A distributed 3D visualization online monitoring system for concrete dam pouring
The distributed 3D visualization online monitoring system for concrete dam pouring utilizes sensing optical fibers and BOTDA devices to demodulate Brillouin backscattered light signals, combined with the LSTM algorithm, to resolve the contradiction between measurement range and accuracy in concrete dam pouring monitoring, achieving efficient and global temperature and strain monitoring and prediction.
Patent Information
- Application Number
- CN202411135104.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-08-19
AI Technical Summary
Existing concrete dam pouring monitoring technologies suffer from limited measurement range, a trade-off between accuracy and cost, and difficulty in achieving comprehensive information acquisition.
A distributed 3D visualization online monitoring system for concrete dam pouring was adopted. The system uses sensing optical fibers tied to the reinforcing bars, combined with BOTDA devices and a data monitoring center. Through Brillouin backscattered light signal demodulation, a 3D BIM engine and LSTM algorithm were integrated to perform real-time monitoring and prediction of temperature and strain.
It achieves temperature and strain monitoring with strong anti-interference ability, wide measurement range and high accuracy, reduces costs, and can predict future trends, providing three-dimensional visualization and analysis early warning.
Smart Images

Figure CN118654778B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to optical fiber sensing and three-dimensional visualization online monitoring technology, and in particular to a distributed three-dimensional visualization concrete dam pouring online monitoring system. Background Art
[0002] The pouring of concrete dams usually faces the problem of temperature control of large-volume concrete. The importance of temperature control of large-volume concrete is mainly reflected in preventing temperature cracks in concrete and ensuring the safety and normal use of building structures. Due to the widespread application of large-volume concrete technology, its volume is large, the surface coefficient is small, the cement hydration heat release is more concentrated, and the internal temperature rises faster, which will lead to a large temperature difference between the inside and outside of the concrete. When this temperature difference is large, concrete is prone to temperature cracks, which affects the safety and normal use of the structure. Therefore, good temperature control of large-volume concrete is the key to preventing temperature cracks.
[0003] At present, the monitoring technology of concrete dam pouring has shifted from traditional monitoring methods to multidisciplinary automated monitoring technology. This technological transformation reflects the development direction of automation, digitalization, expert evaluation, structural integration and efficiency. In the process of concrete dam pouring, monitoring equipment and technology play a vital role and can fully and accurately reflect the performance and status of the dam body.
[0004] In previous monitoring, temperature and strain information is usually arranged at discrete points with a limited perception range. To expand the perception range, the only way is to increase the arrangement of monitoring points, which makes it difficult to obtain global information. Summary of the invention
[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to provide a distributed three-dimensional visualization concrete dam pouring online monitoring system with real-time efficiency, strong anti-interference ability and wide measurement range, which better solves the current contradiction between measurement accuracy, range and cost.
[0006] A distributed three-dimensional visualized concrete dam pouring online monitoring system, the system includes a data monitoring center, a BOTDA device, an optical switch and a sensor optical fiber; the sensor optical fiber is tied to the steel bars of the concrete dam and laid back and forth along the plane in a folded line, and one end extending out of the concrete dam is connected to the optical switch through an optical port; the optical switch is connected to the BOTDA device through an optical port, and the BOTDA device is connected to the data monitoring center through a network;
[0007] The BOTDA device emits narrowband laser light, which is reflected to the BOTDA device after passing through the sensing optical fiber. The backscattered light frequency shift data is detected by demodulating the Brillouin backscattered light signal, and the temperature and strain data along the sensing optical fiber are analyzed at the same time. These data are then sent to the host computer system of the data monitoring center through the network.
