CN114855847B - Intelligent control method for construction site foundation pit dewatering system - Google Patents
Intelligent control method for construction site foundation pit dewatering system Download PDFInfo
- Publication number
- CN114855847B CN114855847B CN202210785079.8A CN202210785079A CN114855847B CN 114855847 B CN114855847 B CN 114855847B CN 202210785079 A CN202210785079 A CN 202210785079A CN 114855847 B CN114855847 B CN 114855847B
- Authority
- CN
- China
- Prior art keywords
- vector
- target
- sample
- loss function
- foundation pit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- E—FIXED CONSTRUCTIONS
- E02—HYDRAULIC ENGINEERING; FOUNDATIONS; SOIL SHIFTING
- E02D—FOUNDATIONS; EXCAVATIONS; EMBANKMENTS; UNDERGROUND OR UNDERWATER STRUCTURES
- E02D19/00—Keeping dry foundation sites or other areas in the ground
- E02D19/06—Restraining of underground water
- E02D19/10—Restraining of underground water by lowering level of ground water
-
- E—FIXED CONSTRUCTIONS
- E03—WATER SUPPLY; SEWERAGE
- E03B—INSTALLATIONS OR METHODS FOR OBTAINING, COLLECTING, OR DISTRIBUTING WATER
- E03B3/00—Methods or installations for obtaining or collecting drinking water or tap water
- E03B3/32—Methods or installations for obtaining or collecting drinking water or tap water with artificial enrichment, e.g. by adding water from a pond or a river
- E03B3/34—Methods or installations for obtaining or collecting drinking water or tap water with artificial enrichment, e.g. by adding water from a pond or a river of underground water
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/15—Correlation function computation including computation of convolution operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A20/00—Water conservation; Efficient water supply; Efficient water use
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Hydrology & Water Resources (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Environmental & Geological Engineering (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- Software Systems (AREA)
- Algebra (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Mining & Mineral Resources (AREA)
- Paleontology (AREA)
- Civil Engineering (AREA)
- Structural Engineering (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Investigation Of Foundation Soil And Reinforcement Of Foundation Soil By Compacting Or Drainage (AREA)
Abstract
The invention relates to an intelligent control method for a foundation pit dewatering system of a construction site, and belongs to the technical field of control or regulation. The method comprises the following steps: inputting the sample comprehensive vector set into a fully-connected neural network, and training the network by using a first loss function, a second loss function, a third loss function, a weight corresponding to the first loss function and a weight corresponding to the second loss function to obtain a target network corresponding to a trained target foundation pit; and inputting the reference comprehensive vector into a target network, and adjusting parameters of each dewatering well water pump and each recharging well water pump during construction of the target foundation pit according to a result output by the target network. According to the control method, the pumping rate of the dewatering well water pump and the irrigation rate of the recharging well water pump are controlled, so that the pumping amount and the irrigation amount are approximately the same when the whole dewatering system operates stably, the stability of the underground water level can be well guaranteed, and the risk of settlement of buildings around the foundation pit is low.
Description
Technical Field
The invention relates to the technical field of control or regulation, in particular to an intelligent control method for a foundation pit dewatering system of a construction site.
Background
Along with the development of urban construction, more and more municipal building engineering is required, when the construction engineering project construction is carried out in an urban area, a soil pit needs to be excavated at a construction design position according to the base elevation and the base plane size, the soil pit is a foundation pit, and after the foundation pit is excavated, water may be accumulated in the foundation pit due to the underground water level; and in order to guarantee engineering quality, prevent the appearance of accidents such as foundation ditch side slope collapse, just need carry out precipitation in to the foundation ditch, if the soil infiltration rate of foundation ditch position is higher, generally can bore and dig precipitation well around the foundation ditch, the passageway is continuously drawn water from the precipitation well and is reached and reduce underground water level in the foundation ditch and make the purpose that the foundation ditch bottom does not outwards ooze water, but if there is the building crowd around the foundation ditch, it can make underground water level take place great change to continue to extract groundwater from the precipitation well, and then perhaps cause the foundation ditch around the foundation ditch to take off bearing capacity reduction after the dehydration, then compressed under the action of gravity, the consolidation, and then form the settlement accident. Therefore, in order to prevent such accidents, a recharge well is drilled and excavated between the nearby building and the foundation pit dewatering well and at a position close to the building, so as to recharge the groundwater and ensure the stability of the groundwater level near the building around the foundation pit.
