Tubular pile reinforced composite foundation bridgehead differential settlement monitoring method and system

Abstract

The invention discloses a tubular pile reinforced composite foundation bridgehead differential settlement monitoring method and system, and belongs to the technical field of foundation treatment and foundation engineering. Comprising the following steps: acquiring soil pressure data, settlement data, displacement data and pore water pressure data of a pipe pile reinforced composite foundation bridgehead, respectively performing feature extraction, and adaptively fusing a plurality of feature extraction results by utilizing correlation to generate a fused feature vector; the fusion feature vector is processed through a trained settlement prediction model, and a settlement prediction result is obtained with spatial-temporal features as guidance; wherein the settlement prediction model comprises an adaptive residual connection LSTM network layer, a hierarchical space-time attention mechanism network layer, a full connection layer and a linear layer which are connected in sequence. Comprehensive monitoring of a complex foundation environment can be realized, and the precision of settlement prediction is improved; the problems that the settlement mechanism is not deeply considered in existing settlement monitoring, and the settlement prediction accuracy needs to be improved are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of foundation treatment and foundation engineering, and in particular to a method and system for monitoring differential settlement of a bridge head on a pipe pile reinforced composite foundation. Background Art

[0002] The statements in this section merely mention background art related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid growth of economic development and people's travel needs, the coverage of the expressway network has been continuously expanded, and expressways have become an important part of the national transportation infrastructure network. After years of development, the driving comfort and safety of expressways have been continuously improved, but there are still some unsatisfactory aspects. For example, there are a large number of "bridge head jump" phenomena caused by differential settlement at the backfill of structures such as bridges and culverts, which seriously threatens the driving comfort and safety of expressways. Differential settlement at the backfill of the abutment is often an important cause of engineering problems such as embankment instability, tilting of bridge abutments and piers, cracking of piers, and cracking and tilting of retaining walls. Therefore, in today's vigorous development of expressways, how to effectively reduce the bridge head jump phenomenon to the greatest extent has become one of the important "chronic diseases" that need to be solved in the construction of high-grade highways.

[0004] In recent years, pile-supported reinforced structures (pile-net structures) have been increasingly used in soft soil foundation treatment, achieving good engineering results. Among them, prestressed concrete pipe pile reinforced composite foundation has a high-rigidity pile body and strong pile body particle bonding, and prestressed concrete pipe piles can transfer loads to deeper soil layers. Compared with flexible piles, they have obvious advantages in improving the bearing capacity of the foundation, so they are widely used in soft soil foundation treatment. Pipe pile reinforced composite foundation is a complex form of foundation treatment, and the research on its working mechanism is still insufficient. By monitoring the differential settlement of the bridge head, the working mechanism of the composite foundation in actual engineering can be deeply studied, and accurate settlement monitoring is of great significance for evaluating the working performance of the pipe pile reinforced composite foundation at the bridge head, predicting potential differential settlement risks, and taking effective prevention and control measures.

[0005] Most traditional manual measurement methods have low accuracy and can no longer meet the needs of high-precision monitoring of modern roadbed projects. Most existing monitoring methods focus on the application of a single technology and lack comprehensive consideration of the complementary advantages of multiple technologies. Secondly, there is insufficient research on the monitoring characteristics of the special structure of pipe pile reinforced composite foundation, which makes it difficult to fully reflect its settlement characteristics. Finally, existing data analysis methods mostly stay at the data processing level and lack in-depth analysis and understanding of the settlement mechanism. Summary of the invention

[0006] In order to address the deficiencies in the prior art, the present invention provides a method, system, electronic equipment, computer-readable storage medium and computer program product for monitoring differential settlement of a bridge head on a pipe-pile reinforced composite foundation, so as to improve the accuracy and reliability of settlement monitoring.

[0007] In a first aspect, the present invention provides a method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation;

[0008] A method for monitoring differential settlement of a bridge head on a pipe pile reinforced composite foundation, comprising:

[0009] The soil pressure data, settlement data, displacement data and pore water pressure data of the pipe pile reinforced composite foundation bridge head are obtained and feature extraction is performed respectively. The multiple feature extraction results are adaptively fused by using correlation to generate a fused feature vector.

[0010] The fused feature vector is processed by a trained settlement prediction model, and the settlement prediction result is obtained by taking the spatiotemporal features as a guide;

[0011] Among them, the sedimentation prediction model includes an adaptive residual connection LSTM network layer, a hierarchical spatiotemporal attention mechanism network layer, a fully connected layer and a linear layer connected in sequence.

[0012] In some implementations, feature extraction is performed on the soil pressure data, the settlement data, the displacement data, and the pore water pressure data respectively as follows:

[0013] The soil pressure data is processed through the CNN sub-network to obtain the pressure space characteristics; the settlement data is processed through the RNN sub-network to obtain the settlement time characteristics; the displacement data is processed through the fully connected layer to obtain the displacement space characteristics; the pore water pressure data is processed through the wavelet transform to obtain the multi-scale time characteristics of the pore water pressure.

[0014] In some implementations, the adaptively fusing multiple feature extraction results using correlation to generate a fused feature vector includes:

[0015] The cosine similarities between the pressure spatial features, displacement spatial features, and multi-scale time features of pore water pressure and the settlement time features are calculated respectively, and normalized by the softmax function to determine the corresponding attention weights.

[0016] According to the corresponding attention weights, the pressure spatial characteristics, displacement spatial characteristics, pore water pressure multi-scale time characteristics and settlement time characteristics are weighted fused.

