Method for resisting non-uniform settlement by additionally installing flexible connection nodes of elevator

Abstract

The invention provides a non-uniform settlement resisting method for flexible connection nodes of an added elevator, and belongs to the technical field of added elevator construction. Multi-point settlement monitoring devices are arranged at the connection positions of an existing building and the added elevator to collect settlement data in real time, and a three-dimensional distribution database is established; an adaptive response model is established based on multi-source monitoring data, geological survey parameters are fused, a machine learning algorithm is utilized to predict a settlement development trend, and a rapid and slow response adjustment mechanism based on a settlement rate change rate threshold value is established; a variable attention optimization model is applied to carry out node damage degree evaluation through frequency vibration modes and damping parameter changes, a slip amount self-adaptive adjustment function is constructed, a slip adjustment value is calculated based on the settlement amount difference shear force change rate and the rotation angle change amount, and the bolt pre-tightening force is automatically adjusted. The technical problem that rigid damage is prone to happening to connection joints of an additionally-installed elevator and an existing building under the effect of non-uniform settlement is solved.

Description

Technical Field

[0001] This invention belongs to the field of elevator installation technology, and specifically relates to a method for resisting non-uniform settlement of flexible connection nodes in elevator installation. Background Technology

[0002] With the widespread implementation of elevator retrofitting projects in existing buildings, connection nodes, as key connection points between the elevator steel structure and the existing building's concrete structure, have seen increasingly mature design and construction technologies. Traditional techniques mainly employ rigid connection methods such as embedded steel plates, chemical anchors, or rebar installation. These methods ensure the load-bearing capacity of the connection node by improving the bond strength and anchorage depth at the connection interface. Under standard geological conditions, these traditional methods can meet the structural safety requirements for normal elevator operation. However, traditional rigid connection nodes exhibit significant technical deficiencies when facing complex geological environments. Particularly when non-uniform settlement occurs in the foundations of the existing building and the retrofitted elevator, rigid connections cannot adapt to the deformation coordination requirements caused by differential settlement. This leads to excessive tensile stress, shear stress, and bending moment at the connection interface, easily causing localized cracking, anchor breakage, or concrete spalling, seriously threatening elevator operational safety and building structural integrity. In current engineering practice, existing buildings, due to their earlier construction dates and relatively lower foundation treatment standards, coupled with the complexity of urban geological conditions and the influence of groundwater level changes, commonly experience non-uniform settlement problems. Traditional rigid connection nodes struggle to adapt to this complex deformation environment, frequently resulting in connection failures and engineering accidents. In other words, existing technologies have a technical problem where the connection between the elevator installation and the existing building is prone to rigid failure under non-uniform settlement. Summary of the Invention

[0003] In view of this, the present invention provides a method for resisting non-uniform settlement of the flexible connection node of the elevator installation, which can solve the technical problem in the prior art that the connection node between the elevator installation and the existing building is prone to rigid failure under non-uniform settlement.

[0004] This invention is implemented as follows: It provides a method for resisting non-uniform settlement at flexible connection nodes for elevator installations. Multiple settlement monitoring devices are deployed at the connection point between the existing building and the elevator installation to collect real-time data on foundation settlement, settlement rate, and soil compression modulus, establishing a three-dimensional settlement distribution database. An adaptive response model is established based on the multi-source monitoring data, integrating shear modulus, elastic modulus, and consolidation coefficient from geological survey data. Machine learning algorithms are used to predict settlement trends, outputting a settlement prediction matrix. A multi-level buffer flexible connection node is designed, comprising an elasto-plastic material layer and a sliding adjustment mechanism. The elasto-plastic material layer uses different compression indices. The material combination and the sliding adjustment mechanism control the friction coefficient through bolt preload to achieve coordination between load-bearing capacity and deformation; real-time monitoring of shear force, bending moment, and displacement changes in multi-level buffer flexible connection nodes, and execution of rapid or slow response adjustment based on the settlement rate change rate and duration; application of variational attention optimization model to perform real-time state assessment of multi-level buffer flexible connection nodes, establishing a grey relational analysis through changes in frequency, mode shape, and damping parameters to construct a node damage assessment system; establishment of an adaptive sliding adjustment function, calculating the sliding adjustment value based on the settlement difference, shear force change rate, and rotation angle change, and executing the corresponding bolt preload adjustment strategy according to the sliding adjustment value range.

[0005] The multi-point settlement monitoring device includes a laser displacement sensor and a tiltmeter. The laser displacement sensor is used to measure the settlement vertically downwards, and the tiltmeter is used to measure the angle change.

[0006] The laser displacement sensor has a measurement accuracy of 0.1 mm, and the settling rate is the first derivative of the settling amount with respect to time.

[0007] The compression modulus is determined by indoor compression tests or field load tests, reflecting the compression deformation characteristics of soil under pressure.

[0008] The shear modulus is obtained through an indoor triaxial shear test and characterizes the soil's ability to resist shear deformation. The elastic modulus is determined through a uniaxial compression test and represents the ratio of stress to strain within the elastic range of the material.

[0009] The consolidation coefficient is obtained through indoor consolidation tests and reflects the time-effect characteristics of soil consolidation deformation.

[0010] The machine learning algorithm predicts the subsidence development trend over the next 72 hours.

[0011] The friction coefficient is obtained through on-site friction tests or material surface roughness tests, and its value ranges from 0.2 to 0.8. The bolt preload is applied by a torque wrench and monitored in real time by a force sensor.

[0012] The shear force and bending moment are measured by a force sensor composed of strain gauges, which is installed at the key stress positions of the multi-level buffer flexible connection node. The displacement is measured by a displacement sensor to measure the linear movement of the multi-level buffer flexible connection node in each direction.

[0013] Specifically, when the rate of change of the settling rate exceeds 0.5% / second and the duration is greater than 30 seconds, a rapid response adjustment is immediately implemented; when the rate of change of the settling rate exceeds 0.2% / second but the duration is less than 30 seconds, a slow response adjustment is implemented.

[0014] The rapid response adjustment is an emergency response mechanism that immediately adjusts the bolt preload and sliding adjustment mechanism parameters when the rate of change of settlement rate is detected to exceed a preset threshold and the duration is relatively long. The adjustment time is controlled to be completed within 5 seconds, and the adjustment range is 20% to 50% of the current parameter value.

[0015] The slow response adjustment is a gradual response mechanism that gradually fine-tunes the bolt preload and sliding adjustment mechanism parameters when the rate of change of settlement rate exceeds a preset threshold but the duration is short. The adjustment time is 30 to 120 seconds and the adjustment range is 5% to 15% of the current parameter value.

[0016] The frequency is obtained by collecting vibration signals from the multi-level buffer flexible connection node using a vibration sensor and performing a fast Fourier transform. The mode shape is obtained by measuring the vibration mode parameters of the multi-level buffer flexible connection node using a modal analyzer. The damping parameter is obtained by measuring the attenuation characteristics of the vibration signal of the multi-level buffer flexible connection node using the free attenuation method.

[0017] The variational attention optimization model is an intelligent assessment system built on the grey relational neural network damage identification algorithm. By establishing a grey relational analysis model of node damage degree and dynamic characteristic parameters, it identifies frequency change rate, mode shape offset and damping ratio change as key damage indicators, and uses the variational attention optimization model to quantitatively assess the damage degree.

