CN113267612B - Soil moisture content detection method, detection system and penetration device - Google Patents
Soil moisture content detection method, detection system and penetration device Download PDFInfo
- Publication number
- CN113267612B CN113267612B CN202110534793.5A CN202110534793A CN113267612B CN 113267612 B CN113267612 B CN 113267612B CN 202110534793 A CN202110534793 A CN 202110534793A CN 113267612 B CN113267612 B CN 113267612B
- Authority
- CN
- China
- Prior art keywords
- soil
- soil body
- bending element
- water content
- element sensor
- 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
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/24—Earth materials
- G01N33/246—Earth materials for water content
Abstract
The invention discloses a soil moisture content detection method, a detection system and a penetration device, wherein the detection method comprises the following steps: obtaining soil body dry densities of different types of soil bodies, measuring the soil body dry densities, the soil penetration depths and the shear wave speeds under vertical pressure of different soil body types and corresponding water contents, and corresponding the water contents to the soil body dry densities, the soil penetration depths, the soil body types, the vertical pressure and the shear wave speeds; and constructing an artificial neural network model by taking the soil body dry density, the soil penetration depth, the shear wave velocity, the soil body type and the vertical pressure as characteristic values and the water content as target values, and obtaining the soil body water content by using the trained artificial neural network model. The detection method can be applied to the measurement of the water content of the compacted road in site construction, the prediction of the water content of different thicknesses is realized, the detection is convenient, the method is suitable for industrial site use, and the detection precision is high.
Description
Technical Field
The invention relates to a method and a system for detecting the moisture content of soil based on a bending element sensor and a penetration device.
Background
In engineering construction, the water content is taken as one of the basic physical properties of the soil body, has important influence on the strength and deformation characteristics of the foundation, and can greatly improve the detection efficiency and save precious time for subsequent construction if the water content of the soil bodies with different depths can be rapidly and accurately measured under the field condition. The current common soil moisture content measuring method comprises a mass method, a ray method, a dielectric method, a remote sensing method, a ground penetrating radar method and the like.
The mass method utilizes the mass change of the soil body to determine the water content, and the measurement result is most accurate, but the time is long, and the method is not suitable for the soil with high organic matter content and cannot meet the condition of on-site rapid measurement; the ray method utilizes the change of rays when the rays pass through the soil body to determine the moisture content of the soil body, can meet the requirements of high-precision and quick measurement on site, but has certain radiation hidden danger and is unfavorable for the health of operators; the dielectric method determines the water content through the relation between the water content and the dielectric property of the soil body, is convenient and accurate to measure, can be used for on-site rapid measurement and in-situ monitoring, is greatly influenced by temperature and salinity, and is mainly imported by the prior equipment, so that the cost is high; the remote sensing method measures the space-time variation of the water content of the surface soil by monitoring the spectral characteristics and the thermal properties of the soil surface, has the advantages of wide range and quick measurement, but only monitors the water content of the surface soil; the ground penetrating radar method can rapidly measure the water content of soil mass on a large scale, is convenient and lossless, can repeatedly detect, and is complex in data extraction and processing.
In summary, the existing detection means are difficult to rapidly measure the water content of soil bodies at different depths on site, and cannot meet the use requirements.
Disclosure of Invention
Aiming at the problems that the existing water content testing technology has long measurement time, hidden radiation hazards, high price, surface layer measurement only and difficult data extraction respectively, the invention provides a soil body water content detection method, a detection system and a penetration device based on a bending element sensor combined with an artificial neural network model.
The technical scheme adopted for solving the technical problems is as follows:
in a first aspect, the present invention provides a method for detecting a soil moisture content, the method comprising:
obtaining soil body dry densities of different types of soil bodies, measuring the soil body dry densities, the soil penetration depths and the shear wave speeds under vertical pressure of different soil body types and the corresponding water contents, and corresponding the water contents to the soil body dry densities, the soil penetration depths, the soil body types, the vertical pressure and the shear wave speeds to form a data set;
constructing an artificial neural network model by taking the soil body dry density, the soil penetration depth, the shear wave velocity, the soil body type and the vertical pressure as characteristic values and the water content as target values, and training the artificial neural network model by utilizing a data set to obtain a trained artificial neural network model;
and inputting the soil body dry density, the soil penetration depth, the shear wave velocity, the soil body type and the vertical pressure of the soil body to be tested into a trained artificial neural network model to obtain the soil body water content.
