CN113191232B - Electro-hydrostatic actuator fault identification method based on multi-mode homologous features and XGboost model - Google Patents
Electro-hydrostatic actuator fault identification method based on multi-mode homologous features and XGboost model Download PDFInfo
- Publication number
- CN113191232B CN113191232B CN202110431264.2A CN202110431264A CN113191232B CN 113191232 B CN113191232 B CN 113191232B CN 202110431264 A CN202110431264 A CN 202110431264A CN 113191232 B CN113191232 B CN 113191232B
- Authority
- CN
- China
- Prior art keywords
- electro
- xgboost model
- hydrostatic actuator
- mode
- training
- 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
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
- F04B—POSITIVE-DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS
- F04B17/00—Pumps characterised by combination with, or adaptation to, specific driving engines or motors
- F04B17/03—Pumps characterised by combination with, or adaptation to, specific driving engines or motors driven by electric motors
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
- F04B—POSITIVE-DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS
- F04B51/00—Testing machines, pumps, or pumping installations
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- 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
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Mechanical Engineering (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention relates to a fault identification method of an electro-hydrostatic actuator based on multi-mode homologous characteristics and an XGboost model, which comprises the steps of firstly carrying out data set segmentation on vibration signals, pressure signals and current signals of the electro-hydrostatic actuator acquired in a big data era to construct an original sample set, and dividing the original sample set into a training set and a testing set; secondly, multi-modal homologous features, namely a time domain modality, a frequency domain modality, a wavelet packet modality and an improved Hilbert-Huang modality, are respectively extracted from the sample set data, a high-dimensional feature vector with higher robustness is finally constructed, and then an XGboost model is utilized to combine the high-dimensional multi-modal homologous features and the XGboost model, so that a hyper-parameter n of the XGboost model is carried out on a training set trees 、n depth And n learning_rates And fine adjustment is carried out, and finally XGboost model generalization performance evaluation is carried out on a training set based on the optimal XGboost model, and the fault identification of the electro-hydrostatic actuator is realized. The invention can improve the fault identification accuracy of the hydraulic pump of the electro-hydrostatic actuator, and has better robustness and higher speed.
Description
Technical Field
The invention relates to an early fault state monitoring and identifying method for an electro-hydrostatic actuator based on multi-mode homologous features and an XGboost model, in particular to a fault identifying method for the electro-hydrostatic actuator based on multi-mode homologous features and the XGboost model.
Background
An Electro-Hydrostatic Actuator (EHA) is a high-pressure and high-speed execution element with high integration such as a motor, a hydraulic pump, a pressurization oil tank, an actuating cylinder, a controller and the like, and is a main control part of an aerocraft. The flight control device is mainly used for controlling the empennage, the wings, the landing gear, the engine and the flaperon, and realizes the rolling, climbing, yawing, taking off and landing flight actions of the aviation aircraft. Therefore, intelligent health monitoring on EHA deployment is an important means for reducing aircraft accidents and guaranteeing safe operation of the aviation aircraft. The vibration of the EHA, the engine and other operating components of the aircraft can interfere with each other, and due to the compression of fluid and the fluid-solid coupling effect of the pump source and the hydraulic circuit, the vibration response mode of the EHA is more complex compared with that of the traditional rotary mechanical equipment, so that the intelligent health monitoring of the early faults of the components of the electro-hydrostatic actuator is necessary to be carried out, and early warning can be provided for air passengers, so that the workers can take relevant measures to avoid causing great loss of lives and properties.
