CN115234006A - Steel structure integral lifting and leveling method based on convolutional neural network and real-time monitoring - Google Patents
Steel structure integral lifting and leveling method based on convolutional neural network and real-time monitoring Download PDFInfo
- Publication number
- CN115234006A CN115234006A CN202210843217.3A CN202210843217A CN115234006A CN 115234006 A CN115234006 A CN 115234006A CN 202210843217 A CN202210843217 A CN 202210843217A CN 115234006 A CN115234006 A CN 115234006A
- Authority
- CN
- China
- Prior art keywords
- steel structure
- neural network
- strain
- lifting
- real
- 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
- 229910000831 Steel Inorganic materials 0.000 title claims abstract description 108
- 239000010959 steel Substances 0.000 title claims abstract description 108
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012544 monitoring process Methods 0.000 title claims abstract description 31
- 238000001514 detection method Methods 0.000 claims abstract description 50
- 238000006073 displacement reaction Methods 0.000 claims abstract description 26
- 230000008569 process Effects 0.000 claims abstract description 26
- 238000004364 calculation method Methods 0.000 claims abstract description 7
- 238000010276 construction Methods 0.000 claims description 22
- 230000001360 synchronised effect Effects 0.000 claims description 10
- 238000012549 training Methods 0.000 claims description 9
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000009434 installation Methods 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 230000009471 action Effects 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 239000000463 material Substances 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 8
- 230000004913 activation Effects 0.000 description 6
- 238000011176 pooling Methods 0.000 description 5
- 230000001537 neural effect Effects 0.000 description 4
- 238000004088 simulation Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000003466 welding Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013529 biological neural network Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000011900 installation process Methods 0.000 description 1
- 238000010422 painting Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- E—FIXED CONSTRUCTIONS
- E04—BUILDING
- E04G—SCAFFOLDING; FORMS; SHUTTERING; BUILDING IMPLEMENTS OR AIDS, OR THEIR USE; HANDLING BUILDING MATERIALS ON THE SITE; REPAIRING, BREAKING-UP OR OTHER WORK ON EXISTING BUILDINGS
- E04G21/00—Preparing, conveying, or working-up building materials or building elements in situ; Other devices or measures for constructional work
- E04G21/14—Conveying or assembling building elements
-
- E—FIXED CONSTRUCTIONS
- E04—BUILDING
- E04G—SCAFFOLDING; FORMS; SHUTTERING; BUILDING IMPLEMENTS OR AIDS, OR THEIR USE; HANDLING BUILDING MATERIALS ON THE SITE; REPAIRING, BREAKING-UP OR OTHER WORK ON EXISTING BUILDINGS
- E04G21/00—Preparing, conveying, or working-up building materials or building elements in situ; Other devices or measures for constructional work
- E04G21/14—Conveying or assembling building elements
- E04G21/16—Tools or apparatus
-
- E—FIXED CONSTRUCTIONS
- E04—BUILDING
- E04G—SCAFFOLDING; FORMS; SHUTTERING; BUILDING IMPLEMENTS OR AIDS, OR THEIR USE; HANDLING BUILDING MATERIALS ON THE SITE; REPAIRING, BREAKING-UP OR OTHER WORK ON EXISTING BUILDINGS
- E04G21/00—Preparing, conveying, or working-up building materials or building elements in situ; Other devices or measures for constructional work
- E04G21/14—Conveying or assembling building elements
- E04G21/16—Tools or apparatus
- E04G21/18—Adjusting tools; Templates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Architecture (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Mechanical Engineering (AREA)
- Civil Engineering (AREA)
- Structural Engineering (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- Biophysics (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Computer Hardware Design (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Conveying And Assembling Of Building Elements In Situ (AREA)
Abstract
The invention discloses a steel structure integral lifting leveling method based on a convolutional neural network and real-time monitoring, which comprises the following steps of: preparing an upper chord member, a lower chord member and a web member of a steel structure, assembling the upper chord member, the lower chord member and the web member on a jig frame to be spliced into a whole, selecting a plurality of hoisting points, and performing integral hoisting by adopting a vertical hoisting mode; selecting a plurality of strain detection points in a steel structure; installing a strain gauge on the strain detection point, wherein the strain gauge is used for acquiring the strain of the structural rod piece; in the lifting process, the strain value of each strain detection point is input into a trained VGG convolutional neural network, the relative vertical displacement of each lifting point is obtained through calculation, the displacement stroke of each hydraulic lifter is output and adjusted, and the position of each lifting point is guaranteed to be on the same level; the invention can monitor the lifting state of the steel structure in real time and then correctly adjust the displacement stroke of each hydraulic lifter, so that the lifted steel structure is always in the optimal stress state.