[0018] The variational attention optimization model is a five-layer neural network architecture consisting of an input layer, a variational coding layer, an attention mechanism layer, a feature fusion layer, and an output layer. The input layer receives three key dynamic parameters: frequency change rate, mode shape offset, and damping ratio change. The variational coding layer maps the input parameters to the latent space for feature abstraction through variational inference. The attention mechanism layer dynamically adjusts the attention weight allocation based on the real-time values ​​of settlement, bolt preload, and friction coefficient.

[0019] Specifically, when the slip adjustment value ∈ [0, 5) mm, the existing bolt preload is maintained; when the slip adjustment value ∈ [5, 15) mm, the bolt preload is increased by 15%; and when the slip adjustment value ∈ [15, +∞) mm, an emergency slip mode is used for adjustment. The slip amount adaptive adjustment function calculates the slip adjustment value based on the settlement difference, shear force change rate, and rotation angle change, with weighting coefficients of 0.5, 0.3, and 0.2, respectively.

[0020] The key locations refer to the junction of the existing building foundation and the elevator installation foundation, the connection between the existing building wall and the elevator shaft, the docking surface between the existing building floor slab and the elevator platform, and the connection point between the existing building structural column and the elevator support structure.

[0021] The key stress-bearing parts refer to the connection interface of the multi-level buffer flexible connection node that bears the maximum shear force, the rotational hinge point that bears the maximum bending moment, the main bearing section that bears the maximum axial force, and the bolt connection assembly surface that bears the maximum torque.

[0022] This invention effectively solves the problem of damage to traditional rigid connections under non-uniform settlement conditions by designing multi-level buffered flexible connection nodes and establishing an intelligent monitoring and adjustment system. The invention employs an innovative combination of elasto-plastic material layers and a sliding adjustment mechanism, enabling the connection node to possess graded deformation capabilities. It can automatically select elastic deformation, plastic deformation, or sliding deformation modes according to the amount of settlement. Through the nonlinear characteristics of the material, it actively absorbs the deformation energy generated by non-uniform settlement, avoiding excessive stress concentration at the connection interface. Simultaneously, the sliding adjustment mechanism achieves dynamic adjustment of the friction coefficient through precise control of bolt preload, ensuring that the connection node maintains stable load-bearing performance under different settlement conditions. The real-time state assessment system established by this invention, based on a variational attention optimization model, can accurately identify the degree of damage to the connection node through changes in dynamic characteristic parameters. Combined with an adaptive sliding adjustment function, it achieves intelligent parameter adjustment, transforming the connection node from passively bearing settlement to actively adapting to settlement changes. This completely solves the technical problem of rigid failure of existing building elevator connection nodes under non-uniform settlement. Attached Figure Description

[0023] Figure 1 This is a flowchart of the method of the present invention.

[0024] Figure 2 The graph shows the time history of the dynamic parameters of the variational attention optimization model in the example.

[0025] Figure 3 This is a diagram showing the change in attention weight allocation for multi-level buffered flexible connection nodes in the embodiment. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0027] like Figure 1 The diagram shown is a flowchart of a method for adding flexible connection nodes to elevators to resist non-uniform settlement, provided by the present invention. This method includes the following steps:

[0028] S01. Install multi-point settlement monitoring devices at the connection between the existing building and the elevator installation location to collect the building foundation settlement, settlement rate and soil compression modulus in real time, and establish a three-dimensional settlement distribution database. The multi-point settlement monitoring devices include laser displacement sensors and tiltmeters.

[0029] S02. Based on multi-source monitoring data, establish an adaptive response model, integrate the shear modulus, elastic modulus and consolidation coefficient in geological exploration data, use machine learning algorithms to predict the settlement development trend in the next 72 hours, and output the settlement prediction matrix.

[0030] S03. Design a multi-level buffer flexible connection node, wherein the multi-level buffer flexible connection node includes an elastic-plastic material layer and a sliding adjustment mechanism. The elastic-plastic material layer adopts a combination of materials with different compression indices. The sliding adjustment mechanism controls the friction coefficient through bolt preload to achieve coordination between load-bearing capacity and deformation.

[0031] S04. Real-time monitoring of shear force, bending moment and displacement changes of multi-level buffer flexible connection nodes. When the settlement rate change rate exceeds 0.5% / second and the duration is greater than 30 seconds, fast response adjustment is immediately executed. When the settlement rate change rate exceeds 0.2% / second but the duration is less than 30 seconds, slow response adjustment is executed.

[0032] S05. Apply variational attention optimization model to perform real-time state assessment of multi-level buffer flexible connection nodes. Establish grey relational analysis through changes in frequency, mode shape and damping parameters, construct a node damage assessment system and identify key damage indicators.

[0033] S06. Establish an adaptive adjustment function for slippage. The adaptive adjustment function for slippage calculates the slippage adjustment value based on the settlement difference, shear force change rate and rotation angle change. When the slippage adjustment value is less than 5 mm, maintain the existing bolt preload. When the slippage adjustment value is ∈ [5, 15) mm, increase the bolt preload by 15%. When the slippage adjustment value is greater than or equal to 15 mm, adopt the emergency slippage mode for adjustment.

[0034] Settlement is measured vertically downwards using a laser displacement sensor with an accuracy of 0.1 mm. The settlement rate is the first derivative of settlement with respect to time. Compression modulus is determined through indoor compression tests or field load tests, reflecting the compressive deformation characteristics of soil under pressure. Shear modulus is obtained through indoor triaxial shear tests, characterizing the soil's resistance to shear deformation. Elastic modulus is determined through uniaxial compression tests, representing the stress-strain ratio within the elastic range. Consolidation coefficient is measured through indoor consolidation tests, reflecting the time-dependent characteristics of soil consolidation deformation. Friction coefficient is obtained through field friction tests or material surface roughness tests, typically ranging from 0.2 to 0.8. Bolt preload is applied using a torque wrench and monitored in real-time by a force sensor, controlling the tightness of the bolt connection. Shear force and bending moment are measured using a force sensor composed of strain gauges, installed at key stress locations in the multi-stage buffer flexible connection node. Displacement is measured by displacement sensors to measure the linear movement of the multi-stage buffer flexible connection node in all directions, and angular change is measured by an inclinometer to measure the angular change of the multi-stage buffer flexible connection node. The frequency was obtained by acquiring vibration signals from the multi-stage buffered flexible connection node using vibration sensors and performing a Fast Fourier Transform. The mode shape was measured using a modal analyzer to determine the vibration modal parameters of the multi-stage buffered flexible connection node. The damping parameters were obtained by measuring the attenuation characteristics of the vibration signals from the multi-stage buffered flexible connection node using the free decay method.

[0035] The rapid response adjustment is an emergency response mechanism that immediately adjusts the bolt preload and sliding adjustment mechanism parameters when the detected settlement rate change rate exceeds a preset threshold and lasts for a prolonged period. The adjustment time is controlled within 5 seconds, and the adjustment range is 20% to 50% of the current parameter value. The slow response adjustment is a gradual response mechanism that progressively fine-tunes the bolt preload and sliding adjustment mechanism parameters when the detected settlement rate change rate exceeds a preset threshold but lasts for a short period. The adjustment time is 30 to 120 seconds, and the adjustment range is 5% to 15% of the current parameter value.

