Real-time monitoring model construction method for uneven settlement of soft soil foundation

Abstract

This invention relates to a method for constructing a real-time monitoring model for uneven settlement of soft soil foundations, belonging to the field of computer-aided design, and more specifically to the field of digital twins. The method includes: uniformly dividing the soft soil foundation area to be monitored into grids, and deploying multiple data sensing units at the center of each grid; using an AI real-time monitoring model to intelligently analyze the settlement depths of each grid in the soft soil foundation area at the next time step based on multi-source data. This invention addresses the technical problem of the inability to effectively analyze the uneven settlement data of each soft soil foundation area at future time steps. By employing different AI real-time monitoring models designed for different soft soil foundation areas, and introducing more types of multi-source data as basic components, it performs intelligent analysis on the uneven settlement data of each soft soil foundation area at future time steps, thereby solving the aforementioned technical problem.

Description

Technical Field

[0001] This invention relates to the field of computer-aided design, more specifically to the field of digital twins, and particularly to a method for constructing a real-time monitoring model for uneven settlement of soft soil foundations. Background Technology

[0002] Digital twins are a commonly used technique in computer-aided design for modeling, simulating, analyzing, and optimizing various application areas. A digital twin refers to a simulation process that fully utilizes physical models, sensor data, operational history, and other data to integrate multiple disciplines and scales. It serves as a virtual mirror image of the physical product, reflecting the entire lifecycle of the corresponding physical product.

[0003] A digital twin is an information model that exists in computer virtual space and is completely equivalent to a physical entity. It allows for the modeling, simulation, analysis, and optimization of physical entities. For example, a dedicated artificial intelligence model can be built for each soft soil foundation area to be monitored, serving as its own digital twin, to model, simulate, analyze, and optimize the real-time or future state of the monitored soft soil foundation area.

[0004] For example, Chinese Utility Model Patent Publication CN205352376U discloses a soft soil foundation settlement monitoring device, including a measuring rod and a relatively movable measuring sleeve sleeved on the measuring rod; a measuring plate is provided at the lower end of the measuring sleeve; the measuring plate is fixed on the soft soil foundation by a fixing device; the measuring plate is provided with a wireless transceiver module, a main control module, an inclination sensor, and a power module, all of which are connected to the main control module; an upper bracket is provided on the measuring rod, and an infrared transmitter is provided on the upper bracket; a lower bracket is provided on the measuring sleeve, and an infrared receiver corresponding to the infrared transmitter is provided on the lower bracket. This utility model discloses a soft soil foundation settlement monitoring device that can effectively measure the settlement of soft soil foundations and achieve intelligent transmission, allowing staff to remotely obtain monitoring information from their office after setting up the monitoring device.

[0005] For example, Chinese invention patent publication CN120850159A proposes an intelligent monitoring method for silty soil foundations based on multi-source data fusion, specifically relating to the field of foundation monitoring and early warning technology. It constructs a dataset of the original soil state of the silty soil foundation and performs anomaly removal and time calibration based on a rheological constitutive model to generate a foundation state feature matrix. A sliding window approach is used to perform dynamic causal correlation tracking analysis on the feature matrix, generating a set of water-soil coupling drift parameters. A nonlinear scaling analysis method is used to extract a set of stress history dependence coefficients. Based on the set of water-soil coupling drift parameters and the set of stress history dependence coefficients, the stability risk level of the silty soil foundation is determined, and corresponding early warning information is selected and output from a pre-set graded early warning instruction database. This improves the accuracy and adaptability of anomaly identification and risk warning, and is suitable for long-term stability monitoring and intelligent risk prevention of soft foundations or silty strata.

[0006] However, none of the aforementioned existing technologies address the prediction of uneven settlement in soft soil foundation areas, nor do they utilize different artificial intelligence models designed for different soft soil foundation areas. This makes it difficult to model, simulate, analyze, and optimize the real-time or future state of the soft soil foundation areas to be monitored. Furthermore, the multi-source data used are all of the same type, such as mechanical or physical data, and the lack of use of more diverse types of multi-source data to predict the uneven settlement in soft soil foundation areas results in insufficient accuracy of the prediction results. Summary of the Invention

[0007] To address technical challenges in this field, this invention provides a method for constructing a real-time monitoring model for uneven settlement of soft soil foundations. This method employs an AI-based real-time monitoring model specifically designed for the soft soil foundation area under monitoring. It incorporates various types of multi-source data as foundational elements, performing intelligent analysis on the settlement depth of each grid within the monitored soft soil foundation area at the next possible moment. Based on the intelligent analysis results, targeted data analysis is conducted to obtain the uneven settlement data of the monitored soft soil foundation area at the next possible moment. This enables a reliable determination of the degree of uneven settlement in the future of the monitored soft soil foundation area based on real-time and historical monitoring data, thereby improving the precision and intelligence of soft soil foundation maintenance.

[0008] According to the present invention, a method for constructing a real-time monitoring model for uneven settlement of soft soil foundation is provided, the method comprising: The soft soil foundation area to be monitored is evenly divided into grids, and a multi-data sensing unit, which is an integrated package of inclinometer, fiber optic strain ring, pore water pressure gauge and temperature sensing chip, is deployed at the center of each grid. A photoelectric imaging unit is installed on the side at the middle position along the length of the soft soil foundation area to be monitored, and the mirror surface of the imaging lens of the photoelectric imaging unit is perpendicular to the horizontal plane. At any given moment, the output content corresponding to each of the multi-data sensing units of each grid and the side imaging image corresponding to the photoelectric imaging unit are captured synchronously. The visualization parameter set of the side imaging image corresponding to the photoelectric imaging unit is analyzed, and the output content and visualization parameter set at any given moment are used as the multimodal fusion data at any given moment. The AI ​​real-time monitoring model is used to intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment, based on the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, multimodal fusion data corresponding to each past moment before the current moment, and settlement data. Among them, the time interval between each past moment before the current moment and the next moment after the current moment on the time axis is a preset time length, and each past moment before the current moment includes the current moment; The AI ​​real-time monitoring model is a superposition of a preset number of graph neural networks.