Description
Technical Field
The invention relates to the technical field of integral lifting of a steel structure of a high-rise or super high-rise building, in particular to a steel structure integral lifting leveling method based on a convolutional neural network and real-time monitoring.
Background
The building shapes of multi-tower interconnection, large-span cantilevers and the like of high-rise or super high-rise buildings are usually completed by adopting a steel structure technology, and the conventional construction mode of single-piece hoisting and high-altitude welding has the difficulties and disadvantages of complex operation, poor safety, delayed construction period and the like. At present, for a large-volume steel structure truss system, a construction unit usually adopts a construction process of firstly assembling the ground and then integrally lifting, and the construction method has the advantages of convenience in operation, high safety, short construction period and the like.
The integral lifting technology of the large-volume steel structure truss system generally adopts a plurality of hydraulic lifters as lifting machines and flexible steel strands as bearing rigging. The hydraulic lifter is of a core-through structure, takes the steel strand as a lifting rigging, and has a series of unique advantages of safety, reliability, light weight of a bearing part, convenient transportation and installation, unnecessary splicing in the middle and the like.
In the whole lifting process, hydraulic lifting workers use a computer to control the displacement stroke of each lifter, and due to the reasons of initial ground height deviation, lifting stroke deviation and the like, the displacement stroke given by a lifting control computer can not ensure that lifting points are at the same height, so that the lifted steel truss is not in an ideal horizontal state, and the stress of components is complex.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, and provides a steel structure integral lifting and leveling method based on a convolutional neural network and real-time monitoring, which can monitor the lifting state of a steel structure in real time and then correctly adjust the displacement stroke of each hydraulic lifter, so that the lifted steel structure is always in the best horizontal stress state, and the safety and the working efficiency of the construction process are improved.
In order to achieve the aim, the invention provides a convolution neural network and real-time monitoring based steel structure integral lifting leveling method, which comprises the following steps:
step S1: preparing and assembling an upper chord member, a lower chord member and a web member of a steel structure, and assembling on the ground of an installation position;
step S2: the upper chord member, the lower chord member and the web members are all arranged on the jig frame and spliced into a whole, a plurality of hoisting points are selected, and the whole hoisting is carried out in a vertical hoisting mode;
and step S3: selecting a plurality of strain detection points for detecting deformation in a steel structure;
and step S4: installing a strain gauge on the strain detection point, wherein the strain gauge is used for acquiring the strain of the structural rod piece;
step S5: in the lifting process of the steel structure, the strain value of each strain detection point is input into the trained VGG convolutional neural network, the relative vertical displacement of each lifting point is obtained through calculation, the displacement stroke of each hydraulic lifter is output and adjusted, and the position of each lifting point is guaranteed to be on the same level.
Preferably, the step S5 further includes performing computer centralized control on the strain gauge and the hydraulic lifters, independently monitoring each hydraulic lifter in the system in real time through data feedback, importing the collected data into a trained VGG convolutional neural network model, and calculating a relative displacement difference between each hoisting point in real time, so as to control and level the steel structure.
Preferably, a hoisting model of a steel structure is established through SAP2000, the type and the material of each rod piece of the hoisting model are synchronous to the real structure, displacement loads of 5mm, 10mm, 15mm, 20mm, 25mm and 30mm are respectively applied to the other three hoisting points by taking one hoisting point as a reference, a plurality of different strain detection points are selected from the upper chord, the lower chord and the web member, the positions of the strain detection points are selected to be consistent with the strain detection points in the step S3, and strain values of the positions of the strain detection points obtained by simulation in the hoisting model are used as a data set to train the VGG convolutional neural network.
