CN117474150A - On-line prediction method based on self-adaptive adjustment of safety state of traction motor of high-speed train - Google Patents
On-line prediction method based on self-adaptive adjustment of safety state of traction motor of high-speed train Download PDFInfo
- Publication number
- CN117474150A CN117474150A CN202311365463.3A CN202311365463A CN117474150A CN 117474150 A CN117474150 A CN 117474150A CN 202311365463 A CN202311365463 A CN 202311365463A CN 117474150 A CN117474150 A CN 117474150A
- Authority
- CN
- China
- Prior art keywords
- model
- online
- data
- traction motor
- speed train
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000012544 monitoring process Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 238000013136 deep learning model Methods 0.000 claims abstract description 4
- 238000010606 normalization Methods 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 4
- 230000004931 aggregating effect Effects 0.000 claims description 2
- 238000011156 evaluation Methods 0.000 claims description 2
- 238000011010 flushing procedure Methods 0.000 claims description 2
- 230000003287 optical effect Effects 0.000 abstract 1
- 238000012216 screening Methods 0.000 abstract 1
- 238000000926 separation method Methods 0.000 abstract 1
- 238000002474 experimental method Methods 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Control Of Electric Motors In General (AREA)
Abstract
The invention discloses an online prediction method based on self-adaptive adjustment of the safety state of a traction motor of a high-speed train, which comprises two stages of separation line and online. In an off-line stage, acquiring high-speed train history monitoring sensor data, adopting a Pearson coefficient to perform characteristic screening on sensor data related to traction motor temperature, and preprocessing; training the ODL by utilizing the preprocessing data, and reasonably initializing model parameters. In an online stage, acquiring real-time sensor data related to the temperature of a high-speed train and a traction motor, and normalizing; inputting the online data into an ODL (optical digital hierarchy) to calculate the loss between the model predicted value and the real value at the current moment, updating the model parameters online based on the hedging strategy and adaptively adjusting the model depth; and inputting the sensor data at the current moment into an ODL model to output a predicted value of the temperature of the future motor, and repeating the online stage process according to the real-time data of the train-mounted end. The online deep learning model constructed by the method is simple and efficient, and the concept drift problem of online data is effectively improved.
Description
Technical Field
The invention relates to the field of high-speed train state prediction, in particular to an online prediction method based on self-adaptive adjustment of the safety state of a traction motor of a high-speed train.
Background
At present, the monitoring of the running state of the traction motor mainly depends on real-time data sent by temperature sensors arranged on a stator core and bearings at two ends, when the temperature of a measuring point of the traction motor exceeds a set specified threshold value, a train can send out early warning information, and train operators can make related measures to prevent the train from accident. However, the monitoring method of judging whether the current temperature exceeds a fixed threshold is a relatively backward means, on the one hand, the sensor data related to the temperature of the traction motor is not effectively utilized; on the other hand, when the temperature exceeds the threshold value and the related operation is performed, the inside of the motor has failed. Therefore, the temperature of the traction motor is accurately predicted, the occurrence of serious faults of the motor can be effectively avoided, and the method has important significance for guaranteeing the safe operation of the train.
The traditional temperature prediction model is generally trained offline through a complete historical data set by means of computing power of a ground end, and then is deployed in a computing platform of a train vehicle-mounted end for application. However, the sensor data on the train-mounted end arrive sequentially in the form of a stream, and the stream data are mostly non-stationary data, i.e. the probability distribution of the data changes with time to generate a concept drift phenomenon. The model of the fixed structure and parameters obtained by offline training is not adapted to the data with varying probability distribution. In order to solve the problem of conceptual drifting of data, an ideal choice is to train a model in an online learning mode, namely, directly deploy the model at the train-mounted end, and then continuously learn and update model parameters for the data streams which arrive in sequence and perform a model prediction task.