The domestic research on the electro-hydrostatic actuator mainly focuses on the aspects of complete machine design, manufacture and power analysis, and the research on the aspects of related early failure identification and maintenance is less. The method for intelligently identifying the faults of the electro-hydrostatic actuator mainly comprises the following steps: a fault identification method based on signal processing, a fault identification method based on a model and a fault identification method based on artificial intelligence. Fault identification based on signal processing is overly dependent on expert experience, and therefore application is greatly limited; the model-based method needs to aim at the existing knowledge of a system mathematical model, and a test scheme is specially designed, so that the method is difficult to popularize and apply. At present, the intelligent health monitoring and identification method for the electro-hydrostatic actuator mainly comprises the following steps: artificial neural networks, support vector machines, decision trees, and the like. The existing method for monitoring the state of the electro-hydrostatic actuator has the problems of increased training time, poor model overfitting capability, poor generalization capability and the like of an artificial neural network, a support vector machine and a decision tree model when the data volume is large and the characteristic dimensionality is high in a big data era. Therefore, in order to meet the requirements of intelligent identification of the electro-hydrostatic actuator in the big data era of the mechanical field, the XGboost model widely applied to the field of internet data mining is introduced by the user, the defects of the traditional machine learning model can be effectively overcome, the relation between the running state of the electro-hydrostatic actuator and the multi-signal characteristics can be better mined by combining the multi-mode homologous characteristics, and the intelligent fault identification of the electro-hydrostatic actuator in the big data era is better solved.
Disclosure of Invention
The invention aims to provide a method for identifying faults of an electro-hydrostatic actuator based on multi-mode homologous features and an XGboost model, aiming at solving the problems of the existing technology for monitoring and identifying early fault states of the electro-hydrostatic actuator.
The invention is realized by adopting the following technical scheme:
the method comprises the steps of firstly, carrying out data set segmentation on vibration signals, pressure signals and current signals of the electro-hydrostatic actuator acquired in a big data era to construct an original sample set, and dividing the original sample set into a training set and a testing set; secondly, multi-modal homologous features, namely a time domain mode, a frequency domain mode, a wavelet packet mode and an improved Hilbert yellow mode, are respectively extracted from the sample set data, a high-dimensional feature vector with higher robustness is finally constructed, and then an XGboost model is used for combining the high-dimensional multi-modal homologous features and the XGboost model to perform the hyper-parameter n of the XGboost model on a training set trees 、n depth And n learning_rates And fine adjustment is carried out, finally, XGboost model generalization performance evaluation is carried out on a training set based on the optimal XGboost model, and the electro-hydrostatic actuator fault identification is realized.
In a further development of the invention, the method comprises the following steps:
step 1, collecting data of a motor current of an electro-hydrostatic actuator, a vibration signal of a motor pump shell and a pressure signal of an oil outlet of a hydraulic pump in a normal state and an abnormal state, performing data set segmentation to construct an original sample set, and dividing a training set and a testing set;
step 2, respectively processing the training set and the test set data of the three signals based on time domain, fourier transform, wavelet packet decomposition and improved Hilbert-Huang transform, and performing characteristic extraction of statistical characteristic parameters on the processing result;
step 3, combining and constructing high-dimensional feature vectors for three signal training data set features of the electro-hydrostatic actuator, using the high-dimensional feature vectors as input sample data of XGboost model training, and performing one-hot coding on multiple faults of the electro-hydrostatic actuator to serve as output samples of model training;
step 4, training an XGboost model of grid search by utilizing an electro-hydrostatic actuator training sample data set, finely adjusting the super parameters of the XGboost model, and finally obtaining an optimal electro-hydrostatic actuator intelligent health monitoring model through cross validation by ten folds;
and 5, performing generalized performance evaluation and fault identification on the intelligent health monitoring model of the electro-hydrostatic actuator on the electro-hydrostatic actuator test set by using the trained grid search optimization XGboost model.
The further improvement of the invention is that in step 1, the vibration, pressure and current signals are respectively subjected to data set segmentation, the segmentation is to cut off the original signals through a window function, the window function length is 2048, the window function is moved to intercept the signals for multiple times, the step length is moved by 200, finally, a sample set is constructed, and the original sample set is randomly disordered to pass through 9: the training set and the test set are extracted at a ratio of 1.
