CN111766067A - Aircraft outfield aircraft engine fault prediction method based on deep learning - Google Patents
Aircraft outfield aircraft engine fault prediction method based on deep learning Download PDFInfo
- Publication number
- CN111766067A CN111766067A CN202010666943.3A CN202010666943A CN111766067A CN 111766067 A CN111766067 A CN 111766067A CN 202010666943 A CN202010666943 A CN 202010666943A CN 111766067 A CN111766067 A CN 111766067A
- Authority
- CN
- China
- Prior art keywords
- engine
- layer
- aircraft
- convolution
- obtaining
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M15/00—Testing of engines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G06N3/045—Combinations of networks
Abstract
The method is mainly applied to the aircraft outfield autonomous security information support system and based on the deep learning of the failure prediction of the aircraft engine; the method comprises the following steps: acquiring a time sequence data set consisting of N airplane parameters sensitive to faults in an airplane engine; acquiring a spectrogram according to time sequence waves of the engine flight parameter data within fixed time; the deep learning algorithm carries out fault prediction on the aircraft engine according to the frequency spectrogram; the method comprises the steps of firstly obtaining a time sequence data set consisting of N airplane parameters sensitive to faults in the airplane engine, then obtaining a frequency spectrogram according to time sequence waves of engine flight parameter data within fixed time, and finally performing fault prediction on the airplane engine according to the frequency spectrogram by adopting a deep learning algorithm based on a convolutional neural network framework. The method can accurately predict the remaining life of the aircraft engine, effectively predict the health condition of the aircraft engine, and avoid serious consequences caused by uncertain faults in actual flight.
Description
Technical Field
The invention belongs to the technical field of computers, and particularly relates to an aircraft engine fault prediction method based on deep learning for a military flight big data aircraft outfield autonomous security information support system.
Background
From the last 90 years to the present, the technology of aviation equipment is rapidly developed, and particularly under the large environment that the operational use style of the aviation equipment is changed, the requirement on the ground guarantee of the airplane is higher and higher, and the guarantee of the engine of the airplane is the most fundamental element. The rapid development of military science and technology puts higher requirements on the guarantee and fault analysis of aircraft engines. However, in long-term development, the technology of securing aircraft engines has always lagged behind other aerospace equipment technologies. The original aircraft engine barrier system has great challenge under the condition of new equipment, and the readiness rate of military aircraft can be greatly reduced if the barrier system cannot meet the guarantee requirement.
The health management of aircraft engines lacks quantitative analysis, and the accumulated experience and data during actual use and maintenance cannot be well combined with design data to make the theory and practice separate. The aircraft engine has no early warning mechanism when breaking down, and the aircraft engine that the aircraft service external field equipment maintainer was equipped with is difficult to accomplish in mind the several, again because the predictability is not enough, and overhauls excessively and overhauls not enough phenomenon and coexist, leads to aircraft engine's intact rate to descend.
When the aircraft engine is in fault, the fault data at the current stage is not structured, and the maintenance personnel in the field of the aircraft service are difficult to perform definite fault diagnosis aiming at the comprehensive analysis of the fault phenomenon, the reliability data, the index data and the like, so that the optimal fault prediction method is difficult to find and replace the aircraft engine. This increases the maintenance costs of the aircraft engine, while the failed engine is not well maintained, resulting in a waste of resources.
The failure prediction of the aircraft engine is the basis of the maintenance of the military aircraft, and influences the operation efficiency and the maintenance guarantee efficiency of the military aircraft all the time, so that the failure prediction of the aircraft engine is very important in the whole army, and the most accurate failure prediction is necessarily developed in the field of ground guarantee while the aviation equipment is vigorously developed.
