CN110866314B - Method for predicting residual life of rotating machinery of multilayer bidirectional gate control circulation unit network - Google Patents
Method for predicting residual life of rotating machinery of multilayer bidirectional gate control circulation unit network Download PDFInfo
- Publication number
- CN110866314B CN110866314B CN201911008163.3A CN201911008163A CN110866314B CN 110866314 B CN110866314 B CN 110866314B CN 201911008163 A CN201911008163 A CN 201911008163A CN 110866314 B CN110866314 B CN 110866314B
- Authority
- CN
- China
- Prior art keywords
- network
- state
- residual life
- gru
- constructing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a method for residual life of a rotary machine of a multilayer bidirectional gating circulation unit network, which comprises the steps of collecting vibration signals; constructing a health index; constructing a network training set; constructing a multi-layer bidirectional gating circulation unit network; training a multi-layer bidirectional gating circulation unit network; network testing, estimation of residual life and acquisition of a confidence interval; and (4) predicting and evaluating the residual life. The method combines the advantages of strong feature extraction capability of deep learning, utilizes the neural network of the bidirectional gating circulation unit to carry out regression prediction, and obtains the confidence interval of the residual life by a Bootstrap method. Aiming at the problems that the model precision of a recurrent neural network model is sensitive to the value of the learning rate in the training process, and the prediction performance of the model is influenced by overhigh and overlow, the natural index is utilized to attenuate the learning efficiency to train the neural network efficiently. The method can accurately predict the residual life and the confidence interval of the rotary machine, can greatly reduce expensive unscheduled maintenance, and avoids the occurrence of disasters.
Description
Technical Field
The invention relates to a technology for predicting the residual life of a rotary machine, in particular to a method for predicting the residual life of the rotary machine of a multilayer bidirectional gating circulation unit network.
Background
Because of the development of advanced sensor and computer technologies, which accumulate large amounts of condition monitoring data in industrial production, data-driven methods have found widespread use in predicting the remaining life of mechanical equipment, because they can utilize condition monitoring data to quantify the degradation process, rather than building an accurate system model that is not readily available.
The fault prediction and health management comprises four stages of fault detection, fault diagnosis, residual life prediction and health management. When a fault is discovered or diagnosed, the machine is typically shut down as soon as possible to avoid catastrophic consequences. Performing such actions often occurs at inconvenient times, often resulting in significant time and economic losses. Therefore, maintenance strategies must be scheduled in a predictive manner rather than a diagnostic manner. Accurate remaining life prediction that allows for efficient mechanical predictive maintenance, thereby reducing costly unplanned mechanical repairs. From this perspective, it is important to consider the remaining life prediction for effective mechanical predictive maintenance. Rotating parts, such as bearings, gears, rotors and the like, which are common in industrial fields, are important components in rotating machinery, and the health condition of the rotating parts directly influences whether the rotating machinery can normally operate. The critical parts are seriously damaged, so that production shutdown is caused, huge economic loss is caused, and therefore, the accurate prediction of the residual service life of mechanical equipment is of great significance to the safe and reliable operation of the equipment.
The existing residual life prediction method mainly has the following problems: (1) The deep learning method can effectively mine hidden characteristics of sensor data, and provides better point estimation for residual life prediction. The prediction results often vary greatly due to measurement noise and model parameters. It is not sufficient to perform only a point estimation of the remaining life. In order to express the uncertainty of the prediction result, not only the determined predicted value of the residual life needs to be calculated, but also a confidence interval of the residual life needs to be calculated; (2) The deep learning method is used for predicting the residual life by adopting a fixed learning efficiency training network, so that the efficiency is lower; (3) Model-based methods attempt to build a mathematical or physical model describing the degradation process of the machine and update the model parameters with measured data, making it difficult to find an accurate model to describe the degradation process of the rotating machine in practice.
Disclosure of Invention
The invention aims to solve the technical problem of the prior art and provides a method for predicting the residual service life of the rotary machine of the multilayer bidirectional gated circulation unit network.
In order to solve the technical problems, the invention adopts the technical scheme that: the method for predicting the residual life of the rotating machinery of the multilayer bidirectional gating circulation unit network comprises the following steps:
step 1, collecting vibration signals: vibration signals of key parts of a rotating machine are acquired.
Step 2, construction of health indexes: a health index describing a degradation process of a key component of the rotating machine is constructed.
