CN116957361A - Ship task system health state detection method based on virtual-real combination - Google Patents

Ship task system health state detection method based on virtual-real combination Download PDF

Info

Publication number
CN116957361A
CN116957361A CN202310948313.9A CN202310948313A CN116957361A CN 116957361 A CN116957361 A CN 116957361A CN 202310948313 A CN202310948313 A CN 202310948313A CN 116957361 A CN116957361 A CN 116957361A
Authority
CN
China
Prior art keywords
health
feature
real
data
fusion
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.)
Granted
Application number
CN202310948313.9A
Other languages
Chinese (zh)
Other versions
CN116957361B (en
Inventor
胡洋
蔡华杰
郭政业
龚盈盈
王剑波
徐文君
刘佳宜
杨继坤
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 Ship Development and Design Centre
Original Assignee
China Ship Development and Design Centre
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 Ship Development and Design Centre filed Critical China Ship Development and Design Centre
Priority to CN202310948313.9A priority Critical patent/CN116957361B/en
Publication of CN116957361A publication Critical patent/CN116957361A/en
Application granted granted Critical
Publication of CN116957361B publication Critical patent/CN116957361B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
    • 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
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Human Resources & Organizations (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • Mathematical Physics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Economics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Educational Administration (AREA)
  • Game Theory and Decision Science (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Medical Informatics (AREA)
  • Development Economics (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a ship task system health state detection method based on virtual-real combination, which relates to the technical field of system health detection and comprises the following steps: acquiring real-time data; virtual data of a digital twin system is obtained, and both real-time data and the virtual data are preprocessed; constructing a neural network frame, wherein the neural network frame comprises a feature extraction module and a feature fusion module; inputting the preprocessed real-time data and virtual data into a deep neural network frame, obtaining real-time features and virtual features through a feature extraction module, and fusing the real-time features and the virtual features by utilizing a feature fusion module to obtain first features; and constructing a health evaluation model, and inputting the first characteristic into the health evaluation model to obtain the health state description of the ship task system. The invention can find out faults in time and ensure the normal operation of the ship task system.

