CN103927458A - Determination method of sensibility of influence factors of anchoring force of soil anchors - Google Patents
Determination method of sensibility of influence factors of anchoring force of soil anchors Download PDFInfo
- Publication number
- CN103927458A CN103927458A CN201410181912.3A CN201410181912A CN103927458A CN 103927458 A CN103927458 A CN 103927458A CN 201410181912 A CN201410181912 A CN 201410181912A CN 103927458 A CN103927458 A CN 103927458A
- Authority
- CN
- China
- Prior art keywords
- anchor force
- influence factor
- soil bolt
- soil
- forecast model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Investigation Of Foundation Soil And Reinforcement Of Foundation Soil By Compacting Or Drainage (AREA)
Abstract
The invention discloses a determination method of sensibility of influence factors of anchoring force of soil anchors. The method comprises the steps that firstly, a soil-anchor anchoring-force prediction model is selected, and input quantities and output quantities of the prediction model are determined, wherein the influence factors of the anchoring force of the M soil anchors and the anchoring force of the soil anchors are used as the M input quantities and the M output quantities of the soil-anchor anchoring-force prediction model respectively; secondly, a training sample set is obtained, wherein in-situ test data of the N soil anchors are collected, and the in-situ test data of each soil anchor are used as one training sample; thirdly, the soil-anchor anchoring-force prediction model is built and trained; fourthly, a multilevel and multi-factor orthogonal test is carried out in the steps of inputting design parameters of the soil anchors, building an orthogonal table, obtaining samples of the orthogonal test, predicting the anchoring force, organizing predicted values of the anchoring force and determining the sensibility. The method has the simple steps, is reasonable in deign, convenient to implement and good in using effect, and can be used for easily, conveniently, rapidly and accurately completing a determining process of the sensibility of the influence factors of the anchoring force of the soil anchors.
Description
Technical field
The invention belongs to soil bolt anchor force analysis of Influential Factors technical field, especially relate to a kind of assay method of soil bolt anchor force influence factor susceptibility.
Background technology
The anchor force of soil bolt depends on length, bolt diameter, aperture, grouting body intensity, the inclination angle of anchor pole, the many factors such as character of the soil body of anchoring section, and these influence factors inevitably exist uncertainty.In actual use procedure, the amplitude of oscillation that above-mentioned uncertain factor may occur has much on the impact of soil bolt anchor force, and anchor-holding force of anchor bolt is on one of major issue of the most responsive care of engineering often of the impact of which kind of factor.
1986, Hohenbichler carried out the Study of Sensitivity of fiduciary level to stochastic variable, Madsen after this, and the people such as Bjerager and Karamchandani have done further research to this.Domestic at present less to the research of the sensitivity to parameter aspect of anchoring engineering, report that visible documents and materials are also mainly confined to the sensitivity analysis of various parameters to anchoring engineering stability and distortion, Study of Sensitivity for soil bolt anchor force is less, almost the fairly perfect research data in this aspect not.Existing Study of Sensitivity method is also confined to finite element method, the obvious shortcoming of this method be exactly when parameter is more calculated amount very large, the combination of a plurality of parameters be difficult to be determined on the other hand.
Summary of the invention
Technical matters to be solved by this invention is for above-mentioned deficiency of the prior art, a kind of assay method of soil bolt anchor force influence factor susceptibility is provided, its method step simple, reasonable in design and realize convenient, result of use is good, can be easy, the sensitivity testing process that fast and accurately completes soil bolt anchor force influence factor.
For solving the problems of the technologies described above, the technical solution used in the present invention is: a kind of assay method of soil bolt anchor force influence factor susceptibility, is characterized in that the method comprises the following steps:
Step 1, the forecast model selection of soil bolt anchor force and input quantity thereof and output quantity are determined: select BP neural network model as soil bolt anchor force forecast model, affect the different affecting factors of soil bolt anchor force as the input quantity of described soil bolt anchor force forecast model using M, and the output quantity using soil bolt anchor force as described soil bolt anchor force forecast model;
The quantity of described input quantity is M, and M described input quantity is denoted as respectively x
1, x
2..., x
m; Described output quantity is denoted as y; M is positive integer and M>=3;
Step 2, training sample set obtain: collect the in-situ test data of N soil bolt, and the in-situ test data of each soil bolt are all stored in the data storage cell joining with data processor; Described in each, the in-situ test data of soil bolt include the numerical value of M influence factor and the measured value of anchor force of this soil bolt, and the in-situ test data of N described soil bolt form training sample set;
Described training sample is concentrated and is comprised N training sample, and each training sample is denoted as { x
1i, x
2i..., x
mi, y
i, x wherein
1i, x
2i..., x
mibe respectively the numerical value of M influence factor of i described soil bolt in N soil bolt, y
iit is the anchor force measured value of i described soil bolt; Wherein, i be positive integer and i=1,2 ..., N;
Step 3, soil bolt anchor force forecast model are set up and training: described data processor calls Matlab software and sets up described soil bolt anchor force forecast model, and adopt the soil bolt anchor force forecast model that the training sample set pair set up in step 2 is set up to train; After having trained, obtain the soil bolt anchor force forecast model training;
Before described soil bolt anchor force forecast model is trained, first frequency of training and training objective error are set;