And 3, constructing a training and testing data set by using a sliding window technology.
Step 4, constructing a network: an MBi-GRU neural network is constructed from a single bidirectional GRU network stack.
And 5, constructing an ensemble learning network by a Bootstrap method to obtain the uncertainty expression of the prediction result. The input and output of the model are the data acquired in the step (3).
Step 6, training the network: training of the entire model is trained by minimizing the loss function through error back transfer, and the invention proposes training the MBi-GRU neural network with natural decay learning efficiency.
And 7, acquiring a residual life mean value and a confidence interval: and (5) inputting the data of the test set into the network trained in the step (5), wherein the corresponding output of the test set is the predicted value and the confidence interval of the residual life which are expected to be obtained.
And 8, evaluating a residual life prediction result: the invention adopts two indexes to evaluate: mean square error and predicted absolute error.
The invention is further improved in that: in the step 2, original features are extracted from the vibration signals. Selecting features with large trend values to form feature subsets, and finally adopting an unsupervised SOM algorithm to construct health indexes for evaluating the whole service life of the bearing from the selected feature subsets;
the invention is further improved in that: in said step 4, the refresh gate controls the current input data x at the same time t And previous memory information h t-1 Outputting a value z between 0 and 1 t The calculation formula is
z t ＝σ(W z [h t-1 ,x t ]+b x ) (1)
Where x is the input data, h is the output of the GRU unit, r is the reset gate, and z is the refresh gateR and z together control how h is hidden from the previous state t-1 Calculating to obtain new hidden state h t 。
z t Determine how much to put h t-1 The next state is passed on, which is obtained from equation (1). Where σ is sigmoid function, W z To update the door weights, b z Is an offset. Reset gate control h t-1 For result h t The degree of importance of. If it was previously memorized h t-1 Completely independent of new memory, the reset gate can function to remove the effect of previous memory, i.e.
r t ＝σ(W r [h t-1 ,x t ]+b r ) (2)
Generating new memory information based on the updated gate
Namely, it is
The output at the current time is h t I.e. by
The basic unit of the Bi-GRU model consists of a forward-propagating GRU unit and a backward-propagating GRU unit together. In a unidirectional neural network architecture, states are always output from front to back. However, in the remaining life prediction, if the output at the present time can be associated with both the state at the previous time and the state at the subsequent time. The current hidden layer state of the Bi-GRU is input by x t Hidden layer state forward at time t-1
And the output of the inverted hidden state
The three parts are jointly determined.
Wherein: the GRU () function represents a non-linear transformation of the input mechanical equipment degradation indicator, encoding the degradation indicator into the corresponding GRU hidden state. w is a t 、v t Respectively represents the forward hidden layer state corresponding to the bidirectional GRU at the time t
And the output of the inverted hidden state
Corresponding weight, b t And showing the bias corresponding to the hidden layer state at the time t.
And 3 Bi-GRU layers and a full-connection regression layer are adopted to carry out regression prediction on the constructed health indexes. By adopting the 3-layer model, the parameters of the model can be increased, and the learning capability of the model is improved. Wherein the hidden state of each layer has 2 information streams, 1 is transmitted to the next time, and 2 is used as the input of the next layer at the current time. The MBi-GRU model can fully utilize the related information of the past and future degradation states of mechanical equipment and effectively transmit the information layer by layer, thereby improving the accuracy of residual life prediction.
The invention is further improved in that: the health indicator in step 5 has N values, namely (z (t) 1 ),z(t 2 ),...,z(t N ) N =1,2, \ 8230;, N. Let L be the length of the sliding window. (z (t) i ),z(t i+1 ),...,z(t i+L-1 ) ) aDegraded state of window length, with corresponding output z (t) i+L ) N samples of the N degradation state data can be generated, and the predicted system state can be expressed by the following functional relation
z(t i+L )＝φ(z(t i ),z(t i+1 ),...,z(t i+L-1 )) (8)
The confidence interval for the prediction result is obtained to quantify the uncertainty of the point estimate. An alternative method is used to resample K times from the original training data. Model phi k (K =1,2.. K.) resampled data S is taken each time k (t 1 :t i ) (K =1,2,. K) training. Finally, an integrated operation of multiple models will produce the mean and variance of the remaining life prediction results. The above description is formulated as follows
Z k (l i +t i )＝φ k (S k (t 1 :t i )) (9)
Wherein Z is k (l i +t i ) Represents a predicted value of the state of the system obtained by a Bootstrap method.
t i Remaining life of time l i The definition is as follows:
l i ＝inf{l i :Z k (l i +t i )≥τ|z(t 1 :t i )} (10)
wherein tau is a preset failure threshold value; z is a radical of formula 0:i Is from t 0 To t i Estimated system state value of time, Z k (t i +l i ) Is estimated t i +l i The system state value at the time. The final confidence interval of the remaining life can be determined by t i Remaining life of time l i The percentile of (a).
