parameter vectors, and taking the multiple groups of characteristic parameter vector sets as sample data sets.
The collected characteristic parameters can be denoised by a plurality of groups of characteristic parameter vector sets by adopting wavelet packet decomposition and other denoising and signal enhancement processing technologies, and background noise in the characteristic parameters is eliminated.
In an exemplary embodiment, before obtaining a plurality of sets of feature parameter vectors of a propeller of an ocean platform as a plurality of sets of sample data sets, the method further includes: determining an arrangement scheme of the propeller sensors by using a preset algorithm; the propeller sensors are arranged by adopting the arrangement scheme, and the characteristic parameters of the running state of the propeller are collected in real time. In this embodiment, the preset algorithm may include: an effective independent method, a sequence method, a random algorithm and the like; and performing optimization processing calculation on the sensor position of the propeller by adopting any one of the preset algorithms to determine an optimized propeller sensor optimization arrangement scheme. The optimized propeller sensor optimal arrangement scheme is used for constructing an ocean platform propeller sensor arrangement scheme which can reflect spatial structure information to the maximum extent and is sensitive enough to state change of the spatial structure, and serves for constructing a subsequent normal sample space.
In an exemplary embodiment, after a plurality of groups of characteristic parameter vector sets of a propeller of an ocean platform are obtained as a plurality of groups of sample data sets, the characteristic parameter vector sets of the running state of the propeller are classified by adopting a preset algorithm, and characteristic parameter vectors of abnormal working conditions are removed to obtain a propeller characteristic parameter vector set of normal working conditions. In this embodiment, after a plurality of sets of feature parameter vector sets of a plurality of propellers in a long time are collected as a plurality of sets of sample data sets, the sample data sets may be classified by using, but not limited to, a clustering algorithm, a gaussian distribution algorithm, and the like, so that abnormal working condition data possibly existing in the sample data sets is removed, and the purity of a normal sample working condition space is ensured. The preset algorithm includes, but is not limited to, a clustering algorithm, a gaussian distribution algorithm, and the like. The preset algorithm is adopted to classify the sample data set, so that the state of the equipment can be judged according to the characteristic parameter values, abnormal data are removed, and only relevant data of the characteristic parameters of the equipment in a normal operation state exist in the sample data set.
And 101, performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrixes according to a preset relevance rule.
In this embodiment, relevance division may be performed on each group of sample data by using a Person correlation coefficient, and a reconstructed sample matrix of each group is obtained according to a predetermined relevance rule, so as to construct a normal sample operating condition space index.
In an exemplary embodiment, the performing relevance division on each group of sample data by using a correlation algorithm, and obtaining a reconstructed sample matrix of each group according to a predetermined relevance rule includes: calculating the correlation of each two groups of characteristic parameter vectors in the sample data according to a Person correlation algorithm; obtaining a reconstructed sample matrix according to the relevance rule; wherein the Person correlation algorithm comprises:
Figure BDA0002362703780000111
wherein: m represents the number of characteristic parameters; xiAnd YiRespectively representing the ith parameter in the characteristic parameter vector X and Y,
Figure BDA0002362703780000112
and
Figure BDA0002362703780000113
respectively representing the parameter means of the feature parameter vectors X and Y,
Figure BDA0002362703780000114
r denotes the correlation between the characteristic parameter vectors X, Y.
In an exemplary embodiment, the association rule includes: and deleting the sample data corresponding to the strong correlation quantity in the sample data correlation table.
In this embodiment, the relevance division is performed on each group of sample data sets by using a correlation algorithm, and the implementation process of obtaining each group of reconstructed sample matrices according to a predetermined relevance rule may include:
the method comprises the following steps of firstly, taking n groups of acquired characteristic parameter vectors of the running state as sample data, and defining the sample data as T, wherein T is an m multiplied by n matrix, and m is the number of characteristic parameters.
And secondly, calculating the correlation between two random groups of characteristic parameter vectors in the sample matrix T by using the Person correlation coefficient.
Assuming that the two sets of state parameter vectors are X and Y, respectively, the correlation is calculated as:
Figure BDA0002362703780000115
in the formula: m represents the number of characteristic parameters; xiAnd YiRespectively representing the ith parameter quantity in the characteristic parameter vectors X and Y;
Figure BDA0002362703780000116
and
Figure BDA0002362703780000117
respectively, representing the parameter means of the feature parameter vectors X and Y, wherein,
Figure BDA0002362703780000118
r denotes the correlation between the characteristic parameter vectors X, Y.
Calculating the correlation of every two groups of characteristic parameter vectors in the sample data T according to the formula to form a (N-1) x (N-1) correlation matrix:
Figure BDA0002362703780000121
1.2 in the formula: n is a radical ofi-NjRepresenting the correlation between the characteristic parameter vectors of the ith and jth lines in the sample data T.
Thirdly, removing the strongly correlated state parameter vectors in the sample data T, and forming a sample data correlation table according to the correlation matrix, as shown in fig. 2, where: 1) n is a radical ofiRepresenting the ith line in the sample data; 2) n is a radical ofi-NjRepresenting the correlation between the characteristic parameter vectors of the ith and jth lines in the sample data T. Can be based on the space structure of the working condition of the normal sampleThe actual requirements are defined as strong correlation, ± (0.8-1.0), medium correlation, ± (0.5-0.8), weak correlation, ± (0.2-0.5), irrelevant, and the sample data correlation table is further processed, wherein the processing process comprises the following steps:
(1) the diagonal is not considered;
(2) deleting rows and columns corresponding to the strong correlation quantity in the sample data correlation table by referring to a correlation rule, wherein the correlation rule refers to that the definition is +/-0.8-1.0 and is strong correlation, +/-0.5-0.8 and is medium correlation, +/-0.2-0.5 and is weak correlation, and +/-0-0.2 and is irrelevant;
(3) and respectively counting rows and columns corresponding to the medium correlation quantity, the weak correlation quantity and the uncorrelated quantity, wherein the statistics cannot be repeated. And obtaining a reconstructed sample matrix of each group through the processing.
And 102, selecting a sample matrix corresponding to the maximum value of the working condition space index in the reconstructed sample matrix as a process memory matrix.
In this embodiment, the reconstructed sample matrix, i.e., the sample matrix from which the strong correlation quantity is removed, includes the medium correlation quantity, the weak correlation quantity, and the uncorrelated quantity. And selecting a sample matrix corresponding to the maximum value of the working condition space index in the plurality of reconstructed sample matrices as a process memory matrix.
In an exemplary embodiment, selecting a sample matrix corresponding to a maximum value of a condition space indicator in a reconstructed sample matrix as a process memory matrix includes: calculating the average distance from each sample in each group of reconstructed sample matrixes to the adjacent sample of the sample; summing the average distances between all samples and adjacent samples to obtain an average value, wherein the average value represents a working condition space index; the reconstructed sample matrix with the largest average value is selected as the process memory matrix.
In the present embodiment, it is assumed that the reconstructed sample matrix is defined as TRefThe size of the sample matrix is defined as m × h, h is the number of reconstructed samples, and m represents the number of characteristic parameters of the sample matrix. Obtaining each group of reconstructed sample matrixes according to the step 101 and a preset relevance rule, and calculating any sample matrix in the reconstructed sample matrixesThe average distance of a sample to its neighboring samples, if a sample has no intermediate correlation samples, the average distance of the sample to its weak correlation volume is calculated, and so on. The following formula can be used to calculate the average distance of any one sample to its neighboring samples:
