CN113742195B - Bayesian neural network-based system health state prediction method - Google Patents
Bayesian neural network-based system health state prediction method Download PDFInfo
- Publication number
- CN113742195B CN113742195B CN202111097887.7A CN202111097887A CN113742195B CN 113742195 B CN113742195 B CN 113742195B CN 202111097887 A CN202111097887 A CN 202111097887A CN 113742195 B CN113742195 B CN 113742195B
- Authority
- CN
- China
- Prior art keywords
- node
- model
- bayesian
- integration
- state
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3466—Performance evaluation by tracing or monitoring
- G06F11/3495—Performance evaluation by tracing or monitoring for systems
-
- 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
- G06N3/047—Probabilistic or stochastic networks
-
- 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
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a system health state prediction method based on a Bayesian neural network, which comprises the following steps: dividing each node of the system structure to obtain a first integration node and a second integration node; establishing an original multi-level system Bayesian network model based on the first integration node to obtain a state distribution prediction result of the first integration node, and establishing a deep learning model and a Bayesian network model based on the second integration node to obtain a state distribution prediction result of the second integration node; and integrating the state distribution prediction result of the first integration node and the state distribution prediction result of the second integration node to realize the overall health state prediction of the system. The health state monitoring data are effectively utilized, and the conditions of data overlapping and coupling are avoided; the Bayesian neural network is utilized to divide and predict the system state, and the operation and maintenance of the subsequent system are effectively guided.
Description
Technical Field
The invention belongs to the technical field of system modeling and state prediction, and particularly relates to a Bayesian neural network-based system health state prediction method.
Background
With the continuous improvement of the system function requirement and the use performance requirement, the structural function complexity of the system shows a trend of rapid increase, which is reflected in that the scale of each component of the system is larger and more complex, the association relationship is more complex, and the structural functions of each component in the system have a hierarchical relationship, so that a multi-level system is formed. The multi-level complex system is widely applied to important fields of aviation, aerospace, navigation, railways, weaponry, production and manufacturing and the like. The enormous scale and complex internal relationships of the multi-level system can cause the fault of the system to be interactively transmitted between the components and the subsystems, and the tiny fault can cause the failure of the system, so that the health state change and the future change trend of the multi-level system need to be grasped, and maintenance and guarantee work needs to be taken at any time.
The method comprises the steps of reasonably utilizing system health state detection data and accurately recognizing the association relation of all components of the system to establish a system health state evaluation model. The multi-hierarchy system hierarchy generally comprises a plurality of hierarchies such as a system, a subsystem, a functional module and a component, wherein each hierarchy comprises a plurality of component units, functional interaction exists among the hierarchies, and the complex structure causes difficulty in effective utilization of multi-feature data of the hierarchy. In the task process of the multi-level system, due to the fact that complex coupling relations exist among component degradation data, and meanwhile incidence relations among nodes can also change dynamically along with degradation of the health state of the system, certain difficulty exists in modeling and describing the health state of the system under the conditions. Aiming at the two problems, the invention provides a system health state prediction method based on a Bayesian neural network, considering the problems of multi-feature data overlapping and node data relation ambiguity on the basis of a common system state prediction model establishing method.
Disclosure of Invention
The invention aims to provide a system health state prediction method based on a Bayesian neural network. The problems of multi-feature data overlapping and node data relation ambiguity can be considered on the common system state prediction model establishing method, the multi-level system complex node relation is effectively processed, and the health state prediction model is established. And a complete process from feature extraction, modeling prediction to state distribution conversion is realized, and finally, the state inference of the whole system is realized.
In order to achieve the purpose, the invention provides the following scheme:
a system health state prediction method based on a Bayesian neural network comprises the following steps:
dividing each node of the system structure to obtain a first integration node and a second integration node;
establishing an original multi-level system Bayesian network model based on the first integration node to obtain a state distribution prediction result of the first integration node, and establishing a deep learning model and a Bayesian network model based on the second integration node to obtain a state distribution prediction result of the second integration node;
and integrating the state distribution prediction result of the first integration node and the state distribution prediction result of the second integration node to realize the overall health state prediction of the system.
