A Reliability Assessment Method for Vibration Reduction Effect of Vibration Isolation Trench Based on Big Data

Abstract

This disclosure relates to the field of big data technology, specifically to a method for reliable evaluation of the vibration reduction effect of vibration isolation trenches based on big data. The method includes: interpolating vibration data to obtain gridded vibration data of a preset scale; performing spatial attention and temporal attention calculations using a spatiotemporal feature model to obtain spatiotemporal feature vectors; obtaining a target vibration propagation map using a graph neural network model based on the spatiotemporal feature vectors, vibration source data, geological data, and an initial vibration propagation map; and evaluating the vibration reduction effect of the vibration isolation trench based on the vibration source data, structural data, environmental data, and the target vibration propagation map to obtain a vibration reduction effect evaluation result. This method can convert discrete vibration monitoring into continuous vibration monitoring, comprehensively reflecting the spatial attenuation law of vibration waves and improving the accuracy of the reliability evaluation of the vibration reduction effect of vibration isolation trenches.

Description

Technical Field

[0001] This disclosure relates to the field of big data technology, specifically to a reliability assessment method for the vibration reduction effect of vibration isolation trenches based on big data. Background Technology

[0002] Vibration isolation trenches are trenches excavated between the vibration source and the protected building or precision instrument. They utilize the wave resistance effect of the medium inside the trench (such as flexible materials like foam) to reflect and scatter vibration waves, thereby achieving vibration reduction.

[0003] In related technologies, the reliability assessment of the vibration reduction effect of vibration isolation trenches mainly utilizes data measured on-site. For example, ground vibration monitoring points are arranged at intervals on both sides of the vibration isolation trench to measure and compare the vibration data before the trench (i.e., the vibration source side) and after the trench (i.e., the protected side), and the amplitude attenuation ratio is calculated to quantify the vibration reduction effect. In this method, the sensors can only capture discrete point data (such as the two sides of the vibration isolation trench), and cannot fully reflect the spatial attenuation law of the vibration wave, thus affecting the accuracy of the reliability assessment of the vibration reduction effect of the vibration isolation trench. Summary of the Invention

[0004] To overcome the problems existing in the related technologies, this disclosure provides a reliability evaluation method for the vibration reduction effect of vibration isolation trenches based on big data, in order to solve the defects in the related technologies.

[0005] According to a first aspect of the present disclosure, a method for reliability evaluation of the vibration reduction effect of vibration isolation trenches based on big data is provided, comprising:

[0006] The vibration data of vibration sensors in the vibration source area, vibration isolation trench and vibration reduction protection area, the vibration source data of the vibration source area, the structural data of the vibration isolation trench, and the environmental data of the vibration source area, the vibration reduction trench and the vibration reduction protection area are acquired. The environmental data includes geological data and meteorological data.

[0007] Interpolation calculations are performed based on the vibration data to obtain gridded vibration data of a preset scale;

[0008] Spatial vibration features are obtained by performing spatial attention calculation based on the environmental data and gridded vibration data at the same time point using a spatiotemporal feature model. Temporal vibration features are obtained by performing temporal attention calculation based on the environmental data and gridded vibration data of the same vibration sensor at different time points. The spatial vibration features, the temporal vibration features, the vibration source data, and the environmental data are then fused into a spatiotemporal feature vector.

[0009] A target vibration propagation map is obtained by using a graph neural network model based on the spatiotemporal feature vector, the vibration source data, the geological data, and the initial vibration propagation map. The initial vibration propagation map is used to characterize the vibration propagation relationship in the vibration source area, the vibration isolation trench, and the vibration reduction protection area.

[0010] The vibration reduction effect of the vibration isolation trench is evaluated based on the vibration source data, the structural data, the environmental data, and the target vibration propagation diagram, and the vibration reduction effect evaluation result is obtained.

[0011] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects:

[0012] The big data-based reliability assessment method for vibration reduction effect of vibration isolation trenches provided in this disclosure first generates gridded vibration data of a preset scale through interpolation calculation, intuitively displaying the entire spatial propagation process of vibration waves. Then, it learns the spatiotemporal dependencies in the vibration field through a spatiotemporal feature model, capturing the propagation mode and abnormal paths of vibration waves to obtain spatiotemporal feature vectors. Next, it constructs a target vibration propagation map by combining the spatiotemporal feature vectors of vibration data, source data, and geological data through a graph neural network, quantitatively assessing the effectiveness of the vibration isolation trench throughout the propagation path. Finally, it evaluates the vibration reduction effect of the vibration isolation trench by leveraging the semantic capabilities of a multimodal large model, combining source data, structural data, environmental data, and the target vibration propagation map, to obtain the vibration reduction effect assessment result. Thus, discrete-point vibration monitoring can be transformed into continuous-field vibration monitoring, comprehensively reflecting the attenuation law of vibration waves in time and space, thereby improving the accuracy of the reliability assessment of the vibration reduction effect of vibration isolation trenches. Attached Figure Description

[0013] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0014] Figure 1 This is a flowchart illustrating an exemplary embodiment of the present disclosure of a method for reliability evaluation of vibration reduction effect of vibration isolation trenches based on big data;

[0015] Figure 2 This is a schematic diagram of a vibration isolation ditch in an exemplary embodiment of the present disclosure, illustrating a method for reliable evaluation of vibration reduction effect of vibration isolation ditch based on big data. Detailed Implementation

[0016] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0017] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0018] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information.