[0008] The BOTDA device includes an optical subsystem, a circuit subsystem and a measurement software subsystem, wherein:
[0009] The optical subsystem includes an optical transmitting module and an optical receiving module. The optical transmitting module includes a semiconductor laser, a first electro-optic modulator, an EDFA and an acousto-optic modulator AOM; the optical receiving module includes a polarization scrambler, a grating filter, a photodetector and a second electro-optic modulator;
[0010] The circuit subsystem includes a drive and temperature control module, a signal receiving module, and a signal processing and power supply module; the drive and temperature control module includes an EOM pulse generator, an AOM pulse generator, a microwave signal source, and a laser temperature control module; the signal receiving module includes a low-noise amplifier module, a bandpass filter, an automatic gain module, and a high-speed acquisition module; the signal processing and power supply module includes an FPGA, a network communication module, and a power management module;
[0011] The measurement software subsystem is an industrial computer, which is controlled by the measurement software, and the measurement software realizes the demodulation of the Brillouin backscattered light signal;
[0012] The demodulation of the Brillouin backscattered light signal specifically comprises the following steps:
[0013] S1, obtaining the Brillouin backscattered light signal data distributed along the optical fiber, and preprocessing the data using a data accumulation weighted average algorithm and a data clustering algorithm;
[0014] S2, using a double Gaussian fitting algorithm to perform peak fitting of the Brillouin spectrum on the data obtained in step S1, analyzing and obtaining temperature and strain data distributed along the optical fiber, and storing the analyzed temperature and strain data in a database.
[0015] The host computer system of the monitoring center uses a three-dimensional BIM engine and WebGL technology to build a three-dimensional model of the concrete dam, arranges the sensor optical fiber on the three-dimensional model, integrates the temperature and strain information collected by the BOTDA device, predicts the temperature and strain information of the concrete dam during operation through the LSTM algorithm, and conducts comprehensive display and analysis and early warning;
[0016] The LSTM algorithm predicts the temperature and strain information of the concrete dam during operation, specifically including the following steps:
[0017] S1, obtaining factors affecting the temperature and strain of the concrete dam during operation, wherein the data include water temperature before and after the dam, ambient temperature, and water level before and after the dam;
[0018] S2, standardize the influencing factors to obtain normalized data;
[0019] S3, taking the normalized influencing factor data as the input variables of the prediction model, taking the temperature and strain of the concrete dam as the output variables, and dividing them into a training set and a test set;
[0020] S4, input the training set into the LSTM prediction model, and model the data in the training set through the LSTM prediction model; input the water temperature in front of and behind the dam, the ambient temperature, and the water level parameters in front of and behind the dam at time t into the trained LSTM prediction model, and obtain the predicted values of the concrete dam temperature and strain at time t+1, which is the output of the model.
[0021] The sensing optical fiber is installed in different concrete dam sections respectively. The optical fiber is a single-mode optical fiber and is armored. The optical fiber is used as both a transmission element and a sensing element to realize distributed measurement of temperature and strain.
[0022] The BOTDA device realizes multi-channel measurement of the sensing optical fiber through optical switch switching.
[0023] The measurement software subsystem also includes a display for displaying the demodulation result of the Brillouin backscattered light signal.
[0024] In the present invention, the realization of the optical subsystem is:
[0025] The BOTDA device optical path polarization-maintaining coupler C1 divides the continuous light emitted by the semiconductor laser LD into two paths, one of which is modulated into a pump pulse light by the first electro-optic modulator and the acousto-optic modulator AOM, and then amplified by the fiber erbium-doped amplifier (EDFA). The grating filter FBG2 filters out the spontaneous radiation noise of the EDFA and connects to port 1 of the optical circulator; the other light is modulated into a continuous detection light by the second electro-optic modulator (EOM). The microwave module provides the modulation signal for the EOM. The continuous detection light produces a frequency shift. PS is a polarization scrambler used to change the polarization state of the detection light. The grating filter FBG1 is used for optical filtering. The filtered continuous detection light enters the single-mode sensing fiber (SM fiber), the pump pulse light and the continuous detection light propagate in opposite directions in the single-mode sensing fiber. When the frequency of the detection light falls within the Brillouin gain spectrum, the detection light is amplified. Port 3 of the circulator is connected to the photodetector PD, which converts the amplified continuous detection light signal into an electrical signal and then sends it to the signal processing part. The signal processing part includes analog signal processing and digital signal processing. The measurement software displays the obtained temperature and strain information distributed along the optical fiber.