Preferably, in a trained VGG convolutional neural network, the strain data of each strain detection point obtained by actual measurement is input, and the relative height difference of hoisting points is obtained through data processing, so that the stroke adjustment of the hydraulic lifter can be performed in a targeted manner.
Preferably, the steel structure is hoisted by adopting a plurality of hydraulic hoists fixedly arranged on the hoisting beam, the flexible steel strand is adopted as a bearing rigging, a reasonable bearing hoisting point is selected on the steel structure, and the hoisting height is not limited.
Preferably, the strain state of the steel structure is real-time monitoring data of a plurality of strain detection points, the strain detection points are arranged on a plurality of rod pieces of the integral structure, and the timeliness of the detection data and the availability of the data are ensured.
Preferably, the training accuracy of the VGG convolutional neural network training model meets requirements, and the number of convolutional layers is not too shallow or too deep to cause under-fitting or over-fitting.
Preferably, the number of the strain detection points is not less than 25, and in the calculation of the VGG convolutional neural network, too small capacity results in larger model training error and under-fitting.
Preferably, the lifting beam has a condition of fixedly mounting a plurality of hydraulic lifters, and both sides of the truss of the steel structure are connected with the tower stiff structure.
Preferably, the hydraulic lifter is driven by a hydraulic circuit, the acceleration is extremely low in the action process, and no additional dynamic load is applied to the steel structure.
Compared with the prior art, the invention has the beneficial effects that:
according to the invention, a technology of monitoring the upper chord member, the lower chord member and the web members in real time is adopted, if the lifting heights of hydraulic lifters of lifting points are different in the lifting process of a steel structure, the stress change information of strain detection points can be fed back in real time through the measurement of a strain gauge, the detection data is input into a trained VGG convolutional neural network to obtain the relative vertical displacement of each lifting point, the displacement stroke of each hydraulic lifter is controlled through output, accurate regulation and control are realized, individual independent regulation and control are realized, the excessive deformation of the steel structure in the lifting process is reduced, the potential safety hazard is caused, and the safe lifting is ensured; and the lifted steel structure is always in the horizontal optimal stress position; in addition, by using the VGG convolutional neural network, the network depth is suitable, the convolutional kernel is large, the parameters are fewer, the method is suitable for complex modes, and when the hoisting steel structure is complex, the number of internal measuring points is large, and the data to be processed is huge, the excellent performance can be embodied.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic diagram of the overall modeling structure of a steel structure provided by the invention;
fig. 2 is a schematic structural diagram of a convolutional neural network provided by the present invention.
The figure comprises the following components:
11. strain detection points; 12. an upper chord; 13. a lower chord; 14. a web member; 15. hoisting points; 21. convolutional layer + RELU activation function; 22. a pooling layer; 23. fully connected layer + RELU activation function; 24. the softmax function outputs the layer.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are one embodiment of the present invention, and not all embodiments of the present invention. All other embodiments obtained by a person skilled in the art based on the embodiments of the present invention without any creative work belong to the protection scope of the present invention.
Referring to fig. 1 and fig. 2, the present embodiment provides a method for integrally lifting and leveling a steel structure based on a convolutional neural network and real-time monitoring.
First, the present invention is introduced in its entirety; the steel structure is generally a truss system which is assembled on the ground, integrally hoisted and installed on one side of a building, as shown in fig. 1, deformation of the steel structure needs to be well controlled in the hoisting process of the steel structure, so that the hoisting process is stable and safe (cannot be disassembled in the air), and meanwhile, the steel structure has a long service life after being installed.
Therefore, during the hoisting process of the steel structure, the invention focuses on the content of research, and based on a numerical simulation technology (simulating and hoisting the steel structure, acquiring strain in the actual hoisting process, and converting the strain into a digital signal for processing), a VGG (convolutional neural network) method and monitoring data acquired in real time in the hoisting process of the steel structure, the invention can acquire the hoisting state of the steel structure in real time, correctly adjust the displacement stroke of each hydraulic hoist, ensure that the hoisting steel structure is always in the horizontal position, and improve the safety and the working efficiency of the construction process.