[0019] As mentioned in the background section, the reliability assessment methods for vibration reduction effect of vibration isolation trenches in related technologies rely on discrete monitoring points on both sides of the trench. These methods can only acquire localized, isolated vibration data and cannot capture the continuous propagation and attenuation patterns of vibration waves along the entire path from the vibration source to the trench and the protected area. Therefore, they cannot accurately reflect the overall effectiveness of the vibration isolation trench. Furthermore, discrete points may miss vibration peaks or abnormal propagation paths (such as diffraction caused by soil fissures or underground pipelines), potentially leading to deviations in the reliability assessment of the vibration reduction effect of the trench.

[0020] Based on this, at least one embodiment of this disclosure provides a reliability evaluation method for the vibration reduction effect of vibration isolation trenches based on big data. Please refer to... Figure 1 It illustrates the process of the method, including steps S101 to S102.

[0021] In step S101, vibration data from vibration sensors in the vibration source area, vibration isolation trench, and vibration reduction protection area, vibration source data of the vibration source area, structural data of the vibration isolation trench, and environmental data of the vibration source area, vibration reduction trench, and vibration reduction protection area are acquired. The environmental data includes geological data and meteorological data.

[0022] In step S102, interpolation calculations are performed based on the vibration data to obtain gridded vibration data of a preset scale.

[0023] In step S103, spatial attention is calculated based on environmental data and gridded vibration data at the same time point using a spatiotemporal feature model to obtain spatial vibration features. Temporal attention is calculated based on environmental data and gridded vibration data from the same vibration sensor at different time points to obtain temporal vibration features. Spatial vibration features, temporal vibration features, vibration source data, and environmental data are then fused into a spatiotemporal feature vector.

[0024] In step S104, a target vibration propagation map is obtained using a graph neural network model based on spatiotemporal feature vectors, vibration source data, geological data, and an initial vibration propagation map. The initial vibration propagation map is used to characterize the vibration propagation relationship within the vibration source region, the isolation trench, and the vibration reduction protection zone.

[0025] In step S105, the vibration reduction effect of the vibration isolation trench is evaluated based on vibration source data, structural data, environmental data, and target vibration propagation diagram, and the vibration reduction effect evaluation result is obtained.

[0026] Therefore, firstly, gridded vibration data of a preset scale can be generated through interpolation calculations to intuitively display the entire spatial propagation process of vibration waves. Then, the spatiotemporal dependencies in the vibration field are learned through a spatiotemporal feature model to capture the propagation modes and anomalous paths of vibration waves, obtaining spatiotemporal feature vectors. Next, a target vibration propagation map is constructed by combining the spatiotemporal feature vectors of vibration data, source data, and geological data using a graph neural network, quantitatively evaluating the effectiveness of the isolation trench throughout the propagation path. Finally, the semantic capabilities of a multimodal large model are used to evaluate the vibration reduction effect of the isolation trench based on source data, structural data, environmental data, and the target vibration propagation map, obtaining the vibration reduction effect evaluation result. Thus, discrete-point vibration monitoring can be transformed into continuous-field vibration monitoring, comprehensively reflecting the spatial attenuation law of vibration waves, thereby improving the accuracy of the reliability assessment of the vibration reduction effect of the isolation trench.

[0027] To facilitate understanding, the steps described above will be further explained below.

[0028] For example, vibration data may include at least one of vibration amplitude, vibration velocity, and vibration acceleration, and can be collected by deploying vibration sensors at intervals in the vibration source area, vibration isolation trench, and vibration reduction protection area.

[0029] For example, the vibration source area could be a construction area, and correspondingly, the vibration source data could be construction data, such as the location of vibration sources like dump trucks and road rollers.

[0030] For example, structural data may include the width, depth, and material of the vibration isolation trench. For instance, such as... Figure 2As shown, a vibration isolation trench, 2m wide and 5m deep, is constructed 5m outside the bottom of the highway embankment slope, and filled with polyethylene foam board. The structural data of this vibration isolation trench may include: {width: 2m, depth: 5m, material: polyethylene foam board}.

[0031] For example, geological data may include soil layer distribution, soil density, and geowave velocity. Geowave velocity can be obtained using ground-penetrating radar. Meteorological data may include humidity and rainfall.

[0032] In some embodiments, the present disclosure is applied to a big data center terminal. The vibration data is sent to the big data center terminal after the vibration data collected by the vibration sensor is compressed by the edge terminal of the vibration source area. The vibration data collected by the vibration sensor includes at least one of the following: vibration data of heavy machinery driving, vibration data of road roller construction, vibration data of soft soil treatment construction, and vibration data of pile driving construction. Different types of vibration data correspond to different compression strategies.