The further improvement of the invention is that in the step 2, the number of wavelet packet decomposition layers is 3, and the complete set empirical mode decomposition of the self-adaptive white noise is introduced to improve the Hilbert-Huang transform mode, so that the problems of aliasing of the empirical mode decomposition mode used by the traditional Hilbert-Huang transform mode and incomplete set empirical mode decomposition process and low decomposition efficiency can be solved.
The further improvement of the invention is that the specific steps of improving the complete set empirical mode decomposition of the adaptive white noise in the Hilbert-Huang transform in the step 2 are as follows:
step 2.1, let S (t) be the original vibration signal sequence, N i (t) is the Gaussian white noise sequence with a normal distribution, ε, added in the ith experiment of a total of I experiments k The signal-to-noise ratio coefficient of the k-th modal component is represented, the initial value k =1, then the k-th modal componentk modal component source signals X i (t) is:
X i (t)=S(t)+ε k N i (t)
step 2.2, based on the i source signals X i (t) determining all maximum and minimum points of the signal, respectively, and determining each source signal X by using cubic spline curve i (t) upper and lower envelope lines, and then obtaining the mean value sequence M of the upper envelope line and the lower envelope line i (t);
Step 2.3, each source signal X i (t) subtracting the respective upper and lower envelope mean sequences M i (t), fundamental modal component sequence H i (t):
H i (t)=X i (t)-M i (t)
Step 2.4, examine each H i (t) whether two conditions for the fundamental modal component are met, if a single H i (t) if not satisfied, then a single H i (t) repeating the above steps 2.2, 2.3 as the signal to be processed until each H i (t) are all a fundamental modal component, the fundamental modal component IMF j (t);
Step 2.5, when k = k +1, j = j +1, the residue signal sequence R (t);
step 2.6, repeat the above operation with R (t) as the new "original" signal S (t) until the iteration terminates.
The further improvement of the invention is that in step 2, the improved Hilbert-Huang mode uses the correlation distance to replace the traditional kurtosis to select the sensitive mode components, and the number of the selected sensitive mode components is 5.
The further improvement of the invention is that based on the relevant distance to replace the traditional kurtosis as the selection standard of the sensitive modal component, the following is concretely realized:
wherein D (X) and D (Y) are variances and E (X) and E (Y) are means.
The further improvement of the invention is that in the step 5, the XGboost model generalization performance is evaluated through the precision P, recall ratio R, accuracy ratio A and F1 score, and the specific method is as follows:
A=(TP+TN)/(TP+FP+TN+FN)
P=(TP)/(TP+FP)
R=(TP)/(TP+FN)
F 1 =(2×P×R)/(P+R)
in the above formula, TP represents the number of faults correctly identified as the category for a certain category of faults, and the category is set to x; TN represents the number of faults of some other class, correctly identified; FP represents the number of other classes that were misidentified as class x; FN represents the number of categories x that were misidentified as other categories.
The invention has at least the following beneficial technical effects:
the invention provides a fault identification method of an electro-hydrostatic actuator based on multi-mode homologous features and an XGboost model. Secondly, multi-mode homologous features, namely a time domain mode, a frequency domain mode, a wavelet packet mode and an improved Hilbert-Huang mode, are respectively extracted from the sample set data, and finally a high-dimensional feature vector with higher robustness is constructed. And then the characteristic that the XGboost model is more effective in being competent for mechanical equipment data mining tasks with large data volume and high feature dimensionality compared with the traditional machine learning model is utilized, and the high-dimensional multi-modal homologous features and the XGboost model are combined to perform the hyper-parameter n of the XGboost model on a training set trees 、n depth And n learning_rates And (3) fine adjustment is carried out to find the XGboost model hyperparameter for realizing the highest fault identification accuracy of the electro-hydrostatic actuator on the electro-hydrostatic actuator sample set, so as to realize accurate and rapid identification of the early fault of the electro-hydrostatic actuator.
