CN110457781B - Passenger comfort-oriented train tunnel-passing time length calculation method - Google Patents
Passenger comfort-oriented train tunnel-passing time length calculation method Download PDFInfo
- Publication number
- CN110457781B CN110457781B CN201910670306.0A CN201910670306A CN110457781B CN 110457781 B CN110457781 B CN 110457781B CN 201910670306 A CN201910670306 A CN 201910670306A CN 110457781 B CN110457781 B CN 110457781B
- Authority
- CN
- China
- Prior art keywords
- train
- tunnel
- sample
- current
- samples
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for calculating the time length of a train passing through a tunnel facing passenger comfort. And the classification of the tunnel air pressure meteorological parameters in the database is realized through a building division and average temperature distribution model. On the basis, typical sequence extraction is carried out on each category, and an RGB color image combined template matching model is constructed. And an LSSVM error model is established for the output error of the RGB color image combined template matching model. And finally, organically fusing the two models to realize the accurate calculation of the time length of the train passing through the tunnel.
Description
Technical Field
The invention relates to a passenger comfort oriented method for calculating the time length of a train passing through a tunnel.
Background
Since the 21 st century, the development of high-speed trains in China is led from foreign countries to absorption and digestion and then to autonomous research and development, and high-speed trains drive the development of the great amount in China. After a great deal of effort is put into research and development technicians to ensure the safety and safety of trains, the comfort of riding becomes a focus of current related research. The passenger riding experience of the whole journey of the passenger train is improved, and the passenger transport competitiveness of the railway is improved. Comprehensive information prompt is favorable for enabling passengers to master the current train running condition, and passenger riding experience is improved. The prompt information in the current train comprises air temperature, train speed per hour and the like, but the prompt of the relevant information of special working conditions such as the fact that the train passes through a tunnel and the like is lacked.
After the train enters the tunnel, the condition of communication information loss can occur, so that people in the internet era can be free from the situation, and the riding comfort of passengers is rapidly reduced. Meanwhile, the dark closed environment in the tunnel, eardrum discomfort caused by poor vehicle tightness and unknown tunnel length can cause people to feel great anxiety. Therefore, an effective and accurate tunnel-passing time length positioning method is designed, and the time length for timely informing passengers that the passengers still need to wait in the tunnel plays an important role in improving the passenger experience. The core of the calculation of the time length of the train passing through the tunnel is the accurate positioning of the train in the tunnel. The common train positioning methods include the following methods:
1. the Beidou GPS positioning mode can provide positioning information in all directions, all weather and all day time by applying Beidou satellite positioning, but is not suitable for positioning trains in tunnels.
2. The wireless base station type uses the wireless base stations at two ends of the tunnel to provide train information, effectively reduces the use of trackside equipment along the line, but is not suitable for positioning the train in the tunnel.
3. The answer mode is that a plurality of trackside equipment are arranged along the railway, and corresponding vehicle-mounted equipment is installed on the train simultaneously, so that the accuracy is improved, but the engineering cost and the maintenance cost are huge, and the practicability is poor.
Disclosure of Invention
The invention aims to solve the technical problem that the method for calculating the time length of the train passing through the tunnel facing passenger comfort is provided aiming at the defects of the prior art, so that the intelligent perception of the distance from the train in the tunnel to the exit of the tunnel is realized, and the residence time of the train in the tunnel is accurately estimated.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows: a passenger comfort-oriented train tunnel-passing time length calculation method estimates the remaining time length of a train passing through a current tunnel by using the following formula: t = O (i)/f Hz (O (i) -O (i-1)); wherein the final output value O (i) = O of the distance between the current position of the train and the tunnel exit M (i) + ε (i); o (i) is the template matching output value of the current sample point; epsilon (i) is a compensation error output result; f. of Hz Is the location update frequency.
Final output value O of distance from current position of train to tunnel exit M (i) The specific determination process of (a) includes:
1) Collecting tunnel air pressure meteorological parameters, and constructing a tunnel air pressure meteorological parameter database;
2) Classifying meteorological parameters in the tunnel meteorological parameter database based on the tunnel meteorological parameter database according to the geographic position of the tunnel and meteorological conditions inside and outside the tunnel to obtain similar samples in the region, and realizing classification of tunnel groups with similar attributes;
3) Further dividing the similar samples in the region to obtain typical samples of the current category;
4) Performing phase space reconstruction on the typical sample to obtain color images of the evolution of the atmospheric pressure meteorological parameters at the head and the tail of the train, namely template images; forming a typical sample combined template library by the template images corresponding to all the typical samples;
5) Respectively collecting sample points at the head and the tail of the current position of the train and a sample sequence of a period of time ahead, carrying out phase space reconstruction and RGB color space combination, and constructing a current position characteristic image { c h ,c f }; performing correlation calculation on the current position characteristic image and an image in a typical sample combined template library, and determining the best matching position of the current position in the typical sample combined template library;
6) The method comprises the steps of training a least square support vector machine by using data of a similar tunnel group, taking a temperature sequence, a humidity sequence and an air pressure sequence as input and taking a prediction error of an optimal matching position as output, and establishing a tunnel mileage prediction error compensation model;
7) Acquiring air pressure meteorological parameters in the running process of the train in real time, constructing a current state characteristic image, and searching and matching a three-dimensional image template; and determining an error compensation input variable and an error compensation model, fusing the output of the template matching model and the output of the error compensation model, and obtaining a final distance value from the tunnel outlet, namely O (i).