Next, each step of the present invention will be described in detail, and includes the following steps.
Step S1: preparing and assembling an upper chord 12, a lower chord 13 and a web member 14 of a steel structure, and assembling on the ground of an installation position; the steel structure ground assembly and integral hoisting mode is adopted, and the steel structure ground assembly and integral hoisting mode is welded on one side of a building.
Furthermore, the safety of the hoisting process is ensured, the main operations of assembling, welding, painting and the like of the steel structure are carried out on the ground, the construction efficiency is high, the construction quality is easy to ensure, the construction operation of the steel structure is centralized on the ground, the influence on the construction of other specialties is small, the parallel construction of multiple operation surfaces can be realized, and the control of the total project period is facilitated.
Step S2: the upper chord member 12, the lower chord member 13 and the web members 14 are all arranged on a jig frame, are supported by the jig frame and are spliced into a whole to form a steel structure; selecting a plurality of hoisting points 15 on the steel structure, and performing integral hoisting by adopting a vertical hoisting mode; furthermore, the hoisting device adopted in the hoisting process is a plurality of hydraulic hoists fixedly arranged on the hoisting beam, flexible steel strands are used as the load bearing rigging, and the hoisting height is not limited as long as a reasonable load bearing hoisting point 15 is selected on a steel structure.
And step S3: selecting a plurality of strain detection points 11 for detecting deformation in a steel structure; the strain detection points 11 should be selected as many as possible at the connection.
And step S4: and arranging a strain gauge on the strain detection point 11, wherein the strain gauge is used for acquiring the strain of the structural rod piece, the strain gauge is connected with a control system, the acquired strain data is transmitted to the control system and is processed by the control system, and a trained VGG convolutional neural network is arranged and operated in the control system.
Step S5: in the lifting process of the steel structure, the strain value of each strain detection point 11 is input into a trained VGG convolutional neural network, the relative vertical displacement of each hoisting point 15 is obtained through calculation, the displacement stroke of each hydraulic lifter is output and adjusted, and the position of each hoisting point 15 is ensured to be on the same level; thereby guarantee that the steel construction is in horizontal position all the time for steel construction deformation is minimum, guarantees the stability and the increase of service life of steel construction.
Further, the step S5 also comprises the steps of carrying out computer centralized control on the strain gauge and the hydraulic lifters, independently monitoring each hydraulic lifter in the system in real time through data feedback, leading the collected data into a trained VGG convolutional neural network model, and calculating the relative displacement difference between each hoisting point 15 in real time so as to control and level the steel structure.
Furthermore, the way of training the VGG convolutional neural network: the hoisting model of the steel structure is built through SAP2000, the types and the materials of all rod pieces (an upper chord 12, a lower chord 13 and a web member 14) of the hoisting model are synchronous to a real structure, one hoisting point 15 is used as a reference, displacement loads of 5mm, 10mm, 15mm, 20mm, 25mm and 30mm are respectively applied to the other three hoisting points 15, a plurality of different strain detection points 11 are selected from the upper chord 12, the lower chord 13 and the web member 14, and the strain detection points 11 are selected from the positions consistent with the strain detection points 11 in the step S3, namely: the simulated strain detection point 11 is consistent with the actual strain detection point 11, so that the trained VGG convolutional neural network is closer to a real state, the accuracy of the VGG convolutional neural network is improved, and the strain value of the position of the strain detection point 11 obtained by simulation in the hoisting model is used as a data set to train the VGG convolutional neural network.
As a preferred scheme of the invention, in a trained VGG convolutional neural network, the strain data of each strain detection point 11 obtained by actual measurement is input, and the relative height difference of the hoisting points 15 is obtained through data processing, so that the stroke adjustment of the hydraulic lifter can be conveniently carried out in a targeted manner.