The invention is further improved in that: in step 6, in the initial stage of the network training, the learning efficiency is set to be larger, and when the optimal solution is gradually approached, the learning efficiency is gradually reduced so as to be closer to the optimal value of the network. The deep cycle neural network is trained by utilizing the natural exponential decay learning efficiency, the neural network can be effectively trained, and the calculation method of the natural exponential decay learning efficiency comprises the following steps:
wherein, lr: current learning rate, lr 0 : initial learning rate, r decay : attenuation rate of learning in each round, 0 < r decay ＜1，S global : current global learning step number, S decay : number of steps per learning round, S decay ＝N sample /N batch I.e. the total number of samples divided by the size of each batch number.
The invention is further improved in that: in step 8, mean square Error, namely Mean Squared Error, MSE, and predicted Absolute Error, namely Mean Absolute Percentage Error, MAPE; the calculation formula is as follows:
wherein, HI Act For true bearing degradation conditions, HI Pre For predicted bearing degradation states, N p Is the number of points predicted.
Has the advantages that:
1. the data-driven residual life prediction method does not rely on a fixed degradation model;
2. the confidence interval of the residual service life can be effectively obtained through a Bootstrap method, and reliable knowledge is provided for operation and maintenance of equipment;
the MBi-GRU model can fully utilize the past and future related information of the degradation state of the mechanical equipment and effectively transmit the information layer by layer, thereby improving the accuracy of residual life prediction;
4. the deep neural network can be trained efficiently by the natural attenuation learning efficiency, so that the time is greatly saved, and the network non-convergence is effectively avoided;
5. the method can accurately predict the residual life of the mechanical equipment, is simple and easy to implement, can be widely applied to the health assessment and residual life prediction of the rotary machinery in the fields of chemical industry, metallurgy, electric power, aviation and the like, can greatly reduce expensive unplanned maintenance, and avoids the occurrence of disasters.
Drawings
FIG. 1 is a flowchart of a method for predicting the remaining life of a rotating machine in a multi-layer bidirectional gated cyclic unit network according to the present invention.
FIG. 2 is a schematic view of 6308 bearing installation;
FIG. 3 is a bearing spalling view;
FIG. 4 is a time domain waveform diagram of a bearing vibration signal;
FIG. 5 is a constructed bearing health indicator diagram;
FIG. 6 is a graph showing the effect of the method of the present invention on remaining life prediction.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific preferred embodiments.
As shown in FIG. 1, the method for predicting the remaining life of a rotating machine of a multi-layer bidirectional gated cyclic unit network of the present embodiment includes the following steps.
Step 1, vibration signal acquisition: vibration signals of key parts of a rotating machine are collected.
Key components of rotary machines include bearings, gears or rotors, etc.; the method for acquiring the vibration signals of the key components is the prior art, in the application, a bearing is taken as an example, and an ABLT-1A bearing service life strengthening test bed is adopted.
The test bench can install 4 bearings simultaneously and carry out accelerated fatigue life test.
The original test system of the test bed consists of 4 thermocouples and 1 acceleration sensor, and the temperature signals of 4 bearing outer rings and the vibration signals of the whole test bed are respectively picked up. In order to effectively monitor the operating conditions of each bearing, the test system was adjusted to add 4 acceleration sensors, each picking up a vibration signal on 3 rigid body shells. The 1 st, 2 nd, 3 rd and 4 th channel vibration sensors respectively correspond to the bearing data of the 1 st, 2 nd, 3 th and 4 th stations. The loading conditions for the test are shown in table 1. The rated dynamic load of the bearings is 42.3kN, and the actual weight is 35kg, namely the rated dynamic load on each bearing is 17.5kN. The radial load loading conditions are shown in table 2. After the bearing is fully loaded and the bearing runs for 16h, the final tester stops because the root mean square of vibration reaches the stop threshold. The vibration rms at full load was 11.0 and the shutdown threshold was set to 45.0. After the bearing is cut by linear cutting, the bearing outer ring at the 3 rd station can be obviously peeled off, as shown in figure 3.