The existing foundation pit dewatering control method is generally realized by monitoring the underground water level, and the method mainly comprises the steps that when a water pump pumps underground water from a dewatering well, if the underground water level in a recharging well is found to be reduced, the recharging well water pump is controlled to recharge the recharging well; if the underground water level in the recharging well is not reduced, stopping recharging the recharging well by the recharging well water pump; although the stability of the underground water level can be ensured to a certain extent by the control mode, the timeliness of maintaining the stability of the underground water level by monitoring the change of the underground water level in the recharging well is poor; because the reduction in groundwater level is related to permeability, when it can be monitored from the recharge well, the groundwater level may have actually dropped elsewhere for a period of time during which there is a risk of settling, and thus the foundation pit precipitation control method is less reliable.
Disclosure of Invention
The invention provides an intelligent control method for a foundation pit dewatering system of a construction site, which is used for solving the problem that the existing method is low in reliability of foundation pit dewatering control, and adopts the following technical scheme:
the invention provides an intelligent control method of a foundation pit dewatering system of a construction site, which comprises the following steps:
acquiring a construction design drawing corresponding to a target foundation pit; the construction design drawing comprises positions of dewatering wells, positions of recharging wells, the edge of a target foundation pit area and the edge of each neighborhood building area; acquiring a sample comprehensive vector set corresponding to a target foundation pit; the sample comprehensive vector is obtained by splicing a sample pumping rate vector and a sample irrigation rate vector;
inputting the sample comprehensive vector set into a fully-connected neural network, and training the network by using a first loss function, a second loss function, a third loss function, a weight corresponding to the first loss function and a weight corresponding to the second loss function to obtain a target network corresponding to a trained target foundation pit;
acquiring a reference comprehensive vector for constructing the target foundation pit, which is given by related workers; and inputting the reference comprehensive vector into a target network, and adjusting parameters of each dewatering well water pump and each recharging well water pump during construction of the target foundation pit according to a result output by the target network.
Has the advantages that: inputting a sample comprehensive vector set into a fully-connected neural network, and training the network by using a first loss function, a second loss function, a third loss function, a weight corresponding to the first loss function and a weight corresponding to the second loss function to obtain a target network corresponding to a trained target foundation pit; then acquiring a reference comprehensive vector for constructing the target foundation pit, which is given by related workers; and inputting the reference comprehensive vector into a target network, and adjusting parameters of each dewatering well water pump and each recharging well water pump during construction of the target foundation pit according to a result output by the target network. The control method provided by the invention ensures that the water pumping quantity and the water irrigation quantity are approximately the same when the whole dewatering system operates stably by controlling the water pumping rate of the dewatering well water pump and the water irrigation rate of the recharging well water pump, thereby better ensuring the stability of the underground water level; and the method also controls the pumping rate of the dewatering well water pump and the irrigation rate of the recharging well water pump to control the position with changed water level in situ at a place far away from the building around the foundation pit as much as possible, so that the risk of the building around the foundation pit sinking is low.
Preferably, the method for obtaining the sample comprehensive vector set corresponding to the target foundation pit includes:
acquiring each pumping speed of each dewatering well water pump and each recharging speed of each recharging well water pump; the pumping speed and the water irrigation speed are randomly set;
randomly selecting any one pumping rate of each dewatering well water pump, and constructing to obtain a pumping rate vector of each sample; the number of parameters in the sample pumping rate vector is the same as the number of precipitation wells;
randomly selecting any one irrigation speed of each recharging well water pump, and constructing to obtain an irrigation speed vector of each sample; the quantity of parameters in the sample irrigation rate vector is the same as that of the recharge wells;
randomly selecting any sample pumping rate vector and any sample irrigation rate vector for splicing to obtain an anisotropic quantity after preset splicing times, and recording as a sample comprehensive vector; splicing once to obtain a sample comprehensive vector; and constructing a sample comprehensive vector set according to the sample comprehensive vectors.
Preferably, the first loss function is:
inputting any sample comprehensive vector into a fully-connected neural network to obtain a target comprehensive vector corresponding to the sample comprehensive vector output by the network; obtaining a target pumping rate vector and a target irrigation rate vector corresponding to the target comprehensive vector according to the target comprehensive vector corresponding to the sample comprehensive vector;
the first loss function corresponding to the sample synthesis vector is:
wherein the content of the first and second substances,
is a first loss function corresponding to the sample synthetic vector, M is the number of parameters in a target pumping rate vector corresponding to the target synthetic vector, N is the number of parameters in a target watering rate vector corresponding to the target synthetic vector,
for the value of the ith parameter in the target pumping rate vector,
the value of the jth parameter in the target-filling rate vector,
and the distance between the precipitation well position corresponding to the ith parameter in the target pumping rate vector and the recharging well position corresponding to the jth parameter in the target recharging rate vector is obtained.
Preferably, after any sample synthesis vector is input into the network, the second loss function corresponding to the sample synthesis vector is:
wherein the content of the first and second substances,
and synthesizing a second loss function corresponding to the vector for the sample.
