CN115419120A - Highway subgrade settlement monitoring and predicting method - Google Patents
Highway subgrade settlement monitoring and predicting method Download PDFInfo
- Publication number
- CN115419120A CN115419120A CN202210652439.7A CN202210652439A CN115419120A CN 115419120 A CN115419120 A CN 115419120A CN 202210652439 A CN202210652439 A CN 202210652439A CN 115419120 A CN115419120 A CN 115419120A
- Authority
- CN
- China
- Prior art keywords
- neural network
- monitoring
- subgrade
- settlement
- predicting
- 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
- 238000000034 method Methods 0.000 title claims abstract description 91
- 238000012544 monitoring process Methods 0.000 title claims abstract description 85
- 238000003062 neural network model Methods 0.000 claims abstract description 9
- 238000013528 artificial neural network Methods 0.000 claims description 100
- 238000012549 training Methods 0.000 claims description 36
- 230000008569 process Effects 0.000 claims description 25
- 230000006870 function Effects 0.000 claims description 20
- 229920001940 conductive polymer Polymers 0.000 claims description 19
- 210000002569 neuron Anatomy 0.000 claims description 19
- 239000002689 soil Substances 0.000 claims description 19
- 238000006073 displacement reaction Methods 0.000 claims description 15
- 238000010276 construction Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 4
- 230000001681 protective effect Effects 0.000 claims description 4
- DMBHHRLKUKUOEG-UHFFFAOYSA-N diphenylamine Chemical compound C=1C=CC=CC=1NC1=CC=CC=C1 DMBHHRLKUKUOEG-UHFFFAOYSA-N 0.000 claims description 3
- 238000004806 packaging method and process Methods 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims 2
- 230000000737 periodic effect Effects 0.000 claims 1
- 238000004364 calculation method Methods 0.000 description 11
- 239000011159 matrix material Substances 0.000 description 10
- 230000004913 activation Effects 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 230000005540 biological transmission Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 239000000463 material Substances 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000002787 reinforcement Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 241000287196 Asthenes Species 0.000 description 1
- 101100391182 Dictyostelium discoideum forI gene Proteins 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002242 deionisation method Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000002045 lasting effect Effects 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- E—FIXED CONSTRUCTIONS
- E02—HYDRAULIC ENGINEERING; FOUNDATIONS; SOIL SHIFTING
- E02D—FOUNDATIONS; EXCAVATIONS; EMBANKMENTS; UNDERGROUND OR UNDERWATER STRUCTURES
- E02D33/00—Testing foundations or foundation structures
-
- E—FIXED CONSTRUCTIONS
- E01—CONSTRUCTION OF ROADS, RAILWAYS, OR BRIDGES
- E01C—CONSTRUCTION OF, OR SURFACES FOR, ROADS, SPORTS GROUNDS, OR THE LIKE; MACHINES OR AUXILIARY TOOLS FOR CONSTRUCTION OR REPAIR
- E01C3/00—Foundations for pavings
-
- E—FIXED CONSTRUCTIONS
- E02—HYDRAULIC ENGINEERING; FOUNDATIONS; SOIL SHIFTING
- E02D—FOUNDATIONS; EXCAVATIONS; EMBANKMENTS; UNDERGROUND OR UNDERWATER STRUCTURES
- E02D1/00—Investigation of foundation soil in situ
Landscapes
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Civil Engineering (AREA)
- Structural Engineering (AREA)
- Mining & Mineral Resources (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Paleontology (AREA)
- General Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Soil Sciences (AREA)
- Architecture (AREA)
- Traffic Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a highway subgrade settlement monitoring and predicting method. According to the invention, the inside and outside combined monitoring system of the highway subgrade is constructed through the intelligent smart geobelt and the Beidou satellite for subgrade deformation monitoring, and the lightweight neural network model is constructed to realize high-efficiency and high-precision prediction of highway subgrade settlement, so that the system is not only suitable for common highway subgrades, but also suitable for widening the highway subgrade, has wide application scenes, can realize full-section full-life distributed monitoring of subgrade deformation, and can realize accurate and high-efficiency prediction of highway subgrade settlement.
Description
Technical Field
The invention relates to the technical field of highway engineering, in particular to a highway subgrade settlement prediction method.
Background
The settlement deformation of the roadbed presents a highly nonlinear characteristic on the change trend, particularly in the road reconstruction and expansion project, the construction time of a newly expanded roadbed and an original old roadbed is usually long, and factors such as filling materials, construction quality, construction parameters and the like all affect the settlement deformation of the roadbed. Therefore, monitoring and predicting the deformation of the roadbed is beneficial to timely taking measures to carry out post-treatment on the roadbed with overlarge deformation, and effective and safe early warning of the catastrophe of the roadbed can be realized.