Step 4, multilevel orthogonal test of multiple factors: by the design parameter of current designed soil bolt is carried out to multilevel orthogonal test of multiple factors, determine the susceptibility of M described influence factor of the current designed soil bolt anchor force of impact, process is as follows:
Step 401, soil bolt design parameter input: by the parameter input unit of joining with described data processor, input the design parameter of current designed soil bolt, this design parameter comprises the design load of M described influence factor of current designed soil bolt;
Step 402, orthogonal arrage are set up: in the orthogonal arrage of setting up, comprise M described influence factor in step 1, and M described influence factor all carried out to S hydraulic test; Wherein, S is positive integer and S >=3;
Described in each, the S of an influence factor test level is the design load of this influence factor is carried out after the adjustment of S different amplitudes, the trial value of acquisition;
Step 403, orthogonal test sample acquisition: according to the design parameter of inputting in step 401, and the orthogonal arrage of setting up in integrating step 402, obtain all test samples used when current designed soil bolt is carried out to multilevel orthogonal test of multiple factors; Described in each, in test sample, include the trial value of M described influence factor of current designed soil bolt;
Step 404, anchor force prediction: the soil bolt anchor force forecast model training in described data processor invocation step three, the corresponding anchor force of all test samples in step 403 is predicted, and obtained the anchor force predicted value of S different tests level of M described influence factor;
Step 405, anchor force predicted value arrange: according to the anchor force predicted value of S different tests level of the M obtaining in step 404 described influence factor, calculate the anchor force predicted value mean value of S different tests level of influence factor described in each; Wherein, in M described influence factor, the anchor force predicted value mean value of t test level of j influence factor is denoted as M
jt, j and t are positive integer, j=1,2 ..., M, t=1,2 ..., S;
Afterwards, according to the anchor force predicted value mean value of S different tests level of the M calculating a described influence factor, calculate the extreme difference of influence factor described in each; Wherein, in M described influence factor, the extreme difference of j influence factor is denoted as Δ M
jt, Δ M
jt=M
jmax-M
jmin, M in formula
jmaxbe the maximal value in the anchor force predicted value mean value of S different tests level of j influence factor, M
jminit is the minimum value in the anchor force predicted value mean value of S different tests level of j influence factor;
Step 5, susceptibility are determined: according to the extreme difference of the M calculating in step 405 described influence factor, the susceptibility of M described influence factor is determined; Wherein, the influence factor that extreme difference is larger is more responsive, larger to the anchor force influence degree of designed soil bolt, and the extreme difference size of each influence factor determines the anchor force influence degree of this influence factor to designed soil bolt.
The assay method of above-mentioned a kind of soil bolt anchor force influence factor susceptibility, is characterized in that: M=5 in step 1.
The assay method of above-mentioned a kind of soil bolt anchor force influence factor susceptibility, it is characterized in that: after the anchor force of soil bolt described in step 3 forecast model has been trained, also need to adopt the soil bolt anchor force forecast model that test sample book set pair trains to test; Described test sample book collection is the sample set being comprised of the concentrated a plurality of training samples of training sample described in step 2 or the sample set re-establishing according to the acquisition methods of training sample set described in step 2.
The assay method of above-mentioned a kind of soil bolt anchor force influence factor susceptibility, is characterized in that: 5 described influence factors are respectively anchoring section length, bolt diameter, aperture, grouting body intensity and anchor pole angle of inclination.
The assay method of above-mentioned a kind of soil bolt anchor force influence factor susceptibility, it is characterized in that: when described soil bolt anchor force forecast model is tested, the sim function that described data processor calls in Matlab Neural Network Toolbox is tested, and generate the anchor force predicted value that each training sample is corresponding, afterwards each is detected to anchor force predicted value that sample is corresponding and the anchor force measured value of this detection sample and carry out difference comparison, draw the predicated error that this detection sample is corresponding; After each training sample has all been tested in described training sample set, obtain predicated error corresponding to each training sample in described training sample set; Afterwards, the predicated error corresponding according to each training sample, analyzes the precision of prediction of the soil bolt anchor force forecast model training.
The assay method of above-mentioned a kind of soil bolt anchor force influence factor susceptibility, it is characterized in that: when described soil bolt anchor force forecast model is trained in step 3, " trainlm " function that described data processor calls in Matlab Neural Network Toolbox is trained.
The assay method of above-mentioned a kind of soil bolt anchor force influence factor susceptibility, it is characterized in that: before in step 3, described soil bolt anchor force forecast model being trained, described data processor first calls normalized module and concentrates the data in each training sample to be normalized to described training sample; After having trained, described data processor also needs to call renormalization processing module to the predicted value generating being carried out to renormalization processing in training process;
Before carrying out anchor force prediction in step 404, described data processor first calls described normalized module the data in all test samples in step 403 is normalized, after calling the corresponding anchor force of all test samples of described soil bolt anchor force forecast model and predicting, also need to call described renormalization processing module generated predicted value is carried out to renormalization processing.
