CN109033521A - A kind of newly built railway ruling grade Study on Decision-making Method for Optimization - Google Patents
A kind of newly built railway ruling grade Study on Decision-making Method for Optimization Download PDFInfo
- Publication number
- CN109033521A CN109033521A CN201810658482.8A CN201810658482A CN109033521A CN 109033521 A CN109033521 A CN 109033521A CN 201810658482 A CN201810658482 A CN 201810658482A CN 109033521 A CN109033521 A CN 109033521A
- Authority
- CN
- China
- Prior art keywords
- layer
- railway
- formula
- locomotive
- ruling grade
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
-
- 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
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of newly built railway ruling grade Study on Decision-making Method for Optimization, the Study on Decision-making Method for Optimization the following steps are included: construct depth convolutional neural networks model first；Then railway case database is established, the every factor that will affect newly built railway ruling grade decision is characterized into grayscale image, and is fused into multichannel image for training network model；Finally propose that a kind of slip sweep, the depth convolutional neural networks model that combined training is completed carry out railway ruling grade decision.Compared with prior art, this method has many advantages, such as that high degree of automation, practical, operational efficiency is high and application prospect is good.
Description
Technical field
The present invention relates to Railway Design methods, and in particular to a kind of newly built railway ruling grade Study on Decision-making Method for Optimization.
Background technique
Ruling grade is the railway main technical standards with global sense, it directly affects the transport energy of route
Power, engineering cost, running cost and traffic safety, in some instances it may even be possible to determine line alignment.With the fast development of Chinese economy,
Railway transportation demand is continuously increased, and simultaneously railway construction gradually changes from Eastern Plain to western mountainous areas, hard and dangerous mountain area
Complex environment makes Environment in Railway Engineering Construction and the contradiction of growing transportation demand further prominent: to better adapt to complicated ground
Shape, geological conditions shorten line length, save engineering-built expense, are a kind of effective using biggish ruling grade
Means；However, route transportation capability is equally influenced by maximum ruling grade, in type of locomotive identical (i.e. traction power is identical)
In the case of, locomotive traction tonnage can be reduced using biggish ruling grade, and then reduce route transportation capability, while also will increase
Running cost and the risk in descending section.In addition, ruling grade belongs to fixed equipment standard, it, will very once railway is built up
Hardly possible change.Therefore, how scientifically and rationally decision is out with locating nature, economy, the ruling grade of social environment best match
The great difficult problem that Alignment Design of Railway Line faces at present.
Newly built railway ruling grade decision is substantially to explore multidimensional influence factor (such as: orographic condition, transportation demand)
With the mapping principle of ruling grade, to select preferred plan.Traditional ruling grade Study on Decision-making Method for Optimization usually first assumes to want
Rule between element meets a certain mathematical model expression formula, then by statistical regression model parameter, obtains mapping principle.As southwest is handed over
King's residence of a high official professor of logical university carries out statistical regression by the design data to thousands of kilometers of China mountain railway, has obtained limitation slope
The universal experience formula (1) of degree and engineering cost mapping principle.A is engineering cost in formula, and I is ruling grade, and a, b, c are to pass through
The model parameter relevant to orographic condition that statistical regression obtains.
However, the mapping principle between multidimensional influence factor and ruling grade is complicated and nonlinear, it is difficult to pass through one
Fixed functional relation is complete, accurately expresses.Therefore, there is an urgent need to one kind can comprehensively, accurately identify multidimensional influence
The method of mapping principle between factor and ruling grade realizes the Optimal Decision-making of newly built railway ruling grade.
Summary of the invention
The technical problems to be solved by the present invention are: provide one kind can comprehensively, accurately identify multidimensional influence factor and limitation
The method of mapping principle between the gradient, and then realize the Optimal Decision-making of newly built railway ruling grade.