In conclusion, the method and the device realize rapid and efficient identification of the faults of the electro-hydrostatic actuator under the conditions of large data volume and high feature dimension, and compared with the conventional identification method of the electro-hydrostatic actuator, the method and the device can improve the fault identification accuracy rate, have better robustness and higher speed of the hydraulic pump of the electro-hydrostatic actuator, are beneficial to reducing life and property losses caused by faults of aerospace aircrafts, and improve the running benefits of the aerospace aircrafts.
Drawings
FIG. 1 is a flow chart of a method for monitoring the state and identifying faults of an electro-hydrostatic actuator based on multi-mode homologous features and an XGboost model.
FIG. 2 is a schematic diagram of segmentation of a data set of an electro-hydrostatic actuator.
Fig. 3 is a flow chart of the modified hilbert yellow algorithm.
FIG. 4 is a flow chart of the XGboost algorithm.
Detailed Description
The invention is described in further detail below with reference to the accompanying drawings.
The invention provides a fault identification method of an electro-hydrostatic actuator based on multi-mode homologous features and an XGboost model. Secondly, multi-modal homologous features, namely a time domain mode, a frequency domain mode, a wavelet packet mode and an improved Hilbert-Huang mode, are respectively extracted from the sample set data, and finally a high-dimensional feature vector with higher robustness is constructed. And then, by utilizing the characteristic that the XGboost model is more effective in being competent for the data mining task of mechanical equipment with large data volume and high feature dimensionality compared with the traditional machine learning model, the XGboost model with the hyper-parameters n is carried out on a training set by combining the high-dimensional multimodal homologous features and the XGboost model trees 、n depth And n learning_rates And fine adjustment is carried out, and finally XGboost model generalization performance evaluation is carried out on a training set based on the optimal XGboost model, and the fault identification of the electro-hydrostatic actuator is realized. Compared with the method of single modal characteristics, artificial neural network, support vector machine and decision tree, the method for identifying the early failure of the electro-hydrostatic actuator has the advantages of higher identification precision, better model robustness and shorter training time.
As shown in fig. 1, the intelligent identification of the electro-hydrostatic actuator, such as pressure, vibration, current signal acquisition, data set segmentation, signal multi-mode homologous feature extraction, grid search fine-tuning model hyper-parameter, feature importance ranking and model generalization performance evaluation, based on the multi-mode homologous feature and the XGBoost model provided by the invention is mainly described in the figure. Data under the normal condition and the abnormal condition of collecting the motor current of the electro-hydrostatic actuator, the vibration signal of a motor pump shell and the pressure signal of an oil outlet of the hydraulic pump, the running state of the electro-hydrostatic actuator is monitored, and early failure intelligent identification is carried out.
The invention discloses a method for identifying faults of an electro-hydrostatic actuator based on multi-mode homologous features and an XGboost model, which mainly comprises the following steps as shown in figure 1:
step 1, collecting data of a motor current of an electro-hydrostatic actuator, a vibration signal of a motor pump shell and a pressure signal of an oil outlet of a hydraulic pump in a normal state and an abnormal state, segmenting a data set to construct an original sample set, and dividing a training set and a testing set;
as shown in fig. 2, when the data set segmentation is performed on the vibration, pressure and current signals, the segmentation is to segment the original signal by a window function, the window function length is 2048, the window function is moved to intercept the signal for multiple times, the step length is moved by 200, and finally a sample set is constructed, and the original sample set is randomly disturbed to pass through 9: the training set and the test set are extracted at a ratio of 1.
Step 2, respectively processing the training set and the test set data of the three signals based on time domain, fourier transform, wavelet packet decomposition and improved Hilbert-Huang transform, and performing characteristic extraction of statistical characteristic parameters on the processing result;
the number of wavelet packet decomposition layers of the wavelet packet mode is 3, and the complete set empirical mode decomposition improved Hilbert-Huang transformation mode of the self-adaptive white noise is introduced to solve the problems that the empirical mode decomposition mode aliasing and the set empirical mode decomposition process used by the traditional Hilbert-Huang transformation mode have no integrity and low decomposition efficiency;
the improved Hilbert-Huang mode uses the correlation distance to replace the traditional kurtosis to select the sensitive mode components, and the number of the selected sensitive mode components is 5.