Preferably, after any sample synthesis vector is input into the network, the third loss function corresponding to the sample synthesis vector is:
wherein the content of the first and second substances,
a third loss function corresponding to the sample synthesis vector,
the value of the ith parameter in the sample pumping rate vector corresponding to the sample comprehensive vector,
and filling the value of the jth parameter in the velocity vector for the sample corresponding to the sample comprehensive vector.
Preferably, the method for obtaining the weight corresponding to the first loss function and the weight corresponding to the second loss function includes:
calculating the mean value of the distance sum between any one dewatering well and any one recharging well according to the positions of the dewatering wells and the positions of the recharging wells, and recording the mean value as a first mean value;
calculating the mean value of the minimum distance sum between the edge of the target foundation pit region and the edge of each neighborhood building region according to the edge of the target foundation pit region and the edge of each neighborhood building region, and recording the mean value as a second mean value;
obtaining a weight value corresponding to a second loss function according to the first average value and the second average value;
and the difference value of subtracting the weight value corresponding to the second loss function from 1 is recorded as the weight value corresponding to the first loss function.
Preferably, the weight value corresponding to the second loss function is calculated according to the following formula:
wherein the content of the first and second substances,
is the weight value corresponding to the second loss function,
is the first average value of the first average value,
is the second mean value.
Drawings
To more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the following description will be made
While the drawings necessary for the embodiment or prior art description are briefly described, it should be apparent that the drawings in the following description are merely examples of the invention and that other drawings may be derived from those drawings by those of ordinary skill in the art without inventive step.
FIG. 1 is a flow chart of an intelligent control method for a construction site foundation pit dewatering system according to the invention;
FIG. 2 is a schematic diagram of the original groundwater level of the present invention without pumping water and irrigating water;
FIG. 3 is a schematic diagram of the variation of groundwater level for pumping water and irrigating according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying 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, rather than all embodiments, and all other embodiments obtained by those skilled in the art based on the embodiments of the present invention belong to the protection scope of the embodiments of the present invention.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The embodiment provides an intelligent control method for a foundation pit dewatering system of a construction site, which is described in detail as follows:
as shown in fig. 1, the intelligent control method for the construction site foundation pit dewatering system comprises the following steps:
s001, acquiring a construction design drawing corresponding to a target foundation pit; the construction design drawing comprises positions of precipitation wells, positions of recharge wells, the edge of a target foundation pit area and the edge of each neighborhood building area; acquiring a sample comprehensive vector set corresponding to a target foundation pit; and the sample comprehensive vector is obtained by splicing a sample pumping rate vector and a sample irrigation rate vector.
The method mainly achieves the purpose of adjusting the reference data for construction of the target foundation pit, which is given by related workers, according to the trained network, so that the pumping speed of each dewatering well water pump and the irrigation speed of each recharging well water pump, which are set when the target foundation pit is constructed, can ensure the stability of the underground water level, and further, the risk of settlement of buildings around the foundation pit is smaller; therefore, in the intelligent control method for the foundation pit dewatering system of the construction site provided by this embodiment, the sample comprehensive vector set is input into the fully-connected neural network, and the network is trained by using the first loss function, the second loss function, the third loss function, the weight corresponding to the first loss function and the weight corresponding to the second loss function, so as to obtain a target network corresponding to a trained target foundation pit; then inputting a reference comprehensive speed vector for constructing the target foundation pit, which is given by related workers, into a trained target network, and outputting the target comprehensive speed vector as the target comprehensive speed vector during construction of the target foundation pit by the network; and then adjusting parameters of each dewatering well water pump and each recharging well water pump during construction of the target foundation pit according to the result output by the target network. According to the control method provided by the embodiment, the pumping rate of the dewatering well water pump and the irrigation rate of the recharging well water pump are controlled, so that the pumping amount and the irrigation amount are approximately the same when the whole dewatering system operates stably, and the stability of the underground water level can be better ensured; and the method also controls the pumping rate of the dewatering well water pump and the irrigation rate of the recharging well water pump to control the position with changed water level in situ at a place far away from the building around the foundation pit as much as possible, so that the risk of the building around the foundation pit sinking is low.
