CN111275136A - Fault prediction system based on small sample and early warning method thereof - Google Patents
Fault prediction system based on small sample and early warning method thereof Download PDFInfo
- Publication number
- CN111275136A CN111275136A CN202010176870.XA CN202010176870A CN111275136A CN 111275136 A CN111275136 A CN 111275136A CN 202010176870 A CN202010176870 A CN 202010176870A CN 111275136 A CN111275136 A CN 111275136A
- Authority
- CN
- China
- Prior art keywords
- data
- module
- characteristic
- feature
- fault
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- 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
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a fault prediction system based on small samples and an early warning method thereof.A data cleaning module is used for carrying out deletion filling, smoothing and redundant sample removing on system data; the characteristic engineering module is used for constructing a first-order difference characteristic of fault occurrence time, abnormal value identification of sensor data and an abnormal statistical characteristic; the characteristic selection module is used for screening out key base characteristics based on the number and algorithm of the division nodes in all the trees of the characteristics; the regression module is used for training a regression model and adjusting parameters by adopting a K-fold cross validation and grid search method so that the regression model reaches preset prediction precision; and the model fusion module is used for constructing a final prediction result by adopting a weighting fusion method for the plurality of models trained by the regressor module. The invention can greatly improve the failure time prediction precision under a small sample data set.
Description
Technical Field
The invention relates to a general fault early warning system, in particular to a fault prediction system based on a small sample and an early warning method thereof.
Background
The machine learning technology has made breakthrough development in the fields of image, medical treatment, recommendation and the like; in the production and manufacturing link, if the time of equipment failure occurrence can be accurately predicted, maintenance means can be provided more actively and ensured, the equipment downtime is reduced, and the maintenance labor cost is reduced. Therefore, the equipment failure prediction is the key for ensuring the efficient operation of the production process and is an important guarantee for realizing intelligent manufacturing.
Aiming at the problem of predicting the fault occurrence time in the production and manufacturing links, deep research has been carried out in academia and industry. Existing prediction methods include traditional methods and modern methods: the traditional method is that the average value of the interval time of the faults is counted and used as the next fault occurrence time; modern methods such as statistical machine learning based methods, neural network based methods, etc. The traditional method has simple logic and is easy to understand, but has lower precision. Modern methods are complex in logic and high in accuracy, but prediction on a small sample data set often generates overfitting and is poor in robustness, which is caused by difficulty in acquiring fault sample data. One problem that is often faced in practice is: the accumulation of raw fault data is insufficient, so that a phenomenon of overfitting to a greater degree is often generated when a target is directly modeled.
The traditional fault early warning system mainly comprises two types, one type is a classification problem, namely whether equipment is about to fail or not is judged according to equipment running state data, the fault early warning system can only obtain two types of 'fault' or 'no fault' of a prediction result, but has higher requirements on the number of positive and negative samples, and has no prediction capability on the occurrence time of the fault; the second type is regression problem, and the occurrence time of the future fault of the equipment is predicted by combining equipment operation state data before the corresponding fault according to the historical fault occurrence time of the equipment.
Disclosure of Invention
The invention aims to provide a fault prediction system based on a small sample, and the invention also aims to provide an early warning method of the fault prediction system. The intelligent algorithm is adopted to carry out discretization processing on the fault occurrence time, and then a sample is converted into data suitable for a classification algorithm, so that the data is input into the model to train the model, and the next fault occurrence time of the equipment is subjected to model prediction.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention relates to a fault prediction system based on a small sample, which comprises the following modules:
the data cleaning module is used for carrying out deletion filling, smoothing and redundant sample removing on the system data and transmitting the cleaned data to the feature engineering module to extract corresponding feature data;
the characteristic engineering module is used for constructing a first-order difference characteristic of fault occurrence time, sensor data abnormal value identification and abnormal statistical type characteristics;
the characteristic selection module is used for screening out key base characteristics based on an L1 regularization characteristic coefficient algorithm, a weighted mean square error reduction algorithm and a characteristic dividing node number and algorithm in all trees, and inputting the base characteristics into the regressor module;
the regression module trains a regression model and adjusts parameters by adopting a K-fold cross validation and grid search method to ensure that the regression model reaches preset prediction precision, and simultaneously predicts test data by using the trained regression module;
and the model fusion module is used for constructing a final prediction result by adopting a weighted fusion method for the plurality of models trained by the regressor module.