[0017] In some embodiments, the soil pressure data, the settlement data, the displacement data and the pore water pressure data are collected through sensor nodes, and the sensor nodes transmit the collected soil pressure data, the settlement data, the displacement data and the pore water pressure data to a sink node within a regional range and perform preprocessing.

[0018] In some embodiments, the adaptive residual connection LSTM network layer includes a multi-layer LSTM network, and the multi-layer LSTM network is residually connected, and the residual connection weight matrix is ​​adaptively adjusted according to the statistical characteristics of the input data.

[0019] In some embodiments, the hierarchical spatiotemporal attention mechanism network layer includes a convolutional layer, a spatial feature weighting module, and a temporal feature weighting module connected in sequence;

[0020] The convolution layer is used to extract the local correlation of the input data and obtain local features; the spatial feature weighting module is used to process the local features through spatial attention and obtain weighted spatial feature data; the temporal feature weighting module is used to process the weighted spatial feature data using temporal attention and generate spatiotemporal weighted feature data.

[0021] In a second aspect, the present invention provides a system for monitoring differential settlement of a bridge head on a pipe pile reinforced composite foundation;

[0022] A pipe pile reinforced composite foundation bridge head differential settlement monitoring system, comprising:

[0023] The multimodal data fusion module is configured to: obtain the soil pressure data, settlement data, displacement data and pore water pressure data of the pipe pile reinforced composite foundation bridge head and perform feature extraction respectively, and adaptively fuse multiple feature extraction results by using correlation to generate a fusion feature vector;

[0024] The settlement prediction module is configured to: process the fused feature vector through a trained settlement prediction model, and obtain a settlement prediction result guided by spatiotemporal features;

[0025] Among them, the sedimentation prediction model includes an adaptive residual connection LSTM network layer, a hierarchical spatiotemporal attention mechanism network layer, a fully connected layer and a linear layer connected in sequence.

[0026] In a third aspect, the present invention provides an electronic device;

[0027] An electronic device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-mentioned method for monitoring differential settlement of a bridge head of a pipe pile reinforced composite foundation.

[0028] In a fourth aspect, the present invention provides a computer-readable storage medium;

[0029] A computer-readable storage medium stores a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation.

[0030] In a fifth aspect, the present invention provides a computer program product;

[0031] A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. The technical solution provided by the present invention can give full play to the advantages of multi-source sensors and wireless sensor networks to achieve comprehensive monitoring of complex ground-based environments; it overcomes the defects of limited monitoring dimensions of a single sensor and forms a rich information set. At the same time, the use of wireless sensor networks effectively avoids the wiring difficulties of traditional wired transmission, improves the flexibility and scalability of the system, and thus improves the accuracy and reliability of monitoring.

[0034] 2. The technical solution provided by the present invention can accurately and comprehensively capture the changes in the internal mechanical state of the composite foundation under load by reasonably distributing different types of sensors at key positions and specific locations of pipe piles, reinforcing materials and bridgehead roadbed; it fully considers the interaction between pipe piles and foundation soil and reinforcing materials, as well as the transmission path and coordinated deformation mechanism of loads in the composite foundation. Compared with the traditional general sensor arrangement method, it has significant scientificity and pertinence, and provides a more accurate and detailed data source for in-depth research on the settlement mechanism of composite foundations.

[0035] 3. The technical solution provided by the present invention can effectively process high-dimensional and multi-modal sensor data, thereby making full use of different types of sensor data, mining feature associations, and obtaining comprehensive and accurate information. It can also enhance the network's ability to learn complex data features, ensuring effective training of deep networks and enhancing feature capture capabilities. Not only can it more accurately learn the spatiotemporal features and complex relationships in multi-dimensional sensor data, thereby improving the accuracy of settlement prediction, but more importantly, it can deeply reveal the settlement mechanism, conduct intelligent risk assessment and decision-making optimization support, and promote the leap of settlement monitoring technology for pipe pile reinforced composite foundations from simple data collection to intelligent analysis and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0037] Figure 1 A schematic diagram of the flow chart of a method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation provided by an embodiment of the present invention;

[0038] Figure 2 Schematic diagram of the design of the pipe pile reinforced composite foundation provided by the embodiment of the present invention; wherein (a) is a schematic diagram of the longitudinal section of the prefabricated pipe pile treatment of the bridgehead section, and (b) is a schematic diagram of the plan view of the prefabricated pipe pile treatment of the bridgehead section;

[0039] Figure 3 A schematic diagram of the plan layout of an earth pressure cell provided in an embodiment of the present invention;

[0040] Figure 4 A schematic diagram of the AA section arrangement of the earth pressure cell provided in an embodiment of the present invention;

[0041] Figure 5 A schematic diagram of the BB section arrangement of the earth pressure cell provided in an embodiment of the present invention;

[0042] Figure 6 A schematic diagram of the plan layout of a sedimentation instrument provided in an embodiment of the present invention;

[0043] Figure 7 A schematic diagram of the arrangement of measuring points for the middle and lower layers of high-strength reinforced geotextile provided in an embodiment of the present invention;

[0044] Figure 8 A schematic diagram of the arrangement of measuring points for the middle and upper layers of high-strength reinforced geotextile provided in an embodiment of the present invention;

[0045] Fig. 9 A schematic diagram of a hierarchical wireless sensor network architecture provided by an embodiment of the present invention;

[0046] Fig.10 A schematic diagram of a deep learning process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0048] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0049] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0050] Embodiment 1

[0051] At present, a single sensor is mostly used to monitor the settlement of pipe pile reinforced composite foundation, but there is a lack of in-depth analysis and understanding of the settlement mechanism, which affects the accuracy and comprehensiveness of settlement monitoring and early warning. Therefore, the present invention provides a method for monitoring differential settlement of bridge heads of pipe pile reinforced composite foundation, arranges sensors and collects data according to the characteristics of pipe pile reinforced composite foundation, and uses deep learning algorithms to achieve real-time monitoring and accurate early warning of settlement.