In a second aspect, the present invention provides a penetration device having a built-in bending element sensor, comprising: the telescopic rod comprises two vertical parallel probe rods (2) and a connecting rod (1) fixedly connected with the two parallel probe rods, and is characterized in that each parallel probe rod is carved with a graduated scale (4), each probe rod (2) is a multi-layer telescopic sleeve rod, the sleeve rods of adjacent layers are connected through spring buckles, the next layer of sleeve rod connected with the spring buckles after the spring buckles are pressed down extends out, graduated scale graduations on the surfaces of the multi-layer telescopic sleeve rods are continuous when the multi-layer telescopic sleeve rods are fully unfolded, the number of unfolding layers of the telescopic sleeve rods is changed according to the depth of entering soil, and the graduation measuring range of the telescopic sleeve rods is ensured to be always larger than the depth of entering soil;
the hollow rod 11 is coaxially arranged in the probe rod 2, the hollow rod 11 is fixedly connected with the inner wall of the innermost loop rod, the screw rod 9 passes through the hollow rod and penetrates out of the probe rod 2, the screw rod is eccentrically arranged, the bottom of the screw rod 9 is fixedly connected with a first triangular slide block, the upper part of the screw rod is provided with a rotary nut 8, and the rotary nut 8 can control the screw rod to move up and down so as to drive the first triangular slide block 12 at the bottom end of the screw rod (9) to move up and down;
the pipe wall on the probe rod at the tail end is reserved with a hole site for pushing out the bending element sensor, the horizontal plane of the hole site is provided with a guide rail (14) for moving the bending element sensor, the guide rail is fixed inside the innermost loop bar and is fixed with the inner wall of the innermost loop bar, the inner side of the bending element sensor is connected with a second triangular slide block, the first triangular slide block is contacted with two inclined planes of the second triangular slide block, the second triangular slide block is extruded by the first triangular slide block through the inclined planes in the up-down moving process of the screw rod, the second triangular slide block moves along the guide rail (14) in the horizontal direction, the bending element sensor is pushed out of the hole site, and the bending element sensors on the two probe rods are used for receiving the bending element sensor and transmitting the bending element sensor.
In a third aspect, the invention provides a soil moisture content detection system based on a bending element sensor, which is characterized by comprising a penetration device with the bending element sensor, an electric signal generating module, an electric signal receiving and processing module and a data management platform (S6); after the penetrating device with the built-in bending element sensor reaches the soil body target depth, applying a regular waveform with set voltage and frequency to the transmitting bending element sensor through the electric signal generating module, wherein the piezoelectric effect forces the transmitting bending element to drive surrounding soil bodies to generate transverse vibration, and shear waves generated in the direction perpendicular to the vibration drive the receiving bending element sensor at the other end to vibrate through the soil body, weak currents generated by the receiving bending element sensor are finally transmitted to the electric signal receiving and processing module, and the electric signal receiving and processing module and the data management platform are in data interconnection and intercommunication;
the electric signal generating module is connected with the transmitting bending element sensor through one end of the BNC connecting wire, one end of the electric signal generating module is connected with the electric signal receiving and processing module, the other end of the electric signal receiving and processing module is connected with the receiving bending element sensor through the BNC connecting wire, and the data management platform realizes data interaction with the electric signal receiving and processing module through a 5G network; in addition, both the transmitting and receiving bending element sensors need to be inserted into the soil body to be measured at the same depth.
The electric signal generation module is used for generating an electric signal;
the electric signal receiving and processing module comprises: a signal amplifier (S2), an oscilloscope (S3) and an embedded computer chip (S4); the input of the signal amplifier is connected with the electric signal of the bending element sensor, the oscilloscope is simultaneously connected with the output of the signal amplifier and the output of the signal generator, the output of the oscilloscope is connected with the embedded computer chip, and the embedded computer chip is communicated with the data management platform (S6);
the data management platform is provided with a database related to soil parameters and has the functions of data import, storage, modification, calculation, calling, export and visualization.
In a fourth aspect, the present invention provides a soil moisture content detection method for soil moisture content detection after compaction of roadbed, which is characterized in that the detection method comprises the following steps:
1) Preparation of test specimens
Selecting different types of soil, screening, removing impurities, drying, weighing, measuring the initial water content and the dry density of the soil, and controlling the water content in the soil by changing the water adding amount to obtain test samples of different soil with different water contents;
2) Collecting raw data
Aiming at test samples, different vertical pressures are applied to the surfaces of different test samples, the penetration depth of a bending element sensor is changed, the measured shear wave speeds of different test samples at different vertical pressures and different soil penetration depths are used as a set of data, and then multiple sets of data of all the test samples are obtained to form a data set;
3) Constructing an artificial neural network model
Taking the soil body dry density, the soil penetration depth, the shear wave velocity, the soil body type and the vertical pressure as characteristic values, taking the water content as a target value, and performing training on all data sets until the error meets a limit value, wherein the limit value is determined according to the precision required by field measurement, so as to obtain a trained artificial neural network model, namely a shear wave velocity-water content prediction model; after data preprocessing, a training set and a testing set are proportionally divided, model parameters are adjusted, training is carried out on all samples until errors meet limit values, the limit values are determined according to the accuracy required by field measurement, the accuracy requirements are higher and lower, and the specific values are related to super-parameter settings (learning rate, hidden layer number, neuron number and the like) of the limit value neural network model, so that a shear wave speed-water content prediction model is obtained.
4) Access data management platform
Writing the trained artificial neural network model into an embedded computer chip, and accessing the embedded computer chip into a data management platform, wherein the soil body dry density, the vertical pressure and the soil body type can be obtained from the data management platform, and the soil body water content of the field construction can be obtained by inputting the soil body shear wave velocity and the soil penetration depth measured on the field;
and after the on-site roadbed is compacted, the trained artificial neural network model is applied to predict the water content of the soil body, meanwhile, the water content spot check is periodically carried out, the characteristic value and the spot check true value of the corresponding sample during the water content spot check are recorded, a new data set is formed, and the new data set is transmitted back to the data management platform for expanding the database of the data management platform and updating the artificial neural network model. The construction site is provided with a density testing instrument, and the measurement data and the data management platform are interconnected and communicated and can be directly called from the data management platform.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention uses the probe with the built-in bending element sensor to measure the shear wave velocity of the soil body, has no limit on the type and depth of the soil body to be measured, has high measuring speed and is harmless to the health of operators. The penetration device is in a telescopic form, is more portable, is convenient to use and has high measurement precision.