The traditional kurtosis index is suitable for selecting sensitive modal components with impact components from vibration signals, and the traditional kurtosis is replaced by the relevant distance to serve as a selection standard of the sensitive modal components, so that the method is more universal, and the current, pressure and vibration signals can be effectively applied, and are specifically realized as follows:
in the formula, D (X) and D (Y) are variance, E (X) and E (Y) are mean, and X and Y are two sections of signals involved in calculating correlation distance respectively.
As shown in FIG. 3, the specific steps of the complete set empirical mode decomposition for adaptive white noise in the modified Hilbert-Huang transform are as follows
Step 2.1, let S (t) be the original vibration signal sequence, N i (t) is the Gaussian white noise sequence with a normal distribution, ε, added in the ith experiment of a total of I experiments k The initial value k =1 of the signal-to-noise ratio coefficient representing the k-th modal component, and then the k-th modal component source signal X i (t) is:
X i (t)=S(t)+ε k N i (t)
step 2.2, based on the i source signals X i (t) determining all maximum and minimum points of the signal, respectively, and determining each source signal X by using cubic spline curve i (t) upper and lower envelope lines, and then obtaining the mean value sequence M of the upper envelope line and the lower envelope line i (t)。
Step 2.3, each source signal X i (t) subtracting the respective upper and lower envelope mean sequences M i (t), fundamental modal component sequence H i (t):
H i (t)=X i (t)-M i (t)
Step 2.4, examine each H i (t) whether two conditions for the fundamental modal component are satisfied. If a single H i (t) if not satisfied, then a single H i (t) repeating the above steps 2.2, 2.3 as the signal to be processed until each H i (t) are all a fundamental modeThe state component, i.e. the fundamental mode component IMF j (t);
Step 2.5, when k = k +1, j = j +1, the residue signal sequence R (t);
step 2.6, repeating the operation by taking R (t) as a new 'original' signal S (t) until the iteration is terminated;
and step 4, as shown in fig. 4, the XGboost model can perform 10-fold cross validation search on the super-parameters by using an electro-hydrostatic actuator training sample data set, continuously fit the residual error output by the previous generation tree model, improve the fitting capability of the model on data, then perform training by using a grid to search the super-parameters of the XGboost model, finely adjust the super-parameters of the XGboost model, and finally obtain the optimal electro-hydrostatic actuator intelligent health monitoring model through ten-fold cross validation.
And 5, performing generalized performance evaluation and fault identification on the intelligent health monitoring model of the electro-hydrostatic actuator on the electro-hydrostatic actuator test set by using the trained grid search optimization XGboost model.
The XGboost model generalization performance is evaluated through the scores of the precision P, the recall rate R, the accuracy A and the F1, and the specific method is as follows:
A=(TP+TN)/(TP+FP+TN+FN)
P=(TP)/(TP+FP)
R=(TP)/(TP+FN)
F 1 =(2×P×R)/(P+R)
in the above formula, TP represents the number of faults correctly identified as the class for a certain class of faults, and the class is assumed to be x; TN represents the number of faults in some other category, correctly identified; FP represents the number of other classes that were misidentified as class x; FN represents the number of classes x that are misidentified as other classes.
Inputting the feature vector data of the electro-hydrostatic actuator vibration signal, pressure signal and current signal test data set into the trained XGboost model to evaluate the generalization performance of the model and identify faults, outputting the precision, recall rate, accuracy and F1 score result of the fault identification and classification of the electro-hydrostatic actuator, and comparing the cross verification result of the training set to obtain the generalization performance evaluation result of the electro-hydrostatic actuator.