TABLE 1 test conditions for full Life test
TABLE 2 radial load conditions
And 2, constructing a health index as shown in figure 5.
Step 21, 6308 bearing vibration signal time domain waveform is shown in fig. 4. The original features extracted from the vibration signal include 16 time domain features, 13 frequency domain features, 17 time-frequency domain features and 2 features based on trigonometric functions, and the features of the trigonometric functions are an inverse-triangular hyperbolic cosine sign and an inverse-triangular hyperbolic sine standard deviation, respectively.
And step 22, selecting the features with the trend value larger than 0.8 to form a feature subset.
Step 23, adopting an unsupervised SOM algorithm to construct a health index for evaluating the total life of the bearing from the selected feature subset, as shown in fig. 5.
And 3, constructing a network training set and a test set, setting the length of a sliding window to be 30, and taking data after 1500 points as the test set.
And 4, stacking the single bidirectional GRU network to construct an MBi-GRU neural network, wherein the layer number is 3, and the number of network neurons is 300.
And 5, obtaining a state predicted value of the system through a Bootstrap method.
Step 6, training the network: training the whole model by minimizing a loss function through error reverse transfer, and training a deep convolutional neural network by using polynomial attenuation learning efficiency; learning efficiency is set as equation 7, initial learning efficiency is 0.01, attenuation rate of learning efficiency is 0.5 decay 40 steps, S global The process is 200 steps.
And 7, acquiring the remaining life mean value and the confidence interval: inputting the data of the test set into the network trained in the step (5), wherein the corresponding output of the test set is the predicted degradation state value of the bearing, and the predicted value of the residual service life of the bearing can be calculated to be 81min through the formulas (8) to (10), the confidence interval is [76.0,85.3], and the error percentage of the predicted value and the true value is 5.88%.
z(t i+L )＝φ(z(t i ),z(t i+1 ),...,z(t i+L-1 )) (8)
Z k (l i +t i )＝φ k (S k (t 1 :t i )) (9)
l i ＝inf{l i :Z k (l i +t i )≥τz(t 1 :t i )} (10)
And 8, evaluating the predicted residual life, and comparing the predicted residual life with the prediction results based on the GRU network, the LSTM network and the fully-connected neural network NN respectively:
the residual life value predicted by the method disclosed by the invention is smaller than those of the rest three methods, namely MSE and MAPE, and the residual life of the bearing can be more accurately predicted. And obtaining a confidence interval of the residual life by a Bootstrap method. The learning efficiency of natural attenuation enables effective training of the network and thus learning of the features in the vibration signal.
In a word, the method combines the advantages of strong feature extraction capability of deep learning, utilizes the neural network of the bidirectional gating circulation unit to carry out regression prediction, and obtains the confidence interval of the residual life by the Bootstrap method. Aiming at the problems that the model precision is sensitive to the value of the learning rate in the training process of the recurrent neural network model, and the prediction performance of the model is influenced by overhigh and overlow, the neural network is efficiently trained by using the natural index attenuation learning efficiency.
The method can accurately predict the residual life and the confidence interval of the rotary machine, can be widely applied to the prediction of the residual life of the rotary machine in the fields of chemical industry, metallurgy, electric power, aviation and the like, can greatly reduce expensive unplanned maintenance, and avoids the occurrence of disasters.
Although the preferred embodiments of the present invention have been described in detail, the present invention is not limited to the details of the embodiments, and various equivalent changes may be made within the technical spirit of the present invention, and the technical scope of the present invention is also covered by the present invention.