2. In the detection method, the soil body dry density, the soil penetration depth, the shear wave velocity, the soil body type and the vertical pressure are used as characteristic values, the water content is used as a target value, an artificial neural network model is constructed, the relation between the characteristic values and the water content of the soil body is obtained through training, the quantitative corresponding relation between the characteristic values and the target value is determined, and the measurement accuracy is high.
3. The detection method can be applied to site construction, the water content after road compaction is measured, the road compaction is laid layer by layer, each layer of soil is different in thickness and depth, the prediction of the water content at different thicknesses is realized, the detection is convenient, and the detection method is suitable for industrial site use.
Drawings
FIG. 1 is an overall flow chart of one embodiment of the measurement method of the present invention.
Fig. 2 is a block diagram of the structure of the measuring system of the present invention.
FIG. 3 is a schematic view of the inner and outer structures of the penetration device with the bending element sensor according to the present invention, wherein the left side is a schematic view of the inner structure, and the right side is a schematic view of the outer structure.
Fig. 4 is a schematic structural diagram of an artificial neural network model in the present invention.
FIG. 5 is a flow chart of the artificial neural network construction of the present invention.
Fig. 6 is a detailed view of a retractable probe spring catch.
Reference numerals:
s1-a penetration device with a built-in bending element sensor, S2-a signal amplifier, S3-an oscilloscope, S4-an embedded computer chip, S5-a signal generator and S6-a data management platform.
The device comprises a 1-connecting rod, a 2-probe rod, a 3-spring buckle, a 4-graduated scale, a 5-bending element sensor protection structure, a 6-transmitting bending element sensor, a 7-conical head, an 8-nut, a 9-screw rod, a 10-connecting cable, a 11-hollow rod, a 12-sliding block, a 13-receiving bending element sensor, a 14-guide rail and a 15-spring.
Detailed Description
The present invention is further explained below with reference to examples and drawings, but is not to be construed as limiting the scope of the present application.
The invention discloses a soil moisture content measuring system based on a bending element sensor combined with an artificial neural network model, which comprises a penetration device S1 internally provided with the bending element sensor, an electric signal generating module, an electric signal receiving and processing module and a data management platform (S6).
After the penetrating device with the built-in bending element sensor reaches the soil body target depth, a regular waveform with certain voltage and frequency is applied to the transmitting bending element sensor through the electric signal generating module, the piezoelectric effect forces the transmitting bending element to drive surrounding soil bodies to generate transverse vibration, the shear wave generated in the direction perpendicular to the vibration drives the receiving bending element sensor at the other end to vibrate through the soil body, weak current generated by the receiving bending element sensor is finally transmitted into the electric signal receiving and processing module, and the electric signal receiving and processing module and the data management platform are in data interconnection and intercommunication. The electric signal generating module is connected with the transmitting bending element sensor through one end of the BNC connecting wire, one end of the electric signal generating module is connected with the electric signal receiving and processing module, the other end of the electric signal receiving and processing module is connected with the receiving bending element sensor through the BNC connecting wire, and the data management platform realizes data interaction with the electric signal receiving and processing module through a 5G network. In addition, both the transmitting and receiving bending elements need to be inserted at the same depth of the soil body to be measured.
The electric signal generation module is used for generating an electric signal, and comprises: a signal generator (S5) capable of generating an electrical signal having a waveform of a predetermined frequency, amplitude and phase;
the electric signal receiving and processing module comprises: a signal amplifier (S2), an oscilloscope (S3) and an embedded computer chip (S4); the input of the signal amplifier is connected with the electric signal of the bending element sensor, the oscilloscope is connected with the output of the signal amplifier and the output of the signal generator, the output of the oscilloscope is connected with the embedded computer chip, and the embedded computer chip and the data management platform (S6) are mutually communicated.
The data management platform is a soil mass site construction data platform with the functions of data import, storage, modification, calculation, calling, export, visualization and the like, the data management platform stores the soil mass dry density, the soil depth, the soil mass type and the like of the soil mass of the construction site, and the data management platform (S6) is internally provided with a database related to soil mass parameters.
