CN113033632A - High-altitude platform fault diagnosis method based on wavelet analysis and multi-layer overrun learning machine - Google Patents

High-altitude platform fault diagnosis method based on wavelet analysis and multi-layer overrun learning machine Download PDF

Info

Publication number
CN113033632A
CN113033632A CN202110258961.2A CN202110258961A CN113033632A CN 113033632 A CN113033632 A CN 113033632A CN 202110258961 A CN202110258961 A CN 202110258961A CN 113033632 A CN113033632 A CN 113033632A
Authority
CN
China
Prior art keywords
layer
output
wavelet
fault
fault diagnosis
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
Application number
CN202110258961.2A
Other languages
Chinese (zh)
Inventor
韩渭辛
许斌
范泉涌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northwestern Polytechnical University
Original Assignee
Northwestern Polytechnical University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Northwestern Polytechnical University filed Critical Northwestern Polytechnical University
Priority to CN202110258961.2A priority Critical patent/CN113033632A/en
Publication of CN113033632A publication Critical patent/CN113033632A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Analysis (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Evolutionary Computation (AREA)
  • Computational Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention relates to a high-altitude platform fault diagnosis method based on wavelet analysis and a multilayer overrun learning machine, and belongs to the field of intelligent fault diagnosis of a high-altitude platform sensor system. The method comprises the following steps: extracting characteristics and denoising of original data of a fault sample with a label based on a system by adopting a wavelet analysis method; constructing a multilayer overrun learning machine, and performing online sequence learning training; and carrying out fault diagnosis on the actual system data. The invention adopts the wavelet analysis method to extract and denoise the fault characteristics, constructs an online multilayer ultralimit learning machine to classify various faults, further diagnoses the fault categories, breaks through the limitation that the existing single-layer ultralimit learning machine has low diagnosis precision and can not diagnose the sensor fault in time, and improves the accuracy of fault diagnosis.

Description

High-altitude platform fault diagnosis method based on wavelet analysis and multi-layer overrun learning machine
Technical Field
The invention relates to an intelligent fault diagnosis method in the field of high-altitude test bed fault diagnosis, in particular to an intelligent fault diagnosis method for a high-altitude test bed based on wavelet analysis and a multilayer overrun learning machine, and belongs to the field of intelligent fault diagnosis for a high-altitude test bed sensor system.
Background
An aerial engine high-altitude simulation test bed (called a high-altitude platform for short) is a ground test bed capable of simulating the aerial working environment condition of an engine. The high-altitude platform can perform engine high-altitude characteristic test measurement, acquire engine high-altitude performance/characteristics, identify the working reliability of engine accessories and systems under different flight environment conditions, and study and examine the structural integrity of the engine under various flight conditions. The high-altitude platform is a national strategic resource and is an essential important means and tool for autonomously developing advanced aeroengines. With the rapid increase of the types of engines and the continuous improvement of the performance, the performance and the function of the engines in the full-envelope range need to be repeatedly debugged, verified and checked in the high-altitude environment, and extremely high requirements are provided for the reliability and the safety of a measurement sensor of an air inlet regulation system of a high-altitude platform. The measurement sensor is easy to break down when operating under severe working conditions of large load, strong vibration and high-frequency use for a long time, a control system is out of order if the measurement sensor is in a light state, major safety accidents occur if the measurement sensor is in a heavy state, and the test safety of the tested engine and the safe operation of the high-altitude platform are seriously threatened. Therefore, it is necessary to develop a method for diagnosing faults of the high-altitude platform sensor, so as to achieve the purposes of real-time online diagnosis of typical fault states and effective avoidance of test risks.
In the bearing fault diagnosis based on the depth wavelet automatic encoder and the extreme learning machine (the Douchi, Guo Shuishen, scientific technology and engineering, volume 29, 20 th of 2020), a bearing fault diagnosis method based on the combination of the depth wavelet automatic encoder and the extreme learning machine is provided. The multilayer overrun learning machine method provided by the invention increases the number of hidden layers, can fully extract the intrinsic characteristic information of the fault, thereby effectively diagnosing and classifying the sensor faults, and has less calculation amount and easier online realization compared with a deep learning classification method.
Disclosure of Invention
Technical problem to be solved
In order to overcome the defect of accuracy of fault diagnosis of a sensor of the conventional high-altitude test bed of the engine, the invention provides a high-altitude test bed fault diagnosis method based on wavelet analysis and a multi-layer overrun learning machine. The method can more fully excavate the fault characteristics of the sensor, accurately diagnose the fault on line and solve the problem of the fault on-line diagnosis of the high-altitude platform sensor
Technical scheme
A high-altitude platform fault diagnosis method based on wavelet analysis and a multilayer overrun learning machine is characterized by comprising the following steps:
step 1: extracting and denoising by adopting a wavelet analysis method based on system labeled fault sample original data x (t);
1) wavelet transform coefficient obtained by adopting binary wavelet transform
Figure RE-GDA0003045230430000021
Where τ is the frequency shift factor and n is (log)2m) -5, m representing the signal length;
2) selecting a threshold value for wavelet coefficients under each decomposition scale to carry out threshold value quantization processing, filtering noise signals with the wavelet coefficients lower than the threshold value, and keeping useful signals with the wavelet coefficients higher than the threshold value; by the formula
Figure RE-GDA0003045230430000022
Calculating to obtain a threshold, wherein N is the sum of the number of wavelet coefficients obtained by performing wavelet transform decomposition on an actual measurement signal x (t), and sigma is the standard deviation of a given additional noise signal; according to Wx(2jTau) is equal to or more than T, and a range A which meets the condition j is equal to or more than j and equal to or less than B;
3) performing wavelet reconstruction of the signal according to the lowest-layer low-frequency wavelet coefficient and each high-frequency wavelet coefficient after wavelet decomposition to obtain a reconstructed signal:
Figure RE-GDA0003045230430000031
step 2: constructing a multilayer overrun learning machine, and performing online sequence learning training;
constructing a first layer of automatic encoder, and selecting N in the initial stage0Group data
Figure RE-GDA0003045230430000032
Wherein
Figure RE-GDA0003045230430000033
For the sample data after feature extraction and denoising, tiIn order to output the target of the output,
Figure RE-GDA0003045230430000034
is the number of hidden layer neurons;
1) randomly generating an input weight matrix wiAnd a bias matrix biWherein, in the step (A),
Figure RE-GDA0003045230430000035
2) computing an initial hidden layer output matrix H0Is provided with
Figure RE-GDA0003045230430000036
Where g (-) is an activation function, here a Sigmiod function;
3) calculating an initial output weight matrix beta0Is provided with
Figure RE-GDA0003045230430000037
Wherein
Figure RE-GDA0003045230430000038
Figure RE-GDA0003045230430000039
Figure RE-GDA00030452304300000310
A matrix composed of output targets;
4) setting k as 0, wherein k is the number of blocks and represents the initial learning stage;
5) let k +1 block sample set
Figure RE-GDA00030452304300000311
6) Computing a hidden layer output matrix Hk+1Is provided with
Figure RE-GDA0003045230430000041
Online sequence learning; recursively updating an output weight matrix beta with new samplesk+1Until k is N;
calculating an output weight matrix betak+1Is provided with
Figure RE-GDA0003045230430000042
Wherein
Figure RE-GDA0003045230430000043
Making k equal to k +1, and turning to the step 1) of the online learning stage until k equal to N is finished;
constructing the next layer of automated encoder with HNAs input to the i-th layer auto-encoder
Figure RE-GDA0003045230430000044
Calculating the output weight matrix of the i-th layer automatic encoder according to the recursive steps 1) to 6)
Figure RE-GDA0003045230430000045
To be provided with
Figure RE-GDA0003045230430000046
The weighted value is used as the weighted value between the ith layer and the (i + 1) th layer of the automatic encoder;
repeating the construction of the automatic encoder until the number of layers reaches q, and calculating the output weight matrix
Figure RE-GDA0003045230430000047
Completing the training of an online sequence multi-layer overrun learning machine;
and step 3: fault diagnosis for actual system data
Checking the actual data X of the system to be checkeddExtracting features and denoising by principal component analysis method to obtain
Figure RE-GDA0003045230430000048
Inputting a multi-layer ultralimit learning machine network on line based on an output weight matrix
Figure RE-GDA0003045230430000049
Calculating the on-line network output value F of the actual datak；
The input layer outputs are:
Figure RE-GDA00030452304300000410
the intermediate coding layer output is:
Figure RE-GDA00030452304300000411
the final layer output value is:
Figure RE-GDA00030452304300000412
where g (-) is an activation function, here a Sigmiod function;
handle conveyerTag values for outcoming and failed samples
Figure RE-GDA0003045230430000051
By comparison, the fault diagnosis logic is:
Figure RE-GDA0003045230430000052
a computer system, comprising: one or more processors, a computer readable storage medium, for storing one or more programs, which when executed by the one or more processors, cause the one or more processors to implement the method as described above.
A computer-readable storage medium storing computer-executable instructions for implementing the method as described above when executed.
A computer program comprising computer executable instructions for implementing a method as described above when executed.
Advantageous effects
The invention provides an intelligent fault diagnosis method based on signals for faults of pressure and temperature sensors of a high-altitude test bed, a wavelet analysis method is adopted for extracting and denoising fault features, an online multi-layer ultralimit learning machine is constructed for classifying various faults, fault categories are further diagnosed, the limitation that the existing single-layer ultralimit learning machine is low in diagnosis precision and cannot diagnose the faults of the sensors in time is broken through, and the accuracy of fault diagnosis is improved.
In addition, the invention aims at the problems of online realization of intelligent fault diagnosis and balance of calculated amount of the high-altitude platform sensor, adopts an online sequence multilayer overrun learning machine to classify the faults, ensures the real-time performance of training, reduces the calculated amount compared with a fault classification algorithm of deep learning, is beneficial to the real-time fault diagnosis in industry and expands the application range of practical engineering.
Drawings
FIG. 1 is a flow chart of the method of the present invention
Detailed Description
The invention will now be further described with reference to the following examples and drawings:
in order to improve the safety and reliability of the high-altitude platform test process, the invention provides a high-altitude platform sensor fault diagnosis method based on wavelet analysis and a multi-layer overrun learning machine, and solves the problem of online fault diagnosis of a high-altitude platform pressure and temperature sensor. The following describes a specific embodiment of the invention in combination with a sensor fault diagnosis process in a high-altitude test bed:
the air inlet pressure system of the high-altitude platform can be regarded as comprising a hydraulic servo system, a flow valve and an air inlet cavity. The gas source control system provides stable gas flow input before the valve, and ensures that the pressure before the valve of the special flow regulating valve is kept stable. Under the effect of the hydraulic servo position control mechanism, the opening degree of the special flow valve is adjusted, the flow of gas in the flow valve is controlled, meanwhile, the exhaust port of the air inlet containing cavity is influenced by the flow of an engine to generate exhaust flow change, and stable air inlet pressure can be obtained, so that the high-altitude simulation requirement of the aircraft engine in the air inlet containing cavity is met. And installing temperature sensors and pressure sensors before and after the flow valve, performing fault injection simulation on the sensors to generate sample data x (t) under different faults, and marking fault labels.
Performing a first step, extracting features and denoising by adopting a wavelet analysis method based on the original data of the system labeled fault sample;
1) wavelet transform coefficient obtained by adopting binary wavelet transform
Figure RE-GDA0003045230430000061
Where τ is the frequency shift factor and n is (log)2m)-5And m represents a signal length.
2) Selecting a threshold value for wavelet coefficient under each decomposition scale to carry out threshold value quantization processing, filtering noise signals with wavelet coefficients lower than the threshold value, and keeping useful signals with wavelet coefficients higher than the threshold value. By the formula
Figure RE-GDA0003045230430000062
And calculating to obtain a threshold, wherein N is the sum of the number of wavelet coefficients obtained by performing wavelet transform decomposition on the actual measurement signal x (t), and sigma is the standard deviation of the given additional noise signal. According to Wx(2jAnd tau) is not less than T, and the range A which meets the condition j is not less than j and not more than B is selected.
3) And performing wavelet reconstruction on the signal according to the lowest-layer low-frequency wavelet coefficient and each high-frequency wavelet coefficient after wavelet decomposition to obtain a reconstructed signal.
Figure RE-GDA0003045230430000063
And (5) executing the step two: constructing a multilayer overrun learning machine, and performing online sequence learning training;
constructing a first layer automatic encoder, and selecting N in an initial stage0Group data
Figure RE-GDA0003045230430000071
Wherein
Figure RE-GDA0003045230430000072
1) Randomly generating an input weight matrix N0And a bias matrix N0Wherein, in the step (A),
Figure RE-GDA0003045230430000073
2) computing an initial hidden layer output matrix H0Is provided with
Figure RE-GDA0003045230430000074
3) Calculating an initial output weight matrix beta0Is provided with
Figure RE-GDA0003045230430000075
Wherein
Figure RE-GDA0003045230430000076
Figure RE-GDA0003045230430000077
4) Let k be 0, k be the number of blocks, represent the initial learning phase.
5) Let k +1 block sample set
Figure RE-GDA0003045230430000078
6) Computing a hidden layer output matrix Hk+1Is provided with
Figure RE-GDA0003045230430000079
And (4) online sequence learning. Recursive updating of the hidden layer output matrix H with new samplesk+1And an output weight matrix betak+1Until k is N.
Calculating an output weight matrix betak+1Is provided with
Figure RE-GDA0003045230430000081
Wherein
Figure RE-GDA0003045230430000082
And (5) making k equal to k +1, and turning to the step (1) of the online learning stage until k equal to N is finished.
Constructing the next layer of automated encoder with HNAs input to the i-th layer auto-encoder
Figure RE-GDA0003045230430000083
Calculating the output weight matrix of the i-th layer automatic encoder by the recursive steps
Figure RE-GDA0003045230430000084
To be provided with
Figure RE-GDA0003045230430000085
As the weight value between the automatic encoders;
repeating the construction of the automatic encoder until the number of layers reaches q, and calculating the output weight matrix
Figure RE-GDA0003045230430000086
And finishing the training of the online sequence multi-layer overrun learning machine.
Step three: and carrying out fault diagnosis on the actual system data.
After the actual data of the system to be detected is subjected to principal component analysis method for feature extraction and denoising, the actual data is input into a multi-layer ultralimit learning machine network on line and is based on an output weight matrix
Figure RE-GDA0003045230430000087
Calculating the on-line network output value F of the actual datakComparing the output value with the tag value of the fault sample
Figure RE-GDA0003045230430000088
By comparison, the fault diagnosis logic is:
Figure RE-GDA0003045230430000089
while the invention has been described with reference to specific embodiments, the invention is not limited thereto, and various equivalent modifications or substitutions can be easily made by those skilled in the art within the technical scope of the present disclosure.

Claims (4)

1.一种基于小波分析和多层超限学习机的高空台故障诊断方法，其特征在于步骤如下：1. a high-altitude platform fault diagnosis method based on wavelet analysis and multi-layer ultra-limit learning machine, is characterized in that step is as follows: 步骤1：基于系统带标签的故障样本原始数据x(t)，采用小波分析法进行提取特征和去噪；Step 1: Based on the original data x(t) of the fault samples with labels in the system, the wavelet analysis method is used to extract features and denoise; 1)采用二进小波变换得到小波变换系数1) Using binary wavelet transform to obtain wavelet transform coefficients
Figure FDA0002968853110000011
Figure FDA0002968853110000011
 其中τ为频移因子，n＝(log2m)-5，m表示信号长度；where τ is the frequency shift factor, n=(log 2 m)-5, m represents the signal length; 2)对各个分解尺度下的小波系数选择一个阈值进行阈值量化处理，过滤小波系数低于阈值的噪声信号，保留小波系数高于阈值的有用信号；通过公式2) Select a threshold for the wavelet coefficients under each decomposition scale to perform threshold quantization processing, filter the noise signal whose wavelet coefficient is lower than the threshold, and retain the useful signal whose wavelet coefficient is higher than the threshold; through the formula
Figure FDA0002968853110000012
Figure FDA0002968853110000012
 计算得出阈值，式中，N为实际测量信号x(t)经过小波变换分解得到小波系数的个数总和，σ为给定的附加噪声信号的标准差；根据Wx(2j，τ)≥T选取符合条件j的范围A≤j≤B；Calculate the threshold, where N is the sum of the number of wavelet coefficients obtained by the actual measurement signal x(t) after wavelet transform decomposition, σ is the standard deviation of the given additional noise signal; according to W x (2 j , τ) ≥T selects the range A≤j≤B that meets the condition j; 3)根据小波分解后的最底层低频小波系数和各个高频小波系数，进行信号的小波重构，得到重构后的信号：3) Carry out wavelet reconstruction of the signal according to the lowest-level low-frequency wavelet coefficients and each high-frequency wavelet coefficient after wavelet decomposition, and obtain the reconstructed signal:
Figure FDA0002968853110000013
Figure FDA0002968853110000013
 步骤2：构造多层超限学习机，进行在线序列学习训练；Step 2: Construct a multi-layer ELM for online sequence learning and training; 构建第一层自动编码器，初始阶段先选取N0组数据
Figure FDA0002968853110000014
其中
Figure FDA0002968853110000015
为经过特征提取和去噪后的样本数据，ti为输出目标，
Figure FDA0002968853110000016
Figure FDA0002968853110000017
为隐含层神经元数目；Build the first layer of auto-encoder, and select N 0 groups of data in the initial stage
Figure FDA0002968853110000014
in
Figure FDA0002968853110000015
is the sample data after feature extraction and denoising, t i is the output target,
Figure FDA0002968853110000016
Figure FDA0002968853110000017
is the number of neurons in the hidden layer; 1)随机产生输入权值矩阵wi和偏置矩阵bi，其中，
Figure FDA0002968853110000018
1) Randomly generate input weight matrix w i and bias matrix b i , where,
Figure FDA0002968853110000018
 2)计算初始的隐含层输出矩阵H0，有2) Calculate the initial hidden layer output matrix H 0 , there are
Figure FDA0002968853110000021
Figure FDA0002968853110000021
 其中g(·)为激活函数，这里为Sigmiod函数；where g( ) is the activation function, here is the sigmiod function; 3)计算初始的输出权重矩阵β0，有3) Calculate the initial output weight matrix β 0 , there are
Figure FDA0002968853110000022
Figure FDA0002968853110000022
 其中in
Figure FDA0002968853110000023
Figure FDA0002968853110000023
Figure FDA0002968853110000024
Figure FDA0002968853110000024
Figure FDA0002968853110000025
为输出目标组成的矩阵；
Figure FDA0002968853110000025
A matrix composed of output targets; 4)设k＝0，k为块数，表示初始化学习阶段；4) Set k=0, k is the number of blocks, indicating the initialization learning stage; 5)设第k+1块样本集
Figure FDA0002968853110000026
5) Set the k+1th sample set
Figure FDA0002968853110000026
 6)计算隐含层输出矩阵Hk+1，有6) Calculate the hidden layer output matrix H k+1 , there are
Figure FDA0002968853110000027
Figure FDA0002968853110000027
 在线序列学习；利用新样本递推更新输出权重矩阵βk+1，直到k＝N；Online sequence learning; use new samples to recursively update the output weight matrix β k+1 until k=N; 计算输出权重矩阵βk+1，有Calculate the output weight matrix β k+1 , we have
Figure FDA0002968853110000028
Figure FDA0002968853110000028
 其中in
Figure FDA0002968853110000031
Figure FDA0002968853110000031
 令k＝k+1，转至在线学习阶段步骤1)直到k＝N结束；Let k=k+1, go to step 1) of the online learning stage until k=N ends; 构造下一层自动化编码器，以HN做为第i层自动编码器的输入
Figure FDA0002968853110000032
Construct the next layer of automatic encoder, taking H N as the input of the i-th layer of automatic encoder
Figure FDA0002968853110000032
 由以上递推步骤1)-6)计算第i层自动编码器输出权重矩阵
Figure FDA0002968853110000033
Calculate the output weight matrix of the i-th layer autoencoder from the above recursive steps 1)-6)
Figure FDA0002968853110000033
 以
Figure FDA0002968853110000034
做为第i层到第i+1层自动编码器之间的权重值；by
Figure FDA0002968853110000034
As the weight value between the i-th layer and the i+1-th layer auto-encoder; 重复构造自动编码器直到层数达到q，计算出输出权重矩阵
Figure FDA0002968853110000035
完成在线序列多层超限学习机训练；Construct the autoencoder repeatedly until the number of layers reaches q, and calculate the output weight matrix
Figure FDA0002968853110000035
Complete online sequence multi-layer EFL training; 步骤3：针对实际系统数据，进行故障诊断Step 3: Perform fault diagnosis against actual system data 把待检测的系统实际数据Xd，进行主元分析法提取特征和去噪后得到
Figure FDA0002968853110000036
在线输入多层超限学习机网络，基于输出权重矩阵
Figure FDA0002968853110000037
计算实际数据的在线网络输出值Fk；The actual data X d of the system to be detected is extracted by principal component analysis and denoised to obtain
Figure FDA0002968853110000036
Online input multi-layer ELM network, based on output weight matrix
Figure FDA0002968853110000037
Calculate the online network output value F k of the actual data; 输入层输出为：
Figure FDA0002968853110000038
The output of the input layer is:
Figure FDA0002968853110000038
 中间编码层输出为：
Figure FDA0002968853110000039
The output of the intermediate coding layer is:
Figure FDA0002968853110000039
 最后一层输出值为：
Figure FDA00029688531100000310
其中g(·)为激活函数，这里为Sigmiod函数；The output value of the last layer is:
Figure FDA00029688531100000310
where g( ) is the activation function, here is the sigmiod function; 把输出值和故障样本的标签值
Figure FDA00029688531100000311
相比较，故障诊断逻辑为：Put the output value and the label value of the fault sample
Figure FDA00029688531100000311
In comparison, the fault diagnosis logic is:
Figure FDA00029688531100000312
Figure FDA00029688531100000312
 2.一种计算机系统，其特征在于包括：一个或多个处理器，计算机可读存储介质，用于存储一个或多个程序，其中，当所述一个或多个程序被所述一个或多个处理器执行时，使得所述一个或多个处理器实现权利要求1所述的方法。2. A computer system, characterized by comprising: one or more processors, and a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more programs When executed by a plurality of processors, the one or more processors are caused to implement the method of claim 1 . 3.一种计算机可读存储介质，其特征在于存储有计算机可执行指令，所述指令在被执行时用于实现权利要求1所述的方法。3. A computer-readable storage medium, characterized by storing computer-executable instructions that, when executed, are used to implement the method of claim 1 . 4.一种计算机程序，其特征在于包括计算机可执行指令，所述指令在被执行时用于实现权利要求1所述的方法。4. A computer program characterized by comprising computer-executable instructions which, when executed, are used to implement the method of claim 1 .
CN202110258961.2A 2021-03-10 2021-03-10 High-altitude platform fault diagnosis method based on wavelet analysis and multi-layer overrun learning machine Pending CN113033632A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110258961.2A CN113033632A (en) 2021-03-10 2021-03-10 High-altitude platform fault diagnosis method based on wavelet analysis and multi-layer overrun learning machine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110258961.2A CN113033632A (en) 2021-03-10 2021-03-10 High-altitude platform fault diagnosis method based on wavelet analysis and multi-layer overrun learning machine

Publications (1)

Publication Number Publication Date
CN113033632A true CN113033632A (en) 2021-06-25

Family

ID=76468906

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110258961.2A Pending CN113033632A (en) 2021-03-10 2021-03-10 High-altitude platform fault diagnosis method based on wavelet analysis and multi-layer overrun learning machine

Country Status (1)

Country Link
CN (1) CN113033632A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114036973A (en) * 2021-10-12 2022-02-11 浙江工业大学 Series arc fault identification method of extreme learning machine based on dynamic online sequence

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108153987A (en) * 2018-01-08 2018-06-12 太原理工大学 A kind of hydraulic pump Multiple faults diagnosis approach based on the learning machine that transfinites
CN109782091A (en) * 2019-01-30 2019-05-21 西华大学 Fault diagnosis method of multilevel inverter based on deep wavelet extreme learning machine
CN110907177A (en) * 2019-12-10 2020-03-24 北京交通大学 Bearing fault diagnosis method based on layered extreme learning machine
CN110927535A (en) * 2019-11-12 2020-03-27 国网天津市电力公司电力科学研究院 Evaluation method of partial discharge severity of power equipment based on extreme learning machine
US20200311544A1 (en) * 2017-09-28 2020-10-01 Agency For Science, Technology And Research Self-assessing deep representational units

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200311544A1 (en) * 2017-09-28 2020-10-01 Agency For Science, Technology And Research Self-assessing deep representational units
CN108153987A (en) * 2018-01-08 2018-06-12 太原理工大学 A kind of hydraulic pump Multiple faults diagnosis approach based on the learning machine that transfinites
CN109782091A (en) * 2019-01-30 2019-05-21 西华大学 Fault diagnosis method of multilevel inverter based on deep wavelet extreme learning machine
CN110927535A (en) * 2019-11-12 2020-03-27 国网天津市电力公司电力科学研究院 Evaluation method of partial discharge severity of power equipment based on extreme learning machine
CN110907177A (en) * 2019-12-10 2020-03-24 北京交通大学 Bearing fault diagnosis method based on layered extreme learning machine

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
PAN DONG ET AL.: "Short-term Load Forecasting Based on EEMD-approximate Entropy and ELM", 2019 IEEE SUSTAINABLE POWER AND ENERGY CONFERENCE (ISPEC) *
何威;何怡刚;罗旗舞;李志刚;邓芳明;张朝龙;: "基于连续小波奇异熵的模拟电路故障诊断新方法", 电子测量与仪器学报, no. 12 *
郝丽娜等: "基于多层超限学习机的滚动轴承故障诊断方法", 科学技术与工程, vol. 17, no. 14, pages 1 - 3 *
陶沙沙等: "基于深度小波自动编码器和极限学习机的轴承故障诊断", 科学技术与工程, vol. 20, no. 29, pages 1 - 4 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114036973A (en) * 2021-10-12 2022-02-11 浙江工业大学 Series arc fault identification method of extreme learning machine based on dynamic online sequence

Similar Documents

Publication Publication Date Title
CN109472110B (en) A Method for Predicting the Remaining Service Life of Aeroengine Based on LSTM Network and ARIMA Model
CN110807257A (en) Method for predicting residual life of aircraft engine
CN113158445B (en) Prediction algorithm for residual service life of aero-engine with convolution memory residual error self-attention mechanism
CN108334907B (en) A kind of complex equipment point method for detecting abnormality and system based on deep learning
CN113469060A (en) Multi-sensor fusion convolution neural network aeroengine bearing fault diagnosis method
CN108256173B (en) Gas circuit fault diagnosis method and system for dynamic process of aircraft engine
CN113869563A (en) A method for predicting the remaining life of aeronautical turbofan engines based on fault feature migration
CN105700518A (en) Fault diagnosis method during industrial process
CN114266278B (en) Dual-attention network-based equipment residual service life prediction method
CN112785091A (en) Method for performing fault prediction and health management on oil field electric submersible pump
CN111637045B (en) Fault diagnosis method for air compressor of ocean platform
CN114297918A (en) Remaining life prediction method of aero-engine based on full attention deep network and dynamic ensemble learning
CN113268833A (en) Migration fault diagnosis method based on deep joint distribution alignment
CN111122811A (en) A fault monitoring method for sewage treatment process based on OICA and RNN fusion model
CN107727333A (en) A kind of diagnostic method for hydraulic cylinder leakage analyzing
CN115081484A (en) Fault diagnosis method of aero-engine sensor based on CRJ-OSELM algorithm
CN116911191A (en) Aeroengine gas circuit system modeling method based on improved PSO optimization BiLSTM
CN112560252A (en) Prediction method for residual life of aircraft engine
Li et al. Structural health monitoring data anomaly detection by transformer enhanced densely connected neural networks
CN113033632A (en) High-altitude platform fault diagnosis method based on wavelet analysis and multi-layer overrun learning machine
CN118820756B (en) Bearing residual life prediction system and method based on large language model
CN115014789A (en) A CNN-GCN-based dual-sensor aero-engine casing fault source acoustic emission localization method
Zhi-hong et al. Sensor fault diagnosis based on wavelet analysis and LSTM neural network
CN114037012A (en) Flight data anomaly detection method based on correlation analysis and deep learning
CN112766408A (en) Aircraft tiny fault diagnosis method based on principal component analysis and multi-layer ultralimit learning machine

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: 20210625