Claims (5)
1. The method for identifying the faults of the electro-hydrostatic actuator based on the multi-mode homologous features and the XGboost model is characterized by comprising the steps of firstly, carrying out data set segmentation on vibration signals, pressure signals and current signals of the electro-hydrostatic actuator acquired in a big data era to construct an original sample set, and dividing the original sample set into a training set and a testing set; secondly, multi-modal homologous features, namely a time domain modality, a frequency domain modality, a wavelet packet modality and an improved Hilbert-Huang modality, are respectively extracted from the sample set data, a high-dimensional feature vector with higher robustness is finally constructed, and then an XGboost model is utilized to combine the high-dimensional multi-modal homologous features and the XGboost model, so that a hyper-parameter n of the XGboost model is carried out on a training set trees 、n depth And n learning_rates Performing fine adjustment, and finally performing XGboost model generalization performance evaluation on a training set based on the optimal XGboost model and realizing the fault identification of the electro-hydrostatic actuator; the method specifically comprises the following steps:
step 1, collecting data of a motor current of an electro-hydrostatic actuator, a vibration signal of a motor pump shell and a pressure signal of an oil outlet of a hydraulic pump in a normal state and an abnormal state, performing data set segmentation to construct an original sample set, and dividing a training set and a testing set;
step 2, respectively processing the training set and the test set data of the three signals based on time domain, fourier transform, wavelet packet decomposition and improved Hilbert-Huang transform, and performing characteristic extraction of statistical characteristic parameters on the processing results; the wavelet packet decomposition layer number is 3, the complete set empirical mode decomposition of the self-adaptive white noise is introduced to improve the Hilbert-Huang transformation mode, and the problems that the empirical mode decomposition mode aliasing and the set empirical mode decomposition process used by the traditional Hilbert-Huang transformation mode have no integrity and low decomposition efficiency can be solved;
the specific steps of the complete set empirical mode decomposition of the self-adaptive white noise in the improved Hilbert-Huang transform are as follows:
step 2.1, let S (t) be the original vibration signal sequence, N i (t) is the standard normal score added in the ith experiment of a total of I experimentsGaussian white noise sequence of cloth, ε k The signal-to-noise ratio coefficient of the k-th modal component is represented, the initial value k =1, and then the k-th modal component source signal X i (t) is:
X i (t)=S(t)+ε k N i (t)
step 2.2, based on the i source signals X i (t) determining all maximum and minimum points of the signal, respectively, and determining each source signal X by using cubic spline curve i (t) upper and lower envelope lines, and then obtaining the mean value sequence M of the upper envelope line and the lower envelope line i (t);
Step 2.3, each source signal X i (t) subtracting the respective upper and lower envelope mean sequences M i (t), fundamental modal component sequence H i (t):
H i (t)=X i (t)-M i (t)
Step 2.4, examine each H i (t) whether two conditions for the fundamental modal component are met, if a single H i (t) if not satisfied, then a single H i (t) repeating the above steps 2.2, 2.3 as the signal to be processed until each H i (t) are all a fundamental modal component, the fundamental modal component IMF j (t);
Step 2.5, when k = k +1, j = j +1, the residue signal sequence R (t);
step 2.6, repeating the operation by taking the R (t) as a new 'original' signal S (t) until the iteration is terminated;
step 3, combining three signal training data set characteristics of the electro-hydrostatic actuator to construct a high-dimensional characteristic vector to be used as input sample data of XGboost model training, and performing independent thermal coding on various faults of the electro-hydrostatic actuator to be used as output samples of the model training;
step 4, training an XGboost model of grid search by utilizing an electro-hydrostatic actuator training sample data set, finely adjusting the super parameters of the XGboost model, and finally obtaining an optimal electro-hydrostatic actuator intelligent health monitoring model through cross validation by ten folds;
and 5, performing generalized performance evaluation and fault identification on the intelligent health monitoring model of the electro-hydrostatic actuator on the electro-hydrostatic actuator test set by using the trained grid search optimization XGboost model.