In order to solve the above-mentioned technical problem, a kind of the technical solution adopted by the present invention are as follows: newly built railway ruling grade optimization
Decision-making technique, comprising the following steps:
S1: building is used for the depth convolutional neural networks model of newly built railway ruling grade Optimal Decision-making；
S2: establish the training dataset D for training depth convolutional neural networkstrainWith validation data set Dvalidate；
S2-1: collect N1Item uses the built mixed passenger and freight railway case of different ruling grades, establishes railway case data collection
D1；
S2-2: it is based on railway case data collection D1In the start, end position of each rail track divide the square of each railway case
Shape survey region, and the grid altitude data information in each rectangle survey region is extracted, establish railway case altitude data collection D2；
S2-3: it is based on D2In each railway case study region grid altitude data information, draw each rectangle survey region
Elevation grayscale image Pelevation, establish the elevation grayscale image for characterizing each railway case study region landform altitude variation characteristic
Collect Delevation；
S2-4: it is based on D2In each railway case study region grid altitude data information, draw each rectangle survey region
Gradient grayscale image Pslope, establish the gradient gray scale atlas D for characterizing each railway case study region terrain slope featureslope；
S2-5: different classifications of rail are characterized as the different grayscale image of gray value, according to D1In each railway case reality
Grade draws classification of rail grayscale image P corresponding with each railway caseclassification, establish classification of rail gray scale atlas
Dclassification；
S2-6: different type of locomotive are characterized as the different grayscale image of gray value, according to D1In each railway case used
Practical type of locomotive, draw type of locomotive grayscale image P corresponding with each railway caselocomotive, establish type of locomotive
Gray scale atlas Dlocomotive；
S2-7: the elevation gray scale atlas D based on foundationelevation, gradient gray scale atlas Dslope, classification of rail gray scale atlas
Dclassification, type of locomotive gray scale atlas Dlocomotive, merge D1In each railway case elevation grayscale image Pelevation、
Gradient grayscale image Pslope, classification of rail grayscale image PclassificationWith type of locomotive grayscale image Plocomotive, formation can characterize respectively
The four-way figure P of railway case informationmerge, and establish data set Dmerge；
S2-8: by data set DmergeIt is 333 × 333 that the four-way figure of each railway case information of middle characterization, which is cut into size,
The picture of pixel, and label data is assigned, the ruling grade value that label data is actually used by each railway case；
S2-9: by S2-8Middle gained tape label datagram presses the ratio cut partition of 4:1, establishes for training depth convolutional Neural net
The training dataset D of networktrainWith validation data set Dvalidate；
S3: use S2The training dataset D of foundationtrainThe constructed network model of training, and use S2The verifying number of foundation
According to collection DvalidateModel accuracy is verified, is obtained by training and the depth convolutional neural networks model verified；
S4: in addition collect N2Item and data set D1Middle different built mixed passenger and freight railway case, and according to step S2-2Extremely
S2-7Generate the four-way figure P of characterization railway case informationmerge, establish test data set Dtest；
S5: it proposes a kind of slip sweep, trained depth convolutional neural networks model is pressed from left to right, by upper
Sequential scan data set D undertestMiddle characterization each railway case elevation information, grade information, classification of rail information, locomotive
The four-way figure of type information determines D according to the output times of each ruling grade valuetestIn each railway case limitation slope
Spend recommendation.
Further, the step S1The depth convolutional neural networks model of middle building include 5 convolutional layers (Conv), 3
A pond layer (Pool), 2 full articulamentums (FC) and 1 Softmax output layer:
The convolution kernel size that (1) first convolutional layer (Conv1) uses is 33 × 33 × 3, step size 4, convolution kernel
Connection amendment linear unit (ReLU) is used as nonlinear activation function after number is 96, Conv1, and model is made to have non-linear spy
Sign；
(2) Conv1 connects first pond layer (Pool1) after Nonlinear Processing, the pond core size that Pool1 is used for
4 × 4, step size 2；
(3) second convolutional layer (Conv2) is connected after Pool1, the convolution kernel size that Conv2 is used walks for 3 × 3 × 96
Width size is 1, and connection amendment linear unit (ReLU) carries out Nonlinear Processing after convolution kernel number is 256, Conv2；
(4) Conv2 connects second pond layer (Pool2) after Nonlinear Processing, the pond core size that Pool2 is used for
3 × 3, step size 2；
(5) third convolutional layer (Conv3) is connected after Pool2, the convolution kernel size that Conv3 is used walks for 3 × 3 × 256
Width size is 1, and connection amendment linear unit (ReLU) carries out Nonlinear Processing after convolution kernel number is 384, Conv3；
(6) Conv3 connects the 4th convolutional layer (Conv4) after Nonlinear Processing, the convolution kernel size that Conv4 is used for
3 × 3 × 384, step size 1, connection amendment linear unit (ReLU) carries out non-linear after convolution kernel number is 384, Conv4
Processing；
(7) Conv4 connects the 5th convolutional layer (Conv5) after Nonlinear Processing, the convolution kernel size that Conv5 is used for
3 × 3 × 384, step size 1, connection amendment linear unit (ReLU) carries out non-linear after convolution kernel number is 256, Conv5
Processing；
(8) Conv5 connects third pond layer (Pool3) after Nonlinear Processing, the pond core size that Pool3 is used for
3 × 3, step size 2；
(9) first full articulamentum (FC1) is connected after Pool3, to prevent over-fitting, arrives FC1 at Pool3 layers
Layer connection uses dropout function, and connection amendment linear unit (ReLU) carries out Nonlinear Processing after FC1；
(10) FC1 connects second full articulamentum (FC2) after Nonlinear Processing, and prevents from producing using dropout function
Over-fitting is given birth to, connection amendment linear unit (ReLU) carries out Nonlinear Processing after FC2；
(11) FC2 connects Softmax output layer after Nonlinear Processing, pushes away for exporting newly built railway ruling grade value
It recommends.
Further, the step S2-1The railway case of middle collection covers different brackets railway and different type of locomotive.
Further, the step S2-2Middle railway rectangle survey region division methods are as follows: setting certain railway case route
Start, end are respectively Si: (xSi,ySi) and Ei: (xEi,yEi), then the survey region of the railway case is with SiAnd EiFor angle steel joint,
With | xEi-xSi| to grow, | yEi-ySi| it is wide rectangular area.
Further, the step S2-3、S2-4In each railway case rectangle survey region elevation grayscale image and the gradient
Grayscale image is all made of Global Mapper Software on Drawing.
Further, the step S2-5In, classification of rail grayscale image PclassificationSize and the railway case square
Shape survey region size is identical.
Further, the step S2-6In, type of locomotive grayscale image PlocomotiveSize and use the type of locomotive iron
The rectangle survey region size of road case is identical.