The specific implementation process of the step 2) comprises the following steps:
1) Classifying the tunnel group into N types;
2) Obtaining the average air temperature distribution in a value range of [10,20] min before the train passes through the tunnel in the current region for 1 year, fitting the average air temperature distribution by adopting a Gaussian distribution function, obtaining a mean value and a variance value, equally dividing the distribution into 10 equal parts according to probability, and defining samples belonging to 1 equal part as the same samples in the current region.
The specific implementation process of the step 3) comprises the following steps:
1) Performing autoregressive differential moving average model modeling on each time sequence, extracting parameters of an autoregressive term, a differential term and a moving regression term of each sequence, and forming a characteristic matrix A by all characteristic quantities of similar samples in a region;
2) Performing dimension reduction on a feature matrix A consisting of all feature quantities of the same type of samples in the region, selecting M principal components with the largest contribution degree to represent the information of the original feature matrix A, and acquiring a transformed matrix A';
3) Defining a kernel function k = ak rbf +βk linear +(1-α-β)k laplace (ii) a Wherein k is rbf Is a radial basis kernel function, k linear Is a linear kernel, k laplace Mapping the eigenvalue in the A' matrix to a feature space corresponding to a kernel function k for the Laplace kernel function;
4) The coefficients α, β and the number of classes n of the kernel function are optimized. The clustering process divides the samples into n classes, each class of samples forming a sample cluster. Determining an optimized objective function
Wherein, avg (C) i ) As a cluster of samples C i Average distance of middle samples, d cen (C i ,C j ) As a cluster of samples C i And cluster of samples C j Distance between the center points;
5) According to the set parameters, the k-means clustering algorithm optimized by the gray wolf optimization algorithm is adopted to realize the clustering of the features after dimensionality reduction, and the clustering center and distance of each clustering sample cluster are obtainedOriginal time series corresponding to 5 samples nearest to cluster center
Wherein
The method comprises the steps of corresponding to a temperature time sequence collected at the head of a train, a temperature time sequence collected at the head of the train, an air pressure time sequence collected at the head of the train, a temperature time sequence collected at the tail of the train and an air pressure time sequence collected at the tail of the train; aggregating the time series
Defined as a representative sample of the current category.
The specific implementation process of the step 5) comprises the following steps:
1) Determining a current sample point and 19 forward sampling points in a time sequence of temperature, humidity and air pressure of the head and the tail of the current position of the train;
2) Phase space reconstruction is carried out by adopting a delayed coordinate method, 6 two-dimensional reconstruction matrixes representing the temperature, humidity and air pressure evolution characteristics of the head and the tail of the train are obtained, the 6 matrixes are combined head and tail according to RGB color space, and a current position characteristic image { c is formed h ,c f }；
3) Performing convolution operation on the current position characteristic image and the image in the template library
Wherein each g i Are all one-dimensional sequences; { h i ,f i I =0,1,2 \ 82305 is a template image;
4) For all g i Sorting the elements in the sequence from big to small, wherein the determined maximum 5 elements are candidate elements, the position where the candidate elements are located before sorting is a candidate position, and the distance from the candidate position to the exit of the tunnel is s j ,j＝1,2,…5；
5) Determining the average value of the mileage values corresponding to the candidate positions to the exit of the tunnel asThe current template matching output value, i.e. the template matching output value
The concrete implementation process of the step 6) comprises the following steps:
1) Define input sample I = (T) h ,H h ,P h ,T f ,H f ,P f ) Wherein T is h ＝(t 1 ,t 2 …,t 19 ,t 20 ) The temperature time sequence of the current sample point and the previous 19 sample points of the train head in the tunnel is obtained; h h For a train head humidity time series of length 20, P h For a train head air pressure time sequence of length 20, T h Is a train tail temperature time sequence with the length of 20, H h Is a train tail humidity time sequence with the length of 20, P h Is a train tail air pressure time sequence with the length of 20. The output sample is a mileage error value epsilon corresponding to the current position; the input and output combination Y = { I, epsilon } constitutes a modeling sample;
2) Dividing a training sample, a verification sample and a test sample to obtain a training set and a test set;
3) Binary coding is carried out on the features of each dimension in the input sample I, when the code value corresponding to the features of a certain dimension is 1, the features are selected as input variables of the LSSVM model, and when the code value corresponding to the features of the certain dimension is 0, the features of the certain dimension are discarded; randomly initializing and coding 60 dimensional features into 0 or 1;
4) Determining input features based on the current feature coding values, and training an LSSVM model by adopting training set data; inputting the data of the verification set into a trained LSSVM model, and obtaining the output sequence of the model as
Defining an optimization objective function
Wherein
Verifying the true output values of the set;
5) And performing iterative optimization operation by adopting a binary ant lion algorithm to determine the optimal input characteristics and an LSSVM model, wherein the model is an LSSVM mileage prediction error compensation model.