[0009] Compared with the prior art, the present invention has at least the following outstanding substantive features: Substantive Feature A: Employing a dedicated AI real-time monitoring model for the soft soil foundation area under monitoring, based on comprehensive screening of various fundamental data, the model intelligently analyzes the settlement depth of each grid within the soft soil foundation area at the next moment from the current moment. It removes the maximum and minimum values ​​from each settlement depth to obtain the remaining multiple settlement depths, and calculates the root mean square error of these remaining depths as the non-uniform settlement data for the soft soil foundation area at the next moment from the current moment. The larger the non-uniform settlement data for the soft soil foundation area at the next moment from the current moment, the more significant the non-uniform settlement. This intelligently determines the degree of non-uniform settlement in the future of the soft soil foundation area based on real-time and past monitoring data, providing sufficient reaction time for maintenance personnel to formulate and implement corresponding countermeasures, thus preventing accidents related to non-uniform settlement in the soft soil foundation area. Substantial Feature B: It provides targeted hardware resources for intelligent analysis of the settlement depth of each grid in the soft soil foundation area to be monitored at the current time and the next time. Specifically, the soft soil foundation area to be monitored is evenly divided into grids. At the center of each grid, a multi-data sensing unit is installed, which is an integrated package of an inclinometer, a fiber optic strain ring, a pore water pressure gauge and a temperature sensing chip. At the same time, a photoelectric imaging unit is set on the side at the middle position of the length direction of the soft soil foundation area to be monitored. The mirror of the imaging lens of the photoelectric imaging unit is perpendicular to the horizontal plane. Substantive Feature C: To perform intelligent analysis of the settlement depth of each grid in the soft soil foundation area to be monitored at the current moment and the next moment, a dedicated AI real-time monitoring model for the soft soil foundation area is constructed. The dedicated AI real-time monitoring model for the soft soil foundation area consists of a preset number of superimposed graph neural networks. The numerical change trend of the preset number follows the numerical change trend of the length of the soft soil foundation area to be monitored. The number of learning operations performed by the AI ​​real-time monitoring model is proportional to the total number of grids in the soft soil foundation area to be monitored. The area of ​​the projection region of each grid on the horizontal plane is equal. Thus, different AI real-time monitoring models with different structures are designed for different soft soil foundation areas, ensuring the reliability and stability of the real-time intelligent results of uneven settlement of soft soil foundation. Substantive Feature D: In each learning operation performed on the AI ​​real-time monitoring model, the settlement depths of each grid in a known soft soil foundation area at the next time step are used as the output data of the AI ​​real-time monitoring model. The length of the soft soil foundation area, daily traffic flow data, total number of grids, resolution of the photoelectric imaging unit used in the soft soil foundation area, preset time length, multimodal fusion data corresponding to each past time step before the given time step, and settlement data are used as the input data of the AI ​​real-time monitoring model. This completes the learning operation performed on the AI ​​real-time monitoring model, thereby ensuring the learning effect of each learning operation of the AI ​​real-time monitoring model. Substantive Feature E: To perform intelligent analysis of the settlement depth of each grid in the monitored soft soil foundation area at the next time step, various basic elements, including multimodal fusion data, are used. These elements include the length of the monitored soft soil foundation area, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, and various multimodal fusion data and settlement data corresponding to each past time step before the current time step. The thorough and comprehensive screening of the above basic elements further ensures the reliability and stability of the real-time intelligent results of uneven settlement of soft soil foundation. Substantive Feature F: Specifically, at any given moment, the system synchronously captures the output content corresponding to each of the multi-data sensing units of each grid and the side imaging image corresponding to the photoelectric imaging unit. It analyzes the visualization parameter set of the side imaging image corresponding to the photoelectric imaging unit and uses the output content and visualization parameter set at any given moment as the multimodal fusion data corresponding to that moment. The output content corresponding to the multi-data sensing unit of each grid synchronously captured at any given moment consists of the tilt angle, strain, pore pressure, and temperature data output by the inclinometer, fiber optic strain ring, pore water pressure gauge, and temperature sensing chip corresponding to that grid, respectively. It also identifies the imaging sub-image of the soft soil foundation area to be monitored in the side imaging image corresponding to the photoelectric imaging unit based on the imaging features of the soft soil foundation area. The imaging depth, coordinate information, and brightness information of each pixel in the imaging sub-image of the soft soil foundation area to be monitored are used as the visualization parameter set of the side imaging image corresponding to the photoelectric imaging unit at any given moment. This provides a customized data structure for the various basic contents used for intelligent analysis. Attached Figure Description

[0010] The embodiments of the present invention will now be described with reference to the accompanying drawings, wherein: Figure 1 This is a schematic diagram illustrating the working scenario of the real-time monitoring model construction method for uneven settlement of soft soil foundation according to the present invention.

[0011] Figure 2 This is a flowchart illustrating the steps of constructing a real-time monitoring model for uneven settlement of soft soil foundation according to Embodiment 1 of the present invention.

[0012] Figure 3 This is a flowchart illustrating the steps of a method for constructing a real-time monitoring model for uneven settlement of soft soil foundation according to Embodiment 2 of the present invention.

[0013] Figure 4 This is a flowchart illustrating the steps of constructing a real-time monitoring model for uneven settlement of soft soil foundation according to Embodiment 3 of the present invention.

[0014] Figure 5 This is a flowchart illustrating the steps of constructing a real-time monitoring model for uneven settlement of soft soil foundation according to Embodiment 4 of the present invention.