The invention relates to an early warning method of a fault prediction system based on a small sample, which comprises the following steps:
step 1, establishing a training sample database; the training sample database contains historical fault data of equipment to be pre-warned, and comprises: the method comprises the following steps of (1) fault code type, fault occurrence time, equipment operation parameters, equipment state parameters, equipment numbers, equipment types, equipment geographic information and equipment installation time;
step 2, performing deletion filling, smoothing and redundant sample elimination on system data through a data cleaning module, and transmitting the cleaned data to a feature engineering module to extract corresponding feature data;
step 3, the characteristic engineering module constructs a first-order difference characteristic of fault occurrence time, abnormal value identification of sensor data and abnormal statistical type characteristic through the characteristic data;
step 4, screening out key base characteristics by adopting a characteristic selection module based on an L1 regularization characteristic coefficient algorithm, a weighted mean square error reduction algorithm and a characteristic dividing node number and algorithm in all trees, and inputting the base characteristics into a regressor module;
step 5, the regressor module trains a regression model and adjusts parameters by adopting a K-fold cross validation and grid search method, so that the regression model reaches preset prediction precision, and the trained regressor module is used for predicting test data;
and 6, constructing a final prediction result by adopting a weighted fusion method for the plurality of models trained by the regressor module through a model fusion module.
In step 3, the characteristic engineering model identifies the abnormal values of the sensor data based on the Isolate Forest algorithm, so that various abnormal values of the sensor are counted and used as statistical characteristics, then the abnormal value data of the sensor are scored based on the maximum entropy algorithm, comprehensive scores of various samples are obtained, and finally a sliding window method is adopted to construct first-order difference characteristics of fault occurrence time.
The Isolate Forest algorithm calculates the number of layers of each sample in each binary tree by traversing a plurality of binary trees, and calculates the average height, so as to judge whether the sample is an abnormal sampling point according to a preset threshold value.
The Isolate Forest algorithm randomly selects a plurality of point samples from training data to be used as sub-samples, and the sub-samples are put into a root node of a tree.
The Isolate Forest algorithm randomly designates a dimension, randomly generates a cutting point p in the current node data, and the cutting point p is generated between the maximum value and the minimum value of the designated dimension in the current node data; a hyperplane is generated according to the cut point p, and then the current node data space is divided into 2 subspaces, namely: a. placing the data smaller than the cutting point p in the specified dimension at the left child node of the current node, and placing the data larger than or equal to the cutting point p at the right child node of the current node; b. and then recursion step a is carried out in the child nodes, and new child nodes are continuously constructed until only one piece of data in the child nodes cannot be cut any more or the child nodes reach a defined height.
The eigen-coefficient calculation is an eigen-coefficient algorithm based on L1 regularization, where the loss function is:
whereinCoefficient of all features, xiRepresents the input of the ith sample, yiRepresents the output of the ith sample and,is a regular term of L1.
The feature coefficient calculation is based on a weighted mean square error reduction algorithm, and the importance of each feature is calculated, so that feature selection is performed according to the feature importance value:
wherein x represents the current node characteristic, y and z represent the leaf node characteristic of the current node, and N (x), N (y), N (z) represent the sample numbers of the nodes x, y, z.
The feature coefficient calculation is to calculate the importance of each feature based on the number of the segmentation nodes of the features in all the trees and an algorithm, so as to select the features according to the value of the feature importance:
where x represents the current node feature, Treei(x) Representing the number of times the current node feature is found on the ith tree.
In step 4, the regressor module obtains a test sample from the training sample database, and performs model training and verification of the relative error of the prediction result according to the candidate features finally selected by the feature selection module.
According to the method, the data characteristic dimensions are enriched through characteristic engineering, the key characteristics are rapidly and accurately screened through characteristic selection, the robustness of fault prediction is enhanced through multi-model fusion, the fault time prediction precision under a small sample data set is greatly improved, the prediction of a fault occurrence time under the small sample by a fault prediction system is met, and meanwhile, the interference of redundant characteristics on a model is greatly reduced.
Drawings
Fig. 1 is a block diagram of a failure prediction system according to the present invention.
Fig. 2 is a flow chart of the warning method according to the present invention.
Detailed Description
The following describes embodiments of the present invention in detail with reference to the drawings, which are implemented on the premise of the technical solution of the present invention, and detailed embodiments and specific operation procedures are provided, but the scope of the present invention is not limited to the following embodiments.