The application number CN201110257149.4 patent discloses a roadbed settlement prediction method based on static sounding and a BP neural network, field data and settlement observation data are collected, and then ordinary roadbed settlement prediction is carried out. However, in the solving process of the ordinary BP neural network, various problems such as under-fitting, slow convergence or over-fitting easily occur along with some influence factors such as the expansion of the scale of the input data set, the increase of the number of layers, the complexity of the structure or the connection form and the like.
The method comprises the following steps of monitoring lateral deformation by adopting an inclinometer in patent application numbers of CN201611126688.3, CN201911367265.4 and the like, but has the disadvantages of high actual installation cost, complex measurement steps, short service life of monitoring equipment, incapability of carrying out permanent monitoring along with the service life of the roadbed and incapability of obtaining full-section horizontal displacement.
Therefore, how to develop a highway subgrade settlement monitoring and predicting method based on a neural network has a full-section full-life distributed deformation monitoring system, is not only applicable to common highway subgrades, but also can realize accurate and efficient prediction of widened subgrade settlement, and is a technical problem to be solved urgently by technical personnel in the field.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a highway subgrade settlement monitoring and predicting method which has wide application scenes, can realize full-section full-life distributed monitoring of subgrade deformation, and has high subgrade settlement predicting precision and high roadbed settlement predicting efficiency.
In order to realize the purpose, the invention provides the following technical scheme:
a highway subgrade settlement monitoring and predicting method starts from a basic feedforward neural network, and establishes a light neural network model for subgrade settlement prediction by means of optimizing a topological structure of the basic feedforward neural network in a form, adding a punishment item to a learning parameter of the basic neural network in an algorithm and the like. In order to further accelerate the calculation of the neural network model and improve the over-fitting problem, the light neural network abandons the current popular algorithm which is easy to cause local optimum such as gradient descent and the like in the algorithm, and a novel quasi-Newton method is adopted to complete the training of the neural network, so that the light neural network can improve the calculation efficiency of the light neural network while ensuring the prediction precision, thereby avoiding the over-fitting problem. And after the design of the light neural network model is completed, monitoring the roadbed displacement by adopting an internal and external combined monitoring method. And preprocessing the data set of the obtained measured data to finish the training process of the light neural network model, and using the light neural network model for highway subgrade settlement monitoring and prediction. The prediction capability, the calculation efficiency and the optimized structure of the light neural network are comprehensively evaluated, so that the highway subgrade settlement prediction method based on the light neural network is provided.
Further, the basic feedforward neural network is a simple 3-layer feedforward neural network, and comprises 1 input layer, 1 hidden layer and 1 output layer, wherein j neurons are contained in the hidden layer. For this basic feedforward neural network, the input quantityI n The neural network is entered from the input layer and then into the hidden layer. In the middle hidden layer, there are j number of neurons, which are finally output through the output layer.
Further, the topology structure of the formally optimized basic neural network is based on the basic feedforward neural network to construct a complex neural network with more hidden layers, and the net input quantity in the learning process is used as a new input quantity to be input into the neuron of the next hidden layer with new weight and bias. Similarly, the layer of neurons can convert the new input quantities again through the activation function, and then input the converted results into the next hidden layer, and the process is circulated until the converted results are transmitted to the output layer to complete the topology optimization of the basic neural network model.
Furthermore, adding a penalty term to the learning parameters of the basic neural network on the basis of the algorithm is to train network model parameters by using a least square error function as a loss function. By using L for the loss function 1 And (4) punishing the norm to obtain a relatively sparse structure of the parameters in advance, so as to complete the lightweight construction of the neural network.
Furthermore, the novel quasi-Newton method is a novel quasi-Newton method for constructing the search direction and the search step length on the basis of a sub-gradient method and an active set method.
Furthermore, the inside and outside combined monitoring method is characterized in that a reference point is arranged at a specified position on a slope outside a monitoring section, and a Beidou monitoring system is adopted to regularly observe the vertical displacement of the surface of the roadbed. Meanwhile, the soil deformation monitoring system and method based on the conductive polymer are adopted, and the smart conductive polymer is buried in the roadbed when the roadbed is constructed, so that the horizontal displacement observation of the full section is carried out.