Further, the step S2-7In each railway case four-way figure PmergeTo use computer vision library
The elevation grayscale image P of merge function in OpenCV to each railway caseelevation, gradient grayscale image Pslope, classification of rail
Grayscale image PclassificationWith locomotive type grayscale image PlocomotiveIt is obtained after being merged.
Further, the step S3Network model constructed by middle training is based on S2The label data collection of foundation
Dtrain, the connection weight of each interlayer in network model is constantly updated by gradient descent algorithm, specific as follows:
(1) Softmax layers of connection weight update
The Softmax layers of ruling grade value recommended for output model, this layer according to the output valve of each neuron of preceding layer,
The output probability of each ruling grade value is calculated, thus the ruling grade for selecting the maximum value of slope of output probability to recommend as model
Value, shown in function representation such as formula (2):
In formula: P (y(i)=j | x(i)；It W) is to select j-th of value using the i-th picture as input data in Softmax layer choosing and make
For the probability of ruling grade, x(i)For Softmax layers of input datas (i.e. the output data of preceding layer), W be Softmax layer and
The connection weight of preceding layer.
Model loss function E is established based on Softmax function, shown in function expression such as formula (3):
In formula: 1 { y(i)=j } it is logical expression, if i input picture marks are j-th of ruling grade, 1 { y(i)=j }=1, otherwise 1 { y(i)=j }=0, λ be weight attenuation coefficient.
Based on loss function E, the residual error of Softmax layers of each neuron can be calculated by formula (4):
The connection weight of Softmax layers of each neuron is updated by formula (5), formula (6):
(2) full articulamentum connection weight updates
Each neuron of full articulamentum is connected with upper one layer of all neurons, and connection weight more new formula is such as
Under:
In formula: WlFor the connection weight matrix of current layer (full articulamentum) each neuron, blFor the company of each neuron of current layer
Bias vector is connect, α is learning rate.
Partial derivative of the loss function to each neuron connection weight of full articulamentumConnect with to each neuron of full articulamentum
Connect the partial derivative of biasingFormula (9) can be pressed respectively and formula (10) calculates.
In formula: xl-1For the output vector of an articulamentum on current layer (full articulamentum), δlIt is each for current layer (full articulamentum)
The residual error of neuron, can be according to the residual error δ of each neuron of articulamentum thereafterl+1It calculates.
In formula: Wl+1For the connection weight matrix of current layer (full articulamentum) each neuron of articulamentum afterwards, f () is ReLU
Activation primitive.
(3) convolutional layer connection weight updates
Each neuron of convolutional layer is connected by convolution kernel with preceding layer, and each convolution kernel connection weight more new formula is as follows:
In formula:For the connection weight matrix of current layer (convolutional layer) d-th of convolution kernel,For current layer (convolutional layer)
The connection bias vector of d-th of convolution kernel, α are learning rate.
Loss function is to each connection weight partial derivative of current layer (convolutional layer) d-th of convolution kernelCalculation formula such as
Under:
In formula:For the output valve of a characteristic pattern of current layer (convolutional layer) previous articulamentum d ', Dl-1For current layer (volume
Lamination) previous articulamentum feature map number,For the residual matrix of current layer (convolutional layer) d-th of characteristic pattern.
Loss function respectively connects biasing partial derivative to current layer (convolutional layer) d-th of convolution kernelCalculation formula such as
Under:
In formula:For the connection bias vector of d-th of characteristic pattern in current layer (convolutional layer),WithIt is respectively current
The line number and columns of d-th of characteristic pattern in layer (convolutional layer),For i row in d-th of characteristic pattern in current layer (convolutional layer),
The residual values of j column.
The residual error of current layer (convolutional layer) is the layer residual computations by backpropagation, based on latter connection.Such as current layer
Latter linked (convolutional layer) is pond layer, then the residual matrix of current layer (convolutional layer) d-th of characteristic pattern is calculated by formula (17).
In formula: Xl-1For the output matrix of current layer (convolutional layer) previous articulamentum,Current layer (convolutional layer) latter company
Connect the residual matrix of d-th of characteristic pattern in layer.
It is convolutional layer as current layer (convolutional layer) is latter linked, then the weight matrix of current layer (convolutional layer) is based on formula (18)
It calculates.
In formula:For the residual matrix of a characteristic pattern of d ' in current layer (convolutional layer) latter articulamentum,For
D layers of weight matrix of a convolution kernel of d " of current layer (convolutional layer) latter articulamentum,D-th of current layer (convolutional layer)
The output matrix of characteristic pattern.
Further, the step S5In slip sweep it is specific as follows: as scan test data collection DtestIn certain
When four-way figure, the ruling grade recommendation in 333 × 333 pixel size regions in the exportable four-way figure is scanned every time, it is complete
After scanning at whole four-way figure, limit of the most value of slope of output times as the characterized railway case of the four-way figure is chosen
Gradient recommendation processed.
