Intelligent road settlement monitoring method

Abstract

The invention discloses an intelligent road settlement monitoring method, and belongs to the technical field of road engineering. Sensors are arranged along a to-be-monitored road section, data collected by the sensors are transmitted to an Internet of Things platform for preprocessing and settlement prediction, the Internet of Things platform deploys a combined model based on two different roadbed settlement prediction models, the preprocessed monitoring data are predicted through the combined model, settlement prediction data are obtained, and the settlement prediction data are subjected to settlement prediction. And intelligent road settlement monitoring is realized. According to the method, an alarm can be given for abnormal states and data in the settlement process based on the obtained settlement prediction data according to needs. According to the invention, the technologies of Internet of Things, artificial intelligence and the like are applied to project engineering, so that the safe operation of engineering project work is ensured to a certain extent, and the safety prevention and control capability of the engineering project is improved.

Description

Technical Field

[0001] The invention relates to the technical field of highway engineering, and in particular to an intelligent monitoring method for highway settlement. Background Art

[0002] In soft foundation highway projects, post-construction settlement of the roadbed and excessive uneven settlement are the main roadbed diseases, so it is very important to monitor the settlement.

[0003] At present, the main problems of settlement monitoring of soft foundation highway projects are: 1. The degree of informatization of construction site monitoring means is not high, and the monitoring method is mainly based on manual monitoring, which has the disadvantages of limited application scope, low monitoring frequency, great influence of human factors and environment, and high labor cost. 2. Insufficient attention is paid to the effective analysis of on-site monitoring data and post-construction monitoring, and the monitoring data cannot be used for effective dynamic design. Therefore, how to conduct real-time, accurate and intelligent settlement monitoring of highways is the core of the problem. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a method for intelligent monitoring of highway subsidence. By utilizing technical means such as the Internet of Things, big data, and artificial intelligence, an intelligent monitoring system for subsidence is developed, which has the functions of full-process information collection, data analysis, real-time monitoring, and abnormal alarm, so as to realize real-time, accurate, and intelligent monitoring of highway subsidence.

[0005] In order to achieve the above technical objectives, the present invention provides a method for intelligent monitoring of highway subsidence, which comprises the following steps:

[0006] Step S1, deploying sensors along the road section to be monitored, wherein the sensors are connected to respective data collectors via RS485 cables;

[0007] Step S2, the data collector is connected to the DTU via wireless communication; the monitoring data is wirelessly transmitted to the DTU;

[0008] Step S3, DTU transmits the data to the IoT platform via wireless communication, and the IoT platform pre-processes the data and performs visualization and early warning configuration;

[0009] Step S4, establishing a Poisson curve settlement prediction model and a settlement prediction model based on the Bayesian update method of Markov chain Monte Carlo; combining the two settlement prediction models to obtain a combined prediction model;

[0010] Step S5, embedding the combined prediction model into the Internet of Things platform, predicting the monitoring data preprocessed in step S3, obtaining settlement prediction data, and realizing intelligent monitoring of highway settlement.

[0011] Compared with the prior art, the present invention has at least the following advantages:

[0012] (1) Compared with conventional monitoring methods, the present invention uses Internet of Things technology to monitor highway settlement, which has a higher level of informatization, a wider range of applicability, and a higher monitoring frequency. Data collection can be carried out throughout the entire process of highway engineering. At the same time, it can timely and effectively feedback abnormal data and issue alarms, which can effectively ensure the safe operation of engineering projects.

[0013] (2) The present invention adds an artificial intelligence model, that is, it proposes to combine the Poisson curve settlement prediction model with the settlement prediction model based on the Bayesian update method of Markov chain Monte Carlo to achieve the prediction of highway settlement. A single model can often only take into account the error of a certain part of the settlement data sequence, and does not consider the overall error enough. Therefore, the above two roadbed settlement prediction models are combined to improve the prediction accuracy. Therefore, preventive measures can be taken in advance according to the prediction results, effectively avoiding potential risks and improving the safety prevention and control capabilities.