Figure BDA0002362703780000131
in formula 1.3:
Figure BDA0002362703780000132
representing a reconstructed sample matrix TRefThe average distance of a sample to its neighboring samples; k represents the number of peripheral samples adjacent to the sample; diRepresenting the euclidean distance between the sample and some neighboring sample, can be calculated by equation (1.4).
Figure BDA0002362703780000133
Calculating a reconstructed sample matrix TRefAverage distance of all samples from their neighbors is added and averaged:
Figure BDA0002362703780000134
in the formula: h is the number of samples after reconstruction.
Figure BDA0002362703780000135
The larger the value of the working condition space index is, the larger the space range for representing the distribution of the samples is, the more comprehensive the covered working condition types are, and the optimal sample data is screened according to the index.
And 103, acquiring a plurality of characteristic parameter vectors of the propeller, and determining an estimation vector corresponding to each characteristic parameter vector according to the process memory matrix.
In this embodiment, a plurality of feature parameter vectors of the propeller are obtained, and an estimation vector corresponding to each feature parameter vector is determined according to the process memory matrix determined in step 102. The plurality of characteristic parameter vectors of the propeller can be acquired from historical data or from current monitoring acquired in real time.
In one exemplary embodiment, obtaining a plurality of eigenvectors of a propeller, determining an estimated vector corresponding to each eigenvector from a process memory matrix, comprises: obtaining a plurality of characteristic parameter vectors of a propeller, and determining similarity measure vectors between the plurality of characteristic vectors and a process memory matrix by using a Mahalanobis distance algorithm; and multiplying the process memory matrix by the similarity measure vector to obtain an estimation vector corresponding to each characteristic parameter vector.
In an exemplary embodiment, calculating a distance similarity between each feature parameter vector and the corresponding estimation vector includes: calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector by adopting a Markov distance algorithm; wherein, the horse-type distance algorithm comprises:
Figure BDA0002362703780000141
in formula 1.6: xmon，XestRespectively representing the characteristic parameter vector and the corresponding estimation vector; s is a covariance matrix,
Figure BDA0002362703780000142
m is the number of feature variables of the feature parameters and the corresponding estimation vectors,
Figure BDA0002362703780000143
and
Figure BDA0002362703780000144
respectively representing a characteristic parameter vector XmonAnd estimate vector XestThe average value of (a) of (b),
Figure BDA0002362703780000145
in this embodiment, the reconstructed sample matrix with the largest average value is selected as the process memory matrix, which is defined as D,
a process memory matrix D:
Figure BDA0002362703780000146
defining the characteristic parameter vector currently monitored by the propeller or the historical characteristic parameter vector as XmonAnd XmonThe corresponding estimated vector is defined as Xest(ii) a Calculation and XmonThe corresponding estimated vector can be obtained by calculating a process memory matrix D:
Xest＝D·W＝D·[w1,w2,…wh]T
＝w1X(t1)+w2X(t2)+…whX(th) (1.8)
in formula 1.8: w represents a similarity measure between the current monitored characteristic parameter vector of the propeller or the historical characteristic parameter vector and the state in the process memory matrix, and the similarity measure can be calculated by the following formula:
W＝(DT·D)-1·(DT·Xmon) (1.9)
to avoid DTD irreversible case, applicable nonlinear operator
Figure BDA0002362703780000155
Instead of the matrix multiplication operator, it can be expressed as:
Figure BDA0002362703780000151
the nonlinear operator in the formula 1.10 can select mahalanobis distance to operate, wherein a mahalanobis distance algorithm is used for determining similarity measure vectors between the plurality of feature vectors and the process memory matrix;
Figure BDA0002362703780000152
in the formula: m is the number of characteristic variables; h is the sample volume; xi＝[x1(ti),x2(ti)…xm(ti)]Wherein x isj(ti) (j ═ 1,2,3 … m) is the difference between the eigenvector and the corresponding characteristic parameter of any sample of the process memory matrix;
Figure BDA0002362703780000153
after calculating to obtain a similarity measure W between the characteristic parameter vector currently monitored by the propeller or the historical characteristic parameter vector and the state in the process memory matrix, calculating an estimation vector corresponding to the characteristic parameter vector currently monitored by the propeller or the historical characteristic parameter vector:
Figure BDA0002362703780000154
and 104, calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector.
In this embodiment, a distance similarity between each feature parameter vector and the corresponding estimation vector may be calculated by using a makerian distance calculation method.
In an exemplary embodiment, calculating the distance similarity between each feature parameter vector and the corresponding estimation vector comprises: calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector by adopting a Markov distance algorithm; wherein, the horse-type distance algorithm comprises:
Figure BDA0002362703780000161
in formula 1.6, Xmon，XestRespectively the characteristic parameter vector and the corresponding estimation vector; s is a covariance matrix,
Figure BDA0002362703780000162
m is the number of feature variables of the feature parameters and the corresponding estimation vectors,
Figure BDA0002362703780000163
and
Figure BDA0002362703780000164
respectively representing a characteristic parameter vector XmonAnd estimate vector XestThe average value of (a) of (b),
Figure BDA0002362703780000165
and 105, determining an early warning threshold value of the early warning of the faults of the propeller according to the distance similarity.
In this embodiment, an early warning threshold for early warning of a propeller fault is determined according to the distance similarity determined in step 104.
In an exemplary embodiment, determining an early warning threshold for early warning of a propeller fault according to the distance similarity includes: calculating the average value of the N distance similarities; and calculating the average value by adopting a 3 sigma principle to obtain an early warning threshold value of the propeller fault early warning, wherein the 3 sigma principle formula comprises:
Figure BDA0002362703780000166
in the above formula, σ is an early warning threshold value of the early warning of the propeller fault, and N is the number of distance similarities;
Figure BDA0002362703780000167
the average value of the distance similarity is obtained; MDiIs distance similarity; and normalizing the early warning threshold value.
In this embodiment, propeller operating data in a certain period of time may be selected as input, an estimation vector corresponding to each of the characteristic parameters is determined through the process memory matrix, and after the estimation vector is determined, N distance similarity sequences are determined as follows:
MD(Xmon,Xest)＝[MD1,MD2,…MDN] (1.14)
calculating the average value of the N continuous distance similarities:
Figure BDA0002362703780000168
in order to eliminate errors caused by interference such as estimation, measurement or calculation, a 3 sigma principle is introduced when an early warning threshold value is set:
Figure BDA0002362703780000169
after introducing the 3 σ principle, the early warning threshold is calculated and specified to [0,1 ]:
Figure BDA0002362703780000171
in the formula: p is an early warning threshold value, k is an early warning threshold value coefficient, and is determined according to field operation experience and generally not more than 1.
And 106, acquiring the current characteristic parameter vector in real time, and judging whether the propeller fails according to the early warning threshold value.
In this embodiment, the current characteristic parameter vector of the propeller is monitored in real time, and whether the propeller fails or not is judged according to the early warning threshold. When the distance similarity between the current characteristic parameter vector and the estimation vector corresponding to the current characteristic parameter vector is smaller than an early warning threshold value, determining that the propeller has a fault; and when the distance similarity between the current characteristic parameter vector and the estimation vector of the current characteristic parameter vector is greater than an early warning threshold value, determining that the propeller is normal.