[0033] For example, in the scenario of highway reconstruction and expansion construction, vibration sources that may generate vibration include:

[0034] (1) Vibration generated during the operation of dump trucks and heavy construction machinery: Heavy vehicles such as dump trucks are prone to vibration when encountering uneven road surfaces. Sudden braking and frequent acceleration and deceleration during operation can also generate vibration. Vibration is also prone to occur when the vehicle's shock absorption system malfunctions.

[0035] (2) Vibration generated during road roller compaction: Road construction requires compaction for both subgrade filling and pavement structure layer construction, and vibration will be generated during road roller compaction.

[0036] (3) Vibration generated during soft soil treatment construction: Hydraulic compaction, impact rolling, dynamic compaction and other soft soil treatment methods will generate vibration.

[0037] (4) Vibration is generated during pile driving.

[0038] It should be understood that, in order to more accurately capture the details of these vibration signal changes and thus more accurately reflect the actual vibration situation, the embodiments of this disclosure employ a high sampling rate, such as 10kHz. This generates a large amount of data. If big data analysis is performed based on an edge terminal on the vibration source side, there may be insufficient computing power, thus affecting the efficiency of the reliability assessment of the vibration reduction effect. Therefore, the embodiments of this disclosure perform big data analysis through a big data center terminal and compress the vibration data before transmission through an edge terminal, which can simultaneously ensure the accuracy and efficiency of the assessment.

[0039] For example, the vibration source type is first determined by analyzing the vibration source data and its temporal and frequency domain characteristics. Then, different compression strategies are applied to vibration data with different source types. For instance, dynamic waveform segments are preserved from heavy machinery driving vibration data to retain the characteristics of sudden braking. Video matrix sparse coding is used for road roller construction vibration data to preserve the dominant frequency energy distribution. Transient impact parameterization is performed on soft soil treatment construction vibration data to preserve peak value, dominant frequency, and pulse width information. Periodic differential coding is used for pile driving construction vibration data to preserve impact rhythm and abnormal waveforms.

[0040] Therefore, different compression strategies can be adopted based on the type of vibration source to fully preserve the specific physical characteristics of different vibration source data, thereby ensuring the accuracy of the reliability assessment of the vibration reduction effect of the vibration isolation trench.

[0041] It should be understood that the vibration data from vibration sensors are discrete point-like distributions. For example, vibration monitoring points before and after the vibration isolation trench are linearly arranged with a spacing of 3-5 meters. However, the actual propagation path of the vibration wave may bypass the vibration isolation trench, forming complex diffraction. These details are difficult to show in the original vibration data. Therefore, in this embodiment of the present disclosure, interpolation calculation is performed based on the vibration data in step S102. For example, Kriging interpolation calculation is used to convert discrete data points into continuous gridded vibration data, resulting in gridded vibration data with a spatial resolution of 0.1m.

[0042] In some embodiments, in step S102, the vibration data can be separated to obtain single vibration data from different vibration sources; the single vibration data from the heavy machinery driving process can be filtered using a pre-established dynamic model to obtain first vibration data; the single vibration data from the road roller construction can be harmonic resonance suppressed to obtain second vibration data; the single vibration data from the soft foundation treatment construction can be extracted by transient impact separation to obtain third vibration data; the single vibration data from the pile driving construction can be tracked by impact envelope to obtain fourth vibration data; the first vibration data, second vibration data, third vibration data and fourth vibration data can be fused to obtain fused vibration data; the fused vibration data can be interpolated to obtain gridded vibration data of a preset scale.

[0043] For example, vibration data from different vibration sources can be obtained by separating the data based on time-domain and frequency-domain characteristics. Then, different noise filtering methods can be applied to the individual vibration data from different sources. For instance, vibration filtering during heavy machinery operation hinges on the application of vehicle dynamics models. First, vehicle parameters and road conditions are obtained as a foundation. Then, dynamic factors such as load changes are processed using extended Kalman filtering. Finally, enhancement processing is applied to sudden braking characteristics. This step-by-step processing effectively separates real vehicle vibrations. Vibration filtering during road roller construction hinges on harmonic characteristic processing. First, the base frequency is updated by calculating the roller's wheel speed. Then, an adaptive notch filter locks the dominant frequency. Next, harmonic comb filters process the harmonic components. Finally, supplementary suppression is applied to ground resonance. The entire process forms a closed loop for frequency characteristic processing. Vibration separation in soft soil treatment relies on impact waveform matching. First, a standard impact waveform library is established as a reference. Then, a matching pursuit algorithm searches for similar waveforms in the signal. The effectiveness is verified based on physical constraints. Finally, the impact signal that meets the requirements is reconstructed and retained. This waveform matching method is particularly suitable for transient impacts. The key to extracting pile driving vibrations is the rhythmic regularity. First, the signal envelope is extracted to highlight impact events. Then, the time intervals between adjacent impacts are detected. Next, physical constraints are used to filter out abnormal events. Finally, only impacts that conform to the rhythmic regularity are retained. This time-based method can effectively extract periodic impacts.