[0052] Next, combine Figure 1-Figure 10 , a method for monitoring differential settlement of a bridge head on a pipe pile reinforced composite foundation disclosed in this embodiment is described in detail. The method for monitoring differential settlement of a bridge head on a pipe pile reinforced composite foundation comprises the following steps:

[0053] S1. Obtain soil pressure data, settlement data, displacement data and pore water pressure data in the monitoring area of ​​the pipe pile reinforced composite foundation bridge head.

[0054] In this embodiment, soil pressure data is collected by a soil pressure box, settlement data is collected by an intelligent settlement meter, displacement data is collected by a flexible displacement meter, and pore water pressure data is collected by a pore water pressure sensor.

[0055] In order to give full play to the advantages of multi-source sensors and reveal the working mechanism of the pipe pile reinforced composite foundation, as an implementation method, before executing S1, it also includes: selecting and arranging sensors according to the characteristics of the pipe pile reinforced composite foundation.

[0056] Specifically, first, the design parameters of the pipe pile reinforced composite foundation are obtained, including the diameter, length, and spacing of the pipe piles, the number of layers of the reinforced geotextile, the laying position and range, the distribution and thickness of each soil layer of the foundation, etc.

[0057] For example, the diameter of the pipe pile is 40cm, the length is 10m-20m, the spacing is 2.5m, 2.8m and 3.0m, the number of layers of reinforced geotextile is 2, and the distribution of the upper soil of the pipe pile is from bottom to top as follows: 15cm thick gravel cushion + 5cm thick medium coarse sand + one layer of geotextile + 5cm thick medium coarse sand + 15cm thick gravel cushion + 5cm thick medium coarse sand + one layer of geotextile + 5cm thick medium coarse sand + 15cm thick gravel cushion.

[0058] Then, the above parameters are input into professional geotechnical engineering numerical simulation software to establish a numerical model of pipe pile reinforced composite foundation. In the model, the geometric shape of the pipe pile is accurately set to be circular, the material is prestressed concrete, and the corresponding elastic modulus is set according to the material; the arrangement of the reinforcement material is horizontal layered laying, and the stiffness is set according to the material properties; the foundation soil is set according to the actual layering situation, and the mechanical parameters of each soil layer, such as gravity, cohesion, internal friction angle, etc., are taken according to the survey report. By simulating the situation under different loads (such as the gradual increase of embankment fill load), the stress-strain transfer law is analyzed, and it is found that when the load is transmitted to the deep foundation through the pipe pile, the stress of the pile body gradually increases, and the stress of the surrounding soil body also changes accordingly, clarifying the principle of pile-soil cooperation and the mechanical behavior characteristics of each part of the structure.

[0059] Finally, based on the working mechanism demonstrated by the above numerical model, various types of sensors are reasonably selected to form a multi-source sensor system according to the different parts and mechanical behaviors of the pipe pile reinforced composite foundation and considering the engineering geological information and construction information.

[0060] Specifically, at the top of the pipe piles and the soil between the piles, it is necessary to monitor the vertical load transfer and settlement differences. A high-precision intelligent settlement meter is selected, which can directly reflect the deformation of the foundation in the vertical direction. As the embankment filling height increases, considering the soil arch effect and stress distribution, an intelligent soil pressure box is selected to monitor the stress conditions of each layer of the embankment and the stress distribution of the soft soil at the pile end and between the piles. An intelligent flexible displacement meter is used to monitor the deformation of the reinforced geotextile. In the soft soil area, pore water pressure sensors are arranged to monitor the changes in pore water pressure to reflect the soil saturation state and the degree of drainage consolidation.

[0061] At the same time, it is necessary to ensure that the performance indicators of the selected sensors, such as measurement accuracy, range, and sensitivity, match the actual monitoring needs of the pipe pile reinforced composite foundation. The range of the settlement meter can be determined according to the maximum settlement estimated by the numerical model, and the range of the soil pressure box can be determined according to the maximum soil pressure that may occur, to ensure that the sensor can accurately capture small changes in actual monitoring and will not be damaged due to insufficient range. These sensors work together to obtain changes in the physical parameters of the pipe pile reinforced composite foundation from different angles, realizing comprehensive monitoring of the multi-dimensional mechanical state of the composite foundation, while a single sensor cannot obtain this comprehensive information at the same time.

[0062] In this embodiment, the soil pressure cell (range: 0.6MPa; accuracy: 0.1%FS), the settlement meter (range: 400mm; accuracy: 0.5%FS), the flexible displacement meter (range: 50mm; accuracy: 0.5%FS; gauge length: 220mm) and the pore water pressure meter (range: 0-0.6MPa, accuracy: 0.1%FS) are mainly used to form a multi-source sensor system.

[0063] Here, the geotechnical engineering numerical simulation software may be an existing finite element software. The above-mentioned numerical model construction, mechanism research and other contents are all built-in functions of the software. Its specific implementation method is well known in the art. This embodiment does not improve it and will not be repeated here.