[0014] (3) The IoT platform of the present invention is deployed based on an open source platform, which reduces development and maintenance costs, facilitates application in various scenarios, and improves efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The figure is a flow chart of the intelligent monitoring method for highway subsidence described in an example of the present invention.

[0016] Figure 2 Schematic diagram of the calculation results of the layered summation method described in the embodiment of the present invention.

[0017] Figure 3 Schematic diagram of the data preprocessing process in an embodiment of the present invention.

[0018] Figure 4 This is a schematic diagram of data visualization described in an embodiment of the present invention.

[0019] Figure 5 This is a diagram of the steps for establishing a highway settlement prediction model based on the Bayesian update method of Markov Chain Monte Carlo in an embodiment of the present invention.

[0020] Figure 6 Schematic diagram of the process of embedding an artificial intelligence data analysis module in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. Note that the following embodiments are merely illustrative in nature, and the present invention is not intended to limit its applicable objects or uses, and the present invention is not limited to the following embodiments.

[0022] Reference Figure 1As shown, the embodiment of the present invention provides a method for intelligent monitoring of highway settlement, including:

[0023] Step S1, deploying sensors along the road section to be monitored, wherein the sensors are connected to respective data collectors via RS485 cables;

[0024] The present invention is based on a certain highway section expansion project, and intelligently monitors the settlement of the highway section. The test section of the project is 56m long, the embankment base is 20m wide, 3.6m high, and the slope is 1:1.5. The soil layers of the highway section are clay layer, silt layer, clay layer and gravel-containing silty clay layer from top to bottom, totaling 28m, and cement mixing piles with a diameter of 0.7m are set every 1.2m below the embankment base. The construction sequence of the highway section is as follows: First, cement mixing pile construction is carried out, the construction time is 2 days, and the construction is stopped for 30 days. After that, the embankment is filled, and the embankment is filled in two layers. The lower layer is a crushed stone cushion layer with a filling height of 0.5m and a filling time of 15 days. Geogrids are laid 0.3m above the base, and the upper layer is fill soil with a filling height of 3.1m and a filling time of 95 days. After the filling is completed, preloading is carried out for 8 months.

[0025] The deepest borehole in the soft soil layer of a typical cross section of the project was selected as the research object to determine the range of the sensor. The monitoring depth of the multi-point displacement meter was determined based on the results of the borehole sampling survey to determine the depth of the soft soil layer. According to the deepest borehole depth of the soft soil layer in the cross section, the monitoring depth of the multi-point displacement meter was selected as 28m below the top surface of the roadbed. Four displacement monitoring points were evenly set along the depth direction to measure the layered compression; the monitoring range of the static level was determined based on the calculation results of the layered summation method. After calculation, the settlement was found to be 0.34m. The calculation results are as follows: Figure 2 Therefore, the maximum range of the static level is selected to be 0.4m.

[0026] According to the length of the project test section, a settlement monitoring profile is set up every 14m along the highway. Four multi-point displacement meters and four static levels (three multi-point displacement meters and three static levels in each profile) are buried on both sides of the roadbed and the center line of the roadbed in each settlement monitoring profile, and they are connected to the data collector through RS485 cable communication. In addition, four settlement plates are set up on both sides of the earth shoulder of the same section, and the total road surface settlement is manually observed once a month.

[0027] In step S2, the data collector is connected to a DTU (Data Transfer Unit, wireless terminal equipment) by wireless communication, and the monitoring data is wirelessly transmitted to the DTU.

[0028] Data acquisition consists of a bus and branch lines. A monitoring branch line is laid out in each monitoring cross section of the roadbed and protected by a PE pipe. Each sensor is connected to its own data collector through the branch line. Then, according to the sensor type, the serial port parameters are set in the DTU configuration software as follows: baud rate 9600, check bit NONE, data bit 8. After the setting, the data collector is successfully connected to the DTU through each branch line through the bus.