According to the invention, a technology of monitoring the upper chord 12, the lower chord 13 and the web members 14 in real time is adopted, if the lifting heights of hydraulic lifters of lifting points 15 are unequal in the lifting process of a steel structure, the stress change information of the strain detection points 11 can be fed back in real time through the measurement of strain gauges, the detection data is input into a trained VGG convolutional neural network to obtain the relative vertical displacement of each lifting point 15, the displacement stroke of each hydraulic lifter is controlled through output, accurate regulation and control are realized, individual independent regulation and control are realized, the excessive deformation of the steel structure in the lifting process is reduced, the potential safety hazard is caused, and the safe lifting is ensured; and the lifted steel structure is always in the best horizontal stress position; in addition, by using the VGG convolutional neural network, the network depth is suitable, the convolutional kernel is large, the parameters are fewer, the method is suitable for complex modes, and when the hoisting steel structure is complex, the number of internal measuring points is large, and the data to be processed is huge, the excellent performance can be embodied.
In this embodiment, the strain state of the steel structure should be real-time monitoring data of a plurality of strain detection points 11, the strain detection points 11 should be arranged on a plurality of rods of the whole structure, and the timeliness of the detection data and the availability of the data are ensured.
The accuracy rate of the VGG convolutional neural network training model meets the requirement, the number of convolutional layers should not be too shallow to cause under-fitting or the number of convolutional layers should not be too deep to cause over-fitting.
Further, the number of the strain detection points 11 is not less than 25, and in the calculation of the VGG convolutional neural network, too small capacity results in larger model training error and under-fitting condition; the excessive capacity causes the large data processing amount of the model, and the overfitting situation occurs.
The lifting beam on the building has the condition of fixedly mounting a plurality of hydraulic lifters, and both sides of the truss of the steel structure are connected with the tower stiff structure; in this embodiment, the steel structure may be fixedly connected to a steel truss extending from a building, and may be welded, bolted, or riveted.
Furthermore, the hydraulic lifter is driven by a hydraulic loop, the acceleration is very low in the action process, and no additional dynamic load is applied to a steel structure; the dynamic load is vibration and impact.
According to the invention, by adopting a hydraulic synchronous lifting construction technology (adopting hydraulic lifters for lifting), sensing monitoring (strain gauge acquisition strain data) and computer centralized control (control system), through data feedback and control instruction transmission, various functions such as synchronous action, load balancing, posture correction, stress control, operation locking, process display, fault alarm and the like can be fully automatically realized, and independent real-time monitoring and adjustment of each hydraulic lifter in the system are realized, so that the synchronous control precision of the hydraulic synchronous lifting process is higher, and the real-time performance is better; through the operation of a computer human-computer interface, automatic control, manual control and inching operation of a single lifting device can be realized, so that the special requirements of synchronous lifting, air attitude adjustment, single-point millimeter-scale fine adjustment and the like in the integral lifting installation process of the lifting unit are met.
Referring to fig. 1 and 2, the invention provides a steel structure lifting and leveling technology based on a convolutional neural network and real-time monitoring data, and solves the problem of integral hoisting of a truss (steel structure). The project is a high-rise public building complex building with a four-tower high-level connected structure. The main structure system is a four-tower + high-position connected structure, the connected structure is located on 14 layers, 15 layers and 16 layers, the connected structure between the tower and the building is composed of a large-span plane truss and steel beams, two ends of the connected structure are welded with stiff steel columns of the tower, the steel columns are mainly box-shaped sections, the steel structures are spliced into a whole on the floor right below the installation position before hoisting, and the whole is lifted to the position by using a vertical hoisting technology. The lifting unit is assembled into a whole on the floor right below the projection surface of the lifting unit, meanwhile, a lifting platform (an upper lifting point) is arranged at the position of 16 layers (elevation +72.35 m) of a main building structure by using stiff columns and corbels of the main building structure, a lower lifting appliance (a lower lifting point) is arranged at the position, corresponding to the upper lifting point, of the steel structure lifting unit, the lower lifting point is 4 end points of 2 upper chords 12 in the position of figure 1, and the upper lifting point and the lower lifting point are connected through special bottom anchors and special steel strands. And lifting the whole steel structure lifting unit to a designed mounting position by using a hydraulic synchronous lifting system, butting the steel structure lifting unit with a bracket at the pre-mounting section, and performing additional mounting on the steel structure lifting unit to finish mounting.