Example two
In order to solve the above problem, as shown in fig. 3, the present invention further provides a device for early warning of a failure of an offshore platform propeller, the device comprising: a memory and a processor; the method is characterized in that:
the memory is used for storing a program for early warning of faults of the marine platform propeller;
the processor is used for reading and executing the program for early warning the failure of the marine platform propeller and executing the following operations:
acquiring a plurality of groups of characteristic parameter vector sets of an ocean platform propeller as a plurality of groups of sample data sets;
performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrixes according to a preset relevance rule;
selecting a sample matrix corresponding to the maximum value of the working condition space index in the reconstructed sample matrix as a process memory matrix;
acquiring a plurality of characteristic parameter vectors of a propeller, and determining an estimation vector corresponding to each characteristic parameter vector according to the process memory matrix;
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector;
determining an early warning threshold value of early warning of the faults of the propeller according to the distance similarity;
acquiring a current characteristic parameter vector in real time, and judging whether the propeller fails according to the early warning threshold;
when the distance similarity between the current characteristic parameter vector and the estimation vector corresponding to the current characteristic parameter vector is smaller than an early warning threshold value, determining that the propeller has a fault;
and when the distance similarity between the current characteristic parameter vector and the estimation vector of the current characteristic parameter vector is greater than an early warning threshold value, determining that the propeller is normal.
In an exemplary embodiment, before the obtaining the sets of vectors of the characteristic parameters of the plurality of sets of marine platform propellers as the sets of sample data, the processor further performs the following operations:
determining an arrangement scheme of the propeller sensors by using a preset algorithm;
the propeller sensor is arranged by adopting the arrangement scheme, and the characteristic parameters of the running state of the propeller are collected in real time.
In an exemplary embodiment, after obtaining the sets of vectors of multiple sets of feature parameters of the marine platform propeller as multiple sets of sample data sets, the processor further performs the following operations:
and classifying the characteristic parameter vector set of the running state of the propeller by adopting a preset algorithm, removing the characteristic parameter vector of the abnormal working condition, and obtaining the characteristic parameter vector set of the propeller under the normal working condition.
In an exemplary embodiment, the performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrices according to a predetermined relevance rule includes:
calculating the correlation of each two groups of characteristic parameter vectors in the sample data according to a Person correlation algorithm;
obtaining a reconstructed sample matrix according to the relevance rule;
the Person correlation algorithm comprises the following steps:
Figure BDA0002362703780000181
wherein: m represents the number of characteristic parameters; xiAnd YiRespectively representing the ith parameter in the characteristic parameter vector X and Y,
Figure BDA0002362703780000182
and
Figure BDA0002362703780000183
respectively, the parameter means of the characteristic parameter vectors X and Y, and r represents the correlation between the characteristic parameter vectors X, Y.
In an exemplary embodiment, the association rule includes:
and deleting the sample data corresponding to the strong correlation quantity in the sample data correlation table.
In an exemplary embodiment, the selecting a sample matrix corresponding to a maximum value of a spatial indicator of an operating condition in a reconstructed sample matrix as a process memory matrix includes:
calculating the average distance from each sample in each group of reconstructed sample matrixes to the adjacent sample of the sample;
summing the average distances between all samples and adjacent samples to obtain an average value, wherein the average value represents a working condition space index;
the reconstructed sample matrix with the largest average value is selected as the process memory matrix.
In an exemplary embodiment, the obtaining a plurality of eigenvectors of the propeller, and determining an estimated vector corresponding to each of the eigenvectors according to the process memory matrix includes:
obtaining a plurality of characteristic parameter vectors of a propeller, and determining similarity measure vectors between the plurality of characteristic vectors and a process memory matrix by using a Mahalanobis distance algorithm;
and multiplying the process memory matrix and the similarity measure vector to obtain an estimation vector corresponding to each characteristic parameter vector.
In an exemplary embodiment, the calculating a distance similarity between each feature parameter vector and the corresponding estimation vector includes:
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector by adopting a Markov distance algorithm;
wherein, the horse-type distance algorithm comprises:
Figure BDA0002362703780000191
in the formula: xmon，XestRespectively representing the characteristic parameter vector and the corresponding estimation vector; s is a covariance matrix,
Figure BDA0002362703780000192
m is the number of feature variables of the feature parameters and the corresponding estimation vectors,
Figure BDA0002362703780000193
and
Figure BDA0002362703780000194
respectively representing a characteristic parameter vector XmonAnd estimate vector XestIs measured.
In an exemplary embodiment, the determining an early warning threshold for early warning of a propeller fault according to the distance similarity includes:
calculating the average value of the N distance similarities;
and calculating the average value by adopting a 3 sigma principle to obtain an early warning threshold value of the propeller fault early warning, wherein the 3 sigma principle formula comprises:
Figure BDA0002362703780000201
in the above formula, σ is an early warning threshold value of the early warning of the propeller fault, and N is the number of distance similarities;
Figure BDA0002362703780000202
the average value of the distance similarity is obtained; MDiIs distance similarity;
and normalizing the early warning threshold value.
An exemplary embodiment
Step 1, determining the arrangement scheme of the propeller sensors by using a preset algorithm.
In this embodiment, the preset algorithm may include: effective independence, sequence, random algorithm, etc.; and performing optimization processing calculation on the sensor position of the propeller by adopting any one of the preset algorithms to determine an optimized propeller sensor optimization arrangement scheme. The optimized propeller sensor optimal arrangement scheme is used for constructing an ocean platform propeller sensor arrangement scheme which can reflect spatial structure information to the maximum extent and is sensitive enough to state change of the spatial structure, and serves for constructing a subsequent normal sample space.
And 2, arranging propeller sensors by adopting the arrangement scheme, and acquiring a plurality of groups of characteristic parameter vector sets of the marine platform propeller as a plurality of groups of sample data sets.
In this embodiment, the characteristic parameter vector may include characteristic parameters of m operating states, such as bearing vibration speed, acceleration signal, temperature values of components such as propeller bearing and motor, and parameter values of related lubricating oil; and taking the collected multiple groups of characteristic parameter vector sets as sample data sets.
The collected characteristic parameter vector set can be subjected to denoising processing by adopting denoising and signal enhancement processing technologies such as wavelet packet decomposition and the like, so that background noise in the characteristic parameters is eliminated.
And 3, classifying the characteristic parameter vector set of the running state of the propeller by adopting a preset algorithm, removing the characteristic parameter vector of the abnormal working condition, and obtaining the characteristic parameter vector set of the propeller under the normal working condition.