[0044] Therefore, by using feature decoupling filtering, precise noise filtering is achieved before interpolation, thereby improving the accuracy of the interpolation results and thus improving the accuracy of the reliability assessment of the vibration reduction effect of the vibration reduction trench.

[0045] In some embodiments, in step S103, spatial attention calculation is performed based on environmental data and gridded vibration data at the same time point using a spatiotemporal feature model to obtain spatial vibration characteristics. This includes: using the spatial attention unit of the spatiotemporal feature model to perform position encoding based on environmental data and the position coordinates of vibration sensors corresponding to each gridded vibration data point to obtain spatial position encoding; performing spatial attention calculation based on the spatial position encoding to obtain initial spatial attention weights; correcting the initial spatial attention weights based on the geological wave velocity in the environmental data to obtain target spatial attention weights; and obtaining spatial vibration characteristics based on the target spatial attention weights.

[0046] For example, the spatiotemporal feature model can be an ST-Transformer model.

[0047] For example, the spatiotemporal feature model can assign a spatial coordinate code to each grid point, enabling the model to perceive the correlation of vibration data in spatial location (e.g., before, during, and after the vibration isolation trench). However, considering that vibration propagation is essentially the physical motion of waves in a medium, if location coding is performed solely using position coordinates in a traditional manner, it may fail to reflect geological variations (e.g., a crack at the end of the vibration isolation trench causing a sudden drop in wave velocity from 150 m / s to 120 m / s) and meteorological influences (e.g., heavy rain saturating the soil and reducing its stiffness by 40%), thus affecting the accuracy of the spatial correlation of vibration data. Therefore, embodiments of this disclosure integrate geological and meteorological data into the location coding, establishing a mapping relationship between spatial coordinates and physical properties, thereby improving the accuracy of the spatial correlation of vibration data. For instance, based on the traditional method, location coordinates can be used for coding first, and then the code corresponding to these location coordinates can be combined with the geological features corresponding to the geological data and the meteorological features corresponding to the meteorological data into a multi-dimensional vector (i.e., spatial location coding).

[0048] For example, the target spatial attention weight can be obtained by correcting the initial spatial attention weight based on the geovelocity in the environmental data, as follows: The corrected weight is determined based on the geovelocity according to the following formula: ,in, This represents the wave velocity difference between the two location coordinates corresponding to the initial spatial attention weight. Then, the product of this corrected weight and the initial spatial attention weight is used as the target spatial attention weight. Thus, the attention mechanism can be forced to follow the vibration propagation law by using a wave velocity difference exponential decay term, thereby improving the accuracy of spatial correlation of vibration data.

[0049] It should be understood that spatial attention is calculated based on spatial location encoding to obtain the initial spatial attention weights. The spatial attention calculation process can refer to relevant techniques, which will not be elaborated here. Additionally, the spatial vibration features can be obtained by multiplying the target spatial attention weights by the original vibration features, or similar techniques can be referenced, which will not be elaborated here.

[0050] In some embodiments, in step S103, time attention calculation is performed based on environmental data and gridded vibration data of the same vibration sensor at different time points to obtain time vibration characteristics. This includes: using the time attention unit of the spatiotemporal feature model to perform time encoding based on the time information of the gridded vibration data corresponding to the same vibration sensor to obtain time location encoding; performing time attention calculation based on the time location encoding to obtain target time attention weight; and obtaining time vibration characteristics based on the target time attention weight. In this case, the time scale factor used to control the time encoding period in the time encoding is determined based on meteorological data, and the time window length corresponding to the time attention calculation is determined based on the geological wave velocity in the environmental data.

[0051] For example, spatiotemporal feature models can be time-encoded using time information, assigning a time coordinate code to each grid point, enabling the model to perceive the correlation of time evolution (such as the propagation delay of vibration waves). For instance, time encoding can be performed using the following formula:

[0052] (1),

[0053] Where TE represents time location code, and t represents time point (unit: milliseconds ms). The time scale factor used to control the time coding period can be obtained based on the following formula:

[0054] (2),

[0055] It should be understood that on sunny days, =1000, corresponding to normal time perception. Heavy rain increases soil moisture content, decreasing wave velocity (e.g., from 150 m / s to 100 m / s), thus prolonging vibration propagation time. By increasing... The time encoding period becomes longer, which is equivalent to slowing down the perception of time by 50% in order to more accurately capture delayed vibrations.

[0056] For example, time-attention calculation may include the following process: First, based on the distance from the source location to the target location and the geological wave velocity data (including wave velocity variations along the path), the expected propagation time of the vibration wave is calculated to obtain the corrected wavefront arrival time. Then, based on the corrected wavefront arrival time, a time window is determined, and within this time window, vibration events that may affect the target point are searched for for time-attention calculation. In the time-attention mechanism, attention is paid to the vibration situation of the target point at the current time (e.g., T), and source vibration events at historical times (e.g., T') are considered. Then, for each historical time T', the time difference between it and the current time (i.e., T-T') is calculated, and this time difference is compared with the corrected wavefront arrival time: if this time difference is close to the corrected wavefront arrival time (i.e., |(T-T')-corrected wavefront arrival time| is small), then the vibration event at that historical time T' should have a greater impact on the target point at the current time, and therefore a higher attention weight is assigned.