The specific implementation process of the step 7) comprises the following steps:
1) Acquiring the average temperature distribution of a train in a value range of [10,20] min before the train passes through the tunnel, determining the sample category to which the current state belongs, and acquiring a current temperature and humidity sequence by using temperature and humidity sensors arranged at the head and the tail of the train;
2) Determining the sample category to which the current state belongs according to the current average air temperature, and extracting a corresponding template library;
3) Determining the best matching position in the template base, and outputting the template matching output value O of the current sample point M (i)；
4) Obtaining a model input vector I in the current state, substituting the model input vector I into the trained LSSVM model, and obtaining a compensation error output result epsilon (I);
5) And fusing the three-dimensional template matching output value and the LSSVM model output value to obtain the final output value of the distance between the current position and the tunnel outlet, namely O (i) = O M (i)+ε(i)。
Compared with the prior art, the invention has the following beneficial effects: the method fully excavates the potential rule that the meteorological parameters of the air pressure in the tunnel change along with the depth of the tunnel by using an artificial intelligence big data analysis technology. The intelligent sensing of the distance from the train to the tunnel outlet in the tunnel is realized through the temperature, humidity and air pressure time sequence data acquired at the two ends of the train. On the basis, effective estimation of the residence time of the train in the tunnel is realized. Accurate information prompt can be provided for passengers, and passenger riding experience is improved. Meanwhile, the method can realize input data acquisition only by vehicle-mounted temperature, humidity and air pressure sensors after the modeling is finished, does not need any trackside equipment, and has great popularization value.
Drawings
FIG. 1 is a schematic diagram of temporal sequence phase space reconstruction and RGB joint template matching;
fig. 2 is a main flow chart of the present invention.
Detailed Description
The method comprises the following specific steps:
step 1: collecting tunnel air pressure meteorological parameters, and constructing a tunnel air pressure meteorological parameter database
The method comprises the steps of collecting temperature, humidity and air pressure sequences in a tunnel when a train passes through in real time through vehicle-mounted sensors distributed at two ends of the train, collecting mileage time sequences of the train in the tunnel through response devices which are installed in the tunnel and on the train in advance, wherein the sampling frequency is 10Hz. And taking the average value of the temperature and the humidity acquired T min before the train enters the tunnel as the estimated value of the local temperature and humidity average value. When a train passes through a tunnel once, temperature sequences, humidity sequences and air pressure sequences acquired by sensors positioned at two ends of the train and estimated values of average air temperature and average humidity of an area where the train is located, which are acquired before the train enters the tunnel, form 1 group of tunnel meteorological parameter samples. And (3) forming a tunnel meteorological parameter database by using tunnel meteorological parameter samples acquired in 1-year operation of all trains in the district. T is in the range of [10,20], and in the present invention, T =10.
And 2, step: tunnel group atmospheric pressure meteorological parameter classification
Based on the tunnel meteorological parameter database, the meteorological parameters in the database are classified according to the geographic position of the tunnel and the meteorological conditions inside and outside the tunnel, so that the classification of tunnel groups with similar attributes is realized. The specific implementation process is as follows:
step A1: and classifying the Chinese climate characteristics according to the Chinese building division, and classifying the tunnel groups into 7 types according to the geographic positions of the tunnel samples in the database. And modeling the tunnel group in each area respectively.
Step A2: and obtaining the average air temperature distribution in the Tmin before the train passes through the tunnel in the current area for 1 year, and fitting the average air temperature distribution by adopting a Gaussian distribution function to obtain a mean value and a variance value. The distribution is divided into 10 equal parts according to probability, and the samples belonging to the same 1 equal part are defined as the same samples in the current area.
And step 3: representative sample characterization of homogeneous samples within a region
And further dividing the same type of samples in the region obtained by roughly dividing the input attribute set based on the tunnel category. The method specifically comprises the following substeps:
step B1: and carrying out evolution feature extraction on the temperature, humidity and air pressure time sequence of the train passing through the tunnel in the sample cluster. The method specifically comprises the steps of carrying out autoregressive differential moving average model (ARIMA) modeling on each time sequence, and extracting parameters of an autoregressive term, a differential term and a moving regression term of each sequence. Specifically, a time series acquired by the train passing through the tunnel once may acquire a 6-column time series, and thus the feature quantity extracted by the train passing through the tunnel once may be represented as a = (p) 1 ,d 1 ,q 1 ,p 2 ,d 2 ,q 2 ,…,p 6 ,d 6 ,q 6 ) And all the characteristic quantities of the same type of samples in the region form a characteristic matrix A.