[0036] The variational attention optimization model is an intelligent assessment system built upon a grey relational-neural network damage identification algorithm. This model establishes a grey relational analysis model linking node damage severity to dynamic characteristic parameters, identifying frequency change rate, mode shape shift, and damping ratio changes as key damage indicators. The model then quantitatively assesses the damage severity. The variational attention optimization model is particularly suitable for this solution because the loads borne by the multi-stage buffer flexible connection nodes of the elevator exhibit significant periodicity and dynamic characteristics. Traditional static detection methods struggle to capture the damage evolution of these nodes under reciprocating loads. Grey relational analysis effectively handles small sample sizes and multi-factor relationships, accurately identifying the intrinsic connection between node damage caused by settlement and changes in dynamic characteristic parameters. The variational attention optimization model, by learning from a large amount of damage pattern data, achieves precise quantitative assessment of the damage severity of the multi-stage buffer flexible connection nodes, providing a scientific basis for maintenance decisions. The variational attention optimization model first uses frequency domain analysis to extract the natural frequencies, mode shape parameters, and modal damping ratios of the multi-level buffer flexible connection nodes. Then, it uses the grey relational analysis method to analyze the correlation strength between each dynamic parameter and the degree of damage, and selects parameters with a correlation degree greater than 0.7 as input features of the variational attention optimization model. Finally, the variational attention optimization model outputs the damage degree assessment value and the remaining bearing capacity prediction value, realizing intelligent monitoring and early warning of the state of the multi-level buffer flexible connection nodes.

[0037] The variational attention optimization model is structured as a five-layer neural network architecture comprising an input layer, a variational coding layer, an attention mechanism layer, a feature fusion layer, and an output layer. The input layer receives three key dynamic parameters: frequency change rate, mode shape offset, and damping ratio change. The variational coding layer maps the input parameters to the latent space for feature abstraction through variational inference. The attention mechanism layer dynamically adjusts the attention weight allocation based on the real-time values ​​of settlement, bolt preload, and friction coefficient. The temperature parameter in the attention weight calculation formula is adjusted in real-time according to the ratio of the current settlement to the design settlement. When the ratio ∈ [0, 0.3), the temperature parameter is set to 0.8 to enhance model sensitivity; when the ratio ∈ [0.3, 0.7], the temperature parameter is set to 1.0 to maintain the standard response; and when the ratio ∈ (0.7, +∞), the temperature parameter is set to 1.5 to reduce model sensitivity and avoid false alarms. The feature fusion layer nonlinearly fuses the encoded features with the attention-weighted features. The output layer generates damage assessment values ​​and remaining bearing capacity prediction values. The steps for establishing the training dataset for the variational attention optimization model include collecting elevator installation engineering cases of existing buildings under different geological conditions, recording the geological parameters, design parameters, and monitoring data during operation of each project, establishing a comprehensive database containing soil physical and mechanical parameters, structural geometric parameters, load characteristic parameters, and damage evolution data, generating dynamic response data of multi-level buffer flexible connection nodes under different damage levels through finite element simulation, and forming a standardized training sample set covering four levels: normal state, minor damage, moderate damage, and severe damage, combined with actual engineering monitoring data. Each sample contains an input dynamic feature vector and a corresponding damage level label. The dataset size reaches 10,000 samples to ensure the model's generalization ability and prediction accuracy. The training steps of the variational attention optimization model include: first, training the variational encoding layer to learn the latent representation of the dynamic parameters using a hierarchical training strategy; then, training the attention mechanism layer to learn the weight allocation rules while fixing the encoding layer parameters; next, jointly training the encoding layer and the attention layer to optimize the feature fusion effect; and finally, training the entire network end-to-end to optimize the damage assessment accuracy. The training process adopts an adaptive learning rate scheduling strategy, with the initial learning rate set to 0.001. Every 1000 iterations, the learning rate is decayed to 0.9 times the original rate. The root mean square error is used as the loss function, and the network parameters are updated through the backpropagation algorithm. An early stopping mechanism is used during training to prevent overfitting. Training is stopped when the validation set loss does not decrease for 50 consecutive iterations. Finally, the model achieves a damage assessment accuracy of over 95% on the test set.

[0038] The adaptive slip adjustment function is calculated based on three key parameters: settlement difference, shear force change rate, and rotation angle change. First, it calculates the settlement difference between adjacent monitoring points as an indicator of uneven settlement. Then, it combines the time change rate of the measured shear force of the multi-level buffer flexible connection node to reflect the dynamic characteristics of the load. Simultaneously, it considers the rotation angle change of the multi-level buffer flexible connection node to characterize the degree of deformation coordination. The slip adjustment value is obtained by weighted fusion of these three parameters, with weighting coefficients of 0.5, 0.3, and 0.2, respectively. When the slip adjustment value calculation results indicate that the multi-level buffer flexible connection node requires significant deformation adjustment, the adaptive slip adjustment function automatically triggers a preload adjustment mechanism, increasing the bolt preload to improve the anti-slip capability of the multi-level buffer flexible connection node. When the slip adjustment value exceeds the design threshold, an emergency slip mode is activated, allowing the multi-level buffer flexible connection node to undergo significant slip deformation under controlled conditions to avoid structural damage.

[0039] The specific implementation methods of the above steps are described in detail below.

[0040] The specific implementation of step S01 involves establishing a distributed monitoring network to achieve comprehensive data collection of building settlement. First, laser displacement sensors are deployed at key locations connecting the existing building and the added elevator. These sensors utilize the triangulation principle, emitting a laser beam and receiving the reflected light signal. The round-trip time difference of the light signal is calculated to obtain precise vertical displacement data, achieving a measurement accuracy of 0.1 mm. The sampling frequency is set to 10 Hz to ensure data timeliness. Simultaneously, tiltmeters are configured at each monitoring point, using gravity sensing to detect minute angular changes in the building structure. The tilt measurement accuracy reaches 0.01°, providing angular deformation parameters for settlement analysis. A three-dimensional coordinate system is established, linking the spatial location information of each monitoring point with the settlement data to form a spatial settlement distribution pattern. The settlement rate is obtained by calculating the first derivative of settlement with respect to time using a numerical differential algorithm. A moving window smoothing algorithm is used to process the raw data to eliminate random noise interference. Combined with soil compression modulus data obtained from geological surveys, a multi-dimensional database containing time series, spatial coordinates, and physical parameters is established to provide fundamental data support for subsequent analysis.

[0041] The specific implementation of step S02 is to construct an intelligent prediction system based on multi-source data fusion technology. The real-time monitoring data obtained in step S01 is spatiotemporally matched with geological exploration data to establish a unified data standardization system. A time series prediction algorithm from machine learning, particularly the Long Short-Term Memory (LSTM) network algorithm, is used to learn the temporal evolution of settlement data. This algorithm can effectively handle long-term dependencies in sequence data and is suitable for capturing the nonlinear characteristics of the settlement process. Input parameters include current settlement amount, settlement rate, soil shear modulus, elastic modulus, and consolidation coefficient. The output is a settlement prediction matrix for the next 72 hours. The prediction model adopts a sliding window training strategy with a window length of 168 hours, updating model parameters every 6 hours to adapt to dynamic changes in settlement behavior. Prediction accuracy is evaluated using the root mean square error (RMSE), and a model retraining mechanism is automatically triggered when the prediction error exceeds 0.05 mm. A confidence interval evaluation system is established, providing a 95% confidence interval for each predicted value to support the reliability judgment of subsequent decisions.

