CN116561528B - RUL prediction method of rotary machine - Google Patents
RUL prediction method of rotary machine Download PDFInfo
- Publication number
- CN116561528B CN116561528B CN202310507580.2A CN202310507580A CN116561528B CN 116561528 B CN116561528 B CN 116561528B CN 202310507580 A CN202310507580 A CN 202310507580A CN 116561528 B CN116561528 B CN 116561528B
- Authority
- CN
- China
- Prior art keywords
- rul
- representing
- feature
- model
- lightgbm
- 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
- 238000000034 method Methods 0.000 title claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 15
- 238000012360 testing method Methods 0.000 claims abstract description 13
- 238000010606 normalization Methods 0.000 claims abstract description 7
- 238000012545 processing Methods 0.000 claims abstract description 7
- 238000002372 labelling Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 12
- 241000287127 Passeridae Species 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 7
- 230000003247 decreasing effect Effects 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 4
- 238000003066 decision tree Methods 0.000 claims description 3
- 238000009826 distribution Methods 0.000 claims description 3
- 230000006835 compression Effects 0.000 claims description 2
- 238000007906 compression Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims 1
- 238000012544 monitoring process Methods 0.000 abstract description 2
- 238000005065 mining Methods 0.000 abstract 1
- 238000013527 convolutional neural network Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 5
- 230000015556 catabolic process Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000006731 degradation reaction Methods 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002431 foraging effect Effects 0.000 description 2
- 238000009776 industrial production Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000004393 prognosis Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- 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
- G06F18/21—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
- 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
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- 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
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the field of monitoring of rotary machinery, and particularly relates to an RUL prediction method of the rotary machinery, which comprises the following steps: carrying out feature selection on sensor data of the rotary mechanical equipment through a Spearman correlation coefficient, carrying out batch standardization and normalization processing on features with higher correlation with a time sequence, and dividing a training set and a testing set and labeling RUL labels of the sequences; building a SE-CNN model to extract features; optimizing a LightGBM model through an ISSA algorithm and training the optimized model; and inputting the divided test set data into a LightGBM model for prediction to obtain the RUL value. According to the invention, SE-CNN is combined with the LightGBM, so that the problems of insufficient processing capacity of high-dimensional complex data and incomplete mining of data characteristics are solved, and the super-parameters of the LightGBM are optimized through ISSA, so that the accuracy of predicting the residual service life of the lightning tGBM is effectively improved.
Description
Technical Field
The invention belongs to the field of monitoring of rotary machinery, and particularly relates to an RUL prediction method of the rotary machinery.
Background
The rotary machine bears the main modern industrial production task, ensures safe and normal work for the research of the rotary machine, and has very important practical significance for industrial production in China. For how to effectively maintain the safe operation of the rotary machine, some domestic and foreign scholars research and put forward a prediction maintenance strategy combining research results in the fields of information technology, artificial intelligence and the like, namely fault prediction and health management (Prognosis and Health Management, PHM), and prediction of the residual service life (Remaining Useful Life, RUL) is a core link thereof. The RUL prediction method based on data driving does not depend on the knowledge of the related field of the equipment system, solves the problem of difficult modeling of the complex equipment system based on a physical model method, and becomes a main stream method in the RUL prediction field under the background of continuous development of a big data background. Ma Hailong and the like utilize the PCA method to fuse the characteristics of root mean square value, peak value, wavelet entropy and the like of the vibration signals to represent the degradation state of the bearing.
The method adopts artificial feature extraction, the modeling mode is complex, and part of the features are easily ignored by the artificial extraction. In recent years, deep learning has been increasingly hot in the field of prediction and health management, and Zhu Jun et al have combined wavelet transform and CNN to predict the RUL of a bearing, first extracting time-frequency features using wavelet transform, and then estimating the RUL using multiscale CNN. Although this approach has a good predictive effect in combination with deep learning, it is still lacking in terms of high dimensionality and complexity relative to the device sensor data.
With the continued development of decision tree integration methods, extreme gradient enhancement algorithms (Extreme Gradient Boosting, XGBoost) have achieved good results in many machine learning challenges. The LightGBM is an improvement on XGBoost in time complexity and prediction precision, and has more excellent performance on high-dimensional complex data processing of a device sensor. Song Hailong et al employ a LightGBM optimized based on a time window feature derived model for residual life prediction (Song Hailong, dawn, gujiang, zhao Qinghe. LightGBM based aeroengine residual life prediction [ J ]. Modern computer, 2021,27 (35): 47-52.). Although the method has higher prediction precision, the depth excavation of the data features is not comprehensive enough, and the super parameters of the LightGBM are difficult to determine and are required to be further optimized and promoted.