Preferably, each node of the system structure is divided according to the internal structure of the system, known health state monitoring data and actual use conditions.
Preferably, the first integration node is a data set which has a definite structure among nodes, clear mutual influence relationship and no coupling phenomenon with corresponding health state data; the second integration node is a data set with complex interaction relationship among nodes, overlapping with corresponding health state data and high coupling degree.
Preferably, the method for constructing the original multi-hierarchy system bayesian network model comprises: constructing a fault tree model of the system based on a fault tree analysis method, selecting a logic gate according to the logic relation of events in the system, connecting the determined top event with all direct events causing the top event to occur, and sequentially constructing step by step; and completing the conversion from the fault tree model to the Bayesian network model through the conversion relation between the event and the node and the conversion relation between the logic gate and the Bayesian network probability distribution to obtain the original multi-level system Bayesian network model.
Preferably, a deep learning model is established for the second integration node, the node characteristic parameters are used as input, health state prediction is performed on the integrated nodes, a system bayesian network model is established, and node health state data are converted into state distribution to obtain a state distribution prediction result of the second integration node.
Preferably, the process of converting the node health status data into the status distribution includes:
and (3) learning by using a variational Bayesian inference algorithm to obtain the health state distribution of the nodes, carrying out corresponding random sampling on the weight parameters of each neuron to form a plurality of groups of model parameters, forward calculating a model prediction result according to a neural network prediction method for each group of model parameters, and obtaining the node state probability distribution after multiple sampling.
Preferably, the original multi-level system bayesian network model is updated based on the state distribution prediction result of the first integration node and the state distribution prediction result of the second integration node to obtain a mixed bayesian network model, and the system state is inferred.
Preferably, based on the mixed Bayesian network model, the Bayesian network parameter learning is performed by using the training data to obtain conditional probability information, so that the whole system health state prediction process is realized.
The invention has the beneficial effects that:
the invention provides a system health state prediction method based on a Bayesian neural network, which comprises the following steps: (1) health state monitoring data are effectively utilized, and the conditions of data overlapping and data coupling are avoided; (2) the Bayesian neural network is utilized to divide and predict the system state, and the operation and maintenance of a subsequent system are effectively guided; (3) the whole method has clear flow, and the key point is to distinguish and utilize data, thereby ensuring the prediction precision and ensuring the simplicity and convenience of the prediction method.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a diagram illustrating relationships between nodes in an embodiment of the present invention;
FIG. 3 is a schematic diagram of a trapezoidal fuzzy membership function according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
A system health state prediction method based on a Bayesian neural network comprises the following steps:
dividing each node of the system structure to obtain a first integration node and a second integration node;
establishing an original multi-level system Bayesian network model based on the first integration node to obtain a state distribution prediction result of the first integration node, and establishing a deep learning model and a Bayesian network model based on the second integration node to obtain a state distribution prediction result of the second integration node;
and integrating the state distribution prediction result of the first integration node and the state distribution prediction result of the second integration node to realize the overall health state prediction of the system.
In this embodiment, the system refers to a final research object of the present invention, which may be an airplane, a group of batteries, a computer, etc.; equipment refers to devices in a system that perform different functions, such as engines in an aircraft, flight controllers, etc.; a node refers to an abstract representation of a low-level device (group) in a system, such as a device/subsystem/component, in a health state prediction model.
Step S1,
As shown in fig. 2, each section is divided into an integrated part with clear data relationship or complex data relationship in consideration of modeling requirements according to the internal structure of the system, known health status monitoring data, actual use conditions, and other contents.
The division principle can be formulated according to system characteristics and prediction requirements, all the nodes have clear structures and clear mutual influence relations, and the health state data corresponding to the nodes do not have obvious coupling phenomenon and can be combed into clear parts. On the contrary, all the parts where the influence relationship between the nodes is complicated, the corresponding health state data is largely overlapped, and the health state features are difficult to be clarified can be regarded as the integrated nodes with unclear data relationship.