In the embodiment, only one foundation pit is analyzed and marked as a target foundation pit; firstly, obtaining a construction design drawing corresponding to a target foundation pit; marking the positions of the recharging wells, the positions of the dewatering wells, the edges of the target foundation pit areas and the edges of the areas of the neighborhood buildings on the construction design drawing, wherein the neighborhood buildings refer to buildings around the target foundation pit, the number of the recharging wells and the number of the dewatering wells are different under general conditions, the number of the dewatering wells is marked as M, and the number of the recharging wells is marked as N; the recharge well is typically located near a neighborhood building; the positions of each recharging well and each dewatering well are respectively provided with a water pump for pumping water from the dewatering wells and irrigating water from the recharging wells; then, related personnel randomly set water pump speed parameters of the water pumps of all well points for many times, so that the pumping speeds of all dewatering well water pumps and the water filling speeds of all recharging well water pumps can be obtained; then randomly selecting any one pumping rate of each dewatering well water pump, recording the pumping rate as a target pumping rate, and constructing and obtaining a pumping rate vector of the dewatering well water pump corresponding to the target foundation pit according to the target pumping rate of each dewatering well water pump, and recording the pumping rate vector as a sample pumping rate vector; after multiple random selections are carried out, the pumping rate vectors of all samples corresponding to the target foundation pit can be constructed and obtained. Similarly, randomly selecting any one irrigation speed of each recharging well water pump, recording the irrigation speed as a target irrigation speed, and constructing and obtaining a recharging well water pump irrigation speed vector corresponding to the target foundation pit according to the target irrigation speed of each recharging well water pump, and recording the vector as a sample irrigation speed vector; after multiple random selections are carried out, the irrigation rate vector of each sample corresponding to the target foundation pit can be constructed and obtained. Randomly selecting any sample pumping rate vector and any sample watering rate vector for splicing to obtain a spliced vector, and recording the spliced vector as a sample comprehensive vector, wherein the sample comprehensive vector is an M + N-dimensional vector; therefore, after multiple splicing, a sample comprehensive vector set can be obtained; the sample comprehensive vector set is input for subsequent network training; the comprehensive sample vector set can be understood as a setting control scheme of water pump water pumping rate parameters in a foundation pit dewatering system, and the comprehensive sample vector set can be understood as a setting control scheme of water pump water pumping rate parameters in a plurality of foundation pit dewatering systems.
And S002, inputting the sample comprehensive vector set into a fully-connected neural network, and training the network by using the first loss function, the second loss function, the third loss function, the weight corresponding to the first loss function and the weight corresponding to the second loss function to obtain a target network corresponding to the trained target foundation pit.
In this embodiment, an auto-supervised neural network is first constructed, the network structure of the auto-supervised neural network is a fully-connected neural network (FC network), a fully-connected layer is 5 layers, and the network input is a sample integrated vector, which corresponds to a sample pumping rate vector and a sample watering rate vector because the sample integrated vector is an M + N-dimensional vector; the network output is also a vector with M + N dimensions and is recorded as a target comprehensive vector; the input data form and the output data form of the network are the same, so that the network output vector corresponding to the sample pumping rate vector corresponding to the sample comprehensive vector can be obtained and recorded as a target pumping rate vector; network output vectors corresponding to the sample irrigation rate vectors corresponding to the sample comprehensive vectors can also be obtained and recorded as target irrigation rate vectors; target comprehensive vectors corresponding to the sample comprehensive vectors output by the network can be obtained by splicing the target pumping rate vectors and the target irrigation rate vectors; the dewatering wells and the recharging wells corresponding to the parameters in the target comprehensive vector correspond to the dewatering wells and the recharging wells corresponding to the parameters in the reference comprehensive vector one by one; and optimizing the parameters of the network by using a random gradient descent method, and preventing the network from generating overfitting by using a regularization method. For example, if any sample comprehensive vector is input into the network, where the a-th parameter in the sample comprehensive vector is data of the i-th dewatering well water pump corresponding to the target foundation pit, the a-th parameter in the target comprehensive vector output by the network is also data of the i-th dewatering well water pump corresponding to the target foundation pit. The specific training process of the fully-connected neural network comprises the following steps:
(a) the process of constructing the first loss function is:
for any dewatering well and any recharging well corresponding to the target foundation pit, the original underground water level is as shown in fig. 2 when water pumping and water recharging are not carried out, namely the heights of the original underground water level at all positions tend to be the same when water pumping and water recharging are not carried out; however, when water is pumped from the dewatering well and water is poured into the recharging well at the same time, the groundwater level changes as shown in fig. 3, that is, due to the underground infiltration, funnel-shaped and inverted funnel-shaped water level layers are formed at the dewatering well and the recharging well respectively, and when the watering rate is the same as the water pumping rate, the groundwater level corresponding to the midpoint between the two wells is ideally as high as the original groundwater level; the dotted lines in fig. 2 and 3 are groundwater table curves, the groundwater table curves when water pumping and irrigation are not performed are recorded as original groundwater table curves, and the groundwater table curves when water pumping and irrigation are performed simultaneously are recorded as characteristic groundwater table curves. If the irrigation rate of the recharge well is greater than the pumping rate of the dewatering well, the position, which is as high as the original underground water level curve, in the characteristic underground water level curve can deviate towards the dewatering well, otherwise, the position deviates towards the recharge well, and the deviation degree is larger if the rate difference is larger; however, in the embodiment, it is desired to control the position of the characteristic groundwater level layer curve, which is as high as the original groundwater level curve, to deviate to the precipitation well, because when the position of the characteristic groundwater level layer curve, which is as high as the original groundwater level curve, deviates to the precipitation well, the width of the water level layer in the precipitation well, which is in a funnel shape, is reduced, the influence of the precipitation well on the surrounding groundwater level when pumping water can be reduced, and further, the settlement of other buildings near the recharge well is prevented, that is, the settlement of each neighborhood building corresponding to the target foundation pit is prevented; therefore, in the embodiment, an underground water level movement loss function is constructed based on the output vector of the sample comprehensive vector, the positions of the dewatering wells and the positions of the recharging wells and is recorded as a first loss function; therefore, when any sample comprehensive vector is input into the network, a target comprehensive vector corresponding to the sample comprehensive vector output by the network can be obtained; obtaining a target pumping rate vector and a target irrigation rate vector of a target comprehensive vector corresponding to the sample comprehensive vector; therefore, the first loss function corresponding to the sample synthesis vector is:
wherein the content of the first and second substances,
is a first loss function corresponding to the sample synthetic vector, M is the number of parameters in a target pumping rate vector corresponding to the target synthetic vector, N is the number of parameters in a target watering rate vector corresponding to the target synthetic vector,
for the value of the ith parameter in the target pumping rate vector,
the value of the jth parameter in the target-filling rate vector,
and the distance between the precipitation well position corresponding to the ith parameter in the target pumping speed vector and the recharging well position corresponding to the jth parameter in the target recharging speed vector is obtained. The smaller the first loss function is, the better the network can learn how to control or set the well point water pumping and filling rate, so that the original underground water level deviates to the precipitation well as much as possible;
has a value range of [ -1, 1]，
Has a value range of [0, 2 ]]；
The moving deviation of the original underground water level during water pumping and irrigation can be reflected, but the representational characteristics of the moving deviation of the original underground water level during water pumping and irrigation reflected by precipitation wells and recharge wells with different distances are different, so that the moving deviation of the original underground water level during water pumping and irrigation is different
As
The influence weight of (c).
Therefore, the first loss function after the comprehensive vector of each sample is input into the network in the training process can be obtained through the method.
(b) The process of constructing the second loss function is:
when the water filling quantity is the same as the water pumping quantity, the change of the underground water level is small, and the water pumping speed and the water filling speed can reflect the water filling quantity and the water pumping quantity; therefore, the underground water quantity change loss function is constructed based on the target pumping rate vector and the target irrigation rate vector and is recorded as a second loss function; therefore, when any sample synthesis vector is input into the network, the second loss function corresponding to the sample synthesis vector is:
wherein the content of the first and second substances,
and synthesizing a second loss function corresponding to the vector for the sample. The smaller the second loss function is, the better the network can learn how to control or set the well point pumping and filling rate, and further the underground water volume change is minimized.
Therefore, the second loss function after the comprehensive vector of each sample is input into the network in the training process can be obtained through the method.
(c) The process of constructing the third loss function is:
in this embodiment, in order to minimize the adjustment cost, a third loss function needs to be constructed, where the minimum adjustment cost means that a difference between an output vector and an input vector is small; therefore, after any sample synthesis vector is input into the network, the third loss function corresponding to the sample synthesis vector is:
wherein the content of the first and second substances,
a third loss function corresponding to the sample synthesis vector,
the first of the water pumping rate vectors of the water pump of the sample dewatering well corresponding to the comprehensive vector of the sampleThe values of the i parameters are such that,
and the value of the jth parameter in the water pump irrigation speed vector of the sample recharging well corresponding to the sample comprehensive vector is obtained. The smaller the value of the third loss function is, the better the network can learn how to control or set the well point pumping and filling rate to minimize the adjustment cost.
Therefore, the third loss function after the comprehensive vector of each sample is input into the network in the training process can be obtained through the method.