As shown in fig. 1, the fault prediction system based on small samples according to the present invention includes the following modules:
the data cleaning module is used for carrying out deletion filling, smoothing and redundant sample removing on the system data and transmitting the cleaned data to the feature engineering module to extract corresponding feature data;
the characteristic engineering module is used for constructing a first-order difference characteristic of fault occurrence time, sensor data abnormal value identification and abnormal statistical type characteristics;
the characteristic selection module is used for screening out key base characteristics based on an L1 regularization characteristic coefficient algorithm, a weighted mean square error reduction algorithm and a characteristic dividing node number and algorithm in all trees, and inputting the base characteristics into the regressor module;
the regression module trains a regression model and adjusts parameters by adopting a K-fold cross validation and grid search method to ensure that the regression model reaches preset prediction precision, and simultaneously predicts test data by using the trained regression module;
and the model fusion module is used for constructing a final prediction result by adopting a weighted fusion method for the plurality of models trained by the regressor module.
As shown in fig. 2, the early warning method of the fault prediction system based on the small sample of the present invention includes the following steps:
step 1, establishing a training sample database; the training sample database contains historical fault data of equipment to be pre-warned, and comprises: the method comprises the following steps of (1) fault code type (fault code 1, 2., 7) fault code), fault occurrence time, equipment operation parameters (pulse pressure value, environment temperature, sensor signal state and motor rotating speed), equipment state parameters (starting time, working time, historical maintenance times and component replacement information), equipment number, equipment type, equipment geographic information (longitude and latitude, province, city, county and district and geographic environment) and equipment installation time;
step 2, carrying out deletion filling on system data through a data cleaning module, specifically, filling by adopting a characteristic mean value aiming at numerical characteristics, filling by adopting a field expressing a vacancy value aiming at discrete characteristics, and filling by adopting an 'unknown' field; smoothing, namely smoothing the interference of the noise data in a log logarithm mode; removing redundant samples, calculating correlation coefficients among the features, screening feature data according to the correlation coefficients, and transmitting the cleaned data to a feature engineering module to extract corresponding feature data;
step 3, the feature engineering module constructs a first-order difference feature of the fault occurrence time of each device according to the feature data, and identifies abnormal values of device operation parameters (pulse pressure value, environment temperature, sensor signal state and motor rotating speed) and abnormal statistical type features by adopting an Isolate Forest algorithm;
step 4, screening out key base characteristics by adopting a characteristic selection module based on an L1 regularization characteristic coefficient algorithm, a weighted mean square error reduction algorithm and a characteristic dividing node number and algorithm in all trees, and inputting the base characteristics into a regressor module;
and 5, the regressor module respectively uses an LASSO regression model, a GBDT regression model or an XGBOOST regression model to obtain training samples from a training sample database, the training samples pass through the data cleaning module, the feature engineering module and the feature selection module, the regression model is trained and parameters are adjusted by adopting a K-fold cross validation and grid search method, the model achieves preset prediction accuracy, and meanwhile the trained regressor module is used for predicting the test data.
And 6, constructing a final prediction result by adopting a weighted fusion method for the plurality of models trained by the regressor module through a model fusion module.
In step 3, the characteristic engineering model identifies the abnormal values of the sensor data based on the Isolate Forest algorithm, so that various abnormal values of the sensor are counted and used as statistical characteristics, then the abnormal value data of the sensor are scored based on the maximum entropy algorithm, comprehensive scores of various samples are obtained, and finally a sliding window method is adopted to construct first-order difference characteristics of fault occurrence time.
The Isolate Forest algorithm calculates the number of layers of each sample in each binary tree by traversing a plurality of binary trees, and calculates the average height, so as to judge whether the sample is an abnormal sampling point according to a preset threshold value.
The Isolate Forest algorithm randomly selects a plurality of point samples from training data to be used as sub-samples, and the sub-samples are put into a root node of a tree.
The Isolate Forest algorithm randomly designates a dimension, randomly generates a cutting point p in the current node data, and the cutting point p is generated between the maximum value and the minimum value of the designated dimension in the current node data; a hyperplane is generated according to the cut point p, and then the current node data space is divided into 2 subspaces, namely: a. placing the data smaller than the cutting point p in the specified dimension at the left child node of the current node, and placing the data larger than or equal to the cutting point p at the right child node of the current node; b. and then recursion step a is carried out in the child nodes, and new child nodes are continuously constructed until only one piece of data in the child nodes cannot be cut any more or the child nodes reach a defined height.
The eigen-coefficient calculation is an eigen-coefficient algorithm based on L1 regularization, where the loss function is:
whereinCoefficient of all features, xiRepresents the input of the ith sample, yiRepresenting the output of the ith sample.