[0026] By using the same laser and coupler to split the light to generate pump light and probe light, the influence of unstable light source power and wavelength drift of the two lasers can be reduced. The polarization scrambler is used to expand the probe light power to the entire polarization state, reducing the fluctuation and attenuation of the signal power.
[0027] In this embodiment, the circuit subsystem is implemented as follows:
[0028] Driving and temperature control module: The EOM pulse generator is used to drive the first electro-optic modulator in the optical transmitting module, the AOM pulse generator is used to drive the acousto-optic modulator in the optical transmitting module, the microwave signal source is used to drive the second electro-optic modulator in the optical receiving module, and the laser temperature control module is used to keep the temperature of the semiconductor laser constant; in the circuit subsystem, the central wavelength of the laser is related to the temperature. If drift occurs, the pulse light will be attenuated when passing through the grating filter, and the pump light power will be weakened, affecting the measurement of the system. In order to maintain the stability of the central wavelength of the laser, the laser temperature control module is used to keep the temperature of the semiconductor laser constant;
[0029] Signal receiving module: The low-noise amplifier module preliminarily amplifies the electrical signal converted by the photodetector, uses a bandpass filter to filter out noise, improves the signal-to-noise ratio of the signal, and finally the high-speed acquisition module collects the analog electrical signal; the central wavelength of the grating filter FBG will drift with changes in temperature, and the central wavelength of the FBG with temperature compensation package is always consistent with the wavelength of the pump light.
[0030] Signal processing and power supply module: FPGA is used to generate precise timing signals. The width of the pump pulse signal is controlled by FPGA and is used to synchronize the pulse drive signals of EOM1 and AOM. The high-speed collected Brillouin backscattering signal data is digitally accumulated and averaged by FPGA to improve the signal-to-noise ratio of the collected signal. The network communication module transmits instructions to control data collection and transmits the collected data to the measurement software subsystem for processing and display. The power management module provides power management functions for all unit modules and has over-power protection function. It cuts off the power supply to protect the device when the circuit is overloaded or short-circuited.
[0031] In the present invention, optical fibers can be distributed along the steel structure as sensing elements. Any point on the continuous optical fiber can be regarded as a separate "sensor", which can provide a large amount of monitoring data and achieve comprehensive coverage of the monitoring area. At the same time, distributed optical fiber sensing technology has higher spatial resolution and measurement accuracy, and can more accurately reflect the physical state of the object being measured.
[0032] The present invention has the following advantages:
[0033] 1. The present invention uses optical fiber as a sensing element to measure concrete temperature and strain information, which is completely different from traditional temperature sensors based on electrical signals and point optical fiber temperature sensors. It has the characteristics of strong anti-interference ability, large measurement range and high measurement accuracy. Traditional measurement methods can only perform point detection. Optical fiber sensors use Brillouin light frequency difference as the sensing quantity, can measure tiny strains, and can measure the temperature and strain values of each point on the entire continuous optical fiber in the form of a distance continuous function.
[0034] 2. The present invention uses ordinary optical fiber as the sensing element. Only one optical fiber is needed to measure the temperature and strain at a distance of tens of kilometers. One host can also measure several optical fibers at the same time, which reduces the data monitoring cost and improves the benefit-cost ratio.
[0035] 3. The optical cable buried in the concrete pouring process of the present invention can be used in the construction process of the concrete dam and can also be used in the operation process of the concrete dam. As a passive component, it has high reliability.
[0036] 4. The present invention utilizes the principle of Brillouin sensing technology to solve the problem that traditional optical fiber sensing can only measure temperature. It can measure temperature and strain at the same time, increasing the perception dimension of data.
[0037] 5. The present invention utilizes a double Gaussian demodulation algorithm to solve the Brillouin frequency shift, which makes the Brillouin frequency shift measurement more accurate and effectively improves the measurement accuracy of temperature and strain.