[0064] The pipe pile reinforced composite foundation bears the load through the synergistic effect of pipe piles, reinforcement materials and foundation soil. The pipe piles provide vertical support and transfer the upper load to the deep foundation. The pile body provides lateral constraints, limiting the lateral deformation of the soil. At the same time, the soil provides passive earth pressure on the pile body, enhancing the stability of the pile. The reinforcement materials (such as geogrids, geotextiles, etc.) increase the shear strength and ductility of the foundation through their tensile strength. They form a kind of tensioning effect in the soil, improve the stress-strain behavior of the soil, increase the bearing capacity and stability of the foundation, and limit the lateral displacement of the soil. The foundation soil bears part of the load and provides lateral constraints. Its structural features include pipe piles arranged at a certain distance and reinforcement materials laid in layers in specific soil layers.

[0065] In view of the structural characteristics and load transfer mechanism described above, this embodiment proposes a targeted sensor arrangement scheme. This scheme takes into account the hierarchical structure of the foundation, the interaction between piles and soil, and the distribution of reinforcement materials, ensuring that the sensors can accurately capture the deformation characteristics of the foundation under different load conditions and better reflect its load transfer and coordinated deformation mechanism.

[0066] Specifically, the sensor layout plan must meet the following requirements:

[0067] (1) The measuring points of different monitoring projects on the same road section need to be arranged on the same section. This arrangement is conducive to the care of the measuring points, centralized observation, unified observation frequency, and more importantly, it is convenient for the comprehensive analysis of the data of various observation projects.

[0068] (2) According to the pile-soil interaction and cooperative deformation mechanism, sensors are reasonably distributed in different areas. Representative monitoring sections are selected along the longitudinal and transverse directions of the roadbed, and sensors are arranged at the pile center and the soil between piles in the longitudinal and transverse directions of the bridge to form a spatial monitoring network, so as to fully understand the deformation characteristics and cooperative working conditions of the foundation in different directions. At the same time, during the filling process, the soil arching effect formed around the pile body can redistribute the load and reduce the stress concentration of the soil between piles. For this reason, considering the soil arching effect, the measurement points are densely arranged at 1.5 times the pile spacing inside the fill roadbed, which can more accurately capture the development process of the soil arching effect and its impact on settlement.

[0069] (3) Sensors are arranged at key locations such as pipe piles, reinforcement materials, and bridgehead roadbed.

[0070] The pipe pile transfers the load of the upper structure to the deeper soil layer through its high-rigidity pile body, thus overcoming the influence of the soft soil layer and reducing the settlement of the foundation. For this reason, the pile top is an important node for load transfer. The sensor can directly monitor the size and change of the load borne by the pile body. Therefore, the earth pressure box and settlement meter are arranged on the pile top of the pipe pile in the foundation soil monitoring area; there is an interaction between the pipe pile and the surrounding soil. The stiffness and strength of the pile body can improve the bearing characteristics of the soil body. At the same time, the soil body provides lateral support to the pile body and jointly bears the load. For this reason, it is necessary to install the earth pressure box and settlement meter in the soft soil at the center of the two piles along the longitudinal bridge and the soft soil at the center of the four piles along the transverse bridge; during the filling process, the soil arch effect formed around the pile body can redistribute the load and reduce the stress concentration of the soil between the piles. In order to monitor the soil arch effect and the load transfer path, it is necessary to arrange the earth pressure box and settlement meter in the fill roadbed, and their arrangement positions correspond to the arrangement positions of the foundation soil.

[0071] The flexible displacement meters are arranged on the tendons of high-strength reinforced geotextiles. The arrangement areas are the pile tops, the centers of two piles and the centers of four piles in the transverse monitoring section of the roadbed. At each position, two flexible displacement meters are arranged perpendicular to each other to measure the longitudinal and transverse displacements of the bridge. At the same time, the number of flexible displacement meters arranged on the section close to the abutment is greater than that on the side away from the abutment. If multi-layer geotextiles are used, the number of flexible displacement meters arranged on the lower geotextile is greater than that on the upper geotextile. Pore water pressure gauges are arranged in the area near the soft soil layer and the groundwater level to monitor the saturation state and consolidation degree of the soil.

[0072] (4) The buried position of the observation element should be accurate and should comply with relevant requirements. During the monitoring period, effective protection measures should be taken for the original parts to prevent collision with construction machinery and damage caused by human factors. It is imperative that the observation work can be completed successfully and achieve satisfactory results.

[0073] For example, Figure 3As shown, the measuring points of different monitoring items (settlement, displacement, soil pressure, etc.) of the same road section are arranged in the same section and one monitoring section is selected along the longitudinal and transverse directions of the roadbed. The longitudinal AA section is selected near the center line of the road. In this embodiment, the earth pressure box and the settlement meter are arranged at the top of the piles in the foundation soil monitoring area, at the soft soil in the center of the two piles along the longitudinal bridge, and at the soft soil in the center of the four piles along the transverse bridge; the arrangement position of the earth pressure box in the fill roadbed corresponds to the position of the foundation soil one by one (such as Figure 4-Figure 6 As shown in the figure, considering that the soil arching effect mainly occurs at 1.5 times the pile spacing inside the fill roadbed, the measurement points in this section are encrypted, and two measurement points are arranged at the part exceeding 1.5 times the pile spacing to roughly observe its deformation trend. Figure 7-Figure 8 As shown, in this embodiment, the flexible displacement meter is arranged on the tendons of the high-strength reinforced geotextile. The arrangement area is the pile top, the center of two piles and the center of four piles in the transverse section of the roadbed. Two flexible displacement meters are arranged perpendicular to each other at each position to measure the longitudinal and transverse displacements of the bridge. The number of flexible displacement meters arranged on the section close to the abutment is greater than that on the side away from the abutment, and the number of flexible displacement meters arranged on the lower geotextile is greater than that on the upper geotextile. When the observation element is buried, a special fixed bracket is used to ensure the accurate position, and a protective cover is used to prevent collision with construction machinery and human damage.