Claims (6)
1. The method for predicting the residual life of the rotating machinery of the multilayer bidirectional gating circulation unit network is characterized by comprising the following steps of: the method comprises the following steps:
step 1, collecting vibration signals: collecting vibration signals of key parts of a rotating machine;
step 2, constructing a health index; constructing a health index describing the degradation process of the key parts of the rotating machinery;
step 3, constructing a network training set: constructing a network training set for the vibration signals acquired in the step 1, taking the degradation state of a key component of the rotary machine as the input of a neural network of a bidirectional gating circulation unit, and constructing a training and testing data set by using a sliding window technology;
step 4, constructing an MBi-GRU (multi-layer bidirectional gated recurrent unit, MBi-GRU) neural network: constructing an MBi-GRU neural network by a single bidirectional GRU network;
step 5, constructing an ensemble learning network through a Bootstrap method to obtain uncertainty expression of a prediction result;
step 6, training the network: training by minimizing a loss function through error reverse transfer, and training the MBi-GRU neural network by using natural attenuation learning efficiency;
and 7, acquiring the remaining life mean value and the confidence interval: inputting the data of the test set into the network trained in the step (2), wherein the output corresponding to the test set is the predicted value of the residual life to be obtained;
and 8, evaluating a residual life prediction result: and evaluating the residual life prediction result by adopting the mean square error and the prediction absolute error.
2. The method for predicting the residual life of a rotating machine in a multi-layer bidirectional gated cyclic unit network according to claim 1, wherein: in the step 2, original features are extracted from the vibration signals, features with large trend values are selected to form feature subsets, and finally the selected feature subsets are used for constructing health indexes for evaluating the whole life of the bearing by adopting an unsupervised SOM algorithm.
3. The method for predicting the residual life of a rotary machine of the multi-layer bidirectional gating cycle unit network according to claim 1, wherein the method comprises the following steps: in step 4, the refresh gate controls the current input data x at the same time t And previous memory information h t-1 Outputting a value z between 0 and 1 t The calculation formula is
z t ＝σ(W z [h t-1 ,x t ]+b x ) (1)
Where x is the input data, h is the output of the GRU unit, r is the reset gate, z is the update gate, and r and z together control how h is hidden from the previous state t-1 Calculating to obtain new hidden state h t ；z t Determine how much h is to be divided t-1 The next state is transmitted, which can be obtained by formula (1);
where σ is sigmoid function, W z To update the door weight, b z Is an offset. Reset gate control h t-1 For the result h t The degree of importance of; if previously memorized h t-1 Completely unrelated to new memory, the reset gate can function to remove the influence of previous memory, i.e.
r t ＝σ(W r [h t-1 ,x t ]+b r ) (2)
Generating new memory information based on the updated gate
Namely, it is
The output at the present moment is h t I.e. by
The basic unit of the Bi-GRU model is formed by a forward propagation GRU unit and a backward propagation GRU unit; in a unidirectional neural network structure, the state is always output from front to back;
however, in the remaining life prediction, if the output at the present time can be associated with both the state at the previous time and the state at the subsequent time; the current hidden layer state of the Bi-GRU is input by x t Hidden layer state forward at time t-1
And output of the hidden state in reverse
Three parts are jointly decided.
Wherein: the GRU () function represents the nonlinear transformation of the degradation index of the input mechanical equipment, and codes the degradation index into a corresponding GRU hidden layer state; w is a t 、v t Respectively represents the forward hidden layer state corresponding to the bidirectional GRU at the time t
And output of the hidden state in reverse
Corresponding weight, b t Indicating the bias corresponding to the hidden layer state at the time t.
4. The method for predicting the residual life of a rotating machine in a multi-layer bidirectional gated cyclic unit network according to claim 1, wherein: the health indicator in step 5 has N values, namely (z (t) 1 ),z(t 2 ),...,z(t N ) N =1,2, \ 8230;, N; let L be the length of the sliding window. (z (t) i ),z(t i+1 ),...,z(t i+L-1 ) A window-length degradation state with a corresponding output z (t) i+L ) N-L samples can be generated from N degradation state data, and the predicted system state can be expressed by the following functional relation
z(t i+L )＝φ(z(t i ),z(t i+1 ),...,z(t i+L-1 )) (8)
The confidence interval of the prediction result is obtained for quantifying the uncertainty of point estimation; resampling K times from the original training data by an alternative method; model phi k (K =1, 2.. K.) data S resampled each time are used k (t 1 :t i ) (K =1,2,. K) training; finally, the integrated operation of the multiple models will produce the mean and variance of the remaining life prediction results; the above description is formulated as follows
Z k (l i +t i )＝φ k (S k (t 1 :t i )) (9)
Wherein Z is k (l i +t i ) Representing a state prediction value of the system obtained by a Bootstrap method; t is t i Remaining life of time l i The definition is as follows:
l i ＝inf{l i :Z k (l i +t i )≥τ|z(t 1 :t i )} (10)
wherein τ is a preset failure threshold; z is a radical of 0:i Is from t 0 To t i Estimated system state value of time, Z k (t i +l i ) Is estimated t i +l i A system state value at a time; the final confidence interval of the remaining life can be determined by t i Remaining life of time l i The percentile of (a).