[0038] 6. The present invention establishes a three-dimensional model of the concrete dam, integrates temperature and strain monitoring data, and can predict the changing trends of temperature and strain in the future, thereby realizing three-dimensional visualization and analysis and early warning of the concrete dam, and providing auxiliary support for decision-making. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 It is the overall structural block diagram of the system in the present invention.
[0040] Figure 2 It is a structural block diagram of the BOTDA device in the present invention.
[0041] Figure 3 This is a diagram of the optical path structure of the BOTDA device in the present invention.
[0042] Figure 4 This is a flow chart of the K-Means clustering algorithm in the present invention.
[0043] Figure 5 This is a flow chart of the double Gaussian fitting algorithm in the present invention.
[0044] Figure 6 This is a flow chart of the LSTM prediction algorithm in the present invention.
[0045] Figure 7 This is a schematic diagram of the installation and layout of the sensor optical cable in the present invention.
[0046] Figure 8 This is a temperature measurement diagram of the 2.5 km sensing optical cable in the present invention.
[0047] Fig. 9 This is the effect diagram of the double Gaussian fitting algorithm in the present invention.
[0048] Fig.10 This is the iterative training diagram of the LSTM algorithm model in the present invention.
[0049] Fig.11 It is a comparison chart of the predicted value and the actual value of concrete temperature in the present invention. DETAILED DESCRIPTION
[0050] The following is a detailed and specific description of a distributed three-dimensional visualized concrete dam pouring online monitoring system of the present invention in conjunction with the accompanying drawings and embodiments.
[0051] Embodiment 1: A distributed three-dimensional visualized concrete dam pouring online monitoring system, the system includes a data monitoring center, a BOTDA device, an optical switch and a sensor optical fiber; the sensor optical fiber is tied to the steel bars of the concrete dam and laid back and forth along the plane in a folded line, and one end extending out of the concrete dam is connected to the optical switch through an optical port; the optical switch is connected to the BOTDA device through an optical port, and the BOTDA device is connected to the data monitoring center through a network;
[0052] The BOTDA device emits narrowband laser, which is reflected to the BOTDA device after passing through the sensing optical fiber, and the backscattered light frequency shift data is detected by demodulating the Brillouin backscattered light signal, and the temperature and strain data along the sensing optical fiber are analyzed at the same time, and then these data are sent to the host computer system of the data monitoring center through the network; the BOTDA device includes an optical subsystem, a circuit subsystem and a measurement software subsystem, wherein,
[0053] The optical subsystem includes an optical transmitting module and an optical receiving module. The optical transmitting module includes a semiconductor laser, a first electro-optic modulator, an EDFA and an acousto-optic modulator AOM; the optical receiving module includes a polarization scrambler, a grating filter, a photodetector and a second electro-optic modulator;
[0054] The circuit subsystem includes a drive and temperature control module, a signal receiving module, and a signal processing and power supply module; the signal receiving module is mainly used for signal amplification and filtering to improve the signal-to-noise ratio of photoelectric conversion; the drive and temperature control module includes an EOM pulse generator, an AOM pulse generator, a microwave signal source, and a laser temperature control module; the signal receiving module includes a low-noise amplifier module, a bandpass filter, an automatic gain module, and a high-speed acquisition module; the signal processing and power supply module includes an FPGA, a network communication module, and a power management module;
[0055] The measurement software subsystem includes an industrial computer and a display. The industrial computer is controlled by the measurement software, which realizes the demodulation of the Brillouin backscattered light signal; controls the start and end of data acquisition, data analysis and processing, data storage, log recording, algorithm fitting, waveform curve display, controls the power management module, and sets system-related parameters.
[0056] The optical switch adopts ZG-mems optical switch of Ziguan Company, with an operating wavelength range of 1260-1650nm, an optical interface type of FC / APC / flange interface, and an optical port of 8 channels;
[0057] The semiconductor laser is a semiconductor laser produced by Beijing Better Optical Communication Technology Co., Ltd., which is an ultra-narrow linewidth semiconductor laser light source with a central wavelength of 1550nm and a spectral linewidth of 1KHz.