The penetrating device with the built-in bending element sensor is improved from a probe rod with a fixed length to a telescopic probe rod, so that portability is improved. Comprising the following steps: the telescopic rod comprises two vertical parallel probe rods (2), a connecting rod (1) fixedly connected with the two parallel probe rods, graduated scales (4) are engraved on each parallel probe rod, the probe rods (2) are multilayer telescopic loop rods, spring buckles (3) are arranged on the outer wall of each layer of loop rod and are connected with loop rods of adjacent layers, the next layer of loop rod connected with the spring buckles after the spring buckles are pressed down can be stretched out, the length of the telescopic probe rods (2) can be adjusted, graduated scale scales on the surfaces of the multilayer telescopic loop rods are continuous when the multilayer telescopic loop rods are fully unfolded, the number of unfolding layers of the telescopic loop rods can be changed according to the depth of entering soil, and the graduated scale range of the telescopic loop rods is ensured to be always larger than the depth of entering soil. The hollow rod 11 is coaxially arranged in the probe rod 2, the hollow rod 11 is fixedly connected with the inner wall of the telescopic probe rod (2) through welding or other modes such as bolt fixing, the screw rod 9 penetrates through the hollow rod and penetrates out of the probe rod 2, the screw rod is eccentrically arranged and provides space for horizontal movement of the triangular slide block, the hollow rod with the inner diameter slightly larger than that of the screw rod is arranged to further limit the movable space of the screw rod, shaking is avoided, meanwhile, space can be reserved for a cable, the first triangular slide block is fixedly connected to the bottom of the screw rod 9, the rotary nut 8 is arranged on the upper portion of the screw rod, the rotary nut 8 can control the screw rod to move up and down, and the first triangular slide block 12 at the bottom end of the screw rod (9) is driven to move up and down;
a guide rail (14) for pushing out the bending element sensor is reserved on the upper pipe wall of the probe rod at the tail end, the guide rail is fixed inside the probe rod and fixed with the inner wall of the probe rod, the inner side of the bending element sensor is connected with a second triangular slide block, a spring is fixedly arranged between the second triangular slide block above the bending element sensor and the inner wall of the probe rod, the first triangular slide block extrudes the second triangular slide block through the inclined surface in the process of moving up and down of the screw rod, so that the second triangular slide block moves along the guide rail (14) in the horizontal direction, namely, horizontal displacement occurs, the bending element sensor is pushed out of the guide rail, good contact with soil is ensured, a nut can be reversely rotated when the bending element sensor is retracted, and the spring (15) pushes the second triangular slide block inwards to reset in the process of releasing the compression state. The bending element sensors on the two probe rods, one receiving the bending element sensor and the other transmitting the bending element sensor.
Two bending element sensor protection structures (5) are symmetrically arranged on the pipe wall of the probe rod of the pushing hole position of the bending element sensor (6) up and down by taking the bending element sensor as the center.
The bending element sensor (6) is connected with the signal generator (S5) through a cable (10), and the cable (10) is placed in the reserved hollow rod (11). The sleeve rod at the bottommost layer of the probe rod (2) is provided with a conical head (7). When the device is used, the spring buckle connected with the lower-layer loop bar is pressed firstly, the lower-layer loop bar is connected with the conical head to pop up, and when the scale of the lower-layer loop bar is insufficient, the spring buckle extending out of the upper-layer loop bar is pressed and controlled, so that the measurement requirements of different depths of entering soil are met.
In this embodiment, three loop bars are nested in sequence to form the whole probe bar, the outer side walls of the upper parts of the two loop bars except the loop bar at the outermost layer are all provided with spring buckles, the upper part and the lower side wall of the loop bar at the upper layer of the loop bar provided with the spring buckles are respectively provided with upper and lower limiting holes, the spring buckles of the loop bar with the spring buckles are located at the limiting holes and can limit the loop bar, when the spring buckles are located at the upper limiting holes, the spring buckles are pressed down, the loop bar at the lowest layer stretches out, stops stretching out at the position of the lower limiting holes and is fixed by the spring buckles again. When the sleeve rod is retracted, the spring buckle at the lower limit hole is pressed, and the sleeve rod can be retracted by pushing the spring buckle upwards with hand.
The invention relates to a soil moisture content detection method, which comprises the following detailed steps:
step one: preparation of test specimens
Collecting different types of soil (generally referred to as sandy soil, silt soil, cohesive soil and the like) outdoors, grinding the soil, sieving the ground soil with a 2mm sieve, and removing impurities such as stones, grass roots and the like in the soil. Weighing the mass of the soil body subjected to screening and impurity removal as M1, measuring the wet density rho of the soil body by using a ring cutter method or other methods, putting the soil body into a baking oven at 110 ℃ and drying until the mass is unchanged, obtaining original samples of different types of soil bodies with water content of 0, weighing the mass as M2, and calculating the initial water content w of the soil body 0 And dry density ρ d 。
Calculation formula of initial water content of soil body
Soil body dry density calculation formula
Step two: collecting raw data
Firstly, taking a part of dried soil mass which is M, adding water into the dried soil mass according to a water adding amount calculation formula, and uniformly mixing the dried soil mass to ensure that the water content of the soil mass is 1%
Water adding amount M w Calculation formula
Wherein: w' is soil body pre-treatmentLet water content, w 1 Is the water content of the soil before water addition.
And then the two probe rods with the built-in bending element sensors (6) are vertically inserted into soil, the same penetration depth is ensured through the graduated scale (4) on the outer wall of the probe, and the penetration depth h is recorded. The two bending element sensor protection structures (5) extending out of the outer wall of the probe rod can ensure that the bending element sensor (6) is not damaged in the insertion process of the probe rod.
Two output ends of the signal generator (S5) are respectively connected with the transmitting bending element sensor (6) and the oscilloscope (S3), and the receiving bending element sensor (13) is connected with the oscilloscope (S3). The signal generator (S5) and the oscilloscope (S3) are turned on, waveform parameters are input, the output wave and the receiving wave are simultaneously displayed by the oscilloscope, and the time difference T between the first wave peaks of the two waveforms is read. The horizontal distance L between the two probes is measured by a graduated scale, the shear wave velocity v of the soil body is calculated by the following formula,
and then uniformly placing weights on the top of the container for holding the soil body, and sequentially applying vertical pressures of 5kPa, 10kPa, 20kPa, 50kPa, 100kPa and 200kPa to the soil body by changing the mass of the weights. And under the same vertical pressure, measuring the shear wave speed of the soil body under the water content by changing the penetration depth of the probe rod. And then removing weights sequentially according to the sequence of 200kPa, 100kPa, 50kPa, 20kPa, 10kPa and 5kPa, changing the depth of the probe rod into soil under the same vertical pressure, measuring the shear wave velocity of the soil under the condition of the water content, and thus completing the data acquisition work of a water content sample. And then adding water into the current soil body according to a water adding amount calculation formula, uniformly mixing, and respectively measuring soil body shear wave velocities with different vertical pressures and depths when the water contents are 2%, 4%, 8%, 16%, 24%, 32%, 48%, 60%, 72%, 84% and 96%.