Coupling refers to the closeness of the relationship between the health state data sets. The higher the coupling degree between data sets, the more the connection between the data sets is, the more the contents reflected by the data are overlapped, the worse the independence of the data sets is, and the larger the mutual influence between the data sets is.
Similarly, the health state data overlapping means that the data capable of reflecting the system health state may partially or completely exist in data sets, the contents that the data can reflect are consistent or similar (which may cause the subsequent calculation to become complex and waste the calculation resources), the independence of the data sets is poor, the mutual influence among the data sets is large, and the like.
Step S2,
And aiming at the first integration node with clear data relation, establishing a traditional system Bayesian network model to obtain a prediction result.
The Bayesian network is a directed acyclic graph description method based on a network structure, a system network topological structure is established by utilizing a directed graph model, and incidence relations and influence degrees among all information elements are expressed by utilizing the network model. Each node in the network represents a corresponding information element, a directional arrow of the correlation direction between the nodes represents the correlation relationship between the information elements, and the influence degree of the correlation relationship is represented by a conditional probability information table.
Step S3:
a second integration node for a complex, ambiguous integration part of the data relationship:
s3.1, carrying out degradation data processing on the nodes by using the LSTM;
s3.11, preparing network model input data;
note SoriThe node health state training data set adopted for prediction is normalized and recorded as S, and the node health state fusion characteristic data is expressed as Sc=[X1,X2,…,Xn,…XN]TN belongs to N, wherein N is the number of single nodes contained in the system; system keyHealth data are expressed as SP=[Y1,Y2,…,Ym,…,YM]TM belongs to M, wherein M is the type of the system health status index, and for simplifying the description, the number of the system health status indexes is assumed to be 1, namely SP=[Y]. Fusing feature data X for health status of each single nodenThe data over the time series T can be expressed as:
wherein T belongs to T as the time mark of the time sequence, and T is the acquisition duration of the time sequence; and P ∈ P is the health status index type (dimension) of the equipment node. Up to this point, system health indicator input-output rules have been given.
S3.12, network model building
The network model structure depends on the data input and output form, the data size and the optimization structure type. The network structure provided by the invention is based on an LSTM basic structure and a Dropout regularization method.
Inputting a layer: taking root node health state degradation fusion characteristic data as model input, wherein the dimensionality of the root node health state degradation fusion characteristic data depends on the number of nodes included in an estimated system and the characteristic number acquired by each node;
hiding the layer: the internal structure of the model consists of a plurality of layers of LSTM network structures, and the number of the layers and the number of the neurons depend on the dimension and the scale of data;
output layer: the model output is a degradation time sequence of a certain formulated index of the system, and the time dimension of the degradation time sequence is consistent with that of the input sequence.
S3.13 model solution
In the model training process, the setting of a Loss function (Loss function) is an important standard for model solution. The method adopts minimum Mean Square Error (MSE) as an optimization target. The specific calculation formula of the mean square error γ is as follows:
wherein, XiFor the actual sequence of samples to be observed,is a sample sequence XiThe mean square error MSE between the two can be obtained from γ in the above equation. Based on the loss function, a suitable optimization algorithm (e.g., RMSprop, Adam, etc.) is employed.
S3.14, model training and result output;
the training of the model is the most critical link for determining the effectiveness and the accuracy of the neural network model, and in order to obtain a satisfactory model training result in the application scene, the following contents need to be well realized:
normalization is carried out: the model training needs to be well normalized to eliminate prediction errors caused by different data scales.
Data set partitioning: generally, according to 8: 1: 1 or 7: 2: the proportion of 1 is used for dividing a training set, a verification set and a test set. The dividing proportion mainly considers that sufficient training data are provided for the model to optimize model parameters, the verification set scale which is not less than the test set is provided to ensure the training effect, and the data set dividing habit of general deep learning is followed.