5. The method for predicting the residual life of a rotary machine of the multi-layer bidirectional gating cycle unit network according to claim 1, wherein the method comprises the following steps: in the step 6, the deep-cycle neural network is trained by using the natural index attenuation learning efficiency, so that the neural network can be effectively trained, and the calculation method of the natural index attenuation learning efficiency comprises the following steps:
wherein, lr: current learning rate, lr 0 : initial learning rate, r decay : attenuation rate of learning in each round, 0 < r decay ＜1，S global : current global learning step number, S decay : number of steps per learning round, S decay ＝N sample /N batch I.e. the total number of samples divided by the size of each batch number.
6. The method for predicting the residual life of the rotary machine of the multi-layer bidirectional gated cyclic unit network according to claim 1, wherein the method comprises the following steps: the mean square error and the prediction absolute error in step 8 are calculated as follows:
wherein, HI Act For true bearing degradation conditions, HI Pre For predicted degradation states of the bearing, N p Is the number of points predicted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911008163.3A CN110866314B (en) | 2019-10-22 | 2019-10-22 | Method for predicting residual life of rotating machinery of multilayer bidirectional gate control circulation unit network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911008163.3A CN110866314B (en) | 2019-10-22 | 2019-10-22 | Method for predicting residual life of rotating machinery of multilayer bidirectional gate control circulation unit network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110866314A CN110866314A (en) | 2020-03-06 |
| CN110866314B true CN110866314B (en) | 2022-11-15 |
Family
ID=69653048
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911008163.3A Active CN110866314B (en) | 2019-10-22 | 2019-10-22 | Method for predicting residual life of rotating machinery of multilayer bidirectional gate control circulation unit network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110866314B (en) |
Families Citing this family (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111289250A (en) * | 2020-02-24 | 2020-06-16 | 湖南大学 | Method for predicting residual service life of rolling bearing of servo motor |
| CN111540193B (en) * | 2020-03-13 | 2022-07-26 | 华南理工大学 | Traffic data restoration method for generating countermeasure network based on graph convolution time sequence |
| CN111597755A (en) * | 2020-04-30 | 2020-08-28 | 北京锐达芯集成电路设计有限责任公司 | Method for predicting remaining life of electronic device based on gated recursion unit neural network |
| CN111459144A (en) * | 2020-05-01 | 2020-07-28 | 西北工业大学 | Airplane flight control system fault prediction method based on deep cycle neural network |
| CN111581888B (en) * | 2020-05-18 | 2022-07-05 | 中车永济电机有限公司 | Construction method of residual service life prediction model of wind turbine bearing |
| CN111595583B (en) * | 2020-05-29 | 2022-02-01 | 重庆大学 | GAU neural network-based bearing residual life prediction method |
| CN112115004B (en) * | 2020-07-29 | 2022-02-11 | 西安交通大学 | Hard disk service life prediction method based on back propagation Bayes deep learning |
| CN112100574A (en) * | 2020-08-21 | 2020-12-18 | 西安交通大学 | Resampling-based AAKR model uncertainty calculation method and system |
| CN112179654B (en) * | 2020-09-28 | 2022-02-01 | 西南交通大学 | Rolling bearing fault identification method based on GAF-CNN-BiGRU network |
| CN112785092B (en) * | 2021-03-09 | 2024-05-07 | 中铁电气化局集团有限公司 | Switch residual life prediction method based on self-adaptive deep feature extraction |
| CN113094822A (en) * | 2021-03-12 | 2021-07-09 | 华中科技大学 | Method and system for predicting residual life of mechanical equipment |
| CN112926273B (en) * | 2021-04-13 | 2023-04-18 | 中国人民解放军火箭军工程大学 | Method for predicting residual life of multivariate degradation equipment |
| CN113379706B (en) * | 2021-06-09 | 2024-04-19 | 苏州大学 | Embryo development prediction method, device, equipment and storage medium |
| CN113722985B (en) * | 2021-08-12 | 2024-04-09 | 武汉科技大学 | Method and system for evaluating health state and predicting residual life of aero-engine |
| CN113895272A (en) * | 2021-10-15 | 2022-01-07 | 青岛科技大学 | Electric vehicle alternating current charging state monitoring and fault early warning method based on deep learning |