The beneficial effects of the present invention are: deep learning simulates the layered structure of brain, can be from mass data automatically
Obtaining, there is the multilayer feature of hierarchy to express, without given mathematic(al) representation, explore input data with it is defeated
Existing potential rule between data out.The present invention program is practical using deep learning algorithm progress newly built railway ruling grade decision
It is feasible.The present invention is based on the convolutional neural networks in deep learning algorithm, propose a kind of newly built railway ruling grade Optimal Decision-making
Method, this method identify the mapping principle between multidimensional influence factor and ruling grade by study manual decision's experience, realize new
Build class people's decision of railway ruling grade.The present invention program uses slip sweep, realizes and limits different railway cases
The decision of the gradient.The method of the present invention high degree of automation, practical, operational efficiency is high, has good popularization and application foreground.
Detailed description of the invention
Fig. 1 is the flow diagram of newly built railway ruling grade Study on Decision-making Method for Optimization of the invention；
Fig. 2 is the depth convolutional neural networks model of the embodiment of the present invention；
Fig. 3 is the structural schematic diagram of the slip sweep of the embodiment of the present invention.
Specific embodiment
To explain the technical content, the achieved purpose and the effect of the present invention in detail, below in conjunction with embodiment and cooperate attached
Figure is explained.
The embodiment of the present invention one is a kind of newly built railway ruling grade Study on Decision-making Method for Optimization, as shown in Figure 1, the Optimal Decision-making
Method the following steps are included:
S1: building is used for the depth convolutional neural networks model of newly built railway ruling grade Optimal Decision-making, constructed net
Network model includes 5 convolutional layers (Conv), 3 pond layers (Pool), 2 full articulamentums (FC) and 1 Softmax output layer:
The convolution kernel size that (1) first convolutional layer (Conv1) uses is 33 × 33 × 3, step size 4, convolution kernel
Connection amendment linear unit (ReLU) is used as nonlinear activation function after number is 96, Conv1, and model is made to have non-linear spy
Sign；
(2) Conv1 connects first pond layer (Pool1) after Nonlinear Processing, the pond core size that Pool1 is used for
4 × 4, step size 2；
(3) second convolutional layer (Conv2) is connected after Pool1, the convolution kernel size that Conv2 is used walks for 3 × 3 × 96
Width size is 1, and connection amendment linear unit (ReLU) carries out Nonlinear Processing after convolution kernel number is 256, Conv2；
(4) Conv2 connects second pond layer (Pool2) after Nonlinear Processing, the pond core size that Pool2 is used for
3 × 3, step size 2；
(5) third convolutional layer (Conv3) is connected after Pool2, the convolution kernel size that Conv3 is used walks for 3 × 3 × 256
Width size is 1, and connection amendment linear unit (ReLU) carries out Nonlinear Processing after convolution kernel number is 384, Conv3；
(6) Conv3 connects the 4th convolutional layer (Conv4) after Nonlinear Processing, the convolution kernel size that Conv4 is used for
3 × 3 × 384, step size 1, connection amendment linear unit (ReLU) carries out non-linear after convolution kernel number is 384, Conv4
Processing；
(7) Conv4 connects the 5th convolutional layer (Conv5) after Nonlinear Processing, the convolution kernel size that Conv5 is used for
3 × 3 × 384, step size 1, connection amendment linear unit (ReLU) carries out non-linear after convolution kernel number is 256, Conv5
Processing；
(8) Conv5 connects third pond layer (Pool3) after Nonlinear Processing, the pond core size that Pool3 is used for
3 × 3, step size 2；
(9) first full articulamentum (FC1) is connected after Pool3, to prevent over-fitting, arrives FC1 at Pool3 layers
Layer uses dropout function, and connection amendment linear unit (ReLU) carries out Nonlinear Processing after FC1；
(10) FC1 connects second full articulamentum (FC2) after Nonlinear Processing, and prevents from producing using dropout function
Over-fitting is given birth to, connection amendment linear unit (ReLU) carries out Nonlinear Processing after FC2；
(11) FC2 connects Softmax output layer after Nonlinear Processing, recommends for exporting newly built railway ruling grade
Value.
S2: establish the training dataset D for training depth convolutional neural networkstrainWith validation data set Dvalidate；
S2-1: it collects and uses 6 ‰, 12 ‰, 24 ‰ for mixed passenger and freight railway case 246 of ruling grade, collected iron
Road case covers I grades, II grades, III level, four kinds of classifications of rail of IV grade, SS 1,3 type of Shaoshan, Shaoshan 4 type, three kinds of locomotive types
Number, establish railway case data collection D1；
S2-2: it is based on railway case data collection D1In the start, end position of each rail track divide the square of each railway case
Shape survey region, and the grid altitude data information in each rectangle survey region is extracted, establish railway case altitude data collection D2；
S2-3: it is based on D2In each railway case study region grid altitude data information, draw each rectangle survey region
Elevation grayscale image Pelevation, establish the elevation grayscale image for characterizing each railway case study region landform altitude variation characteristic
Collect Delevation；
S2-4: it is based on D2In each railway case study region grid altitude data information, draw each rectangle survey region
Gradient grayscale image Pslope, establish the gradient gray scale atlas D for characterizing each railway case study region terrain slope featureslope；
S2-5: four kinds of classifications of rail are characterized with the grayscale image that gray value is 0,40,80,120 respectively, and according to D1In each item
The actual grade of railway case draws classification of rail grayscale image P corresponding with each railway caseclassification, establish iron
Road level gray atlas Dclassification；
S2-6: three kinds of SS 1,3 type of Shaoshan and 4 type of Shaoshan are characterized with the grayscale image that gray value is 160,200,240 respectively
Electric locomotive model, and according to D1In practical type of locomotive used in each railway case, draw and each railway case phase
Corresponding type of locomotive grayscale image Plocomotive, establish type of locomotive gray scale atlas Dlocomotive；
S2-7: the elevation gray scale atlas D based on foundationelevation, gradient gray scale atlas Dslope, classification of rail gray scale atlas
Dclassification, type of locomotive gray scale atlas Dlocomotive, merge D1In each railway case elevation grayscale image Pelevation、
Gradient grayscale image Pslope, classification of rail grayscale image PclassificationWith type of locomotive grayscale image Plocomotive, formation can characterize respectively
The four-way figure P of railway case informationmerge, and establish data set Dmerge；
S2-8: by data set DmergeIt is 333 × 333 that the four-way figure of each railway case information of middle characterization, which is cut into size,
The picture of pixel, and label data is assigned, the ruling grade value that label data is actually used by each railway case；
S2-9: by S2-8Middle gained tape label picture presses the ratio cut partition of 4:1, establishes for training depth convolutional neural networks
Training dataset DtrainWith validation data set Dvalidate；
S3: use S2The training dataset D of foundationtrainThe constructed network model of training, and use S8The verifying number of foundation
According to collection DvalidateModel accuracy is verified, is obtained by training and the depth convolutional neural networks model verified.This is trained and tests
Time-consuming 35 minutes 9 hours (1080 video card of i7 processor, 16G memory and GTX) is demonstrate,proved, the depth convolution that precision is 83.35% is obtained
Neural network model.