Description

Ship task system health state detection method based on virtual-real combination
Technical Field
The invention relates to the technical field of system health detection, in particular to a ship task system health state detection method based on virtual-real combination.
Background
Cabin state monitoring technology plays a vital role in the operation of ships today. The technology is based on historical operation data, and abnormal data is identified in real time by establishing a healthy operation model, so that the reliability of monitored equipment or a monitored system is improved, the failure rate is reduced, the aim of improving the overall safety level of the system is fulfilled, meanwhile, the maintenance-related cost of the monitored equipment or the monitored system is reduced to the greatest extent, and the economical efficiency is improved. The reliability of the equipment is guaranteed.
The state monitoring technology mainly comprises the steps of carrying out data fusion and data mining under different state environments, rapidly calculating all samples given by a research object, avoiding one-sided performance caused by random sampling, obtaining a certain model which is useful for actual engineering, and predicting the future by using the model. The fault diagnosis and prediction algorithm for implementing the state monitoring system is mainly based on two methods: model-based (first principles of physical model) methods or data-driven (using statistics and data mining) methods. Model-based state monitoring techniques focus on the model of the subject and require extensive investigation of detailed physical properties and system functions of the device or system and an empirical library. The method is suitable for feasibility verification of newly developed equipment with good computer models and a part of fault statistics. The data-driven state monitoring technology does not need to know the detailed physical structure and performance of the equipment or the system, but depends more on detailed statistics and measurement data related to the prior faults, and the greater the data quantity, the higher the data quality and the higher the accuracy degree of the model.
The state monitoring technology is an important technology for realizing the intellectualization and the digitalization of ships, as ship equipment gradually tends to be large-scale, multi-system and complex, a diagnosis method based on fault characteristics and fault samples is difficult to obtain practical results because comprehensive fault samples cannot be obtained, and the state of equipment in all working conditions cannot be well represented only by the operation data of target equipment.
The invention patent with the Chinese application number of 202010928713.X discloses a ship technical condition comprehensive evaluation system based on state monitoring, which solves the problem of insufficient precision of equipment characteristic parameters through fault classification, detects potential abnormal conditions through an overall performance measurement method, solves the problem of multiple fault modes, and can timely process equipment faults through supporting collaborative intelligent diagnosis. However, the method in the prior art is relatively simple, and accurate health state evaluation cannot be performed when the data amount is relatively large and complex.
Disclosure of Invention
In view of the above, the invention provides a ship task system health state detection method based on virtual-real combination, which performs deeper data monitoring on a ship task system by fusing real-time data and simulation data of a digital twin system at characteristic level, and performs data-driven health state detection on the ship task system by a set health evaluation function to obtain real-time and accurate health state detection results.
The technical scheme of the invention is realized as follows: the invention provides a ship task system health state detection method based on virtual-real combination, which comprises the following steps:
s1, acquiring original data of a ship task system, constructing a digital twin system according to the ship task system and the original data thereof, and correcting the digital twin system, wherein the original data comprises historical data and real-time data;
s2, virtual data of the digital twin system are obtained, and both the real-time data and the virtual data are preprocessed;
s3, constructing a neural network frame, wherein the neural network frame comprises a feature extraction module and a feature fusion module;
s4, inputting the preprocessed real-time data and virtual data into a deep neural network frame, obtaining real-time features and virtual features through a feature extraction module, and fusing the real-time features and the virtual features by utilizing a feature fusion module to obtain first features;
s5, building a health evaluation model, and inputting the first characteristic into the health evaluation model to obtain the health state description of the ship task system.
Further preferably, step S1 includes:
the ship task system comprises N sub-task systems, each sub-task system comprises a plurality of devices, each device comprises device real-time data and device historical data, all the device real-time data are used as real-time data of the ship task system, all the device historical data are used as historical data of the ship task system, and the real-time data and the historical data form original data;
obtaining elements of physical entities of a ship task system, wherein the elements comprise geometric parameters, physical attributes, operation modes, man-machine interaction, information flow and interface relation, adding assembly constraint relation and rule constraint relation, combining element calculation to generate a combined model, and setting motion constraint on a combined die to obtain a digital twin system;
and (3) performing simulation verification on the single sub-task system, inputting the original data of the sub-task system and the elements of the physical entity of the sub-task system into the digital twin system, and adjusting the parameters of the digital twin system according to the deviation value of the simulation result and the actual result until the deviation value is within a preset range, so as to obtain the corrected digital twin system.
Further preferably, step S4 includes:
s41, the feature extraction module comprises a multi-scale convolution network, real-time data and virtual data are sequentially input into the multi-scale convolution network to be extracted to obtain convolution features of a plurality of scales, the convolution features of the plurality of scales form a feature sequence, the feature sequence of the real-time data is used as a real-time feature, and the feature sequence of the virtual data is used as a virtual feature;
the S42 feature fusion module comprises a first fusion network and a second fusion network, the first fusion network is utilized to respectively fuse the convolution features of the same scale of the real-time features and the virtual features, the obtained same-scale convolution fusion feature sequences are input into the second fusion network, and the same-scale convolution feature sequences are sequentially fused in pairs according to the sequence from small scale to large scale, so that the first features are obtained.
Further preferably, step S41 includes:
the multi-scale convolution network comprises a 7*7 convolution layer and 4 convolution blocks;
respectively inputting the real-time data x and the virtual data y into a multi-scale convolution network, and extracting convolution characteristics x of a first scale according to a 7*7 convolution layer 1 And y 1 Then sequentially inputting 4 convolution blocks, and respectively extracting to obtain 4-scale convolution characteristics x 2 、x 3 、x 4 、x 5 And y 2 、y 3 、y 4 、y 5 The scale of the convolution features is sequentially increased according to the data flow direction;
the convolution characteristics of 5 scales of real-time data are formed into a characteristic sequence, and the real-time characteristic X= { X is obtained 1 ,x 2 ,x 3 ,x 4 ,x 5 The convolution characteristics of 5 scales of the virtual data are formed into a characteristic sequence, and the virtual characteristic Y= { Y is obtained 1 ,y 2 ,y 3 ,y 4 ,y 5 }。
Further preferably, step S42 includes:
the real-time feature x= { X 1 ,x 2 ,x 3 ,x 4 ,x 5 Sum virtual feature y= { Y 1 ,y 2 ,y 3 ,y 4 ,y 5 Input into the first converged network, respectively x 1 And y is 1 Fusion to obtain c 2 、x 2 And y is 2 Fusion to obtain c 2 、x 3 And y is 3 Fusion to obtain c 3 、x 4 And y is 4 Fusion to obtain c 4 、x 5 And y is 5 Fusion to obtain c 5 C, adding 1 、c 2 、c 3 、c 4 、c 5 Constitute the same-scale convolution fusion characteristic sequence C= { C 1 ,c 2 ,c 3 ,c 4 ,c 5 }；