The assay method of above-mentioned a kind of soil bolt anchor force influence factor susceptibility, it is characterized in that: while setting up described soil bolt anchor force forecast model in step 3, the soil bolt anchor force forecast model of setting up comprises an input layer, a hidden layer and an output layer, described input layer comprises M corresponding with the individual described influence factor of M respectively node, described hidden layer comprises (2M+1) individual node, and described output layer comprises a node.
The assay method of above-mentioned a kind of soil bolt anchor force influence factor susceptibility, is characterized in that: S=5 in step 402.
The assay method of above-mentioned a kind of soil bolt anchor force influence factor susceptibility, it is characterized in that: in step 402 described in each 5 test level of influence factor be to the design load of this influence factor raise 20%, raise 10%, raise 0%, lower 10% and lower 20% after, the trial value of acquisition.
The present invention compared with prior art has the following advantages:
1, method step simple, reasonable in design and realize convenient.
2, reasonable in design, the present invention is based on soil bolt anchor force in-situ test, consider the anchor force influence factors such as bolt diameter, anchoring section length, inclination angle and grouting body intensity, aperture, utilize error back propagation (the back propagation of artificial neural network (artificial neural network), BP) algorithm and MATLAB artificial neural network tool box, set up anchor force forecast model; On this basis, adopt orthogonal arrage test theory to analyze the susceptibility of anchor force to each influence factor, can provide reference frame for the practical application of corresponding anchoring reinforcing engineering.
3, result of use is good and practical value is high, and the BP algorithm based on artificial neural network considers the influence factors such as bolt diameter, grouting body intensity, anchoring section length, aperture, anchor pole inclination angle, has set up the model of mind of soil bolt anchor force prediction.In conjunction with the in-situ test data of bolt anchorage engineering, to network train, error analysis, thereby verified applicability and the feasibility of the forecast model of setting up.The forecast model that utilization trains, and in conjunction with orthogonal arrage test theory, designed many groups orthogonal test, the Sensitivity Analysis to soil bolt anchor force influence factor.The neural network method for designing of employing based on orthogonal test, the method can utilize principle of orthogonal test to improve network structure from arbitrary network, improves network performance, until final design goes out more excellent or optimum neural network.And utilize the extremum difference analysis of orthogonal test to obtain optimizing decision, disclose the interactive power of each interlayer of neural network.
In sum, the inventive method step simple, reasonable in design and realize convenient, result of use is good, can be easy, the sensitivity testing process that fast and accurately completes soil bolt anchor force influence factor.
Below by drawings and Examples, technical scheme of the present invention is described in further detail.
Accompanying drawing explanation
Fig. 1 is method flow block diagram of the present invention.
Fig. 2 is the structural representation that the present invention sets up soil bolt anchor force forecast model.
Fig. 3 is the present invention's change curve that training error changes with frequency of training while carrying out model training.
Embodiment
The assay method of a kind of soil bolt anchor force influence factor susceptibility as shown in Figure 1, comprises the following steps:
Step 1, the forecast model selection of soil bolt anchor force and input quantity thereof and output quantity are determined: select BP neural network model as soil bolt anchor force forecast model, affect the different affecting factors of soil bolt anchor force as the input quantity of described soil bolt anchor force forecast model using M, and the output quantity using soil bolt anchor force as described soil bolt anchor force forecast model.
The quantity of described input quantity is M, and M described input quantity is denoted as respectively x
1, x
2..., x
m; Described output quantity is denoted as y; M is positive integer and M>=3.
During actual use, a selected M described influence factor is the larger important factor in order of soil bolt anchor force impact.
In the present embodiment, M=5 in step 1.5 described influence factors are respectively anchoring section length, bolt diameter, aperture, grouting body intensity and anchor pole angle of inclination.Wherein, the unit in anchoring section length, bolt diameter and aperture is mm, and the unit of grouting body intensity is MPa, and the unit at anchor pole angle of inclination is °, the anchor hole aperture that aperture is soil bolt.
During actual use, can according to specific needs, the value size of M and M described influence factor be adjusted respectively.That is to say, can according to specific needs selected influence factor be adjusted accordingly, can select other influence factor.
Step 2, training sample set obtain: collect the in-situ test data of N soil bolt, and the in-situ test data of each soil bolt are all stored in the data storage cell joining with data processor; Described in each, the in-situ test data of soil bolt include the numerical value of M influence factor and the measured value of anchor force of this soil bolt, and the in-situ test data of N described soil bolt form training sample set.
Described training sample is concentrated and is comprised N training sample, and each training sample is denoted as { x
1i, x
2i..., x
mi, y
i, x wherein
1i, x
2i..., x
mibe respectively the numerical value of M influence factor of i described soil bolt in N soil bolt, y
iit is the anchor force measured value of i described soil bolt; Wherein, i be positive integer and i=1,2 ..., N.
The numerical value of M influence factor in N described training sample forms input matrix, and described input matrix is N * Metzler matrix.