The existing online learning method is to use an online gradient descent algorithm to perform back propagation training of the deep neural network on each individual data sample. However, this method requires setting the structure of the model before training, and fails to provide a test set for verification for an online scene, and if the initial model depth is set unreasonably, the model prediction accuracy will be reduced or convergence will be difficult. How to combine the advantages of online learning and deep learning, dynamically adjusting parameters and structures of a model, and being applied to online prediction of temperature of a traction motor of a high-speed train is a problem to be solved.
Disclosure of Invention
Aiming at the problems, the online deep learning model ODL is used for real-time prediction of the safety state of the traction motor, and the ODL can dynamically adjust the structure and parameters of the model to adapt to streaming data with continuously changing probability distribution. The ODL model combines the advantages of deep learning and online learning, adopts a Deep Neural Network (DNN) as a basic structure thereof, and realizes online learning of model parameters through an opposite strategy. The result shows that the ODL model can be effectively applied to the real-time prediction of the temperature of the train end, and is beneficial to early finding out the abnormal state of the motor.
An online prediction method for the safety state of a traction motor of a high-speed train based on self-adaption adjustment is characterized by comprising the following steps:
step 1: according to collected historical sensor monitoring data of the high-speed train, sensor signal data most relevant to the temperature change of a traction motor stator are screened by utilizing a Pearson coefficient, and the sensor signal data are preprocessed;
step 2: constructing an on-line deep learning model ODL by taking a deep neural network DNN as a basic structure, setting the hidden layer number of the DNN, namely the maximum depth L of the model and the number of hidden layer units, connecting each hidden layer with a predictor, and then training the model through the preprocessed data in the step 1 to generate initial parameters of the on-line stage ODL model;
step 3: acquiring sensor signal data Z related to temperature change of traction motor in real time t And to Z t Carrying out on-line normalization;S t ∈R m×1 represents the m-dimensional signal value at time t, < >>Representative length t s +t p Time period m-dimensional signal, t s Representation ofThe length of the sequence time window, t p Representing a model predicted time step;
step 4: normalized Z t Acquisition through sliding windowAs input to the model, wherein m-dimensional sensor signal time series +.>For online training of models, normalized X t Inputting an ODL model to predict to generate a predicted value +.>And is matched with the true value Y of the temperature at the moment t t Comparing, and calculating the loss of the model at the current moment according to the loss function;
step 5: parameters of the ODL model are updated on line according to a counter-flushing strategy for the predictors with different depths by a despreading counter-propagation algorithm, the weight of each hidden layer predictor is updated according to the scaling dot product of the output of the predictors and the actual value, and the depth of the model is adjusted through weight value self-adaption;
step 6: aggregating samplesInput into ODL model to generate corresponding prediction set
Preferably, preprocessing the sensor signal data in step 1 includes filling in data missing values, aligning data time points, and normalizing data.
Preferably, in step 1, the pearson coefficient is used to screen the sensor signal data most relevant to the temperature change of the traction motor stator, specifically, the pearson coefficient is introduced as an evaluation standard, the degree of correlation between the collected historical sensor data of the high-speed train and the temperature characteristic of the motor stator is quantitatively evaluated, and the absolute value of the coefficient is considered to be the sensor signal data most relevant to the temperature change of the traction motor stator when the absolute value of the coefficient falls in a value interval [0.7,1 ].
Preferably, DNNs in the model structure of the ODL connect each hidden layer and a predictor, where each predictor has a consistent size with the output data of the hidden layer, and the prediction values of the L predictors are weighted and combined to obtain the prediction result of the final model.
Preferably, the weight of each predictor is alpha l Representing 0.ltoreq.alpha l And the depth of the model can be adjusted by adjusting the numerical value of the output weight of the predictor.
Preferably, step 3 is for Z t Performing online normalization comprises performing data normalization once every time a batch of data arrives in an online scene, and acquiring Z by using a maximum and minimum normalization method for data normalization t Each dimension of the sequenceMaximum and minimum Max of (2) i And Min i The normalization formula is as follows:
preferably, the loss function in step 4 is:
the first term of the formula represents the MSE loss of the model, and the second two terms are L1 regularization term and lambda w ,λ θ For the regularization coefficient(s), W 1 ,||θ|| 1 Refers to the sum of the absolute values of the elements in the weight vector.