The feature coefficient calculation is based on a weighted mean square error reduction algorithm, and the importance of each feature is calculated, so that feature selection is performed according to the feature importance value:
wherein x represents the current node characteristic, y and z represent the leaf node characteristic of the current node, and N (x), N (y), N (z) represent the sample numbers of the nodes x, y, z.
The feature coefficient calculation is to calculate the importance of each feature based on the number of the segmentation nodes of the features in all the trees and an algorithm, so as to select the features according to the value of the feature importance:
where x represents the current node feature, Treei(x) Representing the number of times the current node feature is found on the ith tree.
In step 4, the regressor module obtains a test sample from the training sample database, and performs model training and verification of the relative error of the prediction result according to the candidate features finally selected by the feature selection module.
When the method is applied to different equipment and different fault codes, the performance comparison of the models with and without feature selection is shown in table 1.
TABLE 1
As can be seen from table 1, when data is input into the LASSO model, the GBDT model, and the XGBOOST model and applied to training of different devices and different fault codes, after feature selection is performed on the input data, a relative error of 94% of the models is smaller than that of the models without feature selection performed on the data in the test set. This shows that the feature selection method adopted by the invention can improve the accuracy of the model.
When the method is applied to different equipment and different sample quantities, the model fusion results after feature selection are compared and shown in table 2.
TABLE 2
As can be seen from table 2, after feature selection, in the test set, under all different devices and different fault code combinations, the relative error of 73% of the weighted fusion model is smaller than the respective relative error of the corresponding combination on the three base models (LASSO model, GBDT model, XGBOOST model). This shows that the accuracy of the model to the prediction result is greatly improved by the method of the weighted fusion model adopted by the invention.
Claims (10)
1. A fault prediction system based on small samples is characterized in that: the system comprises the following modules:
the data cleaning module is used for carrying out deletion filling, smoothing and redundant sample removing on the system data and transmitting the cleaned data to the feature engineering module to extract corresponding feature data;
the characteristic engineering module is used for constructing a first-order difference characteristic of fault occurrence time, sensor data abnormal value identification and abnormal statistical type characteristics;
the characteristic selection module is used for screening out key base characteristics based on an L1 regularization characteristic coefficient algorithm, a weighted mean square error reduction algorithm and a characteristic dividing node number and algorithm in all trees, and inputting the base characteristics into the regressor module;
the regression module trains a regression model and adjusts parameters by adopting a K-fold cross validation and grid search method to ensure that the regression model reaches preset prediction precision, and simultaneously predicts test data by using the trained regression module;
and the model fusion module is used for constructing a final prediction result by adopting a weighted fusion method for the plurality of models trained by the regressor module.
2. The early warning method of the fault prediction system based on the small sample as claimed in claim 1, wherein: the method comprises the following steps:
step 1, establishing a training sample database; the training sample database contains historical fault data of equipment to be pre-warned, and comprises: the method comprises the following steps of (1) fault code type, fault occurrence time, equipment operation parameters, equipment state parameters, equipment numbers, equipment types, equipment geographic information and equipment installation time;
step 2, performing deletion filling, smoothing and redundant sample elimination on system data through a data cleaning module, and transmitting the cleaned data to a feature engineering module to extract corresponding feature data;
step 3, the characteristic engineering module constructs a first-order difference characteristic of fault occurrence time, abnormal value identification of sensor data and abnormal statistical type characteristic through the characteristic data;
step 4, screening out key base characteristics by adopting a characteristic selection module based on an L1 regularization characteristic coefficient algorithm, a weighted mean square error reduction algorithm and a characteristic dividing node number and algorithm in all trees, and inputting the base characteristics into a regressor module;
step 5, the regressor module trains a regression model and adjusts parameters by adopting a K-fold cross validation and grid search method, so that the regression model reaches preset prediction precision, and the trained regressor module is used for predicting test data;
and 6, constructing a final prediction result by adopting a weighted fusion method for the plurality of models trained by the regressor module through a model fusion module.
3. The method for early warning of the fault prediction system based on the small sample as claimed in claim 2, wherein: in step 3, the characteristic engineering model identifies the abnormal values of the sensor data based on the Isolate Forest algorithm, so that various abnormal values of the sensor are counted and used as statistical characteristics, then the abnormal value data of the sensor are scored based on the maximum entropy algorithm, comprehensive scores of various samples are obtained, and finally a sliding window method is adopted to construct first-order difference characteristics of fault occurrence time.