In the present embodiment, N=15.During actual use, can according to specific needs, the value size of N be adjusted accordingly.
In the present embodiment, the training sample set of setting up refers to table 1:
The tables of data of each training sample in table 1 training sample set
Step 3, soil bolt anchor force forecast model are set up and training: described data processor calls Matlab software and sets up described soil bolt anchor force forecast model, and adopt the soil bolt anchor force forecast model that the training sample set pair set up in step 2 is set up to train; After having trained, obtain the soil bolt anchor force forecast model training.
Before described soil bolt anchor force forecast model is trained, first frequency of training and training objective error are set.
In the present embodiment, while setting up described soil bolt anchor force forecast model, the soil bolt anchor force forecast model of setting up comprises an input layer, a hidden layer and an output layer, described input layer comprises M corresponding with the individual described influence factor of M respectively node, described hidden layer comprises (2M+1) individual node, and described output layer comprises a node.
When reality is determined the number of nodes in described hidden layer, according to Kolmogorov theorem, having under the condition of Rational structure and appropriate weights, three layers of feedforward network can approach continuous function arbitrarily.Three layers of BP neural network model that the present invention sets up, can complete n arbitrarily and tie up the mapping that m ties up, and refer to Fig. 2.
In the present embodiment, in described hidden layer, each node adopts the asymmetry S type function of one of sigmoid function: f (x)=1/1+e
-x, as neuron excitation function.During actual use, each node in described hidden layer all receives all nodes in described input layer.
When reality is set up described soil bolt anchor force forecast model, " Newff " function that described data processor calls in Matlab Neural Network Toolbox is set up.
In the present embodiment, when described soil bolt anchor force forecast model is trained in step 3, " trainlm " function that described data processor calls in Matlab Neural Network Toolbox is trained.Wherein, " trainlm " function is the training function based on L-M algorithm.
And, before described soil bolt anchor force forecast model is trained, need first set up soil bolt anchor force forecast model to be carried out to initialization.Actual while carrying out initialization, initwb function is according to initiation parameter initializes weights matrix and the biasing of every one deck oneself.
In the present embodiment, before described soil bolt anchor force forecast model being trained in step 3, described data processor first calls normalized module and concentrates the data in each training sample to be normalized to described training sample, specifically each element in described input matrix is normalized; After having trained, described data processor also needs to call renormalization processing module to the predicted value generating being carried out to renormalization processing in training process.
Be normalized in process, while concentrating arbitrary numerical value that affects element to be normalized on described training sample, all according to formula x '=(x-MinValuex)/(MaxValuex-MinValuex) be normalized, in formula, x and x ' are respectively that this affect the numerical value of element before and after normalized, and MaxValuex and MinValuex are respectively concentrated this of described training sample affects greatest measure and the minimum value of element.
After having trained, N anchor force predicted value corresponding to described training sample of described soil bolt anchor force forecast model output, and the N exporting an anchor force predicted value forms output matrix.The actual renormalization that carries out is when process, according to formula y '=(y-MinValuey)/(MaxValuey-MinValuey) be normalized, in formula, y and y ' are respectively the anchor force predicted value that renormalization is processed front and back, and MaxValuey and MinValuey are respectively maximal value and the minimum value that renormalization is processed front anchor force predicted value.
In the present embodiment, before actual training, speed of convergence for accelerating network, improve counting yield, the normalized function premnmx that utilizes Matlab Neural Network Toolbox to provide, input matrix corresponding to described training sample set and output matrix are normalized, make data after processing between [1,1].Utilize the data after normalization to train set up soil bolt anchor force forecast model, training function adopts " trainlm " function of Matlab nerve network system acquiescence.Training shows that interval number of times is set to 50, and maximum cycle (being frequency of training) is set to 1000, and performance objective value (being training objective error) is set to 0.01, and learning coefficient is set to 0.01.As shown in Figure 3, when accumulative total train epochs reaches 14 step, it is 0.01 that training error converges to aimed at precision.After training finishes, to described training sample, concentrate the data of each training sample to predict, generate anchor force predicted value, and predicted value is carried out to renormalization and process and obtain final predicted value.
In the present embodiment, after the anchor force of soil bolt described in step 3 forecast model has been trained, also need to adopt the soil bolt anchor force forecast model that test sample book set pair trains to test; Described test sample book collection is the sample set being comprised of the concentrated a plurality of training samples of training sample described in step 2 or the sample set re-establishing according to the acquisition methods of training sample set described in step 2.
When described soil bolt anchor force forecast model is tested, the sim function that described data processor calls in Matlab Neural Network Toolbox is tested, and generate the anchor force predicted value that each training sample is corresponding, afterwards each is detected to anchor force predicted value that sample is corresponding and the anchor force measured value of this detection sample and carry out difference comparison, draw the predicated error that this detection sample is corresponding; After each training sample has all been tested in described training sample set, obtain predicated error corresponding to each training sample in described training sample set; Afterwards, the predicated error corresponding according to each training sample, analyzes the precision of prediction of the soil bolt anchor force forecast model training.