[0029] Step S3, DTU transmits the data to the IoT platform via wireless communication, and the IoT platform pre-processes and visualizes the data;

[0030] S301, the IoT platform deploys a monitoring solution based on the ThingsBoard platform and receives raw monitoring data through the MQTT protocol;

[0031] Deploy the ThingsBoard-based IoT platform locally. Import the DTU device information on the IoT platform, select the MQTT protocol connection method, set the device credentials: client ID, user name, password, and then select the MQTT protocol connection method in the MQTT node of Node-RED, enter the parameters: server: 127.0.0.1, port: 8080, and enter the above device credentials, and then select the data upload format: json with IMEI. After the settings are completed, the MQTT node will successfully connect to the IoT platform through the MQTT protocol "publish-subscribe", upload the monitoring data to the IoT platform, and store it in the database as raw monitoring data.

[0032] S302, using Node-RED as an edge computing node, pre-process the original monitoring data to remove abnormal data and process noise data. The process is as follows Figure 3 As shown. First, establish a TCP connection with the IoT platform database through the TCP node, set the address and port, and select the TCP connection method. After the connection is successful, the original monitoring data stored in the database is transferred to the TCP node.

[0033] Then write the abnormal data removal code in the Function1 node: Set the settlement values ​​to be independent x1, x2..., x n , calculate its arithmetic mean x and residual error v i =x i -x(i=1,2,...,n), and calculate the standard deviation σ according to Bessel formula. If a certain sedimentation value x b The residual error v b (1<=b<=n), satisfying the following formula |v b |=|x b -x|>3σ, then x b It is a bad value containing large error values ​​and should be eliminated.

[0034] Then connect the TCP node to the Function1 node to obtain the monitoring data without abnormal data. Then connect to the Function2 node and write the noise data processing code in the Function2 node: Moving average = (data point 1 + data point 2 + data point 3 + .... + data point n) / n, where n is the time window size of the moving average, which means taking the first n data points for average calculation. Starting from the first data point of the data sequence, calculate the moving average in sequence. For the first moving average, the first n data points need to be used for calculation; for subsequent moving averages, slide back one data point each time and recalculate the average. Record the calculated moving average as the smoothed data sequence.

[0035] Finally, the preprocessed monitoring data is connected to the IoT platform through the MQTT node and stored in the database as effective monitoring data.

[0036] S303, visualize and configure early warning for the effective monitoring data after preprocessing, such as Figure 4 As shown. Generate a time curve chart with selectable time periods according to the changes in the monitoring point data, which is convenient for retrospective viewing and analysis. At the same time, generate differential settlement curves according to the settlement of different measuring points. The differential settlement at different times can be displayed on the same page, which is convenient for comparative analysis and real-time update. Set up the early warning module, set the range of settlement, and set three-level alarm thresholds according to the characteristics of the project. The first level alarm is the most serious, the second level is the second, and the third level alarm is the lightest. The second-level alarm and the third-level alarm are displayed in the automatically pushed daily, weekly, and monthly reports. Once the first-level alarm is reached, it will not only be displayed in the report, but also the alarm information will be pushed to the user in time. EXCEL and WORD reports can be exported on the platform. The report has functions such as settlement statistics of measuring points, curves, alarm levels and alarm number statistics, and abnormal data type statistics. Perform a more comprehensive processing and analysis of the data to better guide the construction of the project.

[0037] Step S4, establishing two artificial intelligence models for roadbed settlement prediction, and combining the two artificial intelligence models for roadbed prediction.

[0038] The change trend of the Poisson curve is more consistent with the law of the late stage of roadbed settlement, so the prediction of roadbed settlement has a good accuracy. First, a Poisson curve settlement prediction model is established. The expression of the Poisson curve settlement prediction model is:

[0039]

[0040] Where: S tis the settlement of the roadbed at a certain predicted position at time t; k, a, and b are unknown coefficients (where a>0, b>0).