<xnotran> VGG VGG16 , 2 , - - - - - - - - - - - - - - - - - - - - , , VGG16 6 , 2 + RELU 21 22; </xnotran> The first 2 block structures are 2 convolutional layers + RELU activation function 21 and pooling layer 22, the next 3 block structures are 3 convolutional layers + RELU activation function 21 and pooling layer 22, and the last block structure is fully-connected layer + RELU activation function 23. Since both convolutional layers and fully-connected layers have weight coefficients, also referred to as weight layers, where convolutional layers 13, fully-connected 3, pooling layers do not involve weights. So there are 13+3=16 layers, and the activation units used are all RELU functions. For the VGG16 convolutional neural network, 13 convolutional layers and 5 pooling layers are responsible for extracting features, and the last 3 fully-connected layers are responsible for completing classification tasks.
The VGG convolution neural network is adopted to calculate and process real-time monitoring data, a larger convolution kernel is used in series in convolution, the convolution kernel has fewer parameters than a single convolution kernel, and meanwhile, more nonlinear changes are possessed than a single convolution kernel, so that the method is suitable for more complex modes; the convolution kernels are connected in series, and the features extracted for multiple times are finer than those extracted by a single convolution kernel; the convolution step is smaller than the kernel size, so that the feature extraction can be covered, and the feature fineness is improved.
Furthermore, the complexity of a network model is reduced by 3 strategies of local receptive field, weight sharing and down sampling of the VGG convolutional neural network, and the neural units of different layers are connected locally, namely the neural unit of each layer is only connected with part of the neural units of the previous layer. Each neural unit responds only to regions within the receptive field and is completely indifferent to regions outside the receptive field. Such a local connected mode ensures that the learned spatial local mode of the convolution kernel has the strongest response to the input. The weight sharing network structure is more similar to a biological neural network, the complexity of a network model is reduced, and the number of weights is reduced. Such a network structure is highly invariant to translation, scaling, tilting or other forms of deformation. And the convolutional neural network can effectively learn corresponding features from a large number of samples, thereby avoiding a complex feature extraction process.
In the invention, the bed-jig of each member (the upper chord 12, the lower chord 13 and the web member 14) on the ground is assembled to form a whole (a steel structure or a hoisting unit), and a plurality of strain detection points 11 (shown in figure 1) for detecting deformation are selected from the steel structure; installing strain gauges on the strain detection points 11, wherein the strain gauges are used for acquiring strain of the structural rod piece, inputting strain values of the strain detection points 11 into a trained VGG convolutional neural network, calculating to obtain relative vertical displacement of each hoisting point 15, outputting and adjusting displacement stroke of each hydraulic lifter, and ensuring the positions of the hoisting points 15 to be on the same level; thereby guarantee that the steel construction is in horizontal position all the time for steel construction deformation is minimum, guarantees the stability and the increase of service life of steel construction. The strain gauge and the hydraulic lifters are subjected to computer centralized control, and independent real-time monitoring is performed on each hydraulic lifter in the system through data feedback, so that the data precision of the hydraulic lifters in the synchronous lifting process is higher, and the real-time performance is better. And importing the collected data into a trained VGG convolutional neural network model, and calculating the relative displacement difference between each hoisting point 15 in real time so as to control and level the steel structure.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.
Claims (10)
1. A steel structure integral lifting leveling method based on a convolutional neural network and real-time monitoring is characterized in that: the method comprises the following steps:
step S1: preparing an upper chord member, a lower chord member and a web member of an assembled steel structure, and assembling on the ground of an installation position;
step S2: the upper chord member, the lower chord member and the web members are all arranged on the jig frame and spliced into a whole, a plurality of hoisting points are selected, and the whole is hoisted in a vertical hoisting mode;
and step S3: selecting a plurality of strain detection points for detecting deformation in a steel structure;
and step S4: installing a strain gauge on the strain detection point, wherein the strain gauge is used for acquiring the strain of the structural rod piece;
step S5: in the lifting process of the steel structure, the strain value of each strain detection point is input into a trained VGG convolutional neural network, the relative vertical displacement of each lifting point is obtained through calculation, the displacement stroke of each hydraulic lifter is output and adjusted, and the position of each lifting point is guaranteed to be on the same level.