And step B2: and (3) performing dimensionality reduction on a characteristic matrix A formed by all characteristic quantities of the same type of samples in the region by adopting a Principal Component Analysis (PCA) algorithm. And selecting the information of the original characteristic matrix A represented by the 5 principal components with the largest contribution degree to obtain a transformed matrix A'.
And step B3: defining kernel functions
k＝αk rbf +βk linear +(1-α-β)k laplace (1)
In the formula k rbf Is a radial basis kernel function, k linear Is a linear kernel, k laplace Is a laplacian kernel. And mapping the characteristic values in the A' matrix to a characteristic space corresponding to the kernel function k.
And step B4: and determining an optimized object, and optimizing coefficients alpha and beta and the category number of the kernel function by adopting a wolf optimization algorithm (GWOO). Wherein alpha, beta epsilon [0,1], the number of classes is a positive integer less than 20.
And step B5: determining an optimized objective function
Wherein avg (C) i ) Is a cluster C i Average distance of middle samples, d cen (C i ,C j ) Is a cluster C i And are in cluster C j The distance between the center points.
Step B6: and according to the set parameters, realizing the clustering result of the reduced-dimension features by adopting a k-means clustering algorithm optimized by a grey wolf optimization algorithm. Obtaining the clustering center of each clustering sample cluster and the original time sequence corresponding to the 5 samples nearest to the clustering center
Wherein
The temperature time sequence collected corresponding to the train head, the air pressure time sequence collected corresponding to the train head, the temperature time sequence collected corresponding to the train tail, and the air pressure time sequence collected corresponding to the train tail. Aggregating time series
Defined as a typical sample of the current category.
And 4, step 4: construction of a canonical sample joint template library
Performing phase space reconstruction on a typical sample, setting the delay time to be 1, setting the window length to be 5, performing phase space reconstruction on a temperature, humidity and air pressure time sequence by adopting a delay coordinate method, obtaining three two-dimensional reconstruction matrixes representing evolution characteristics of the temperature, the humidity and the air pressure of the head of the train and three two-dimensional reconstruction matrixes representing evolution characteristics of the temperature, the humidity and the air pressure of the tail of the train, combining the three matrixes according to RGB color space, and forming a color image of evolution of meteorological parameters of the head and the tail of the train, namely the template image { h } i ,f i },i＝0,1,2…5。
And 5: training template RGB color joint matching model
Respectively collecting sample points at the head and the tail of the current position of the train and a sample sequence of a period of time ahead, carrying out phase space reconstruction and RGB color space combination, and constructing the current positionCharacteristic image { c } h ,c f }. And performing correlation calculation on the current position feature module and the images in the template library to determine the best matching position of the current position in the template library. The method specifically comprises the following steps:
step C1: the current sample point and the previous 19 sampling points in the time sequence of the temperature, the humidity and the air pressure at the head and the tail of the current position of the train.
And step C2: setting the delay time to be 1 and the window length to be 5, performing phase space reconstruction by adopting a delay coordinate method, acquiring 6 two-dimensional reconstruction matrixes representing the temperature, humidity and air pressure evolution characteristics of the head and the tail of the train, and combining the 6 matrixes head and tail according to RGB color space to form a current position characteristic image { c } h ,c f }。
And C3: performing convolution operation on the current position characteristic image and the image in the template library
Wherein each g i Are all one-dimensional sequences.
And C4: for all g i Sorting the elements in the sequence from big to small, wherein the determined maximum 5 elements are candidate elements, the position where the candidate elements are located before sorting is a candidate position, and the distance from the candidate position to the exit of the tunnel is s j ,j＝1,2,…5。
And C5: taking the average value of the mileage value from the exit of the tunnel corresponding to the candidate position to determine the current template matching output value, namely the template matching output value
And 6: establishing a joint template matching error compensation model
The method comprises the steps of training a Least Square Support Vector Machine (LSSVM) by using data of the same tunnel group, taking a temperature sequence, a humidity sequence and an air pressure sequence as input and taking a prediction error of a template matching model as output, and establishing a tunnel mileage prediction error compensation model. The method specifically comprises the following steps:
step D1: defining training samples, defining input samples I = (T) h ,H h ,P h ,T f ,H f ,P f ) Wherein T is h ＝(t 1 ,t 2 …,t 19 ,t 20 ) The temperature time sequence of the current sample point and the previous 19 sample points of the train head in the tunnel is shown. Similarly, H h Is a train head humidity time sequence with a length of 20, P h For a train head air pressure time sequence of length 20, T h Is a train tail temperature time sequence with the length of 20, H h Is a train tail humidity time sequence with the length of 20, P h Is a train tail air pressure time sequence with the length of 20. The output sample is the mileage error value epsilon corresponding to the current position. The input and output combination Y = { I, epsilon } constitutes a modeling sample. And 3000 samples are selected for each similar tunnel group to establish a mileage prediction error compensation model.