[0058] The first and second electro-optic modulators (EOMs) are MXER-LN series intensity electro-optic modulators from IXBLUE, with an extinction ratio of more than 35dB;
[0059] The EDFA adopts the KY-EDFA-BI series erbium-doped fiber amplifier produced by Beijing Keyang Optoelectronics;
[0060] The acousto-optic modulator (AOM) is a 1550nm fiber acousto-optic modulator SGTF200-1550 produced by Chongqing Shangmao Technology Development Co., Ltd.
[0061] The polarization scrambler (PS) is PSC-LN-10 produced by Suzhou Bofu Optoelectronics Technology Co., Ltd.
[0062] The grating filter (FBG) adopts the ultra-narrow bandpass filter module of Shanghai Haoliang Optoelectronic Equipment Co., Ltd., which is a symmetrical flat filter with a bandwidth as low as 1GHz (8pm);
[0063] The photoelectric detector adopts PQW10A-L produced by Suzhou Bofu Optoelectronics Technology Co., Ltd., and the frequency can reach DC-40GHz;
[0064] The microwave signal source adopts the 3910 series RF driver of Shanghai Haoliang Optoelectronic Equipment Co., Ltd., with a rise time as low as 4ns, adjustable RF power, and optional external clock synchronization;
[0065] The EOM1 and EOM2 pulse generators adopt the pulse selector SmartPicker-EOM launched by Shanghai Haoliang Optoelectronics;
[0066] The AOM pulse generator adopts the pulse selector SmartPicker AOM launched by Shanghai Haoliang Optoelectronics;
[0067] The laser temperature control module adopts the ONLD low-noise laser driver of Wuhan Optical Valley Interconnect Technology Co., Ltd.
[0068] The low noise amplifier module adopts the THS4304 chip of Texas Instruments;
[0069] The bandpass filter adopts the BFCN-1445 product of Fuzhou Weibo Telecommunications Co., Ltd.;
[0070] The automatic gain module adopts the amplifier PGA849 chip of Texas Instruments;
[0071] The high-speed acquisition module adopts the high-speed data acquisition chip ADC32RF54 of Texas Instruments;
[0072] The FPGA data acquisition module adopts Intel® Stratix® 10 NX 2100 FPGA and has 2,073,000 logic units;
[0073] The network communication module adopts the DM9161BI / CI network communication chip of DAVICOM;
[0074] The power management module adopts the power management module designed by TPSM861257, TPSM861252, and TPSM828302 chips of Texas Instruments.
[0075] The demodulation of the Brillouin backscattered light signal adopts a data accumulation weighted average algorithm and a data clustering algorithm to pre-process the data; a double Gaussian fitting algorithm is used to perform peak fitting of the Brillouin spectrum on the pre-processed data, and the temperature and strain data distributed along the optical fiber are obtained by analysis, and the analyzed temperature and strain data are stored in a database;
[0076] The specific implementation steps of the K-means algorithm for data cumulative weighted average are as follows, where K=2, that is, the collected data sets are divided into 2 categories:
[0077] S1, randomly select the initial centers of k classes from the n data spaces that have been accumulated and averaged;
[0078] S2, in each iteration, calculates the distance between each object and the center object calculated last time, and then classifies each object into the class with the shortest distance;
[0079] S3, use the mean method to recalculate and update the central object of each class;
[0080] S4, for k cluster centers, use steps S2 and S3 to update and iterate each cluster center, calculate the standard measurement function, if the function converges, the iteration ends, otherwise continue the iteration;
[0081] S5, after the data is classified by 2-means algorithm, the data is divided into two groups. For each group of data P∈[(x1,y1),...,(x m ,y m )] and Q∈[(x1,y1),...,(x n ,y n )] Each of them fits the Gaussian function f(x)=a*exp(-(xb)² / (2c²)) to obtain the Gaussian function f 1 ( x )and f 2 ( x ), find the mean y of arrays P and Q respectively pk and qk , calculate the weighted coefficient p 1. p 2, where P1=y pk / (y pk +y qk) 、P2=y qk / (y pk +y qk) , and then get the fitting function of the whole curve ,like Figure 5 shown.