[0057] It should be understood that the propagation speed of vibration waves in a medium is greatly affected by geological conditions. For example, in a homogeneous soil layer, the wave velocity may be relatively stable; however, at the junction of fissures, cavities, or different soil layers, the propagation speed will change (e.g., the propagation speed decreases in fissure areas). If the wavefront arrival time is not corrected and the average or ideal wave velocity is used directly to calculate the expected propagation time, the resulting propagation time will deviate from the actual situation (e.g., in fissure areas, the actual wave velocity is slower, and the propagation time is longer). Consequently, the time attention mechanism will search for vibration events within the wrong time window, leading to an inability to accurately correlate the source event with the target point vibration, thus reducing the model's prediction accuracy. Therefore, in this embodiment, the wavefront arrival time is corrected based on the geological wave velocity to ensure that the source event affecting the target point is found within the correct time window, enabling the model to adapt to complex geological conditions.

[0058] After obtaining the spatial vibration characteristics and temporal vibration characteristics, the spatial vibration characteristics, temporal vibration characteristics, vibration source characteristics corresponding to the vibration source data, and environmental characteristics corresponding to the environmental data can be concatenated into a spatiotemporal feature vector.

[0059] In some embodiments, the spatiotemporal feature model is trained as follows: Spatial attention is calculated based on sample environmental data from the vibration source region, isolation ditch, and vibration reduction protection zone, as well as sample gridded vibration data at the same time point, to obtain sample spatial vibration features. Temporal attention is then calculated based on sample gridded vibration data from the same vibration sensor at different time points to obtain sample temporal vibration features. The sample spatial vibration features, sample temporal vibration features, sample source data from the vibration source region, and sample environmental data are fused into a sample spatiotemporal feature vector. The sample gridded vibration data is obtained by interpolating historical measured vibration data from vibration sensors in the vibration source region, isolation ditch, and vibration reduction protection zone. Vibration data is reconstructed based on the sample spatiotemporal feature vector to obtain predicted vibration data. A first loss value is calculated based on the predicted vibration data and the corresponding historical measured vibration data. A predicted vibration displacement is calculated based on the predicted vibration data. A second loss value is calculated based on the wave equation, the predicted vibration displacement, and the sample geological wave velocity in the sample environmental data. The parameters of the spatiotemporal feature model are adjusted based on the first and second loss values.

[0060] For example, the sample spatiotemporal feature vector can be obtained based on the initial spatiotemporal feature model as described above. Then, the sample spatiotemporal feature vector output by the spatiotemporal feature model can be reconstructed into a vibration field using a lightweight decoder. The first loss value (i.e., reconstruction loss) is calculated using the measured vibration field grid data of historical projects and the reconstructed predicted vibration data, thereby ensuring the integrity of the feature output.

[0061] For example, embodiments of this disclosure also introduce a second loss value (i.e., physical equation constraint loss) to force the characteristics to satisfy the vibration propagation law. For example, the second loss value can be obtained by the following formula:

[0062] (3),

[0063] in, This represents the second loss value. Indicates the predicted vibration displacement. Indicates the geological wave velocity of the sample. Represents a time variable. This represents the Laplace operator.

[0064] It should be understood that, as shown in the following calculation, the wave equation describes the fundamental physical laws governing the propagation of vibration waves:

[0065] (4),

[0066] When the displacement field output by the model When the calculation formula (3) is not satisfied, the second loss value increases, thereby guiding the model to adjust the parameters so that the decoded feature vector satisfies the calculation formula (3). Thus, the displacement field output by the model can satisfy the real physical laws.

[0067] In this way, no manual annotation of feature vectors is required during the training of the spatiotemporal feature model. The training can supervise the integrity of the output feature information through vibration field reconstruction loss and supervise the output feature to meet the real physical laws through physical equation constraints. This achieves automatic mapping between feature vectors and physical laws (such as diffraction path dimension), and the annotation cost is low, which improves the training efficiency of the spatiotemporal feature model and thus improves the reliability evaluation efficiency of the vibration reduction effect of the vibration isolation ditch.

[0068] In some embodiments, in step S104, obtaining a target vibration propagation map using a graph neural network model based on spatiotemporal feature vectors, vibration source data, geological data, and an initial vibration propagation map includes: updating the node features and edge weights in the initial vibration propagation map using a graph neural network model based on spatiotemporal feature vectors and environmental data to obtain an intermediate vibration propagation map; determining a node update strategy for the intermediate vibration propagation map based on geological data and the vibration source data, wherein the node update strategy includes a node addition strategy and a node deletion strategy, the node addition strategy being used to add a first node in the intermediate vibration propagation map and initialize the first node based on geological data, and the node deletion strategy being used to delete a second node in the intermediate vibration propagation map and the edge connected to the second node; and executing the node update strategy on the intermediate vibration propagation map to obtain the target vibration propagation map.