Step D2: and dividing the training sample, the verification sample and the test sample. 70% of 3000 samples are selected as a training set and 30% are selected as a verification set in a non-return random sampling mode.
And D3: and determining an optimized object and initializing an optimized value. And (3) optimizing the input characteristics of the model by adopting a binary ant lion algorithm, namely performing binary coding on the characteristics of each dimension in the input sample I, selecting the characteristics as input variables of the LSSVM model when the coding value corresponding to the characteristics of a certain dimension is 1, and discarding the characteristics of the dimension when the coding value corresponding to the characteristics of the certain dimension is 0. The 60 dimensional features are randomly initialized and coded as 0 or 1.
Step D4: an optimization objective function is determined. And determining input features based on the current feature coding value, and training the LSSVM model by adopting training set data. Inputting the data of the verification set into a trained LSSVM model, and obtaining the output sequence of the model as
Defining an optimization objective function
In the formula
The true output values of the set are verified.
Step D5: and outputting the optimized prediction model. And performing iterative optimization operation by adopting a binary ant lion algorithm to determine the optimal input characteristics and an LSSVM model, wherein the model is an LSSVM mileage prediction error compensation model.
And 7: obtaining test data and calling mileage prediction model
In the running process of the train, collecting real-time atmospheric pressure meteorological parameters, constructing a current state characteristic image, and searching and matching a three-dimensional image template; and determining an error compensation input variable and an error compensation model. And fusing the template matching output and the error compensation output to obtain a final distance value from the tunnel exit. The method specifically comprises the following steps:
step E1: and acquiring the average air temperature distribution in the Tmin before the train passes through the tunnel. The sample class to which the current state belongs is determined. And acquiring a current temperature and humidity sequence by using temperature and humidity sensors arranged at the head and the tail of the train.
Step E2: and determining the sample type of the current state according to the current average air temperature, and extracting a corresponding template library.
Step E3: referring to the flow in step 3, the best matching position is determined in the template library, and the template matching output value O of the current sample point is output M (i)。
Step E4: and 6, referring to the flow of the flow C1 in the step 6, obtaining a model input vector I in the current state, substituting the model input vector I into the trained LSSVM model, and obtaining a compensation error output result epsilon (I).
And E5: fusing the three-dimensional template matching output value and the LSSVM model output value to obtain the final output value of the distance from the current position to the tunnel outlet
O(i)＝O M (i)+ε(i) (5)
And 8: calculation of predicted residual time for exiting tunnel
The remaining time of the train passing through the current tunnel can be estimated according to the following formula:
t＝O(i)/f Hz (O(i)-O(i-1)) (6)
in the formula f Hz Is the location update frequency.
Claims (6)
1. A passenger comfort-oriented train tunnel-passing time length calculation method is characterized in that the remaining time length of a train passing through a current tunnel is estimated by the following formula: t = O (i)/f Hz (O (i) -O (i-1)); wherein the final output value O (i) = O of the distance between the current position of the train and the tunnel exit M (i) + ε (i); o (i) is the final output value of the distance between the current position of the train and the exit of the tunnel; o (i-1) is the final output value of the distance between a position on the train and the tunnel outlet; epsilon (i) is the compensation error output result; f. of Hz A location update frequency; o is M (i) Matching an output value for the template of the current sample point;
the specific determination process of the final output value O (i) of the distance from the current position of the train to the tunnel exit comprises the following steps:
1) Collecting tunnel air pressure meteorological parameters and constructing a tunnel air pressure meteorological parameter database;
2) Classifying meteorological parameters in the tunnel meteorological parameter database based on the tunnel meteorological parameter database according to the geographic position of the tunnel and meteorological conditions inside and outside the tunnel to obtain similar samples in the region, and realizing classification of tunnel groups with similar attributes;
3) Further dividing the similar samples in the region to obtain typical samples of the current category;
4) Performing phase space reconstruction on the typical sample to obtain color images of the evolution of the atmospheric pressure meteorological parameters at the head and the tail of the train, namely template images; template images corresponding to all the typical samples form a typical sample combined template library;
5) Respectively collecting sample points at the head and the tail of the current position of the train and a sample sequence of a period of time ahead, carrying out phase space reconstruction and RGB color space combination, and constructing a current position characteristic image { c h ,c f }; comparing the current position feature image with the dictionaryPerforming correlation calculation on the images in the sample combined template library, and determining the best matching position of the current position in the typical sample combined template library;
6) Adopting data of the same kind of tunnel group, taking a temperature sequence, a humidity sequence and an air pressure sequence as input, taking the prediction error of the optimal matching position as output, training a least square support vector machine, and establishing a tunnel mileage prediction error compensation model;
7) Acquiring air pressure meteorological parameters in the running process of the train in real time, constructing a current state characteristic image, and searching and matching a three-dimensional image template; and determining an error compensation input variable and an error compensation model, fusing the output of the template matching model and the output of the error compensation model, and obtaining a final distance value from the tunnel outlet, namely O (i).