Disclosure of Invention
In order to solve the technical problems, the invention provides a RUL prediction method of a rotary machine, which comprises the following steps:
s1: acquiring sensor data of the rotating mechanical equipment;
s2: carrying out feature selection on sensor data through a Spearman correlation coefficient, carrying out batch standardization and normalization processing on features with higher correlation with a time sequence, and dividing a training set and a test set and labeling RUL labels of the sequences;
s3: constructing an SE-CNN model to extract characteristics of the processed data;
s4: optimizing the number of cotyledons, the maximum depth and the learning rate in the LightGBM model through an ISSA algorithm, inputting data in a training set into the optimized LightGBM model to train according to the RUL label marked by the optimized LightGBM model, and completing training of the model when the loss function value of the LightGBM model is minimum;
s5: and inputting the divided test set data into the trained LightGBM model, and outputting the predicted RUL value.
The invention has the beneficial effects that:
according to the invention, SE-CNN with strong data feature depth extraction capability is combined with the LightGBM with good high-dimensional complex data prediction effect, so that the problem that a deep learning method is insufficient in high-dimensional complex data processing capability is solved, the problem that the data feature is not fully mined by using only the LightGBM, and meanwhile, important super parameters of the LightGBM are optimized by ISSA, so that the accuracy of residual service life prediction is effectively improved.
Drawings
FIG. 1 is an overall flow chart of the present invention;
FIG. 2 is a graph showing the characteristics of an engine with a pre-processed C-MAPSS according to the present invention;
FIG. 3 is a diagram of the SE-CNN structure of the present invention;
fig. 4 is a diagram of the ISSA-LightGBM structure of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in the overall flowchart of fig. 1, the RUL prediction method for a rotary machine includes:
s1: acquiring sensor data of the rotating mechanical equipment;
s2: carrying out feature selection on sensor data through a Spearman correlation coefficient, carrying out batch standardization and normalization processing on features with higher correlation with a time sequence, and dividing a training set and a test set and labeling RUL labels of the sequences;
s3: constructing an SE-CNN model to extract characteristics of the processed data;
s4: optimizing the number of cotyledons, the maximum depth and the learning rate in the LightGBM model through an ISSA algorithm, inputting data in a training set into the optimized LightGBM model to train according to the RUL label marked by the optimized LightGBM model, and completing training of the model when the loss function value of the LightGBM model is minimum;
s5: and inputting the divided test set data into the trained LightGBM model, and outputting the predicted RUL value.
Taking one engine in the C-MAPSS aeroengine data set as an example, firstly calculating 24 Spearman correlation coefficients capable of reflecting equipment degradation information parameters, selecting 14 features with higher correlation degree with a time sequence, and constructing a time sequence feature sequence. The Spearman correlation coefficient is calculated as follows:
wherein X is n A time-based sequence is represented and,mean value of time series>A sequence representing the selected feature,/->Mean values representing feature sequences (n=1, 2, … N); ρ m Represents the mth characteristic sequence->With time sequence X n Is used for the correlation coefficient of the (c).
Batch normalization and normalization of the selected feature sequences are performed as shown in formula (2) and formula (3), respectively:
wherein Y' represents data after batch normalization; μ represents the mean of the feature sequence; sigma represents the standard deviation of the feature sequence; y is Y * Representing the value of normalized data, Y max And Y min Respectively a maximum value and a minimum value in each sampling period.
Dividing the processed features into a training set and a test set, and using the formula (4) as the RUL label of the monotonically decreasing feature sequence and the formula (5) as the RUL label of the monotonically increasing feature sequence:
wherein x is td And x tr RUL labels respectively representing a monotonically decreasing feature sequence and a monotonically increasing feature sequence at a t-th sampling point, and l represents a full life cycle.
The feature data of a certain engine of the C-MAPSS after pretreatment is shown in fig. 2, and the selected features have higher correlation with time sequence, so that the degradation information of the rotary machine can be reflected.
As shown in FIG. 3, "SE-CNN model building", is mainly composed of SE modules and basic CNN modules. Firstly, the preprocessed features are input into a CNN module to perform feature extraction, a and b are the feature number and the sampling point number of original features respectively, the extraction result is input into an SE module, and the feature weight is learned according to a loss function through a network, so that the weight of an effective feature map is larger, and the weight of an ineffective or less effective feature map is smaller, thereby enabling a training model to achieve a better result.
