CN114510778A - Track irregularity prediction method based on hybrid intelligent optimization LSTM - Google Patents

Track irregularity prediction method based on hybrid intelligent optimization LSTM Download PDF

Info

Publication number
CN114510778A
CN114510778A CN202210021594.9A CN202210021594A CN114510778A CN 114510778 A CN114510778 A CN 114510778A CN 202210021594 A CN202210021594 A CN 202210021594A CN 114510778 A CN114510778 A CN 114510778A
Authority
CN
China
Prior art keywords
lstm
model
track irregularity
prediction
pso
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210021594.9A
Other languages
Chinese (zh)
Inventor
孟海宁
李维
童新宇
姬文江
张嘉薇
杨哲
黑新宏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian University of Technology
Original Assignee
Xian University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian University of Technology filed Critical Xian University of Technology
Priority to CN202210021594.9A priority Critical patent/CN114510778A/en
Publication of CN114510778A publication Critical patent/CN114510778A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/15—Vehicle, aircraft or watercraft design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/25—Design optimisation, verification or simulation using particle-based methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • 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)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a track irregularity prediction method based on hybrid intelligent optimization LSTM, which comprises the steps of preprocessing time sequence data, and then optimizing the hyperparameter of an LSTM model by utilizing a PSO algorithm so as to determine the network structure of the LSTM model; and optimizing the initial weight threshold of the LSTM model by using a GA algorithm, and determining the weight threshold of the LSTM model. And finally, training and predicting the track irregularity data by using the determined hyper-parameters and weight threshold values. The method for predicting the track irregularity data based on the LSTM-PSO-GA model solves the problem that the accuracy is low in the prediction process of the traditional prediction method, optimizes LSTM parameters by using PSO and GA algorithms, avoids the problem that the model falls into a local optimal solution, and improves the prediction convergence speed. Finally, the prediction of the track irregularity data is realized, and the track irregularity phenomenon is predicted more accurately.