2. The passenger comfort oriented method for calculating the time length for the train to pass through the tunnel according to claim 1, wherein the step 2) is realized by the following specific steps:
1) Classifying the tunnel group into N types;
2) The method comprises the steps of obtaining average air temperature distribution in Tmin before a train passes through a tunnel in the current region for 1 year, fitting the average air temperature distribution by adopting a Gaussian distribution function, obtaining a mean value and a variance value, equally dividing the distribution into 10 equal parts according to probability, and defining samples which belong to the same 1 equal part as the same samples in the current region as the same samples.
3. The passenger comfort oriented train tunnel-passing time length calculation method according to claim 1, wherein the concrete implementation process of the step 3) comprises the following steps:
1) Performing autoregressive differential moving average model modeling on each time sequence, extracting parameters of an autoregressive term, a differential term and a moving regression term of each sequence, and forming a characteristic matrix A by all characteristic quantities of similar samples in a region;
2) Performing dimension reduction processing on a feature matrix A consisting of all feature quantities of similar samples in the region, selecting M principal components with the largest contribution degree to represent the information of an original feature matrix A, and acquiring a transformed matrix A';
3) Defining a kernel function k = α k rbf +βk linear +(1-α-β)k laplace (ii) a Wherein k is rbf Is a radial basis kernel function, k linear Is a linear kernel, k laplace Mapping the eigenvalue in the A' matrix to the eigenspace corresponding to the kernel function k for the Laplace kernel function;
4) Optimizing coefficients alpha and beta of the kernel function and the class number n, dividing the samples into n classes, and forming a sample cluster by each class of samples; determining an optimized objective function
Wherein, avg (C) i ) As a cluster of samples C i Average distance of middle samples, d cen (C i ,C j ) As a cluster of samples C i And sample cluster C j Distance between the center points;
5) According to the set parameters, the k-means clustering algorithm optimized by the gray wolf optimization algorithm is adopted to realize the clustering of the characteristics after dimensionality reduction, and the clustering center of each clustering sample cluster and the original time sequence corresponding to the 5 samples closest to the clustering center are obtained
Wherein
Corresponding to a temperature time sequence collected at the head of the train, a humidity time sequence collected at the head of the train, an air pressure time sequence collected at the head of the train, a temperature time sequence collected at the tail of the train, a humidity time sequence collected at the tail of the train and an air pressure time sequence collected at the tail of the train; aggregating time series
Defined as a typical sample of the current category.
4. The passenger comfort oriented train tunnel-passing time length calculation method according to claim 1, wherein the concrete implementation process of the step 5) comprises the following steps:
1) Determining a current sample point and 19 forward sampling points in a time sequence of temperature, humidity and air pressure of the head and the tail of the current position of the train;
2) Phase space reconstruction is carried out by adopting a delay coordinate method, 6 two-dimensional reconstruction matrixes representing the temperature, humidity and air pressure evolution characteristics of the head and the tail of the train are obtained, and the 6 matrixes are combined head and tail according to RGB color space to form a current position characteristic image { c } h ,c f }；
3) Performing convolution operation on the current position characteristic image and the image in the template library
Wherein each g i Are all one-dimensional sequences; { h i ,f i I =0,1,2 \ 82305 is a template image;
4) For all g i Sorting the elements in the sequence from big to small, wherein the determined maximum 5 elements are candidate elements, the position where the candidate elements are located before sorting is a candidate position, and the distance from the candidate position to the exit of the tunnel is s j ,j＝1,2,…5；
5) Taking the average value of the mileage value from the exit of the tunnel corresponding to the candidate position to determine the current template matching output value, namely the template matching output value
5. The passenger comfort oriented train tunnel-passing time length calculation method according to claim 1, wherein the concrete implementation process of the step 6) comprises the following steps:
1) Define input sample I = (T) h ,H h ,P h ,T f ,H f ,P f ) Wherein T is h ＝(t 1 ,t 2 …,t 19 ,t 20 ) The temperature time sequence of the current sample point and the previous 19 sample points of the train head in the tunnel is obtained; h h For a train head humidity time series of length 20, P h For a train head air pressure time sequence, T, of length 20 h Is a train tail temperature time sequence with the length of 20, H h Is a train tail humidity time sequence with the length of 20, P h Is a train tail air pressure time sequence with the length of 20; the output sample is a mileage error value epsilon corresponding to the current position; the input and output combination Y = { I, epsilon } constitutes a modeling sample;
2) Dividing a training sample, a verification sample and a test sample to obtain a training set and a test set;
3) Binary coding is carried out on the features of each dimension in the input sample I, when the code value corresponding to the features of a certain dimension is 1, the features are selected as input variables of the LSSVM model, and when the code value corresponding to the features of the certain dimension is 0, the features of the certain dimension are discarded; randomly initializing and coding 60 dimensional features into 0 or 1;
4) Determining input features based on the current feature coding values, and training an LSSVM model by adopting training set data; inputting the data of the verification set into a trained LSSVM model, and obtaining the output sequence of the model as
Defining an optimization objective function
Wherein
Verifying the true output values of the set;
5) And performing iterative optimization operation by adopting a binary ant lion algorithm to determine the optimal input characteristics and an LSSVM model, wherein the model is an LSSVM mileage prediction error compensation model.