Repeating the steps after the whole process is finished and changing the soil body types until the soil body shear wave speed corresponding to each soil body type, the water content, the dry density, the soil penetration depth and the vertical pressure is measured.
Step three: constructing an artificial neural network model
(1) The invention adopts BP neural network model, and the input layer has five characteristic values, namely soil body dry density, soil penetration depth, shear wave velocity, soil body type and vertical pressure; the output layer only has one target value, namely the water content of the soil body.
And (3) respectively carrying out normalization treatment on the characteristic value soil body dry density, the soil body shear wave speed, the soil penetration depth and the vertical pressure measured in the first step and the second step to enable the characteristic value soil body dry density, the soil body shear wave speed, the soil penetration depth and the vertical pressure to be in a [0,1] interval, and attaching different data labels to different soil body types. And taking the characteristic value and the corresponding target value as a group of data, and forming a data set by a plurality of groups of data. The normalization formula is as follows:
wherein:
is the characteristic value x i Values after normalization, x max Is equal to x i Maximum value, x in the similar characteristic values min Is equal to x i The minimum value among the similar eigenvalues.
The dataset was then assembled using the set aside method according to 4: the proportion of 1 is divided into a training set and a testing set which are mutually exclusive, and meanwhile, in order to ensure the consistency of data distribution, the training set and the testing set are divided by adopting a hierarchical sampling method, so that the influence of extra deviation introduced in the data dividing process on the model prediction effect is avoided.
Theoretical analysis proves that the BP neural network model with a single hidden layer can map all continuous functions, the characteristic value and the target value of the model are fewer, and the requirement can be met by only setting one hidden layer. In order to avoid the phenomenon of 'over fitting' when the number of hidden nodes is too large, the neural network model has low prediction precision when the number of hidden nodes is too small, the hidden node number is determined by adopting a trial-and-error method, and the empirical formula is as follows
Wherein: m is the hidden node number, n is the input layer node number, l is the output layer node number, alpha is a constant between 1 and 10, and the hidden node number of the neural network model is between 3 and 12 and finally determined to be 10.
(2) Training neural networks
1. Random numbers are given to a weight matrix v= (V1, V2, vj, V10) between the input layer and the hidden layer and a weight matrix w= (W1) between the hidden layer and the output layer, a sample pattern counter p and a training number counter q are set to 1, an error E is set to 0, and a learning rate η is set to (0, 1)]Decimal, accuracy E after network training MIN Set to a positive decimal.
2. With current sample X p 、d p ，X p For the p-th group of input vectors, d p For the p-th group of desired output vectors, O k Is the output of the kth neuron of the output layer, y j Is the output of the jth neuron of the hidden layer, x i Is the i-th eigenvalue of the input layer;
assigning a value to the vector array X, d, calculating each component in the output vector Y of the hidden layer and the output vector O of the output layer using the following formula
o k ＝f(netk)
y j ＝f(netj)
Wherein: f (x) is a monopolar Sigmoid function, v ij Representing the weight of the ith eigenvalue and the jth neuron connecting channel of the hidden layer,w jk the weight of the channel connecting the jth neuron of the hidden layer and the kth target value is represented, and net is an intermediate variable of the function f (x).
3. Calculating error signals of each layer and adjusting weight of each layer
The error is defined as follows:
hidden layer:
input layer:
calculating error signals of each layer
Wherein:
error signal for the kth neuron of the output layer,/->
For the error signal of the jth neuron of the hidden layer, net is an intermediate variable of the function f (x);
adjusting weight of each layer
k＝1，2，…，l j＝1，2，…，m
j＝1，2，…，m i＝0，1，，…，n
Wherein: n is the number of input layer characteristic values, m is the number of hidden layer nodes, l is the number of output layer target values, and the number of the output layer target values is equal to the number of the output layer nodes; η is a learning rate coefficient and is generally 0.001 to 0.01.
4. Calculate the total output error E of all samples RME
Wherein: p is the total number of training samples,
is the desired output vector of the kth target value of the p-th group,>
is the output vector of the kth target value of the output layer of the p-th group,
5. cycling steps 2-4 ensures that all samples complete the training (i.e., steps p=p+ 1;q =q+1 in the flowchart).
6. Checking whether the total error of the network meets the precision requirement, if E RME ＜E MIN Ending training, otherwise, setting the total error E RME Setting 0, setting the sample count p to 1, and retraining all samples according to the steps 2-5 until the error meets the precision requirement.
Step four: access data management platform
Writing the trained artificial neural network model into an embedded computer chip (S4), and connecting the input and output ends of the embedded computer chip with a data management platform (S6) interface. The data management platform (S6) has the functions of data import, storage, modification, calculation, calling, export, visualization and the like.