[0042] The specific implementation of step S03 involves designing a connecting component with adaptive deformation capabilities using the principles of materials mechanics. The multi-level buffer flexible connection node adopts a layered design concept, sequentially arranging a high-stiffness steel layer, a medium-stiffness rubber layer, and a low-stiffness buffer material layer from the inside out. The elastoplastic material layers utilize combinations of materials with different compression indices: the inner layer uses high-strength polyurethane with a compression index of 0.15, the middle layer uses natural rubber with a compression index of 0.35, and the outer layer uses foam buffer material with a compression index of 0.55, forming a gradient stiffness distribution to achieve gradual load transfer and dispersion. The sliding adjustment mechanism is designed based on the principles of tribomechanics, controlling the friction coefficient between contact surfaces through high-strength bolts with adjustable preload. The friction coefficient can be adjusted within the range of 0.2 to 0.8. The bolt preload is applied using torque control, and a force sensor monitors the preload value in real time to ensure optimal matching between connection stiffness and deformation coordination. The entire node adopts a modular design, facilitating adjustments to material combinations and geometric parameters according to different engineering conditions.

[0043] The specific implementation of step S04 involves establishing an intelligent control system combining continuous monitoring and graded response. Force sensors composed of strain gauges are installed at key stress-bearing components of the multi-level buffer flexible connection nodes. Shear force and bending moment values ​​are calculated by measuring strain values ​​and combining them with the elastic modulus. Displacement sensors, employing resistive or capacitive principles, measure the linear displacement of the node in all directions, with a measurement range of ±50mm and an accuracy of 0.01mm. A real-time data processing system is established, using digital signal processing algorithms to filter and extract features from the acquired signals. A dual-threshold judgment mechanism is set up: when the settlement rate change exceeds 0.5% / second and lasts for more than 30 seconds, the system determines it to be in a dangerous state and immediately executes rapid response adjustment, with the adjustment time controlled within 5 seconds and the adjustment range being 20% ​​to 50% of the current parameter value. When the settlement rate change is between 0.2% and 0.5% / second but lasts for less than 30 seconds, slow response adjustment is executed, with an adjustment time of 30 to 120 seconds and an adjustment range of 5% to 15% of the current parameter value. The response adjustment achieves precise control of the bolt preload through an automated actuator.

[0044] The specific implementation of step S05 involves constructing a deep learning-based intelligent assessment system for node status. Vibration sensors are used to collect dynamic response signals from multi-level buffered flexible connection nodes. Frequency domain features are extracted using a Fast Fourier Transform algorithm to obtain natural frequencies, mode shape parameters, and modal damping ratios. A five-layer neural network architecture is established. The input layer receives three key dynamic parameters: frequency change rate, mode shape offset, and damping ratio change. The variational coding layer uses variational inference theory to map the input parameters to a latent feature space, achieving data dimensionality reduction and feature abstraction. The attention mechanism layer dynamically adjusts the weight allocation based on the ratio of the current settlement to the design settlement. When the ratio is less than 0.3, the temperature parameter is set to 0.8 to enhance sensitivity; when the ratio is between 0.3 and 0.7, it is set to 1.0 to maintain a standard response; and when the ratio is greater than 0.7, it is set to 1.5 to reduce sensitivity. The feature fusion layer uses a nonlinear activation function to effectively combine encoded features and attention features. The output layer generates damage assessment values ​​and remaining bearing capacity prediction values, providing a quantitative basis for maintenance decisions. The model was trained using a standardized dataset containing 10,000 samples, covering four levels: normal, minor, moderate, and severe damage, with a prediction accuracy of over 95%.

[0045] The specific implementation of step S06 involves establishing an adaptive adjustment algorithm based on multi-parameter fusion. The slip adjustment function uses the settlement difference, shear force change rate, and rotation angle change as input parameters, and calculates the slip adjustment value using a weighted fusion strategy. The settlement difference reflects the degree of uneven settlement between adjacent monitoring points, and a continuous settlement gradient distribution is obtained through spatial interpolation. The shear force change rate is obtained by numerical differentiation of the time-series data, characterizing the dynamic changes in the load. The rotation angle change is measured by an inclinometer, reflecting the deformation coordination state of the node. The weighting coefficients for the three parameters are set to 0.5, 0.3, and 0.2, respectively, and the weight allocation is determined based on the sensitivity analysis of the parameters' impact on node performance. When the slip adjustment value is in the range of 0 to 5 mm, the existing bolt preload is maintained unchanged. When the slip adjustment value is in the range of 5 to 15 mm, the bolt preload is increased by 15% to improve the anti-slip capability. When the slip adjustment value exceeds 15mm, the emergency slip mode is adopted, which allows the node to produce large slip deformation under controlled conditions. Stress concentration is released through deformation to avoid sudden structural failure.

[0046] The key technical ideas of this invention are mainly reflected in the following three aspects. First, there is the multi-source data fusion prediction technology based on machine learning. This technology combines real-time monitoring data with geological parameters and uses a long short-term memory network algorithm to learn the complex nonlinear laws of settlement evolution. Compared with traditional linear prediction methods, it can more accurately capture the time dependence and spatial correlation in the settlement process, significantly improving the accuracy and reliability of settlement prediction and providing a scientific basis for active protection. Second, there is the adaptive design technology of multi-level buffer flexible connection nodes. Through a combination of gradient stiffness materials and a sliding mechanism with an adjustable friction coefficient, it achieves the step-by-step transfer of loads and stress dispersion. Compared with traditional rigid connections, it can effectively absorb the deformation energy caused by settlement, avoid structural damage caused by stress concentration, and maintain the overall stability and load-bearing capacity of the connection. Third, there is the intelligent damage assessment technology based on a variational attention optimization model. This technology combines frequency domain analysis and grey relational theory to intelligently identify damage characteristics from dynamic response signals. Compared with traditional static detection methods, it has higher sensitivity and accuracy, enabling early warning and quantitative assessment of damage.

[0047] The synergistic effect of these three key technological approaches forms a complete intelligent protection system, achieving closed-loop control from prediction and protection to assessment, and possessing significant technological advantages compared to existing technologies. Predictive technology provides forward-looking information to the protection system, enabling proactive rather than reactive protective measures, greatly improving the system's response efficiency. Adaptive connection technology implements precise adjustments based on prediction results, effectively mitigating the adverse effects of uneven settlement through the coordinated deformation of materials and structures. This supports the continuous optimization of the prediction model and the dynamic adjustment of protection strategies. These three elements mutually reinforce each other, forming a positive cycle that endows the entire system with self-learning and self-optimization capabilities, enabling it to adapt to complex changes in different geological conditions and engineering environments, achieving a fundamental transformation from traditional passive protection to intelligent active protection.