S4: in addition collect 36 and data set D1Middle different built mixed passenger and freight railway case, and according to step S2-2Extremely
S2-7Establish the four-way figure P of characterization railway case informationmerge, establish test data set Dtest；
S5: it proposes a kind of slip sweep, trained depth convolutional neural networks model is pressed from left to right, by upper
Sequential scan data set D undertestMiddle characterization each railway case elevation information, grade information, classification of rail information, locomotive
The four-way figure of type information, and according to the output times of each ruling grade value, determine DtestIn each railway case limit
Gradient recommendation processed.In 36 railway cases of this test, the ruling grade of 34 railway cases has been obtained accurately certainly
Plan (i.e. the ruling grade value of model recommendation is identical as the ruling grade value of manual decision), accuracy rate is up to 94.44%.
Slip sweep alleged by the present invention refers to by scanning whole picture, according to the output times of different ruling grades
Decision ruling grade value.
In conclusion the present invention provides a kind of newly built railway ruling grade Study on Decision-making Method for Optimization, depth convolution is constructed first
Then neural network model establishes railway case database, will affect every factor characterization of newly built railway ruling grade decision
At grayscale image, and multichannel image is fused into for training network model；Finally propose a kind of slip sweep, combined training
The depth convolutional neural networks model of completion carries out newly built railway ruling grade decision.
The above description is only an embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalents made by bright specification and accompanying drawing content are applied directly or indirectly in relevant technical field, similarly include
In scope of patent protection of the invention.
Claims (9)
1. a kind of newly built railway ruling grade Study on Decision-making Method for Optimization, it is characterised in that: the following steps are included:
S1: building is used for the depth convolutional neural networks model of newly built railway ruling grade Optimal Decision-making；
S2: establish the training dataset D for training depth convolutional neural networkstrainWith validation data set Dvalidate；
S2-1: collect N1Item uses the built mixed passenger and freight railway case of different ruling grades, establishes railway case data collection D1；
S2-2: it is based on the railway case data collection D1In the start, end position of each rail track divide the square of each railway case
Shape survey region, and the grid altitude data information in each rectangle survey region is extracted, establish railway case altitude data collection D2；
S2-3: it is based on D2In each railway case study region grid altitude data information, draw the elevation of each rectangle survey region
Grayscale image Pelevation, establish the elevation gray scale atlas for characterizing each railway case study region landform altitude variation characteristic
Delevation；
S2-4: it is based on D2In each railway case study region grid altitude data information, draw the gradient of each rectangle survey region
Grayscale image Pslope, establish the gradient gray scale atlas D for characterizing each railway case study region terrain slope featureslope；
S2-5: different classifications of rail are characterized as the different grayscale image of gray value, according to D1In each railway case actual grade,
Draw classification of rail grayscale image P corresponding with each railway caseclassification, establish classification of rail gray scale atlas
Dclassification；
S2-6: different type of locomotive are characterized as the different grayscale image of gray value, according to D1In it is real used in each railway case
Border type of locomotive draws type of locomotive grayscale image P corresponding with each railway caselocomotive, establish type of locomotive gray scale
Atlas Dlocomotive；
S2-7: the elevation gray scale atlas D based on foundationelevation, gradient gray scale atlas Dslope, classification of rail gray scale atlas
Dclassification, type of locomotive gray scale atlas Dlocomotive, merge D1In each railway case elevation grayscale image Pelevation、
Gradient grayscale image Pslope, classification of rail grayscale image PclassificationWith type of locomotive grayscale image Plocomotive, formation can characterize respectively
The four-way figure P of railway case informationmerge, and establish data set Dmerge；
S2-8: by data set DmergeIt is 333 × 333 pixels that the four-way figure of each railway case information of middle characterization, which is cut into size,
Picture, and assign label data, the ruling grade value that label data is actually used by each railway case；
S2-9: by S2-8Middle gained tape label datagram presses the ratio cut partition of 4:1, establishes for training depth convolutional neural networks
Training dataset DtrainWith validation data set Dvalidate；
S3: use S2The training dataset D of foundationtrainThe constructed network model of training, and use S2The validation data set of foundation
DvalidateModel accuracy is verified, is obtained by training and the depth convolutional neural networks model verified；
S4: in addition collect N2Item and data set D1Middle different built mixed passenger and freight railway case, and according to step S2-2To S2-7It is raw
At the four-way figure P of characterization railway case informationmerge, establish test data set Dtest；
S5: a kind of slip sweep is proposed, by trained depth convolutional neural networks model by from left to right, from top to bottom
Sequential scan data set DtestMiddle characterization each railway case elevation information, grade information, classification of rail information, type of locomotive
The four-way figure of information determines D according to the output times of each ruling grade valuetestIn the ruling grade of each railway case push away
Recommend value.