4. The early warning method of the fault prediction system based on the small samples as claimed in claim 3, wherein: the Isolate Forest algorithm calculates the number of layers of each sample in each binary tree by traversing a plurality of binary trees, and calculates the average height, so as to judge whether the sample is an abnormal sampling point according to a preset threshold value.
5. The early warning method of the fault prediction system based on the small samples as claimed in claim 3, wherein: the Isolate Forest algorithm randomly selects a plurality of point samples from training data to be used as sub-samples, and the sub-samples are put into a root node of a tree.
6. The early warning method of the fault prediction system based on the small samples as claimed in claim 3, wherein: the Isolate Forest algorithm randomly designates a dimension, randomly generates a cutting point p in the current node data, and the cutting point p is generated between the maximum value and the minimum value of the designated dimension in the current node data; a hyperplane is generated according to the cut point p, and then the current node data space is divided into 2 subspaces, namely: a. placing the data smaller than the cutting point p in the specified dimension at the left child node of the current node, and placing the data larger than or equal to the cutting point p at the right child node of the current node; b. and then recursion step a is carried out in the child nodes, and new child nodes are continuously constructed until only one piece of data in the child nodes cannot be cut any more or the child nodes reach a defined height.
7. The method for early warning of the fault prediction system based on the small sample as claimed in claim 2, wherein: the eigen-coefficient calculation is an eigen-coefficient algorithm based on L1 regularization, where the loss function is:
8. The method for early warning of the fault prediction system based on the small sample as claimed in claim 2, wherein: the feature coefficient calculation is based on a weighted mean square error reduction algorithm, and the importance of each feature is calculated, so that feature selection is performed according to the feature importance value:
wherein x represents the current node characteristic, y and z represent the leaf node characteristic of the current node, and N (x), N (y), N (z) represent the sample numbers of the nodes x, y, z.
9. The method for early warning of the fault prediction system based on the small sample as claimed in claim 2, wherein: the feature coefficient calculation is to calculate the importance of each feature based on the number of the segmentation nodes of the features in all the trees and an algorithm, so as to select the features according to the value of the feature importance:
where x represents the current node feature, Treei(x) Representing the number of times the current node feature is found on the ith tree.
10. The method for early warning of the fault prediction system based on the small sample as claimed in claim 2, wherein: in step 4, the regressor module obtains a test sample from the training sample database, and performs model training and verification of the relative error of the prediction result according to the candidate features finally selected by the feature selection module.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010176870.XA CN111275136B (en) | 2020-03-13 | 2020-03-13 | Fault prediction system based on small sample and early warning method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010176870.XA CN111275136B (en) | 2020-03-13 | 2020-03-13 | Fault prediction system based on small sample and early warning method thereof |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111275136A true CN111275136A (en) | 2020-06-12 |
CN111275136B CN111275136B (en) | 2023-05-02 |
Family
ID=70999638
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010176870.XA Active CN111275136B (en) | 2020-03-13 | 2020-03-13 | Fault prediction system based on small sample and early warning method thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111275136B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112905572A (en) * | 2021-01-29 | 2021-06-04 | 铁道警察学院 | Data anomaly information studying and judging model and method |
CN114004425A (en) * | 2021-12-29 | 2022-02-01 | 北京京东振世信息技术有限公司 | Article circulation information prediction model generation method, information generation method and device |
CN114511113A (en) * | 2022-01-28 | 2022-05-17 | 国网上海市电力公司 | Fault positioning method and system for TTU (time to Unit) equipment in transformer area, electronic equipment and storage medium |
CN117753795A (en) * | 2024-02-07 | 2024-03-26 | 东北大学 | feedforward control method for hot rolled products with multiple steel types and specifications |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140214733A1 (en) * | 2013-01-30 | 2014-07-31 | Siemens Aktiengesellschaft | Method And Apparatus For Deriving Diagnostic Data About A Technical System |
WO2015158198A1 (en) * | 2014-04-17 | 2015-10-22 | 北京泰乐德信息技术有限公司 | Fault recognition method and system based on neural network self-learning |