[0041] The steps to predict roadbed settlement using the Poisson curve settlement prediction model are:

[0042] Obtain the sensor monitoring data within the set time range before the current moment, and take the measured settlement values ​​with equal monitoring time intervals. Let the measured roadbed settlement values ​​corresponding to each time point t=1, 2, 3, ..., n be S1, S2, S3, ..., S n ;

[0043] Divide the time range into three monitoring periods, let i = n / 3; then, the time point in the first monitoring period is t = 1, 2, 3, ..., i, the time point in the second monitoring period is t = i+1, i+2, i+3, ..., 2i, and the time point in the third monitoring period is t = 2i+1, 2i+2, 2i+3, ..., 3i; the roadbed settlement characteristic parameters corresponding to the three monitoring time periods are y1, y2, y3;

[0044]

[0045]

[0046] The undetermined coefficients in the Poisson curve settlement prediction model expression are updated according to the roadbed settlement characteristic parameters, where:

[0047]

[0048] Substitute the three coefficients a, b, and k obtained into the Poisson curve settlement prediction model, let the time t tend to infinity, and obtain the final roadbed settlement S.

[0049] After the Poisson curve model is established, a highway settlement prediction model based on the Bayesian update method of Markov chain Monte Carlo is established. The steps are as follows: Figure 5 As shown. Finite element software was used to establish a finite element model. The soil parameters were regarded as random variables, and 2000 sets of parameter samples were generated by Latin hypercube sampling, which were input into the finite element model to calculate the corresponding settlement results. Based on the parameter samples, a roadbed settlement data set was constructed, and then the random forest method was used to construct a proxy model. The roadbed settlement data set constructed above was randomly divided into two parts, 80% of which was used as a training set and the remaining 20% ​​as a test set. The proxy model was trained using the training set, and then the soil parameters in the test set were input into the proxy model and the roadbed settlement calculation results were output. The calculation results of the proxy model were compared with those of the original finite element model. After the fitting effect met the requirements, the proxy model was used to replace the original finite element model.

[0050] According to the properties of soft soil, a soil parameter data set is generated; the soil parameter data set is input into the proxy model to calculate the settlement, and according to the influence of each parameter on the settlement value of soft soil roadbed, the highly sensitive soil parameters are determined and regarded as random variables, and other soil parameters are regarded as constants.

[0051] According to geological survey reports, existing literature data and engineering experience, the prior information of highly sensitive soil parameters is obtained and their prior distribution is determined. Based on the proxy model and combined with monitoring data, the Markov Chain Monte Carlo Bayesian Update Algorithm (MCMC) is used to obtain the posterior samples of soil parameters; the posterior samples of the updated soil parameters are substituted into the proxy model to obtain the predicted amount of soft foundation settlement.

[0052] After the two settlement prediction models are established, the weights of different models are determined by the inverse error method to form a combined prediction model:

[0053] Let y represent the actual settlement monitoring value, represents the predicted value of the i-th model at time t, and y(t) represents the predicted value of the combined model at time t;

[0054] Calculate the absolute error e of the i-th model at time t i (t), the sum of the reciprocal of the absolute error d i (t), the weight wi of the i-th model at time t, and then the prediction value of the combined model at time t is obtained. The calculation formula is as follows:

[0055]

[0056] Step S5 embeds the combined prediction model into the Internet of Things platform, predicts the monitoring data preprocessed in step S3, obtains settlement prediction data, and realizes intelligent monitoring of highway settlement.

[0057] The above artificial intelligence model is embedded with the help of Node-RED, and the process is shown in Figure 6. First, the pre-processed effective monitoring data in the database is retrieved through the TCP node, and the developed artificial intelligence analysis module code is written into the Function node. The monitoring data is connected to the Function node, and the predicted data (predicted settlement) is generated after calculation and processing by the Function node code. Then the predicted data is connected to the Internet of Things platform through the MQTT node. According to the actual situation of the project, a safety range of the predicted data is set. If the safety range is exceeded, an early warning is generated. The early warning message is fed back to the remote terminals of the design, construction, construction, owner and other units through SMS and other means. The management personnel will determine whether there is a potential risk and adjust the construction method accordingly.