In this embodiment, after a plurality of sets of feature parameter vector sets of a plurality of propellers in a long time are collected as a plurality of sets of sample data sets, the sample data sets may be classified by using, but not limited to, a clustering algorithm, a gaussian distribution algorithm, and the like, so that abnormal working condition data possibly existing in the sample data sets is removed, and the purity of a normal sample working condition space is ensured. The preset algorithm includes, but is not limited to, a clustering algorithm, a gaussian distribution algorithm, and the like. The preset algorithm is adopted to classify the sample data set, so that the state of the equipment can be judged according to the characteristic parameter values, abnormal data are removed, and only the relevant data of the characteristic parameters of the equipment in the normal operation state exist in the sample data set.
And 4, carrying out relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrixes according to a preset relevance rule.
And 41, calculating the correlation of each two groups of characteristic parameter vectors in the sample data according to a Person correlation algorithm.
In this step, the Person correlation algorithm includes:
Figure BDA0002362703780000211
wherein: m represents the number of characteristic parameters; xiAnd YiRespectively representing the ith parameter in the characteristic parameter vector X and Y,
Figure BDA0002362703780000212
and
Figure BDA0002362703780000213
respectively, the parameter means of the characteristic parameter vectors X and Y, and r represents the correlation between the characteristic parameter vectors X, Y.
And 42, obtaining a reconstructed sample matrix according to the relevance rule.
The association rule includes: and deleting the sample data corresponding to the strong correlation quantity in the sample data correlation table.
In this embodiment, the relevance division is performed on each group of sample data sets by using a correlation algorithm, and the implementation process of obtaining each group of reconstructed sample matrices according to a predetermined relevance rule may include:
the method comprises the following steps of firstly, taking n groups of acquired characteristic parameter vectors of the running state as sample data, and defining the sample data as T, wherein T is an m multiplied by n matrix, and m is the number of characteristic parameters.
And secondly, calculating the correlation between two random groups of characteristic parameter vectors in the sample matrix T by using the Person correlation coefficient.
Assuming that the two sets of state parameter vectors are X and Y, respectively, the correlation is calculated as:
Figure BDA0002362703780000221
in the formula: m represents the number of state characterization parameters; xiAnd YiRespectively representing the ith parameter quantity in the characteristic parameter vectors X and Y;
Figure BDA0002362703780000222
and
Figure BDA0002362703780000223
respectively, representing the parameter means of the feature parameter vectors X and Y, wherein,
Figure BDA0002362703780000224
r denotes the correlation between the characteristic parameter vectors X, Y.
Calculating the correlation of every two groups of characteristic parameter vectors in the sample data T according to the formula to form a (N-1) x (N-1) correlation matrix:
Figure BDA0002362703780000225
1.2 in the formula: n is a radical ofi-NjRepresenting the correlation between the characteristic parameter vectors of the ith and jth lines in the sample data T.
And thirdly, removing the strongly correlated state parameter vectors in the sample data T, and forming a sample data correlation table according to the correlation matrix, as shown in FIG. 2. According to the actual requirement on the working condition space structure of the normal sample, defining that +/-0.8-1.0 is strong correlation, +/- (0.5-0.8) is medium correlation, +/- (0.2-0.5) is weak correlation and +/-0-0.2 is irrelevant, and further processing the sample data correlation table, wherein the processing process comprises the following steps:
(1) the diagonal is not considered;
(2) deleting rows and columns corresponding to the strong correlation quantity in the sample data correlation table by referring to a correlation rule, wherein the correlation rule refers to that the definition is +/-0.8-1.0 and is strong correlation, +/-0.5-0.8 and is medium correlation, +/-0.2-0.5 and is weak correlation, and +/-0-0.2 and is irrelevant;
(3) and respectively counting rows and columns corresponding to the medium correlation quantity, the weak correlation quantity and the uncorrelated quantity, wherein the statistics cannot be repeated. And obtaining the reconstructed sample matrix of each group through the processing procedure.
And 5, selecting a sample matrix corresponding to the maximum value of the working condition space index in the reconstructed sample matrix as a process memory matrix.
Step 51, calculating the average distance from each sample in each group of reconstructed sample matrix to the adjacent sample of the sample;
step 52, summing the average distances between all samples and adjacent samples to obtain an average value, wherein the average value represents a working condition space index;
and 53, selecting the reconstructed sample matrix with the maximum average value as a process memory matrix.
In this embodiment, the reconstructed sample matrix, i.e., the sample matrix from which the strong correlation quantity is removed, includes the medium correlation quantity, the weak correlation quantity, and the uncorrelated quantity. And selecting a sample matrix corresponding to the maximum value of the working condition space index in the plurality of reconstructed sample matrices as a process memory matrix.
In the present embodiment, it is assumed that the reconstructed sample matrix is defined as TRefThe size of the sample matrix is defined as m × h, h is the number of reconstructed samples, and m represents the number of characteristic parameters of the sample matrix. And obtaining each group of reconstructed sample matrixes according to a preset relevance rule, calculating the average distance from any sample in the reconstructed sample matrixes to the adjacent samples, calculating the average distance from a sample to the weak correlation quantity if the sample has no medium correlation sample, and the like. The following formula can be used to calculate the average distance of any one sample to its neighboring samples:
Figure BDA0002362703780000231
in formula 1.3:
Figure BDA0002362703780000232
representing a reconstructed sample matrix TRefThe average distance of a sample to its neighboring samples; k represents the number of peripheral samples adjacent to the sample; diIndicating that the sample is in a certain neighborhood with respect to itThe euclidean distance between the near samples can be calculated by the formula (1.4).
Figure BDA0002362703780000233
Calculating a reconstructed sample matrix TRefAverage distance of all samples from their neighbors is added and averaged:
Figure BDA0002362703780000234
in the formula: h is the number of samples after reconstruction.
Figure BDA0002362703780000241
The larger the value of the working condition space index is, the larger the space range for representing the distribution of the samples is, the more comprehensive the covered working condition types are, and the optimal sample data is screened according to the index.
And 6, acquiring a plurality of characteristic parameter vectors of the propeller, and determining an estimation vector corresponding to each characteristic parameter vector according to the process memory matrix.
Step 61, obtaining a plurality of characteristic parameter vectors of the propeller, and determining similarity measure vectors between the plurality of characteristic vectors and a process memory matrix by using a Mahalanobis distance algorithm;
and step 62, multiplying the process memory matrix and the similarity measure vector to obtain an estimation vector corresponding to each characteristic parameter vector.
In this embodiment, calculating the distance similarity between each feature parameter vector and the corresponding estimation vector includes: calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector by adopting a Markov distance algorithm; wherein, the horse-type distance algorithm comprises:
Figure BDA0002362703780000242
in formula 1.6: xmon，XestRespectively representing the characteristic parameter vector and the corresponding estimation vector; s is a covariance matrix,
Figure BDA0002362703780000243
m is the number of feature variables of the feature parameters and the corresponding estimation vectors,
Figure BDA0002362703780000244
and
Figure BDA0002362703780000245
respectively representing feature parameter vectors XmonAnd estimate vector XestThe average value of (a) of (b),
Figure BDA0002362703780000246
in this embodiment, the reconstructed sample matrix with the largest average value is selected as the process memory matrix, which is defined as D,
a process memory matrix D:
Figure BDA0002362703780000247
