Summary of the invention
One of object of the present invention, be in order to overcome the type selecting of shield machine in prior art, be subject to the impact of expert's subjective factor or individual preference, do not consider each engineering geologic factors relevance, cause the indefinite unified defect of type selecting result, a kind of shield machine Intelligent Model Selection method based on engineering geologic factors relevance is provided, this Intelligent Model Selection method has been considered the relevance between each engineering geologic factors, make type selecting result clear and definite, objective, and the method is easy and simple to handle, quick.
Two of object of the present invention is for a kind of device of the shield machine Intelligent Model Selection based on engineering geologic factors is provided.This device simple in structure, easy to operate, has considered the relevance between each geologic agent, makes type selecting result clear and definite, objective.
One of object of the present invention can be achieved through the following technical solutions:
Shield machine Intelligent Model Selection method based on engineering geologic factors relevance, is characterized in that comprising the steps:
1) structure data for projection first according to the criteria for classification of shield structure type selecting classification indicators, produces at random the sample set of Shield Machine Selection in each rate range of each shield machine type
with corresponding experience level y
i(i=1,2 ..., n; J=1,2 ..., p),
and y
ibe respectively each shield machine type and corresponding experience level, n, p is respectively number of samples and index number;
2) form projecting direction and calculate projection value, establishing a={a
1, a
2..., a
pbe projecting direction, p dimension data x
ijcomprehensively become with a
jone Dimensional Projection value z for projecting direction
i, projection value is calculated by following expression:
3) set up projection target function, extract projection index x
ijvariation information, by projection value z
ipartial projection point be condensed into several some groups, to determine z
ithe D of local density
zand z
istandard deviation S
z, the expression formula of setting up projection target function is as follows:
Q(a)=S
zD
z (2)
4) optimize projection target function, according to projection target function Q (a) along with the variation of projecting direction a changes, optimal projection direction is exactly the projecting direction that maximum possible exposes high dimensional data category feature, by solving projection target function maximization problems, estimates best projection direction a
*,
maxQ(a)=S
zD
z (3)
Formation is with { a
j| j=1,2 ..., the complex nonlinear that p} is optimized variable is optimized relation;
5) set up shield structure type selecting disaggregated model, the best projection direction a that step 4) is tried to achieve
*bring the projection value that (2) formula is calculated each sample into
, according to
with y
iscatter diagram set up Shield Machine Selection disaggregated model, Shield Machine Selection disaggregated model with logistic curve function representation is:
In formula,
the calculated value of the-the i Shield Machine Selection sample; The maximum classification grade of N-shield machine is the higher limit of this curve;
6) to be incorporated into Shield Machine Selection disaggregated model obtained above be formula (5) to the engineering geology parameters of input engineering, obtains the shield machine type of applicable employing of this project.
One of object of the present invention can also be achieved through the following technical solutions:
Further, abovementioned steps 1) during structure data for projection, set up shield structure type selecting
and y
ibetween nonlinear relationship, in order to eliminate the impact of dimension, use expression formula
will
be converted into x
ij.