When the water content of the soil body of a certain area needs to be measured, parameters such as the soil body type, the dry density and the like of the area are called through an interface of a data management platform (S6).
Inserting a probe rod into soil body to measure shear wave speed, and reading out the depth of the soil through a graduated scale (4) on the surface of the probe rod (2);
the vertical pressure can be obtained by conversion according to construction parameters of foundation treatment machines such as a road roller, the vertical pressure corresponding to the foundation treatment machines such as the road roller is stored in the data management platform, and can be directly called when needed, and also can be manually calculated and input.
And inputting the obtained soil type, soil dry density, soil shear wave speed, soil penetration depth and vertical stress into a trained artificial neural network model as characteristic values. After the artificial neural network model in the embedded computer chip (S4) obtains all the characteristic values, the water content of the soil body is automatically calculated, and all the characteristic values and the target values predicted by the model are uploaded to the data management platform (S6) through the interface for storage.
The measuring method can be applied to the water content measurement of soil after the compaction of the subgrade on site, so that the introduction of vertical pressure is necessary, and the acquisition of the vertical pressure can be obtained according to actual construction.
The soil moisture content detection system based on the bending element sensor is provided with manual input, data offline storage and model update, when communication with a database related to soil parameters of a data management platform (S6) is temporarily unavailable due to data transmission and the like, a relevant characteristic value can be manually input through an external input device connected with an embedded computer chip, the model calculates a target value moisture content and then stores the data offline in a memory of the computer chip, a data set obtained in a measurement process is stored after the data transmission is recovered, the data set is uploaded to the data management platform (S6) together with a real value of a moisture content sampling sample for a period of time, the expanded data can be used for training an artificial neural network model in real time, the updated artificial neural network model is written into the embedded computer chip (S4) again, the model prediction precision is improved, and the self-updating of the database and the artificial neural network model is realized. In the construction process of the on-site roadbed, the water content of the soil body in the construction process is measured in time by using the artificial neural network model, a detection department can randomly sample part of the predicted soil body on site to obtain a real water content value corresponding to the characteristic value of the soil sample, and the sampled characteristic value and the real value are recorded to be used as a new data set for updating the artificial neural network model.
The invention is applicable to the prior art where it is not described.
Claims (8)
1. The detection method can be applied to the water content measurement of soil after the compaction of the on-site roadbed, the road is compacted layer by layer, each layer of soil is compacted layer by layer, the thickness of each layer of soil is different, the depth is different, and the water content prediction at different depths is realized; the detection method comprises the following steps:
collecting different types of soil outdoors to obtain soil dry densities of different types of soil, controlling the water content in the soil by changing the water adding amount to obtain test samples of different soil with different water content, applying different vertical pressures on the surfaces of different test samples aiming at the test samples, changing the penetration depth of a bending element sensor, and taking the measured shear wave speeds of different test samples with different vertical pressures and different soil penetration depths as a group of data to obtain multiple groups of data of all test samples so as to form a data set;
constructing an artificial neural network model by taking the soil body dry density, the soil penetration depth, the shear wave velocity, the soil body type and the vertical pressure as characteristic values and the water content as target values, and training the artificial neural network model by utilizing a data set to obtain a trained artificial neural network model;
inputting the soil body dry density, the soil penetration depth, the shear wave speed, the soil body type and the vertical pressure of the soil body to be tested into a trained artificial neural network model to obtain the soil body water content of site construction; the soil shear wave speed and the soil penetration depth of the soil to be measured are measured on site, and the vertical pressure of the soil to be measured is obtained through conversion according to the construction parameters of the foundation treatment machine.
2. The method according to claim 1, wherein the shear wave velocity of the soil is measured by a penetration device with a built-in bending element sensor, the penetration device comprising: the telescopic sleeve rod comprises two vertical parallel probe rods (2) and a connecting rod (1) fixedly connected with the two parallel probe rods, wherein each parallel probe rod is carved with a graduated scale (4), each probe rod (2) is a multilayer telescopic sleeve rod, sleeve rods of adjacent layers are connected through spring buckles, the next layer of sleeve rod connected with the adjacent layers of sleeve rods after the spring buckles are pressed down extends out, graduated scale graduations on the surfaces of the multilayer telescopic sleeve rods are continuous when the multilayer telescopic sleeve rods are fully unfolded, the number of layers of the expansion layers of the telescopic sleeve rods is changed according to the depth of the earth, and the graduated range of the telescopic sleeve rods is ensured to be always larger than the depth of the earth;
the hollow rod is coaxially arranged in the probe rod, the hollow rod is fixedly connected with the inner wall of the innermost loop rod, the screw rod penetrates through the hollow rod and penetrates out of the probe rod, the screw rod is eccentrically arranged, the bottom of the screw rod is fixedly connected with a first triangular slide block, a rotating nut is arranged on the upper part of the screw rod, the rotating nut can control the screw rod to move up and down, and the first triangular slide block at the bottom end of the screw rod (9) is driven to move up and down;
the pipe wall on the probe rod at the tail end is reserved with a hole site for pushing out the bending element sensor, the horizontal plane of the hole site is provided with a guide rail (14) for moving the bending element sensor, the guide rail is fixed inside the innermost loop bar and is fixed with the inner wall of the innermost loop bar, the inner side of the bending element sensor is connected with a second triangular slide block, the first triangular slide block is contacted with two inclined planes of the second triangular slide block, the second triangular slide block is extruded by the first triangular slide block through the inclined planes in the up-down moving process of the screw rod, the second triangular slide block moves along the guide rail (14) in the horizontal direction, the bending element sensor is pushed out of the hole site, and the bending element sensors on the two probe rods are used for receiving the bending element sensor and transmitting the bending element sensor.