2. The method for identifying the faults of the electro-hydrostatic actuator based on the multi-modal homologous features and the XGboost model according to claim 1, wherein in the step 1, the vibration, pressure and current signals are respectively subjected to data set segmentation, the segmentation is to cut off the original signals through a window function, the window function length is 2048, the signals are cut off by moving the window function for multiple times, the step length is moved by 200, finally, a sample set is constructed, and the original sample set is randomly disordered to pass through 9: the training set and the test set are extracted at a ratio of 1.
3. The electro-hydrostatic actuator fault identification method based on the multi-mode homologous features and the XGboost model as claimed in claim 1, wherein in the step 2, sensitive modal components are selected by using the improved Hilbert-Huang mode instead of the traditional kurtosis, and the number of the selected sensitive modal components is 5.
4. The method for identifying the fault of the electro-hydrostatic actuator based on the multi-mode homologous feature and the XGboost model as claimed in claim 3, wherein the method is implemented based on the selection criterion that the traditional kurtosis is replaced by the related distance as the sensitive modal component, and comprises the following steps:
wherein D (X) and D (Y) are variances and E (X) and E (Y) are means.
5. The method for identifying the fault of the electro-hydrostatic actuator based on the multi-mode homologous features and the XGboost model according to claim 1, wherein in the step 5, the generalization performance of the XGboost model is evaluated through scores of precision P, recall rate R, accuracy A and F1, and the method comprises the following specific steps:
A=(TP+TN)/(TP+FP+TN+FN)
P=(TP)/(TP+FP)
R=(TP)/(TP+FN)
F1=(2×P×R)/(P+R)
in the above formula, TP represents the number of faults correctly identified as the category for a certain category of faults, and the category is set to x; TN represents the number of faults in some other category, correctly identified; FP represents the number of other classes that were misidentified as class x; FN represents the number of classes x that are misidentified as other classes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110431264.2A CN113191232B (en) | 2021-04-21 | 2021-04-21 | Electro-hydrostatic actuator fault identification method based on multi-mode homologous features and XGboost model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110431264.2A CN113191232B (en) | 2021-04-21 | 2021-04-21 | Electro-hydrostatic actuator fault identification method based on multi-mode homologous features and XGboost model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113191232A CN113191232A (en) | 2021-07-30 |
CN113191232B true CN113191232B (en) | 2023-04-18 |
Family
ID=76977903
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110431264.2A Active CN113191232B (en) | 2021-04-21 | 2021-04-21 | Electro-hydrostatic actuator fault identification method based on multi-mode homologous features and XGboost model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113191232B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114987773A (en) * | 2022-05-20 | 2022-09-02 | 成都飞机工业(集团)有限责任公司 | Method, device, equipment and medium for identifying performance abnormity of aircraft hydraulic system |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110837851A (en) * | 2019-10-25 | 2020-02-25 | 西安交通大学 | Fault diagnosis method for hydraulic pump of electro-hydrostatic actuator |
Family Cites Families (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106092574B (en) * | 2016-05-30 | 2018-04-17 | 西安工业大学 | Based on the Method for Bearing Fault Diagnosis for improving EMD decomposition and sensitive features selection |
US10113552B2 (en) * | 2016-10-13 | 2018-10-30 | Caterpillar Inc. | System, method, and apparatus to monitor compressor health |
CN107358156B (en) * | 2017-06-06 | 2020-05-19 | 华南理工大学 | Feature extraction method for ultrasonic tissue characterization based on Hilbert-Huang transform |
CN109100143A (en) * | 2018-07-06 | 2018-12-28 | 华中科技大学 | Fault Diagnosis of Roller Bearings and equipment based on CEEMDAN and CFSFDP |