Preferably, the parameters of the online updating of the ODL model in step 5 include hidden layer-to-hidden layer parameters W l And parameter θ between hidden layer and predictor l ;θ l And W is l The updating of (a) is completed by adopting an online gradient descent algorithm, and theta l Gradient descent, W, depending on losses generated by a single predictor only l Considering the losses generated by the predictors, the update formulas are respectively as follows:
the weight of the predictor at the next moment is updated according to the scaling dot product of the predictor output and the actual value, and the formula is as follows:
f l and outputting the corresponding predictor for the first hidden layer.
Preferably, the predicted value in step 6 is calculated as follows:
α l the weight of the predictor corresponding to the first hidden layer.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a diagram of an ODL model structure for on-line deep learning in the method of the present invention;
FIG. 3 is a real-time temperature prediction framework based on an ODL model in the method of the invention;
FIG. 4 is a MAE comparison of experimental results predicted by the method of the present invention with other methods;
fig. 5 is a comparison of experimental results predicted using the method of the present invention with OGD model predicted results.
Detailed Description
The invention is further described below with reference to the drawings and examples.
FIG. 1 is a block flow diagram of the method of the present invention. In the method, firstly, high-speed train historical sensor monitoring data are collected, then sensor signals most relevant to temperature change of a traction motor stator are screened by utilizing a Pearson coefficient according to the collected data, and the sensor signal data are preprocessed. The preprocessing comprises missing value filling, data time point alignment and data normalization.
The data used in the embodiment of the invention comes from the sensor monitoring data collected by the wireless transmission system of the high-speed train in the actual running process of the high-speed train. In the experiment, data of 16 days of continuous operation of the train was used, 12 days of data being used as a training set of the model and 4 days of data being used as a test set.
Then, 18 signals, which are strongly related to the temperature of the stator of the traction motor of the single-section motor car, in WTDS data are selected as input through a Pearson coefficient method, wherein the selected sensor signals are shown in a table 1, and the last column shows the Pearson coefficients of the input signals and the temperature of the stator.
Table 1 input signal description
The sampling period of the different sensors on the high speed train is different, for example, the period of the temperature sensor is 15 seconds, and the period of the speed sensor is 30 seconds. To ensure that the lengths of the model input data are the same, the sampling period of the sensor needs to be unified. In order to fully utilize the latest sensor monitoring data as much as possible under the condition of ensuring real-time requirements, the sensor acquisition period of the invention is 1min. Since the train has a sunroof time per day, only the operation data of 8 to 22 points per day period are used for the model test. The data samples usable per day were 14×60=840 in length and 16×840 in size, as shown in table 2.
Table 2 data example
The sensor sampling period is 1min, so the predicted duration is t p And (3) minutes. Will t in the experiment p Set to 5, 10 and 15, representing predicted motor temperatures after 5, 10 and 15 minutes, respectively. The sensor data acquired online is then normalized.
And then, inputting the normalized data into an ODL model for online learning. The model structure of the ODL is shown in fig. 2. The ODL model constructed by the method adopts a Deep Neural Network (DNN) as a basic structure, sets the hidden layer number (the maximum depth L of the model) and the hidden layer unit number of the DNN, connects each hidden layer with a predictor, and trains the model through preprocessing data to rationalize the initial parameters of the model in an online stage. Unlike the output strategy of the traditional DNN model, DNN in the ODL model does not pass through the last layer h L Deriving a prediction result, connecting each hidden layer with the predictors, and then carrying out weighted combination on the prediction values of the L predictors to obtain the prediction result of the final model, wherein the sizes of the output data of each predictor and each hidden layer are consistent, namely the number of neurons of each hidden layer is the same. Weight reuse α for each predictor l Wherein 0.ltoreq.alpha l And is less than or equal to 1. Therefore, the prediction result of the model is based on predictors connected with each hidden layer, and the depth of the model can be adjusted by adjusting the numerical value of the output weight of the predictors. And training the model offline, generating initial parameters of the online model, and then carrying out online learning training, updating and prediction of the model.