Furthermore, the Beidou monitoring system carries out positioning through three second-generation Beidou satellites, a relative positioning method is adopted, the vertical displacement of the surface of the section of the roadbed is monitored through a global satellite system, and the horizontal displacement inside the roadbed is monitored through an intelligent smart geotechnical belt by the soil body deformation monitoring system based on the conductive polymer. The Beidou monitoring system comprises a Beidou receiving antenna, a fixed power supply or a solar cell panel, a receiver, a wireless transmission module and the like. The Beidou receiving antenna is connected with a datum point at a specified position through a connecting support, and the datum point is respectively arranged at the slope toe position of a new roadbed, the road shoulder position and the top of a combining part of the new roadbed and the old roadbed of the test section. The receiver, the wireless transmission module and the like are all placed in an equipment protection box. And carrying out accurate positioning by using the second-generation Beidou three satellites, and calculating the relative position relation between the point to be measured and the datum point by using a global satellite system baseline by adopting a relative positioning method.
Furthermore, the soil deformation monitoring system based on the conductive polymer adopts the smart conductive polymer as a soil internal monitoring element. Connecting cables at different measuring points on the strip-shaped smart conductive polymer, extending the cables at a plurality of measuring points out of the tail end of the strip, and then packaging the strip with an insulating protective sleeve to prepare the smart geobelt. Inside smart geotechnical zone level was buried the road bed, its end was exposed from the domatic department of road bed side slope, was connected with the cable and is connected with the data acquisition station, and the high in the clouds server is passed to through the network in the acquisition station, realizes the automatic collection monitoring of monitoring data with the help of the high in the clouds server.
Further, the data set preprocessing is to process the training data set by a normalization method, and the training data set can be mapped into the range of [0,1], so that the generalization capability of the light neural network is enhanced.
Further, in the training process of the light neural network model, a training data set is divided into a verification set and a training set, limited data is randomly extracted to serve as the verification set, and the rest of the data is used as the training set.
Compared with the prior art, the invention has the beneficial effects that:
1. the inside and outside combined monitoring method can realize high-precision real-time intelligent monitoring of roadbed deformation data.
Specifically, the method adopts a short-baseline double-difference Beidou monitoring scheme to observe the deformation of the roadbed, and adopts a method combining double-difference non-geometric distance combination and a method combining a deionization layer observation value minus Wei Deju to process the observation value for the collected data. The influence of the geometric distance of the station satellite, the atmospheric layer delay and various noises on the positioning is weakened or eliminated, so that the measurement error is reduced, and the positioning accuracy of the Beidou global satellite system is improved.
In particular, the smart conductive polymer adopted in the patent has good durability, high strength and low manufacturing cost. The rib material is provided with a self-detection technology without embedding a sensor in the material. And the detection is accurate in time, the performance is stable, through extracting the deformation of the reinforcement material, the stress characteristic information and the like, the full-section positioning soil body fracture surface of the cross section of the roadbed can be realized, the damage state of the reinforced soil body can be diagnosed, and more reasonable technical means can be provided for soil body safety early warning.
It can be seen that the integrated intelligent monitoring can be remotely realized by adopting an internal and external combined detection method, the measurement personnel do not need to manually operate and read monitoring data on site, and the monitoring system can not only perform high-precision real-time monitoring, but also realize lasting monitoring along with the service life of the roadbed.
2. Using a loss function and L 1 Norm punishment realizes the lightweight construction of the neural network.
Specifically, a least squares error function is used as a loss function to train network parameters, using L for the loss function 1 And (4) punishing the norm to obtain a relatively sparse structure of the parameters in advance, so as to complete the lightweight construction of the neural network. The method avoids various problems such as under-fitting, slow convergence or over-fitting caused by the transformation of a series of influencing factors such as the enlargement of the input data set, the increase of the number of layers, the complication of the structure or the connection form and the like.
3. And a novel quasi-Newton method based on a sub-gradient method and an active set method is adopted to realize the driving of the light neural network.
In particular, an approximation matrix is used
To replace the Heisel matrix in Newton's method
Avoiding the need to compute the inverse of the hessel matrix. The calculation process is very complicated and can not ensure that the hessel matrix in the iterative process keeps positive nature, so that the learning of the neural network is easily trapped into local optimum, and an accurate prediction result can not be obtained. On the basis of the quasi-Newton method, the super-parameters are secondarily planned by adopting an active set method, and the sparse structure of the light neural network is thoroughly obtained, so that the overfitting risk of the light neural network is effectively reduced, and the learning efficiency of the light neural network can be obviously improved.
4. And the light neural network is adopted to realize high-precision and high-efficiency prediction of subgrade settlement.
Based on one hand of data measured by an internal and external joint monitoring method, the settlement value obtained by predicting by using the light neural network is mostly matched with the corresponding observed settlement value, the prediction accuracy of the light neural network on subgrade settlement is better, the prediction capability is stable, and meanwhile, the training efficiency can be greatly improved by the light neural network.