[0074] Through the above-mentioned sensor layout scheme, a collaborative relationship between sensors is established, so that different types of sensors form an interconnected monitoring network in space, that is, earth pressure boxes, settlement meters, flexible displacement meters and pore water pressure sensors are formed into "sensor clusters" at key locations such as pipe piles, reinforcement materials and bridgehead roadbed. Through comprehensive analysis of the data of these "sensor clusters", the settlement mechanism and deformation pattern of the monitored area can be judged more accurately.

[0075] Based on the above sensor deployment scheme, further, this embodiment adopts a hierarchical wireless sensor network architecture, including sensor nodes, aggregation nodes and a monitoring center, the sensor nodes are communicatively connected to the aggregation nodes, and the aggregation nodes are communicatively connected to the monitoring center.

[0076] In the hierarchical wireless sensor network architecture, a hybrid network of Bluetooth and ZigBee is used in areas close to the monitoring center and with dense sensor nodes, that is, within 100m from the monitoring center. Bluetooth is used for short-distance high-speed data transmission between nodes to quickly collect data from sensor nodes in the same small area (within 10m). ZigBee builds a stable self-organizing network to reliably aggregate data to upper-level nodes. Sensor nodes use low-power microcontrollers to collect multi-source sensor data, and monitor network signal strength, transmission delay and data traffic in real time through intelligent network switching modules. When the ZigBee network load is higher than 80% (the threshold can be set), part of the data traffic is automatically switched to Bluetooth.

[0077] The aggregation node uses high-performance embedded devices to receive sensor node data and run median filtering to remove noise. Distributed storage technology is used to store data in different aggregation nodes according to the monitoring area, and a distributed database management system is established. The monitoring center configures the server, manages the entire network, stores data and performs complex analysis.

[0078] At the same time, based on the results of edge computing, the aggregation node can make some simple decisions locally. For example, when an abnormal change in the settlement data in a certain area is detected, a local warning is immediately triggered to notify nearby construction personnel to conduct a preliminary inspection without waiting for the data to be transmitted to the monitoring center for processing. This achieves real-time monitoring and rapid response, and enhances the security and reliability of the system.

[0079] Exemplarily, the installation steps of the hierarchical wireless sensor network architecture are as follows: First, according to the monitoring range and sensor distribution of the pipe pile reinforced composite foundation, the wireless sensor network nodes are reasonably arranged at key locations such as around the pipe piles and on the foundation surface to ensure that each sensor can establish a stable connection with the nearest node. Secondly, set the network's frequency band, channel, transmission rate and other parameters. Finally, after completing the network node deployment and parameter setting, initialize the wireless sensor network to put each node into normal working state. Then perform a network connectivity test to check whether the communication between each sensor node and the data processing center is normal and whether the data transmission is accurate. If a problem is found, troubleshoot the fault point in time and make adjustments to ensure that the entire wireless sensor network can operate stably and reliably.

[0080] According to the monitoring range (monitoring length is 40m, width is 35.6m) and sensor distribution, wireless sensor network nodes are arranged around the piles and on the foundation surface every 10m, and the network frequency band is set to 2.4GHz, the channel is 11, and the transmission rate is 250kbps. After initializing the nodes, a network test tool is used to perform a connectivity test to ensure that each sensor node communicates normally with the data processing center and that data transmission is accurate.

[0081] In this embodiment, the multi-dimensional sensor collects data at a set frequency (e.g., once per hour). During data preprocessing, if the measured value of the sedimentation instrument has an abnormal value of 10 mm (outside the reasonable range), it is corrected or eliminated. Then, all data are normalized to the minimum and maximum values.

[0082] S2. Feature extraction is performed on soil pressure data, settlement data, displacement data and pore water pressure data respectively; specifically including:

[0083] S201. Process the soil pressure data through the CNN sub-network to obtain pressure spatial characteristics.

[0084] In this embodiment, the CNN subnetwork processes the soil pressure data through a 3×3 convolutional layer to extract local features. The convolutional layer is set with 6 channels, and each channel performs a convolution operation with the input data through a convolution kernel to generate multiple feature maps; the pooling layer uses a maximum pooling operation to process multiple feature maps, the pooling window size is 2×2, the step size is 2, and the feature map is reduced in dimensionality to retain key information.

[0085] After convolution and pooling operations, the soil pressure data is converted into a series of representative feature maps, namely pressure space features, which reflect the spatial distribution characteristics and change trends of soil pressure, such as pressure concentration areas, pressure change gradients, etc. Finally, the feature map is flattened and converted into feature vector form to obtain pressure space features, so as to facilitate subsequent fusion based on the attention mechanism.

[0086] S202. Process the sedimentation data through the RNN sub-network to obtain sedimentation time features.

[0087] Specifically, the RNN subnetwork includes an input layer, a hidden layer, and an output layer. The hidden layer is the core part of the RNN subnetwork for data processing and feature learning. It is responsible for receiving input layer data, extracting data features through internal calculations and state updates, and passing the processed information to the output layer.

[0088] The hidden layer is the key part to realize the memory and learning function of time series data. In the application of sedimentation data feature extraction, the disadvantage of the number of hidden layers depends on the complexity of the sedimentation data. The memory unit, as a key component of the hidden layer, realizes the memory and utilization function of the hidden layer for historical information, enabling RNN to process data with time series characteristics, such as sedimentation data.