When reality is analyzed the precision of prediction of the soil bolt anchor force forecast model training, the mean value of the predicated error that all training samples are corresponding is the precision of prediction of this soil bolt anchor force forecast model.
In the present embodiment, after the soil bolt anchor force forecast model training of setting up finishes, this soil bolt anchor force forecast model has reached the performance index of appointment, in each layer network, stored the inherent Nonlinear Mapping relation that training sample provides, and weight and threshold value between clear and definite each input layer and output layer.Herein, 5 training samples that training sample described in selecting step two is concentrated form test sample book collection, and described soil bolt anchor force forecast model is tested.During actual test, first with postmnmx function, the data of each training sample are normalized, recycling sim function generates corresponding anchor force predicted value, with postmnmx function, the predicted value generating is carried out to renormalization processing afterwards, obtains final predicted value; Finally predicted value and in-situ test value (being anchor force measured value) are compared, its comparative result refers to table 2.Test result shows, the average relative error of the soil bolt anchor force forecast model training is 1.95%, meets requirement of engineering, illustrates that this model has certain feasibility and practicality aspect anchor-holding force of anchor bolt prediction.
Table 2 soil bolt anchor force forecast model table with test results
Step 4, multilevel orthogonal test of multiple factors: by the design parameter of current designed soil bolt is carried out to multilevel orthogonal test of multiple factors, determine the susceptibility of M described influence factor of the current designed soil bolt anchor force of impact, process is as follows:
Step 401, soil bolt design parameter input: by the parameter input unit of joining with described data processor, input the design parameter of current designed soil bolt, this design parameter comprises the design load of M described influence factor of current designed soil bolt.
Step 402, orthogonal arrage are set up: in the orthogonal arrage of setting up, comprise M described influence factor in step 1, and M described influence factor all carried out to S hydraulic test; Wherein, S is positive integer and S >=3.
Described in each, the S of an influence factor test level is the design load of this influence factor is carried out after the adjustment of S different amplitudes, the trial value of acquisition.
Orthogonal experiment, utilize exactly orthogonality principle and mathematical statistics, from a large amount of testing sites, select typical, representational testing site, application quadrature form rationally scientifically arranges test, obtains a kind of test design method of optimum result with few number of times of trying one's best.In test, need to investigate, controllable condition is called factor or the factor, each state of factor or grade are called a level (being test level) of factor.
In the present embodiment, in the present embodiment, S=5 in step 402.
In step 402 described in each 5 test level of influence factor be to the design load of this influence factor raise 20%, raise 10%, raise 0%, lower 10% and lower 20% after, the trial value of acquisition, refers to table 3.
5 test level detail lists of table 35 influence factor
In the present embodiment, for 5 levels of 5 factor, as comprehensive test needs 3125 tests, adopt orthogonal arrage L25 (5
6) only needing 25 Marshall Tests, table 4 is because of the selected L25 (5 of prime number according to position sum of series
6) orthogonal arrage.
Table 4L25 (5
6) orthogonal arrage
Step 403, orthogonal test sample acquisition: according to the design parameter of inputting in step 401, and the orthogonal arrage of setting up in integrating step 402, obtain all test samples used when current designed soil bolt is carried out to multilevel orthogonal test of multiple factors; Described in each, in test sample, include the trial value of M described influence factor of current designed soil bolt.
Step 404, anchor force prediction: the soil bolt anchor force forecast model training in described data processor invocation step three, the corresponding anchor force of all test samples in step 403 is predicted, and obtained the anchor force predicted value of S different tests level of M described influence factor.
Step 405, anchor force predicted value arrange: according to the anchor force predicted value of S different tests level of the M obtaining in step 404 described influence factor, calculate the anchor force predicted value mean value of S different tests level of influence factor described in each; Wherein, in M described influence factor, the anchor force predicted value mean value of t test level of j influence factor is denoted as M
jt, j and t are positive integer, j=1,2 ..., M, t=1,2 ..., S.
Afterwards, according to the anchor force predicted value mean value of S different tests level of the M calculating a described influence factor, calculate the extreme difference of influence factor described in each; Wherein, in M described influence factor, the extreme difference of j influence factor is denoted as Δ M
jt, Δ M
jt=M
jmax-M
jmin, M in formula
jmaxbe the maximal value in the anchor force predicted value mean value of S different tests level of j influence factor, M
jminit is the minimum value in the anchor force predicted value mean value of S different tests level of j influence factor.
Step 5, susceptibility are determined: according to the extreme difference of the M calculating in step 405 described influence factor, the susceptibility of M described influence factor is determined; Wherein, the influence factor more responsive (susceptibility that is the influence factor that extreme difference is larger is stronger) that extreme difference is larger, anchor force influence degree to designed soil bolt is larger, and the extreme difference size of each influence factor determines the anchor force influence degree of this influence factor to designed soil bolt.
In the present embodiment, before carrying out anchor force prediction in step 404, described data processor first calls described normalized module the data in all test samples in step 403 is normalized, after calling the corresponding anchor force of all test samples of described soil bolt anchor force forecast model and predicting, also need to call described renormalization processing module generated predicted value is carried out to renormalization processing.