Description

Track irregularity prediction method based on hybrid intelligent optimization LSTM
Technical Field
The invention belongs to the technical field of time series prediction, and particularly relates to a track irregularity prediction method based on hybrid intelligent optimization (LSTM).
Background
The continuous development of the high speed and the heavy load of the railway traffic in China has the advantages that the track repeatedly bears the load of rolling stocks, and the track gauge, the level, the height and other geometric shapes and spatial positions of the track are gradually changed, so that the track is unsmooth. The irregularity of the railway track directly threatens the running safety of the train. The track irregularity state development trend is predicted based on the historical data of the track irregularity, the track safety is checked, and an important theoretical basis can be provided for a track maintenance strategy.
In recent years, many scholars at home and abroad study the problem of track irregularity prediction. Yonganti predicts the Track Quality state, establishes a linear regression equation of a Track Quality Index (TQI), and solves the linear regression equation by using a least square method. The song construction army adopts a grey system uncertainty theory, and establishes a grey nonlinear orbit irregularity prediction model by taking TQI time series data as an index. The creep wave is provided based on a non-equidistant gray linear weighted combination prediction model, and a weighted combination function is adopted to correct a prediction residual sequence, so that the prediction precision is improved. And the Jiachaolong proposes a segmented neural network-ARMA (autoregressive moving average) recursion model, adopts the neural network recursion model for a low-frequency approximate sequence after wavelet decomposition, adopts the ARMA model for a high-frequency detail sequence, and predicts the track irregularity state. The horse takes a rough trend of the predicted TQI sequence by utilizing a non-equidistant gray model, and then carries out residual error correction on the preliminary prediction result by utilizing an Elman neural network, thereby predicting the TQI sequence more accurately. Guoholong proposed a combined model based on the gray theory and a support vector machine to predict TQI data. The Yao sub-peak provides a combined prediction model based on a gray model and a recurrent neural network to predict the track irregularity trend. The Sterculia sinensis is provided with a track irregularity estimation Algorithm based on combination of a Genetic Algorithm (GA) and a vehicle track coupling model, and the track irregularity estimation Algorithm is used for converting a track irregularity problem into an estimation problem of model parameters to solve. However, the above method does not consider the situation of interference with the prediction result when the fluctuation variation of the data is large, and does not sufficiently mine the characteristic information of the track irregularity data, so that it is difficult to form a stable and accurate track irregularity prediction model.
The track irregularity data has non-linearity, randomness, and burstiness. The Long Short-Term Memory (LSTM) recurrent neural network can detect the nonlinear characteristics of data according to the time dependence relationship of sequence data and establish a prediction model of nonlinear data under different characteristics. Particle Swarm Optimization (PSO) can be used for optimizing hyper-parameters of a neural network model, and GA can optimize initial weight threshold of the neural network model, so that the problems that the model prediction process is easy to fall into a local optimal solution and the convergence speed is low are solved. Therefore, some prediction methods of hybrid intelligent algorithms become effective ways for predicting track irregularity, and better performance can be obtained through combination of various intelligent algorithms.
Disclosure of Invention
The invention aims to provide a track irregularity prediction method based on hybrid intelligent optimization LSTM. The method solves the problem that the traditional prediction method has low prediction precision on the track irregularity sequence data with large fluctuation.
The technical scheme adopted by the invention is that,
the track irregularity prediction method based on the hybrid intelligent optimization LSTM comprises the following steps:
step 1, collecting track irregularity sequence data;
step 2, acquiring track irregularity sequence data, and converting the track irregularity sequence data into a track quality index time sequence;
step 3, preprocessing the track quality index time series data acquired in the step 2;
step 4, establishing an LSTM model according to the data obtained in the step 3, and simultaneously optimizing the hyper-parameters of the LSTM model by utilizing a PSO algorithm;
step 5, establishing an LSTM model according to the data obtained in the step 3 and the optimal hyper-parameter obtained in the step 4, and optimizing a weight threshold of the LSTM model by using a GA algorithm to obtain an LSTM-PSO-GA model;
and 6, predicting future data by using an LSTM-PSO-GA model.
The invention is also characterized in that:
in step 3, the sequence data is preprocessed by a normalization processing method, and the original sequence data is mapped to [0,1 ]]The specific method comprises the following steps: calculating the maximum value and the minimum value of the sequence data, and respectively recording as XmaxAnd Xmin(ii) a X is then subtracted from each of the sequence dataminIs then divided by Xmax-Xmin。