[0015] Figure 6 This is a flowchart illustrating the steps of constructing a real-time monitoring model for uneven settlement of soft soil foundation according to Embodiment 5 of the present invention. Detailed Implementation

[0016] like Figure 1The diagram illustrates a working scenario for the real-time monitoring model construction method for uneven settlement of soft soil foundations according to the present invention. This invention relates to the field of computer-aided design, and more specifically to the field of digital twins.

[0017] The specific technical process of this invention is as follows: Technical Process 1: It provides specially designed hardware resources for intelligent analysis of the settlement depth of each grid in the soft soil foundation area to be monitored at the current time and the next time. Specifically, the soft soil foundation area to be monitored is evenly divided into grids. At the center of each grid, a multi-data sensing unit is installed, which is an integrated package of an inclinometer, a fiber optic strain ring, a pore water pressure gauge and a temperature sensing chip. At the same time, a photoelectric imaging unit is installed on the side at the middle position of the length direction of the soft soil foundation area to be monitored. The mirror surface of the imaging lens of the photoelectric imaging unit is perpendicular to the horizontal plane. Technical Process Two: To perform intelligent analysis of the settlement depth of each grid in the soft soil foundation area to be monitored at the current moment and the next moment, a dedicated AI real-time monitoring model for the soft soil foundation area to be monitored is constructed, such as... Figure 1 As shown; Specifically, the AI ​​real-time monitoring model specifically designed for the soft soil foundation area to be monitored, and the structural customization mainly involves the following aspects: First: The AI ​​real-time monitoring model specifically for the soft soil foundation area to be monitored is a superposition of multiple graph neural networks with a preset number of parameters; Second: The numerical change trend of the preset quantity follows the numerical change trend of the length of the soft soil foundation area to be monitored; For example, when the length of the soft soil foundation area to be monitored is 10 kilometers, the preset quantity is 5; when the length of the soft soil foundation area to be monitored is 8 kilometers, the preset quantity is 4; when the length of the soft soil foundation area to be monitored is 6 kilometers, the preset quantity is 3; when the length of the soft soil foundation area to be monitored is 4 kilometers, the preset quantity is 2, and so on. Third: The number of learning operations performed by the AI ​​real-time monitoring model is directly proportional to the total number of grids in the soft soil foundation area to be monitored, and the area of ​​the projected area of ​​each grid on the horizontal plane is equal. For example, the total number of grids in the soft soil foundation area to be monitored is 16, and the number of learning operations performed by the AI ​​real-time monitoring model is 800; the total number of grids in the soft soil foundation area to be monitored is 32, and the number of learning operations performed by the AI ​​real-time monitoring model is 1600; the total number of grids in the soft soil foundation area to be monitored is 64, and the number of learning operations performed by the AI ​​real-time monitoring model is 3200, and so on. Fourth: In each learning operation performed on the AI ​​real-time monitoring model, the settlement depths of each grid in a known soft soil foundation area at the next time step are used as the output data of the AI ​​real-time monitoring model. The length of the soft soil foundation area, daily traffic flow data and total number of grids, resolution of the photoelectric imaging unit used in the soft soil foundation area, preset time length, multimodal fusion data corresponding to each past time step before the given time step, and settlement data are used as the input data of the AI ​​real-time monitoring model. This completes the learning operation performed on the AI ​​real-time monitoring model, thereby ensuring the learning effect of each learning operation of the AI ​​real-time monitoring model. In this way, by designing AI real-time monitoring models with different structures for different soft soil foundation areas, the reliability and stability of real-time intelligent results of uneven settlement of soft soil foundations are ensured. Technical Process 3: In order to perform intelligent analysis of the settlement depth of each grid in the soft soil foundation area to be monitored at the current time and the next time, various basic contents including multimodal fusion data are used. These basic contents, which are more types of multi-source data, are derived from the hardware resources specifically designed in Technical Process 1. Specifically, the multimodal fusion data includes the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, multimodal fusion data corresponding to each past time before the current time, and settlement data. Among them, the output content corresponding to each grid's multi-data sensing unit at any time and the visualization parameter set at any time are taken as the multimodal fusion data corresponding to any time. like Figure 1 As shown, the multimodal fusion data can be divided into four data sources, including the output content from the multi-data sensing unit, which can transmit its corresponding output content wirelessly; the visualization parameter set from the photoelectric imaging unit; the settlement data from the local storage terminal; and the remaining auxiliary information, which includes the length of the soft soil foundation area to be monitored, the daily traffic flow data and the total number of grids, the resolution of the photoelectric imaging unit, and the preset time length. More specifically, at any given moment, the output content corresponding to each of the multi-data sensing units of each grid and the side imaging image corresponding to the photoelectric imaging unit are captured synchronously. Based on the imaging characteristics of the soft soil foundation area, the imaging sub-image of the soft soil foundation area to be monitored in the side imaging image corresponding to the photoelectric imaging unit is identified. The imaging depth, coordinate information and brightness information of each pixel in the imaging sub-image of the soft soil foundation area to be monitored are used as the visualization parameter set of the side imaging image corresponding to the photoelectric imaging unit at any given moment. More specifically, the output content of each grid's multi-data sensing unit, which is synchronously captured at any given time, is the tilt angle, strain, pore pressure, and temperature data output by the tilt meter, fiber strain ring, pore pressure gauge, and temperature sensing chip corresponding to the grid, respectively. In this way, through the thorough and comprehensive screening of the above-mentioned basic elements, the reliability and stability of the real-time intelligent results of uneven settlement of soft soil foundation are further guaranteed. Technical Process Four: Utilizing the AI ​​real-time monitoring model specifically designed for the soft soil foundation area under monitoring, as constructed using Technical Process Two, and based on the more comprehensive and diverse multi-source data filtered in Technical Process Three, intelligent analysis is performed on the settlement depth of each grid within the soft soil foundation area under monitoring at the current moment and the next moment. For example... Figure 1 As shown; Technical Process Five: Targeted data analysis is performed on the intelligent analysis results from Technical Process Four to obtain the uneven settlement data of the monitored soft soil foundation area at the next time step, such as... Figure 1 As shown, the obtained uneven settlement data can be wirelessly transmitted to a remote real-time monitoring server; Specifically, intelligent analysis obtains the settlement depth of each grid in the soft soil foundation area to be monitored at the next time step of the current time step. The maximum and minimum values ​​of each settlement depth step are removed to obtain the remaining multiple settlement depths. The root mean square error of the remaining multiple settlement depths is calculated as the non-uniform settlement data of the soft soil foundation area to be monitored at the next time step of the current time step. The larger the non-uniform settlement data of the soft soil foundation area to be monitored at the next time step of the current time step, the more obvious the degree of non-uniform settlement of the soft soil foundation area to be monitored at the next time step of the current time step. Therefore, through the coordinated operation of the above five technical processes, this invention can employ a dedicated AI real-time monitoring model for the soft soil foundation area to be monitored. Based on a comprehensive screening of various fundamental elements, it performs intelligent analysis on the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment from the current moment, and performs targeted data analysis on the intelligent analysis results to obtain the uneven settlement data of the soft soil foundation area to be monitored at the next moment from the current moment. This completes the intelligent determination of the degree of uneven settlement of the soft soil foundation area to be monitored at future moments based on real-time monitoring data and past monitoring data, providing sufficient reaction time for the maintenance party of the soft soil foundation area to formulate and implement corresponding countermeasures, and avoiding the occurrence of accidents related to uneven settlement of the soft soil foundation area.