Fusing the same-scale convolution feature sequence C= { C 1 ,c 2 ,c 3 ,c 4 ,c 5 Inputting into a second fusion network, and fusing the characteristic sequence C= { C for the same-scale convolution by using a screening function 1 ,c 2 ,c 3 ,c 4 ,c 5 Feature screening is carried out, redundant information is removed, and a same-scale convolution screening feature sequence C ' = { C ' is obtained ' 1 ,c' 2 ,c' 3 ,c' 4 ,c' 5 }；
The feature sequence C ' = { C ' is screened by co-scale convolution ' 1 ,c' 2 ,c' 3 ,c' 4 ,c' 5 Performing layer-by-layer fusion according to the order of the scale from small to large, and firstly fusing c' 5 And c' 4 Fusion to give c' 5,4 And c 'is carried out again' 5,4 And c' 3 Fusion to give c' 5,4,3 After which c' 5,4,3 And c' 2 Fusion to give c' 5,4,3,2 Finally c' 5,4,3,2 And c' 1 And fusing to obtain a first characteristic.
Further preferably, the screening function is:
F＝gelu(GeM(conv1(c i )))
wherein F represents a screening function, c i For the co-scale convolution fusion feature, i=1, 2,3,4,5, gelu is the activation function, geM is the average pooling layer, conv1 represents the 1*1 convolution layer.
Further preferably, the formula for performing layer-by-layer fusion on the co-scale convolution screening feature sequences is as follows:
c' 5,4 ＝σ 4 (F 4 ·c' 4 )+F 5 ·c' 5
c' 5,4,3 ＝σ 3 (F 3 ·c' 3 )+F 5,4 ·c' 5,4
c' 5,4,3,2 ＝σ 2 (F 2 ·c' 2 )+F 5,4,3 ·c' 5,4,3
A＝σ 1 (F 1 ·c' 1 )+F 5,4,3,2 ·c' 5,4,3,2
wherein A is a first feature, F is a screening function,c' i for co-scale convolution screening features, i=1, 2,3,4,5, a is a constant and tanh represents a hyperbolic function.
Further preferably, step S5 includes:
s51, constructing a health assessment model comprising a health grade and a health assessment function;
s52, dividing the first features according to subordinate sub-task systems to obtain N first feature subsets, inputting the N first feature subsets into a health evaluation model, performing health judgment on the first feature subsets of each sub-task system according to health evaluation functions and health grades of the sub-task systems to obtain health state evaluation results of the N sub-task systems, and obtaining health state description of the whole ship task system according to the health state evaluation results of the N sub-task systems.
Further preferably, step S52 includes:
s521, classifying the health status into three health grades of health, slight abnormality and serious abnormality;
s522 divides the first feature subset of the single sub-task system into a plurality of first feature clusters { { A by the device of the sub-task system 1 },{A 2 },...{A n -n is the number of devices, a represents the first feature;
s523 for a single device, its first feature cluster is { A j Determining a health parameter Λ= { λ of the device 1 ,λ 2 ,...,λ m -wherein m is the number of health parameters, which refer to the presence of a location node in the device reflecting the health status of the device;
s524, determining a threshold set of each health parameter through a threshold selection method, wherein the threshold set comprises three thresholds which respectively correspond to three health levels;
s525 determines the influence degree of each health parameter on the health state of the equipment, quantifies the influence degree into weight values, and forms a weight value set B= { B 1 ,B 2 ,...,B m }；
S526 in { A j Determining a first feature corresponding to the health parameter according to the feature information, taking the first feature as a health parameter feature set, calculating a feature mean value of each health parameter feature set, and taking the feature mean value as a feature value of each health parameter;
s527 builds a health evaluation function according to the characteristic value, the threshold value set and the weight value set of each health parameter, and obtains the health grade of the equipment according to the health evaluation function;
s528, repeating the steps S523-S527 to obtain the health grade of all the devices in the single sub-task system, and forming a health state evaluation result of the sub-task system according to the health grade of the devices;
and S529, obtaining the health state description of the whole ship task system according to the health state evaluation results of all the sub-task systems.
Further preferably, the health assessment function is:
wherein G is a healthy state,is the characteristic value of the kth health parameter, T is the threshold set, < >>Health class discriminant function representing kth health parameter, B k The weight value of the kth health parameter, m is the number of health parameters, τ 1 、τ 2 、τ 3 A first index, a second index and a third index respectively, wherein the first index is an index value of health of the health class,the second index is an index value with a health grade being slightly abnormal, the third index is an index value with a health grade being severely abnormal, G 1 Indicating that the health grade is healthy, G 2 Indicating a slight abnormality in health grade, G 3 Indicating that the health grade is severely abnormal.
Compared with the prior art, the method has the following beneficial effects:
(1) The method comprises the steps of constructing a digital twin system and correcting to obtain real-time data and virtual data, fusing the two data in characteristic level to enhance the expression capacity of characteristics, and removing redundant information of the characteristics through characteristic screening so as to improve the accuracy of detection when the ship task system is healthily detected through the characteristics;
(2) The set feature fusion module comprises two fusion networks, firstly, extracting multi-scale features of two kinds of data, and for the same kind of data, carrying out self-fusion on the multi-scale features of the two kinds of data, then, gradually fusing the fused features of the two kinds of data from small to large in scale, and introducing a screening function in the feature fusion to increase the performance of the features so that the features contain rich information;
(3) The health state of the individual equipment is judged according to the health parameters of each equipment, a health evaluation function is constructed by combining the characteristic values, the threshold value sets and the weight value sets of the health parameters, and the health indexes of the three equipment are introduced to evaluate the health state of the equipment in multiple layers and multiple directions.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method according to an embodiment of the invention.
Detailed Description
The following description of the embodiments of the present invention will clearly and fully describe the technical aspects of the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, are intended to fall within the scope of the present invention.
As shown in fig. 1, the invention provides a ship task system health state detection method based on virtual-real combination, which comprises the following steps:
s1, acquiring original data of a ship task system, constructing a digital twin system according to the ship task system and the original data thereof, and correcting the digital twin system, wherein the original data comprises historical data and real-time data;
s2, virtual data of the digital twin system are obtained, and both the real-time data and the virtual data are preprocessed;
s3, constructing a neural network frame, wherein the neural network frame comprises a feature extraction module and a feature fusion module;
s4, inputting the preprocessed real-time data and virtual data into a deep neural network frame, obtaining real-time features and virtual features through a feature extraction module, and fusing the real-time features and the virtual features by utilizing a feature fusion module to obtain first features;
s5, building a health evaluation model, and inputting the first characteristic into the health evaluation model to obtain the health state description of the ship task system.
Specifically, in an embodiment of the present invention, step S1 includes:
the ship task system comprises N sub-task systems, each sub-task system comprises a plurality of devices, each device comprises device real-time data and device historical data, all the device real-time data are used as real-time data of the ship task system, all the device historical data are used as historical data of the ship task system, and the real-time data and the historical data form original data;
obtaining elements of physical entities of a ship task system, wherein the elements comprise geometric parameters, physical attributes, operation modes, man-machine interaction, information flow and interface relation, adding assembly constraint relation and rule constraint relation, combining element calculation to generate a combined model, and setting motion constraint on a combined die to obtain a digital twin system;