In step 4, the LSTM model comprises an input layer, a hidden layer, an output layer, network training, network prediction and a module. The input layer is responsible for carrying out preliminary processing on an original track irregularity sequence to meet network input requirements, the hidden layer adopts LSTM cells to build a single-layer cyclic neural network, the output layer provides a prediction result network, and network prediction adopts an iterative method to predict point by point.
In step 4, the specific method for optimizing the hyper-parameters of the LSTM model by using the PSO algorithm comprises the following steps: firstly, determining hyper-parameters to be optimized in an LSTM model as a time window and the number of nodes of a hidden layer; parameters are then initialized, including population size (number of particles) M, maximum number of iterations NmaxB, inertial weight w, boundary X of particle positionmaxAnd XminRange of particle velocity VmaxAnd VminAcceleration factor c1And c2(ii) a A population of particles is randomly generated, each individual comprising two hyper-parameters: time window and number of hidden layer nodes. And in the PSO algorithm iteration process, the positions of the current individuals are adjusted by utilizing the global optimal individuals and the historical optimal individuals. And after the PSO algorithm iteration is finished, acquiring the optimal hyper-parameter of the LSTM model.
In step 5, the specific method for optimizing the LSTM model weight threshold by using the GA algorithm is as follows: initializing a population, and coding a weight threshold to be optimized; if the current maximum fitness value has no significant change or reaches the maximum iteration times of the population in the iteration process, stopping optimization; otherwise, updating the hyperparameter chromosome by using selection, crossing and mutation operations, and assigning the new hyperparameter to the LSTM neural network.
The LSTM-PSO-GA model comprises an input layer, a hidden layer and an output layer, wherein the sequence data obtained in the step 3 and the prediction result of the LSTM model in the step 4 are used as the input of the input layer, and the output layer is the prediction result of the LSTM-PSO-GA model; the hidden layer uses tanh as the activation function.
The beneficial effects of the invention are:
the invention not only solves the problem that the traditional prediction method has low prediction precision on the track irregularity sequence data with large fluctuation. And the time series data calculation method for respectively optimizing the neural network hyperparameter and the weight threshold by utilizing the PSO algorithm and the GA algorithm is provided, and the problems that the model is easy to fall into a local optimal solution, and the convergence speed is low and unstable in the model prediction process are solved. The prediction method can extract the characteristic change of the track irregularity data, finally realize high-accuracy prediction and analysis of the track irregularity sequence data, and more accurately predict the track irregularity phenomenon.
Drawings
FIG. 1 is a flow chart of the method for predicting track irregularity based on hybrid intelligent optimization LSTM according to the present invention;
FIG. 2 is a flowchart of GA-optimized LSTM weight threshold in the method for predicting track irregularity based on hybrid intelligent optimization LSTM of the present invention;
FIG. 3 is a diagram of a single LSTM cell structure in the method for predicting the track irregularity based on the hybrid intelligent optimization LSTM according to the present invention;
FIG. 4 is a track quality index data diagram of the track irregularity prediction method based on hybrid intelligent optimization LSTM;
FIG. 5 is a track quality index prediction result diagram in the track irregularity prediction method based on the hybrid intelligent optimization LSTM.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention discloses a track irregularity prediction method based on hybrid intelligent optimization LSTM, and a flow chart of the method is shown in figure 1. The method comprises the following steps:
step 1, collecting rail irregularity sequence data;
step 2, acquiring track irregularity time sequence data and converting the track irregularity time sequence data into a track quality index time sequence; the non-compliant time series data includes: the track quality index, the track gauge standard deviation, the height standard deviation and the horizontal standard deviation are four items of data;
step 3, preprocessing the track irregularity sequence data acquired in the step 2;
step 4, establishing an LSTM network model according to the track irregularity data obtained in the step 3, and simultaneously optimizing the hyper-parameters of the LSTM model by utilizing a PSO algorithm;
step 5, establishing an LSTM network model according to the track irregularity data obtained in the step 3 and the hyperparameter obtained in the step 4, and optimizing a weight threshold of the LSTM model by using a GA algorithm to obtain an LSTM-PSO-GA model;
and 6, predicting the future track irregularity sequence data by using the model obtained in the step 5 and an LSTM-PSO-GA model.
In step 3, preprocessing the track irregularity sequence data by adopting a normalization processing method, and mapping the original sequence data to [0,1 ]]The interval is specifically as follows: calculating the maximum value and the minimum value of the sequence data, and respectively recording as XmaxAnd Xmin(ii) a X is then subtracted from each of the sequence dataminIs then divided by Xmax-Xmin。
In step 4, the specific method for optimizing the hyper-parameters of the LSTM model by using the PSO algorithm is as follows:
and 4.1, determining hyper-parameters to be optimized in the LSTM model as a time window and the number of nodes of a hidden layer. Initializing parameters: population size (number of particles) M, maximum number of iterations NmaxInertial weight w, boundary X of particle positionmaxAnd XminRange of particle velocity VmaxAnd VminAcceleration factor c1And c2；