[0082] The host computer system of the monitoring center uses a three-dimensional BIM engine and WebGL technology to build a three-dimensional model of the concrete dam, arranges the sensor optical fiber on the three-dimensional model, integrates the temperature and strain information collected by the BOTDA device, predicts the temperature and strain information of the concrete dam during operation through the LSTM algorithm, and conducts comprehensive display and analysis and early warning;
[0083] The monitoring center predicts the temperature and strain of the concrete dam based on the LSTM algorithm, including the following steps:
[0084] Obtain the factors that affect the temperature and strain of the concrete dam during operation, wherein the data include the water temperature before and after the dam, the ambient temperature, and the water level before and after the dam; standardize the influencing factors to obtain normalized data; use the normalized influencing factor data as the input variables of the prediction model, and the temperature and strain of the concrete dam as the output variables, and divide them into a training set and a test set according to the data volume of 6:4; input the training set into the LSTM prediction model, and model the data in the training set through the LSTM prediction model; input the water temperature before and after the dam, the ambient temperature, and the water level parameters before and after the dam at time t into the trained LSTM prediction model, and obtain the predicted values of the temperature and strain of the concrete dam at time t+1, which is the output of the model;
[0085] Furthermore, during the training process, the number of input nodes, the number of hidden layer neurons and the output results of the LSTM prediction model are determined, and the output temperature or strain of the network is obtained by multiplying the outputs of the tanh and sigmoid functions. In each layer, the BPTT algorithm is used to improve the learning ability of the LSTM prediction model for time series. The log-likelihood function is selected as the loss function, the number of neurons in the input layer is selected as 5, and the input time series variables are {x1, x2, x3, …, x T}, where the input time series of the neuron at sampling time t is x t =[I t (1),I t (2),…,I t (5)]; the output variable is temperature; the neural network selects 2 layers of LSTM, and finally connects a fully connected layer and passes through an activation function as the final output.
[0086] Furthermore, the LSTM prediction model includes a forget gate, an input gate and an output gate, wherein the forget gate layer reads h t-1 and x t , output a value f between 0 and 1 t ,Right now
[0087] f t =σ(W xf x t +W hf h t -1+b f );
[0088] Input gate layer, with the input x of the current layer t and the output h of the hidden unit at the previous moment t-1 As input, the output result i t As the information to be updated, i.e.
[0089] i t =σ(W xi xi +W hi h t-1 +b i );
[0090] Then the cell state is updated, and the new cell state C t It is determined by the old state and the output results of the forget gate and the input gate, that is,
[0091] ;
[0092] ;
[0093] The output gate layer runs a sigmoid layer to determine which part of the cell state will be output. Then, the cell state is processed through a tanh and multiplied by the output of the sigmoid gate as the output, that is,
[0094] O t =σ(W xo x t +W ho h t-1 +b o );
[0095] H t =O t *tanh(C t );
[0096] Where σ and tanh are activation functions; Wx represents the weight from the input layer to the hidden layer; Wh represents the weight of the recursive connection;
[0097] b represents the bias term.
[0098] Furthermore, in order to evaluate the deviation between the predicted value and the actual value during the training process, the root mean square error is selected as the loss function to evaluate the training effect; the neural network parameters are continuously trained and optimized with the minimum loss function value as the goal until the root mean square error is less than the set threshold and the training model is saved. The calculation formula of the root mean square error is:
[0099] ;
[0100] Where n represents the number of samples, yi and yi′ are the actual and predicted values of temperature or strain at time i.