and (3) performing simulation verification on the single sub-task system, inputting the original data of the sub-task system and the elements of the physical entity of the sub-task system into the digital twin system, and adjusting the parameters of the digital twin system according to the deviation value of the simulation result and the actual result until the deviation value is within a preset range, so as to obtain the corrected digital twin system.
In this embodiment, the real-time data of the device may be data that is currently detected by using each element of the device to select points, and the historical data of the device may be all data that is detected in the past.
In this embodiment, the material attribute is divided into a static physical attribute and a dynamic physical attribute, where specific parameters of the static physical attribute include system setting parameters such as weight, displacement, draft, speed, material, execution range of each subtask system, and the dynamic physical attribute is a system parameter that dynamically changes according to the process of executing the task by the ship.
In this embodiment, the digital twin system is obtained by modeling in three-dimensional software, the accuracy of the digital twin system can reach the actual requirement through a data exchange technology, when data acquisition is performed on a certain device, the ship task system performs field acquisition according to the actual device, and the digital twin system relies on a simulation technology to acquire virtual data.
In this embodiment, since the digital twin system is obtained by three-dimensional modeling, a step of model verification may be further provided, specifically, a deviation value may be preset, and the deviation value may be set according to actual experience, and a simulation result of the digital twin system is obtained by using three-dimensional simulation, and is compared with an actual result of the ship mission system to obtain a deviation value, and correction is performed based on the deviation value. In order to facilitate the verification step and reduce the calculation amount, a subtask system or a certain device thereof can be selected as a correction unit to correct the model.
Specifically, in step S2, the preprocessing includes: data cleansing and data correction, wherein:
the data cleaning comprises data denoising, redundant data removal and data standardization;
data correction includes missing value processing, error value correction, and outlier processing.
There are many ways to pre-process the data, and the invention is not limited in this regard, as this step is to improve the quality of the data.
Specifically, in an embodiment of the present invention, step S4 includes:
s41, the feature extraction module comprises a multi-scale convolution network, real-time data and virtual data are sequentially input into the multi-scale convolution network to be extracted to obtain convolution features of a plurality of scales, the convolution features of the plurality of scales form a feature sequence, the feature sequence of the real-time data is used as a real-time feature, and the feature sequence of the virtual data is used as a virtual feature;
the S42 feature fusion module comprises a first fusion network and a second fusion network, the first fusion network is utilized to respectively fuse the convolution features of the same scale of the real-time features and the virtual features, the obtained same-scale convolution fusion feature sequences are input into the second fusion network, and the same-scale convolution feature sequences are sequentially fused in pairs according to the sequence from small scale to large scale, so that the first features are obtained.
Specifically, in the present embodiment, step S41 includes:
the multi-scale convolution network comprises a 7*7 convolution layer and 4 convolution blocks;
respectively inputting the real-time data x and the virtual data y into a multi-scale convolution network, and extracting convolution characteristics x of a first scale according to a 7*7 convolution layer 1 And y 1 Then sequentially inputting 4 convolution blocks, and respectively extracting to obtain 4-scale convolution characteristics x 2 、x 3 、x 4 、x 5 And y 2 、y 3 、y 4 、y 5 The scale of the convolution features is sequentially increased according to the data flow direction;
composing convolution characteristics of 5 scales of real-time data into characteristics thereofSequence, get real-time feature x= { X 1 ,x 2 ,x 3 ,x 4 ,x 5 The convolution characteristics of 5 scales of the virtual data are formed into a characteristic sequence, and the virtual characteristic Y= { Y is obtained 1 ,y 2 ,y 3 ,y 4 ,y 5 }。
Accordingly, step S42 includes:
the real-time feature x= { X 1 ,x 2 ,x 3 ,x 4 ,x 5 Sum virtual feature y= { Y 1 ,y 2 ,y 3 ,y 4 ,y 5 Input into the first converged network, respectively x 1 And y is 1 Fusion to obtain c 2 、x 2 And y is 2 Fusion to obtain c 2 、x 3 And y is 3 Fusion to obtain c 3 、x 4 And y is 4 Fusion to obtain c 4 、x 5 And y is 5 Fusion to obtain c 5 C, adding 1 、c 2 、c 3 、c 4 、c 5 Constitute the same-scale convolution fusion characteristic sequence C= { C 1 ,c 2 ,c 3 ,c 4 ,c 5 }；
Fusing the same-scale convolution feature sequence C= { C 1 ,c 2 ,c 3 ,c 4 ,c 5 Inputting into a second fusion network, and fusing the characteristic sequence C= { C for the same-scale convolution by using a screening function 1 ,c 2 ,c 3 ,c 4 ,c 5 Feature screening is carried out, redundant information is removed, and a same-scale convolution screening feature sequence C ' = { C ' is obtained ' 1 ,c' 2 ,c' 3 ,c' 4 ,c' 5 }；
The feature sequence C ' = { C ' is screened by co-scale convolution ' 1 ,c' 2 ,c' 3 ,c' 4 ,c' 5 Performing layer-by-layer fusion according to the order of the scale from small to large, and firstly fusing c' 5 And c' 4 Fusion to give c' 5,4 And c 'is carried out again' 5,4 And c' 3 Fusion to give c' 5,4,3 After which c' 5,4,3 And c' 2 Fusion to give c' 5,4,3,2 Finally c' 5,4,3,2 And c' 1 And fusing to obtain a first characteristic.
In this embodiment, the screening function is:
F＝gelu(GeM(conv1(c i )))
wherein F represents a screening function, c i For the co-scale convolution fusion feature, i=1, 2,3,4,5, gelu is the activation function, geM is the average pooling layer, conv1 represents the 1*1 convolution layer.
Specifically, the formula for carrying out layer-by-layer fusion on the co-scale convolution screening characteristic sequences is as follows:
c' 5,4 ＝σ(F 4 ·c' 4 )+F 5 ·c' 5
c' 5,4,3 ＝σ(F 3 ·c' 3 )+F 5,4 ·c' 5,4
c' 5,4,3,2 ＝σ(F 2 ·c' 2 )+F 5,4,3 ·c' 5,4,3
A＝σ(F 1 ·c' 1 )+F 5,4,3,2 ·c' 5,4,3,2
wherein A is a first feature, F is a screening function,a is a constant and tanh represents a hyperbolic function.
In a specific example of the present invention, the multi-scale convolution network may be a residual network, where the convolution blocks are residual blocks, and the first scale convolution feature is extracted by the 7*7 convolution layer, and then four different scale convolution features are obtained sequentially according to the four residual blocks.
It should be noted that, in this embodiment, the virtual features of the real-time features are fused between the same scales, and after the fusion step is performed, the same-scale convolution fusion feature sequence is input into the second fusion network to perform the first screening, so as to avoid the condition that the redundancy or fusion abnormality occurs when two different features are fused in the same scale to cause the feature performance to be reduced, and one screening is performed on the same-scale convolution feature to improve the quality of the feature.
It should be noted that, in this embodiment, in the process of performing layer-by-layer fusion on the same-scale convolution screening feature sequences according to the order of the scale from small to large, a screening function is added, which is to screen the features just fused and to be fused in the fusion process, and belongs to the second screening.
Specifically, in an embodiment of the present invention, step S5 includes:
s51, constructing a health assessment model comprising a health grade and a health assessment function;
s52, dividing the first features according to subordinate sub-task systems to obtain N first feature subsets, inputting the N first feature subsets into a health evaluation model, performing health judgment on the first feature subsets of each sub-task system according to health evaluation functions and health grades of the sub-task systems to obtain health state evaluation results of the N sub-task systems, and obtaining health state description of the whole ship task system according to the health state evaluation results of the N sub-task systems.