3. The method for detecting the moisture content of the soil body according to claim 2, wherein a spring is fixedly arranged between the second triangular slide block above the bending element sensor and the inner wall of the probe rod, the nut is reversely rotated first when the bending element sensor is retracted, and the second triangular slide block is pushed inwards in the process of releasing the compression state of the spring (15), so that the bending element sensor is reset.
4. The soil body moisture content detection method according to claim 2, characterized in that two bending element sensor protection structures (5) are symmetrically arranged on the pipe wall of the probe rod of the pushing hole position of the bending element sensor (6) with the bending element sensor as a center.
5. The method for detecting the water content of the soil body according to claim 2, wherein the inner diameter of the hollow rod is larger than the diameter of the screw rod and can accommodate the cable.
6. The soil body moisture content detection method according to claim 1, wherein the detection method is implemented by using a soil body moisture content detection system based on a bending element sensor, and the detection system comprises a penetration device with a built-in bending element sensor, an electric signal generating module, an electric signal receiving and processing module and a data management platform (S6); after the penetrating device with the built-in bending element sensor reaches the soil body target depth, applying a regular waveform with set voltage and frequency to the transmitting bending element sensor through the electric signal generating module, wherein the piezoelectric effect forces the transmitting bending element to drive surrounding soil bodies to generate transverse vibration, and shear waves generated in the direction perpendicular to the vibration drive the receiving bending element sensor at the other end to vibrate through the soil body, weak currents generated by the receiving bending element sensor are finally transmitted to the electric signal receiving and processing module, and the electric signal receiving and processing module and the data management platform are in data interconnection and intercommunication;
the electric signal generating module is connected with the transmitting bending element sensor through one end of the BNC connecting wire, one end of the electric signal generating module is connected with the electric signal receiving and processing module, the other end of the electric signal receiving and processing module is connected with the receiving bending element sensor through the BNC connecting wire, and the data management platform realizes data interaction with the electric signal receiving and processing module through a 5G network; in addition, the transmitting and receiving bending element sensors are required to be inserted into the same depth of the soil body to be detected;
the electric signal generation module is used for generating an electric signal;
the electric signal receiving and processing module comprises: a signal amplifier (S2), an oscilloscope (S3) and an embedded computer chip (S4); the input of the signal amplifier is connected with the electric signal of the bending element sensor, the oscilloscope is simultaneously connected with the output of the signal amplifier and the output of the signal generator, the output of the oscilloscope is connected with the embedded computer chip, and the embedded computer chip is communicated with the data management platform (S6);
the data management platform is provided with a database related to soil parameters and has the functions of data import, storage, modification, calculation, calling, export and visualization.
7. The detection method can be applied to the water content measurement of soil after the compaction of the on-site roadbed, the road is compacted layer by layer, each layer of soil is compacted layer by layer, the thickness of each layer of soil is different, the depth is different, and the water content prediction at different depths is realized; the detection method is characterized by comprising the following steps:
1) Preparation of test specimens
Collecting different types of soil outdoors, selecting different types of soil, screening, removing impurities, drying, weighing, measuring initial water content and dry density of the soil, and controlling the water content in the soil by changing water adding amount to obtain test samples of different soil with different water contents;
2) Collecting raw data
Aiming at test samples, different vertical pressures are applied to the surfaces of different test samples, the penetration depth of a bending element sensor is changed, the measured shear wave speeds of different test samples at different vertical pressures and different soil penetration depths are used as a set of data, and then multiple sets of data of all the test samples are obtained to form a data set;
3) Constructing an artificial neural network model
Taking the soil body dry density, the soil penetration depth, the shear wave velocity, the soil body type and the vertical pressure as characteristic values, taking the water content as a target value, and performing training on all data sets until the error meets a limit value, wherein the limit value is determined according to the precision required by field measurement, so as to obtain a trained artificial neural network model, namely a shear wave velocity-water content prediction model;
4) Access data management platform
Writing the trained artificial neural network model into an embedded computer chip, and accessing the embedded computer chip into a data management platform, wherein the soil body dry density, the vertical pressure and the soil body type can be obtained from the data management platform, and the soil body water content of the field construction can be obtained by inputting the soil body shear wave velocity and the soil penetration depth measured on the field;
and after the on-site roadbed is compacted, the trained artificial neural network model is applied to predict the water content of the soil body, meanwhile, the water content spot check is periodically carried out, the characteristic value and the spot check true value of the corresponding sample during the water content spot check are recorded, a new data set is formed, and the new data set is transmitted back to the data management platform for expanding the database of the data management platform and updating the artificial neural network model.