[0058] Through the above steps, an intelligent highway monitoring system is established. The schematic diagram of the intelligent monitoring system is as follows: Figure 6 shown.

[0059] Although the present invention is described based on limited engineering example monitoring data, professionals can understand through the above description that there may be other application examples within the scope of the present invention. In addition, it should be noted that the language used in this document is mainly for the purpose of ease of reading and teaching, rather than for explaining or limiting the content of the present invention. Therefore, many modifications and changes are obvious to those of ordinary skill in the art without departing from the scope and spirit of the attached claims. The scope of the present invention is intended to be illustrative rather than restrictive, and the actual scope is determined by the attached claims.

[0060] The above contents only represent the embodiments of the present invention. It should be emphasized that: based on the principles of the present invention, ordinary technicians in this field can still make some improvements and adjustments. These improvements and adjustments should also be considered as the protection scope of the present invention.

Claims

1. A method for intelligent monitoring of highway subsidence, characterized in that: The steps include: Step S1, deploying sensors along the road section to be monitored, wherein the sensors are connected to respective data collectors via RS485 cables; Step S2, the data collector is connected to the DTU via wireless communication; the monitoring data is wirelessly transmitted to the DTU; Step S3, DTU transmits the data to the IoT platform via wireless communication, and the IoT platform pre-processes the data and performs visualization and early warning configuration; Step S4, establishing a Poisson curve settlement prediction model and a settlement prediction model based on the Bayesian update method of Markov chain Monte Carlo; combining the two settlement prediction models to obtain a combined prediction model; Step S5, embedding the combined prediction model into the Internet of Things platform, predicting the monitoring data preprocessed in step S3, obtaining settlement prediction data, and realizing intelligent monitoring of highway settlement.

2. The method for intelligent monitoring of highway subsidence according to claim 1, characterized in that: The sensors in step S1 are a static level and a multi-point displacement meter; in step S1, the settlement of the road section to be monitored is estimated based on geological survey reports, literature and engineering experience, and the range and accuracy of the sensor are selected based on the estimation results.

3. The method for intelligent monitoring of highway subsidence according to claim 1, characterized in that: In step S1, a settlement monitoring section is selected at a set distance along the highway section to be monitored, and a multi-point displacement meter and a static level are buried on both sides of the roadbed and the center line of the roadbed of each settlement monitoring section. The static level is used to obtain the vertical displacement value at the roadbed surface, and the multi-point displacement meter is arranged below the static level to obtain the settlement values ​​at different depths of the layers.

4. The method for intelligent monitoring of highway subsidence according to claim 1, characterized in that: The DTU in step S2 is selected as a 4G DTU with low bandwidth, low power consumption, long distance, and support for multiple connections, and the DTU parameters are configured according to the sensor type to achieve the reception of monitoring data.

5. The method for intelligent monitoring of highway subsidence according to claim 1, characterized in that: Step S3 specifically includes the following steps: S301, the IoT platform deploys a monitoring solution based on the ThingsBoard platform and receives raw monitoring data through the MQTT protocol; S302, using Node-RED as an edge computing node, pre-processing the original monitoring data to remove abnormal data and process noise data; S303, visualize and configure early warning for the effective monitoring data after preprocessing.

6. A method for intelligent monitoring of highway subsidence according to claim 5, characterized in that: The abnormal data elimination is as follows: the mean μ and standard deviation σ of the single-day monitoring data are calculated. If the deviation of a single measurement data exceeds 3σ, the measurement data is deemed to be abnormal data and is eliminated. Then, the mean, standard deviation and single measurement data deviation of the remaining data are calculated, and the data elimination process is repeated until the deviation of each data is less than 3 times the standard deviation. The noise data processing adopts a data smoothing method, that is, for the data after the abnormal data is eliminated, the mean of n items of data in the window is calculated by moving the time window, where n is the size of the moving time window, and the mean is used to replace the original value to eliminate random fluctuations.