Drawings
FIG. 1 is a technical flow chart of a highway subgrade settlement monitoring and predicting method;
FIG. 2 is a diagram of a basic feedforward neural network topology;
FIG. 3 is a comparison graph of measured and predicted values of settlement of the horizontally filled and widened roadbed in example 1;
FIG. 4 is a graph comparing the calculation efficiency of the general BP neural network and the light neural network in example 1;
FIG. 5 is a comparison graph of measured and predicted values of subgrade settlement in the partial filling and widening of example 2;
FIG. 6 is a graph comparing the calculation efficiency of the general BP neural network and the light neural network in example 2;
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention relates to a highway subgrade settlement monitoring and predicting method, which comprises the following processes as shown in figure 1:
optimizing the topological structure S1 of a basic neural network, adding a punishment item S2 to a learning parameter, training the neural network by adopting a novel quasi-Newton method S3, monitoring the roadbed displacement by adopting an internal and external combined monitoring method S4, preprocessing an actually measured data set S5 and training a light neural network S6.
S1, optimizing the topological structure of a basic neural network:
in the basic neural network shown in fig. 2, a signal intensity (weight) represented by an unknown coefficient w and an intensity adjustment term (bias) represented by b for each input quantity are connected to each neuron element with signs
Representing the input of the nth quantity in the input layer with a weight w to the jth neuron in the hidden layer; by symbol
A bias term representing the jth neuron in the ith hidden layer. The j neuron then converts the input quantity by an activation function (neuron activation) which is a specific non-linear function, the converted input quantity being called net input quantity, symbolized
Representing the net input to the j-th activated neuron in the i-th hidden layer. Finally, the activated neuron will pass the net input to the output layer output.
Based on the basic feedforward neural network shown in fig. 2, a complex neural network with more hidden layers is constructed, and the net input quantity in the learning process is input as a new input quantity to the neuron of the next hidden layer with a new weight and bias. Similarly, the layer of neurons will again convert these "new inputs" through the activation function, and then input the converted results into the next hidden layer, and so on until they are passed to the output layer for output. The basic neural network topology optimization process can be expressed as follows:
wherein, I ij Is a matrix of input quantities x, the subscripts indicating the input of the ith quantity to the jth neuron of the 1 st hidden layer; i is the number of input quantities; j is the number of neurons per hidden layer; e.g. of the type i Is a residual error; w is a k Is the weight of the kth neuron of the 1 st hidden layer; b is a mixture of k Is the bias of the kth neuron of the 1 st hidden layer;
the k-th preset parameter is input into the neural network in advance, consists of weight and bias, and is obtained by training the neural network in advance; a is k Is an activation function, and a hyperbolic tangent function is adopted in the research, namely a k ＝(e 2x -1)/(e 2x +1)。
Input to the next hidden layer and so on until passed to the output layer output. The basic neural network topology optimization process can be expressed as follows:
s2, adding a penalty term to the learning parameters:
training network parameters using a least squares error function as a loss function, said loss function being expressed as follows:
the neural network parameter training process is to make the input value x ij Through the process of minimization of the loss function. By using L for the loss function 1 And (4) punishing the norm to obtain a relatively sparse structure of the parameters in advance, so as to complete the lightweight construction of the neural network. The process described can be expressed as follows:
wherein λ is k,j And λ k Are all determined
w k 、b k Sparsity Lagrange multiplier factor, beta j [k] The hyper-parameter is a series of adjustment values obtained during the training process.
S3, training the neural network by adopting a novel quasi-Newton method:
if order
Then apply L 1 The learning process of the norm-penalized light neural network can be expressed as follows:
in this patent, for
Is selected for use
The direction of the sub-gradient of (a) is taken as the steepest descent direction,
for psi r The sub-gradients of the components can be expressed as:
wherein:
is obtained by
After the sub-gradients for each parameter, pair
And processing by adopting the novel quasi-Newton algorithm based on the sub-gradient. First using Taylor expansion to obtain
In phi tau]The approximation form of (c):
wherein,
is a sea seoul matrix.
In minimizing the approximation form, the sub-gradient-based search direction can be expressed as:
the whole iteration form can be expressed as:
ψ[τ]+ηh[τ]→ψ[τ+1]
wherein: η is the advance value obtained by line search.
Then using the approximation matrix
To replace the sea seire matrix
Thereby avoiding complex calculations, i.e. the iterative form can be expressed as:
wherein,
ζ[τ]＝ψ[τ+1]-ψ[τ]；
is an identity matrix.