Data sorting: according to the consistency of the front and rear characteristics of the data, the data are reordered, so that the training set contains various types of characteristic data, and the trained model has a good effect.
Training the model after finishing the input data processing, and judging specific model parameters according to the training data effect; the model output is set as the system state of health quantity predicted value.
S3.2, on the basis of the result of the step S3.1, carrying out state distribution conversion on the obtained node health state degradation data by using a Bayesian neural network model;
system multi-state partitioning based on fuzzy function:
for example, for p root nodes, the reliability obtained according to the prediction model is discretized in a time series manner in an unsupervised equal-width interval manner, and assuming that a system prediction result is required to be returned every unit time after the prediction model is established, the system state in the future T time is predicted according to the prediction step of the width, so as to obtain a p × T order matrix:
element R in the matrixi,TEach of the nodes X represents a node (i 1, 2.. times.p; and T1, 2.. times.t)iReliability prediction at discrete time τ. And then respectively dividing the reliability of the p nodes at different moments by fuzzy numbers on the basis of a fuzzy theory to the states possibly experienced in the degradation process. The whole process of decreasing the reliability is divided into processes by selecting proper fuzzy numbers and fuzzy membership functions, and the trapezoidal fuzzy number is taken as an example for the node XiAt R1The degree of membership of the "light fault" state is defined and the corresponding fuzzy membership function is shown in fig. 3.
When the reliability is between (R)b,Rc) Degree of membership in internal timeMeaning that the device must be in light failure, with Ri,τGradually trend towards RaOr RbThe probability of membership to this state is reduced to 0 and the corresponding membership function is shown below.
Sequentially marking the N states of the node from the worst state to the best state by 1-N, and then marking the node X at a certain timeiDegree of membership to each state 1,2, N has the following relationship:
step S4:
and constructing a Bayesian network model of the whole system, and finally realizing the whole health state prediction work of the system by combining two models with definite data relation and indefinite data relation.
This step still uses the traditional system Bayesian network model to obtain the prediction result. And the first integration node and the second integration node are used as nodes in the Bayesian network model of the system for training.
Based on the result of the step S3, obtaining a state distribution corresponding to the ambiguous integration node, and then performing system state inference by using a bayesian network parameter learning and state inference algorithm, wherein the relevant flow algorithm comprises:
inputting: node multi-feature fusion data;
s4.1, initialization:
a) establishing a system Bayesian network model BN based on the system model0;
b) Performing Node integration on the undefined nodes to form a virtual NodexIncorporating a system bayesian network;
s4.2, integrating node state distribution calculation
a) Establishing a deep learning prediction model with X ═ X1,...,xm) As input, predicting an integration node health state Y;
b) establishing a Bayesian neural network model BNN, taking Y as model input, and predicting the state distribution of the integration nodes
S4.3, model integration
a) Updating original network structure BN by using integrated node as new network node0To obtain a mixed system model BN1;
b) In BN1Based on the training data, the correlated Bayes is carried out by using the training dataAnd the network parameter learning obtains the conditional probability information.
And (3) outputting: the system health state is mixed with the predictive model.
The invention provides a system health state prediction method based on a Bayesian neural network, which comprises the following steps: (1) health state monitoring data are effectively utilized, and the conditions of data overlapping and data coupling are avoided; (2) the Bayesian neural network is utilized to divide and predict the system state, and the operation and maintenance of a subsequent system are effectively guided; (3) the whole method has clear flow, and the key point is to distinguish and utilize data, thereby ensuring the prediction precision and ensuring the simplicity and convenience of the prediction method.
The above-described embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solutions of the present invention can be made by those skilled in the art without departing from the spirit of the present invention, and the technical solutions of the present invention are within the scope of the present invention defined by the claims.