In the present embodiment, in step 401, the design parameter of the current designed soil bolt of inputting is as follows: anchoring section length 9.3m, diameter 30mm, aperture 130mm, grouting body intensity 3.0Mpa, 15 °, anchor pole inclination angle.
According to selected L25 (5
6) orthogonal arrage, in his-and-hers watches 3, each test level changes and is combined to form 25 test samples.Afterwards, with the soil bolt anchor force forecast model training, above-mentioned 25 test samples are predicted respectively, predicted the outcome in Table 5.
The table 5 test sample table that predicts the outcome
Five experimental results of each influence factor par are averaging, just can obtain the mean value of each factor each index in varying level situation, poor to the index maximizing of same influence factor varying level and minimum value, can obtain this influence factor and change corresponding extreme difference, refer to table 6.
The orthogonal experiments table of each influence factor of table 6
As shown in Table 6: the susceptibility of the anchor force influence factor of current designed soil bolt is followed successively by anchoring section length, grouting body intensity, bolt diameter, aperture and anchor pole pitch angle.Each extreme difference size that affects key element is the influence degree to soil bolt anchor force, and its extreme difference is larger, larger to the influence degree of soil bolt anchor force.
To sum up, descending anchoring section length, grouting body intensity, bolt diameter, aperture and the anchor pole pitch angle of being arranged as of susceptibility of current designed soil bolt anchor force influence factor, this sensitivity analysis result can provide reference frame for the practical application of similar reinforcing engineering.
The above; it is only preferred embodiment of the present invention; not the present invention is imposed any restrictions, every any simple modification of above embodiment being done according to the technology of the present invention essence, change and equivalent structure change, and all still belong in the protection domain of technical solution of the present invention.
Claims (10)
1. an assay method for soil bolt anchor force influence factor susceptibility, is characterized in that the method comprises the following steps:
Step 1, the forecast model selection of soil bolt anchor force and input quantity thereof and output quantity are determined: select BP neural network model as soil bolt anchor force forecast model, affect the different affecting factors of soil bolt anchor force as the input quantity of described soil bolt anchor force forecast model using M, and the output quantity using soil bolt anchor force as described soil bolt anchor force forecast model;
The quantity of described input quantity is M, and M described input quantity is denoted as respectively x
1, x
2..., x
m; Described output quantity is denoted as y; M is positive integer and M>=3;
Step 2, training sample set obtain: collect the in-situ test data of N soil bolt, and the in-situ test data of each soil bolt are all stored in the data storage cell joining with data processor; Described in each, the in-situ test data of soil bolt include the numerical value of M influence factor and the measured value of anchor force of this soil bolt, and the in-situ test data of N described soil bolt form training sample set;
Described training sample is concentrated and is comprised N training sample, and each training sample is denoted as { x
1i, x
2i..., x
mi, y
i, x wherein
1i, x
2i..., x
mibe respectively the numerical value of M influence factor of i described soil bolt in N soil bolt, y
iit is the anchor force measured value of i described soil bolt; Wherein, i be positive integer and i=1,2 ..., N;
Step 3, soil bolt anchor force forecast model are set up and training: described data processor calls Matlab software and sets up described soil bolt anchor force forecast model, and adopt the soil bolt anchor force forecast model that the training sample set pair set up in step 2 is set up to train; After having trained, obtain the soil bolt anchor force forecast model training;
Before described soil bolt anchor force forecast model is trained, first frequency of training and training objective error are set;
Step 4, multilevel orthogonal test of multiple factors: by the design parameter of current designed soil bolt is carried out to multilevel orthogonal test of multiple factors, determine the susceptibility of M described influence factor of the current designed soil bolt anchor force of impact, process is as follows:
Step 401, soil bolt design parameter input: by the parameter input unit of joining with described data processor, input the design parameter of current designed soil bolt, this design parameter comprises the design load of M described influence factor of current designed soil bolt;
Step 402, orthogonal arrage are set up: in the orthogonal arrage of setting up, comprise M described influence factor in step 1, and M described influence factor all carried out to S hydraulic test; Wherein, S is positive integer and S >=3;
Described in each, the S of an influence factor test level is the design load of this influence factor is carried out after the adjustment of S different amplitudes, the trial value of acquisition;
Step 403, orthogonal test sample acquisition: according to the design parameter of inputting in step 401, and the orthogonal arrage of setting up in integrating step 402, obtain all test samples used when current designed soil bolt is carried out to multilevel orthogonal test of multiple factors; Described in each, in test sample, include the trial value of M described influence factor of current designed soil bolt;
Step 404, anchor force prediction: the soil bolt anchor force forecast model training in described data processor invocation step three, the corresponding anchor force of all test samples in step 403 is predicted, and obtained the anchor force predicted value of S different tests level of M described influence factor;
Step 405, anchor force predicted value arrange: according to the anchor force predicted value of S different tests level of the M obtaining in step 404 described influence factor, calculate the anchor force predicted value mean value of S different tests level of influence factor described in each; Wherein, in M described influence factor, the anchor force predicted value mean value of t test level of j influence factor is denoted as M
jt, j and t are positive integer, j=1,2 ..., M, t=1,2 ..., S;
Afterwards, according to the anchor force predicted value mean value of S different tests level of the M calculating a described influence factor, calculate the extreme difference of influence factor described in each; Wherein, in M described influence factor, the extreme difference of j influence factor is denoted as Δ M
jt, Δ M
jt=M
jmax-M
jmin, M in formula
jmaxbe the maximal value in the anchor force predicted value mean value of S different tests level of j influence factor, M
jminit is the minimum value in the anchor force predicted value mean value of S different tests level of j influence factor;
Step 5, susceptibility are determined: according to the extreme difference of the M calculating in step 405 described influence factor, the susceptibility of M described influence factor is determined; Wherein, the influence factor that extreme difference is larger is more responsive, larger to the anchor force influence degree of designed soil bolt, and the extreme difference size of each influence factor determines the anchor force influence degree of this influence factor to designed soil bolt.