S4, monitoring the roadbed displacement by adopting an internal and external combined monitoring method:
the Beidou monitoring system and the soil deformation monitoring system based on the conductive polymer are adopted to realize the real-time monitoring of the roadbed displacement by an internal and external combined monitoring method.
The Beidou monitoring system comprises a Beidou receiving antenna, a fixed power supply or a solar cell panel, a receiver, a wireless transmission module and the like. Receiver, wireless transmission module etc. all place in an equipment protecting box, and this equipment protecting box accessible fixed bolster or expansion bolts are fixed in near big dipper receiving antenna to reach the purpose of protective apparatus.
The Beidou receiving antenna is connected with a datum point at a specified position through a connecting support, and the datum point is respectively arranged at the slope foot of a new roadbed, the road shoulder and the top of a joint of the new roadbed and the old roadbed on the test section.
The real-time monitoring is to use three second-generation Beidou satellites for accurate positioning, and a relative positioning method is adopted to obtain the relative position relation between the point to be measured and the datum point through global satellite system baseline calculation.
The relative positioning method is characterized in that more than two receivers are used for simultaneously measuring, and common errors of signals of the two receivers are eliminated through differential operation, so that relative coordinates with higher accuracy between the two receivers are obtained.
By adopting the soil deformation monitoring system and method (CN 201310312664.7) based on the conductive polymer, the smart conductive polymer is buried in the roadbed when the roadbed is widened to be constructed, and the horizontal displacement observation of the full section is carried out. The soil deformation monitoring system based on the conductive polymer adopts the smart conductive polymer as a soil internal monitoring element. Connecting cables at different measuring points on the strip-shaped conductive polymer, extending the cables of a plurality of measuring points out of the tail end of the strip, packaging the strip with an insulating protective sleeve to realize water resistance, abrasion resistance and static electricity resistance, and further preparing the intelligent smart geobelt.
The intelligent smart geotechnical belt is horizontally embedded into the roadbed, the tail end of the smart geotechnical belt is exposed out of the slope of the roadbed slope, a cable is connected with the data acquisition station, and the acquisition station uploads the data to the cloud server through the network.
The real-time monitoring is realized by realizing the detection of the deformation inside the soil body in the whole life cycle through the distributed self-checking technology of the deformation of the reinforcement, persistently and effectively positioning the sliding surface of the potential crack inside the soil body, diagnosing the deformation state inside the soil body, and realizing the automatic acquisition and monitoring of monitoring data by virtue of a cloud server, thereby providing a reasonable technical means for realizing the safety early warning of the soil body.
S5, actual measurement data set preprocessing:
the settlement observed values of the subgrade are respectively expressed as
S i 、
The strain values of the smart geotechnical belt are respectively recorded as
Then there are:
wherein: the superscript "1" indicates the first monitoring section;
representing an amount of time, with other marker meanings as indicated above. The data set used to train the lightweight neural network is as follows:
wherein:
S i ＝(S 1 ,S 2 ,...,S 180 ) T
in order to improve the generalization capability of the light neural network, the training data set needs to be normalized, and the normalization process is to process each column of vectors in the training data set according to the following method:
wherein: z represents normalized data; x represents an element in each column vector; x min 、X max The minimum and maximum elements in each column vector are represented separately.
The training data set is processed by adopting the normalization method, and the training data set can be mapped into the range of [0,1], so that the generalization capability of the light neural network is enhanced.
S6, training a light neural network:
the training data set is divided into a verification set and a training set, 10% of data is randomly extracted to serve as the verification set, and the rest of data is used as the training set. Meanwhile, in order to quantitatively measure the prediction capability of the light neural network and the common feedforward neural network, besides visually comparing a predicted value with an actual observed value, the prediction result is analyzed by adopting an index of a Root Mean Square Error (RMSE). The RMSE calculation method is as follows:
wherein:
representing a predicted value; s i Representing the corresponding actual observations.
For the neural network, when the predicted value is completely matched with the observed value, namely the prediction accuracy is 100%, the corresponding RMSE value is 0. In other words, the more predictive capability of the neural network, the higher the accuracy, the lower the RMSE value, and vice versa.
Example 1: horizontal filling broadening roadbed settlement monitoring prediction
The method is based on the reconstruction and expansion project of a certain highway in Shandong province, the site is basically yellow river plain alluvial silt, the project property is poor, the underground water level is high, and a weak layer exists. After 20 years of service operation, the overall deformation condition of the old roadbed is basically stable. The height of an old roadbed in the road section is 6m, the right half width of the top surface is 14m, the widening width is 7m, and the slope ratio is 1.75. The traditional horizontal layered filling is adopted, the width of a filling part of a newly-built roadbed is 5m, and the height of the filling part is 2m. The highway subgrade settlement monitoring and predicting method is adopted to conduct horizontal filling and widening subgrade settlement monitoring and predicting.