The SE module mainly comprises the following three operations:
(1) Squeeze: feature compression is carried out along the space dimension, each two-dimensional feature channel is changed into a real number, the output dimension is matched with the input feature channel number, and the expression is as follows:
wherein z is c An output representing a Squeeze operation; f (F) sq Representing a Squeeze operation; u (u) c A c-th feature map representing the input matrix; w represents the width of the input feature map; h represents the height of the input feature map; i and j are the locations of the feature map across the width and height, respectively.
(2) The specification: a mechanism similar to gates in a recurrent neural network generates weights for each characteristic channel expressed as:
S′＝F ex (z,W)＝σ(W 2 δ(W 1 z)) (7)
s' represents the output of the specification operation, namely the weight matrix of each feature map; f (F) ex Representing an expression operation; z is the result of the previous step of Squeeze operation; w represents a weight; w (W) 1 ，W 2 Are weight matrixes; sigma represents Sigmoid function transformation; delta represents the Relu function transformation.
(3) Reweight: the output weight of the specification is weighted to the previous feature channel by channel through multiplication, and the recalibration of the original feature in the channel dimension is completed, wherein the expression is as follows:
wherein,representing a characteristic diagram after being subjected to a weight operation; f (F) scale Representing a Reweight operation; s is S c And the output weight of the c-th characteristic diagram is represented.
Features extracted by SE-CNN are resized by flattening layers as inputs to the ISSA-LightGBM model.
As shown in the ISSA-LightGBM structure diagram of FIG. 4, the output of SE-CNN is used as the input of the model, the number of cotyledons, the maximum depth and the learning rate in the ISSA optimized LightGBM model are first performed, and then RUL prediction is performed.
SSA forms an explorator-follower-alerter model by deducting the foraging and early warning actions of sparrow groups in the face of hazards. The explorer generally has better fitness and carries more dominant information, and the search range and the foraging position of the explorer are better. The traditional seeker position update formula is as follows:
where M is the current iteration number and M is the total iteration number.Representing the position information of the kth sparrow in the d-th dimension at the mth iteration,/>Is a random number, W is 0,1]Representing early warning value, S epsilon [0.5,1 ]]Representing the security value, X is a random number subject to a standard normal distribution, and Y is a row, multi-dimensional, all-matrix.
For traditional seekers, sparrows are in safe areas when W < S, and the degree of discovery should be enhanced to find the optimum. However, there is a lack of bridge in the original formula linking the updated location and the current fitness value. The invention thus proposes an exploration factor (Exploration factor, ef) for improving the position of an explorer within a safety area, expressed in the following manner:
wherein f B The optimal fitness value is the current optimal fitness value; f (f) W The current worst fitness value;
the improved seeker position updating formula is as follows:
wherein,representing the updated position information of the kth sparrow in the d dimension, M representing the current iteration number, M representing the total iteration number, ef representing the exploration factor,>representing the position information of the kth sparrow in the d-th dimension at the mth iteration,/>Is expressed in the interval (0, 1)]W represents the random number in interval [0,1 ]]S represents the early warning value in interval [0.5,1 ]]X represents a random number subject to a standard normal distribution, and Y represents a row of a multidimensional all-one matrix.
Flow of ISSA optimization of cotyledon number, maximum depth and learning rate in LightGBM model: firstly, ISSA and LightGBM parameters are initialized, and output results of SE-CNN are divided into a training set and a test set input model. And secondly, calculating an initial fitness value, updating the positions of a follower and an early warning person in the ISSA, and calculating the latest fitness value and the optimal position. Finally judging whether the maximum iteration times are reached, if so, outputting optimal parameters to update the LightGBM model; if not, updating the follower position again for optimizing.
The loss function of the LightGBM is as follows:
wherein x is i For input of the model, i.e. data in the test set, y i Is the true value of the RUL,for the predicted value of RUL, K is the total number of decision trees, F is oneGroup regression tree, f k The score value of the leaf node is l, the trained loss function is l, and omega is a regularization function.