[0089] In this embodiment, the number of hidden layers is 128, and the sedimentation data is processed through 128 hidden layers to learn the time series characteristics; the RNN sub-network memorizes and learns the historical sedimentation data through the memory units in the hidden layer, and can capture the change pattern of sedimentation over time, such as the changing trend and periodic changes of the sedimentation rate.

[0090] 128 hidden layer units are used to process the settlement data and learn the time series characteristics. RNN memorizes and learns the historical settlement data through memory units, and can capture the change patterns of settlement over time, such as the changing trend and periodic changes of the settlement rate.

[0091] S203: Process the displacement data through a fully connected layer to obtain displacement space features.

[0092] The displacement data is collected by a flexible displacement meter, which is installed on the tendons of the high-strength reinforced geotextile. The displacement data in the longitudinal and transverse directions of the bridge are collected at the pile top, the center of two piles and the center of four piles in the transverse section of the roadbed.

[0093] Each neuron in the fully connected layer is connected to all neurons in the previous layer, and the connection is realized through the weight matrix. When inputting data, the fully connected layer multiplies the input vector by the weight matrix, adds the bias vector, and then performs a nonlinear transformation through the activation function to obtain the output vector.

[0094] The flexible displacement meter data reflects the deformation of the geotextile, and its data dimension and characteristics are different from those of the earth pressure and settlement data. The fully connected layer is based on the input data dimension. For example, each position has two displacement values ​​in the longitudinal direction and transverse direction. Assuming that n flexible displacement meters are arranged in the transverse section of the roadbed, the number of input layer nodes is 2n.

[0095] Specifically, the fully connected layer extracts features according to the input data dimension (each position is displaced in two directions, there are n positions in total, and the number of input layer nodes is 2n) and the preset hidden layer structure (such as two hidden layers, 8 nodes in the first layer, and 4 nodes in the second layer); the weight matrix (the weight matrix W1 dimension from the input layer to the first hidden layer is 8×6, and the weight matrix dimension W2 from the first hidden layer to the second hidden layer is 4×8) is multiplied with the input data and added with the bias vector (the bias vector b1 dimension is 8×1, and the b2 dimension is 4×1), and then processed by the ReLU activation function, the features in the displacement data are gradually extracted, and the final output of the second hidden layer is a 4-dimensional vector, that is, the displacement space feature; this vector reflects the complex relationship between the displacements of the geotextile reinforcement at different positions and the overall deformation trend and other characteristics.

[0096] S204. Process the pore water pressure data by wavelet transform to obtain multi-scale time characteristics of the pore water pressure.

[0097] When studying the uneven settlement of soil, the changes in pore water pressure at different locations and depths have different time-scale characteristics. Therefore, the pore water pressure data is processed by wavelet transform, the pore water pressure data is decomposed into different scales, and the wavelet coefficients at different scales are obtained; the statistical characteristics of the wavelet coefficients at different scales (such as mean, variance, energy, etc.) are calculated to obtain the multi-scale time characteristics of pore water pressure that can fully describe the pore water pressure data.

[0098] These statistical features can summarize the information contained in the wavelet coefficients from different angles, making the data features more concise and clear, facilitating subsequent fusion analysis with other types of data (such as earth pressure, settlement data, etc.), and input into prediction models (such as neural network models) for prediction of uneven soil settlement.

[0099] Here, S201 - S204 can be executed in parallel to improve processing efficiency.

[0100] S3. Adaptively fuse multiple feature extraction results using correlation to generate a fused feature vector.

[0101] Specifically, the cosine similarities between the pressure spatial features, displacement spatial features, multi-scale time features of pore water pressure and settlement time features are calculated respectively, and normalized by the softmax function to determine the corresponding attention weights; according to the corresponding attention weights, the pressure spatial features, displacement spatial features, multi-scale time features of pore water pressure and settlement time features are weighted fused to obtain a fused feature vector.

[0102] For example, it is assumed that the pressure space characteristic vector is expressed as P = (p 1 ,p 2 ,…p m ), the displacement space eigenvector is expressed as D = (d 1 ,d 2 ,…d m ) The multi-scale time characteristic vector of pore water pressure is expressed as W = (w 1 ,w 2 ,…w m ) Sedimentation time characteristics

[0103] ,, vector is represented as S=(s 1 ,s 2 ,…s m ), and the dimensions of each feature vector are the same.

[0104] First, the cosine similarity between the pressure spatial characteristics, displacement spatial characteristics, and multi-scale time characteristics of pore water pressure and the settlement time characteristics is calculated respectively. The cosine similarity between the pressure spatial characteristics and the settlement time characteristics is expressed as:

[0105]

[0106] The cosine similarity between the displacement space feature and the settlement time feature is expressed as:

[0107]

[0108] The cosine similarity between the multi-scale time characteristics of pore water pressure and the settlement time characteristics is expressed as:

[0109]

[0110] Then, the cosine similarities calculated above are combined into a vector X = (sim PS ,sim DS ,simWS ), also expressed as X=(x 1 ,x 2 ,x 3 ), based on this, the formula for calculating the attention weight through the softmax function is expressed as:

[0111]

[0112] Through the above steps, the cosine similarity between the pressure spatial features, displacement spatial features, and multi-scale time features of pore water pressure and the settlement time features can be calculated respectively, and normalized by the softmax function to determine the corresponding attention weights. These attention weights can be used in the subsequent feature weighted fusion process, so that the model can reasonably fuse multiple features according to the correlation between different features and settlement time features, so as to better perform tasks such as settlement prediction.