2. The convolution neural network and real-time monitoring based steel structure overall lifting leveling method according to claim 1, characterized in that: and the step S5 also comprises the steps of carrying out computer centralized control on the strain gauge and the hydraulic lifters, independently monitoring each hydraulic lifter in the system in real time through data feedback, leading the collected data into a trained VGG convolutional neural network model, and calculating the relative displacement difference between each hoisting point in real time so as to control and level the steel structure.
3. The convolution neural network and real-time monitoring based steel structure overall lifting leveling method according to claim 2, characterized in that: building a hoisting model of a steel structure through SAP2000, wherein the model and the material of each rod piece of the hoisting model are synchronous to the real structure, applying displacement loads of 5mm, 10mm, 15mm, 20mm, 25mm and 30mm to the other three hoisting points respectively by taking one hoisting point as a reference, selecting a plurality of different strain detection points on the upper chord, the lower chord and the web member, selecting the positions of the strain detection points to be consistent with the strain detection points in the step S3, and training the VGG convolutional neural network by taking the strain values of the strain detection points simulated in the hoisting model as a data set.
4. The convolutional neural network and real-time monitoring based steel structure overall lifting leveling method as claimed in claim 3, wherein: and inputting the strain data of each strain detection point obtained by actual measurement in the trained VGG convolutional neural network, and obtaining the relative height difference of the hoisting points through data processing, thereby being convenient for pertinently adjusting the stroke of the hydraulic lifter.
5. The convolutional neural network and real-time monitoring based steel structure integral lifting leveling method as claimed in claim 1, characterized in that: the steel construction adopts the fixed a plurality of hydraulic lifting wares of installing on the lifting beam to hoist, adopts flexible steel strand wires as the bearing rigging, selects reasonable bearing hoist and mount point on the steel construction, and it is unrestricted to promote the height.
6. The convolution neural network and real-time monitoring based steel structure overall lifting leveling method according to claim 1, characterized in that: the strain state of the steel structure is real-time monitoring data of a plurality of strain detection points, the strain detection points are arranged on a plurality of rod pieces of the whole structure, and the timeliness of the detection data and the availability of the data are guaranteed.
7. The convolutional neural network and real-time monitoring based steel structure overall lifting leveling method as claimed in claim 3, wherein: the accuracy rate of the VGG convolutional neural network training model meets the requirement, and the number of convolutional layers is not too shallow or too deep to cause under-fitting or over-fitting.
8. The convolutional neural network and real-time monitoring based steel structure integral lifting leveling method as claimed in claim 7, wherein: the number of the strain detection points is not less than 25, and in the calculation of the VGG convolutional neural network, the model training error is large due to the fact that the capacity is too small, and under-fitting occurs.
9. The convolutional neural network and real-time monitoring based steel structure overall lifting leveling method as claimed in claim 5, wherein: the lifting beam has the condition of fixedly mounting a plurality of hydraulic lifters, and both sides of the truss of the steel structure are connected with the tower stiff structure.