Because the positions of the target foundation pit dewatering well and the recharging well near the neighborhood building have different attention degrees to the first loss function and the second loss function, if the distance between the recharging well and the dewatering well is generally short, the change loss (the second loss function) of the underground water quantity is more important, because the infiltration effect of the underground water between the dewatering well and the recharging well is small and the underground water level is increased or reduced more quickly if the distance between the dewatering well and the recharging well is short; and if the target foundation pit is close to the adjacent buildings, the underground water level movement loss (first loss function) is more concerned so as to ensure that the adjacent buildings are not influenced by the sedimentation effect as much as possible. Therefore, in order to achieve the purpose of minimizing the change of the underground water level and enabling the original underground water level to deviate towards the precipitation well as far as possible, the control of the pumping speed of the precipitation well and the recharge well pump needs to set attention weights for two loss terms. The method specifically comprises the following steps:
calculating the mean value of the distance sum between any one dewatering well and any one recharging well according to the positions of the dewatering wells and the positions of the recharging wells, and recording the mean value as a first mean value; calculating the mean value of the minimum distance sum between the edge of the target foundation pit region and the edge of each neighborhood building region, and recording as a second mean value; obtaining a weight value corresponding to the second loss function according to the first average value and the second average value; calculating a weight value corresponding to the second loss function according to the following formula:
wherein the content of the first and second substances,
is the weight value corresponding to the second loss function,
is the first average value of the first average value,
is the second mean value. And the difference value of subtracting the weight value corresponding to the second loss function from 1 is recorded as the weight value corresponding to the first loss function.
Therefore, according to the first loss function, the second loss function, the third loss function, the weight corresponding to the first loss function and the weight corresponding to the second loss function after the comprehensive vector of each sample is input into the network in the training process, the comprehensive loss function corresponding to the comprehensive vector of each sample in the training process is obtained; after any sample comprehensive vector is input into the network, the comprehensive loss function corresponding to the sample comprehensive vector is as follows:
wherein, the first and the second end of the pipe are connected with each other,
for the synthetic loss function corresponding to the sample synthetic vector,
and the weight value of the first loss function corresponding to the sample comprehensive vector is obtained.
S003, acquiring a reference comprehensive vector for constructing the target foundation pit, which is given by a relevant worker; and inputting the reference comprehensive vector into a target network, and adjusting parameters of each dewatering well water pump and each recharging well water pump during construction of the target foundation pit according to a result output by the target network.
The method comprises the steps of adjusting parameters of each dewatering well and parameters of each recharging well during construction of a target foundation pit according to a result output by a target network; the method specifically comprises the following steps:
firstly, acquiring a reference comprehensive vector for constructing a target foundation pit, which is given by related workers, wherein the acquisition mode of the reference comprehensive vector is the same as that of the sample comprehensive vector, namely the reference comprehensive vector is obtained by splicing a reference pumping rate vector and a reference irrigation rate vector; the reference pumping rate vector and the reference irrigation rate vector are values set by related personnel, each parameter value in the reference pumping rate vector is a reference pumping rate of the precipitation well water pump corresponding to the corresponding parameter, each parameter value in the reference irrigation rate vector is a reference irrigation rate of the recharge well water pump corresponding to the corresponding parameter, and the reference pumping rate and the reference irrigation rate are set by a preliminary test person; one parameter in the reference pumping speed vector corresponds to one dewatering well, and one parameter in the reference irrigation speed vector corresponds to one recharging well. Inputting the reference comprehensive vector into a trained target network, outputting by the network to obtain a target comprehensive vector corresponding to the reference comprehensive vector, wherein the precipitation wells and the recharge wells corresponding to the parameters in the target comprehensive vector correspond to the precipitation wells and the recharge wells corresponding to the parameters in the reference comprehensive vector one by one; therefore, according to the target comprehensive vector, the target pumping rate of each dewatering well water pump and the target irrigation rate of each recharging well water pump corresponding to the target foundation pit can be obtained, and the target pumping rate and the target irrigation rate are final set values of the pumping rates of each dewatering well and each recharging well water pump during construction of the target foundation pit; therefore, the pumping speed of each dewatering well water pump and the irrigation speed of each recharging well water pump in the foundation pit dewatering process can be made to be the corresponding target pumping speed and target irrigation speed, and intelligent control of the foundation pit dewatering system is achieved.
Has the advantages that: in the embodiment, a sample comprehensive vector set is input into a fully-connected neural network, and the network is trained by using a first loss function, a second loss function, a third loss function, a weight corresponding to the first loss function and a weight corresponding to the second loss function, so as to obtain a target network corresponding to a trained target foundation pit; then acquiring a reference comprehensive vector for constructing the target foundation pit, which is given by related workers; and inputting the reference comprehensive vector into a target network, and adjusting parameters of each dewatering well water pump and each recharging well water pump during construction of the target foundation pit according to a result output by the target network. According to the control method provided by the embodiment, the pumping rate of the dewatering well water pump and the irrigation rate of the recharging well water pump are controlled, so that the pumping amount and the irrigation amount are approximately the same when the whole dewatering system operates stably, and the stability of the underground water level can be better ensured; and the method also controls the pumping rate of the dewatering well water pump and the irrigation rate of the recharging well water pump to control the position with changed water level in situ at a place far away from the building around the foundation pit as much as possible, so that the risk of the building around the foundation pit sinking is low.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.