Specifically, in the present embodiment, step S52 includes:
s521, classifying the health status into three health grades of health, slight abnormality and serious abnormality;
s522 divides the first feature subset of the single sub-task system into a plurality of first feature clusters { { A by the device of the sub-task system 1 },{A 2 },...{A n -n is the number of devices, a represents the first feature;
s523 for a single device, its first feature cluster is { A j Determining a health parameter Λ= { λ of the device 1 ,λ 2 ,...,λ m -wherein m is the number of health parameters, which refer to the presence of a location node in the device reflecting the health status of the device;
s524, determining a threshold set of each health parameter through a threshold selection method, wherein the threshold set comprises three thresholds which respectively correspond to three health levels;
s525 determines the influence degree of each health parameter on the health state of the equipment, quantifies the influence degree into weight values, and forms a weight value set B= { B 1 ,B 2 ,...,B m }；
S526 in { A j Determining a first feature corresponding to the health parameter according to the feature information, taking the first feature as a health parameter feature set, calculating a feature mean value of each health parameter feature set, and taking the feature mean value as a feature value of each health parameter;
s527 builds a health evaluation function according to the characteristic value, the threshold value set and the weight value set of each health parameter, and obtains the health grade of the equipment according to the health evaluation function;
s528, repeating the steps S523-S527 to obtain the health grade of all the devices in the single sub-task system, and forming a health state evaluation result of the sub-task system according to the health grade of the devices;
and S529, obtaining the health state description of the whole ship task system according to the health state evaluation results of all the sub-task systems.
Specifically, the health assessment function is:
where G is a health evaluation value,is the characteristic value of the kth health parameter, T is the threshold set, < >>Health class discriminant function representing kth health parameter, B k The weight value of the kth health parameter, m is the number of health parameters, τ 1 、τ 2 、τ 3 The first index, the second index and the third index are respectively, wherein the first index is an index value of health, the second index is a health grade and is lightThe third index is that the health grade is the index of serious abnormality, G 1 Indicating that the health grade is healthy, G 2 Indicating a slight abnormality in health grade, G 3 Indicating that the health grade is severely abnormal.
Step S5 is described in a specific embodiment:
1. the health grade according to which the health state is detected in this embodiment is determined, and three health grades are set in this embodiment, namely health, slight abnormality and serious abnormality.
2. In this embodiment, the detection of the health status is performed by using a device as a unit, the device includes a plurality of elements, real-time data of the device is obtained by testing and collecting the elements, and the corresponding digital devices are modeled in the digital twin system. Therefore, the first features can be divided into a plurality of first feature clusters { { A according to the equipment according to the data of different feature membership 1 },{A 2 },...{A n I.e. each feature cluster is affiliated with a single device.
3. The health state of the element can be reflected to a certain degree by selecting the element which can directly represent the health state of the device from all elements of the device, namely, detecting the health state of the element through the data tested and collected by the element. The location nodes refer to these elements. These elements are selected as health parameters, and in particular, the number of health parameters is m, and the value of m is determined according to the actual elements.
4. The threshold set of each health parameter is determined by a threshold selection method, specifically, the threshold selection method may be: combining the law of large numbers and the center limit theory, determining a threshold benchmark based on a 3 sigma criterion according to the data distribution expectations and variances of health parameters, and setting three thresholds according to the threshold benchmark, wherein the three thresholds respectively correspond to three health levels.
5. Each health parameter can reflect the health state of the equipment to a certain extent, but the influence degree of the health parameter on the equipment is different, and each health parameter is judged to be set according to the actual operation or the experience valueThe specific influence degree of the equipment is obtained, numerical quantization is carried out according to the influence degree of all health parameters on the equipment, a weight value is formed, and the weight value is formed into a set B= { B 1 ,B 2 ,...,B m }. Specifically, the influence degree can be comprehensively determined according to the importance degree of the health parameter for ensuring the normal operation of the equipment, the key degree of the data collected by testing the health parameter, and the like, in this embodiment, the influence degree is scored in percentage form according to the result of the comprehensive determination, and then normalized to obtain a weight value, wherein the weight value is [0,1]And the sum of all weight values in the weight value set B is 1.
6. In the case of collecting real-time data of a device and virtual data of the device, the data of the device are actually collected one by one, and the first feature has characteristic information which describes the original source of the feature, so that the data can be obtained in { A } j Selecting first features corresponding to the health parameters, taking the first features as a health parameter feature set corresponding to the health parameters, adding and averaging the feature values in each health parameter feature set, and taking the obtained values as the feature values of each health parameter, wherein the feature values can represent the actual conditions of the health parameters to a certain extent.
7. The health evaluation function is constructed according to the characteristic value, the threshold value set and the weight value set of each health parameter, the health state of the health parameter is judged by a health grade judging function firstly, specifically, the characteristic value and the threshold value set of the health parameter can be compared, the threshold value in the threshold value set is selected according to the data of the health parameter, so that the threshold value can be directly used for judging the characteristic value of the health parameter, in the embodiment, each health parameter is provided with a specific threshold value set, the health grade of the health parameter can be obtained according to the judging result of the characteristic value and the threshold value set, then the health grade of the whole equipment can be obtained by combining the weight value of the health parameter, at the moment, the health grade of the equipment is judged according to the calculating result of the health evaluation function and the health index, and the health index is classified into a first index tau according to the health grade 1 Second index τ 2 And a third index τ 3 . For example, the health evaluation value of the whole apparatus is at τ 1 The health rating of the device is considered healthy when it is.
8. And repeating the steps to evaluate the health grade of each device in the single sub-task system, forming a table of the health grade of all the devices, and forming a health state evaluation result of the sub-task system by the table name of the sub-task system.
9. Repeating the steps for each sub-task system to obtain the health state evaluation results of all the sub-task systems, namely, the health state tables of all the sub-task systems, and summarizing the tables to form a folder which is used as the health state description of the ship task system. The file folder can be adjusted by a follow-up professional to carry out health check on the whole ship task system, determine which elements or equipment need to be debugged, maintained and the like, and after the ship task system is updated, the updated content is synchronized to the digital twin system by utilizing a data exchange technology.
The invention utilizes the digital twin technology, detects the health state of the ship task system through virtual-real combined data driving, finds faults in time and ensures the normal operation of the ship task system.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.