[0069] In some embodiments, the initial vibration propagation map is established as follows: taking the vibration sensor, the specific engineering component in the vibration isolation trench, and the specific protection location in the vibration reduction protection zone as nodes, respectively, the edges between the nodes are constructed based on the distance and geological properties between the vibration sensor, the specific engineering component, and the specific protection location, and the initial vibration propagation map is obtained. The weight of the edge in the initial vibration propagation map is obtained based on the preset engineering attenuation coefficient, the distance between the nodes connected by the edge, the average geological wave velocity of the path corresponding to the edge, and the average geological density of the path corresponding to the edge.

[0070] For example, the initial vibration propagation diagram may include sensor nodes, vibration isolation trench structure nodes, and protected nodes. Sensor nodes correspond to vibration monitoring points, vibration isolation trench structure nodes correspond to specific vibration isolation engineering components, and protected nodes correspond to the vibration propagation endpoint and a specific protection location. Specific engineering components may be, for example, the deep steel sheet piles of the vibration isolation trench, and protected nodes may be, for example, the west-side column foundation of the protected factory building.

[0071] For example, the edge construction rules in the initial vibration propagation map include: forced connection of adjacent spaces, cross-structure attenuation connection, geological anomaly path connection, and long-distance attenuation blocking. Forced connection of adjacent spaces ensures the continuity of energy transfer. Since isolation trenches and building foundations alter wave impedance, cross-structure attenuation connection quantifies the vibration reduction efficiency of the engineering structure. Since cracks or pipes can form waveguides, geological anomaly path connection can capture anomalous propagation risks. Furthermore, considering geometric diffusion and material attenuation, long-distance attenuation blocking avoids unnecessary calculations and improves efficiency.

[0072] In some embodiments, the weights of the edges in the initial vibration propagation graph are obtained based on the following formula:

[0073] (5),

[0074] in, Nodes in the initial vibration propagation diagram With nodes The weight of the edges between them. This represents the preset engineering attenuation coefficient. Represents a node With nodes The distance between, Represents a node With nodes The average geological wave velocity between Represents a node With nodes The average geological density between.

[0075] For example, engineering attenuation coefficient It can be pre-calibrated to 0.05.

[0076] It should be understood that, according to the vibration wave energy attenuation equation: ,in, , This represents the time decay coefficient, which can be obtained through shaking table testing. A medium inertia term is introduced. Correction: ,in, The density attenuation coefficient can be obtained through resonant column tests on soil samples. Then, based on energy transfer experimental data, it was found that density... Increasing this will amplify the time decay effect; therefore, it is possible to make ,in, This represents the proportionality constant. Accordingly, we can obtain: Then, calibration was performed through experiments. and The relationship was found to be: .

[0077] Therefore, through the above calculation formula (5), the attenuation law of the wave equation is transformed into a computable topological parameter, so that the edge weights between nodes in the vibration propagation diagram change with geological data, better reflecting the actual vibration propagation law. Furthermore, through a single parameter (i.e., the engineering attenuation coefficient)... This control facilitates real-time engineering calculations.

[0078] For example, in the process of updating the node features in the initial vibration propagation map using a graph neural network model based on spatiotemporal feature vectors and geological data, different update methods can be used for sensor nodes, vibration isolation trench nodes, and protected nodes. For instance, the node features of sensor nodes in the vibration propagation map include vibration values ​​and vibration spectrum characteristics; the node features of vibration isolation trench nodes include transmission coefficients and integrity; and the node features of protected nodes include safety thresholds and vibration exceedance risks. The transmission coefficient characterizes the energy transmittance of vibration waves through the vibration isolation trench and can be obtained through geological density and geological wave velocity; integrity is used to quantify the structural integrity of the vibration isolation trench and can be obtained based on methods such as crack identification. In the actual update process, the node features of sensor nodes can be updated based on spatiotemporal feature vectors, the node features of vibration isolation trench nodes can be updated based on geological data, and the node features of protected nodes can be updated based on both spatiotemporal feature vectors and geological data.

[0079] For example, different update methods can be used for updating edge weights for sensor nodes, vibration isolation ditch nodes, and protected nodes. For instance, in the process of calculating edge weights based on the above formula (5), the edge weight between sensor nodes and vibration isolation ditch nodes is calculated as follows: and Take the values ​​before the trench is formed, and determine the edge weights between the nodes of the vibration isolation trench and the protected nodes. and Take the value after the ditch.

[0080] For example, based on geological data and vibration source data, a node update strategy is determined for the intermediate vibration propagation map. For instance, if a new fracture is discovered through geological data analysis, requiring the insertion of a node, a first node is added to the intermediate vibration propagation map and connected to neighboring nodes. Then, initial node characteristic values ​​are generated based on geological data and physical rules. As another example, if the sheet piles in the vibration isolation trench are being removed, requiring the removal of a node, the second node and the edges connected to it in the intermediate vibration propagation map are deleted.