[0048] Furthermore, this invention also solves the technical problems of insufficient accuracy and lack of predictive capability in settlement monitoring of connection nodes. Traditional monitoring methods mainly rely on manual measurement and simple instruments, resulting in low monitoring frequency and limited accuracy. They cannot capture the dynamic changes in settlement and lack the ability to predict future settlement trends. This invention achieves real-time settlement monitoring with an accuracy of 0.1 mm by establishing a multi-point monitoring device including a laser displacement sensor and a tiltmeter. By combining shear modulus, elastic modulus, and consolidation coefficient from geological survey data and using machine learning algorithms to establish an adaptive response model, it can predict the settlement development trend within the next 72 hours and output a settlement prediction matrix. This provides scientific data support for the active adjustment of connection nodes and significantly improves the accuracy of settlement monitoring and the reliability of prediction.

[0049] Furthermore, this invention addresses the technical problems of outdated methods for identifying damage to connection nodes and the high subjectivity of assessment results. Existing technologies primarily rely on visual inspection and experience-based judgment to assess the damage state of connection nodes, lacking objective quantitative analysis methods and making it difficult to accurately identify early damage and assess its severity. This invention establishes a quantitative relationship between damage severity and dynamic characteristic parameters by constructing a variational attention optimization model based on grey relational analysis and neural networks. Utilizing frequency change rate, mode shape offset, and damping ratio change as key damage indicators, and training on a standardized dataset containing 10,000 samples, it achieves precise quantitative assessment of the damage severity of connection nodes. The model achieves an assessment accuracy of over 95% on the test set, providing an objective and reliable technical basis for scientific maintenance decisions regarding connection nodes and fundamentally changing the subjectivity and uncertainty inherent in traditional damage assessment.

[0050] Specifically, the principle of this invention is as follows: The technical principle behind solving the problem of rigid failure of connection nodes lies in achieving the active adaptation capability of connection nodes to non-uniform settlement through a combination of flexible design and intelligent control. The design principle of the multi-level buffer flexible connection node is based on the stress dispersion and energy dissipation theory in materials mechanics. Through the graded response mechanism of the elastoplastic material layer, the concentrated stress generated by non-uniform settlement is transformed into distributed stress. The elastic deformation of the material absorbs small-amplitude settlement, and the plastic deformation dissipates moderate-level settlement. When the settlement exceeds the material deformation limit, the sliding adjustment mechanism releases excessive deformation constraints through frictional sliding, preventing brittle failure of the connection node. The working principle of the intelligent monitoring and adjustment system is based on real-time data acquisition and predictive control theory. A three-dimensional settlement distribution database is established through a multi-point settlement monitoring device. Machine learning algorithms are used to analyze the settlement development law and predict future trends, providing forward-looking guidance for the active adjustment of the connection node. The variational attention optimization model establishes a nonlinear mapping relationship between the degree of damage and dynamic characteristic parameters through deep learning technology. It can identify early damage signals of the connection node from minute changes in frequency, mode shape, and damping parameters, achieving quantitative assessment of the damage state. The slip adaptive adjustment function is based on a multi-parameter fusion control strategy. By calculating the settlement difference, shear force change rate and rotation angle change in real time, it automatically adjusts the bolt preload and slip parameters to ensure that the connection node achieves a dynamic balance between load-bearing capacity and deformation coordination. This intelligent adjustment mechanism enables the connection node to automatically optimize its working state according to the actual working conditions, thereby effectively avoiding rigid failure.

[0051] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0052] The specific implementation of step S01 is to achieve comprehensive collection of building settlement data by establishing a distributed monitoring network. The settlement rate calculation formula is as follows:

[0053] ;

[0054] In the formula, Let be the settlement rate at time t, in mm / s; The cumulative settlement at time t is expressed in mm. This is a time variable, with the unit being seconds (s).

[0055] The parameter acquisition method is as follows: The displacement data is obtained using a laser displacement sensor, comprising the following steps: Step 1: A laser beam with a wavelength of 650nm is emitted and perpendicularly illuminates the surface of the measurement point; Step 2: A receiver detects the reflected light signal and calculates the optical path difference to obtain displacement data, with a measurement accuracy of 0.1mm; Step 3: A time-series database is established to record continuous measurement values. Settlement rate The sampling interval was set to 0.1s and was obtained through a numerical differentiation algorithm.

[0056] The spatial interpolation function expression for establishing the three-dimensional settlement distribution database is:

[0057] ;

[0058] In the formula, Spatial coordinates The settlement at time t, in mm; , These are the horizontal coordinates, in meters (m). Let be the spatial weight function for the i-th monitoring point, which is dimensionless; The measured settlement at time t for the i-th monitoring point is expressed in mm. This represents the total number of monitoring points. The weighting function is calculated using the inverse distance weighting method: ,in Let be the Euclidean distance from the target point to the i-th monitoring point. The Euclidean distance from the target point to the j-th monitoring point is calculated using the following formulas:

[0059] ;

[0060] ;

[0061] In the formula, Let i be the coordinates of the i-th monitoring point. Let be the coordinates of the j-th monitoring point, in meters.

[0062] The specific implementation of step S02 is to construct an intelligent prediction system based on multi-source data fusion technology. The expression for the settlement prediction matrix is:

[0063] ;

[0064] In the formula, This is the settlement prediction matrix for the 72-hour period, with dimensions [missing information]. ,in Number of monitoring points; For prediction functions of Long Short-Term Memory (LSTM) networks; The input feature matrix contains the current settlement, settlement rate, and soil shear modulus. Elastic modulus and consolidation coefficient ; This is the network parameter matrix.

[0065] The parameter acquisition method is as follows: shear modulus The results were obtained using indoor triaxial shear tests, including the following steps: Step 1: preparing standard soil samples and mounting them in a triaxial apparatus; Step 2: applying confining pressure and conducting the shear test; Step 3: calculating the stress-strain relationship. The value, typically ranging from 10 to 100 MPa. Elastic modulus. The consolidation coefficient was determined using a uniaxial compression test. The typical range was determined by indoor consolidation tests. ~ .

[0066] The specific implementation method of step S03 is the same as described above, and will not be repeated in detail here.

[0067] The specific implementation of step S04 is to establish an intelligent control system that combines continuous monitoring with graded response. The formula for calculating the rate of change of settling rate is:

[0068] ;

[0069] In the formula, denoted as the rate of change of settlement velocity at time t, in mm / s².

[0070] The expression for the response discriminant function is:

[0071] ;

[0072] In the formula, The response order function is dimensionless. Indicates rapid response adjustment. This indicates a slow response adjustment. This indicates a desire to maintain the status quo; The duration is in seconds; the thresholds of 0.005 and 0.002 are in units of seconds. , with the rate of change of sedimentation rate The units should be kept consistent.

[0073] The specific implementation of step S05 involves constructing a node state intelligent evaluation system based on deep learning. The temperature parameter adjustment function for attention weights is:

[0074] ;

[0075] In the formula, For temperature parameters; The ratio of the current settlement to the design settlement is expressed as follows: ,in This represents the current measured settlement. To determine the allowable settlement amount.