8. The method for detecting the water content of the soil body according to claim 7, wherein the process of collecting the original data is as follows:
taking a part of dried soil mass with the mass of M, adding water into the dried soil mass according to a water adding amount calculation formula, and uniformly mixing the dried soil mass to ensure that the water content of the soil mass is 1%;
water adding amount M w Calculation formula
Wherein: w' is the preset water content of the soil body, w 1 The water content of the soil body before water addition;
then two probe rods with built-in bending element sensors (6) are vertically inserted into soil, the same penetration depth is ensured through a graduated scale (4) on the outer wall of the probe, and the penetration depth h is recorded;
two output ends of a signal generator (S5) are respectively connected with a transmitting bending element sensor (6) and an oscilloscope (S3), and a receiving bending element sensor (13) is connected with the oscilloscope (S3); the signal generator (S5) and the oscillograph (S3) are turned on, waveform parameters are input, the oscillograph simultaneously displays output waves and receiving waves, the time difference T between the first wave peaks of the two waveforms is read out, the horizontal distance L between the two probe rods is measured by using a graduated scale, the soil shear wave velocity v is calculated by the following formula,
then uniformly placing weights on the top of a container for holding the soil body, and sequentially applying vertical pressures of 5kPa, 10kPa, 20kPa, 50kPa, 100kPa and 200kPa to the soil body by changing the mass of the weights; under the same vertical pressure, measuring the shear wave speed of the soil body under the water content by changing the penetration depth of the probe rod;
then removing weights sequentially according to the sequence of 200kPa, 100kPa, 50kPa, 20kPa, 10kPa and 5kPa, changing the depth of the probe rod into soil under the same vertical pressure, measuring the shear wave velocity of the soil under the condition of the water content, and calculating to complete the data acquisition work of a water content sample;
adding water into the current soil body according to a water adding amount calculation formula, uniformly mixing, and respectively measuring soil body shear wave velocities with different vertical pressures and depths when the water contents are 2%, 4%, 8%, 16%, 24%, 32%, 48%, 60%, 72%, 84% and 96%;
repeating the steps after the whole process is finished and changing the soil body types until the soil body shear wave speed corresponding to each soil body type, the water content, the dry density, the soil penetration depth and the vertical pressure is measured.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110534793.5A CN113267612B (en) | 2021-05-17 | 2021-05-17 | Soil moisture content detection method, detection system and penetration device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110534793.5A CN113267612B (en) | 2021-05-17 | 2021-05-17 | Soil moisture content detection method, detection system and penetration device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113267612A CN113267612A (en) | 2021-08-17 |
| CN113267612B true CN113267612B (en) | 2023-05-16 |
Family
ID=77231290
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110534793.5A Active CN113267612B (en) | 2021-05-17 | 2021-05-17 | Soil moisture content detection method, detection system and penetration device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113267612B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116229285B (en) * | 2023-05-06 | 2023-08-04 | 深圳大学 | Soil water content monitoring method integrating Internet of things data and space scene |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108007798B (en) * | 2017-10-24 | 2020-03-31 | 湖南大学 | Penetration type soil body strength and wave velocity combined test device and use method thereof |
| CN111105332B (en) * | 2019-12-19 | 2020-09-08 | 河北工业大学 | Highway intelligent pre-maintenance method and system based on artificial neural network |
-
2021
- 2021-05-17 CN CN202110534793.5A patent/CN113267612B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113267612A (en) | 2021-08-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103353462B (en) | A kind of rock nonuniformity method for quantitatively evaluating based on Magnetic resonance imaging | |
| CN103174122B (en) | Lateral stress pore pressure probe used for testing soil static lateral pressure coefficient | |
| WO2009117784A1 (en) | System, apparatus and method for measuring soil moisture content | |
| CN102680413B (en) | Device and method for hyperspectral rapid detection on content of field soil organic matters by using panoramic annular belt | |
| CN109164033B (en) | Method for acquiring hydraulic conductivity of unsaturated soil based on resistivity method | |
| CN113267612B (en) | Soil moisture content detection method, detection system and penetration device | |
| Evett et al. | Neutron scattering, capacitance, and TDR soil water content measurements compared on four continents | |
| Fatehnia et al. | Correlation between shear wave velocity and SPT-N values for North Florida soils | |
| CN206378448U (en) | Polluted Soil detection arrangement of measuring-line structure based on comprehensive geophysical prospecting methods | |
| CN203361120U (en) | Pore pressure static sounding test device | |
| CN202809591U (en) | Static cone penetration system | |
| CN108593525A (en) | A kind of rock three-dimensional permeability real-time determining device and method | |
| CN205981293U (en) | Humiture measurement instrument | |
| CN203101660U (en) | In-situ continuous penetration sounding probe applied to NAPLs contamination sandy soil field survey | |
| Tian et al. | Combined penetrometer and standing wave ratio probe to measure compactness and moisture content of soils. | |
| CN104634317B (en) | Suitable for the apparatus and method of the tunnel surrounding delaminating deposition test of simulated experiment | |
| CN111257947B (en) | Cross-hole resistivity karst cave detection method | |
| Heitor et al. | Assessment of the post-compaction characteristics of a silty sand | |
| Yan et al. | Application of Spatial Time Domain Reflectometry for investigating moisture content dynamics in unsaturated sand | |
| CN113419288A (en) | Data detection and preprocessing method for underground shelter inversion | |
| Xie et al. | Characterizing plant root parameters with deep learning-based heat pulse method | |
| CN208283233U (en) | A kind of rock three-dimensional permeability real-time determining device | |
| Chen et al. | A frequency-domain full waveform inversion method of elastic waves in quantitative defection investigation | |
| DE102021116885B4 (en) | Method for the non-destructive examination of a sample using a radar, in particular geo-radar device, radar device, computer program and computer program product | |
| CN109238369A (en) | A kind of stress and strain testing method based on MEMS sensor |
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 |