In order to accurately construct an on-line traction motor temperature prediction model, the input of the model is a sensor signal related to four traction motor temperature signals on a single-section motor car, and the output of the model is the stator temperature of the four motors. Fig. 3 shows the overall flow of real-time temperature prediction, where the model needs to complete the training, updating and predicting process of the model at time t. Consider an m-dimensional sensor signal time sequenceAs an input signal to the temperature prediction model, Y t Is the true value of the temperature at time t, wherein +.>S t ∈R m×1 Represents the m-dimensional signal value at the time t, corresponds to m monitoring variables related to the temperature of the stator, and is t s Representing the length of the sequence time window, t p Representing the time step of the model prediction.
During the training phase at time t, all sensor information before time t is available, at time t sequence X is entered t And the stator temperature Y measured at time t t Then the model obtains the error (loss) between the predicted value and the true value; and then the model is updated on line by the hedging strategy. In the prediction phase, sample setIs input into the model and generates the corresponding prediction set +.>At the moment, the temperature prediction result can be fed back to train operators in real time, so that the train operators can grasp the working state of the traction motor in advance.
In the experiments of the embodiments of the present invention, two performance indexes are used to evaluate the prediction effect of the ODL model, namely, the Mean Absolute Error (MAE) and the Root Mean Square Error (RMSE), where the MAE can evaluate the absolute error of the model prediction result and the RMSE can measure the degree of similarity between the predicted sequence and the actual sequence. The two calculations are as follows:
wherein Y is t ,The real temperature value and the predicted value at the time T are respectively, and T is the length of a certain group of predicted data.
The ODL model has 4 super parameters, namely the maximum depth L of the DNN model, the sliding time window size m, the initial learning rate eta of the online gradient descent algorithm and the number H of hidden nodes. The superparameter selection is also different when predicting temperatures of different time steps, and the model is set to (16,20,0.001,210) according to the superparameter experiment when performing the predicted task for 5 minutes into the motor temperature, (L, m, η, H), in order to make the temperature predicted task more meaningful, the model is then also performing the predicted task for 10 minutes into the motor temperature and 15 minutes into the motor temperature, the superparameter settings of the two tasks being (16,30,0.001,210) and (16,45,0.001,220), respectively.
The experiment adopts 7 groups of control models, motor temperature predictions of 5, 10 and 15 minutes in the future are carried out, and an MAE error index of 5, 10 and 15 steps of predicting the motor temperature is shown in FIG. 4, wherein the model is carried out for 4 continuous days. As the prediction step increases, the prediction error of each model increases. From the results, the prediction error of the ODL model per day is far smaller than that of other models under the same prediction step length, and the model prediction effect under the online learning is obviously better than that of the offline learning. The ODL model can well process the streaming data and effectively reduce the concept drift phenomenon of the data.
The real-time prediction result and the actual value are displayed, and as shown in fig. 5, the comparison situation of the prediction value and the actual value when the model continuously performs the temperature for four days for 5 steps (5 min), 10 steps (10 min) and 15 steps (15 min) is given. For ease of observation, only the best predictive effect in the offline and OGD models is compared to the ODL, and it can be observed that the predictive result of the ODL model is more closely to the true value.
Experiments show that the prediction result of the ODL model has very excellent performance, and the temperature mutation condition in streaming data can be sensitively tracked. ODL is deployed on a train, so that the on-the-way prediction of the temperature of the traction motor can be effectively realized, and measures are taken in time for motor abnormality.
The present invention is not limited to the preferred embodiments, and any changes or substitutions that would be apparent to one skilled in the art within the scope of the present invention are intended to be included in the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.