CN110220709B (en) * | 2019-06-06 | 2020-04-21 | 北京科技大学 | Rolling bearing fault diagnosis method based on CNN model and transfer learning |
CN111222458A (en) * | 2020-01-06 | 2020-06-02 | 浙江工业大学 | Rolling bearing fault diagnosis method based on ensemble empirical mode decomposition and convolutional neural network |
CN111428418A (en) * | 2020-02-28 | 2020-07-17 | 贵州大学 | Bearing fault diagnosis method and device, computer equipment and storage medium |
CN111272429B (en) * | 2020-03-04 | 2021-08-17 | 贵州大学 | Bearing fault diagnosis method |
CN112200016A (en) * | 2020-09-17 | 2021-01-08 | 东北林业大学 | Electroencephalogram signal emotion recognition based on ensemble learning method AdaBoost |
CN112347588A (en) * | 2020-11-26 | 2021-02-09 | 中国舰船研究设计中心 | Rotary machine fault diagnosis method based on wavelet packet decomposition |
CN112668500B (en) * | 2020-12-30 | 2023-12-29 | 太原科技大学 | Xgboost-based rolling mill multi-target vibration prediction method |
CN112651385A (en) * | 2021-01-18 | 2021-04-13 | 西北工业大学青岛研究院 | Magnetic anomaly multi-feature information extraction method based on Hilbert-Huang transform |
-
2021
- 2021-04-21 CN CN202110431264.2A patent/CN113191232B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110837851A (en) * | 2019-10-25 | 2020-02-25 | 西安交通大学 | Fault diagnosis method for hydraulic pump of electro-hydrostatic actuator |
Also Published As
Publication number | Publication date |
---|---|
CN113191232A (en) | 2021-07-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhang et al. | Instance-based ensemble deep transfer learning network: A new intelligent degradation recognition method and its application on ball screw | |
CN112149316A (en) | Aero-engine residual life prediction method based on improved CNN model | |
CN109213121B (en) | Method for diagnosing clamping cylinder fault of fan braking system | |
CN109460618A (en) | A kind of rolling bearing remaining life on-line prediction method and system | |
CN106886660A (en) | EEMD Hilbert envelope spectrums are with Rolling Bearing Status recognition methods under the varying load that DBN is combined | |
CN107368809A (en) | A kind of bearing fault sorting technique based on rarefaction representation and dictionary learning | |
CN113505655B (en) | Intelligent bearing fault diagnosis method for digital twin system | |
CN105469138A (en) | Control system performer fault diagnosis method based on particle swarms and support vector machine | |
CN112378660A (en) | Intelligent fault diagnosis method for aero-engine bearing based on data driving | |
CN104614166A (en) | Method for identifying failure state of rotor vibration signal of aircraft engine | |
CN113191232B (en) | Electro-hydrostatic actuator fault identification method based on multi-mode homologous features and XGboost model | |
CN104832418A (en) | Hydraulic pump fault diagnosis method based on local mean conversion and Softmax | |
CN114880925A (en) | Equipment life prediction method based on time convolution network and multi-layer self-attention | |
CN110737976A (en) | mechanical equipment health assessment method based on multi-dimensional information fusion | |
CN115587290A (en) | Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network | |
CN110020637B (en) | Analog circuit intermittent fault diagnosis method based on multi-granularity cascade forest | |
CN112163474B (en) | Intelligent gearbox diagnosis method based on model fusion | |
BENKEDJOUH et al. | Deep Learning for Fault Diagnosis based on short-time Fourier transform | |
Wang | Research on the fault diagnosis of mechanical equipment vibration system based on expert system | |
CN112748331A (en) | Circuit breaker mechanical fault identification method and device based on DS evidence fusion | |
Korovin et al. | Real-time diagnostics of oil production equipment using data mining | |
CN105279553B (en) | A kind of height adds to water system fault degree recognition methods | |
Xiao et al. | Health assessment for piston pump using LSTM neural network | |
CN110222390A (en) | Gear crack recognition methods based on wavelet neural network | |
CN114563671A (en) | High-voltage cable partial discharge diagnosis method based on CNN-LSTM-Attention neural network |
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 |