2. newly built railway ruling grade Study on Decision-making Method for Optimization according to claim 1, it is characterised in that: the step S1In,
The depth convolutional neural networks model includes 5 convolutional layers, 3 pond layers, 2 full articulamentums and 1 Softmax output
Layer:
1) the convolution kernel size that first convolutional layer uses is 33 × 33 × 3, step size 4, and convolution kernel number is 96, first
Connection amendment linear unit makes model have nonlinear characteristic as nonlinear activation function after a convolutional layer；
2) first convolutional layer connects first pond layer, the pond core size that first pond layer uses after Nonlinear Processing
It is 4 × 4, step size 2；
3) second convolutional layer is connected after first pond layer, the convolution kernel size that second convolutional layer uses for 3 × 3 × 96,
Step size is 1, and convolution kernel number is 256, and connection amendment linear unit carries out Nonlinear Processing after second convolutional layer；
4) second convolutional layer connects second pond layer, the pond core size that second pond layer uses after Nonlinear Processing
It is 3 × 3, step size 2；
5) third convolutional layer is connected after second pond layer, the convolution kernel size that third convolutional layer uses for 3 × 3 × 256,
Step size is 1, and convolution kernel number is 384, and connection amendment linear unit carries out Nonlinear Processing after third convolutional layer；
6) third convolutional layer connects the 4th convolutional layer, the convolution kernel size that the 4th convolutional layer uses after Nonlinear Processing
It is 3 × 3 × 384, step size 1, convolution kernel number is 384, and connection amendment linear unit carries out non-after the 4th convolutional layer
Linear process；
7) the 4th convolutional layer connects the 5th convolutional layer, the convolution kernel size that the 5th convolutional layer uses after Nonlinear Processing
It is 3 × 3 × 384, step size 1, convolution kernel number is 256, and connection amendment linear unit carries out non-after the 5th convolutional layer
Linear process；
8) the 5th convolutional layer connects third pond layer, the pond core size that third pond layer uses after Nonlinear Processing
It is 3 × 3, step size 2；
9) first full articulamentum is connected after the layer of third pond, to prevent over-fitting, layer is arrived in third pond
First full articulamentum connection uses dropout function, and connection amendment linear unit carries out non-linear after first full articulamentum
Processing；
10) first full articulamentum connects second full articulamentum after Nonlinear Processing, and prevents from producing using dropout function
Over-fitting is given birth to, connection amendment linear unit carries out Nonlinear Processing after second full articulamentum；
11) second full articulamentum connects Softmax output layer after Nonlinear Processing, for exporting newly built railway ruling grade
Recommendation.
3. newly built railway ruling grade Study on Decision-making Method for Optimization according to claim 1, it is characterised in that: the step S2-1
In, the railway case of collection covers different brackets railway and different type of locomotive.
4. newly built railway ruling grade Study on Decision-making Method for Optimization according to claim 1, it is characterised in that: the step S2-2
In, the method for dividing rectangle survey region based on rail track start, end position is as follows:
If certain railway case route start, end are respectively Si: (xSi,ySi) and Ei: (xEi,yEi), then the research of the railway case
Region is with SiAnd EiFor angle steel joint, with | xEi-xSi| to grow, | yEi-ySi| it is wide rectangular area.
5. newly built railway ruling grade Study on Decision-making Method for Optimization according to claim 1, it is characterised in that: the step S2-5
In, classification of rail grayscale image PclassificationSize it is identical as the rectangle survey region size of the railway case.
6. newly built railway ruling grade Study on Decision-making Method for Optimization according to claim 1, it is characterised in that: the step S2-6
In, type of locomotive grayscale image PlocomotiveSize with using the type of locomotive railway case rectangle survey region size phase
Together.
7. newly built railway ruling grade Study on Decision-making Method for Optimization according to claim 1, it is characterised in that: the step S2-7
In each railway case four-way figure PmergeTo use the merge function in the OpenCV of computer vision library to each railway case
The elevation grayscale image P of exampleelevation, gradient grayscale image Pslope, classification of rail grayscale image PclassificationWith locomotive type gray scale
Scheme PlocomotiveIt is obtained after being merged.