Step 4.2, randomly generating a particle population, wherein each individual comprises two hyper-parameters: time window and number of hidden layer nodes.
And 4.3, iterating by adopting a PSO algorithm, wherein the specific method for constructing the LSTM model in each iteration comprises 5 functional modules of an input layer, a hidden layer, an output layer, network training and network prediction. The input layer is responsible for carrying out preliminary processing on the original track irregularity sequence to meet network input requirements, the hidden layer establishes a single-layer cyclic neural network, the output layer provides a prediction result, and the network prediction module carries out point-by-point prediction by adopting an iterative method. Firstly, defining the original track uneven sequence column after normalization as F in the input layero＝{f1,f2,…,fnH, the divided training set and test set can be represented as Ftr＝{f1,f2,…,fmAnd Fte＝{fm+1,fm+2,…,fnMeet the constraint condition m<N and m, N ∈ N. In order to adapt to the characteristic of hidden layer input, a data segmentation method is applied to FtrProcessing is performed, and if the division length is L, the model after division is X ═ X1,X2,…,XL}，Xp＝{fp,fp+1,…,fm-L+p-1P is more than or equal to 1 and less than or equal to L; p, L ∈ N. The corresponding desired output is Y ═ Y1,Y2,…,YL},YP＝{fp+1,fp+2,…,fm-L+p}. Then, X is input into a hidden layer, which contains L isomorphic LSTM cells connected at successive times, and the output of X after passing through the hidden layer can be expressed as P ═ { P ═ P1,P2,…,PL}，Pp＝LSTMforward(Xp,Cp-1,Hp-1) In the formula Cp-1And Hp-1The state and output of the previous LSTM cell, respectively; LSTMforwardThe LSTM forward cell calculation method is shown. Setting the magnitude of the cell state vector to SstateThen C isp-1And Hp-1Both vectors are Sstate. Hidden layer output P, model input X and theoretical output Y are all two-dimensional arrays with dimensions (m-L, L). Selecting mean square error
Figure BDA0003462593300000071
As an error calculation formula, the loss function of the training process can be defined as:
Figure BDA0003462593300000072
and setting the minimum loss function as an optimization target, and continuously updating the network weight to further obtain a final hidden layer network.
And 4.4, adjusting the position of the current individual by using the globally optimal individual and the historically optimal individual in the PSO algorithm iterative process.
And 4.5, obtaining the optimal hyper-parameter of the LSTM model after the PSO algorithm iteration is finished.
In step 5, the process of optimizing the weight threshold of the LSTM model by using the GA algorithm is shown in fig. 2, and the specific method is as follows:
step 5.1, determining the time window size and the number of hidden layer nodes of the LSTM model according to the optimal hyper-parameter obtained in the step 4;
step 5.2, GA algorithm coding: encoding an initial weight threshold of the LSTM model, and taking a model error as a fitness value of each chromosome;
step 5.3, generating a specified number of populations;
step 5.4, starting iteration, updating the hyperparameter chromosome by utilizing selection, intersection and mutation operations, and assigning a new hyperparameter to the LSTM model;
step 5.5, if the current maximum fitness value reaches the maximum iteration times of the population, stopping optimization;
step 5.6, calculating the current prediction error, storing the current optimal hyperparameter and the corresponding LSTM network model, and completing the prediction of the time series data of the track irregularity;
constructing an LSTM model, training and predicting the existing data; the specific method for constructing the LSTM model comprises 5 functional modules of an input layer, a hidden layer, an output layer, network training and network prediction. The input layer is responsible for carrying out preliminary processing on the original response time sequence to meet the network input requirement, the hidden layer adopts the LSTM cell represented by the figure 3 to build a single-layer cyclic neural network, the output layer provides a prediction result network, and the network prediction adopts an iterative method to predict point by point.
The LSTM-PSO-GA model comprises an input layer, a hidden layer and an output layer, wherein the sequence data obtained in the step 3 and the prediction result of the LSTM model in the step 4 are used as the input of the input layer, and the output layer is the prediction result of the LSTM-PSO-GA model; the hidden layer uses tanh as the activation function.
The present embodiment uses the data from Kyoto line uplink K449+300-K449+600 sector within 1000 days. The rail inspection vehicle acquires rail irregularity detection data every 0.25 m along the track mileage direction, and can acquire various index data of 1200 detection points in the section in a detection process, wherein the index data comprises a plurality of indexes such as track gauge, level, height and the like. The 1200 data collected by each index each time are counted to obtain track quality index data (TQI) of the day, a TQI sequence is shown in figure 4, a prediction result of an LSTM-PSO-GA model in the TQI sequence is shown in figure 5, error pairs of different models are shown in table 1, a Root Mean Square Error (RMSE), a Mean Absolute Error (MAE) and a Mean Absolute Percentage Error (MAPE) are respectively adopted as evaluation indexes and are respectively shown in formulas (3), (4) and (5), wherein RMSE is a standard deviation, N is the number of data samples, and y is the number of data samplespredictiveTo predict value, ytrueIs the actual value.
Figure BDA0003462593300000091
Figure BDA0003462593300000092
Figure BDA0003462593300000101
TABLE 1 comparison of prediction errors for different models
Figure BDA0003462593300000102