[0018] The key points of this invention are: hardware resource design based on uniform grid division, targeted arrangement of multiple data sensing units and photoelectric imaging units, directional customized design of AI real-time monitoring models for different structures in different soft soil foundation areas, full and comprehensive screening of multiple types of data with extended categories, and targeted data analysis for calculating uneven settlement data.

[0019] The following will describe in detail the method for constructing a real-time monitoring model for uneven settlement of soft soil foundations according to the present invention through examples.

[0020] Example 1 Figure 2 This is a flowchart illustrating the steps of constructing a real-time monitoring model for uneven settlement of soft soil foundation according to Embodiment 1 of the present invention.

[0021] like Figure 2 As shown, the method for constructing a real-time monitoring model for uneven settlement of soft soil foundation includes the following specific steps: Step S21: Divide the soft soil foundation area to be monitored into grids evenly, and deploy a multi-data sensing unit at the center of each grid, which is an integrated package of inclinometer, fiber optic strain ring, pore water pressure gauge and temperature sensing chip. For example, for each grid in the soft soil foundation area to be monitored, an inclinometer, a fiber optic strain gauge, a pore water pressure gauge, and a temperature sensing chip are arranged from top to bottom at the center of the grid in a direction perpendicular to the horizon, and the inclinometer, fiber optic strain gauge, pore water pressure gauge, and temperature sensing chip arranged from top to bottom are integrated into a single package. Step S22: Set up a photoelectric imaging unit on the side at the middle position along the length of the soft soil foundation area to be monitored, with the mirror surface of the imaging lens of the photoelectric imaging unit perpendicular to the horizontal plane. For example, the photoelectric imaging unit includes an imaging lens, a support, an angle detection unit, an angle correction unit, a microcontroller, a flexible circuit board, a photoelectric sensor, and a filter; As a further example, the angle detection unit is used to detect whether the mirror surface of the imaging lens of the photoelectric imaging unit is perpendicular to the horizontal plane. The angle correction unit is connected to the angle detection unit and is used to drive the imaging lens to keep the mirror surface of the imaging lens perpendicular to the horizontal plane when the angle detection unit detects that the mirror surface of the imaging lens of the photoelectric imaging unit is not perpendicular to the horizontal plane. As a further example, in the horizontal direction, the imaging lens, filter, photoelectric sensor, microcontroller and bracket are arranged in sequence, and the imaging lens, filter, photoelectric sensor and microcontroller are integrated into a single package; Step S23: At any given moment, simultaneously capture the output content corresponding to each of the multi-data sensing units of each grid and the side imaging image corresponding to the photoelectric imaging unit, analyze the visualization parameter set of the side imaging image corresponding to the photoelectric imaging unit, and use the output content and visualization parameter set at any given moment as the multimodal fusion data corresponding to any given moment. Specifically, this step completes the synchronous acquisition of visualized data and multi-sensor data, effectively expanding the data categories for various basic contents used in intelligent analysis; Step S24: Using an AI real-time monitoring model, based on the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, multimodal fusion data corresponding to each past time before the current time, and each settlement data, intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next time after the current time. For example, the current time is 5:00 PM. The previous times before the current time are 5:00 PM, 4:50 PM, 4:40 PM, 4:30 PM, 4:20 PM, 4:10 PM, 4:00 PM, 3:50 PM, 3:40 PM, 3:30 PM, 3:20 PM, 3:10 PM, 3:00 PM, 2:50 PM, 2:40 PM, and 2:30 PM, for a total of 16 times. Accordingly, when the current time is 5:00 PM, the next time is 5:10 PM; Among them, the time interval between each past moment before the current moment and the next moment after the current moment on the time axis is a preset time length, and each past moment before the current moment includes the current moment; For example, if the current time is 5:00 PM, and the previous times are 5:00 PM, 4:50 PM, 4:40 PM, 4:30 PM, 4:20 PM, 4:10 PM, 4:00 PM, 3:50 PM, 3:40 PM, 3:30 PM, 3:20 PM, 3:10 PM, 3:00 PM, 2:50 PM, 2:40 PM, and 2:30 PM (a total of 16 times), the preset time length is 10 minutes. The AI ​​real-time monitoring model is a superposition of a preset number of graph neural networks, where the preset number is the total number of graph neural networks within the AI ​​real-time monitoring model. Among them, the output content of each grid's multi-data sensing unit, which is synchronously captured at any time, is the tilt angle, strain, pore pressure and temperature data output by the tilt meter, fiber strain ring, pore water pressure gauge and temperature sensing chip corresponding to the grid, respectively. Specifically, the inclinometer, fiber optic strain gauge, pore water pressure gauge and temperature sensing chip corresponding to the grid synchronously output the corresponding various types of sensing data at any given time. Among them, the visualization parameter set of the side imaging image corresponding to the photoelectric imaging unit includes: based on the imaging features of the soft soil foundation area, the imaging sub-image of the soft soil foundation area to be monitored in the side imaging image of the photoelectric imaging unit is identified, and the imaging