[0101] Embodiment 2: A distributed three-dimensional visualization concrete dam pouring online monitoring system, such as Figure 7As shown in the figure, the sensing optical fiber is laid along the steel bars on a concrete dam. The sensing optical fiber is tied to the steel bars and laid back and forth in a zigzag line along the plane. A layer of optical cable is laid every 1 meter in elevation, and then the optical cable is led up to the upper layer for laying. The laying is cyclical until the optical cable is stretched out of the dam surface. The sensing optical fiber is then connected to the BOTDA device. The BOTDA device emits a pulse laser signal to the optical fiber buried in the concrete. When the optical signal propagates in the optical fiber, it is affected by the optical fiber material to generate a backscattered optical signal. When the temperature and strain of the concrete change, the optical fiber microstructure changes, causing the central wavelength of the backscattered light to shift. By detecting the return time of the backscattered light, the position where the temperature and strain change can be calculated, the spatial positioning of the optical fiber in the concrete dam can be realized, and the temperature and strain at each position along the optical fiber can be demodulated. The BOTDA device transmits the obtained temperature and strain information to the host computer system of the monitoring center through the network, and completes the data storage, management, calculation, result display and analysis alarm, so as to achieve the purpose of real-time monitoring of the distributed temperature and strain of the concrete dam.
[0102] The temperature of the concrete dam section measured by the 2.5 km sensor optical cable is about 25°C with a resolution of 1°C. Figure 8 shown.
[0103] A set of data was taken from the 2.5 km sensor fiber data set. According to the measurement time and measurement accuracy requirements, a total of 10 frequency points were collected for each spatial position. The obtained data were divided into two categories for fitting. The specific steps are as follows:
[0104] S1, using the data cumulative weighted average algorithm to process the data, the initial centers of the two classes are the first data and the second data (corresponding to Fig. 9 The horizontal coordinates are x=10.76 and x=10.772), and then the distances of each point to the two initial centers are calculated. If the distance to the initial center of x=10.76 is close, it is classified as the same class as x=10.76. After completing this step, the means of the two classes are calculated, the difference between the two means and the initial center is calculated, and the two means are used as the new centers of the two classes. Then the above process is repeated until the difference between the mean and the center of the new class data is less than the set value. The classification after the iteration is completed is shown in Table 1 below. The map value in Table 1 indicates the class to which the 10 data belong. The value 1 represents the first class of data, and the value 2 represents the second class of data. The four points x=10.76, 10.808, 10.820, and 10.832 are classified as the second class, and the remaining points are classified as the first class. The final two cluster centers are 63.7220 and 227.1473;
[0105] Table 1 Grouping after clustering algorithm
[0106] X / Frequency 10.76 10.772 10.784 10.796 10.808 10.820 10.832 10.844 10.856 10.868 Y / Brillouin gain 31.495 37.858 56.858 98.36 200.19 288.702 192.55 94.056 52.722 42.478 Map / Grouping 2 1 1 1 2 2 2 1 1 1
[0107] S2, after the data classification is completed, Gaussian fitting is performed on the data in each class, and the size of the fitted center frequency is 10.8197GHz, as shown in Fig. 9 As shown;
[0108] S3, the prediction model based on the LSTM algorithm is established in the Python 3.10.0 environment and function libraries such as Keras, Scikit-learn, and Tensorflow; the state monitoring information in the training sample set is used as the input of the network, and the network model is trained offline. First, the parameters are initialized, the learning step size is set to 0.02, the input dimension is 5, the number of output layer nodes is 1, all elements of the weight parameter matrix are initialized to (-1,1), and the error limit is set to 1×10 -2 , set the number of iterations to 250, perform offline training on the samples, and thus obtain the temperature prediction model;
[0109] like Fig.10 As shown in the figure, taking the temperature change of a concrete dam over a period of time as an example, the loss value of the training data set converges very quickly. After 50 iterations, the loss value is close to the expected level. After 100 iterations, the loss value stabilizes at the lowest level of 0.0002, meeting the minimum error requirement. The error of all training samples is less than the threshold.
[0110] like Fig.11 As shown in the figure, the actual value and predicted value of the concrete dam temperature during the operation period. It can be seen that the waveforms and trends of the actual value and the predicted value are basically consistent, the degree of overlap is high, and the prediction has a high accuracy.