7. The method for intelligent monitoring of highway subsidence according to claim 1, characterized in that: The Poisson curve settlement prediction model expression in step S4 is: Where: S t is the settlement of the roadbed at a certain predicted position at time t; k, a, b are unknown coefficients; The steps to predict roadbed settlement using the Poisson curve settlement prediction model are: Obtain the sensor monitoring data within the set time range before the current moment, and take the measured settlement values ​​with equal monitoring time intervals. Let the measured roadbed settlement values ​​corresponding to each time point t=1, 2, 3, ..., n be S1, S2, S3, ..., S n ; Divide the time range into three monitoring periods, let i = n / 3; then, the time point in the first monitoring period is t = 1, 2, 3, ..., i, the time point in the second monitoring period is t = i+1, i+2, i+3, ..., 2i, and the time point in the third monitoring period is t = 2i+1, 2i+2, 2i+3, ..., 3i; the roadbed settlement characteristic parameters corresponding to the three monitoring time periods are y1, y2, y3; The undetermined coefficients in the Poisson curve settlement prediction model expression are updated according to the roadbed settlement characteristic parameters, where: Substitute the three coefficients a, b, and k obtained into the Poisson curve settlement prediction model, let the time t tend to infinity, and obtain the final roadbed settlement S.

8. The method for intelligent monitoring of highway subsidence according to claim 1, characterized in that: In step S4, a highway settlement prediction model based on the Bayesian update method of Markov Chain Monte Carlo is established, which specifically includes: Finite element software is used to establish a finite element model based on highway section survey data; soil parameters are regarded as random variables, multiple sets of parameter samples are generated through Latin hypercube sampling, and they are input into the finite element model to calculate the corresponding settlement results; Based on the parameter samples, a roadbed settlement data set is constructed, and then a proxy model is constructed using the random forest method. The roadbed settlement data set constructed above is randomly divided into a training set and a test set. The proxy model is trained using the training set, and then the soil parameters in the test set are input into the proxy model and the roadbed settlement calculation results are output. The calculation results of the proxy model are compared with those of the original finite element model. When the comparison effect meets the requirements, the proxy model is used to replace the original finite element model. According to the properties of soft soil, a soil parameter data set is generated; the soil parameter data set is input into the proxy model to calculate the settlement, and according to the influence of each parameter on the settlement value of the soft soil roadbed, the highly sensitive soil parameters are determined and regarded as random variables, and other soil parameters are regarded as constants; The prior information of highly sensitive soil parameters is obtained and their prior distribution is determined. Based on the proxy model and combined with the actual settlement monitoring data of the sensor, the Markov Chain Monte Carlo Bayesian update algorithm is used to obtain the posterior samples of the soil parameters. The posterior samples of the updated soil parameters are substituted into the proxy model to obtain the predicted amount of soft foundation settlement.

9. The method for intelligent monitoring of highway subsidence according to claim 1, characterized in that: In step S4, the two settlement prediction models are combined to obtain a combined prediction model, which specifically includes: After the two settlement prediction models are established, the weights of different models are determined using the inverse error method to form a combined prediction model; Let y represent the actual settlement monitoring value, represents the predicted value of the i-th model at time t, and y(t) represents the predicted value of the combined model at time t; Calculate the absolute error e of the i-th model at time t i (t), the sum of the reciprocal of the absolute error d i (t), the weight w of the i-th model at time t i , and then get the combined model prediction value at time t, the calculation formula is as follows:

10. The method for intelligent monitoring of highway subsidence according to claim 1, characterized in that: Step S5 also includes: feeding back the settlement prediction results to the user's remote terminal through cloud services, and issuing early warnings for excessive settlement and stability risks.

Images