8. newly built railway ruling grade Study on Decision-making Method for Optimization according to claim 1, it is characterised in that: the step S3In
The constructed network model of training is based on S2The label data collection D of foundationtrain, network is constantly updated by gradient descent algorithm
The connection weight of each interlayer in model, specific as follows:
1) Softmax layers of connection weight update:
The Softmax layers of ruling grade value recommended for output model, this layer are calculated according to the output valve of each neuron of preceding layer
The output probability of each ruling grade value, thus the ruling grade value for selecting the maximum value of slope of output probability to recommend as model,
Shown in its function representation such as formula (2):
In formula: P (y(i)=j | x(i)；It W) is to select j-th of value as limit in Softmax layer choosing using the i-th picture as input data
The probability of the gradient processed, x(i)For Softmax layers of input data, W is the connection weight of Softmax layers with preceding layer；
Model loss function E is established based on Softmax function, shown in function expression such as formula (3):
In formula: 1 { y(i)=j } it is logical expression, if i-th input picture mark is j-th of ruling grade value, 1 { y(i)=j }=1, otherwise 1 { y(i)=j }=0, λ be weight attenuation coefficient；
Based on loss function E, the residual error of Softmax layers of each neuron is calculated by formula (4):
The connection weight of Softmax layers of each neuron is updated by formula (5), formula (6):
2) full articulamentum connection weight updates:
Each neuron of full articulamentum is connected with upper one layer of all neurons, and connection weight more new formula is as follows:
In formula: WlFor the connection weight matrix of each neuron of current layer, blFor the connection bias vector of each neuron of current layer, α is
Learning rate；
Partial derivative of the loss function to each neuron connection weight of full articulamentumIt is connected partially with to each neuron of full articulamentum
The partial derivative setIt is calculated respectively by formula (9) and formula (10)；
In formula: xl-1For the output vector of an articulamentum on current layer, δlIt, can be according to connecting thereafter for the residual error of each neuron of current layer
Meet the residual error δ of each neuron of layerl+1It calculates；
δl=(Wl+1)δl+1⊙f′(Wlxl-1+bl) (11)
In formula: Wl+1For the connection weight matrix of each neuron of articulamentum after current layer, f () is ReLU activation primitive；
3) convolutional layer connection weight updates:
Each neuron of convolutional layer is connected by convolution kernel with preceding layer, and each convolution kernel connection weight more new formula is as follows:
In formula:For the connection weight matrix of d-th of convolution kernel of current layer,It is biased for the connection of d-th of convolution kernel of current layer
Vector, α are learning rate；
Loss function is to each connection weight partial derivative of d-th of convolution kernel of current layerCalculation formula it is as follows:
In formula:For the output valve of the previous a characteristic pattern of articulamentum d ' of current layer, Dl-1For the spy of the previous articulamentum of current layer
Map number is levied,For the residual matrix of d-th of characteristic pattern of current layer；
Loss function respectively connects biasing partial derivative to d-th of convolution kernel of current layerCalculation formula it is as follows:
In formula:For the connection bias vector of d-th of characteristic pattern in current layer,WithD-th of feature respectively in current layer
The line number and columns of figure,For i row in d-th of characteristic pattern in current layer, the residual values of j column；
The residual error of current layer is the layer residual computations by backpropagation, based on latter connection；If latter linked current layer is pond
Change layer, then the residual matrix of d-th of characteristic pattern of current layer is calculated by formula (17)；
In formula: Xl-1For the output matrix of the previous articulamentum of current layer,D-th characteristic pattern is residual in the latter articulamentum of current layer
Poor matrix；
If latter linked current layer is convolutional layer, the weight matrix of current layer is calculated by formula (18):
In formula:For the residual matrix of a characteristic pattern of d ' in the latter articulamentum of current layer,For the latter company of current layer
D layers of weight matrix of a convolution kernel of d " of layer are connect,The output matrix of d-th of characteristic pattern of current layer.