Claims (6)

1. The track irregularity prediction method based on the hybrid intelligent optimization LSTM is characterized by comprising the following steps:
step 1, collecting track irregularity sequence data;
step 2, acquiring track irregularity sequence data and converting the track irregularity sequence data into a track quality index time sequence;
step 3, preprocessing the track quality index time series data acquired in the step 2;
step 4, establishing an LSTM model according to the data obtained in the step 3, and simultaneously optimizing the hyper-parameters of the LSTM model by utilizing a PSO algorithm;
step 5, establishing an LSTM model according to the data obtained in the step 3 and the optimal hyper-parameter obtained in the step 4, and optimizing a weight threshold of the LSTM model by using a GA algorithm to obtain an LSTM-PSO-GA model;
and 6, predicting future data by using an LSTM-PSO-GA model.
2. The method of claim 1, wherein in step 3, the sequence data is preprocessed by normalization to map the raw sequence data to [0,1 ]]The specific method comprises the following steps: calculating the maximum value and the minimum value of the sequence data, and respectively recording as XmaxAnd Xmin(ii) a X is then subtracted from each of the sequence dataminIs then divided by Xmax-Xmin。
3. The method of claim 1, wherein in step 4, the LSTM model comprises an input layer, a hidden layer, an output layer, a network training and network prediction module, and a module. The input layer is responsible for carrying out preliminary processing on an original track irregularity sequence to meet network input requirements, the hidden layer adopts LSTM cells to build a single-layer cyclic neural network, the output layer provides a prediction result network, and network prediction adopts an iterative method to predict point by point.
4. The method for predicting the track irregularity based on the hybrid intelligent optimization LSTM according to claim 1, wherein in the step 4, the specific method for optimizing the hyper-parameters of the LSTM model by using the PSO algorithm is as follows: firstly, determining hyper-parameters to be optimized in an LSTM model as a time window and the number of nodes of a hidden layer; parameters are then initialized, including population size (number of particles) M, maximum number of iterations NmaxB, inertial weight w, boundary X of particle positionmaxAnd XminRange of particle velocity VmaxAnd VminAcceleration factor c1And c2(ii) a A population of particles is randomly generated, each individual comprising two hyper-parameters: time window and number of hidden layer nodes. And in the PSO algorithm iteration process, the positions of the current individuals are adjusted by utilizing the global optimal individuals and the historical optimal individuals. And after the PSO algorithm iteration is finished, acquiring the optimal hyper-parameter of the LSTM model.
5. The method for predicting the track irregularity based on the hybrid intelligent optimization LSTM according to claim 1, wherein in the step 5, the specific method for optimizing the weight threshold of the LSTM model by using the GA algorithm is as follows: initializing a population, and coding a weight threshold to be optimized; if the current maximum fitness value has no significant change or reaches the maximum iteration times of the population in the iteration process, stopping optimization; otherwise, updating the hyperparameter chromosome by using selection, crossing and mutation operations, and assigning the new hyperparameter to the LSTM neural network.
6. The LSTM-based hybrid intelligent optimization (LSTM) -based track irregularity prediction method of claim 1, wherein the LSTM-PSO-GA model comprises an input layer, a hidden layer and an output layer, wherein the sequence data obtained in step 3 and the prediction result of the LSTM model in step 4 are used as input of the input layer, and the output layer is the prediction result of the LSTM-PSO-GA model; the hidden layer uses tanh as the activation function.
CN202210021594.9A 2022-01-10 2022-01-10 Track irregularity prediction method based on hybrid intelligent optimization LSTM Pending CN114510778A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210021594.9A CN114510778A (en) 2022-01-10 2022-01-10 Track irregularity prediction method based on hybrid intelligent optimization LSTM