Further, abovementioned steps 3) set up expression formula Q (a)=S of projection target function
zd
zin, S
zfor projection value z
istandard deviation, represent the overall dispersion of data for projection; D
zfor projection value z
ilocal density, expression formula is as follows:
In formula, E
z-projection value z
imean value; r
ij=| z
i-z
j|, be the distance between sample; U (R-r
ij)-unit-step function, works as R>=r
ijtime, its value is 1, works as R<r
ijtime, its value is 0; R is the windows radius of local density, can determine according to test, is generally taken as
While further, abovementioned steps 4) optimizing projection target function, form with { a
j| j=1,2 ..., the complex nonlinear that p} is optimized variable is optimized relation and is solved as follows:
(1) initialization population, arranges number of particles M, maximum iteration time T, gets R=p, according to the criteria for classification of Shield Machine Selection classification indicators, produces at random sample set
and y
i, will
be converted into x
ij, and determine at random initial position and the speed of each particle;
(2) calculate projection value and the fitness of each particle, according to fitness, determine the individual optimal value p of each particle
i,jglobal optimum p with colony
g,j;
(3) enter major cycle, upgrade speed and the position ,Bing unitization projecting direction of a particle according to formula (8), (9), expression formula is as follows
v
i,j(t+1)=wv
i,j(t)+c
1r
1[p
i,j-x
i,j(t)]+c
2r
2[p
g,j-x
i,j(t)] (8)
x
i,j(t+1)=x
i,j(t)+v
i,j(t+1),j=1,2,…,d (9)
In formula: v
i,j(t)-particle rapidity during the t time iteration; x
i,j(t)-particle position during the t time iteration; W-Inertia weight factor; c
1, c
2-positive study the factor; r
1, r
2equally distributed random number between-0 to 1, wherein:
In formula: the random number of N (0,1)-standardized normal distribution; Random number between rand (0,1)-0 to 1; μ
max-random weight galassing mean-max; μ
min-random weight galassing mean-min; The variance of sigma-random weight,
(4) calculate projection value and the fitness of the t time iteration, upgrade the individual optimal solution p of each particle
i,j(t), upgrade globally optimal solution p
g,j(t),
(5) judge whether to meet error condition, if do not met error requirements and not reaching maximum iteration time, proceed to (3) step, otherwise the optimal location of end loop output population, i.e. optimum solution.
While further, abovementioned steps 5) setting up shield structure type selecting disaggregated model, the c in logistic curve function expression (1) and c (2) are undetermined parameter, solve, by following minimization problem
To solving of expression formula (11), adopt random weight heavy particle group algorithm RandWPSO to solve above-mentioned minimization problem, method is by solving best projection direction a
*step.
Further, described shield machine comprises grid type shield machine, earth pressure balance shield machine, muddy water balancing earth-pressure shielding machine and slurry type shield machine.
The inventive method, first according to the criteria for classification of Shield Machine Selection classification indicators, produces at random the sample set of Shield Machine Selection and corresponding experience level, thereby has constructed data for projection; Then by RandWPSO, optimize projecting direction, the projecting direction obtaining according to optimization and PP method calculate optimum projection value; The optimum projection value of last substitution obtains Shield Machine Selection disaggregated model, and the undetermined parameter in its Shield Machine Selection disaggregated model is optimized and obtained by RandWPSO.The engineering geology parameters of input engineering is incorporated into Shield Machine Selection disaggregated model obtained above, can obtain the shield machine type of applicable employing of this project.
Two of object of the present invention can be achieved through the following technical solutions:
The device of the shield machine Intelligent Model Selection based on engineering geologic factors, its design feature is: comprise central controller, man-machine interface and power module, described central controller comprises data acquisition module, data processing module and data memory module, the output terminal of described man-machine interface is connected with the input end of data acquisition module, man-machine interface has input message end and control signal input end, and data acquisition module is by man-machine interface gathering project geologic parameter; The I/O port connection data acquisition module of described data processing module, the information of sending into according to data acquisition module produces the sample set of Shield Machine Selection
with corresponding experience level data for projection y
i, obtain and set projecting direction a
*, and setting projecting direction a
*basis on construct shield structure type selecting disaggregated model, in conjunction with the information of data collecting module collected, form the parameter value of Shield Machine Selection sample, according to described parameter value, obtain engineering and be applicable to the shield machine type adopting; Parameter when described data memory module is used for storing shield structure type selection calculation, and export corresponding parameter according to the request of data processing module; Power module is central controller and man-machine interface power supply.
Two of object of the present invention can also be achieved through the following technical solutions:
Further, the data processing module of central controller comprises controlled processing unit, receiving element, operation processing unit, transmitting element, storage element and I/O port.
Further, described data processing module, first according to the criteria for classification of Shield Machine Selection classification indicators, produces the sample set of Shield Machine Selection at random
with corresponding experience level y
i, set up projection target function, adopt random weight heavy particle group algorithm to solve and optimize projection target function maximization problems, obtain best projection direction a
*thereby, obtain optimum projection value
; Then pass through
with y
iscatter diagram set up Shield Machine Selection disaggregated model; Finally, by the engineering geology parameters combinatorial input Shield Machine Selection disaggregated model of data acquisition module, obtain this project shield machine type of applicable employing.