Claims (7)
1. A system health state prediction method based on a Bayesian neural network is characterized by comprising the following steps:
dividing each node of the system structure to obtain a first integration node and a second integration node; the division principle is formulated according to system characteristics and prediction requirements, and each section is divided into an integrated part with definite data relation or complex data relation;
the first integration node is a data set which has a definite structure among nodes, clear mutual influence relation and no coupling phenomenon with corresponding health state data; the second integration node is a data set with complex mutual influence relationship among nodes, overlapping with corresponding health state data and high coupling degree;
establishing an original multi-level system Bayesian network model based on the first integration node to obtain a state distribution prediction result of the first integration node, and establishing a deep learning model and a Bayesian network model based on the second integration node to obtain a state distribution prediction result of the second integration node;
integrating the state distribution prediction result of the first integration node and the state distribution prediction result of the second integration node to realize the overall health state prediction of the system;
wherein, the integration process includes:
obtaining the state distribution corresponding to the ambiguous integration node based on the state distribution prediction results of the first integration node and the second integration node, and performing system state inference by using a Bayesian network parameter learning and state inference algorithm, wherein the flow algorithm comprises:
inputting: node multi-feature fusion data;
s4.1, initialization:
a) establishing a system Bayesian network model BN based on the system model0;
b) Performing Node integration on the undefined nodes to form a virtual NodexIncorporating a system bayesian network;
s4.2, integrating node state distribution calculation:
a) establishing a deep learning prediction model with X ═ X1,...,xm) As input, predicting an integration node health state Y; wherein X is (X)1,...,xm) Fusing data for the node multi-features;
b) establishing a Bayesian neural network model BNN, taking Y as model input, and predicting the state distribution of the integration nodes;
s4.3, model integration:
a) updating original network structure BN by using integrated node as new network node0To obtain a mixed system model BN1;
b) In BN1On the basis, the training data is utilized to learn the related Bayesian network parameters to obtain conditional probability information;
and (3) outputting: the system health state is mixed with the predictive model.
2. The Bayesian neural network-based system health status prediction method according to claim 1, wherein each node of the system structure is divided according to the internal structure of the system, known health status monitoring data, and actual usage.
3. The Bayesian neural network-based system health status prediction method according to claim 1, wherein the original multi-level system Bayesian network model construction method comprises: constructing a fault tree model of the system based on a fault tree analysis method, selecting a logic gate according to the logic relation of events in the system, connecting the determined top event with all direct events causing the top event to occur, and sequentially constructing step by step; and completing the conversion from the fault tree model to the Bayesian network model through the conversion relation between the event and the node and the conversion relation between the logic gate and the Bayesian network probability distribution to obtain the original multi-level system Bayesian network model.
4. The Bayesian neural network-based system health status prediction method as recited in claim 1, wherein a deep learning model is established for the second integration node, the node characteristic parameters are used as input, health status prediction is performed on the integrated nodes, a system Bayesian network model is established again, and node health status data are converted into status distribution to obtain a status distribution prediction result of the second integration node.
5. The Bayesian neural network-based system health status prediction method of claim 4, wherein the process of transforming the node health status data into a state distribution comprises:
and (3) learning by using a variational Bayesian inference algorithm to obtain the health state distribution of the nodes, carrying out corresponding random sampling on the weight parameters of each neuron to form a plurality of groups of model parameters, forward calculating a model prediction result according to a neural network prediction method for each group of model parameters, and obtaining the node state probability distribution after multiple sampling.
6. The Bayesian neural network-based system health status prediction method according to claim 5, wherein the original multi-level system Bayesian network model is updated based on the status distribution prediction result of the first integration node and the status distribution prediction result of the second integration node, so as to obtain a mixed Bayesian network model, and system status inference is performed.