[0076] The specific implementation of step S06 is to establish an adaptive adjustment algorithm based on multi-parameter fusion. The specific expression of the slip adaptive adjustment function is:

[0077] ;

[0078] In the formula, This is the sliding adjustment value, in mm; This represents the difference in settlement between adjacent monitoring points, in mm. This represents the rate of change of shear force, expressed in N / s. This represents the change in angle, expressed in rad.

[0079] The parameter acquisition method is as follows: The settlement difference between adjacent monitoring points is obtained through spatial interpolation. The calculation formula is as follows: ,in and , respectively, represent the settlement at time t of the i-th and (i+1)-th adjacent monitoring points, in mm; shear force change rate. The shear force time series data measured by strain gauges is obtained by numerical differentiation, where The shear force borne by the multi-stage buffer flexible connection node is expressed in N, and the rate of change of shear force is expressed in N / s; the change in rotation angle is also expressed. The measurement was obtained directly by an inclinometer with an accuracy of 0.01°, and the unit is rad.

[0080] The expression for the bolt preload adjustment function is as follows:

[0081] ;

[0082] In the formula, Preload force for the bolts at the next moment; This represents the bolt preload at the current moment. The preload for emergency slip mode is typically set to 50% of the initial preload. Optionally, the time interval can be 30 minutes to 72 hours.

[0083] The specific implementation of step S07 involves constructing a system stability guarantee mechanism based on matrix theory. The matrix rank deficiency detection employs a singular value decomposition algorithm, and the settlement prediction matrix... The singular value decomposition expression is:

[0084] ;

[0085] In the formula, It is a left singular vector matrix; It is a diagonal singular value matrix; It is the transpose of the right singular vector matrix.

[0086] The function for determining the rank of a matrix is:

[0087] ;

[0088] In the formula, For matrix rank; Let i be the i-th singular value; To determine the threshold, the value is set to... ; This is an indicator function; it returns 1 if the condition is met, and 0 otherwise. and These represent the number of rows and columns of the matrix, respectively.

[0089] The matrix health assessment indicators include the condition number and the minimum singular value, expressed as follows:

[0090] ;

[0091] ;

[0092] In the formula, It is a condition number; It is the maximum singular value; It is the smallest singular value; Let be the rank of the matrix.

[0093] The principles and effects of each formula are explained below. The settlement rate calculation formula is based on differential theory. It obtains velocity information by performing time-domain differentiation on the displacement signal. Compared with the traditional difference method, it can provide a continuous rate change trend, providing accurate input parameters for subsequent dynamic response, and significantly improving the system's sensitivity to settlement changes and response speed.

[0094] ;

[0095] The three-dimensional settlement distribution function uses the principle of spatial interpolation to extend discrete monitoring point data into a continuous spatial distribution. Compared with single-point monitoring, it can comprehensively reflect the settlement distribution characteristics of the entire connected area, providing a spatial basis for accurately assessing the degree of uneven settlement and effectively avoiding judgment errors caused by local monitoring blind spots.

[0096] ;

[0097] The weight function Based on the inverse distance weighting principle, the closer the monitoring point is, the greater its influence on the target location, thus ensuring the rationality and accuracy of spatial interpolation.

[0098] The slip adaptive adjustment function is based on the principle of multi-parameter weighted fusion, which comprehensively considers three key factors: settlement difference, dynamic load change and angular deformation. Through reasonable weight allocation, it realizes a comprehensive evaluation of the node state. Compared with the single parameter judgment method, it can more accurately reflect the true stress and deformation state of the node, and significantly improve the scientificity and reliability of the adjustment decision.

[0099] ;

[0100] The bolt preload adjustment function adopts a segmented control strategy, which implements differentiated adjustment according to different ranges of the slip adjustment value. This ensures structural stability under normal working conditions and provides emergency protection under abnormal conditions. Compared with the fixed parameter control method, it has stronger adaptability and safety.

[0101] ;

[0102] The temperature parameter adjustment function adaptively adjusts the sensitivity of the attention mechanism to achieve differentiated damage detection strategies under different settlement levels. Compared with the fixed threshold method, it can effectively reduce false alarms and false negatives, and improve the accuracy and practicality of damage identification.

[0103] ;

[0104] The ratio function By comparing the current settlement with the design allowable value, a quantitative assessment of the system's operating status was achieved, providing a reliable numerical basis for intelligent decision-making.

[0105] In this embodiment, the bolts can be adjusted manually or by automated equipment.

[0106] To better understand and implement this invention, a specific application scenario is provided below as Example 2: A building is located in a soft soil foundation area. Preliminary geological surveys revealed that the foundation soil layer mainly consists of silty clay and clayey silt, with a high groundwater level, posing a significant risk of differential settlement. The technical team decided to use the elevator installation flexible connection node anti-non-uniform settlement method of this invention to solve this technical problem.

[0107] First, the technical team deployed laser displacement sensors at key locations connecting the existing building and the added elevator. Four laser displacement sensors were installed at the interface between the existing building foundation and the elevator foundation, with a measurement accuracy of 0.1 mm; six laser displacement sensors were installed at the interface between the existing building wall and the elevator shaft; eight laser displacement sensors were installed at the interface between the existing building floor slab and the elevator platform; and two laser displacement sensors were installed at the connection points between the existing building structural columns and the elevator support structure. Simultaneously, strain gauge-based force sensors were installed at key stress-bearing locations in the multi-level buffer flexible connection nodes. Twelve strain gauges were installed at the connection interface bearing the maximum shear force; eight strain gauges were installed at the rotational hinge points bearing the maximum bending moment; sixteen strain gauges were installed at the main load-bearing section bearing the maximum axial force; and six strain gauges were installed at the bolted connection assembly surface bearing the maximum torque. The technical team also deployed inclinometers at the four corners of the multi-level buffer flexible connection nodes to measure angle changes.

[0108] Next, the technical team obtained detailed soil layer parameters through geological exploration. The silty clay layer has a compression modulus of 3.2 MPa, a shear modulus of 1.8 MPa, an elastic modulus of 4.5 MPa, and a consolidation coefficient of [missing value]. / s. The silty clay layer has a compression modulus of 8.7 MPa, a shear modulus of 5.2 MPa, an elastic modulus of 12.6 MPa, and a consolidation coefficient of [missing value]. Based on this multi-source monitoring data, the technical team established an adaptive response model, integrating geological survey data and using machine learning algorithms to predict the settlement trend over the next 72 hours. Monitoring data showed that the initial settlement of the existing building foundation was 2.3 mm, with a settlement rate of 0.08 mm / d, while the initial settlement of the elevator installation foundation was 1.7 mm, with a settlement rate of 0.12 mm / d, indicating significant differential settlement characteristics.

[0109] The technical team designed a multi-level buffer flexible connection node using a three-layer elasto-plastic material structure. The first layer, 25mm thick, is a polyurethane elastomer with a compression index of 0.15, primarily responsible for buffering small deformations. The second layer, 40mm thick, is a rubber composite material with a compression index of 0.28, absorbing moderate deformations. The third layer, 60mm thick, is a metal spring composite material with a compression index of 0.42, adjusting for large deformations. The sliding adjustment mechanism uses high-strength bolts (M24) with an initial preload of 180kN, applied via a torque wrench and monitored in real-time by a force sensor. The coefficient of friction, measured in field friction tests, is 0.35, falling within the design range of 0.2 to 0.8.