Further, described shield machine comprises grid type shield machine, earth pressure balance shield machine, muddy water balancing earth-pressure shielding machine and slurry type shield machine.
The present invention has following outstanding beneficial effect:
The present invention, first according to the criteria for classification of Shield Machine Selection classification indicators, produces at random the sample set of Shield Machine Selection and corresponding experience level, thereby has constructed data for projection; Then by RandWPSO, optimize projecting direction, the projecting direction obtaining according to optimization and PP method calculate optimum projection value; The optimum projection value of last substitution obtains Shield Machine Selection disaggregated model, undetermined parameter in its Shield Machine Selection disaggregated model is optimized and is obtained by RandWPSO, the engineering geology parameters of finally inputting engineering is incorporated into Shield Machine Selection disaggregated model obtained above, obtains this project shield machine type of applicable employing.The shortcomings such as personal experience that overcome and do not considered each engineering geologic factors relevance in current Shield Machine selection method, rely on merely expert; Innovate the technology of Shield Machine Selection, there is technical characterstic rapidly and efficiently; By each shield structure type selecting result has been carried out to quantification treatment, avoided the subjectivity of expertise type selecting, make type selecting result clearly unified, specifically strong, the open strong and abundant beneficial effect of Internet resources of easy to operate, applied widely, function.
Embodiment
Below in conjunction with specific embodiment, the present invention is further illustrated.
Specific embodiment 1:
With reference to Fig. 1, the device of the shield machine Intelligent Model Selection based on engineering geologic factors that the present embodiment relates to, comprise
central controller 1, man-machine interface 2 and power module 3, described
central controller 1 comprises data acquisition module 1-1, data processing module 1-2 and data memory module 1-3, the output terminal of described man-machine interface 2 is connected with the input end of data acquisition module 1-1, man-machine interface 2 has input message end and control signal input end, and data acquisition module 1-1 is by man-machine interface 2 gathering project geologic parameters; The I/O port connection data acquisition module 1-1 of described data processing module 1-2, the information of sending into according to data acquisition module 1-1 produces the sample set of Shield Machine Selection
with corresponding experience level data for projection y
i, obtain and set projecting direction a
*, and setting projecting direction a
*basis on construct shield structure type selecting disaggregated model, the information gathering in conjunction with data acquisition module 1-1 forms the parameter value of Shield Machine Selection sample, obtains engineering be applicable to the shield machine type adopting according to described parameter value; Parameter when described data memory module 1-3 is used for storing shield structure type selection calculation, and export corresponding parameter according to the request of data processing module 1-2; Power module 3 is
central controller 1 and man-machine interface power supply source.
In the present embodiment:
The data processing module 1-2 of central controller 1 comprises controlled processing unit, receiving element, arithmetic element, transmitting element, storage element and I/O port, described controlled processing unit, receiving element, arithmetic element, transmitting element, storage element and I/O port can adopt conventional singlechip chip, signal receiving chip, Arithmetic Processing Chip, signal sends chip, storage chip and I/O chip, singlechip chip and signal receiving chip, Arithmetic Processing Chip, signal sends chip, the connected mode of storage chip and I/O chip can adopt the connected mode of routine techniques, utilize known singlechip chip, signal receiving chip, Arithmetic Processing Chip, signal sends chip, storage chip and I/O chip structure, function and for the method for technical solution problem, structure has gathering project geologic parameter, the information of sending into according to data acquisition module 1-1 produces the sample set of Shield Machine Selection
with corresponding experience level data for projection y
i, obtain and set projecting direction a
*, and setting projecting direction a
*basis on construct shield structure type selecting disaggregated model, the information gathering in conjunction with data acquisition module 1-1 forms the parameter value of Shield Machine Selection sample, according to described parameter value, obtains the shield machine Intelligent Model Selection device based on engineering geologic factors that engineering is applicable to the shield machine type of functionality of employing.