Fig. 3 is a comparison between the measured value and the predicted value of the settlement of the horizontal filling widening roadbed, and it can be seen that the difference between the predicted value and the actual value of the light neural network is 0.00721mm at the minimum and 0.42719mm at the maximum. From the aspect of prediction accuracy, the lowest accuracy of the light neural network in the working process is 88.67%, the highest accuracy is 99.87%, and the average accuracy is 96.16%, which shows that the light neural network can keep a higher accuracy in predicting the settlement of the horizontally filled and widened roadbed.
And (3) comparing the light neural network with the common BP neural network under the same computing environment, and testing the computing efficiency of the light neural network and the common feedforward neural network. The time required to record each neural network starting near the same larger RMSE and initially trained to a smaller RMSE value is shown in fig. 4. The time taken for the two neural networks to fall to 35 is uniformly calculated. For the BP neural network, the time for the RMSE value to drop to a lower level is longer and 18.78 times of that of the light neural network, so that the light neural network can greatly improve the training efficiency.
Example 2: partial filling broadening roadbed settlement monitoring prediction
The method is based on the reconstruction and expansion project of a certain highway in Shandong province, the site is basically yellow river plain alluvial silt, the project property is poor, the underground water level is high, and a weak layer exists. After 20 years of service operation, the overall deformation condition of the old roadbed is basically stable. The height of an old road base in the road section is 6m, the right half width of the top surface is 14m, the widening width is 7m, and the slope ratio is 1.75. And selecting a test section 20m long on the road section, and widening the roadbed by filling in parts, wherein the width of the part of the newly-built roadbed filled in advance is 5m, and the height of the part of the newly-built roadbed is 2m. The highway subgrade settlement monitoring and predicting method is adopted to conduct fractional filling and widening subgrade settlement monitoring and predicting.
Fig. 5 is a comparison between the measured value and the predicted value of the settlement of the partial filling and widening roadbed, and it can be seen that the difference between the predicted value and the actual value of the light neural network is 0.00012mm at the minimum and 0.38246mm at the maximum. From the aspect of prediction accuracy, the lowest accuracy of the light neural network in the working process is 89.00%, the highest accuracy is 99.99%, and the average accuracy is 96.51%, which shows that the light neural network can maintain a higher accuracy in predicting the settlement of the partial filling and widening roadbed.
And (3) testing the calculation efficiency of the light neural network and the common feedforward neural network by using the common BP neural network as a reference under the same calculation environment. The time required to record each neural network starting near the same larger RMSE and initially trained to a smaller RMSE value is shown in fig. 6. The time taken for the two neural networks to fall to 35 is uniformly calculated. For the BP neural network, the time for the RMSE value to drop to a lower level is longer and is 20.16 times of that of the light neural network, so that the light neural network can greatly improve the training efficiency.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (10)
1. A highway subgrade settlement monitoring and predicting method is characterized in that a light neural network model for subgrade settlement prediction is established by optimizing a topological structure of a basic feedforward neural network in form and adding punishment items to learning parameters of the basic neural network in algorithm, training of the neural network is completed by adopting a novel quasi-Newton method, subgrade displacement monitoring is performed by adopting an internal and external combined monitoring method after the light neural network is designed, and data set preprocessing is performed on obtained actual measurement data to complete a training process of the light neural network and use the training process for highway subgrade settlement monitoring and prediction.
2. The method for monitoring and predicting the subgrade settlement of the highway according to claim 1, wherein the basic feedforward neural network is a 3-layer feedforward neural network comprising 1 input layer, 1 hidden layer and 1 output layer, wherein the hidden layer comprises j neurons; for this basic feedforward neural network, input quantity I n The data can enter a neural network from an input layer and then enter a hidden layer; in the middle hidden layer, there are j number of neurons, which are finally output through the output layer.
3. The method for monitoring and predicting the subgrade settlement of the highway according to claim 1, wherein the topological structure for formally optimizing the basic neural network is based on the basic feedforward neural network to construct a complex neural network with more hidden layers.
4. The method for monitoring and predicting the subgrade settlement of the highway according to claim 1, wherein the penalty term is algorithmically added to the learning parameters of the basic neural network by training network model parameters by using a least square error function as a loss function; by using L for the loss function 1 And (4) punishing the norm to obtain a relatively sparse structure of the parameters in advance, so as to complete the lightweight construction of the neural network.