[0081] This enables the physical-driven topological evolution of vibration propagation diagrams, better adapting to actual assessment needs and improving the applicability of scenarios for assessing the reliability of vibration reduction effects of isolation trenches.

[0082] In some embodiments, in step S104, the vibration reduction effect of the vibration isolation trench can be evaluated based on vibration source data, structural data, environmental data and target vibration propagation diagram using a multimodal large model to obtain the vibration reduction effect evaluation result. The vibration reduction effect evaluation result includes the overall vibration reduction efficiency of the vibration isolation trench, the confidence level of the overall vibration reduction efficiency, the probability of vibration exceeding the standard at the location of the vibration sensor and the specific protection location in the vibration reduction protection zone, and the risk interpretation information of the probability of vibration exceeding the standard.

[0083] For example, the multimodal large model can evaluate the overall vibration reduction efficiency based on vibration source data, structural data, environmental data, and target vibration propagation maps, and uses a Beta distribution to characterize the efficiency uncertainty, outputting a 95% confidence interval. For instance, the multimodal large model could output: "At a 95% confidence level, the current vibration isolation efficiency is ≥76.2%." Furthermore, the multimodal large model can calculate the probability of vibration exceeding the standard at the location of vibration sensors and specific protected locations within the vibration reduction protection zone, and provide risk interpretation information for the probability of vibration exceeding the standard. For example, the multimodal large model could output: "The probability of vibration exceeding the standard in the precision laboratory in the western area of ​​the factory is greater than 60%, mainly because an abnormal pattern was identified. Under the combination of H-300 pile hammer and soil moisture >25%, vibration is easily propagated around the vibration isolation trench, resulting in a decrease in vibration reduction effect of approximately 40%."

[0084] Therefore, a reliability assessment method for the vibration reduction effect of vibration isolation trenches can be developed, which enables continuous perception throughout the entire space, fusion of multi-source data, and dynamic and probabilistic reliability assessment using artificial intelligence technology, thereby improving the accuracy of reliability assessment for the vibration reduction effect of vibration isolation trenches.

[0085] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0086] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0087] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. A reliability assessment method for vibration reduction effect of vibration isolation trenches based on big data, characterized in that, include: The vibration data of vibration sensors in the vibration source area, vibration isolation trench and vibration reduction protection area, the vibration source data of the vibration source area, the structural data of the vibration isolation trench, and the environmental data of the vibration source area, the vibration reduction trench and the vibration reduction protection area are acquired. The environmental data includes geological data and meteorological data. Interpolation calculations are performed based on the vibration data to obtain gridded vibration data of a preset scale; Spatial vibration features are obtained by performing spatial attention calculation based on the environmental data and gridded vibration data at the same time point using a spatiotemporal feature model. Temporal vibration features are obtained by performing temporal attention calculation based on the environmental data and gridded vibration data of the same vibration sensor at different time points. The spatial vibration features, the temporal vibration features, the vibration source data, and the environmental data are then fused into a spatiotemporal feature vector. A target vibration propagation map is obtained by using a graph neural network model based on the spatiotemporal feature vector, the vibration source data, the geological data, and the initial vibration propagation map. The initial vibration propagation map is used to characterize the vibration propagation relationship in the vibration source area, the vibration isolation trench, and the vibration reduction protection area. The vibration reduction effect of the vibration isolation trench is evaluated based on the vibration source data, the structural data, the environmental data, and the target vibration propagation diagram, and the vibration reduction effect evaluation result is obtained.

2. The reliability assessment method for vibration reduction effect of vibration isolation trenches based on big data according to claim 1, characterized in that, The spatial vibration characteristics are obtained by performing spatial attention calculation based on the environmental data and gridded vibration data at the same time point using a spatiotemporal feature model, including: The spatial attention unit of the spatiotemporal feature model performs position encoding based on the environmental data and the position coordinates of the vibration sensor corresponding to each gridded vibration data to obtain a spatial position code. Based on the spatial position code, spatial attention is calculated to obtain an initial spatial attention weight. The initial spatial attention weight is then corrected based on the geological wave velocity in the environmental data to obtain a target spatial attention weight. Based on the target spatial attention weight, the spatial vibration characteristics are obtained.

3. The reliability assessment method for vibration reduction effect of vibration isolation trenches based on big data according to claim 1, characterized in that, The time-attention calculation based on the environmental data and the gridded vibration data of the same vibration sensor at different time points yields time-vibration characteristics, including: The time attention unit of the spatiotemporal feature model performs time encoding based on the time information of the gridded vibration data corresponding to the same vibration sensor to obtain time location encoding. Time attention is then calculated based on the time location encoding to obtain the target time attention weight. Based on the target time attention weight, the time vibration feature is obtained. The time scale factor used to control the time encoding period in the time encoding is determined based on the meteorological data. The time window length corresponding to the time attention calculation is determined after correcting the wavefront arrival time based on the geological wave velocity in the environmental data.