Claims
1. A distributed 3D visualization concrete dam pouring online monitoring system, characterized by The system includes a data monitoring center, a BOTDA device, an optical switch and a sensor optical fiber; the sensor optical fiber is tied to the steel bars of the concrete dam and laid back and forth along the plane in a zigzag pattern, and one end extending out of the concrete dam is connected to the optical switch through an optical port; the optical switch is connected to the BOTDA device through an optical port, and the BOTDA device is connected to the data monitoring center through a network; The BOTDA device emits narrowband laser, which is reflected to the BOTDA device after passing through the sensing optical fiber, and the backscattered light frequency shift data is detected by demodulating the Brillouin backscattered light signal, and the temperature and strain data along the sensing optical fiber are analyzed at the same time, and then these data are sent to the host computer system of the data monitoring center through the network; the BOTDA device includes an optical subsystem, a circuit subsystem and a measurement software subsystem, wherein, The optical subsystem includes an optical transmitting module and an optical receiving module. The optical transmitting module includes a semiconductor laser, an electro-optical modulator 1, an EDFA and an acousto-optic modulator AOM; the optical receiving module includes a polarization scrambler, a grating filter, a photodetector and an electro-optical modulator 2; The circuit subsystem includes a drive and temperature control module, a signal receiving module, and a signal processing and power supply module; the drive and temperature control module includes an EOM pulse generator, an AOM pulse generator, a microwave signal source, and a laser temperature control module; the signal receiving module includes a low-noise amplifier module, a bandpass filter, an automatic gain module, and a high-speed acquisition module; the signal processing and power supply module includes an FPGA, a network communication module, and a power management module; The measurement software subsystem includes an industrial computer and a display. The industrial computer is controlled by the measurement software, which realizes the demodulation of the Brillouin scattered backscattered light signal. The demodulation of the Brillouin backscattered light signal specifically comprises the following steps: S1, obtaining the Brillouin backscattered light signal data distributed along the optical fiber, and preprocessing the data using a data accumulation weighted average algorithm and a data clustering algorithm; S2, using a double Gaussian fitting algorithm to perform peak fitting of the Brillouin spectrum on the data obtained in step S1, analyzing and obtaining temperature and strain data distributed along the optical fiber, and storing the analyzed temperature and strain data in a database; The host computer system of the monitoring center uses a three-dimensional BIM engine and WebGL technology to build a three-dimensional model of the concrete dam, arranges the sensor optical fiber on the three-dimensional model, integrates the temperature and strain information collected by the BOTDA device, predicts the temperature and strain information of the concrete dam during operation through the LSTM algorithm, and conducts comprehensive display and analysis and early warning; The LSTM algorithm predicts the temperature and strain information of the concrete dam during operation, specifically including the following steps: S1, obtaining factors affecting the temperature and strain of the concrete dam during operation, wherein the data include water temperature before and after the dam, ambient temperature, and water level before and after the dam; S2, standardize the influencing factors to obtain normalized data; S3, taking the normalized influencing factor data as the input variable of the prediction model, taking the temperature and strain of the concrete dam as the output variables, and dividing them into a training set and a test set according to a certain ratio; S4, input the training set into the LSTM prediction model, and model the data in the training set through the LSTM prediction model; input the water temperature in front of and behind the dam, the ambient temperature, and the water level parameters in front of and behind the dam at time t into the trained LSTM prediction model, and obtain the predicted values of the concrete dam temperature and strain at time t+1, which is the output of the model.
2. A distributed three-dimensional visualization concrete dam pouring online monitoring system as claimed in claim 1, characterized in that The sensing optical fiber is installed in different concrete dam sections respectively. The optical fiber is a single-mode optical fiber and is armored. The optical fiber is used as both a transmission element and a sensing element to realize distributed measurement of temperature and strain.
3. A distributed three-dimensional visualization concrete dam pouring online monitoring system as claimed in claim 1, characterized in that The BOTDA device realizes multi-channel measurement of the sensing optical fiber through optical switch switching.