Disclosure of Invention
The invention provides a method for predicting the failure of an aircraft outfield aircraft engine based on deep learning, aiming at the problems in the prior art, which specifically comprises the following steps:
step 1: obtaining a time series data set composed of N airplane parameters sensitive to faults in an engine
N airplane parameters sensitive to faults in the engine comprise the total temperature of gas behind a low-pressure turbine, a vibration value, the rotating speed of a low-pressure rotor of the engine, the rotating speed of a high-pressure rotor of the engine, the position of an accelerator, the rotating angle of a low-pressure inlet blade, the rotating angle of a high-pressure inlet blade, the position of a nozzle fish scale, the total temperature of air at the inlet of the engine, the pressure of a lubricating oil inlet, the duty ratio S1, the duty ratio S8 and the APII-39 secondary power;
step 2, acquiring a spectrogram according to time sequence waves of the engine flight parameter data within fixed time
The method specifically comprises the following steps:
step 2.1, obtaining time sequence data consisting of N airplane parameters sensitive to faults in an engine, and partitioning the time sequence data according to t milliseconds fixed time;
step 2.2, drawing the engine flight parameter data in the time t into a time sequence wave;
step 2.3, decomposing the time sequence wave by utilizing Fourier transform operation, solving the energy value of each frequency band, and obtaining the time sequence wave frequency spectrogram of N airplane parameters sensitive to faults in the engine;
step 3, predicting the faults of the aircraft engine by using a deep learning algorithm according to the spectrogram;
step 3.1, off-line training a convolutional neural network framework by utilizing a flight parameter data spectrogram obtained by Fourier transform during normal operation and abnormal operation of an aircraft engine;
step 3.1.1, constructing a convolutional neural network, which comprises the following specific steps:
A. constructing an input layer: taking the flight parameter data spectrograms with the same size of the aircraft engine in normal operation and abnormal operation as an input layer;
B. constructing a rolling layer: the convolution layer is composed of K convolution filters, the K filters represent that K convolution kernels and an input image are subjected to convolution operation, each convolution kernel corresponds to one feature map, feature extraction is carried out through the convolution operation, new K feature image matrixes s1 are obtained, and the new K feature image matrixes s1 are results of feature extraction on the spectrogram; the convolution kernel is a feature matrix;
C. constructing a down-sampling layer: the down-sampling layer samples the feature map to reduce dimensionality and computational complexity; dividing K characteristic image matrixes s1 obtained through convolution layer operation into non-overlapping matrix areas with the size of t multiplied by t, wherein the operation of taking the maximum value of each matrix area is called maximum pooling, and the operation of taking the mean value is called mean pooling; the operation of weighted summation or maximum value of the divided matrix area is carried out, the frequency of the data is larger, the influence of the frequency on the average of the whole data is larger, and the weight is larger; multiplying the matrix area with large data frequency number by a weight matrix, adding offset, and then obtaining a final down-sampling characteristic image matrix c1 through the operation of an activation function;
D. and D, repeating the step B and the step C: taking the downsampled characteristic image matrix C1 as an input layer again, obtaining a second convolution layer s2 through the step B, obtaining a second downsampled layer C2 through the step C, repeating the two steps to reduce the size of the characteristic image matrix s1 until the characteristic image matrix s1 is rasterized into one-dimensional data after the last filtering; constructing a dense connection layer: arranging data rasterized into one dimension into a line to form a characteristic vector, wherein the characteristic vector is fully connected with an output layer and is used as a full connection layer, namely, the convolution layer is multiplied by a weight matrix of a down sampling layer, and then offset is added, and then an activation function sigmoid is used for the result obtained by the operation to obtain the full connection layer;
E. constructing an output layer: the output layer adopts a softmax layer, is connected with the full-connection layer and outputs a detection result, and the softmax is used for a multi-classification process, maps the output of a plurality of neurons into a (0, 1) interval and considers the mapping result as prediction of a probability theory;
3.1.2, performing steepest descent optimization on the error gradient of the convolutional neural network by using a flight parameter data spectrogram in normal operation and abnormal operation of an aircraft engine by adopting an Adam algorithm, and training the convolutional neural network off line;
and 3.2, performing online prediction on the fault condition of the aircraft engine by using the convolutional neural network framework trained in the step 3.1 and according to the time sequence wave spectrum diagram of the flight parameter data obtained in the step 2.3, wherein the online prediction is specifically implemented by inputting new flight parameter data into a training network, extracting data characteristics through a convolution kernel, and obtaining a prediction result through downsampling and activating a function, wherein the result is the online prediction result of the fault of the aircraft engine.
In a specific embodiment of the invention, in step 2.1, t 15000 milliseconds.
In another specific embodiment of the present invention, in step C, the activation function uses a pointwise sigmoid activation function, where the sigmoid activation function is:
in another embodiment of the present invention, in step E, if the input layer is a set of 28 × 28 spectrograms of flight parameter data in a certain time, the first layer convolutional layer uses 6 convolutional kernels to obtain 6 24 × 4 feature maps, and 6 sampled 12 × 12 feature maps are obtained through the sampling layer; obtaining 12 characteristic graphs of 8 multiplied by 8 by 12 through a second layer of convolution layer which is composed of 12 convolution kernels, and obtaining 12 characteristic graphs of 4 multiplied by 4 through a sampling layer; and connecting the 12 4 x 4 characteristic graphs with the full-connection layer after passing through the grating, extracting the integral characteristic information of the spectrogram, obtaining a high-level characteristic vector of the spectrogram, and using the high-level characteristic vector to input a subsequent sotfmax classifier, thereby predicting the fault of the aircraft engine.
The invention provides a method for predicting the faults of an aircraft engine based on deep learning, which is applied to an aircraft service outfield autonomous security information support system and can effectively predict the faults of the aircraft engine. The invention discloses a method for predicting the fault of an aircraft engine based on deep learning, which is applied to an aircraft service outfield autonomous security information support system and comprises the following steps: the method comprises the steps of firstly, obtaining a time sequence data set consisting of N airplane parameters sensitive to faults in an airplane engine, then obtaining a frequency spectrogram according to time sequence waves of engine flight parameter data within fixed time, and finally performing fault prediction on the airplane engine according to the frequency spectrogram by adopting a deep learning algorithm based on a convolutional neural network framework. The method can accurately predict the residual life of the aircraft engine, effectively predict the health condition of the aircraft engine, and avoid serious consequences caused by uncertain faults in actual flight.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic flow diagram illustrating a method for aircraft engine fault prediction based on deep learning for use in a flight service outfield autonomous security information support system in accordance with the present invention;
FIG. 2 is a block diagram of a convolutional neural network in accordance with an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without inventive step, are within the scope of the present invention.
Fig. 1 is a flowchart of a first embodiment of the method for predicting a failure of an aircraft outfield aircraft engine based on deep learning, which specifically includes the following steps:
step 1, acquiring a time sequence data set consisting of N airplane parameters sensitive to faults in an engine;
in the embodiment of the invention, N airplane parameters sensitive to faults in the engine comprise the total temperature of gas after a low-pressure turbine, a vibration value, the rotating speed of a low-pressure rotor of the engine, the rotating speed of a high-pressure rotor of the engine, the position of an accelerator, the rotating angle of a low-pressure inlet blade, the rotating angle of a high-pressure inlet blade, the position of a nozzle flake, the total temperature of air at the inlet of the engine, the pressure of a lubricating oil inlet, the duty ratio S1, the duty ratio S8 and the APII-39. The time series data set of the N aircraft parameters in the engine that are sensitive to faults is obtained by conventional methods well known to those skilled in the art, such as by data download, etc.
Step 2, acquiring a spectrogram according to time sequence waves of the engine flight parameter data within fixed time;
the method specifically comprises the following steps:
step 2.1, the time sequence data formed by N types of airplane parameters sensitive to faults in the engine is obtained and is partitioned according to the fixed time length of t milliseconds, wherein t is 15000 milliseconds for example;
step 2.2, drawing the engine flight parameter data in 15000 millisecond time into a time sequence wave;
the method for drawing the engine flight parameter data into time sequence waves is well known by the technical personnel in the field and is not described in detail;
and 2.3, decomposing the time sequence wave by utilizing Fourier transform operation, solving the energy value of each frequency band, and obtaining the time sequence wave frequency spectrum of N airplane parameters sensitive to faults in the engine. For the specific implementation, please refer to the signal and linear system in various versions.
And 3, predicting the faults of the aircraft engine by using a deep learning algorithm according to the spectrogram.
The specific implementation of this step can be found in "Shao H, Jiang H, Zhang X, et al.
Step 3.1, off-line training a convolutional neural network framework by utilizing a flight parameter data spectrogram obtained by Fourier transform during normal operation and abnormal operation of an aircraft engine;
step 3.1.1, constructing a convolutional neural network, which comprises the following specific steps:
A. constructing an input layer: taking the flight parameter data spectrograms with the same size of the aircraft engine in normal operation and abnormal operation as an input layer;
B. constructing a rolling layer: the convolution layer is composed of K convolution filters, the K filters represent convolution operation between K convolution kernels and an input image, and each convolution kernel corresponds to a feature map. The convolution operation is to use a convolution kernel to walk around the image matrix, multiply the corresponding position elements, add the multiplication results, and finally form new K characteristic image matrixes s 1. Feature extraction is performed through convolution operation, and obtaining new K feature image matrixes s1 is a result of feature extraction performed on the spectrogram. An important feature of the convolution operation is that the original signal features can be enhanced and the noise can be reduced by the convolution operation. Therefore, the convolution layer mainly has the function of feature extraction, and the convolution kernel is a feature matrix.
C. Structure of downsampling layer (pooling layer): the down-sampling layer (pooling layer) samples the feature map to reduce dimensionality and computational complexity. The K feature image matrices s1 obtained by convolutional layer operation are divided into non-overlapping matrix regions of size t × t, and the operation of taking the maximum value for each matrix region is called maximum pooling, and the operation of taking the average value is called average pooling. And carrying out weighted summation or maximum value calculation on the divided matrix area, wherein the frequency of the data is larger, which shows that the influence of the frequency on the average of the whole data is larger, and the weight is larger. Multiplying the matrix area with large data frequency by a weight matrix, adding an offset, wherein the weight matrix and the offset are well known to those skilled in the art and are not described in detail, and then obtaining a final downsampling characteristic image matrix c1 through operation of an activation function, wherein the activation function is generally completed by a point-by-point sigmoid activation function, and other appropriate functions can also be adopted; the sigmoid activation function is:
D. and D, repeating the step B and the step C: taking the downsampled feature image matrix C1 as an input layer again, obtaining a second convolution layer s2 through the step B, obtaining a second downsampled layer C2 through the step C, repeating the two steps, and reducing the size (namely the length and the width of the picture) of the feature image matrix s1 until the feature image matrix s1 is rasterized into one-dimensional data after the last filtering; constructing a dense connection layer: arranging data rasterized into one dimension into a line to form a characteristic vector, wherein the characteristic vector is fully connected with an output layer and is used as a full connection layer, namely, a convolution layer is multiplied by a weight matrix of a down sampling layer, and then offset is added, and then an activation function sigmoid is used for a result obtained by the operation to obtain a final full connection layer;
(Steps A-D refer to Shao H, Jiang H, Zhang X, et al. Rolling bearing fault finding using an optimization delay network [ J ]. Measurement Science & Technology, 2015, 26 (11);)
E. Constructing an output layer: the output layer adopts a softmax layer, is connected with the full-connection layer and outputs the final detection result, and the softmax is used for the multi-classification process, maps the output of a plurality of neurons into a (0, 1) interval and can regard the mapping result as the prediction of probability; specific implementation details are described in (Chen Y, Lin Z, ZHao X, et al. deep Learning-based Classification of Hyperspectral Data [ J ]. F Selected Topics in Applied Earth orbit and Remote Sensing, 2014, 7 (6): 2094-
As shown in fig. 2, if the input layer is a set of spectrogram with a size of 28 × 28 of flight parameter data in a certain time, the first layer convolution layer adopts 6 convolution kernels, so as to obtain 6 characteristic image matrices of 24 × 4, and 6 characteristic image matrices of 12 × 12 after sampling are obtained through the sampling layer; obtaining 12 characteristic image matrixes of 8 multiplied by 8 through a second layer of convolution layer which consists of 12 convolution kernels, and obtaining 12 characteristic image matrixes of 4 multiplied by 4 through a sampling layer; connecting 12 4 x 4 feature image matrixes with a full-connection layer after passing through a grating, extracting the overall feature information of the spectrogram to obtain a high-layer feature vector of the final spectrogram, and using the high-layer feature vector to input a subsequent sotfmax classifier so as to predict the faults of the aircraft engine; the convolutional neural network is only used for illustration, the size of an input image, the number of channels can be changed, the number of convolutional kernels and the depth of the network can be changed.
And 3.1.2, performing steepest descent optimization on the error gradient of the convolutional neural network by using the flight parameter data spectrogram of the aircraft engine in normal operation and abnormal operation and adopting an Adam algorithm, and training the convolutional neural network off line. The detailed implementation method should be referred to the "Python and machine learning actual war" authored by he yu jian (electronic industry press, published in 2017 month).
And 3.2, performing online prediction on the fault condition of the aircraft engine by using the convolutional neural network framework trained in the step 3.1 according to the time sequence wave spectrum diagram of the flight parameter data obtained in the step 2.3, wherein the online prediction is specifically implemented by inputting new flight parameter data into a training network (the data can be obtained through the step 3.1.1A), extracting data characteristics through a convolution kernel (the data can be obtained through the step 3.1.1B), and obtaining a prediction result through downsampling and activating a function, wherein the result is the online prediction result of the fault of the aircraft engine (the results can be obtained through the step 3.1.1C and D).
According to the method, a time sequence data set composed of N airplane parameters sensitive to faults in the airplane engine is obtained, a frequency spectrum graph is obtained according to time sequence waves of the airplane parameter data in a fixed time, and finally, the airplane engine is subjected to fault prediction according to the frequency spectrum graph by adopting a deep learning algorithm based on a convolutional neural network framework, so that the residual life of the airplane engine is accurately predicted, the health condition of the airplane engine can be effectively predicted, and serious consequences caused by uncertain faults in actual flight are avoided.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (4)
1. The method for predicting the engine fault of the aircraft service outfield aircraft based on deep learning is characterized by comprising the following steps
Step 1: obtaining a time series data set composed of N airplane parameters sensitive to faults in an engine
N airplane parameters sensitive to faults in the engine comprise the total temperature of gas behind a low-pressure turbine, a vibration value, the rotating speed of a low-pressure rotor of the engine, the rotating speed of a high-pressure rotor of the engine, the position of an accelerator, the rotating angle of a low-pressure inlet blade, the rotating angle of a high-pressure inlet blade, the position of a nozzle fish scale, the total temperature of air at the inlet of the engine, the pressure of a lubricating oil inlet, the duty ratio S1, the duty ratio S8 and the APII-39 secondary power;
step 2, acquiring a spectrogram according to time sequence waves of the engine flight parameter data within fixed time
The method specifically comprises the following steps:
step 2.1, obtaining time sequence data consisting of N airplane parameters sensitive to faults in an engine, and partitioning the time sequence data according to t milliseconds fixed time;
step 2.2, drawing the engine flight parameter data in the time t into a time sequence wave;
step 2.3, decomposing the time sequence wave by utilizing Fourier transform operation, solving the energy value of each frequency band, and obtaining the time sequence wave frequency spectrogram of N airplane parameters sensitive to faults in the engine;
step 3, predicting the faults of the aircraft engine by using a deep learning algorithm according to the spectrogram;
step 3.1, off-line training a convolutional neural network framework by utilizing a flight parameter data spectrogram obtained by Fourier transform during normal operation and abnormal operation of an aircraft engine;
step 3.1.1, constructing a convolutional neural network, which comprises the following specific steps:
A. constructing an input layer: taking the flight parameter data spectrograms with the same size of the aircraft engine in normal operation and abnormal operation as an input layer;
B. constructing a rolling layer: the convolution layer is composed of K convolution filters, the K filters represent that K convolution kernels and an input image are subjected to convolution operation, each convolution kernel corresponds to one feature map, feature extraction is carried out through the convolution operation, new K feature image matrixes s1 are obtained, and the new K feature image matrixes s1 are results of feature extraction on the spectrogram; the convolution kernel is a feature matrix;
C. constructing a down-sampling layer: the down-sampling layer samples the feature map to reduce dimensionality and computational complexity; dividing K characteristic image matrixes s1 obtained through convolution layer operation into non-overlapping matrix areas with the size of t multiplied by t, wherein the operation of taking the maximum value of each matrix area is called maximum pooling, and the operation of taking the mean value is called mean pooling; the operation of weighted summation or maximum value of the divided matrix area is carried out, the frequency of the data is larger, the influence of the frequency on the average of the whole data is larger, and the weight is larger; multiplying the matrix area with large data frequency number by a weight matrix, adding offset, and then obtaining a final down-sampling characteristic image matrix c1 through the operation of an activation function;
D. and D, repeating the step B and the step C: taking the downsampled characteristic image matrix C1 as an input layer again, obtaining a second convolution layer s2 through the step B, obtaining a second downsampled layer C2 through the step C, repeating the two steps to reduce the size of the characteristic image matrix s1 until the characteristic image matrix s1 is rasterized into one-dimensional data after the last filtering; constructing a dense connection layer: arranging data rasterized into one dimension into a line to form a characteristic vector, wherein the characteristic vector is fully connected with an output layer and is used as a full connection layer, namely, the convolution layer is multiplied by a weight matrix of a down sampling layer, and then offset is added, and then an activation function sigmoid is used for the result obtained by the operation to obtain the full connection layer;
E. constructing an output layer: the output layer adopts a softmax layer, is connected with the full-connection layer and outputs a detection result, and the softmax is used for a multi-classification process, maps the output of a plurality of neurons into a (0, 1) interval and considers the mapping result as prediction of a probability theory;
3.1.2, performing steepest descent optimization on the error gradient of the convolutional neural network by using a flight parameter data spectrogram in normal operation and abnormal operation of an aircraft engine by adopting an Adam algorithm, and training the convolutional neural network off line;
and 3.2, performing online prediction on the fault condition of the aircraft engine by using the convolutional neural network framework trained in the step 3.1 and according to the time sequence wave spectrum diagram of the flight parameter data obtained in the step 2.3, wherein the online prediction is specifically implemented by inputting new flight parameter data into a training network, extracting data characteristics through a convolution kernel, and obtaining a prediction result through downsampling and activating a function, wherein the result is the online prediction result of the fault of the aircraft engine.
2. The method of predicting outboard aircraft engine faults as claimed in claim 1 wherein in step 2.1, t 15000 milliseconds.
3. The method for predicting the engine fault of the aircraft outboard farm of claim 1, wherein in the step C, the activating function is a pointwise sigmoid activating function, and the sigmoid activating function is:
4. the method of claim 1, wherein in step E, if the input layer is a set of 28 × 28 spectrograms of flight parameter data in a certain time, the first layer convolution layer uses 6 convolution kernels to obtain 6 24 × 4 feature maps, and 6 sampled 12 × 12 feature maps are obtained through the sampling layer; obtaining 12 characteristic graphs of 8 multiplied by 8 by 12 through a second layer of convolution layer which is composed of 12 convolution kernels, and obtaining 12 characteristic graphs of 4 multiplied by 4 through a sampling layer; and connecting the 12 4 x 4 characteristic graphs with the full-connection layer after passing through the grating, extracting the integral characteristic information of the spectrogram, obtaining a high-level characteristic vector of the spectrogram, and using the high-level characteristic vector to input a subsequent sotfmax classifier, thereby predicting the fault of the aircraft engine.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010666943.3A CN111766067A (en) | 2020-07-10 | 2020-07-10 | Aircraft outfield aircraft engine fault prediction method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010666943.3A CN111766067A (en) | 2020-07-10 | 2020-07-10 | Aircraft outfield aircraft engine fault prediction method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111766067A true CN111766067A (en) | 2020-10-13 |
Family
ID=72726347
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010666943.3A Pending CN111766067A (en) | 2020-07-10 | 2020-07-10 | Aircraft outfield aircraft engine fault prediction method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111766067A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112613431A (en) * | 2020-12-28 | 2021-04-06 | 中北大学 | Automatic identification method, system and device for leaked gas |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6240343B1 (en) * | 1998-12-28 | 2001-05-29 | Caterpillar Inc. | Apparatus and method for diagnosing an engine using computer based models in combination with a neural network |
| CN109115501A (en) * | 2018-07-12 | 2019-01-01 | 哈尔滨工业大学（威海） | A kind of Civil Aviation Engine Gas path fault diagnosis method based on CNN and SVM |
| CN109992916A (en) * | 2019-04-09 | 2019-07-09 | 山东超越数控电子股份有限公司 | A kind of engineering outfield aircraft engine failure prediction method based on flight data, terminal and readable storage medium storing program for executing |
| CN109992915A (en) * | 2019-04-09 | 2019-07-09 | 山东超越数控电子股份有限公司 | A kind of engineering outfield aircraft fuel system failure prediction method based on deep learning, terminating machine and readable storage medium storing program for executing |
| CN110068760A (en) * | 2019-04-23 | 2019-07-30 | 哈尔滨理工大学 | A kind of Induction Motor Fault Diagnosis based on deep learning |
-
2020
- 2020-07-10 CN CN202010666943.3A patent/CN111766067A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6240343B1 (en) * | 1998-12-28 | 2001-05-29 | Caterpillar Inc. | Apparatus and method for diagnosing an engine using computer based models in combination with a neural network |
| CN109115501A (en) * | 2018-07-12 | 2019-01-01 | 哈尔滨工业大学（威海） | A kind of Civil Aviation Engine Gas path fault diagnosis method based on CNN and SVM |
| CN109992916A (en) * | 2019-04-09 | 2019-07-09 | 山东超越数控电子股份有限公司 | A kind of engineering outfield aircraft engine failure prediction method based on flight data, terminal and readable storage medium storing program for executing |
| CN109992915A (en) * | 2019-04-09 | 2019-07-09 | 山东超越数控电子股份有限公司 | A kind of engineering outfield aircraft fuel system failure prediction method based on deep learning, terminating machine and readable storage medium storing program for executing |
| CN110068760A (en) * | 2019-04-23 | 2019-07-30 | 哈尔滨理工大学 | A kind of Induction Motor Fault Diagnosis based on deep learning |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112613431A (en) * | 2020-12-28 | 2021-04-06 | 中北大学 | Automatic identification method, system and device for leaked gas |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Han et al. | An adaptive spatiotemporal feature learning approach for fault diagnosis in complex systems | |
| Gou et al. | Aeroengine control system sensor fault diagnosis based on CWT and CNN | |
| CN107066759B (en) | Steam turbine rotor vibration fault diagnosis method and device | |
| Zhao et al. | Highly imbalanced fault diagnosis of mechanical systems based on wavelet packet distortion and convolutional neural networks | |
| CN113469060A (en) | Multi-sensor fusion convolution neural network aeroengine bearing fault diagnosis method | |
| CN114282579A (en) | Aviation bearing fault diagnosis method based on variational modal decomposition and residual error network | |
| CN108182452A (en) | Aero-engine fault detection method and system based on grouping convolution self-encoding encoder | |
| EP1717659A1 (en) | Kernel-Based Fault Detection System and Method | |
| Deng et al. | An improved deep residual network with multiscale feature fusion for rotating machinery fault diagnosis | |
| CN110334948B (en) | Power equipment partial discharge severity evaluation method and system based on characteristic quantity prediction | |
| CN111125841A (en) | Health assessment method for aircraft engine, computer device and storage medium | |
| CN110987439A (en) | Aeroengine fault prediction method based on Logitics regression and Xgboost model | |
| CN114355240A (en) | Power distribution network ground fault diagnosis method and device | |
| CN111766067A (en) | Aircraft outfield aircraft engine fault prediction method based on deep learning | |
| CN115204368A (en) | Aircraft engine fault diagnosis method based on intelligent chip technology | |
| Lingli et al. | Fault diagnosis of a planetary gearbox based on a local bi-spectrum and a convolutional neural network | |
| Wu et al. | A planetary gearbox fault diagnosis method based on time-series imaging feature fusion and a transformer model | |
| Ge et al. | A deep condition feature learning approach for rotating machinery based on MMSDE and optimized SAEs | |
| Calderano et al. | An enhanced aircraft engine gas path diagnostic method based on upper and lower singleton type-2 fuzzy logic system | |
| DiBiano et al. | Automated diagnostics for manufacturing machinery based on well-regularized deep neural networks | |
| Sharanya et al. | Predicting remaining useful life of turbofan engines using degradation signal based echo state network | |
| Zurita et al. | A review of vibration machine diagnostics by using artificial intelligence methods | |
| CN115238833B (en) | Method for extracting performance degradation characteristics of air turbine starter bearing | |
| Hassan et al. | Wavelet-based multiresolution bispectral analysis for detection and classification of helicopter drive-shaft problems | |
| Nie et al. | A novel framework using gated recurrent unit for fault diagnosis of rotary machinery with noisy labels |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20201013 |