6. The passenger comfort oriented train tunnel-passing time length calculation method according to claim 1, wherein the concrete implementation process of the step 7) comprises the following steps:
1) Acquiring the average air temperature distribution of a train in T min before the train passes through a tunnel, determining the sample category to which the current state belongs, and acquiring a current temperature and humidity sequence by using temperature and humidity sensors arranged at the head and the tail of the train; wherein, the value range of T is [10,20];
2) Determining the sample category to which the current state belongs according to the current average air temperature, and extracting a corresponding template library;
3) Determining the best matching position in the template library, and outputting the template matching output value O of the current sample point M (i)；
4) Obtaining a model input vector I in the current state, substituting the model input vector I into the trained LSSVM model, and obtaining a compensation error output result epsilon (I);
5) And fusing the three-dimensional template matching output value and the LSSVM model output value to obtain the final output value of the distance between the current position and the tunnel outlet, namely O (i) = O M (i)+ε(i)。
Priority Applications (5)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910670306.0A CN110457781B (en) | 2019-07-24 | 2019-07-24 | Passenger comfort-oriented train tunnel-passing time length calculation method |
| JP2021541170A JP7226855B2 (en) | 2019-07-24 | 2020-07-22 | Calculation method, system and storage medium for train passing through tunnel |
| SG11202103099TA SG11202103099TA (en) | 2019-07-24 | 2020-07-22 | Method, system and storage medium for calculating tunnel passing duration of train |
| AU2020317917A AU2020317917B2 (en) | 2019-07-24 | 2020-07-22 | Method for calculating length of time taken for train to pass through tunnel, system, and storage medium |
| PCT/CN2020/103578 WO2021013192A1 (en) | 2019-07-24 | 2020-07-22 | Method for calculating length of time taken for train to pass through tunnel, system, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910670306.0A CN110457781B (en) | 2019-07-24 | 2019-07-24 | Passenger comfort-oriented train tunnel-passing time length calculation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110457781A CN110457781A (en) | 2019-11-15 |
| CN110457781B true CN110457781B (en) | 2022-12-23 |
Family
ID=68483278
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910670306.0A Active CN110457781B (en) | 2019-07-24 | 2019-07-24 | Passenger comfort-oriented train tunnel-passing time length calculation method |
Country Status (5)
| Country | Link |
|---|---|
| JP (1) | JP7226855B2 (en) |
| CN (1) | CN110457781B (en) |
| AU (1) | AU2020317917B2 (en) |
| SG (1) | SG11202103099TA (en) |
| WO (1) | WO2021013192A1 (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110457781B (en) * | 2019-07-24 | 2022-12-23 | 中南大学 | Passenger comfort-oriented train tunnel-passing time length calculation method |
| CN112669238B (en) * | 2020-12-31 | 2022-04-29 | 齐鲁工业大学 | Method for accurately restoring original image of digital image after color correction |
| CN114723106B (en) * | 2022-03-16 | 2023-04-18 | 西南交通大学 | Inter-station goods train cooperative flow distribution method based on fixed-point aggregation mode under mixed condition |
| CN114820623B (en) * | 2022-06-30 | 2022-09-09 | 鲁冉光电(微山)有限公司 | Imaging quality detection method for vehicle-mounted camera |
| CN115230772B (en) * | 2022-08-22 | 2023-07-28 | 南京恩瑞特实业有限公司 | Urban rail train speed and distance measurement redundancy protection method |
| CN116754022B (en) * | 2023-08-22 | 2023-10-27 | 山东浪潮新基建科技有限公司 | On-line detection emergency early warning method and system for cable tunnel |
| CN117251736A (en) * | 2023-11-13 | 2023-12-19 | 杭州海康威视数字技术股份有限公司 | Internet of things data aggregation method and device based on neural network and space-time correlation degree |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102043871A (en) * | 2009-10-26 | 2011-05-04 | 中南大学 | Method for solving aerodynamic effect of long tunnel or tunnel group |
| CN107563425A (en) * | 2017-08-24 | 2018-01-09 | 长安大学 | A kind of method for building up of the tunnel operation state sensor model based on random forest |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020194759A1 (en) * | 1998-04-24 | 2002-12-26 | Badaracco Juan M. | Cinema-like still pictures display for travelling spectators |