5. The method for monitoring and predicting the settlement of the roadbed of claim 1, wherein the novel quasi-Newton method is a novel quasi-Newton method for constructing the search direction and the search step size on the basis of a sub-gradient method and an active set method.
6. The method for monitoring and predicting the settlement of the road subgrade as claimed in claim 1, wherein the inside and outside combined monitoring method is to set reference points at specified positions on the side slope outside the monitored section, and use a Beidou monitoring system to perform periodic observation of the vertical displacement of the surface of the subgrade, and simultaneously use a soil deformation monitoring system and method based on the conductive polymer to embed the smart conductive polymer into the subgrade during widening the subgrade construction to perform horizontal displacement observation of the whole section.
7. The inside and outside combined monitoring method according to claim 6, wherein the big dipper monitoring system is positioned by three second generation big dipper satellites, the vertical displacement of the roadbed section surface is monitored by a global satellite system by adopting a relative positioning method, and the horizontal displacement in the roadbed is monitored by the soil deformation monitoring system based on the conductive polymer through an intelligent smart geotechnical belt.
8. The inside-outside combination monitoring method according to claim 6, wherein the soil deformation monitoring system based on conductive polymers adopts smart conductive polymers as soil interior monitoring elements; connecting cables at different measuring points on the banded smart conductive polymer, extending the cables at a plurality of measuring points out of the tail end of the band, and then packaging the band with an insulating protective sleeve to prepare an intelligent smart geoband; inside intelligent smart geotechnical zone level buried the road bed, its terminal domatic department from the road bed side slope exposes, is connected with the cable and is connected with the data acquisition station, and the high in the clouds server is passed to through the network in the acquisition station, realizes the automatic collection monitoring of monitoring data with the help of the high in the clouds server.
9. The method for monitoring and predicting the subgrade settlement of the highway according to claim 1, wherein the preprocessing of the data set is to process a training data set by a normalization method, and map the training data set into the range of [0,1] so as to enhance the generalization capability of the lightweight neural network.
10. The method for monitoring and predicting the settlement of the road subgrade as claimed in claim 1, wherein the training process of the light neural network is to divide a training data set into a validation set and a training set, randomly extract limited data as the validation set, and use the rest as the training set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210652439.7A CN115419120B (en) | 2022-06-08 | 2022-06-08 | Highway subgrade settlement monitoring and predicting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210652439.7A CN115419120B (en) | 2022-06-08 | 2022-06-08 | Highway subgrade settlement monitoring and predicting method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115419120A true CN115419120A (en) | 2022-12-02 |
| CN115419120B CN115419120B (en) | 2024-07-19 |
Family
ID=84196565
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210652439.7A Active CN115419120B (en) | 2022-06-08 | 2022-06-08 | Highway subgrade settlement monitoring and predicting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115419120B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116050624A (en) * | 2023-02-03 | 2023-05-02 | 广西海川路桥建设有限公司 | Comprehensive monitoring method and system for highway construction |
| CN117970285A (en) * | 2024-03-28 | 2024-05-03 | 湖南省交通科学研究院有限公司 | Roadbed settlement monitoring method based on airborne laser radar |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102425148A (en) * | 2011-09-02 | 2012-04-25 | 铁道第三勘察设计院集团有限公司 | Rapid sub-grade settlement predicting method based on static sounding and BP (Back Propagation) neural network |
| CN102926368A (en) * | 2012-10-31 | 2013-02-13 | 西安理工大学 | Device and method for monitoring differential settlement of roadbed of road |
| CN103334462A (en) * | 2013-07-22 | 2013-10-02 | 山东大学 | Conductive polymer-based soil deformation monitoring system and method |
| CN108897010A (en) * | 2018-06-21 | 2018-11-27 | 中国化学工程第六建设有限公司 | Ground based method, system and storage medium are dug using the measurement of GNSS measuring system |
| CN109829577A (en) * | 2019-01-17 | 2019-05-31 | 北京交通大学 | Track train operating status prediction technique based on deep neural network structural model |
| CN110631547A (en) * | 2019-09-26 | 2019-12-31 | 成都理工大学 | Roadbed monitoring and early warning system based on Internet of things |
| JP2021008729A (en) * | 2019-06-28 | 2021-01-28 | 大和ハウス工業株式会社 | Ground subsidence prediction system |