defining the characteristic parameter vector currently monitored by the propeller or the historical characteristic parameter vector as XmonAnd XmonThe corresponding estimated vector is defined as Xest(ii) a Calculation and XmonThe corresponding estimated vector can be obtained by calculating a process memory matrix D:
Xest＝D·W＝D·[w1,w2,…wh]T
＝w1X(t1)+w2X(t2)+…whX(th) (1.8)
in formula 1.8: w represents a similarity measure between the current monitored characteristic parameter vector of the propeller or the historical characteristic parameter vector and the state in the process memory matrix, and the similarity measure can be calculated by the following formula:
W＝(DT·D)-1·(DT·Xmon) (1.9)
to avoid DTD irreversible cases, applicable nonlinear operators
Figure BDA0002362703780000255
Instead of the matrix multiplication operator, it can be expressed as:
Figure BDA0002362703780000251
the nonlinear operator in the formula 1.10 can select mahalanobis distance to operate, wherein a mahalanobis distance algorithm is used for determining similarity measure vectors between the plurality of feature vectors and the process memory matrix;
Figure BDA0002362703780000252
in the formula: m is the number of characteristic variables; h is the sample capacity; xi＝[x1(ti),x2(ti)…xm(ti)]Wherein x isj(ti) (j ═ 1,2,3 … m) is the difference between the eigenvector and the corresponding characteristic parameter of any sample of the process memory matrix;
Figure BDA0002362703780000253
after calculating to obtain a similarity measure W between the characteristic parameter vector currently monitored by the propeller or the historical characteristic parameter vector and the state in the process memory matrix, calculating an estimation vector corresponding to the characteristic parameter vector currently monitored by the propeller or the historical characteristic parameter vector:
Figure BDA0002362703780000254
and 7, calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector, wherein the specific implementation process comprises the following steps:
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector by adopting a Markov distance algorithm;
wherein, the horse-type distance algorithm comprises:
Figure BDA0002362703780000261
in the formula: xmon，XestRespectively the characteristic parameter vector and the corresponding estimation vector; s is a covariance matrix,
Figure BDA0002362703780000262
m is the number of feature variables of the feature parameters and the corresponding estimation vectors,
Figure BDA0002362703780000263
and
Figure BDA0002362703780000264
respectively representing feature parameter vectors XmonAnd estimate vector XestIs measured.
And 8, determining an early warning threshold value of the early warning of the faults of the propeller according to the distance similarity.
Step 81, calculating the average value of the N distance similarities;
and 82, calculating the average value by adopting a 3 sigma principle to obtain an early warning threshold value of the propeller fault early warning, wherein the 3 sigma principle formula comprises:
Figure BDA0002362703780000265
in the above formula, σ is an early warning threshold value of the propeller fault early warning, and N is the number of distance similarities;
Figure BDA0002362703780000266
the average value of the distance similarity is obtained; MDiIs distance similarity;
and 83, normalizing the early warning threshold value.
In this embodiment, propeller operating data in a certain period of time may be selected as input, an estimation vector corresponding to each of the characteristic parameters is determined through the process memory matrix, and after the estimation vector is determined, N distance similarity sequences are determined as follows:
MD(Xmon,Xest)＝[MD1,MD2,…MDN] (1.14)
calculating the average value of the N continuous distance similarities:
Figure BDA0002362703780000267
in order to eliminate errors caused by interference such as estimation, measurement or calculation, a 3 sigma principle is introduced when an early warning threshold value is set:
Figure BDA0002362703780000268
after introducing the 3 σ principle, the early warning threshold is calculated and specified to [0,1 ]:
Figure BDA0002362703780000271
in the formula: p is an early warning threshold value, k is an early warning threshold value coefficient, and is determined according to field operation experience and generally not more than 1.
And 9, acquiring the current characteristic parameters of the propeller in real time, and judging whether the propeller is in fault or not according to the early warning threshold value.
When the distance similarity between the current characteristic parameter vector and the estimation vector corresponding to the current characteristic parameter vector is smaller than an early warning threshold value, determining that the propeller has a fault;
and when the distance similarity between the current characteristic parameter vector and the estimation vector of the current characteristic parameter vector is greater than an early warning threshold value, determining that the propeller is normal.
In the embodiment, a plurality of groups of characteristic parameter vector sets of the marine platform propeller are obtained as a plurality of groups of sample data sets; performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrixes according to a preset relevance rule; selecting a sample matrix corresponding to the maximum value of the working condition space index in the reconstructed sample matrix as a process memory matrix; acquiring a plurality of characteristic parameter vectors of a propeller, and determining an estimation vector corresponding to each characteristic parameter vector according to a process memory matrix; calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector; determining an early warning threshold value of the early warning of the faults of the propeller according to the distance similarity; and acquiring the current characteristic parameters of the propeller in real time, and judging the fault state of the propeller according to the early warning threshold value. Therefore, this embodiment achieves the following technical effects:
1. the arrangement of the propeller sensors is optimized, so that the characteristic parameters of the collected operation state can represent the operation state of the propeller to the maximum extent, the optimized sensor arrangement can reflect the space structure information to the maximum extent and the propeller sensor arrangement scheme which is sensitive enough to the state change of the space structure, and the effectiveness of data collection is improved.
2. The content of abnormal working condition data in the sample data is reduced. The reduction of the data content of the abnormal working condition benefits from the following steps: historical operation data of the propeller are screened by using a clustering algorithm and a Gaussian distribution algorithm, and the purity of a normal sample working condition space is improved, so that a method model constructed on the basis is ensured to have good performance, and the accuracy of propeller fault early warning is improved.
3. The repetition rate of the working conditions in the normal working condition sample space is reduced. The reduction of the repetition rate of the working condition benefits from: and classifying the sample data correlation through the Person correlation coefficient, and removing strong correlation quantity, thereby reducing the data redundancy in the sample space, further constructing a normal sample working condition space index to judge the reasonable degree of the sample space, and ensuring that the sample space has large-range working condition coverage.
4. The construction mode of the process memory matrix D is optimized. Optimization of the process memory matrix D benefits from: the Mahalanobis distance which is not influenced by dimensions and can give consideration to the connection among various characteristics is adopted for operation, so that the calculation result is more practical.
5. The error caused by interference such as estimation, measurement or calculation and the like when the early warning threshold value is set is eliminated, and the elimination of the error benefits from the following steps: the 3 sigma principle is introduced to eliminate the error, so that the early warning threshold value is ensured to have a good early warning effect, and the aim of reducing the false warning rate is fulfilled.
It will be understood by those of ordinary skill in the art that all or some of the steps of the methods, systems, functional modules/units in the devices disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof. In a hardware implementation, the division between functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be performed by several physical components in cooperation. Some or all of the components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as is well known to those of ordinary skill in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, Digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can accessed by a computer. In addition, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media as known to those skilled in the art.