depth, coordinate information and brightness information of each pixel in the imaging sub-image of the soft soil foundation area to be monitored are used as the visualization parameter set of the side imaging image of the photoelectric imaging unit at any time. For example, the imaging features of the soft soil foundation area can be the color imaging features of the soft soil foundation area. The pixels in the side imaging image corresponding to the photoelectric imaging unit that match the color imaging features of the soft soil foundation area are used as the constituent pixels of the imaging sub-image of the soft soil foundation area to be monitored, so as to complete the identification of the imaging sub-image of the soft soil foundation area to be monitored. Among them, the AI ​​real-time monitoring model is a superposition of a preset number of graph neural networks, including: a preset number of numerical change trends following the numerical change trend of the length of the soft soil foundation area to be monitored; For example, the numerical trend of the preset quantity follows the numerical trend of the length of the soft soil foundation area to be monitored, including: when the length of the soft soil foundation area to be monitored is 10 kilometers, the preset quantity is 5; when the length of the soft soil foundation area to be monitored is 8 kilometers, the preset quantity is 4; when the length of the soft soil foundation area to be monitored is 6 kilometers, the preset quantity is 3; when the length of the soft soil foundation area to be monitored is 4 kilometers, the preset quantity is 2, and so on. Among them, the AI ​​real-time monitoring model is a superposition of multiple graph neural networks with a preset number of parameters, and the number of learning operations performed by the AI ​​real-time monitoring model is proportional to the total number of grids in the soft soil foundation area to be monitored, and the area of ​​the projection area of ​​each grid on the horizontal plane is equal. For example, the number of learning operations performed by the AI ​​real-time monitoring model is proportional to the total number of grids in the soft soil foundation area to be monitored, and the area of ​​the projected area of ​​each grid on the horizontal plane is equal, including: the total number of grids in the soft soil foundation area to be monitored is 16, the number of learning operations performed by the AI ​​real-time monitoring model is 800, the total number of grids in the soft soil foundation area to be monitored is 32, the number of learning operations performed by the AI ​​real-time monitoring model is 1600, the total number of grids in the soft soil foundation area to be monitored is 64, the number of learning operations performed by the AI ​​real-time monitoring model is 3200, and so on; In each learning operation performed on the AI ​​real-time monitoring model, the settlement depths of each grid in a known soft soil foundation area at the next time step are taken as the output data of the AI ​​real-time monitoring model. The length of the soft soil foundation area, daily traffic flow data and total number of grids, resolution of the photoelectric imaging unit used in the soft soil foundation area, preset time length, multimodal fusion data corresponding to each past time step before the given time step, and settlement data are taken as the input data of the AI ​​real-time monitoring model, thus completing the learning operation performed on the AI ​​real-time monitoring model.

[0022] Example 2 Figure 3 This is a flowchart illustrating the steps of a method for constructing a real-time monitoring model for uneven settlement of soft soil foundation according to Embodiment 2 of the present invention.

[0023] like Figure 3 As shown, with Figure 2 Unlike the previous embodiment, in the method for constructing a real-time monitoring model for uneven settlement of soft soil foundation, after using an AI real-time monitoring model to intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment based on the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, preset time length, multimodal fusion data corresponding to each past moment before the current moment, and each settlement data, i.e. after step S24, the method further includes: Step S25: Obtain the settlement depth of each grid in the soft soil foundation area to be monitored at the next time of the current time. Remove the maximum and minimum values ​​from each settlement depth to obtain the remaining settlement depths. Calculate the root mean square error of the remaining settlement depths to use as the non-uniform settlement data of the soft soil foundation area to be monitored at the next time of the current time. For example, the maximum and minimum values ​​of each settlement depth are removed to obtain the remaining multiple settlement depths. The root mean square error of the remaining multiple settlement depths is calculated as the non-uniform settlement data of the soft soil foundation area to be monitored at the current time and the next time. The number of maximum values ​​and the number of minimum values ​​can be set by the user, and the number of maximum values ​​and the number of minimum values ​​are generally equal. The calculation of the root mean square error of the remaining multiple settlement depths to serve as the non-uniform settlement data of the soft soil foundation area to be monitored at the current time to the next time includes: the larger the non-uniform settlement data of the soft soil foundation area to be monitored at the current time to the next time, the more obvious the degree of non-uniform settlement of the soft soil foundation area to be monitored at the current time to the next time. Example

[0024] Figure 4This is a flowchart illustrating the steps of constructing a real-time monitoring model for uneven settlement of soft soil foundation according to Embodiment 3 of the present invention.