Claims (2)
1. An intelligent control method for a foundation pit dewatering system of a construction site is characterized by comprising the following steps:
acquiring a construction design drawing corresponding to a target foundation pit; the construction design drawing comprises positions of precipitation wells, positions of recharge wells, the edge of a target foundation pit area and the edge of each neighborhood building area; acquiring a sample comprehensive vector set corresponding to a target foundation pit; the sample comprehensive vector is obtained by splicing a sample pumping rate vector and a sample irrigation rate vector;
inputting the sample comprehensive vector set into a fully-connected neural network, and training the network by using a first loss function, a second loss function, a third loss function, a weight corresponding to the first loss function and a weight corresponding to the second loss function to obtain a target network corresponding to a trained target foundation pit;
acquiring a reference comprehensive vector for constructing the target foundation pit, which is given by related workers; inputting the reference comprehensive vector into a target network, and adjusting parameters of each dewatering well water pump and each recharging well water pump during construction of a target foundation pit according to a result output by the target network;
the first loss function is:
inputting any sample comprehensive vector into a fully-connected neural network to obtain a target comprehensive vector corresponding to the sample comprehensive vector output by the network; obtaining a target pumping rate vector and a target irrigation rate vector corresponding to the target comprehensive vector according to the target comprehensive vector corresponding to the sample comprehensive vector;
the first loss function corresponding to the sample synthesis vector is:
wherein the content of the first and second substances,
is a first loss function corresponding to the sample synthetic vector, M is the number of parameters in a target pumping rate vector corresponding to the target synthetic vector, N is the number of parameters in a target watering rate vector corresponding to the target synthetic vector,
for the value of the ith parameter in the target pumping rate vector,
the value of the jth parameter in the target-filling rate vector,
the distance between the position of the precipitation well corresponding to the ith parameter in the target pumping rate vector and the position of the recharge well corresponding to the jth parameter in the target watering rate vector is obtained;
the method for obtaining the sample comprehensive vector set corresponding to the target foundation pit comprises the following steps:
acquiring each pumping speed of each dewatering well water pump and each recharging speed of each recharging well water pump; the pumping speed and the water irrigation speed are randomly set;
randomly selecting any one pumping rate of each dewatering well water pump, and constructing to obtain a pumping rate vector of each sample; the number of parameters in the sample pumping rate vector is the same as the number of precipitation wells;
randomly selecting any one irrigation speed of each recharging well water pump, and constructing to obtain an irrigation speed vector of each sample; the quantity of parameters in the sample irrigation rate vector is the same as that of the recharge wells;
randomly selecting any sample pumping rate vector and any sample irrigation rate vector for splicing to obtain an anisotropic quantity after preset splicing times, and recording as a sample comprehensive vector; splicing once to obtain a sample comprehensive vector; constructing a sample comprehensive vector set according to the sample comprehensive vectors;
after any sample comprehensive vector is input into the network, a second loss function corresponding to the sample comprehensive vector is as follows:
wherein the content of the first and second substances,
a second loss function corresponding to the sample synthesis vector;
after any sample comprehensive vector is input into the network, a third loss function corresponding to the sample comprehensive vector is as follows:
wherein the content of the first and second substances,
a third loss function corresponding to the sample synthesis vector,
the value of the ith parameter in the sample pumping rate vector corresponding to the sample comprehensive vector,
filling water for the value of the jth parameter in the sample irrigation rate vector corresponding to the sample comprehensive vector;
the method for obtaining the weight corresponding to the first loss function and the weight corresponding to the second loss function comprises the following steps:
calculating the mean value of the distance sum between any one dewatering well and any one recharging well according to the positions of the dewatering wells and the positions of the recharging wells, and recording the mean value as a first mean value;
calculating the mean value of the minimum distance sum between the edge of the target foundation pit region and the edge of each neighborhood building region according to the edge of the target foundation pit region and the edge of each neighborhood building region, and recording the mean value as a second mean value;
obtaining a weight value corresponding to a second loss function according to the first average value and the second average value;
and the difference value of subtracting the weight value corresponding to the second loss function from 1 is recorded as the weight value corresponding to the first loss function.