Claims (9)
1. An online prediction method for the safety state of a traction motor of a high-speed train based on self-adaption adjustment is characterized by comprising the following steps:
step 1: according to collected historical sensor monitoring data of the high-speed train, sensor signal data most relevant to the temperature change of a traction motor stator are screened by utilizing a Pearson coefficient, and the sensor signal data are preprocessed;
step 2: constructing an on-line deep learning model ODL by taking a deep neural network DNN as a basic structure, setting the hidden layer number of the DNN, namely the maximum depth L of the model and the number of hidden layer units, connecting each hidden layer with a predictor, and then training the model through the preprocessed data in the step 1 to generate initial parameters of the on-line stage ODL model;
step 3: acquiring sensor signal data Z related to temperature change of traction motor in real time t And to Z t Carrying out on-line normalization;S t ∈R m×1 represents the m-dimensional signal value at time t, < >>Representative length t s +t p Time period m-dimensional signal, t s Representing the length of the sequence time window, t p Representing a model predicted time step;
step 4: normalized Z t Acquisition through sliding windowAs input to the model, wherein m-dimensional sensor signal time series +.>For online training of models, normalizedX of (2) t Inputting an ODL model to predict to generate a predicted value +.>And is matched with the true value Y of the temperature at the moment t t Comparing, and calculating the loss of the model at the current moment according to the loss function;
step 5: parameters of the ODL model are updated on line according to a counter-flushing strategy for the predictors with different depths by a despreading counter-propagation algorithm, the weight of each hidden layer predictor is updated according to the scaling dot product of the output of the predictors and the actual value, and the depth of the model is adjusted through weight value self-adaption;
step 6: aggregating samplesInput into ODL model to generate corresponding prediction set
2. The method for online prediction of the traction motor safety state of the high-speed train based on the self-adaption adjustment according to claim 1, wherein the preprocessing of the sensor signal data in the step 1 comprises filling of data missing values, alignment of data time points and normalization of data.
3. The method for online prediction of the traction motor safety state of the high-speed train based on the self-adaptation adjustment according to claim 1 is characterized in that in the step 1, sensor signal data most relevant to the temperature change of the traction motor stator is screened by using pearson coefficients, specifically, pearson correlation coefficients are introduced as evaluation criteria, the degree of correlation between collected historical sensor data of the high-speed train and the temperature characteristics of the motor stator is quantitatively evaluated, and the absolute value of the coefficients is regarded as the sensor signal data most relevant to the temperature change of the traction motor stator in a value interval [0.7,1 ].
4. The online prediction method based on the self-adaptive adjustment of the safety state of the traction motor of the high-speed train according to claim 1, wherein DNN in the model structure of the ODL connects each hidden layer with a predictor, the output data of each predictor is consistent with that of the hidden layer, and the prediction values of the L predictors are weighted and combined to obtain the prediction result of the final model.
5. The online prediction method based on self-adaptive adjustment of high-speed train traction motor safety state according to claim 4, wherein the weight of each predictor is alpha l Representing 0.ltoreq.alpha l And the depth of the model can be adjusted by adjusting the numerical value of the output weight of the predictor.
6. The online prediction method based on self-adaptive adjustment of the safety state of the traction motor of the high-speed train according to claim 1, wherein the step 3 is characterized in that t Performing online normalization comprises performing data normalization once every time a batch of data arrives in an online scene, and acquiring Z by using a maximum and minimum normalization method for data normalization t Each dimension of the sequenceMaximum and minimum Max of (2) i And Min i The normalization formula is as follows:
7. the online prediction method for the traction motor safety state of the high-speed train based on the self-adaption adjustment according to claim 1, wherein the loss function in the step 4 is as follows:
the first term of the formula represents the MSE loss of the model, and the second two terms are L1 regularization term and lambda w ,λ θ For the regularization coefficient(s), W 1 ,||θ|| 1 Refers to the sum of the absolute values of the elements in the weight vector.