After the training of the above steps, the ISSA-LightGBM model already has the ability to predict the rotary machine RUL. And inputting the divided test set data into a model, outputting a predicted RUL value, and carrying out result evaluation to finally realize a rotary machine residual service life prediction method of SE-CNN and ISSA-LightGBM.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (5)
1. A RUL prediction method for a rotary machine, comprising:
s1: acquiring sensor data of the rotating mechanical equipment;
s2: the sensor data is subjected to feature selection through a Spearman correlation coefficient, features with higher correlation with a time sequence are subjected to standardization and normalization processing, and a training set and a testing set are divided and RUL labels of the sequences are marked;
s3: constructing an SE-CNN model to extract characteristics of the processed data;
s4: optimizing the number of cotyledons, the maximum depth and the learning rate in the LightGBM model through an ISSA algorithm, inputting data in a training set into the optimized LightGBM model to train according to the RUL label marked by the optimized LightGBM model, and completing training of the model when the loss function value of the LightGBM model is minimum;
the ISSA algorithm comprises: a heuristic factor is proposed to improve the position of the seeker within the secure area;
the exploration factor includes:
wherein Ef represents the exploration factor, f B Represents the current optimal fitness value, f W Representing a current worst fitness value, M representing a current iteration number, and M representing a total iteration number;
the improved seeker position update formula comprises:
wherein,indicating the updated position information of the kth sparrow in the d-th dimension +.>Representing the position information of the kth sparrow in the d-th dimension at the mth iteration,/>Is expressed in the interval (0, 1)]W represents the random number in interval [0,1 ]]S represents the early warning value in interval [0.5,1 ]]X represents a random number subject to a standard normal distribution, Y represents a row of a multidimensional all-matrix;
a loss function of the LightGBM model, comprising:
wherein y is i Representing the true value of the RUL,representing the predicted value of RUL,>k represents the total number of decision trees, f k Representing leavesScore value, x, of child node i Representing data in a test set, F representing a set of regression trees, l representing a trained loss function, Ω representing a regularized function;
s5: and inputting the divided test set data into the trained LightGBM model, and outputting the predicted RUL value.
2. The RUL prediction method of a rotary machine according to claim 1, wherein said Spearman correlation coefficient calculation comprises:
wherein ρ is m Representing the mth signature sequenceWith time sequence X n Correlation coefficient, X n Representing a time sequence,/->Mean value of time series>Represents the mth signature sequence,/->Representing the mean of the feature sequences, n=1, 2, … N, N representing the number of sequences.
3. The RUL prediction method of a rotary machine according to claim 1, wherein labeling the RUL tag of the sequence comprises:
the RUL label comprises an RUL label of a monotonically decreasing characteristic sequence and an RUL label of a monotonically increasing characteristic sequence;
the RUL label of the monotonically decreasing feature sequence:
the RUL tag of the monotonically increasing feature sequence:
wherein x is td And x tr RUL labels respectively representing a monotonically decreasing feature sequence and a monotonically increasing feature sequence at a t-th sampling point, and l represents a full life cycle.
4. The RUL prediction method of a rotary machine according to claim 1, wherein the SE-CNN model comprises: SE module, basic CNN module, and leveling layer;
the preprocessed features are input into a CNN module for feature extraction, the extraction result is input into an SE module, and feature weights are learned, so that the effective feature images have larger weights, the ineffective or less effective feature images have smaller weights, and the size is adjusted through a flattening layer.
5. The RUL prediction method of a rotary machine of claim 4, wherein the SE module learns a feature weight process comprising:
performing feature compression along the space dimension, changing each two-dimensional feature channel into a real number, and keeping the output dimension matched with the input feature channel number;
generating a weight for each characteristic channel;
and weighting the generated weights to the characteristics of the input SE module channel by channel through multiplication, and completing recalibration of the original characteristics in the channel dimension.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310507580.2A CN116561528B (en) | 2023-05-08 | 2023-05-08 | RUL prediction method of rotary machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310507580.2A CN116561528B (en) | 2023-05-08 | 2023-05-08 | RUL prediction method of rotary machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116561528A CN116561528A (en) | 2023-08-08 |
| CN116561528B true CN116561528B (en) | 2024-03-01 |
Family
ID=87499505