[0110] During operation, the multi-point settlement monitoring device continuously collected data. At 10:32 AM one morning, the monitoring system detected a settlement rate change of 0.6% / second for 45 seconds, exceeding the threshold condition for rapid response adjustment. The system immediately executed the rapid response adjustment mechanism, adjusting the bolt preload within 3.8 seconds, increasing the preload from 180kN to 234kN, an adjustment range of 30%. After adjustment, the settlement rate change rapidly decreased to below 0.15% / second, the shear force of the multi-stage buffer flexible connection node decreased from 127kN to 89kN, the bending moment decreased from 45kN·m to 31kN·m, and the displacement change was controlled within the allowable range.

[0111] like Figure 2 As shown, the technical team applied a variational attention optimization model to perform real-time status assessment of the multi-level buffer flexible connection node. The variational attention optimization model, using vibration signals collected by vibration sensors and obtained through Fast Fourier Transform, yielded a natural frequency of 28.7Hz, a 5.0% decrease from the initial design frequency of 30.2Hz. Modal analysis parameters showed a mode shape shift of 3.2mm for the first mode and 1.8mm for the second mode. Damping parameters measured using the free decay method showed a damping ratio of 0.047, an 11.9% increase from the initial value of 0.042.

[0112] The technical team established a node damage assessment system based on grey relational analysis. By calculating the correlation strength between various dynamic parameters and damage level, the correlation degree of frequency change rate was 0.85, that of mode shape offset was 0.78, and that of damping ratio change was 0.72, all exceeding the screening threshold of 0.7, and were used as input features for the variational attention optimization model. The five-layer neural network architecture of the variational attention optimization model includes 32 input neurons, 64 variational coding neurons, 128 attention mechanism neurons, 96 feature fusion neurons, and 16 output neurons. The model evaluation results show that the current node damage level is minor, with a damage level assessment value of 0.23, and the predicted remaining bearing capacity is 87.4% of the original design bearing capacity.

[0113] Table 1 shows the key parameter data collected during the monitoring period:

[0114] Table 1. Monitoring Data of Key Parameters for Multi-Level Buffer Flexible Connection Nodes

[0115]

[0116] Based on the data in Table 1, the technical team applied the adaptive slip adjustment function for calculation. The adaptive slip adjustment function is based on three parameters: settlement difference, shear force change rate, and rotation angle change, with weighting coefficients of 0.5, 0.3, and 0.2, respectively. At 11:30, the slip adjustment value was 12.4 mm, falling within the range of [5, 15) mm. The system automatically increased the bolt preload by 15%, adjusting it from 234 kN to 269 kN. Monitoring data after the adjustment showed that the slip adjustment value decreased to 8.3 mm at 12:00, indicating that the adjustment strategy was effective.

[0117] like Figure 3 As shown, the attention weight allocation mechanism of the variational attention optimization model is dynamically adjusted based on real-time monitoring data. When the ratio of settlement to design settlement is 0.25, the temperature parameter is set to 0.8 to enhance model sensitivity, and the attention weights are tilted towards the rate of change of frequency, with weight allocations of 0.52 for the rate of change of frequency, 0.28 for the mode shape shift, and 0.20 for the change in damping ratio. When the ratio increases to 0.45, the temperature parameter is adjusted to 1.0 to maintain the standard response, and the attention weights tend to be evenly distributed, with weight allocations of 0.38, 0.34, and 0.28 respectively. When the ratio further increases to 0.82, the temperature parameter is set to 1.5 to reduce model sensitivity and avoid false alarms, and the attention weights are tilted towards the change in damping ratio, with weight allocations adjusted to 0.25, 0.31, and 0.44.

[0118] On the third day of continuous monitoring, the system detected an abnormal settlement event. At 14:25, the settlement rate suddenly reached 1.2% / second, far exceeding the rapid response threshold of 0.5% / second, and lasted for 78 seconds. The system immediately triggered the emergency response mechanism, increasing the bolt preload from 269kN to 404kN within 4.2 seconds, an adjustment range of 50%. Simultaneously, the calculated slip adjustment value was 18.7mm, exceeding the 15mm threshold. The system activated the emergency slip mode, allowing the multi-stage buffer flexible connection node to undergo moderate slip deformation under controlled conditions. In the emergency slip mode, the friction coefficient was dynamically adjusted from 0.35 to 0.28, and the damping coefficient of the slip adjustment mechanism increased to 1.8 times its original value, ensuring the stability and controllability of the slip process.

[0119] Table 2 shows the changes in system parameters during the emergency response:

[0120] Table 2 System Parameter Changes During Emergency Response

[0121]

[0122] The data in Table 2 show that the emergency response mechanism effectively controlled the stress state and deformation of the nodes. The maximum shear force and maximum bending moment after the response were both lower than the values ​​before the response, and the node displacement was also within an acceptable range, proving the effectiveness of the emergency slip mode.

[0123] The variational attention optimization model continuously evaluates the state throughout the monitoring period. The model training dataset contains 10,000 samples, covering four levels: normal state, minor impairment, moderate impairment, and severe impairment. A hierarchical training strategy is employed, with an initial learning rate of 0.001, which decays to [a higher learning rate] after 8500 iterations. The final model achieved a damage assessment accuracy of 96.3% on the test set, exceeding the design target of 95%. In this engineering application, the real-time damage assessment value output by the model fluctuated between 0.15 and 0.31, consistently remaining within the range of minor damage, while the predicted remaining bearing capacity remained stable between 85% and 92% of the original design bearing capacity.

[0124] The slip adjustment function exhibited good adaptive performance throughout the monitoring period. During the function calculation, the settlement difference contributed the most, with a weighting coefficient of 0.5, reflecting the dominant influence of differential settlement on the connection node. The shear force change rate had a weighting coefficient of 0.3, demonstrating the significant impact of dynamic loads on the node state. The rotation angle change had a weighting coefficient of 0.2, characterizing the adjustment requirements for deformation coordination. Over the 72 hours of monitoring, the slip adjustment values ​​were distributed as follows: 67.2% of the time, the slip adjustment value ∈ [0, 5) mm, maintaining the existing bolt preload; 28.5% of the time, the slip adjustment value ∈ [5, 15) mm, increasing the bolt preload by 15%; and 4.3% of the time, the slip adjustment value ∈ [15, +∞) mm, employing an emergency slip mode.

[0125] A three-dimensional settlement distribution database recorded the settlement distribution characteristics of the connecting area. The cumulative settlement at the southwest corner of the existing building was 8.7 mm, the southeast corner 6.2 mm, the northwest corner 7.1 mm, and the northeast corner 5.8 mm, showing an uneven distribution pattern with higher settlement in the southwest and lower settlement in the northeast. The settlement at the four corners of the elevator installation foundation was 4.3 mm, 3.9 mm, 4.7 mm, and 4.1 mm, respectively, which was relatively uniform. The largest difference in settlement occurred between the southwest corner of the existing building and the southeast corner of the elevator installation, with a difference of 4.8 mm, making this a key monitoring area.