Claims (10)

1. A ship task system health state detection method based on virtual-real combination is characterized by comprising the following steps:
s1, acquiring original data of a ship task system, constructing a digital twin system according to the ship task system and the original data thereof, and correcting the digital twin system, wherein the original data comprises historical data and real-time data;
s2, virtual data of the digital twin system are obtained, and both the real-time data and the virtual data are preprocessed;
s3, constructing a neural network frame, wherein the neural network frame comprises a feature extraction module and a feature fusion module;
s4, inputting the preprocessed real-time data and virtual data into a deep neural network frame, obtaining real-time features and virtual features through a feature extraction module, and fusing the real-time features and the virtual features by utilizing a feature fusion module to obtain first features;
s5, building a health evaluation model, and inputting the first characteristic into the health evaluation model to obtain the health state description of the ship task system.
2. The method of claim 1, wherein step S1 comprises:
the ship task system comprises N sub-task systems, each sub-task system comprises a plurality of devices, each device comprises device real-time data and device historical data, all the device real-time data are used as real-time data of the ship task system, all the device historical data are used as historical data of the ship task system, and the real-time data and the historical data form original data;
obtaining elements of physical entities of a ship task system, wherein the elements comprise geometric parameters, physical attributes, operation modes, man-machine interaction, information flow and interface relation, adding assembly constraint relation and rule constraint relation, combining element calculation to generate a combined model, and setting motion constraint on a combined die to obtain a digital twin system;
and (3) performing simulation verification on the single sub-task system, inputting the original data of the sub-task system and the elements of the physical entity of the sub-task system into the digital twin system, and adjusting the parameters of the digital twin system according to the deviation value of the simulation result and the actual result until the deviation value is within a preset range, so as to obtain the corrected digital twin system.
3. The method of claim 1, wherein step S4 comprises:
s41, the feature extraction module comprises a multi-scale convolution network, real-time data and virtual data are sequentially input into the multi-scale convolution network to be extracted to obtain convolution features of a plurality of scales, the convolution features of the plurality of scales form a feature sequence, the feature sequence of the real-time data is used as a real-time feature, and the feature sequence of the virtual data is used as a virtual feature;
the S42 feature fusion module comprises a first fusion network and a second fusion network, the first fusion network is utilized to respectively fuse the convolution features of the same scale of the real-time features and the virtual features, the obtained same-scale convolution fusion feature sequences are input into the second fusion network, and the same-scale convolution feature sequences are sequentially fused in pairs according to the sequence from small scale to large scale, so that the first features are obtained.
4. A method as claimed in claim 3, wherein step S41 comprises:
the multi-scale convolution network comprises a 7*7 convolution layer and 4 convolution blocks;
respectively inputting the real-time data x and the virtual data y into a multi-scale convolution network, and extracting convolution characteristics x of a first scale according to a 7*7 convolution layer 1 And y 1 Then sequentially inputting 4 convolution blocks, and respectively extracting to obtain 4-scale convolution characteristics x 2 、x 3 、x 4 、x 5 And y 2 、y 3 、y 4 、y 5 The scale of the convolution features is sequentially increased according to the data flow direction;
the convolution characteristics of 5 scales of real-time data are formed into a characteristic sequence, and the real-time characteristic X= { X is obtained 1 ，x 2 ，x 3 ，x 4 ，x 5 The convolution characteristics of 5 scales of the virtual data are formed into a characteristic sequence, and the virtual characteristic Y= { Y is obtained 1 ，y 2 ，y 3 ，y 4 ，y 5 }。
5. The method of claim 4, wherein step S42 comprises:
the real-time feature x= { X 1 ，x 2 ，x 3 ，x 4 ，x 5 Sum virtual feature y= { Y 1 ，y 2 ，y 3 ，y 4 ，y 5 Input into the first converged network, respectively x 1 And y is 1 Fusion to obtain c 2 、x 2 And y is 2 Fusion to obtain c 2 、x 3 And y is 3 Fusion to obtain c 3 、x 4 And y is 4 Fusion to obtain c 4 、x 5 And y is 5 Fusion to obtain c 5 C, adding 1 、c 2 、c 3 、c 4 、c 5 Constitute the same-scale convolution fusion characteristic sequence C= { C 1 ，c 2 ，c 3 ，c 4 ，c 5 }；
Fusing the same-scale convolution feature sequence C= { C 1 ，c 2 ，c 3 ，c 4 ，c 5 Inputting into a second fusion network, and fusing the characteristic sequence C= { C for the same-scale convolution by using a screening function 1 ，c 2 ，c 3 ，c 4 ，c 5 Feature screening is carried out, redundant information is removed, and a same-scale convolution screening feature sequence C ' = { C ' is obtained ' 1 ，c′ 2 ，c′ 3 ，c′ 4 ，c′ 5 }；
The feature sequence C ' = { C ' is screened by co-scale convolution ' 1 ，c′ 2 ，c′ 3 ，c′ 4 ，c′ 5 Performing layer-by-layer fusion according to the order of the scale from small to large, and firstly fusing c' 5 And c' 4 Fusion to give c' 5，4 And c 'is carried out again' 5，4 And c' 3 Fusion to give c' 5，4，3 After which c' 5，4，3 And c' 2 Fusion to give c' 5，4，3，2 Finally c' 5，4，3，2 And c' 1 And fusing to obtain a first characteristic.
6. The method of claim 5, wherein the screening function is:
F＝gelu(GeM(conv1(c i )))
wherein F represents a screening function, c i For the co-scale convolution fusion feature, i=1, 2,3,4,5, gelu is the activation function, geM is the average pooling layer, conv1 represents the 1*1 convolution layer.
7. The method of claim 6, wherein the formula for layer-by-layer fusion of the co-scale convolution screening feature sequences is:
c′ 5，4 ＝σ 4 (F 4 ·c′ 4 )+F 5 ·c′ 5
c′ 5，4，3 ＝σ 3 (F 3 ·c′ 3 )+F 5，4 ·c′ 5，4
c′ 5，4，3，2 ＝σ 2 (F 2 ·c′ 2 )+F 5，4，3 ·C′ 5，4，3
A＝σ 1 (F 1 ·c′ 1 )+F 5，4，3，2 ·c′ 5，4，3，2
wherein A is a first feature, F is a screening function,c′ i for co-scale convolution screening features, i=1, 2,3,4,5, a is a constant and tanh represents a hyperbolic function.
8. The method of claim 2, wherein step S5 comprises:
s51, constructing a health assessment model comprising a health grade and a health assessment function;
s52, dividing the first features according to subordinate sub-task systems to obtain N first feature subsets, inputting the N first feature subsets into a health evaluation model, performing health judgment on the first feature subsets of each sub-task system according to health evaluation functions and health grades of the sub-task systems to obtain health state evaluation results of the N sub-task systems, and obtaining health state description of the whole ship task system according to the health state evaluation results of the N sub-task systems.
9. The method of claim 8, wherein step S52 comprises:
s521, classifying the health status into three health grades of health, slight abnormality and serious abnormality;
s522 divides the first feature subset of the single sub-task system into a plurality of first feature clusters { { A by the device of the sub-task system 1 }，{A 2 }，...{A n -n is the number of devices, a represents the first feature;
s523 for a single device, its first feature cluster is { A j Determining a health parameter Λ= { λ of the device 1 ，λ 2 ，...，λ m -wherein m is the number of health parameters, which refer to the presence of a location node in the device reflecting the health status of the device;
s524, determining a threshold set of each health parameter through a threshold selection method, wherein the threshold set comprises three thresholds which respectively correspond to three health levels;
s525 determines the influence degree of each health parameter on the health state of the equipment, quantifies the influence degree into weight values, and forms a weight value set B= { B 1 ,B 2 ,...,B m }；
S526 in { A j Determining a first feature corresponding to the health parameter according to the feature information, taking the first feature as a health parameter feature set, calculating a feature mean value of each health parameter feature set, and taking the feature mean value as a feature value of each health parameter;
s527 builds a health evaluation function according to the characteristic value, the threshold value set and the weight value set of each health parameter, and obtains the health grade of the equipment according to the health evaluation function;
s528, repeating the steps S523-S527 to obtain the health grade of all the devices in the single sub-task system, and forming a health state evaluation result of the sub-task system according to the health grade of the devices;
and S529, obtaining the health state description of the whole ship task system according to the health state evaluation results of all the sub-task systems.
10. The method of claim 9, wherein the health assessment function is:
where G is a health evaluation value,is the characteristic value of the kth health parameter, T is the threshold set, < >>Health class discriminant function representing kth health parameter, B k The weight value of the kth health parameter, m is the number of health parameters, τ 1 、τ 2 、τ 3 The first index is an index value of health, the second index is an index value of slight abnormality, the third index is an index value of serious abnormality, G 1 Indicating that the health grade is healthy, G 2 Indicating a slight abnormality in health grade, G 3 Indicating that the health grade is severely abnormal.
CN202310948313.9A 2023-07-27 2023-07-27 Ship task system health state detection method based on virtual-real combination Active CN116957361B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310948313.9A CN116957361B (en) 2023-07-27 2023-07-27 Ship task system health state detection method based on virtual-real combination