CN110417011A (en) * | 2019-07-31 | 2019-11-05 | 三峡大学 | A kind of online dynamic secure estimation method based on mutual information Yu iteration random forest |
CN110806743A (en) * | 2019-12-05 | 2020-02-18 | 成都天玙兴科技有限公司 | Equipment fault detection and early warning system and method based on artificial intelligence |
-
2020
- 2020-03-13 CN CN202010176870.XA patent/CN111275136B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140214733A1 (en) * | 2013-01-30 | 2014-07-31 | Siemens Aktiengesellschaft | Method And Apparatus For Deriving Diagnostic Data About A Technical System |
WO2015158198A1 (en) * | 2014-04-17 | 2015-10-22 | 北京泰乐德信息技术有限公司 | Fault recognition method and system based on neural network self-learning |
CN110417011A (en) * | 2019-07-31 | 2019-11-05 | 三峡大学 | A kind of online dynamic secure estimation method based on mutual information Yu iteration random forest |
CN110806743A (en) * | 2019-12-05 | 2020-02-18 | 成都天玙兴科技有限公司 | Equipment fault detection and early warning system and method based on artificial intelligence |
Non-Patent Citations (1)
Title |
---|
邓森;景博;周宏亮;朱海鹏;刘小平;: "一种改进的自适应增强-支持向量回归机的故障预测方法" * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112905572A (en) * | 2021-01-29 | 2021-06-04 | 铁道警察学院 | Data anomaly information studying and judging model and method |
CN114004425A (en) * | 2021-12-29 | 2022-02-01 | 北京京东振世信息技术有限公司 | Article circulation information prediction model generation method, information generation method and device |
CN114004425B (en) * | 2021-12-29 | 2022-06-07 | 北京京东振世信息技术有限公司 | Article circulation information prediction model generation method, information generation method and device |
CN114511113A (en) * | 2022-01-28 | 2022-05-17 | 国网上海市电力公司 | Fault positioning method and system for TTU (time to Unit) equipment in transformer area, electronic equipment and storage medium |
CN117753795A (en) * | 2024-02-07 | 2024-03-26 | 东北大学 | feedforward control method for hot rolled products with multiple steel types and specifications |
CN117753795B (en) * | 2024-02-07 | 2024-05-31 | 东北大学 | Feedforward control method for hot rolled products with multiple steel types and specifications |
Also Published As
Publication number | Publication date |
---|---|
CN111275136B (en) | 2023-05-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111275136B (en) | Fault prediction system based on small sample and early warning method thereof | |
CN109102005B (en) | Small sample deep learning method based on shallow model knowledge migration | |
CN111222549B (en) | Unmanned aerial vehicle fault prediction method based on deep neural network | |
CN106168799B (en) | A method of batteries of electric automobile predictive maintenance is carried out based on big data machine learning | |
CN105279365B (en) | For the method for the sample for learning abnormality detection | |
KR102274389B1 (en) | Method for building anomaly pattern detection model using sensor data, apparatus and method for detecting anomaly using the same | |
CN111507376A (en) | Single index abnormality detection method based on fusion of multiple unsupervised methods | |
CN113378990B (en) | Flow data anomaly detection method based on deep learning | |
CN111325410B (en) | Universal fault early warning system based on sample distribution and early warning method thereof | |
CN115293280A (en) | Power equipment system anomaly detection method based on space-time feature segmentation reconstruction | |
CN111738308A (en) | Dynamic threshold detection method for monitoring index based on clustering and semi-supervised learning | |
CN114201374A (en) | Operation and maintenance time sequence data anomaly detection method and system based on hybrid machine learning | |
CN114385614A (en) | Water quality early warning method based on Informmer model | |
CN117421684A (en) | Abnormal data monitoring and analyzing method based on data mining and neural network | |
CN113255591A (en) | Bearing fault diagnosis method based on random forest and fusion characteristics | |
CN116915582A (en) | Diagnosis and analysis method and device for fault root cause of communication terminal | |
CN115758908A (en) | Alarm online prediction method under alarm flooding condition based on deep learning | |
CN114841412A (en) | Method for predicting pH value of sea cucumber growing water | |
CN114529067A (en) | Method for performing predictive maintenance on electric vehicle battery based on big data machine learning | |
CN114048546A (en) | Graph convolution network and unsupervised domain self-adaptive prediction method for residual service life of aircraft engine | |
CN117095247A (en) | Numerical control machining-based machining gesture operation optimization method, system and medium | |
CN117216713A (en) | Fault delimiting method, device, electronic equipment and storage medium | |
CN111934903A (en) | Docker container fault intelligent prediction method based on time sequence evolution genes | |
CN111885146B (en) | Data transmission method for industrial data cloud service platform | |
CN115392710A (en) | Wind turbine generator operation decision method and system based on data filtering |
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 |