[0113] S4. Process the fused feature vector through the trained settlement prediction model, and obtain the settlement prediction result guided by the spatiotemporal features. Specifically including:

[0114] S401. Process the fused feature vector through an adaptive residual connection LSTM network to obtain a time series feature vector.

[0115] The adaptive residual connection LSTM network mainly extracts and processes the time series features of the fused feature vector. It learns the time dependency and long-term features in the input data through the operation of the LSTM network (including the forget gate, input gate, output gate and candidate memory cells) and the adaptive residual connection.

[0116] Specifically, the adaptive residual connection LSTM network includes a 3-layer LSTM network, which includes a forget gate, an input gate, an output gate, and candidate memory cells; in the LSTM network, the hidden unit is the basic component of the hidden layer, which is used to process the input data and extract features. In this embodiment, the first layer of the LSTM network has 256 hidden units, the second layer of the LSTM network has 128 hidden units, and the third layer of the LSTM network has 64 hidden units. These hidden units receive information from the input data (such as feature vectors after multimodal fusion) at each time step, as well as their own hidden state at the previous time step. They update their own hidden states through the operations of the forget gate, input gate, output gate, and candidate memory cells, thereby learning the time series features and long-term dependencies in the input data; these hidden states can capture the characteristics of the data's temporal change trend, cycle, and so on.

[0117] For example, for the input data at a certain moment, 256 hidden units (in the first layer) will process it simultaneously, and each hidden unit will calculate an output based on its own parameters (weight matrix and bias vector) and input data. These outputs are combined to form the hidden state of this layer, which is then passed to the next layer or used to output prediction results. The number of hidden units determines the complexity of the features that the network can learn. The more hidden units there are, the stronger the network's expressive power is, but it may also bring problems such as overfitting.

[0118] Furthermore, the weight matrix of the adaptive residual connection LSTM network is dynamically adjusted during the training process according to the input data characteristics.

[0119] Specifically, during the training process, the adjustment of the weight matrix is ​​achieved through the back-propagation algorithm.

[0120] In the actual prediction process, the weight matrix is ​​usually fixed; because the weight matrix has been adjusted through a large amount of training data during the training phase, so that it can effectively extract features and predict input data. When predicting, the new input data (such as the new feature vector after multimodal fusion) is input into the trained LSTM network, and the network performs forward propagation calculations according to the trained weight matrix and structure, and outputs the prediction results.

[0121] Furthermore, a residual connection is added between the input and output of each layer of LSTM, that is, the input of a layer of LSTM is x t , the output is h t , the next layer input x t+1 =h t ×W res +x t .

[0122] Here, the adaptive residual connection weight matrix is ​​adjusted by learning the mean and variance of the input data through a small neural network.

[0123] Specifically, first, for the input data (such as the feature vector after multimodal fusion), its mean and variance need to be calculated; a small neural network (such as a simple multi-layer perceptron, MLP) can be constructed to adjust the adaptive residual connection weight matrix according to the mean and variance. This small neural network takes the mean and variance as input and outputs the adjusted weight matrix. Use the training data to train this small neural network. The goal of training is to enable the small neural network to output a suitable weight matrix based on the mean and variance of the input data to optimize the performance of the LSTM network. During the training process, a loss function is defined to measure the effectiveness of the weight matrix output by the small neural network.

[0124] S402: Input the processed fusion feature vector into the hierarchical spatiotemporal attention mechanism network for processing, and output the settlement prediction result.

[0125] After the processed fused feature vector is input into the hierarchical spatiotemporal attention mechanism network, it first enters the convolution layer, and the convolution kernel of the convolution layer is set to 3×3 to extract the local correlation of the input data; the feature map output by the convolution layer is expanded into a vector as the input of the fully connected layer; the fully connected layer outputs the spatial attention weight corresponding to each position by learning the overall information of the input feature map, and reshapes the spatial attention weight vector output by the fully connected layer into a spatial attention weight matrix, in which each element of the matrix represents the importance weight of the corresponding position in space; then, the original spatial feature data (i.e., the feature map output by the convolution layer) is multiplied by the matrix to obtain the weighted spatial feature data. When calculating time attention, the weighted spatial feature data is used as input, and the hidden unit of the LSTM network is set to 128 learning time series features. At each time step, the LSTM unit receives the weighted spatial feature data as input, and combines the hidden state and memory cell state of the previous time step for calculation. The hidden state of the last time step of the LSTM network is used as the input of the fully connected layer. The fully connected layer calculates the time attention weight through the weight matrix and the bias vector; the weighted spatial feature data is multiplied by the time attention weight in the time dimension to obtain the spatiotemporal weighted feature data; the spatiotemporal weighted feature data is input into the fully connected layer, processed by the weight matrix and the ReLU activation function, and finally, the linear layer outputs the settlement prediction result.

[0126] In this embodiment, the settlement prediction result may be a settlement value at a certain moment in the future.

[0127] Furthermore, with the help of visualization software, the settlement monitoring data (historical settlement curve), prediction results (future settlement trend chart) and settlement mechanism analysis results (settlement contribution ratio chart of different regions) are displayed in an intuitive graphical interface, that is, a curve of settlement changing with time in a certain period of time and a cloud map of settlement amount at different locations are drawn.