9. newly built railway ruling grade Study on Decision-making Method for Optimization according to claim 1, it is characterised in that: the step S5In,
As scan test data collection DtestIn certain four-way figure when, it is big to scan 333 × 333 pixels in the exportable four-way figure every time
The ruling grade recommendation of zonule selects the value of slope conduct for then taking output times most after completing whole four-way figure scanning
The ruling grade recommendation of the characterized railway case of the four-way figure.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810658482.8A CN109033521B (en) | 2018-06-25 | 2018-06-25 | Newly-built railway slope-limiting optimization decision method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810658482.8A CN109033521B (en) | 2018-06-25 | 2018-06-25 | Newly-built railway slope-limiting optimization decision method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109033521A true CN109033521A (en) | 2018-12-18 |
| CN109033521B CN109033521B (en) | 2021-04-20 |
Family
ID=64610700
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810658482.8A Active CN109033521B (en) | 2018-06-25 | 2018-06-25 | Newly-built railway slope-limiting optimization decision method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109033521B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118070217A (en) * | 2024-01-29 | 2024-05-24 | 中南大学 | Railway main technical standard recommendation method, computer storage medium and equipment |
| CN118643578A (en) * | 2024-08-14 | 2024-09-13 | 兰州交通大学 | Plateau area railway line design method based on deep convolution generation countermeasure network |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105243233A (en) * | 2015-11-04 | 2016-01-13 | 中南大学 | Line-station collaborative optimization method for railway in complicated mountain area |
| CN106777752A (en) * | 2016-12-30 | 2017-05-31 | 华东交通大学 | A kind of bullet train follows the trail of operation curve Optimal Setting method |
| CN106844949A (en) * | 2017-01-18 | 2017-06-13 | 清华大学 | A kind of training method for realizing the controllable two-way LSTM models of locomotive section |
| CN106910185A (en) * | 2017-01-13 | 2017-06-30 | 陕西师范大学 | A kind of DBCC disaggregated models and construction method based on CNN deep learnings |
| CN107133960A (en) * | 2017-04-21 | 2017-09-05 | 武汉大学 | Image crack dividing method based on depth convolutional neural networks |
| CN107689035A (en) * | 2017-08-30 | 2018-02-13 | 广州华多网络科技有限公司 | A kind of homography matrix based on convolutional neural networks determines method and device |
| US20180068216A1 (en) * | 2015-03-13 | 2018-03-08 | Institute Of Acoustics, Chinese Academy Of Sciences | Big data processing method based on deep learning model satisfying k-degree sparse constraint |
-
2018
- 2018-06-25 CN CN201810658482.8A patent/CN109033521B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180068216A1 (en) * | 2015-03-13 | 2018-03-08 | Institute Of Acoustics, Chinese Academy Of Sciences | Big data processing method based on deep learning model satisfying k-degree sparse constraint |
| CN105243233A (en) * | 2015-11-04 | 2016-01-13 | 中南大学 | Line-station collaborative optimization method for railway in complicated mountain area |
| CN106777752A (en) * | 2016-12-30 | 2017-05-31 | 华东交通大学 | A kind of bullet train follows the trail of operation curve Optimal Setting method |
| CN106910185A (en) * | 2017-01-13 | 2017-06-30 | 陕西师范大学 | A kind of DBCC disaggregated models and construction method based on CNN deep learnings |
| CN106844949A (en) * | 2017-01-18 | 2017-06-13 | 清华大学 | A kind of training method for realizing the controllable two-way LSTM models of locomotive section |
| CN107133960A (en) * | 2017-04-21 | 2017-09-05 | 武汉大学 | Image crack dividing method based on depth convolutional neural networks |
| CN107689035A (en) * | 2017-08-30 | 2018-02-13 | 广州华多网络科技有限公司 | A kind of homography matrix based on convolutional neural networks determines method and device |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118070217A (en) * | 2024-01-29 | 2024-05-24 | 中南大学 | Railway main technical standard recommendation method, computer storage medium and equipment |
| CN118643578A (en) * | 2024-08-14 | 2024-09-13 | 兰州交通大学 | Plateau area railway line design method based on deep convolution generation countermeasure network |
| CN118643578B (en) * | 2024-08-14 | 2024-10-18 | 兰州交通大学 | Plateau area railway line design method based on deep convolution generation countermeasure network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109033521B (en) | 2021-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gigović et al. | Application of the GIS-DANP-MABAC multi-criteria model for selecting the location of wind farms: A case study of Vojvodina, Serbia | |
| Yu et al. | Prediction of highway tunnel pavement performance based on digital twin and multiple time series stacking | |
| CN113255811B (en) | BIM-based beam bridge multi-source heterogeneous data fusion decision system | |
| CN106780089B (en) | Permanent basic farmland planning method based on neural network cellular automaton model | |
| CN109636150A (en) | A kind of method for building up and its system of smart city " more rule unifications " appraisement system | |
| CN106991374A (en) | Handwritten Digit Recognition method based on convolutional neural networks and random forest | |
| Lawal et al. | A geographic information system and multi-criteria decision analysis in proposing new recreational park sites in universiti teknologi malaysia | |
| Srisawat et al. | Development of decision support system for evaluating spatial efficiency of regional transport logistics | |
| CN105931116A (en) | Automated credit scoring system and method based on depth learning mechanism | |
| CN102360458A (en) | Fuzzy evaluation method for selection of immigrant settlement areas | |
| CN107527133A (en) | Group of cities National land space utilizes increased quality aid decision-making system | |
| CN108876487A (en) | A kind of industrial plot estimation method based on big data and intelligent decision mechanism | |
| CN109033521A (en) | A kind of newly built railway ruling grade Study on Decision-making Method for Optimization | |
| CN103258143A (en) | Evaluation method for implementation effect of city scape planning | |
| CN113360989A (en) | City green roof planning method | |
| Jianwen et al. | Land-use classification using ASTER data and self-organized neutral networks | |
| CN104361255A (en) | Simulation method for urban expansion through modified cellular automaton | |
| Lin et al. | An expert-knowledge-based model for evaluating cultural tourism strategies: A case of Tainan City, Taiwan | |
| Li | Suitability evaluation method of urban and rural spatial planning based on artificial intelligence | |
| Feng et al. | Prediction of Chinese automobile growing trend considering vehicle adaptability based on Cui–Lawson model | |
| Gou et al. | Analysis on the influence factors of college students about the willingness to work in rural e-commerce | |
| Jun et al. | Decision-making system and verification of pavement diseases treatment scheme for highway reconstruction and extension | |
| CN109741597A (en) | A kind of bus section runing time prediction technique based on improvement depth forest | |
| CN114648153A (en) | Dense oil dessert optimization method and system based on gray matrix weighting method | |
| Amado | Expanding the use of pavement management data |
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 |