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310948313.9A CN116957361B (en) 2023-07-27 2023-07-27 Ship task system health state detection method based on virtual-real combination

Publications (2)

Publication Number Publication Date
CN116957361A true CN116957361A (en) 2023-10-27
CN116957361B CN116957361B (en) 2024-02-06

Family

ID=88452710

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310948313.9A Active CN116957361B (en) 2023-07-27 2023-07-27 Ship task system health state detection method based on virtual-real combination

Country Status (1)

Country Link
CN (1) CN116957361B (en)

Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111817880A (en) * 2020-06-17 2020-10-23 安徽创米信息技术有限公司 Oil and gas field production equipment health management system and implementation method
CN112418451A (en) * 2020-11-05 2021-02-26 武汉大学 Transformer fault diagnosis positioning system based on digital twinning
CN112508373A (en) * 2020-11-27 2021-03-16 江苏科技大学 Marine diesel engine performance evaluation system based on digital twin technology
CN112685949A (en) * 2020-11-25 2021-04-20 江苏科技大学 Transformer health prediction method based on digital twinning
CN113221338A (en) * 2021-04-25 2021-08-06 南京邮电大学 Digital twin-based complex equipment system model verification method and system
CN113247041A (en) * 2021-05-12 2021-08-13 上海工程技术大学 Wheel set axle box device state monitoring method based on digital twin model
CN113488997A (en) * 2021-07-06 2021-10-08 浙江工业大学 Micro-grid critical energy evaluation method based on disturbance response multi-scale features
CN113569475A (en) * 2021-07-21 2021-10-29 上海工程技术大学 Subway axle box bearing fault diagnosis system based on digital twinning technology
CN113835060A (en) * 2021-11-23 2021-12-24 武汉格蓝若智能技术有限公司 Power transformer online state monitoring method and system based on digital twinning
CN114740969A (en) * 2022-02-09 2022-07-12 北京德信电通科技有限公司 Interaction system and method for fusing digital twin and virtual reality
WO2022226034A1 (en) * 2021-04-21 2022-10-27 Northwestern University Hierarchical deep learning neural networks-artificial intelligence: an ai platform for scientific and materials systems innovation
WO2023020388A1 (en) * 2022-03-14 2023-02-23 中国长江三峡集团有限公司 Gearbox fault diagnosis method and apparatus, gearbox signal collection method and apparatus, and electronic device
KR20230050812A (en) * 2021-10-08 2023-04-17 울산대학교 산학협력단 Remaining useful life prediction method of fuel cell system and digital twin device performing the same
CN116246366A (en) * 2021-09-18 2023-06-09 南京航空航天大学 Unmanned aerial vehicle healthy operation assessment method based on digital twinning