7. The Bayesian neural network-based system health status prediction method according to claim 6, wherein a Bayesian network parameter learning is performed by using training data based on the mixed Bayesian network model to obtain conditional probability information, so that a whole system health status prediction process is realized.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111097887.7A CN113742195B (en) | 2021-09-18 | 2021-09-18 | Bayesian neural network-based system health state prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111097887.7A CN113742195B (en) | 2021-09-18 | 2021-09-18 | Bayesian neural network-based system health state prediction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113742195A CN113742195A (en) | 2021-12-03 |
CN113742195B true CN113742195B (en) | 2022-03-29 |
Family
ID=78739891
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111097887.7A Active CN113742195B (en) | 2021-09-18 | 2021-09-18 | Bayesian neural network-based system health state prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113742195B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114280491B (en) * | 2021-12-23 | 2024-01-05 | 中山大学 | Retired battery residual capacity estimation method based on active learning |
CN116501027B (en) * | 2023-06-29 | 2023-10-03 | 中南大学 | Distributed braking system health assessment method, system, equipment and storage medium |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8762321B2 (en) * | 2011-07-18 | 2014-06-24 | Siemens Aktiengesellschaft | Method, system and computer program product for automatic generation of Bayesian networks from system reliability models |
CN103745108B (en) * | 2014-01-10 | 2016-08-17 | 电子科技大学 | The Reliability assessment method that multi-level Condition Monitoring Data merges |
JP6041331B1 (en) * | 2016-02-26 | 2016-12-07 | 国立大学法人山口大学 | Information processing apparatus, information processing program, and information processing method |
CN105913124B (en) * | 2016-04-08 | 2018-08-24 | 北京航空航天大学 | System health status prediction technique based on Bayesian network and base layer data |
CN110097100A (en) * | 2019-04-19 | 2019-08-06 | 太原理工大学 | A kind of Bridge State Assessment method based on discrete dynamic Bayesian network |
CN110489898B (en) * | 2019-08-26 | 2021-05-28 | 北京航空航天大学 | Dynamic multi-level system modeling and state prediction method based on hybrid cognition |
-
2021
- 2021-09-18 CN CN202111097887.7A patent/CN113742195B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN113742195A (en) | 2021-12-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111383452A (en) | Method for estimating and predicting short-term traffic running state of urban road network | |
CN113742195B (en) | Bayesian neural network-based system health state prediction method | |
CN109766583A (en) | Based on no label, unbalanced, initial value uncertain data aero-engine service life prediction technique | |
CN110321603A (en) | A kind of depth calculation model for Fault Diagnosis of Aircraft Engine Gas Path | |
CN106096830B (en) | Relay protection state evaluation method and system based on generalized evidence theory | |
CN104881609A (en) | Credibility evaluation method of software unit of complex software system | |
Minli et al. | Research on the application of artificial neural networks in tender offer for construction projects | |
CN109471698A (en) | System and method for detecting abnormal behavior of virtual machine in cloud environment | |
CN109919236A (en) | A kind of BP neural network multi-tag classification method based on label correlation | |
CN113076686A (en) | Aircraft trajectory prediction method based on social long-term and short-term memory network | |
CN113688253A (en) | Hierarchical perception temporal knowledge map representation learning method | |
CN115114128A (en) | Satellite health state evaluation system and evaluation method | |
CN110289987B (en) | Multi-agent system network anti-attack capability assessment method based on characterization learning | |
CN115757103A (en) | Neural network test case generation method based on tree structure | |
CN117131979A (en) | Traffic flow speed prediction method and system based on directed hypergraph and attention mechanism | |
Singaravel et al. | Explainable deep convolutional learning for intuitive model development by non–machine learning domain experts | |
CN113887125B (en) | Method for evaluating operation effectiveness of complex simulation system | |
Li et al. | Model construction for complex product design change with improved dendritic neural network | |
Lyu et al. | Examination on avionics system fault prediction technology based on ashy neural network and fuzzy recognition | |
CN115237091A (en) | Electromechanical device fault tracing method and system | |
Nikolos | On the use of multiple surrogates within a differential evolution procedure for high-lift airfoil design | |
Aziz et al. | Employing Recurrent Neural Networks to Forecast the Dollar Exchange Rate in the Parallel Market of Iraq | |
Li et al. | Application of combination forecasting model in aircraft failure rate forecasting | |
Patil et al. | Develop efficient technique of cost estimation model for software applications | |
CN114372627A (en) | Urban vehicle travel time estimation method based on hybrid deep learning framework |
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 |