2. according to the assay method of a kind of soil bolt anchor force influence factor susceptibility claimed in claim 1, it is characterized in that: M=5 in step 1.
3. according to the assay method of a kind of soil bolt anchor force influence factor susceptibility described in claim 1 or 2, it is characterized in that: after the anchor force of soil bolt described in step 3 forecast model has been trained, also need to adopt the soil bolt anchor force forecast model that test sample book set pair trains to test; Described test sample book collection is the sample set being comprised of the concentrated a plurality of training samples of training sample described in step 2 or the sample set re-establishing according to the acquisition methods of training sample set described in step 2.
4. according to the assay method of a kind of soil bolt anchor force influence factor susceptibility claimed in claim 2, it is characterized in that: 5 described influence factors are respectively anchoring section length, bolt diameter, aperture, grouting body intensity and anchor pole angle of inclination.
5. according to the assay method of a kind of soil bolt anchor force influence factor susceptibility claimed in claim 3, it is characterized in that: when described soil bolt anchor force forecast model is tested, the sim function that described data processor calls in Matlab Neural Network Toolbox is tested, and generate the anchor force predicted value that each training sample is corresponding, afterwards each is detected to anchor force predicted value that sample is corresponding and the anchor force measured value of this detection sample and carry out difference comparison, draw the predicated error that this detection sample is corresponding; After each training sample has all been tested in described training sample set, obtain predicated error corresponding to each training sample in described training sample set; Afterwards, the predicated error corresponding according to each training sample, analyzes the precision of prediction of the soil bolt anchor force forecast model training.
6. according to the assay method of a kind of soil bolt anchor force influence factor susceptibility described in claim 1 or 2, it is characterized in that: when described soil bolt anchor force forecast model is trained in step 3, " trainlm " function that described data processor calls in Matlab Neural Network Toolbox is trained.
7. according to the assay method of a kind of soil bolt anchor force influence factor susceptibility described in claim 1 or 2, it is characterized in that: before in step 3, described soil bolt anchor force forecast model being trained, described data processor first calls normalized module and concentrates the data in each training sample to be normalized to described training sample; After having trained, described data processor also needs to call renormalization processing module to the predicted value generating being carried out to renormalization processing in training process;
Before carrying out anchor force prediction in step 404, described data processor first calls described normalized module the data in all test samples in step 403 is normalized, after calling the corresponding anchor force of all test samples of described soil bolt anchor force forecast model and predicting, also need to call described renormalization processing module generated predicted value is carried out to renormalization processing.
8. according to the assay method of a kind of soil bolt anchor force influence factor susceptibility described in claim 1 or 2, it is characterized in that: while setting up described soil bolt anchor force forecast model in step 3, the soil bolt anchor force forecast model of setting up comprises an input layer, a hidden layer and an output layer, described input layer comprises M corresponding with the individual described influence factor of M respectively node, described hidden layer comprises (2M+1) individual node, and described output layer comprises a node.
9. according to the assay method of a kind of soil bolt anchor force influence factor susceptibility described in claim 1 or 2, it is characterized in that: S=5 in step 402.