8. The online prediction method based on self-adaptive adjustment of the safety state of a traction motor of a high-speed train according to claim 1, wherein the parameters of the online update ODL model in step 5 include hidden layer-to-hidden layer parameter W l And parameter θ between hidden layer and predictor l ;θ l And W is l The updating of (a) is completed by adopting an online gradient descent algorithm, and theta l Gradient descent, W, depending on losses generated by a single predictor only l Considering the losses generated by the predictors, the update formulas are respectively as follows:
the weight of the predictor at the next moment is updated according to the scaling dot product of the predictor output and the actual value, and the formula is as follows:
f l and outputting the corresponding predictor for the first hidden layer.
9. The online prediction method for the traction motor safety state of the high-speed train based on the self-adaption adjustment according to claim 8, wherein the predicted value in the step 6 is calculated as follows:
α l the weight of the predictor corresponding to the first hidden layer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311365463.3A CN117474150A (en) | 2023-10-20 | 2023-10-20 | On-line prediction method based on self-adaptive adjustment of safety state of traction motor of high-speed train |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311365463.3A CN117474150A (en) | 2023-10-20 | 2023-10-20 | On-line prediction method based on self-adaptive adjustment of safety state of traction motor of high-speed train |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117474150A true CN117474150A (en) | 2024-01-30 |
Family
ID=89623060
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311365463.3A Pending CN117474150A (en) | 2023-10-20 | 2023-10-20 | On-line prediction method based on self-adaptive adjustment of safety state of traction motor of high-speed train |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117474150A (en) |
-
2023
- 2023-10-20 CN CN202311365463.3A patent/CN117474150A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108960303B (en) | Unmanned aerial vehicle flight data anomaly detection method based on LSTM | |
WO2019080367A1 (en) | Method for evaluating health status of mechanical device | |
CN109343505A (en) | Gear method for predicting residual useful life based on shot and long term memory network | |
CN111160620B (en) | Short-term wind power prediction method based on end-to-end memory network | |
US9934470B2 (en) | Production equipment including machine learning system and assembly and test unit | |
CN110309537B (en) | Intelligent health prediction method and system for aircraft | |
CN111337244B (en) | Method and device for monitoring and diagnosing faults of input shaft of fan gearbox | |
CN109709483A (en) | A kind of variable-pitch system of wind turbine generator method for diagnosing faults | |
CN114325395A (en) | Method and device for determining battery state | |
CN108961460B (en) | Fault prediction method and device based on sparse ESGP (Enterprise service gateway) and multi-objective optimization | |
Huang et al. | An adversarial learning approach for machine prognostic health management | |
CN110991776A (en) | Method and system for realizing water level prediction based on GRU network | |
CN114943179A (en) | Reliability evaluation and residual life prediction method based on multi-source degradation data fusion | |
CN116010900A (en) | Multi-scale feature fusion gearbox fault diagnosis method based on self-attention mechanism | |
CN114548375A (en) | Cable-stayed bridge main beam dynamic deflection monitoring method based on bidirectional long-short term memory neural network | |
CN114002597A (en) | Motor fault diagnosis method and system based on GRU network stator current analysis | |
CN115017826A (en) | Method for predicting residual service life of equipment | |
CN112257914A (en) | Aviation safety cause and effect prediction method based on random forest | |
CN117474150A (en) | On-line prediction method based on self-adaptive adjustment of safety state of traction motor of high-speed train | |
CN112149896A (en) | Attention mechanism-based mechanical equipment multi-working-condition fault prediction method | |
CN113779724B (en) | Intelligent fault prediction method and system for filling packaging machine | |
CN113468720B (en) | Service life prediction method for digital-analog linked random degradation equipment | |
CN114925476A (en) | Method for predicting quaternion of rolling bearing degradation track augmentation and storage medium | |
CN114819350A (en) | RUL prediction method under sensor fault condition based on multiple tasks | |
CN114692738A (en) | Lightweight real-time series anomaly detection 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 |