2. The intelligent control method for the foundation pit dewatering system of the construction site as claimed in claim 1, wherein the weight value corresponding to the second loss function is calculated according to the following formula:
wherein the content of the first and second substances,
is the weight value corresponding to the second loss function,
is the first average value of the first average value,
is the second mean value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210785079.8A CN114855847B (en) | 2022-07-06 | 2022-07-06 | Intelligent control method for construction site foundation pit dewatering system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210785079.8A CN114855847B (en) | 2022-07-06 | 2022-07-06 | Intelligent control method for construction site foundation pit dewatering system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114855847A CN114855847A (en) | 2022-08-05 |
| CN114855847B true CN114855847B (en) | 2022-09-20 |
Family
ID=82626753
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210785079.8A Active CN114855847B (en) | 2022-07-06 | 2022-07-06 | Intelligent control method for construction site foundation pit dewatering system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114855847B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102383412A (en) * | 2010-08-27 | 2012-03-21 | 中铁二十二局集团第三工程有限公司 | Construction method by adopting dewatering and water recharging to control sedimentation surrounding deep foundation pit |
| CN104899682A (en) * | 2015-05-19 | 2015-09-09 | 上海市建工设计研究院有限公司 | Evaluation method for construction risk of antiseepage waterproof curtain of deep foundation fit in coastal area |
| CN107989055A (en) * | 2017-11-01 | 2018-05-04 | 中国核工业第二二建设有限公司 | A kind of intelligence control system and control method for architectural engineering deep-well precipitation |
| CN111475884A (en) * | 2020-04-24 | 2020-07-31 | 福州大学 | Foundation pit dewatering optimization method based on particle swarm algorithm and groundwater model |
| CN113779835A (en) * | 2021-09-11 | 2021-12-10 | 浙江永欣联科信息科技股份有限公司 | AI and intelligent monitoring system based deep and large foundation pit safety early warning method |
-
2022
- 2022-07-06 CN CN202210785079.8A patent/CN114855847B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102383412A (en) * | 2010-08-27 | 2012-03-21 | 中铁二十二局集团第三工程有限公司 | Construction method by adopting dewatering and water recharging to control sedimentation surrounding deep foundation pit |
| CN104899682A (en) * | 2015-05-19 | 2015-09-09 | 上海市建工设计研究院有限公司 | Evaluation method for construction risk of antiseepage waterproof curtain of deep foundation fit in coastal area |
| CN107989055A (en) * | 2017-11-01 | 2018-05-04 | 中国核工业第二二建设有限公司 | A kind of intelligence control system and control method for architectural engineering deep-well precipitation |
| CN111475884A (en) * | 2020-04-24 | 2020-07-31 | 福州大学 | Foundation pit dewatering optimization method based on particle swarm algorithm and groundwater model |
| CN113779835A (en) * | 2021-09-11 | 2021-12-10 | 浙江永欣联科信息科技股份有限公司 | AI and intelligent monitoring system based deep and large foundation pit safety early warning method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114855847A (en) | 2022-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10831164B2 (en) | Optimized hydromodification management with active stormwater controls | |
| CN108549774B (en) | Method for determining fixed water level water pumping quantity of suspension type curtain foundation pit with heterogeneous confined water layer | |
| CN110387896B (en) | Foundation pit dewatering well design method | |
| CN107587516A (en) | A kind of efficient sealing of Follow-up type hinders sand precipitation hybrid system and its construction method | |
| CN110219277A (en) | The dam bursting flood analysis system of dam system | |
| CN116029232B (en) | Complex well type water invasion front prediction method for bottom water and gas reservoir | |
| CN115408886B (en) | Water lowering and draining system and method for controlling underground structure settlement through hydraulic engineering | |
| CN110232479A (en) | A kind of city flood control by reservoir regulation compensation optimizing dispatching method | |
| CN107916646A (en) | A kind of rain-flood resources amount of taping the latent power evaluation method | |
| CN114855847B (en) | Intelligent control method for construction site foundation pit dewatering system | |
| CN115907429A (en) | PSO algorithm-based combined overflow optimization control method and device | |
| JP2024086775A (en) | How to determine the capacity of the regulating reservoir | |
| CN106202911B (en) | River ecological flow groundwater supply water demand calculation method | |
| CN102680027B (en) | Method for calculating displacement and discharging time in dynamical precipitation process of pressure-load fully penetrating well or well group | |
| WO2023241211A1 (en) | Landscape channeling using autonomous robotic soil dredger | |
| CN115730739A (en) | Urban waterlogging prediction method based on LSTM neural network | |
| Addis et al. | Comparison of beaching slopes from two centrally discharging tailings storage facilities | |
| KR101970196B1 (en) | Method for Calculating Flux of Runoff Considering Storm Sewer Systems | |
| Donaldson | Rehabilitation and modernisation of irrigation schemes | |
| CN209227555U (en) | Wet row's Tailings Dam dry type discharge structure | |
| CN113806901A (en) | Method for optimizing rainwater pipe system designed based on rainwater management model and storage device | |
| CN207392415U (en) | A kind of automatic storage capacity regulating gate system of caisson type | |
| KR102579259B1 (en) | Intelligent rainwater management system | |
| Rastogi et al. | Influence of artificial recharge basin shapes on the mound underneath in phreatic aquifers | |
| CN109183816A (en) | Wet row's Tailings Dam dry type discharge structure and raising Tailings Dam effective storage method |
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 |