Patent Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111817880A (en) * 2020-06-17 2020-10-23 安徽创米信息技术有限公司 Oil and gas field production equipment health management system and implementation method
CN112418451A (en) * 2020-11-05 2021-02-26 武汉大学 Transformer fault diagnosis positioning system based on digital twinning
CN112685949A (en) * 2020-11-25 2021-04-20 江苏科技大学 Transformer health prediction method based on digital twinning
CN112508373A (en) * 2020-11-27 2021-03-16 江苏科技大学 Marine diesel engine performance evaluation system based on digital twin technology
WO2022226034A1 (en) * 2021-04-21 2022-10-27 Northwestern University Hierarchical deep learning neural networks-artificial intelligence: an ai platform for scientific and materials systems innovation
CN113221338A (en) * 2021-04-25 2021-08-06 南京邮电大学 Digital twin-based complex equipment system model verification method and system
CN113247041A (en) * 2021-05-12 2021-08-13 上海工程技术大学 Wheel set axle box device state monitoring method based on digital twin model
CN113488997A (en) * 2021-07-06 2021-10-08 浙江工业大学 Micro-grid critical energy evaluation method based on disturbance response multi-scale features
CN113569475A (en) * 2021-07-21 2021-10-29 上海工程技术大学 Subway axle box bearing fault diagnosis system based on digital twinning technology
CN116246366A (en) * 2021-09-18 2023-06-09 南京航空航天大学 Unmanned aerial vehicle healthy operation assessment method based on digital twinning
KR20230050812A (en) * 2021-10-08 2023-04-17 울산대학교 산학협력단 Remaining useful life prediction method of fuel cell system and digital twin device performing the same
CN113835060A (en) * 2021-11-23 2021-12-24 武汉格蓝若智能技术有限公司 Power transformer online state monitoring method and system based on digital twinning
CN114740969A (en) * 2022-02-09 2022-07-12 北京德信电通科技有限公司 Interaction system and method for fusing digital twin and virtual reality
WO2023020388A1 (en) * 2022-03-14 2023-02-23 中国长江三峡集团有限公司 Gearbox fault diagnosis method and apparatus, gearbox signal collection method and apparatus, and electronic device

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
丁恩杰;俞啸;廖玉波;吴传龙;陈伟;郁万里;王威;: "基于物联网的矿山机械设备状态智能感知与诊断", 煤炭学报, no. 06, pages 2308 - 2319 *
王安邦;孙文彬;段国林;: "基于数字孪生与深度学习技术的制造加工设备智能化方法研究", 工程设计学报, no. 06, pages 666 - 674 *
陈志鼎;梅李萍;: "基于数字孪生技术的水轮机虚实交互系统设计", 水电能源科学, no. 09, pages 167 - 170 *

Also Published As

Publication number Publication date
CN116957361B (en) 2024-02-06

Similar Documents

Publication Publication Date Title
CN105279365B (en) For the method for the sample for learning abnormality detection
Chien et al. A system for online detection and classification of wafer bin map defect patterns for manufacturing intelligence
CN108900546A (en) The method and apparatus of time series Network anomaly detection based on LSTM
CN106950945B (en) A kind of fault detection method based on dimension changeable type independent component analysis model
CN107153759B (en) A kind of aviation turbofan engine method for predicting residual useful life of multisource data fusion
CN108537259A (en) Train control on board equipment failure modes and recognition methods based on Rough Sets Neural Networks model
JP2006318263A (en) Information analysis system, information analysis method and program
CN115454778A (en) Intelligent monitoring system for abnormal time sequence indexes in large-scale cloud network environment
CN112416662A (en) Multi-time series data anomaly detection method and device
CN111638988A (en) Cloud host fault intelligent prediction method based on deep learning
CN117076171A (en) Abnormality detection and positioning method and device for multi-element time sequence data
CN111949496B (en) Data detection method and device
CN117411192B (en) Monitoring method and system for uninterruptible power supply parallel operation system
CN115858606A (en) Method, device and equipment for detecting abnormity of time series data and storage medium
CN117319223A (en) Digital twinning technology-based pilot cockpit visualization method and system
CN110320802B (en) Complex system signal time sequence identification method based on data visualization
CN116957361B (en) Ship task system health state detection method based on virtual-real combination
CN115904916A (en) Hard disk failure prediction method and device, electronic equipment and storage medium
US20240152818A1 (en) Methods for mitigation of algorithmic bias discrimination, proxy discrimination and disparate impact
CN106681791A (en) Incremental virtual machine anomaly detection method based on symmetric neighbor relation
CN113505850A (en) Boiler fault prediction method based on deep learning
CN113011748A (en) Recommendation effect evaluation method and device, electronic equipment and readable storage medium
CN109492913A (en) A kind of modularization Risk Forecast Method of distribution terminal, device and can storage medium
CN117371861B (en) Digital-based household service quality intelligent analysis method and system
CN113779011B (en) Data restoration method and device based on machine learning and computer equipment

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