[0128] The working status of the foundation is evaluated based on the monitoring results. If the predicted settlement value is close to the design allowable settlement value, the foundation is judged to be in a critical state. This provides a basis for engineering decision-making and determines whether to carry out local reinforcement of the foundation (such as increasing the number of pipe piles in areas with large settlement) or adjust construction parameters (such as controlling the filling rate). At the same time, the actual settlement data is regularly compared with the predicted results. If the prediction error is large (such as the mean square error exceeds 0.5mm), the foundation is judged to be in a critical state. 2 ), then readjust the model parameters (such as adjusting the neural network weights) to improve the model prediction accuracy.

[0129] Embodiment 2

[0130] This embodiment discloses a differential settlement monitoring system for a bridge head of a pipe pile reinforced composite foundation, comprising:

[0131] The multimodal data fusion module is configured to: obtain the soil pressure data, settlement data, displacement data and pore water pressure data of the pipe pile reinforced composite foundation bridge head and perform feature extraction respectively, and adaptively fuse multiple feature extraction results by using correlation to generate a fusion feature vector;

[0132] The settlement prediction module is configured to: process the fused feature vector through a trained settlement prediction model, and obtain a settlement prediction result guided by spatiotemporal features;

[0133] Among them, the sedimentation prediction model includes an adaptive residual connection LSTM network layer, a hierarchical spatiotemporal attention mechanism network layer, a fully connected layer and a linear layer connected in sequence.

[0134] It should be noted that the multimodal data fusion module and the settlement prediction module correspond to the steps in Embodiment 1, and the examples and application scenarios implemented by the modules and the corresponding steps are the same, but are not limited to the contents disclosed in Embodiment 1. It should be noted that the modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0135] Embodiment 3

[0136] Embodiment 3 of the present invention provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps of the above-mentioned method for monitoring differential settlement of a bridge head of a pipe pile reinforced composite foundation are completed.

[0137] Embodiment 4

[0138] Embodiment 4 of the present invention provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the above-mentioned method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation are completed.

[0139] Embodiment 5

[0140] Embodiment 5 of the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation.

[0141] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0142] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0143] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0144] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0145] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for monitoring differential settlement of a bridge head on a pipe pile reinforced composite foundation, characterized in that: include: The soil pressure data, settlement data, displacement data and pore water pressure data of the pipe pile reinforced composite foundation bridge head are obtained and feature extraction is performed respectively. The multiple feature extraction results are adaptively fused by using correlation to generate a fused feature vector. The fused feature vector is processed by a trained settlement prediction model, and the settlement prediction result is obtained by taking the spatiotemporal features as a guide; Among them, the sedimentation prediction model includes an adaptive residual connection LSTM network layer, a hierarchical spatiotemporal attention mechanism network layer, a fully connected layer and a linear layer connected in sequence.

2. The method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation according to claim 1, characterized in that: The feature extraction of the soil pressure data, the settlement data, the displacement data and the pore water pressure data is specifically performed as follows: The soil pressure data is processed by the CNN sub-network to obtain pressure space features; the settlement data is processed by the RNN sub-network to obtain settlement time features; the displacement data is processed by the fully connected layer to obtain displacement space features; The pore water pressure data are processed by wavelet transform to obtain multi-scale time characteristics of pore water pressure.

3. The method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation according to claim 2, characterized in that: Adaptively fusing multiple feature extraction results using correlation to generate a fused feature vector includes: The cosine similarities between the pressure spatial features, displacement spatial features, and multi-scale time features of pore water pressure and the settlement time features are calculated respectively, and normalized by the softmax function to determine the corresponding attention weights. According to the corresponding attention weights, the pressure spatial characteristics, displacement spatial characteristics, pore water pressure multi-scale time characteristics and settlement time characteristics are weighted fused.

4. The method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation according to claim 1, characterized in that: The soil pressure data, the settlement data, the displacement data and the pore water pressure data are collected through sensor nodes, and the sensor nodes transmit the collected soil pressure data, the settlement data, the displacement data and the pore water pressure data to a sink node within a regional range and perform preprocessing.

5. The method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation according to claim 1, characterized in that: The adaptive residual connection LSTM network layer includes a multi-layer LSTM network, and the multi-layer LSTM networks are residually connected, and the residual connection weight matrix is ​​adaptively adjusted according to the statistical characteristics of the input data.

6. The method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation according to claim 1, characterized in that: The hierarchical spatiotemporal attention mechanism network layer includes a convolutional layer, a spatial feature weighting module and a temporal feature weighting module connected in sequence; The convolution layer is used to extract the local correlation of the input data and obtain local features; the spatial feature weighting module is used to process the local features through spatial attention and obtain weighted spatial feature data; the temporal feature weighting module is used to process the weighted spatial feature data using temporal attention and generate spatiotemporal weighted feature data.

7. A pipe pile reinforced composite foundation bridge head differential settlement monitoring system, characterized in that: include: The multimodal data fusion module is configured to: obtain the soil pressure data, settlement data, displacement data and pore water pressure data of the pipe pile reinforced composite foundation bridge head and perform feature extraction respectively, and adaptively fuse multiple feature extraction results by using correlation to generate a fusion feature vector; The settlement prediction module is configured to: process the fused feature vector through a trained settlement prediction model, and obtain a settlement prediction result guided by spatiotemporal features; Among them, the sedimentation prediction model includes an adaptive residual connection LSTM network layer, a hierarchical spatiotemporal attention mechanism network layer, a fully connected layer and a linear layer connected in sequence.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation as described in any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the method for monitoring differential settlement of a bridge head in a pipe pile reinforced composite foundation as described in any one of claims 1 to 6 are implemented.

Images