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310507580.2A Active CN116561528B (en) | 2023-05-08 | 2023-05-08 | RUL prediction method of rotary machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116561528B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013152655A (en) * | 2012-01-26 | 2013-08-08 | Hitachi Ltd | Abnormality diagnostic method and health management method for plant or facility |
| CN112149316A (en) * | 2019-11-04 | 2020-12-29 | 中国人民解放军国防科技大学 | Aero-engine residual life prediction method based on improved CNN model |
| CN113486585A (en) * | 2021-07-06 | 2021-10-08 | 新智数字科技有限公司 | Method and device for predicting remaining service life of equipment, electronic equipment and storage medium |
| CN113516316A (en) * | 2021-07-29 | 2021-10-19 | 昆明理工大学 | Attention-GRU short-term load prediction method based on sparrow search optimization |
| CN115236520A (en) * | 2022-07-20 | 2022-10-25 | 山东工商学院 | Battery remaining service life prediction method and system based on ISSA-LSTM algorithm |
| US11527786B1 (en) * | 2022-03-28 | 2022-12-13 | Eatron Technologies Ltd. | Systems and methods for predicting remaining useful life in batteries and assets |
| CN116010884A (en) * | 2022-11-21 | 2023-04-25 | 杭州电力设备制造有限公司 | Fault diagnosis method of SSA-LightGBM oil-immersed transformer based on principal component analysis |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113723007B (en) * | 2021-09-08 | 2023-09-15 | 重庆邮电大学 | Equipment residual life prediction method based on DRSN and sparrow search optimization |
-
2023
- 2023-05-08 CN CN202310507580.2A patent/CN116561528B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013152655A (en) * | 2012-01-26 | 2013-08-08 | Hitachi Ltd | Abnormality diagnostic method and health management method for plant or facility |
| CN112149316A (en) * | 2019-11-04 | 2020-12-29 | 中国人民解放军国防科技大学 | Aero-engine residual life prediction method based on improved CNN model |
| CN113486585A (en) * | 2021-07-06 | 2021-10-08 | 新智数字科技有限公司 | Method and device for predicting remaining service life of equipment, electronic equipment and storage medium |
| CN113516316A (en) * | 2021-07-29 | 2021-10-19 | 昆明理工大学 | Attention-GRU short-term load prediction method based on sparrow search optimization |
| US11527786B1 (en) * | 2022-03-28 | 2022-12-13 | Eatron Technologies Ltd. | Systems and methods for predicting remaining useful life in batteries and assets |
| CN115236520A (en) * | 2022-07-20 | 2022-10-25 | 山东工商学院 | Battery remaining service life prediction method and system based on ISSA-LSTM algorithm |
| CN116010884A (en) * | 2022-11-21 | 2023-04-25 | 杭州电力设备制造有限公司 | Fault diagnosis method of SSA-LightGBM oil-immersed transformer based on principal component analysis |
Non-Patent Citations (2)
| Title |
|---|
| Short-term load forecasting based on CEEMDAN-FE-ISSA-LightGBM model;Zhihao Li.et.;《Frontiers in Energy Research 》;1-14 * |
| 绝缘子污秽等级的图像识别关键技术研究;刘芃良;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》(第12期);C042-101 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116561528A (en) | 2023-08-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wu et al. | An adaptive deep transfer learning method for bearing fault diagnosis | |
| Che et al. | Hybrid multimodal fusion with deep learning for rolling bearing fault diagnosis | |
| CN113486578B (en) | Method for predicting residual life of equipment in industrial process | |
| CN115018021B (en) | Machine room abnormity detection method and device based on graph structure and abnormity attention mechanism | |
| Liang et al. | Multi-scale dynamic adaptive residual network for fault diagnosis | |
| CN108596327B (en) | Seismic velocity spectrum artificial intelligence picking method based on deep learning | |
| CN114218872B (en) | DBN-LSTM semi-supervised joint model-based residual service life prediction method | |
| Miao et al. | A novel real-time fault diagnosis method for planetary gearbox using transferable hidden layer | |
| CN115618296A (en) | Dam monitoring time sequence data anomaly detection method based on graph attention network | |
| CN110083125A (en) | A kind of machine tool thermal error modeling method based on deep learning | |
| WO2022010731A1 (en) | Compact representation and time series segment retrieval through deep learning | |
| CN116028876A (en) | Rolling bearing fault diagnosis method based on transfer learning | |
| Qin et al. | RCLSTMNet: A Residual-convolutional-LSTM Neural Network for Forecasting Cutterhead Torque in Shield Machine | |
| CN116796275A (en) | Multi-mode time sequence anomaly detection method for industrial equipment | |
| CN115456044A (en) | Equipment health state assessment method based on knowledge graph multi-set pooling | |
| CN117636477A (en) | Multi-target tracking matching method based on radial basis function fuzzy neural network | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| CN112149896A (en) | Attention mechanism-based mechanical equipment multi-working-condition fault prediction method | |
| CN112560252B (en) | Method for predicting residual life of aeroengine | |
| Liu et al. | Semi-supervised deep learning recognition method for the new classes of faults in wind turbine system | |
| CN114297795A (en) | Mechanical equipment residual life prediction method based on PR-Trans | |
| CN118132934A (en) | Real-time state analysis method and system for machine tool spindle | |
| CN117909881A (en) | Fault diagnosis method and device for multi-source data fusion pumping unit | |
| CN116561528B (en) | RUL prediction method of rotary machine | |
| CN115048873B (en) | Residual service life prediction system for aircraft engine |
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 |