| CN114205251B (en) * | 2021-12-09 | 2022-12-02 | 西安电子科技大学 | Switch link resource prediction method based on space-time characteristics |
| CN114720129B (en) * | 2022-03-25 | 2023-06-27 | 山东大学 | Rolling bearing residual life prediction method and system based on bidirectional GRU |
| CN115017937B (en) * | 2022-04-28 | 2024-05-14 | 重庆大学 | Gear residual service life prediction method based on double-thread gating circulation unit |
| CN114841078B (en) * | 2022-05-26 | 2024-09-06 | 重庆大学 | Method and system for predicting residual service life of bearing |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110108456A (en) * | 2019-04-16 | 2019-08-09 | 东南大学 | A kind of rotating machinery health evaluating method of depth convolutional neural networks |
| CN110245745A (en) * | 2019-05-30 | 2019-09-17 | 华中科技大学 | Equipment remaining life prediction technique based on integrated bi-directional Recognition with Recurrent Neural Network |
-
2019
- 2019-10-22 CN CN201911008163.3A patent/CN110866314B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110108456A (en) * | 2019-04-16 | 2019-08-09 | 东南大学 | A kind of rotating machinery health evaluating method of depth convolutional neural networks |
| CN110245745A (en) * | 2019-05-30 | 2019-09-17 | 华中科技大学 | Equipment remaining life prediction technique based on integrated bi-directional Recognition with Recurrent Neural Network |
Non-Patent Citations (1)
| Title |
|---|
| 基于集成神经网络的剩余寿命预测;张永峰等;《工程科学学报》(第10期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110866314A (en) | 2020-03-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110866314B (en) | Method for predicting residual life of rotating machinery of multilayer bidirectional gate control circulation unit network | |
| Wu et al. | Degradation-aware remaining useful life prediction with LSTM autoencoder | |
| CN109141881B (en) | Rotary machine health assessment method of deep self-coding network | |
| He et al. | Long short-term memory network with multi-resolution singular value decomposition for prediction of bearing performance degradation | |
| Cai et al. | Artificial intelligence enhanced two-stage hybrid fault prognosis methodology of PMSM | |
| Sankararaman et al. | Bayesian methodology for diagnosis uncertainty quantification and health monitoring | |
| WO2017094267A1 (en) | Abnormality detection system, abnormality detection method, abnormality detection program, and method for generating learned model | |
| Yan et al. | A dynamic multi-scale Markov model based methodology for remaining life prediction | |
| Liu et al. | A switching hidden semi-Markov model for degradation process and its application to time-varying tool wear monitoring | |
| Benmoussa et al. | Remaining useful life estimation without needing for prior knowledge of the degradation features | |
| Wang et al. | Reliability estimation and remaining useful lifetime prediction for bearing based on proportional hazard model | |
| Chang et al. | Remaining useful life prediction for rolling bearings using multi-layer grid search and LSTM | |
| CN114297918A (en) | Aero-engine residual life prediction method based on full-attention depth network and dynamic ensemble learning | |
| CN112966400B (en) | Centrifugal fan fault trend prediction method based on multi-source information fusion | |
| Liu et al. | Investigation on rolling bearing remaining useful life prediction: A review | |
| Chen et al. | Gated adaptive hierarchical attention unit neural networks for the life prediction of servo motors | |
| CN111143934A (en) | Structural deformation prediction method based on time convolution network | |
| Wang et al. | Bearing degradation evaluation using improved cross recurrence quantification analysis and nonlinear auto-regressive neural network | |
| Senanayaka et al. | Autoencoders and recurrent neural networks based algorithm for prognosis of bearing life | |
| Feng et al. | Online residual useful life prediction of large-size slewing bearings—A data fusion method | |
| CN117664575A (en) | Rolling bearing residual service life prediction method based on double-graph attention and GRU | |
| Abdelli et al. | Machine learning based data driven diagnostic and prognostic approach for laser reliability enhancement | |
| Sohaidan et al. | Remaining useful life prediction of turbofan engine using long-short term memory | |
| Sun et al. | Prognostics of multilayer ceramic capacitors via the parameter residuals | |
| Hu et al. | Anomaly detection and remaining useful lifetime estimation based on degradation state for bearings |
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 |