Described data processing module 1-2, first according to the criteria for classification of Shield Machine Selection classification indicators, produces the sample set of Shield Machine Selection at random
with corresponding experience level y
i, set up projection target function, adopt random weight heavy particle group algorithm to solve and optimize projection target function maximization problems, obtain best projection direction a
*thereby, obtain optimum projection value
; Then pass through
with y
iscatter diagram set up Shield Machine Selection disaggregated model; Finally, by the engineering geology parameters combinatorial input Shield Machine Selection disaggregated model of data acquisition module 1-1, obtain this project shield machine type of applicable employing.
Parameter when described data memory module 1-3 is used for storing shield structure type selection calculation, and export corresponding parameter according to the request of data processing module 1-2.
Described man-machine interface 2 is for inputting engineering geology parameters to data acquisition module 1-1, and data processing module 1-2 obtains engineering geology parameters from data acquisition module 1-1, and the information of sending into according to data acquisition module 1-1 produces the sample set of Shield Machine Selection
with corresponding experience level data for projection y
i, obtain and set projecting direction a
*, and setting projecting direction a
*basis on construct shield structure type selecting disaggregated model, the information gathering in conjunction with data acquisition module 1-1 forms the parameter value of Shield Machine Selection sample, according to described parameter value, obtaining the shield machine type that engineering be applicable to adopt, is the engineering shield machine type of applicable employing.
Power module 3 is central controller 1 and man-machine interface 2 power supplies.
The Shield Machine Selection that the present embodiment is applicable to, comprises grid type shield machine, earth pressure balance shield machine, muddy water balancing earth-pressure shielding machine and slurry type shield machine.
With reference to Fig. 2-Fig. 4, the shield machine Intelligent Model Selection method based on engineering geologic factors relevance that the present embodiment relates to, comprises the steps:
1) structure data for projection
First according to shield structure type selecting, divide the criteria for classification of * class index, the random sample set that produces the raw Shield Machine Selection of * in each rate range of each shield machine type
with corresponding experience level y
i(i=1,2 ..., n; J=1,2 ..., p),
and y
ibe respectively each shield machine type and corresponding experience level, n, p is respectively number of samples and index number;
According to the criteria for classification of Shield Machine Selection classification indicators, produce at random the sample set of Shield Machine Selection
with corresponding experience level y
i, structure data for projection, calculates projection value;
Structure data for projection, first the criteria for classification according to shield structure type selecting classification indicators is shown in Table 1, the random sample set that produces Shield Machine Selection in each rate range of each shield machine type
with corresponding experience level y
i, i=1,2 ..., n; J=1,2 ..., p,
and y
ibe respectively classification indicators and corresponding experience level, n, p is respectively number of samples and index number, sets up the RandWPSO-PP model of shield structure type selecting and sets up exactly
and y
ibetween nonlinear relationship, in order to eliminate the impact of dimension, will
be converted into x
ij, as follows:
Table 1 is the criteria for classification of Shield Machine Selection classification indicators
2) calculate projection value
If a={a
1, a
2..., a
pbe projecting direction, projection Pursuit Method is exactly p dimension data x
ijcomprehensively become with a
jone Dimensional Projection value z for projecting direction
i
Secondly, optimize projecting direction, set up projection target function, adopt random weight heavy particle group algorithm to solve and optimize projection target function maximization problems, obtain best projection direction a
*;
3) set up projection target function
When comprehensive projection index, need maximum extraction projection index x
ijvariation information, just require projection value z
ipartial projection point intensive as far as possible, be preferably condensed into several some groups, i.e. z
ithe D of local density
zlarge as far as possible, between subpoint group, scatter as far as possible on the whole, i.e. z simultaneously
istandard deviation S
zalso large as far as possible, adopt Friedman-Tukey type projection index (being to adopt projection Pursuit Method), its function expression is as follows:
Q(a)=S
zD
z (3)
In formula: S
zfor projection value z
istandard deviation represent the dispersion that data for projection is overall; D
zfor projection value z
ilocal density, S
zexpression formula as follows:
In formula, E
z-projection value z
imean value; r
ij=| z
i-z
j|, be the distance between sample;
D
zexpression formula as follows:
U (R-r in formula
ij) be unit-step function, work as R>=r
ijtime, its value is 1, works as R<r
ijtime, its value is 0; R is the windows radius of local density, according to test, determines, is generally taken as
4) optimize projection target function
Projection target function Q (a) changes along with the variation of projecting direction a, and optimal projection direction is exactly the projecting direction that maximum possible exposes high dimensional data category feature, by solving projection target function, maximizes to estimate best projection direction a
*, that is:
maxQ(a)=S
zD
z (6)
Adopt random weight heavy particle group algorithm to solve { a
j| j=1,2 ..., the complex nonlinear optimization problem that p} is optimized variable;
Described random weight heavy particle group algorithm, comprises the steps:
(1) initialization population, arranges number of particles M, maximum iteration time T, gets R=p, according to the criteria for classification of Shield Machine Selection classification indicators, produces at random sample set
and y
i, will
be converted into x
ij, and determine at random initial position and the speed of each particle;
(2) calculate projection value and the fitness of each particle, according to fitness, determine the individual optimal value p of each particle
i,jglobal optimum p with colony
g,j;
(3) enter major cycle, according to formula (8), (9), upgrade speed and the position ,Bing unitization projecting direction of a particle.
v
i,j(t+1)=wv
i,j(t)+c
1r
1[p
i,j-x
i,j(t)]+c
2r
2[p
g,j-x
i,j(t)] (8)
x
i,j(t+1)=x
i,j(t)+v
i,j(t+1),j=1,2,…,d (9)
In formula: v
i,j(t) particle rapidity while being the t time iteration; x
i,j(t) particle position while being the t time iteration; W is Inertia weight factor; c
1, c
2for the positive study factor; r
1, r
2be equally distributed random number between 0 to 1, wherein:
In formula: the random number that N (0,1) is standardized normal distribution; Rand (0,1) is the random number between 0 to 1; μ
maxfor random weight galassing mean-max; μ
minfor random weight galassing mean-min; Sigma is the variance of random weight;
(4) calculate projection value and the fitness of the t time iteration, upgrade the individual optimal solution p of each particle
i,j(t), upgrade globally optimal solution p
g,j(t);
(5) judge whether to meet error condition, if do not met error requirements and not reaching maximum iteration time, proceed to (3) step, otherwise the optimal location of end loop output population, i.e. optimum solution.
Then, best projection direction is brought into, adopt projection Pursuit Method to calculate optimum projection value
, by
with y
iscatter diagram can set up Shield Machine Selection disaggregated model:
5) set up shield structure type selecting disaggregated model
The best projection direction a that step 4) is tried to achieve
*bring the optimum projection value of calculating each sample in formula (2) into
, according to
with y
iscatter diagram set up Shield Machine Selection disaggregated model, adopt logistic curve function (Logistic CurveFunction) simulation,
Undetermined parameter c (1) in formula and c (2), solve by following minimization problem,
Adopt random weight heavy particle group algorithm to solve above-mentioned minimization problem, the method is with solving best projection direction a in step 4
*method.
Finally, the engineering geology parameters of step 6, input engineering is incorporated into Shield Machine Selection disaggregated model obtained above, i.e. (11) obtain the shield machine type of applicable employing of this project.
The present embodiment is applicable to the shield machine of type selecting, comprises grid type shield machine, earth pressure balance shield machine, muddy water balancing earth-pressure shielding machine and slurry type shield machine.
Described is existing algorithms most in use by random weight heavy particle group algorithm (RandWPSO) and projection Pursuit Method (PP), and logistic curve function (Logistic Curve Function) and Friedman-Tukey type projection index are prior aries.
The method adopts matlab language compilation calculation procedure to calculate, and system and device and the operation platform of matlab are as follows:
Hardware environment:
CPU:Pentium III and above, AMD Athlon, Athlon XP, Athlon MP;
Disk space: 400MB and more than;
Internal memory: 256MB and more than;
Video card: 16 and more than.
Software environment: the different platforms such as Windows2000/NT4.0/XP/XP SP/7, UNIX, Mac OS X.