| US9578495B2 (en) | 2011-12-19 | 2017-02-21 | Qualcomm Incorporated | Handling impaired wireless connection in a communication system |
| JP5924733B2 (en) | 2012-03-30 | 2016-05-25 | 株式会社Ｊｓｏｌ | Interactive in-car content providing system and content providing method |
| CN105205224B (en) * | 2015-08-28 | 2018-10-30 | 江南大学 | Time difference Gaussian process based on fuzzy curve analysis returns soft-measuring modeling method |
| CN109029372A (en) * | 2018-07-16 | 2018-12-18 | 银河水滴科技（北京）有限公司 | A kind of detection vehicle localization method and device |
| CN109029373A (en) * | 2018-07-16 | 2018-12-18 | 银河水滴科技（北京）有限公司 | The synthesized positioning method and device of vehicle are detected in a kind of tunnel |
| CN110457781B (en) * | 2019-07-24 | 2022-12-23 | 中南大学 | Passenger comfort-oriented train tunnel-passing time length calculation method |
-
2019
- 2019-07-24 CN CN201910670306.0A patent/CN110457781B/en active Active
-
2020
- 2020-07-22 WO PCT/CN2020/103578 patent/WO2021013192A1/en active Application Filing
- 2020-07-22 AU AU2020317917A patent/AU2020317917B2/en active Active
- 2020-07-22 SG SG11202103099TA patent/SG11202103099TA/en unknown
- 2020-07-22 JP JP2021541170A patent/JP7226855B2/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102043871A (en) * | 2009-10-26 | 2011-05-04 | 中南大学 | Method for solving aerodynamic effect of long tunnel or tunnel group |
| CN107563425A (en) * | 2017-08-24 | 2018-01-09 | 长安大学 | A kind of method for building up of the tunnel operation state sensor model based on random forest |
Also Published As
| Publication number | Publication date |
|---|---|
| SG11202103099TA (en) | 2021-04-29 |
| CN110457781A (en) | 2019-11-15 |
| WO2021013192A1 (en) | 2021-01-28 |
| JP7226855B2 (en) | 2023-02-21 |
| JP2022517633A (en) | 2022-03-09 |
| AU2020317917B2 (en) | 2022-11-24 |
| AU2020317917A1 (en) | 2021-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110457781B (en) | Passenger comfort-oriented train tunnel-passing time length calculation method | |
| CN110796168B (en) | Vehicle detection method based on improved YOLOv3 | |
| CN109145939B (en) | Semantic segmentation method for small-target sensitive dual-channel convolutional neural network | |
| CN103106265B (en) | Similar image sorting technique and system | |
| CN110427993B (en) | High-speed train navigation blind area positioning method based on meteorological parameters | |
| Yang et al. | Image-based visibility estimation algorithm for intelligent transportation systems | |
| CN110163069B (en) | Lane line detection method for driving assistance | |
| CN111127499A (en) | Security inspection image cutter detection segmentation method based on semantic contour information | |
| CN105718866A (en) | Visual target detection and identification method | |
| CN103488993A (en) | Crowd abnormal behavior identification method based on FAST | |
| CN111209808B (en) | Unmanned aerial vehicle image semantic segmentation recognition method based on layering processing | |
| CN103426004B (en) | Model recognizing method based on error correcting output codes | |
| CN112465199A (en) | Airspace situation evaluation system | |
| CN113610035B (en) | Rice tillering stage weed segmentation and identification method based on improved coding and decoding network | |
| CN104200226A (en) | Particle filtering target tracking method based on machine learning | |
| CN104408731A (en) | Region graph and statistic similarity coding-based SAR (synthetic aperture radar) image segmentation method | |
| CN116109986A (en) | Vehicle track extraction method based on laser radar and video technology complementation | |
| CN111191510B (en) | Relation network-based remote sensing image small sample target identification method in complex scene | |
| CN111340001B (en) | Traffic jam recognition method based on remote sensing image deep learning | |
| CN116012709B (en) | High-resolution remote sensing image building extraction method and system | |
| CN116946183A (en) | Commercial vehicle driving behavior prediction method considering driving capability and vehicle equipment | |
| CN116468935A (en) | Multi-core convolutional network-based stepwise classification and identification method for traffic signs | |
| CN116311380A (en) | Skin typing method for autonomous learning of small sample data | |
| CN105160336A (en) | Sigmoid function based face recognition method | |
| CN112101132B (en) | Traffic condition prediction method based on graph embedding model and metric learning |
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 |