Claims (10)

1. A method for early warning of faults of a propeller of an ocean platform is characterized by comprising the following steps:
acquiring a plurality of groups of characteristic parameter vector sets of an ocean platform propeller as a plurality of groups of sample data sets;
performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrixes according to a preset relevance rule; wherein the association rule comprises: deleting sample data corresponding to the strong correlation quantity in the sample data correlation table;
selecting a sample matrix corresponding to the maximum value of the working condition space index in the reconstructed sample matrix as a process memory matrix, wherein the process memory matrix comprises the following steps: calculating the average distance from each sample in each group of reconstructed sample matrixes to the adjacent sample of the sample;
summing the average distances between all samples and adjacent samples to obtain an average value, wherein the average value represents a working condition space index;
selecting a reconstructed sample matrix with the largest average value as a process memory matrix;
acquiring a plurality of characteristic parameter vectors of a propeller, and determining an estimation vector corresponding to each characteristic parameter vector according to the process memory matrix;
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector;
determining an early warning threshold value of the early warning of the propeller fault according to the distance similarity, comprising: calculating the average value of the N distance similarities;
and calculating the average value by adopting a 3 sigma principle to obtain an early warning threshold value of the propeller fault early warning, wherein the 3 sigma principle formula comprises:
Figure FDA0003478351360000011
in the above formula, σ is the early warning threshold value of the early warning of the propeller fault, N is the number of distance similarities,
Figure FDA0003478351360000012
as an average of distance similarities, MDiIs distance similarity;
normalizing the early warning threshold;
acquiring the current characteristic parameters of the propeller in real time, and judging whether the propeller fails according to the early warning threshold;
when the distance similarity between the current characteristic parameter vector and the estimation vector corresponding to the current characteristic parameter vector is smaller than an early warning threshold value, determining that the propeller has a fault;
when the distance similarity between the current characteristic parameter vector and the estimation vector of the current characteristic parameter vector is greater than an early warning threshold value, determining that the propeller is normal;
the method for performing relevance division on each group of sample data by using a correlation algorithm and obtaining each group of reconstructed sample matrix according to a preset relevance rule comprises the following steps:
calculating the correlation of each two groups of characteristic parameter vectors in the sample data according to a Person correlation algorithm;
obtaining a reconstructed sample matrix according to the relevance rule;
the Person correlation algorithm comprises:
Figure FDA0003478351360000021
wherein: m represents the number of characteristic parameters; xiAnd YiRespectively representing the ith parameter in the characteristic parameter vector X and Y,
Figure FDA0003478351360000022
and
Figure FDA0003478351360000023
respectively, the parameter means of the characteristic parameter vectors X and Y, and r represents the correlation between the characteristic parameter vectors X, Y.
2. The method of claim 1, wherein before the obtaining the sets of vectors of the characteristic parameters of the plurality of groups of the marine platform propeller as the sets of sample data, the method further comprises:
determining an arrangement scheme of the propeller sensors by using a preset algorithm;
the propeller sensor is arranged by adopting the arrangement scheme, and the characteristic parameters of the running state of the propeller are collected in real time.
3. The method of claim 1, wherein after the obtaining of the sets of vectors of the characteristic parameters of the plurality of sets of the marine platform propeller as the sets of sample data, the method further comprises:
and classifying the characteristic parameter vector set of the running state of the propeller by adopting a preset algorithm, removing the characteristic parameter vector of the abnormal working condition, and obtaining the characteristic parameter vector set of the propeller under the normal working condition.
4. The method for early warning of failure of an offshore platform propeller according to claim 1, wherein the obtaining a plurality of eigenvector vectors of the propeller and determining an estimated vector corresponding to each of the eigenvector vectors according to the process memory matrix comprises:
obtaining a plurality of characteristic parameter vectors of a propeller, and determining similarity measurement vectors between the plurality of characteristic vectors and a process memory matrix by using a Mahalanobis distance algorithm;
and multiplying the process memory matrix and the similarity measure vector to obtain an estimation vector corresponding to each characteristic parameter vector.
5. The method of claim 4, wherein the calculating the distance similarity between each feature parameter vector and the corresponding estimation vector comprises:
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector by adopting a Markov distance algorithm;
wherein, the horse-type distance algorithm comprises:
Figure FDA0003478351360000031
in the formula: xmon，XestRespectively representing the characteristic parameter vector and the corresponding estimation vector; s is a covariance matrix,
Figure FDA0003478351360000032
m is the number of feature variables of the feature parameters and the corresponding estimation vectors,
Figure FDA0003478351360000033
and
Figure FDA0003478351360000034
respectively representing a characteristic parameter vector XmonAnd estimate vector XestIs measured.
6. An ocean platform propeller fault early warning device, the device comprising: a memory and a processor; the method is characterized in that:
the memory is used for storing a program for early warning of faults of the marine platform propeller;
the processor is used for reading and executing the program for early warning the failure of the marine platform propeller and executing the following operations:
acquiring a plurality of groups of characteristic parameter vector sets of an ocean platform propeller as a plurality of groups of sample data sets;
performing relevance division on each group of sample data sets by using a correlation algorithm, and obtaining each group of reconstructed sample matrixes according to a preset relevance rule, wherein the relevance rule comprises the following steps: deleting sample data corresponding to the strong correlation quantity in the sample data correlation table;
selecting a sample matrix corresponding to the maximum value of the working condition space index in the reconstructed sample matrix as a process memory matrix, wherein the process memory matrix comprises the following steps:
calculating the average distance from each sample in each group of reconstructed sample matrixes to the adjacent sample of the sample;
summing the average distances between all samples and adjacent samples to obtain an average value, wherein the average value represents a working condition space index;
selecting a reconstructed sample matrix with the maximum average value as a process memory matrix;
acquiring a plurality of characteristic parameter vectors of a propeller, and determining an estimation vector corresponding to each characteristic parameter vector according to the process memory matrix;
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector;
determining an early warning threshold value of the early warning of the propeller fault according to the distance similarity, comprising: calculating the average value of the N distance similarities;
and calculating the average value by adopting a 3 sigma principle to obtain an early warning threshold value of the propeller fault early warning, wherein the 3 sigma principle formula comprises:
Figure FDA0003478351360000041
in the 3 sigma principle formula, sigma is an early warning threshold value of the early warning of the faults of the propeller, N is the number of distance similarities,
Figure FDA0003478351360000042
as an average of distance similarities, MDiIs distance similarity;
normalizing the early warning threshold;
acquiring the current characteristic parameters of the propeller in real time, and judging whether the propeller fails according to the early warning threshold;
when the distance similarity between the current characteristic parameter vector and the estimation vector corresponding to the current characteristic parameter vector is smaller than an early warning threshold value, determining that the propeller has a fault;
when the distance similarity between the current characteristic parameter vector and the estimation vector of the current characteristic parameter vector is greater than an early warning threshold value, determining that the propeller is normal;
the method for performing relevance division on each group of sample data by using a correlation algorithm and obtaining each group of reconstructed sample matrix according to a preset relevance rule comprises the following steps:
calculating the correlation of each two groups of characteristic parameter vectors in the sample data according to a Person correlation algorithm;
obtaining a reconstructed sample matrix according to the relevance rule;
the Person correlation algorithm comprises:
Figure FDA0003478351360000051
wherein: m represents the number of characteristic parameters, XiAnd YiRespectively representing the ith parameter in the characteristic parameter vector X and Y,
Figure FDA0003478351360000052
and
Figure FDA0003478351360000053
respectively, the parameter means of the characteristic parameter vectors X and Y, and r represents the correlation between the characteristic parameter vectors X, Y.
7. The device of claim 6, wherein before the obtaining of the sets of vectors of characteristic parameters of the plurality of sets of marine platform propulsion units as the sets of sample data, the processor further performs the following operations:
determining an arrangement scheme of the propeller sensors by using a preset algorithm;
the propeller sensor is arranged by adopting the arrangement scheme, and the characteristic parameters of the running state of the propeller are collected in real time.
8. The device of claim 6, wherein after the obtaining of the sets of vectors of the characteristic parameters of the plurality of sets of the marine platform propeller as the sets of sample data, the processor further performs the following operations:
and classifying the characteristic parameter vector set of the running state of the propeller by adopting a preset algorithm, removing the characteristic parameter vector of the abnormal working condition, and obtaining the characteristic parameter vector set of the propeller under the normal working condition.
9. The device of claim 6, wherein the obtaining a plurality of characteristic parameter vectors of the thruster and determining an estimation vector corresponding to each of the characteristic parameter vectors according to the process memory matrix comprises:
obtaining a plurality of characteristic parameter vectors of a propeller, and determining similarity measure vectors between the plurality of characteristic vectors and a process memory matrix by using a Mahalanobis distance algorithm;
and multiplying the process memory matrix and the similarity measure vector to obtain an estimation vector corresponding to each characteristic parameter vector.
10. The device of claim 9, wherein the calculating the distance similarity between each of the characteristic parameter vectors and the corresponding estimation vector comprises:
calculating the distance similarity between each characteristic parameter vector and the corresponding estimation vector by adopting a Markov distance algorithm;
wherein, the horse-type distance algorithm comprises:
Figure FDA0003478351360000061
in the formula: xmon，XestRespectively representing the characteristic parameter vector and the corresponding estimation vector; s is a covariance matrix,
Figure FDA0003478351360000062
m is the number of feature variables of the feature parameters and the corresponding estimation vectors,
Figure FDA0003478351360000063
and
Figure FDA0003478351360000064
respectively representing a characteristic parameter vector XmonAnd estimate vector XestIs measured.
CN202010029817.7A 2020-01-10 2020-01-10 Fault early warning method and device for ocean platform propeller Active CN111260893B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010029817.7A CN111260893B (en) 2020-01-10 2020-01-10 Fault early warning method and device for ocean platform propeller