[0025] like Figure 4 As shown, with Figure 2 Unlike the previous embodiment, in the method for constructing a real-time monitoring model for uneven settlement of soft soil foundation, after using an AI real-time monitoring model to intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment based on the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, preset time length, multimodal fusion data corresponding to each past moment before the current moment, and each settlement data, i.e. after step S24, the method further includes: Step S26: Obtain the settlement depth of each grid in the soft soil foundation area to be monitored at the next time of the current time, and perform synchronous on-site display of the settlement depth of each grid in the soft soil foundation area to be monitored at the next time of the current time; For example, an LCD display array can be used to obtain the settlement depth of each grid in the soft soil foundation area to be monitored at the next time from the current time, and to perform synchronous on-site display of the settlement depth of each grid in the soft soil foundation area to be monitored at the next time from the current time.

[0026] Example 4 Figure 5 This is a flowchart illustrating the steps of constructing a real-time monitoring model for uneven settlement of soft soil foundation according to Embodiment 4 of the present invention.

[0027] like Figure 5 As shown, with Figure 4 Unlike the previous embodiment, in the method for constructing a real-time monitoring model for uneven settlement of soft soil foundation, after using an AI real-time monitoring model to intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment based on the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, preset time length, multimodal fusion data corresponding to each past moment before the current moment, and each settlement data, i.e. after step S24, the method further includes: Step S27: Obtain the settlement depth of each grid in the soft soil foundation area to be monitored at the next time of the current time, and wirelessly transmit the settlement depth of each grid in the soft soil foundation area to be monitored at the next time of the current time into the same network data packet to the remote real-time monitoring server. For example, the remote real-time monitoring server can be a big data server, a blockchain server, or a cloud computing server; The process of wirelessly transmitting the settlement depths corresponding to each grid in the soft soil foundation area to be monitored at the next moment of the current moment to a remote real-time monitoring server after inputting the data into the same network data packet includes: the wireless transmission is based on time-division duplex communication mode or frequency-division duplex communication mode.

[0028] Example 5 Figure 6 This is a flowchart illustrating the steps of constructing a real-time monitoring model for uneven settlement of soft soil foundation according to Embodiment 5 of the present invention.

[0029] like Figure 6 As shown, with Figure 2 Unlike the previous embodiment, in the method for constructing a real-time monitoring model for uneven settlement of soft soil foundation, after uniformly dividing the soft soil foundation area to be monitored into various grids and deploying a multi-data sensing unit integrated with an inclinometer, fiber optic strain ring, pore water pressure gauge, and temperature sensing chip at the center of each grid, i.e. after step S21, the method further includes: Step S28: Perform multiple learning operations on the AI ​​real-time monitoring model to complete multiple reconstructions of each model parameter of the AI ​​real-time monitoring model; Among them, performing multiple learning operations on the AI ​​real-time monitoring model to complete multiple reconstructions of each model parameter of the AI ​​real-time monitoring model includes: using different physical storage addresses to store each model parameter of the AI ​​real-time monitoring model after multiple reconstructions. For example, TF memory chips or MMC memory chips can be selected to store the various model parameters of the AI ​​real-time monitoring model after multiple reconstructions at different physical storage addresses.

[0030] Next, the various method embodiments of the present invention will be described in detail.

[0031] In the method for constructing a real-time monitoring model for uneven settlement of soft soil foundation according to various method embodiments of the present invention: The trend of the preset quantity's numerical change follows the trend of the length of the soft soil foundation area to be monitored, including: using a quantity change curve to represent the trend of the preset quantity's numerical change, and using a length change curve to represent the trend of the length of the soft soil foundation area to be monitored. For example, numerical change functions and length change functions can be used to represent the quantity change curve and the length change curve, respectively; The method of following the numerical change trend of the preset quantity with the numerical change trend of the length of the soft soil foundation area to be monitored also includes: performing numerical normalization processing on the curve length of the quantity change curve and the length change curve respectively to obtain the first normalized curve and the second normalized curve. Among them, the numerical change trend of the preset quantity following the numerical change trend of the length of the soft soil foundation area to be monitored also includes: the curve shapes of the first normalized curve and the second normalized curve completely overlap.

[0032] In the method for constructing a real-time monitoring model for uneven settlement of soft soil foundation according to various method embodiments of the present invention: An AI real-time monitoring model is used to intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment, based on the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, multimodal fusion data corresponding to each time before the current moment, and settlement data. This includes inputting the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, multimodal fusion data corresponding to each time before the current moment, and settlement data into the AI ​​real-time monitoring model in parallel. The AI ​​real-time monitoring model intelligently analyzes the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment based on the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, multimodal fusion data corresponding to each past moment before the current moment, and settlement data. It also includes: executing the AI ​​real-time monitoring model to obtain the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment of the current moment output by the AI ​​real-time monitoring model. For example, the MATLAB toolbox can be used to execute the AI ​​real-time monitoring model to simulate and test the data processing process of each grid in the monitored soft soil foundation area at the next time step, which is output by the AI ​​real-time monitoring model. The parallel input of the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of the photoelectric imaging unit, preset time length, multimodal fusion data corresponding to each past time before the current time, and settlement data into the AI ​​real-time monitoring model includes: using an ASIC chip to complete the parallel input of the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of the photoelectric imaging unit, preset time length, multimodal fusion data corresponding to each past time before the current time, and settlement data.

[0033] And in the method for constructing a real-time monitoring model for uneven settlement of soft soil foundation according to various method embodiments of the present invention: The number of learning operations performed by the AI ​​real-time monitoring model is proportional to the total number of grids in the soft soil foundation area to be monitored, and the area of ​​the projected area of ​​each grid on the horizontal plane is equal. This includes: using a numerical mapping function to represent the numerical mapping relationship between the number of learning operations performed by the AI ​​real-time monitoring model and the total number of grids in the soft soil foundation area to be monitored. For example, programmable logic devices can be used to test and simulate the implementation of numerical mapping functions; The number of learning operations performed by the AI ​​real-time monitoring model is proportional to the total number of grids in the soft soil foundation area to be monitored, and the area of ​​the projected area of ​​each grid on the horizontal plane is equal. It also includes the following: in the numerical mapping function, the total number of grids in the soft soil foundation area to be monitored is the input value of the numerical mapping function. Furthermore, the number of learning operations performed by the AI ​​real-time monitoring model is proportional to the total number of grids in the soft soil foundation area to be monitored, and the area of ​​the projected area of ​​each grid on the horizontal plane is equal. Additionally, in the numerical mapping function, the number of learning operations performed by the AI ​​real-time monitoring model corresponding to the total number of grids in the soft soil foundation area to be monitored is the output value of the numerical mapping function.