10. The convolutional neural network and real-time monitoring based steel structure integral lifting leveling method as claimed in claim 9, wherein: the hydraulic lifter is driven by a hydraulic loop, the acceleration is extremely low in the action process, and no additional dynamic load is applied to the steel structure.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210843217.3A CN115234006B (en) | 2022-07-18 | 2022-07-18 | Steel structure integral lifting and leveling method based on convolutional neural network and real-time monitoring |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210843217.3A CN115234006B (en) | 2022-07-18 | 2022-07-18 | Steel structure integral lifting and leveling method based on convolutional neural network and real-time monitoring |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115234006A true CN115234006A (en) | 2022-10-25 |
| CN115234006B CN115234006B (en) | 2023-11-24 |
Family
ID=83672648
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210843217.3A Active CN115234006B (en) | 2022-07-18 | 2022-07-18 | Steel structure integral lifting and leveling method based on convolutional neural network and real-time monitoring |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115234006B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103790234A (en) * | 2014-01-03 | 2014-05-14 | 永升建设集团有限公司 | Large-span truss hydraulic multi-point synchronous lifting construction method |
| CN103835511A (en) * | 2014-02-28 | 2014-06-04 | 中天建设集团有限公司 | Integral lifting method for ultra-long light thin giant steel tube truss |
| CN104088467A (en) * | 2014-07-17 | 2014-10-08 | 成都市第四建筑工程公司 | High-altitude construction deformation monitoring method of large-cantilever steel truss structure |
-
2022
- 2022-07-18 CN CN202210843217.3A patent/CN115234006B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103790234A (en) * | 2014-01-03 | 2014-05-14 | 永升建设集团有限公司 | Large-span truss hydraulic multi-point synchronous lifting construction method |
| CN103835511A (en) * | 2014-02-28 | 2014-06-04 | 中天建设集团有限公司 | Integral lifting method for ultra-long light thin giant steel tube truss |
| CN104088467A (en) * | 2014-07-17 | 2014-10-08 | 成都市第四建筑工程公司 | High-altitude construction deformation monitoring method of large-cantilever steel truss structure |
Non-Patent Citations (1)
| Title |
|---|
| 杨露菁: "智能图像处理及应用", 中国铁道出版社有限公司, pages: 58 - 59 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115234006B (en) | 2023-11-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101634604B (en) | Static strength test method of large vane used for wind power and electrical power and test system thereof | |
| CN108226399A (en) | A kind of beam-string structure damage combined recognising method based on modal parameter | |
| CN102071814A (en) | High-rise long-span door type truss integral lifting device and construction method thereof | |
| CN115467421B (en) | Hoisting construction method for complex large-span steel roof | |
| CN112195797B (en) | Closure construction method for continuous beam bridge | |
| CN111749447A (en) | Tall and big space fastener type full scaffold monitoring structure and monitoring method | |
| CN108842954A (en) | A kind of huge different section Honeycomb Beam structural system construction method of installation of aerofoil profile roofing | |
| CN115234006B (en) | Steel structure integral lifting and leveling method based on convolutional neural network and real-time monitoring | |
| CN111535446B (en) | Construction method of large-span stadium | |
| CN217580579U (en) | Gymnasium steel roof hoist mechanism with basement structure | |
| CN101672743B (en) | Simulation test estimation system and test method for construction of inner stayguy suspended holding rod split tower | |
| CN104863259B (en) | Upside-down construction method for large-span ultra-high signal-layer cylindrical architectural steel structure | |
| CN111395144A (en) | Method for monitoring and judging health state of long-span continuous rigid frame bridge | |
| CN113868925A (en) | Reduction coefficient determining method and bracket structure design method | |
| CN102877657A (en) | Large-span H-shaped plane composite structure beam string upper-air cable replacement construction method | |
| CN107451336B (en) | Robust design method for optimizing product performance change boundary | |
| CN112065043B (en) | Large cantilever safe intelligent construction system and method for finite element synchronous analysis | |
| CN213418410U (en) | Tall and big space fastener type full scaffold monitoring structure | |
| CN112255938A (en) | Intelligent detection control system for high pier hydraulic creeping formwork | |
| CN114934596B (en) | Gymnasium steel roof lifting construction method and mechanism with basement structure | |
| CN115924740B (en) | Asynchronous lifting control method for large-span steel structure | |
| CN114476943B (en) | Method for determining hoisting point of large wind power generation component | |
| CN220960174U (en) | Super-large-span heavy steel gallery upper conjoined construction monitoring system | |
| CN220542392U (en) | Synchronous testing device for pneumatic characteristics of tower crane-bridge tower combined system | |
| Chen et al. | Study on structural checking method of inclined elevator for tower group of 500 kV Jiangyin second crossing of Yangtze River |
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 |