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010029817.7A CN111260893B (en) 2020-01-10 2020-01-10 Fault early warning method and device for ocean platform propeller

Publications (2)

Publication Number Publication Date
CN111260893A CN111260893A (en) 2020-06-09
CN111260893B true CN111260893B (en) 2022-05-03

Family

ID=70950457

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010029817.7A Active CN111260893B (en) 2020-01-10 2020-01-10 Fault early warning method and device for ocean platform propeller

Country Status (1)

Country Link
CN (1) CN111260893B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112113595B (en) * 2020-09-25 2021-03-30 西门子交通技术(北京)有限公司 Sensor fault detection method, device and computer readable medium
CN113687421B (en) * 2021-08-23 2022-10-21 中国石油大学(北京) Data processing method and device for seismic signals, electronic equipment and storage medium
CN114088388A (en) * 2021-12-10 2022-02-25 华润电力技术研究院有限公司 Fault diagnosis method and fault diagnosis device for gearbox
CN115027634A (en) * 2022-06-24 2022-09-09 中海油田服务股份有限公司 Power maintenance supply ship fault early warning method and system
CN116720150B (en) * 2023-08-09 2023-10-20 山东晋工科技有限公司 Mechanical refrigeration system fault diagnosis method and system
CN117347093B (en) * 2023-12-04 2024-02-23 张家港市胜港机械制造有限公司 Fault detection system based on drilling and production equipment measuring instrument

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2011090382A (en) * 2009-10-20 2011-05-06 Mitsubishi Heavy Ind Ltd Monitoring system
CN105114348A (en) * 2015-09-09 2015-12-02 长春合成兴业能源技术有限公司 Device and method for induced draft fan fault early warning based on power station operation data
CN105758633A (en) * 2016-02-26 2016-07-13 中国航空工业集团公司上海航空测控技术研究所 Method for evaluating health conditions of various components of gearbox
CN107168205A (en) * 2017-06-07 2017-09-15 南京航空航天大学 A kind of online health monitoring data collection and analysis method of civil aircraft air-conditioning system
CN108171271A (en) * 2018-01-11 2018-06-15 湖南大唐先科技有限公司 A kind of equipment deteriorates early warning method and system
CN109241669A (en) * 2018-10-08 2019-01-18 成都四方伟业软件股份有限公司 A kind of method for automatic modeling, device and its storage medium
CN109857079A (en) * 2018-12-05 2019-06-07 上海交通大学 The intelligent diagnosing method and device of machining center axis system working condition exception
CN110108486A (en) * 2018-01-31 2019-08-09 阿里巴巴集团控股有限公司 Bearing fault prediction technique, equipment and system
CN110275879A (en) * 2019-05-16 2019-09-24 浙江浙能技术研究院有限公司 A method of Trouble Match and early warning are carried out based on fault data state matrix

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6609036B1 (en) * 2000-06-09 2003-08-19 Randall L. Bickford Surveillance system and method having parameter estimation and operating mode partitioning
US6892163B1 (en) * 2002-03-08 2005-05-10 Intellectual Assets Llc Surveillance system and method having an adaptive sequential probability fault detection test
US7568628B2 (en) * 2005-03-11 2009-08-04 Hand Held Products, Inc. Bar code reading device with global electronic shutter control
JP4730451B2 (en) * 2009-03-16 2011-07-20 富士ゼロックス株式会社 Detection data processing apparatus and program

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2011090382A (en) * 2009-10-20 2011-05-06 Mitsubishi Heavy Ind Ltd Monitoring system
CN105114348A (en) * 2015-09-09 2015-12-02 长春合成兴业能源技术有限公司 Device and method for induced draft fan fault early warning based on power station operation data
CN105758633A (en) * 2016-02-26 2016-07-13 中国航空工业集团公司上海航空测控技术研究所 Method for evaluating health conditions of various components of gearbox
CN107168205A (en) * 2017-06-07 2017-09-15 南京航空航天大学 A kind of online health monitoring data collection and analysis method of civil aircraft air-conditioning system
CN108171271A (en) * 2018-01-11 2018-06-15 湖南大唐先科技有限公司 A kind of equipment deteriorates early warning method and system
CN110108486A (en) * 2018-01-31 2019-08-09 阿里巴巴集团控股有限公司 Bearing fault prediction technique, equipment and system
CN109241669A (en) * 2018-10-08 2019-01-18 成都四方伟业软件股份有限公司 A kind of method for automatic modeling, device and its storage medium
CN109857079A (en) * 2018-12-05 2019-06-07 上海交通大学 The intelligent diagnosing method and device of machining center axis system working condition exception
CN110275879A (en) * 2019-05-16 2019-09-24 浙江浙能技术研究院有限公司 A method of Trouble Match and early warning are carried out based on fault data state matrix

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
基于MSET模型的风力发电机故障预警;李大中,常永亮,赵杰,刘建屏;《华北电力技术》;20161225;43-48 *
基于MSET的电站风机故障预警技术研究;杨婷婷，张蓓,吕游，邸小慧;《热能动力工程》;20170920;63-69 *
基于多元状态估计和偏离度的电厂风机故障预警;刘涛,刘吉臻,吕游,崔超;《动力工程学报》;20160615;454-460 *
多元状态估计的记忆矩阵选取及风机故障预警方法;李锋,潘凤萍,廖宏楷,吕游,黄鑫;《自动化仪表》;20190120;74-78 *

Also Published As

Publication number Publication date
CN111260893A (en) 2020-06-09

Similar Documents

Publication Publication Date Title
CN111260893B (en) Fault early warning method and device for ocean platform propeller
US6957172B2 (en) Complex signal decomposition and modeling
US7567878B2 (en) Evaluating anomaly for one class classifiers in machine condition monitoring
CN112911627B (en) Wireless network performance detection method, device and storage medium
CN111538311B (en) Flexible multi-state self-adaptive early warning method and device for mechanical equipment based on data mining
Harkat et al. Machine learning-based reduced kernel PCA model for nonlinear chemical process monitoring
CN110807245B (en) Automatic modeling method and system for equipment fault early warning
CN111967535A (en) Fault diagnosis method and device for temperature sensor in grain storage management scene
CN112070154A (en) Time series data processing method and device
CN113579851A (en) Non-stationary drilling process monitoring method based on adaptive segmented PCA
CN110858072A (en) Method and device for determining running state of equipment
KR20200010671A (en) System and method for fault diagnosis of equipment based on machine learning
CN116401137B (en) Core particle health state prediction method and device, electronic equipment and storage medium
CN112486722A (en) System fault detection method and related device
CN116400244B (en) Abnormality detection method and device for energy storage battery
CN116908684A (en) Motor fault prediction method and device, electronic equipment and storage medium
CN117031294A (en) Battery multi-fault detection method, device and storage medium
CN113449809A (en) Cable insulation on-line monitoring method based on KPCA-NSVDD
CN115700553A (en) Anomaly detection method and related device
CN113283157A (en) System, method, terminal and medium for predicting life cycle of intelligent stamping press part
CN114112390A (en) Early fault diagnosis method for nonlinear complex system
CN116700213B (en) Industrial equipment abnormality detection method and related device based on gating circulation unit
CN113609207B (en) Data preprocessing method for slope deformation monitoring data
CN117668684B (en) Power grid electric energy data anomaly detection method based on big data analysis
KR102621106B1 (en) Apparatus and method for fault diagnosis of ship engines

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant