CN113869266B - Centrifugal compressor rotating stall early warning method based on big data analysis - Google Patents
Centrifugal compressor rotating stall early warning method based on big data analysis Download PDFInfo
- Publication number
- CN113869266B CN113869266B CN202111187895.0A CN202111187895A CN113869266B CN 113869266 B CN113869266 B CN 113869266B CN 202111187895 A CN202111187895 A CN 202111187895A CN 113869266 B CN113869266 B CN 113869266B
- Authority
- CN
- China
- Prior art keywords
- neural network
- centrifugal compressor
- layer
- signal
- sparse
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000007405 data analysis Methods 0.000 title claims abstract description 6
- 238000013528 artificial neural network Methods 0.000 claims abstract description 57
- 230000009467 reduction Effects 0.000 claims abstract description 22
- 238000012545 processing Methods 0.000 claims abstract description 4
- 230000010349 pulsation Effects 0.000 claims description 29
- 210000002569 neuron Anatomy 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 2
- 238000012423 maintenance Methods 0.000 abstract description 5
- 238000007781 pre-processing Methods 0.000 abstract description 3
- 238000003745 diagnosis Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 239000012530 fluid Substances 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000011065 in-situ storage Methods 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000008531 maintenance mechanism Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005272 metallurgy Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000010992 reflux Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Signal Processing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Control Of Positive-Displacement Air Blowers (AREA)
- Structures Of Non-Positive Displacement Pumps (AREA)
Abstract
The invention provides a centrifugal compressor rotating stall early warning method based on big data analysis, which comprises the steps of collecting real-time operation parameter information of a centrifugal compressor, carrying out signal preprocessing, adopting a stack sparse noise reduction self-coding neural network, simultaneously extracting signal deep characteristic information by adding Gaussian white noise with different intensities, adding a Softmax classifier at the last layer of the neural network, identifying flow at different rotating speeds, determining critical flow values of rotating stall, and realizing early warning of rotating stall early states of the centrifugal compressor at different rotating speeds. According to the state identification and prediction results, the processing of massive centrifugal compressor data in engineering is realized, and a foundation is laid for engineering application. The invention can monitor the running state of the centrifugal compressor in real time, provides intelligent operation and maintenance management decisions for the centrifugal compressor for enterprises, improves the running efficiency of equipment and realizes the maximization of the benefits of the enterprises.
Description
Technical Field
The invention relates to a centrifugal compressor rotating stall early warning method based on big data analysis, and belongs to the technical field of intelligent diagnosis of important equipment.
Background
The centrifugal compressor has the advantages of simple structure, large flow, stable operation, small volume, convenient maintenance, no pollution to gas and the like, and is widely applied to the fields of petrochemical industry, metallurgy, aviation, electric power and the like. Compressors are generally divided into two categories, according to the different ways in which compressed gas increases in gas pressure: positive displacement compressors and turbine compressors. In general, the positive displacement compressor is applied to occasions with small and medium flow, and the working condition with large flow is more suitable for using the turbine compressor. The turbine compressor can be further divided into the following four types: centrifugal compressors, axial compressors, diagonal compressors and compound compressors. However, with the continuous improvement of the industrial level in recent years, the requirements of the actual production on the performance parameters of the centrifugal compressor are higher and higher, and the compressor is developed towards the directions of high pressure ratio, high rotating speed, large flow and slimness, and a series of problems affecting the stability of the compressor are accompanied.
Based on in-situ statistical analysis, most accidents are caused by fatigue failure of the blade due to fluid pulsation or unsteady fluid excitation. When the intake air flow rate of the large centrifugal compressor is reduced to a certain degree, the internal flow field of the large centrifugal compressor has a strong unstable pressure pulsation phenomenon, and the large vibration of the rotor part is often caused, so that fatigue failure occurs. Such pressure pulsation phenomenon is called a condition unsteady, i.e., a flow instability problem, including phenomena such as rotating stall and surge. Rotating stall is an unstable flow phenomenon that significantly reduces compressor performance, and as testing techniques continue to develop, many methods of detecting rotating stall characteristics of compressors have emerged, however, stall characteristics also appear to be diverse, depending on compressor type. In the face of the problems existing in the identification of the rotating stall characteristics of the current compressor, the method for detecting the stall frequency spectrum characteristics of the compressor has great significance in establishing an early failure early warning mechanism of the rotating stall, exploring an intelligent diagnosis method and improving the safe operation capability of important equipment.
Disclosure of Invention
The invention provides an early warning method for rotating stall of an industrial centrifugal compressor at different rotating speeds, which aims to solve the technical problems that: in engineering application, the signal processing method of mass centrifugal compressor data replaces the intelligent diagnosis method of original pressure pulsation data and rotating stall critical flow state with a small amount of effective characteristic information.
The technical scheme of the invention is as follows:
According to the early warning method of rotating stall of the centrifugal compressor based on big data analysis, real-time operation parameter information of the centrifugal compressor is collected and signal preprocessing is carried out, a stack sparse noise reduction self-coding neural network (SSDAE) is adopted, meanwhile, deep signal characteristic information is extracted by adding Gaussian white noise with different intensities, a Softmax classifier is added at the last layer of the neural network, the flow at different rotating speeds is identified, and the critical flow value of rotating stall is determined; the method comprises the following specific steps:
Step 1: and (3) processing mass engineering data:
and collecting pressure pulsation signals of the centrifugal compressor at different flow rates through a pressure pulsation sensor. The collected pressure pulsation signal is divided into signal segments with equal time intervals, and one signal segment is used as one sample, so that the critical speed of the centrifugal compressor at which rotating stall occurs is completely described.
Step 2: building a depth stack sparse noise reduction self-coding neural network:
A sparse automatic encoder neural network (SAE) is used as a characteristic representation method for learning original data, and consists of an input layer, an implicit layer and an output layer. The input layer neurons are collected centrifugal compressor pressure pulsation signals, the output layer neurons are centrifugal compressor signals reconstructed by a neural network, and the number of the middle hidden layer neurons is smaller than that of the input layer neurons and the output layer neurons. During network training, each training sample is passed through the network to generate a new signal at the output layer, and the purpose of network learning is to minimize the difference between the output layer reconstructed centrifugal compressor signal and the input original centrifugal compressor signal. In a self-encoder neural network, a mean square error function is typically chosen to represent the difference between the output signal and the input signal. The mean square error function is expressed as:
wherein x is an input signal, z is a neural network output signal, and m is the number of samples.
In order to prevent the over-fitting problem, noise is added to the input data of the input layer on the basis of the SAE network, so that a sparse noise reduction self-coding neural network (SDAE) is formed, and the characteristic learning capability of the neural network is further enhanced. And setting the value of the input layer node to 0 according to a certain probability for the original input data, so as to obtain a signal containing noise. In order to obtain higher-level feature expression, SDAEs are stacked layer by layer in a deep network structure mode to form a model structure composed of two or more SDAEs which are connected up and down, namely a stacked noise reduction automatic encoder neural network (SSDAE) is formed, and the features of signals are extracted more completely.
To meet the sparsity constraint, SSDAE neural networks average activation is taken to be a value near zero. The sparse limitation has various forms, and the penalty factors of neurons of an implicit layer generally select KL sparse penalty terms, wherein the penalty terms are as follows:
Where beta is the sparse penalty, s is the second layer neuron, Is the average activation quantity of the neuron, a j represents the activation quantity of the neuron, and the object of the sparse term is to make/>With/>The difference between p and p increases and the value of the penalty factor will rise sharply. The weight decay function is used to avoid overfitting, as shown in the following equation,
Where λ is a weight adjustment factor, W is a neural network weight parameter, and s is a current neural network layer number.
Therefore, the loss function of the stack sparse noise reduction self-encoding neural network (SSDAE) of the centrifugal compressor pressure pulsation signal is written as:
L＝E+Jweight+Jsparse
step 3: training of neural networks:
The parameter updating of the stack sparse noise reduction self-coding neural network adopts a quasi-Newton method (L-BFGS). The quasi-Newton algorithm is suitable for large-scale numerical calculation, has the characteristic of high convergence rate of Newton method, does not need to store a Hesse matrix like the Newton method, saves a large amount of space and calculation resources, enables parameters of a neural network to be continuously and iteratively updated through an L-BFGS back propagation algorithm, deeply learns and extracts the characteristics of pressure pulsation signals of the centrifugal compressor, and ensures the accuracy of critical flow identification of rotating stall of the centrifugal compressor.
Step 4: and (3) carrying out centrifugal compressor rotating stall critical flow state identification by using the trained neural network:
And (2) taking the centrifugal compressor pressure pulsation signals acquired in the step (1) and containing various different flow working conditions as input samples, inputting the input samples into the stack sparse noise reduction self-coding neural network obtained in the step (2), and further improving the characteristic learning capability of the neural network on the signals by adding Gaussian white noise signals with different intensities. Initializing neural network parameters, and determining the optimal parameters of each layer of network by reverse iterative optimization of a quasi-Newton method. The stack sparse noise reduction self-coding neural network takes the minimum difference value between an input sample and an output sample as an objective function. And finally, adding a softmax classification layer at the last layer of the neural network, and intelligently identifying pressure pulsation signals of the centrifugal compressor under different flow rates. And according to the classification result, the pressure pulsation signals of the centrifugal compressor under different flow rates are intelligently identified, the critical flow rate value of rotating stall is judged, the utilization rate and the action efficiency of the centrifugal compressor are improved, the downtime is reduced, and the maximum utilization of enterprises is realized.
The beneficial effects of the invention are as follows: the invention collects pressure pulsation signals of the centrifugal compressor under different flow conditions, and prepares one-dimensional time domain signals into two-dimensional signal samples through processing the pressure pulsation signals, thereby covering various flow conditions. And a double hidden layer stack sparse noise reduction self-coding neural network is built again, gaussian white noise signals with different intensities are added, the characteristic learning capacity of the neural network is further improved, and intelligent operation and maintenance of the centrifugal compressor are realized. The data acquisition part establishes a mass database of different flow working conditions of the centrifugal compressor, and provides a data basis for intelligent analysis of big data of the centrifugal compressor; the intelligent flow state diagnosis part intelligently identifies the rotating stall critical flow state of the centrifugal compressor through the characteristic learning capability of the neural network on the signals, thereby establishing an operation and maintenance mechanism with the addition of predictive operation and maintenance and post-operation and maintenance, improving the production efficiency of enterprises and saving economic cost.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
FIG. 2 is a workflow diagram of a stacked sparse noise reduction self-encoding neural network of the present invention.
FIG. 3 is a schematic diagram of identification accuracy of rotating stall critical speed of a centrifugal compressor in an embodiment of the invention.
Detailed Description
The following describes the embodiments of the present invention further with reference to the drawings and technical schemes.
The flow of the method of the invention is shown in figure 1, and the embodiment carries out experiments under different flow rates of a large number of centrifugal compressors, and comprises the following steps:
Step 1: sample data acquisition and preprocessing:
Collecting pressure pulsation signals of the centrifugal compressor under different flow rates by using a pressure pulsation sensor; dividing the signals of the mass centrifugal compressors after normalization pretreatment to obtain a plurality of time series sample data with equal length; and combining the plurality of sample data obtained by segmentation with corresponding different flows to form a training set and a testing set. The experiment is carried out on a model-level centrifugal compressor test bed. The centrifugal compressor consists of an inlet pipeline, a guide vane adjusting device, an impeller, a diffuser, a reflux device and the like, wherein the impeller is a semi-open impeller, the diffuser is a full-height blade diffuser, the inlet guide vane adjusting device is uniformly distributed symmetrical wing type guide vanes, pressure pulsation signals under 11 flow rates of the centrifugal compressor are collected, 200 samples are produced at each flow rate, each sample contains 2048 sample points, and 2200 multiplied by 2048 samples are formed in a conformal mode.
TABLE 1 working condition parameters
Step 2: building a stack sparse noise reduction automatic encoder model;
As shown in fig. 2, the stack sparse noise reduction self-coding neural network is composed of a signal input layer, a feature learning layer and a state classification layer;
The input layer is used for preparing 2200×2048 two-dimensional samples of the acquired mass pressure pulsation signals and inputting the samples into the neural network. The number of neurons in the input layer of the network was set to 2048, consistent with the number of sample points per sample input. The self-coding network adopts double hidden layers, the number of the hidden layer neurons is set according to a neuron decreasing mode, parameters of the neural network are initialized, and the intensity of Gaussian white noise is set.
Step 3: training a model;
The training process of the whole model is divided into two stages of unsupervised pre-training and supervised fine tuning;
And in the stage of unsupervised pre-training, the difference value of the input signal and the output signal is used as a target value, the mean square error is used as a loss function to carry out unsupervised training on the stack sparse noise reduction self-coding built in the step 2, and meanwhile, gaussian white noise signals with different intensities are added to enhance the characteristic learning capability of the network. After the end of the unsupervised pre-training phase, the neural network is subjected to supervised fine tuning according to the labels of the samples. In the whole training process of the model, taking the difference value between the output signal and the input signal as the judgment basis of the model identification effect; after the training process is finished, the model with the minimum loss value and the highest recognition accuracy is selected as the final model.
Step 4: carrying out intelligent recognition on the rotating stall critical flow state of the centrifugal compressor at different rotating speeds;
the method is characterized in that a Softmax classifier is added to the last layer of the neural network constructed in the step 2, the Softmax classifier is a generalized variant of a logistic regression classifier (LR) facing multi-classification tasks, the Softmax is simple to calculate, the result is easy to understand, and the method is widely applied to machine learning and convolutional neural networks. The sample is subjected to intelligent recognition on the pressure pulsation signals of the centrifugal compressor under different flow rates through the last classification layer, the recognition accuracy is shown in figure 3, the critical flow value of the rotating stall of the centrifugal compressor is determined, and early warning of the rotating stall of the centrifugal compressor is realized.
Claims (1)
1. The early warning method for rotating stall of the centrifugal compressor based on big data analysis is characterized in that real-time operation parameter information of the centrifugal compressor is collected and signal pretreatment is carried out, a stack sparse noise reduction self-coding neural network SSDAE is adopted, meanwhile, deep signal characteristic information is extracted by adding Gaussian white noise with different intensities, a Softmax classifier is added at the last layer of the neural network, the flow at different rotating speeds is identified, and the critical flow value of the rotating stall is determined; the method comprises the following specific steps:
Step 1: and (3) processing mass engineering data:
Collecting pressure pulsation signals of the centrifugal compressor at different flow rates through a pressure pulsation sensor; dividing the collected pressure pulsation signal into signal segments with equal time intervals, wherein one signal segment is used as a sample, so that the critical speed of the centrifugal compressor for rotating stall is completely described;
Step 2: building a depth stack sparse noise reduction self-coding neural network:
The sparse automatic encoder neural network SAE is used as a characteristic representation method for learning original data, and consists of an input layer, an implicit layer and an output layer; the input layer neurons are collected centrifugal compressor pressure pulsation signals, the output layer neurons are centrifugal compressor signals reconstructed by a neural network, and the number of the middle hidden layer neurons is smaller than that of the input layer neurons and the output layer neurons; during network training, each training sample generates a new signal at the output layer through the network, and the purpose of network learning is to minimize the difference between the centrifugal compressor signal reconstructed at the output layer and the original centrifugal compressor signal input; in the self-encoder neural network, a mean square error function is selected to represent the difference between an output signal and an input signal; the mean square error function is expressed as:
Wherein x is an input signal, z is a neural network output signal, and m is the number of samples;
In order to prevent the over-fitting problem, noise is added to the input data of the input layer on the basis of an SAE network, so that a sparse noise reduction self-coding neural network SDAE is formed, and the characteristic learning capacity of the neural network is further enhanced; for the original input data, setting the value of the input layer node to 0 with a certain probability, and obtaining a signal containing noise; in order to obtain higher-level feature expression, SDAEs are stacked layer by layer in a deep network structure form to form a model structure consisting of two or more SDAEs which are connected up and down, namely a stacked noise reduction automatic encoder neural network SSDAE is formed, and the features of signals are extracted more completely;
To meet the sparsity constraint, the SSDAE neural network average activation is taken to be a value near zero; the sparse limitation has various forms, the penalty factors of the neurons of the hidden layer select KL sparse penalty terms, and the penalty terms are as follows:
Where beta is the sparse penalty, s is the second layer neuron, Is the average activation quantity of the neuron, a j represents the activation quantity of the neuron, and the object of the sparse term is to make/>With/>The difference between p and p increases and the value of the penalty factor will rise sharply; the weight decay function is used to avoid overfitting, as shown in the following equation,
Wherein lambda is a weight adjusting factor, W is a neural network weight parameter, and s is the current neural network layer number;
therefore, the loss function of the stack sparse noise reduction self-encoding neural network SSDAE for the centrifugal compressor pressure pulsation signal is written as:
L＝E+Jweight+Jsparse
step 3: training of neural networks:
the parameter updating of the stack sparse noise reduction self-coding neural network adopts a quasi-Newton method L-BFGS, and the parameter of the neural network is continuously and iteratively updated through an L-BFGS back propagation algorithm, so that the characteristics of the pressure pulsation signal of the centrifugal compressor are deeply learned and extracted, and the accuracy of the critical flow identification of the rotating stall of the centrifugal compressor is ensured;
step 4: and (3) carrying out centrifugal compressor rotating stall critical flow state identification by using the trained neural network:
Taking the centrifugal compressor pressure pulsation signals acquired in the step 1 and containing various different flow working conditions as input samples, inputting the input samples into the stack sparse noise reduction self-coding neural network obtained in the step 2, and further improving the characteristic learning capacity of the neural network on the signals by adding Gaussian white noise signals with different intensities; initializing neural network parameters, and determining the optimal parameters of each layer of network by reverse iterative optimization of a quasi-Newton method;
the stack sparse noise reduction self-coding neural network takes the minimum difference value between an input sample and an output sample as an objective function;
Finally, a softmax classification layer is added to the last layer of the neural network, and pressure pulsation signals of the centrifugal compressor under different flow rates are intelligently identified; and according to the classification result, intelligently identifying pressure pulsation signals of the centrifugal compressor at different flow rates, and judging the critical flow value of the rotating stall.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111187895.0A CN113869266B (en) | 2021-10-12 | 2021-10-12 | Centrifugal compressor rotating stall early warning method based on big data analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111187895.0A CN113869266B (en) | 2021-10-12 | 2021-10-12 | Centrifugal compressor rotating stall early warning method based on big data analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113869266A CN113869266A (en) | 2021-12-31 |
| CN113869266B true CN113869266B (en) | 2024-05-10 |
Family
ID=78999088
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111187895.0A Active CN113869266B (en) | 2021-10-12 | 2021-10-12 | Centrifugal compressor rotating stall early warning method based on big data analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113869266B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101887479A (en) * | 2010-07-23 | 2010-11-17 | 华南理工大学 | Rapid diagnosis method for rotating stall of axial flow compressor based on determined learning theory |
| CN107016241A (en) * | 2017-04-05 | 2017-08-04 | 重庆交通大学 | Based on the rotating machinery lifetime stage recognition methods for adding sample enhancing depth own coding learning network of making an uproar |
| EP3483799A1 (en) * | 2017-11-10 | 2019-05-15 | General Electric Company | Systems and methods to generate an asset workscope |
| WO2021128510A1 (en) * | 2019-12-27 | 2021-07-01 | 江苏科技大学 | Bearing defect identification method based on sdae and improved gwo-svm |
| CN113358346A (en) * | 2021-06-07 | 2021-09-07 | 沈阳理工大学 | Gas valve fault diagnosis method based on wavelet packet decomposition and BP neural network |
-
2021
- 2021-10-12 CN CN202111187895.0A patent/CN113869266B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101887479A (en) * | 2010-07-23 | 2010-11-17 | 华南理工大学 | Rapid diagnosis method for rotating stall of axial flow compressor based on determined learning theory |
| CN107016241A (en) * | 2017-04-05 | 2017-08-04 | 重庆交通大学 | Based on the rotating machinery lifetime stage recognition methods for adding sample enhancing depth own coding learning network of making an uproar |
| EP3483799A1 (en) * | 2017-11-10 | 2019-05-15 | General Electric Company | Systems and methods to generate an asset workscope |
| WO2021128510A1 (en) * | 2019-12-27 | 2021-07-01 | 江苏科技大学 | Bearing defect identification method based on sdae and improved gwo-svm |
| CN113358346A (en) * | 2021-06-07 | 2021-09-07 | 沈阳理工大学 | Gas valve fault diagnosis method based on wavelet packet decomposition and BP neural network |
Non-Patent Citations (2)
| Title |
|---|
| 基于MPE-CEEMD算法的轴流压缩机旋转失速状态识别;姚金鑫;杨树华;王培林;太兴宇;李宏坤;;风机技术;20200226(第01期);全文 * |
| 基于改进堆叠降噪自编码的滚动轴承故障分类;侯文擎;叶鸣;李巍华;;机械工程学报;20171211(第07期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113869266A (en) | 2021-12-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110849626B (en) | Self-adaptive sparse compression self-coding rolling bearing fault diagnosis system | |
| CN114997051B (en) | Aero-engine life prediction and health assessment method based on transfer learning | |
| Li et al. | Fault diagnosis of rotating machinery based on combination of deep belief network and one-dimensional convolutional neural network | |
| Sakthivel et al. | Vibration based fault diagnosis of monoblock centrifugal pump using decision tree | |
| CN110006649B (en) | Bearing fault diagnosis method based on improved ant lion algorithm and support vector machine | |
| CN107305601B (en) | Method for analyzing efficiency factors of oil pumping well system | |
| CN108549962B (en) | Wind power prediction method based on historical segmented sequence search and time sequence sparsification | |
| CN110263834B (en) | Method for detecting abnormal value of new energy power quality | |
| CN113705869B (en) | Electromechanical equipment few-sample degradation trend prediction method of unsupervised meta-learning network | |
| CN108961468B (en) | Ship power system fault diagnosis method based on integrated learning | |
| CN112901449A (en) | Air compressor system energy consumption optimization method based on machine learning | |
| CN114218872A (en) | Method for predicting remaining service life based on DBN-LSTM semi-supervised joint model | |
| CN112288157A (en) | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning | |
| CN112749840A (en) | Method for acquiring reference value of energy efficiency characteristic index of thermal power generating unit | |
| CN113535515A (en) | High-efficiency dynamic acquisition method based on Zabbix monitoring system | |
| Yu et al. | Fault diagnosis method for submersible reciprocating pumping unit based on deep belief network | |
| CN113869266B (en) | Centrifugal compressor rotating stall early warning method based on big data analysis | |
| CN114718861A (en) | Intelligent diagnosis method for working condition of screw pump well based on deep learning | |
| Shan et al. | Semisupervised fault diagnosis of gearbox using weighted graph-based label propagation and virtual adversarial training | |
| CN116821828A (en) | Multi-dimensional time sequence prediction method based on industrial data | |
| CN116011351A (en) | Oil well reasonable sinking degree determining method based on clustering algorithm and Widedeep network | |
| CN116070126A (en) | Aviation plunger pump oil distribution disc abrasion detection method and system based on countermeasure self-supervision | |
| CN116127831A (en) | Soft measurement method for difficult-to-measure parameters of heavy gas turbine | |
| CN117370870B (en) | Knowledge and data compound driven equipment multi-working condition identification and performance prediction method | |
| CN115906591B (en) | XGBoost network-based oil well working fluid level calculation method |
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 |