| CN113157030A (en) * | 2021-05-07 | 2021-07-23 | 西安石油大学 | Greenhouse temperature and humidity monitoring and self-adaptive adjusting system based on machine learning |
-
2022
- 2022-06-08 CN CN202210652439.7A patent/CN115419120B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102425148A (en) * | 2011-09-02 | 2012-04-25 | 铁道第三勘察设计院集团有限公司 | Rapid sub-grade settlement predicting method based on static sounding and BP (Back Propagation) neural network |
| CN102926368A (en) * | 2012-10-31 | 2013-02-13 | 西安理工大学 | Device and method for monitoring differential settlement of roadbed of road |
| CN103334462A (en) * | 2013-07-22 | 2013-10-02 | 山东大学 | Conductive polymer-based soil deformation monitoring system and method |
| CN108897010A (en) * | 2018-06-21 | 2018-11-27 | 中国化学工程第六建设有限公司 | Ground based method, system and storage medium are dug using the measurement of GNSS measuring system |
| CN109829577A (en) * | 2019-01-17 | 2019-05-31 | 北京交通大学 | Track train operating status prediction technique based on deep neural network structural model |
| JP2021008729A (en) * | 2019-06-28 | 2021-01-28 | 大和ハウス工業株式会社 | Ground subsidence prediction system |
| CN110631547A (en) * | 2019-09-26 | 2019-12-31 | 成都理工大学 | Roadbed monitoring and early warning system based on Internet of things |
| CN113157030A (en) * | 2021-05-07 | 2021-07-23 | 西安石油大学 | Greenhouse temperature and humidity monitoring and self-adaptive adjusting system based on machine learning |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116050624A (en) * | 2023-02-03 | 2023-05-02 | 广西海川路桥建设有限公司 | Comprehensive monitoring method and system for highway construction |
| CN116050624B (en) * | 2023-02-03 | 2023-08-22 | 广西海川路桥建设有限公司 | Comprehensive monitoring method and system for highway construction |
| CN117970285A (en) * | 2024-03-28 | 2024-05-03 | 湖南省交通科学研究院有限公司 | Roadbed settlement monitoring method based on airborne laser radar |
| CN117970285B (en) * | 2024-03-28 | 2024-06-04 | 湖南省交通科学研究院有限公司 | Roadbed settlement monitoring method based on airborne laser radar |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115419120B (en) | 2024-07-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11333796B2 (en) | Spatial autocorrelation machine learning-based downscaling method and system of satellite precipitation data | |
| Yuan et al. | Estimating surface soil moisture from satellite observations using a generalized regression neural network trained on sparse ground-based measurements in the continental US | |
| CN115419120A (en) | Highway subgrade settlement monitoring and predicting method | |
| Kiany et al. | Evaluation of the TRMM 3B42 product for extreme precipitation analysis over southwestern Iran | |
| Muneer et al. | Modeling of runoff in the arid regions using remote sensing and geographic information system (GIS) | |
| CN111401602A (en) | Assimilation method for satellite and ground rainfall measurement values based on neural network | |
| CN106021868A (en) | Multi-rule algorithm-based remote sensing data downscaling method | |
| CN113554849A (en) | Air-ground monitoring slope system and instability risk assessment method thereof | |
| CN109886456B (en) | Debris flow prediction method based on PCA and mixed kernel function LSSVR | |
| CN105550423A (en) | CMORPH satellite precipitation data downscaling method based on Fuzzy-OLS (Ordinary Least Squares) | |
| CN103778305A (en) | k-VNN- and LS-SVM-based modelling method for icing of electric transmission line | |
| CN105528523A (en) | Soil thickness inversion method based on remote sensing data | |
| CN114663840B (en) | Tunnel environment inspection equipment monitoring method and system | |
| CN114858238A (en) | Underground water storage and transformation monitoring and early warning system | |
| CN115980745A (en) | Underground mining subsidence area intelligent identification method and device, electronic equipment and storage medium | |
| CN115204032A (en) | ENSO prediction method and device based on multi-channel intelligent model | |
| CN117893362B (en) | Multi-time-space-scale offshore wind power characteristic screening and enhanced power prediction method | |
| CN115099416A (en) | Visual post-explanation method of integrated hydrological meteorological machine learning forecasting model | |
| CN115270608A (en) | Coastal zone ground settlement prediction method based on ARIMA and LSTM | |
| Su et al. | An approach using multi-factor combination to evaluate high rocky slope safety | |
| Kaloop et al. | A novel AI approach for modeling land surface temperature of Freetown, Sierra Leone, based on land-cover changes | |
| CN115392128B (en) | Method for simulating river basin runoff by utilizing space-time convolution LSTM network | |
| CN115906640A (en) | Power transmission line soil erosion grade regression method based on deep learning model | |
| CN114912075A (en) | Soil humidity data missing filling method fusing spatial-temporal three-dimensional information of site and satellite remote sensing observation | |
| Peddinti et al. | Efficient pavement monitoring for south korea using unmanned aerial vehicles |
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 |