10. according to the assay method of a kind of soil bolt anchor force influence factor susceptibility claimed in claim 9, it is characterized in that: in step 402 described in each 5 test level of influence factor be to the design load of this influence factor raise 20%, raise 10%, raise 0%, lower 10% and lower 20% after, the trial value of acquisition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410181912.3A CN103927458A (en) | 2014-04-30 | 2014-04-30 | Determination method of sensibility of influence factors of anchoring force of soil anchors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410181912.3A CN103927458A (en) | 2014-04-30 | 2014-04-30 | Determination method of sensibility of influence factors of anchoring force of soil anchors |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103927458A true CN103927458A (en) | 2014-07-16 |
Family
ID=51145676
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410181912.3A Pending CN103927458A (en) | 2014-04-30 | 2014-04-30 | Determination method of sensibility of influence factors of anchoring force of soil anchors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103927458A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106501465A (en) * | 2016-12-23 | 2017-03-15 | 石家庄铁道大学 | A kind of detection method for detecting Detection of Bolt Bonding Integrity |
| CN109187770A (en) * | 2018-10-23 | 2019-01-11 | 四川升拓检测技术股份有限公司 | A kind of anchor pole AI detection method |
| CN111365051A (en) * | 2020-03-06 | 2020-07-03 | 广西交通设计集团有限公司 | Method for estimating stress of carbonaceous rock tunnel anchor rod based on transfer function of feedback algorithm |
-
2014
- 2014-04-30 CN CN201410181912.3A patent/CN103927458A/en active Pending
Non-Patent Citations (2)
| Title |
|---|
| 李东波 等: "基于 ANN的碳纤维楠竹锚杆锚固力预测研究", 《力学与实践》 * |
| 李东波: "基于人工神经网络的碳纤维楠竹锚杆锚固力预测研究", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106501465A (en) * | 2016-12-23 | 2017-03-15 | 石家庄铁道大学 | A kind of detection method for detecting Detection of Bolt Bonding Integrity |
| CN106501465B (en) * | 2016-12-23 | 2018-11-13 | 石家庄铁道大学 | A kind of detection method for detecting Detection of Bolt Bonding Integrity |
| CN109187770A (en) * | 2018-10-23 | 2019-01-11 | 四川升拓检测技术股份有限公司 | A kind of anchor pole AI detection method |
| CN111365051A (en) * | 2020-03-06 | 2020-07-03 | 广西交通设计集团有限公司 | Method for estimating stress of carbonaceous rock tunnel anchor rod based on transfer function of feedback algorithm |
| CN111365051B (en) * | 2020-03-06 | 2021-04-06 | 广西交通设计集团有限公司 | Method for estimating stress of carbonaceous rock tunnel anchor rod based on transfer function of feedback algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Momeni et al. | Application of artificial neural network for predicting shaft and tip resistances of concrete piles | |
| RU2014124177A (en) | SYSTEM AND METHOD OF APPLICATION OF INDEPENDENT DATA SUBSETS IN THE SPACE FOR DETERMINING UNCERTAINTY OF DISPLACEMENT OF UNRELIABLE PROPERTIES OF DISTRIBUTION OF THE DISTRIBUTION OF CORRELATED DATA IN THE SPACE ON THE COLLECTOR LAYER | |
| CN102147479B (en) | Modelling method of reservoir space physical property parameters | |
| CN103077267A (en) | Parameter sound source modeling method based on improved BP (Back Propagation) neural network | |
| Grimstad et al. | A non-associated creep model for structured anisotropic clay (n-SAC) | |
| CN101847171A (en) | Back analysis method of slope displacement based on safety monitoring | |
| CN108876021A (en) | A kind of Medium-and Long-Term Runoff Forecasting method and system | |
| CN107657393A (en) | The Seismic Evaluation method of the lower bridge of near-fault ground motion effect | |
| CN109726902A (en) | A kind of Slope Stability Evaluation method | |
| CN103856951A (en) | Coverage hole eliminating method based on multi-sensor probability perceptual model | |
| CN103927458A (en) | Determination method of sensibility of influence factors of anchoring force of soil anchors | |
| Venanzi et al. | Robust optimization of a hybrid control system for wind-exposed tall buildings with uncertain mass distribution | |
| Sivakumar | Application of Electronic Spread Sheet and Water Balance Error Optimization Technique in Ground Water Model Study to Improve the Ground Water System in Restricted Area | |
| Jafariavval et al. | New formulae for capacity energy-based assessment of liquefaction triggering | |
| CN102495944B (en) | Time series forecasting method and equipment and system adopting same | |
| CN103898890A (en) | Soil layer quantization layering method based on double-bridge static sounding data of BP neural network | |
| CN105844089A (en) | Construction method of damage estimation model of RC (Reinforced Concrete) frame structure layer | |
| Esmaeilabadi et al. | Prediction of site response spectrum under earthquake vibration using an optimized developed artificial neural network model | |
| Zhou et al. | Soft soil foundation settlement prediction and economic cost management analysis based on new algorithm | |
| Li et al. | Bird mating optimizer in structural damage identification | |
| Shooshpasha et al. | AN INVESTIGATION OF FRICTION ANGLE CORRELATION WITH GEOTECHNICAL PROPERTIES FOR GRANULAR SOILS USING GMDH TYPE NEURAL NETWORKS (RESEARCH NOTE) | |
| Tanaka et al. | Estimation of geological data using assisted history matching in the pilot-scale CO2 injection site, Iwanohara, Japan | |
| Bari et al. | Effects of soil spatial variability on axisymmetric versus plane strain analyses of ground improvement by prefabricated vertical drains | |
| Baltaji et al. | Non-linear time domain site response and soil structure interaction analyses for nuclear facilities using moose | |
| Sharafi et al. | Assessment of Cyclic Resistance Ratio (CRR) in Silty Sands Using Artificial Neural Networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140716 |