[0034] In addition, the following technical content can be cited to further highlight the essential features of the present invention: Before the length of the soft soil foundation area to be monitored, the daily traffic flow data and the total number of grids, the preset time length, the multimodal fusion data corresponding to each past time before the current time, and the settlement data need to be processed by numerical normalization respectively. Among them, the settlement depths of each grid in the soft soil foundation area to be monitored, output by the AI ​​real-time monitoring model, are in the form of numerical normalization after the numerical normalization process. For example, the numerical normalization process is an octal numerical conversion process. That is, the length of the soft soil foundation area to be monitored, the daily traffic flow data and the total number of grids, the preset time length, the multimodal fusion data corresponding to each past time before the current time, and the settlement data of each time are all in the numerical representation after octal numerical conversion. At the same time, the settlement depth of each grid of the soft soil foundation area to be monitored in the next time after the current time, output by the AI ​​real-time monitoring model, is also in the numerical representation after octal numerical conversion.

[0035] Although the present invention has been described with reference to the accompanying drawings and various preferred embodiments, it will be apparent to those skilled in the art that various modifications and variations can be made to the invention without departing from its spirit and scope. Therefore, all modifications and variations of the invention are covered by the appended claims and their equivalents.

Claims

1. A method for constructing a real-time monitoring model for uneven settlement of soft soil foundations, characterized in that, The method includes: The soft soil foundation area to be monitored is evenly divided into grids, and a multi-data sensing unit, which is an integrated package of inclinometer, fiber optic strain ring, pore water pressure gauge and temperature sensing chip, is deployed at the center of each grid. A photoelectric imaging unit is installed on the side at the middle position along the length of the soft soil foundation area to be monitored, and the mirror surface of the imaging lens of the photoelectric imaging unit is perpendicular to the horizontal plane. At any given moment, the output content corresponding to each of the multi-data sensing units of each grid and the side imaging image corresponding to the photoelectric imaging unit are captured synchronously. The visualization parameter set of the side imaging image corresponding to the photoelectric imaging unit is analyzed, and the output content and visualization parameter set at any given moment are used as the multimodal fusion data at any given moment. The AI ​​real-time monitoring model is used to intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment, based on the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, multimodal fusion data corresponding to each past moment before the current moment, and settlement data. Among them, the time interval between each past moment before the current moment and the next moment after the current moment on the time axis is a preset time length, and each past moment before the current moment includes the current moment; The AI ​​real-time monitoring model is a superposition of a preset number of graph neural networks.

2. The method for constructing a real-time monitoring model for uneven settlement of soft soil foundation as described in claim 1, characterized in that: The output of each grid's multi-data sensing unit, which is synchronously captured at any given time, consists of the tilt angle, strain, pore pressure, and temperature data output by the tilt meter, fiber optic strain ring, pore water pressure gauge, and temperature sensing chip corresponding to that grid, respectively. The visualization parameter set for analyzing the side imaging image corresponding to the photoelectric imaging unit includes: identifying the imaging sub-image of the soft soil foundation area to be monitored in the side imaging image corresponding to the photoelectric imaging unit based on the imaging features of the soft soil foundation area, and using the imaging depth, coordinate information and brightness information of each pixel in the imaging sub-image of the soft soil foundation area to be monitored as the visualization parameter set for the side imaging image corresponding to the photoelectric imaging unit at any time.

3. The method for constructing a real-time monitoring model for uneven settlement of soft soil foundations as described in claim 2, characterized in that: The AI ​​real-time monitoring model is a superposition of a preset number of graph neural networks, including: a preset number of numerical change trends that follow the numerical change trends of the length of the soft soil foundation area to be monitored; Among them, the AI ​​real-time monitoring model is a superposition of multiple graph neural networks with a preset number of parameters, and the number of learning operations performed by the AI ​​real-time monitoring model is proportional to the total number of grids in the soft soil foundation area to be monitored, and the area of ​​the projection area of ​​each grid on the horizontal plane is equal. In each learning operation performed on the AI ​​real-time monitoring model, the settlement depths of each grid in a known soft soil foundation area at the next time step are used as the output data of the AI ​​real-time monitoring model. The length of the soft soil foundation area, daily traffic flow data, total number of grids, resolution of the photoelectric imaging unit used in the soft soil foundation area, preset time length, multimodal fusion data corresponding to each past time step before the given time step, and settlement data are used as the input data of the AI ​​real-time monitoring model to complete the current learning operation performed on the AI ​​real-time monitoring model.

4. The method for constructing a real-time monitoring model for uneven settlement of soft soil foundations as described in claim 3, characterized in that, After employing an AI real-time monitoring model to intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment, based on the length of the area to be monitored, daily traffic flow data, total number of grids, preset time length, multimodal fusion data corresponding to each past moment before the current moment, and settlement data, the method further includes: The settlement depths of each grid in the soft soil foundation area to be monitored are obtained at the next time step of the current time step. The maximum and minimum values ​​of each settlement depth are removed to obtain the remaining settlement depths. The root mean square error of the remaining settlement depths is calculated as the non-uniform settlement data of the soft soil foundation area to be monitored at the next time step of the current time step. The calculation of the root mean square error of the remaining multiple settlement depths to serve as the non-uniform settlement data of the soft soil foundation area to be monitored at the current time to the next time includes: the larger the non-uniform settlement data of the soft soil foundation area to be monitored at the current time to the next time, the more obvious the degree of non-uniform settlement of the soft soil foundation area to be monitored at the current time to the next time.