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210021594.9A CN114510778A (en) 2022-01-10 2022-01-10 Track irregularity prediction method based on hybrid intelligent optimization LSTM

Publications (1)

Publication Number Publication Date
CN114510778A true CN114510778A (en) 2022-05-17

Family

ID=81548907

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210021594.9A Pending CN114510778A (en) 2022-01-10 2022-01-10 Track irregularity prediction method based on hybrid intelligent optimization LSTM

Country Status (1)

Country Link
CN (1) CN114510778A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116050814A (en) * 2023-04-03 2023-05-02 北京交通大学 Self-adaptive correction method for tamping scheme of ballasted track
CN116307302A (en) * 2023-05-23 2023-06-23 西南交通大学 Inversion method, system and storage medium for track irregularity dynamic and static detection data
WO2024060441A1 (en) * 2022-09-21 2024-03-28 中建材创新科技研究院有限公司 Control optimization method for high-precision cutting of gypsum plasterboard

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2024060441A1 (en) * 2022-09-21 2024-03-28 中建材创新科技研究院有限公司 Control optimization method for high-precision cutting of gypsum plasterboard
CN116050814A (en) * 2023-04-03 2023-05-02 北京交通大学 Self-adaptive correction method for tamping scheme of ballasted track
CN116050814B (en) * 2023-04-03 2023-08-04 北京交通大学 Self-adaptive correction method for tamping scheme of ballasted track
CN116307302A (en) * 2023-05-23 2023-06-23 西南交通大学 Inversion method, system and storage medium for track irregularity dynamic and static detection data
CN116307302B (en) * 2023-05-23 2023-07-25 西南交通大学 Inversion method, system and storage medium for track irregularity dynamic and static detection data

Similar Documents

Publication Publication Date Title
CN111915059B (en) Attention mechanism-based Seq2Seq berth occupancy prediction method
CN109492822B (en) Air pollutant concentration time-space domain correlation prediction method
CN114510778A (en) Track irregularity prediction method based on hybrid intelligent optimization LSTM
CN109142171B (en) Urban PM10 concentration prediction method based on feature expansion and fusing with neural network
CN108388962B (en) Wind power prediction system and method
CN112734097B (en) Unmanned train energy consumption prediction method, system and storage medium
CN112149230B (en) Method for predicting comfort deterioration of wind-induced train of strong wind railway
CN112734094B (en) Intelligent city intelligent rail vehicle fault gene prediction method and system
CN112966853A (en) Urban road network short-term traffic flow prediction method based on space-time residual error mixed model
CN112149915B (en) Charging load prediction method and system for charging station
CN116757057A (en) Air quality prediction method based on PSO-GA-LSTM model
CN113674524A (en) LSTM-GASVR-based multi-scale short-time traffic flow prediction modeling and prediction method and system
CN110766190A (en) Power distribution network load prediction method
CN116187835A (en) Data-driven-based method and system for estimating theoretical line loss interval of transformer area
CN116311921A (en) Traffic speed prediction method based on multi-spatial scale space-time converter
CN115907122A (en) Regional electric vehicle charging load prediction method
CN113536665B (en) Road surface temperature short-term prediction method and system based on characteristic engineering and LSTM
CN112098869B (en) Self-adaptive electric vehicle SOC estimation method based on big data
CN116167465A (en) Solar irradiance prediction method based on multivariate time series ensemble learning
CN114463978B (en) Data monitoring method based on track traffic information processing terminal
CN115907228A (en) Short-term power load prediction analysis method based on PSO-LSSVM
CN112907971B (en) Urban road network short-term traffic flow prediction method based on genetic algorithm optimization space-time residual error model
CN115056829A (en) Train motion state estimation method for multi-vehicle type continuous learning
CN113947904A (en) Multi-scale short-term traffic flow prediction method based on S-G filtering and deep belief network
Wang et al. Research on parking space prediction based on long short-term memory

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