[0126] Based on historical monitoring data and geological parameters, a machine learning algorithm successfully predicted the settlement trend over the next 72 hours. The prediction results show that the settlement rate of existing buildings will gradually increase from 0.08 mm / d to 0.11 mm / d within the next 24 hours, stabilize at around 0.10 mm / d within 48 hours, and is expected to decrease to 0.07 mm / d after 72 hours. The predicted settlement rate for elevator installation foundations is 0.12 mm / d to 0.09 mm / d within the first 24 hours, and then remain relatively stable. The settlement prediction matrix outputs a 20×20 spatial grid, with each grid point containing three predicted values: settlement amount, settlement rate, and settlement acceleration.

[0127] The multi-level buffer flexible connection node exhibits excellent adaptive adjustment capabilities in actual operation. The three-layer structure of the elasto-plastic material layer effectively absorbs the deformation energy generated by differential settlement. The compression deformation of the first layer of polyurethane elastomer material is 3.8 mm, the compression deformation of the second layer of rubber composite material is 7.2 mm, and the compression deformation of the third layer of metal spring composite material is 11.5 mm, with a total deformation of 22.5 mm, which is within the design allowable range. The sliding adjustment mechanism achieves precise control of the friction coefficient within the range of 0.28 to 0.42 by dynamically adjusting the bolt preload, ensuring the coordination and unity of load-bearing capacity and deformation.

[0128] The real-time evaluation function of the variational attention optimization model provides reliable assurance for engineering safety. The model extracts dynamic characteristic parameters using frequency domain analysis and filters key features through grey relational analysis, achieving accurate quantification of node damage levels. Throughout the monitoring period, the model identified three minor abnormal states and one emergency response event, all of which were handled promptly and effectively. The model's damage assessment system establishes a comprehensive early warning mechanism: a yellow warning is issued when the damage assessment value exceeds 0.4, an orange warning when it exceeds 0.6, and a red warning when it exceeds 0.8, triggering the emergency response procedure.

[0129] This invention represents a significant technological advancement over traditional methods. Traditional rigid connections are prone to stress concentration and connection failure when facing differential settlement, while this invention, through the design of multi-level buffered flexible connection nodes, achieves graded absorption of deformation and uniform stress distribution. Traditional passive monitoring methods can only detect damage after it occurs, while the adaptive response model established in this invention can predict settlement trends 72 hours in advance, realizing a shift from passive response to proactive prevention. Traditional fixed-parameter designs cannot adapt to changes in soil properties and dynamic load characteristics, while this invention, through an adaptive slip adjustment function and variational attention optimization model, achieves real-time optimization and intelligent adjustment of system parameters. Traditional single-index evaluation methods are prone to misjudgment, while the multi-parameter fusion evaluation system established in this invention identifies key damage indicators through grey relational analysis, improving the accuracy and reliability of damage identification. Traditional manual adjustment methods have slow response speeds and limited accuracy, while this invention achieves millisecond-level automatic response and micron-level precision control, significantly improving the system's response efficiency and adjustment accuracy.

[0130] It should be noted that the variables involved in this invention are explained in detail in Table 3 below.

[0131] Table 3. Variable Explanation Table

[0132]

[0133] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for adding flexible connection nodes to elevators to resist non-uniform settlement, characterized in that, Multiple settlement monitoring devices are deployed at the connection points between existing buildings and newly installed elevators to collect real-time data on foundation settlement, settlement rate, and soil compression modulus, establishing a three-dimensional settlement distribution database. An adaptive response model is built based on multi-source monitoring data, integrating shear modulus, elastic modulus, and consolidation coefficient from geological survey data. Machine learning algorithms are used to predict settlement trends, outputting a settlement prediction matrix. Multi-level buffer flexible connection nodes are designed, each comprising an elasto-plastic material layer and a sliding adjustment mechanism. The elasto-plastic material layer uses a combination of materials with different compression indices, and the sliding adjustment mechanism controls the friction coefficient through bolt preload to achieve coordination between bearing capacity and deformation. The shear force, bending moment, and displacement changes of the multi-level buffer flexible connection nodes are monitored in real-time, and rapid or slow response adjustments are implemented based on the settlement rate change and duration. A variational attention optimization model is applied to perform real-time state assessment of multi-level buffer flexible connection nodes. Grey relational analysis is established through changes in frequency, mode shape, and damping parameters to construct a node damage assessment system. An adaptive adjustment function for slip is established, and the slip adjustment value is calculated based on the settlement difference, shear force change rate, and rotation angle change. The corresponding bolt preload adjustment strategy is executed according to the slip adjustment value range.

2. The method for adding flexible connection nodes to elevators to resist non-uniform settlement according to claim 1, characterized in that, The multi-point settlement monitoring device includes a laser displacement sensor and a tiltmeter. The laser displacement sensor is used to measure the settlement vertically downwards, and the tiltmeter is used to measure the angle change.

3. The method for resisting non-uniform settlement by adding flexible connection nodes to elevators according to claim 2, characterized in that, The laser displacement sensor has a measurement accuracy of 0.1 mm, and the settling rate is the first derivative of the settling amount with respect to time.

4. The method for resisting non-uniform settlement by adding flexible connection nodes to elevators according to claim 3, characterized in that, The compression modulus is determined by indoor compression tests or field load tests, reflecting the compression deformation characteristics of soil under pressure.

5. The method for resisting non-uniform settlement by adding flexible connection nodes to elevators according to claim 4, characterized in that, The shear modulus is obtained through an indoor triaxial shear test and characterizes the soil's ability to resist shear deformation. The elastic modulus is determined through a uniaxial compression test and represents the ratio of stress to strain within the elastic range of the material.

6. The method for resisting non-uniform settlement by adding flexible connection nodes to elevators according to claim 5, characterized in that, The consolidation coefficient is obtained through indoor consolidation tests and reflects the time-effect characteristics of soil consolidation deformation.

7. The method for adding flexible connection nodes to elevators to resist non-uniform settlement according to claim 6, characterized in that, The variational attention optimization model is an intelligent assessment system built on the grey relational neural network damage identification algorithm. By establishing a grey relational analysis model of node damage degree and dynamic characteristic parameters, it identifies frequency change rate, mode shape offset and damping ratio change as key damage indicators, and uses the variational attention optimization model to quantitatively assess the damage degree.

8. The method for resisting non-uniform settlement by adding flexible connection nodes to elevators according to claim 7, characterized in that, The coefficient of friction is obtained through on-site friction tests or material surface roughness tests, and the value ranges from 0.2 to 0.

8. The bolt preload is applied by a torque wrench and monitored in real time by a force sensor.

9. The method for adding flexible connection nodes to elevators to resist non-uniform settlement according to claim 8, characterized in that, The shear force and bending moment are measured by a force sensor composed of strain gauges, which is installed at the key stress positions of the multi-level buffer flexible connection node. The displacement is measured by a displacement sensor to measure the linear movement of the multi-level buffer flexible connection node in each direction.

10. The method for adding flexible connection nodes to elevators to resist non-uniform settlement according to claim 9, characterized in that, When the rate of change of the settling rate exceeds 0.5% / second and the duration is greater than 30 seconds, a fast response adjustment is immediately executed; when the rate of change of the settling rate exceeds 0.2% / second but the duration is less than 30 seconds, a slow response adjustment is executed.

Images