5. The method for constructing a real-time monitoring model for uneven settlement of soft soil foundations as described in claim 3, characterized in that, After employing an AI real-time monitoring model to intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment, based on the length of the area to be monitored, daily traffic flow data, total number of grids, preset time length, multimodal fusion data corresponding to each past moment before the current moment, and settlement data, the method further includes: The system acquires the settlement depth of each grid in the soft soil foundation area to be monitored at the next time step, and performs synchronous on-site display of the settlement depth of each grid in the soft soil foundation area to be monitored at the next time step.

6. The method for constructing a real-time monitoring model for uneven settlement of soft soil foundations as described in claim 3, characterized in that, After employing an AI real-time monitoring model to intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment, based on the length of the area to be monitored, daily traffic flow data, total number of grids, preset time length, multimodal fusion data corresponding to each past moment before the current moment, and settlement data, the method further includes: The settlement depths of each grid in the soft soil foundation area to be monitored at the next time point of the current time are obtained, and the settlement depths of each grid in the soft soil foundation area to be monitored at the next time point of the current time point are entered into the same network data packet and wirelessly transmitted to the remote real-time monitoring server. The process of wirelessly transmitting the settlement depths corresponding to each grid in the soft soil foundation area to be monitored at the next moment of the current moment to a remote real-time monitoring server after inputting the data into the same network data packet includes: the wireless transmission is based on time-division duplex communication mode or frequency-division duplex communication mode.

7. The method for constructing a real-time monitoring model for uneven settlement of soft soil foundations as described in claim 3, characterized in that, After uniformly dividing the soft soil foundation area to be monitored into grids, and deploying a multi-data sensing unit—integrated with an inclinometer, fiber optic strain gauge, pore water pressure gauge, and temperature sensing chip—at the center of each grid, the method further includes: Multiple learning operations are performed on the AI ​​real-time monitoring model to reconstruct the various model parameters of the AI ​​real-time monitoring model multiple times; The process of performing multiple learning operations on the AI ​​real-time monitoring model to reconstruct the various model parameters of the AI ​​real-time monitoring model includes storing the various model parameters of the AI ​​real-time monitoring model after multiple reconstructions in different physical storage addresses.

8. The method for constructing a real-time monitoring model for uneven settlement of soft soil foundation as described in any one of claims 3-7, characterized in that: The trend of the preset quantity's numerical change follows the trend of the length of the soft soil foundation area to be monitored, including: using a quantity change curve to represent the trend of the preset quantity's numerical change, and using a length change curve to represent the trend of the length of the soft soil foundation area to be monitored. The method of following the numerical change trend of the preset quantity with the numerical change trend of the length of the soft soil foundation area to be monitored also includes: performing numerical normalization processing on the curve length of the quantity change curve and the length change curve respectively to obtain the first normalized curve and the second normalized curve. Among them, the numerical change trend of the preset quantity following the numerical change trend of the length of the soft soil foundation area to be monitored also includes: the curve shapes of the first normalized curve and the second normalized curve completely overlap.

9. The method for constructing a real-time monitoring model for uneven settlement of soft soil foundation as described in any one of claims 3-7, characterized in that: An AI real-time monitoring model is used to intelligently analyze the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment, based on the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, multimodal fusion data corresponding to each time before the current moment, and settlement data. This includes inputting the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, multimodal fusion data corresponding to each time before the current moment, and settlement data into the AI ​​real-time monitoring model in parallel. The AI ​​real-time monitoring model intelligently analyzes the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment based on the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of photoelectric imaging units, preset time length, multimodal fusion data corresponding to each past moment before the current moment, and settlement data. It also includes: executing the AI ​​real-time monitoring model to obtain the settlement depth of each grid in the soft soil foundation area to be monitored at the next moment of the current moment output by the AI ​​real-time monitoring model. The parallel input of the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of the photoelectric imaging unit, preset time length, multimodal fusion data corresponding to each past time before the current time, and settlement data into the AI ​​real-time monitoring model includes: using an ASIC chip to complete the parallel input of the length of the soft soil foundation area to be monitored, daily traffic flow data and total number of grids, resolution of the photoelectric imaging unit, preset time length, multimodal fusion data corresponding to each past time before the current time, and settlement data.

10. The method for constructing a real-time monitoring model for uneven settlement of soft soil foundation as described in any one of claims 3-7, characterized in that: The number of learning operations performed by the AI ​​real-time monitoring model is proportional to the total number of grids in the soft soil foundation area to be monitored, and the area of ​​the projected area of ​​each grid on the horizontal plane is equal. This includes: using a numerical mapping function to represent the numerical mapping relationship between the number of learning operations performed by the AI ​​real-time monitoring model and the total number of grids in the soft soil foundation area to be monitored. The number of learning operations performed by the AI ​​real-time monitoring model is proportional to the total number of grids in the soft soil foundation area to be monitored, and the area of ​​the projected area of ​​each grid on the horizontal plane is equal. It also includes the following: in the numerical mapping function, the total number of grids in the soft soil foundation area to be monitored is the input value of the numerical mapping function. The number of learning operations performed by the AI ​​real-time monitoring model is proportional to the total number of grids in the soft soil foundation area to be monitored, and the area of ​​the projected area of ​​each grid on the horizontal plane is equal. Furthermore, in the numerical mapping function, the number of learning operations performed by the AI ​​real-time monitoring model corresponding to the total number of grids in the soft soil foundation area to be monitored is the output value of the numerical mapping function.

Images