4. The reliability assessment method for vibration reduction effect of vibration isolation trenches based on big data according to claim 1, characterized in that, The spatiotemporal feature model is trained in the following manner: Spatial attention is calculated based on sample environmental data from the vibration source region, the vibration isolation ditch, and the vibration reduction protection zone, as well as sample gridded vibration data at the same time point, using an initial spatiotemporal feature model to obtain sample spatial vibration characteristics. Temporal attention is then calculated based on sample gridded vibration data from the same vibration sensor at different time points to obtain sample temporal vibration characteristics. The sample spatial vibration characteristics, the sample temporal vibration characteristics, the sample vibration source data from the vibration source region, and the sample environmental data are fused into a sample spatiotemporal feature vector. The sample gridded vibration data is obtained by interpolating historical measured vibration data from vibration sensors in the vibration source region, the vibration isolation ditch, and the vibration reduction protection zone. The sample spatiotemporal feature vector is reconstructed into predicted vibration data, and a first loss value is calculated based on the predicted vibration data and the corresponding historical measured vibration data. The predicted vibration displacement is calculated based on the predicted vibration data, and a second loss value is calculated based on the wave equation, the predicted vibration displacement, and the sample geological wave velocity in the sample environmental data. Based on the first loss value and the second loss value, the parameters of the spatiotemporal feature model are adjusted.

5. The reliability assessment method for vibration reduction effect of vibration isolation trenches based on big data according to claim 1, characterized in that, The process of obtaining the target vibration propagation map using a graph neural network model based on the spatiotemporal feature vector, the vibration source data, the geological data, and the initial vibration propagation map includes: The node features and edge weights in the initial vibration propagation graph are updated using a graph neural network model based on the spatiotemporal feature vector and the environmental data to obtain an intermediate vibration propagation graph. Based on the geological data and the vibration source data, a node update strategy for the intermediate vibration propagation map is determined. The node update strategy includes a node addition strategy and a node deletion strategy. The node addition strategy is used to add a first node to the intermediate vibration propagation map and initialize the first node based on the geological data. The node deletion strategy is used to delete a second node and the edge connected to the second node in the intermediate vibration propagation map. The node update strategy is applied to the intermediate vibration propagation map to obtain the target vibration propagation map.

6. The reliability assessment method for vibration reduction effect of vibration isolation trenches based on big data according to any one of claims 1-5, characterized in that, The initial vibration propagation diagram was established in the following manner: Using the vibration sensor, the specific engineering component in the vibration isolation trench, and the specific protected location in the vibration reduction protection zone as nodes, edges between the nodes are constructed based on the distance and geological properties between the vibration sensor, the specific engineering component, and the specific protected location to obtain an initial vibration propagation map. The weight of the edges in the initial vibration propagation map is obtained based on a preset engineering attenuation coefficient, the distance between the nodes connected by the edges, the average geological wave velocity of the path corresponding to the edge, and the average geological density of the path corresponding to the edge.

7. The reliability assessment method for vibration reduction effect of vibration isolation trenches based on big data according to claim 6, characterized in that, The weights of the edges in the initial vibration propagation graph are obtained based on the following formula: ， in, The nodes in the initial vibration propagation diagram are represented. With nodes The weight of the edges between them. This represents the preset engineering attenuation coefficient. Represents the node With the node The distance between, Represents the node With the node The average geological wave velocity between Represents the node With the node The average geological density between.

8. The reliability assessment method for vibration reduction effect of vibration isolation trench based on big data according to any one of claims 1-5, characterized in that, The step of interpolating the vibration data to obtain gridded vibration data of a preset scale includes: The vibration data is separated to obtain individual vibration data from different vibration sources; The first vibration data is obtained by filtering the single vibration data from the heavy machinery driving process through a pre-established dynamic model, the second vibration data is obtained by harmonic resonance suppression processing of the single vibration data from the road roller construction, the third vibration data is obtained by transient impact separation extraction of the single vibration data from the soft foundation treatment construction, and the fourth vibration data is obtained by impact envelope tracking of the single vibration data from the pile driving construction. The first vibration data, the second vibration data, the third vibration data, and the fourth vibration data are fused to obtain fused vibration data; Interpolation calculations are performed on the fused vibration data to obtain gridded vibration data of a preset scale.

9. The reliability assessment method for vibration reduction effect of vibration isolation trenches based on big data according to any one of claims 1-5, characterized in that, The method is applied to a big data center terminal. The vibration data is sent to the big data center terminal after the edge terminal of the vibration source area compresses the vibration data collected by the vibration sensor. The vibration data collected by the vibration sensor includes at least one of the vibration data from heavy machinery operation, vibration data from road roller construction, vibration data from soft soil treatment construction, and vibration data from pile driving construction. Different compression strategies are corresponding to vibration data of different vibration source types.

10. The reliability assessment method for vibration reduction effect of vibration isolation trenches based on big data according to any one of claims 1-5, characterized in that, The vibration reduction effect of the isolation trench is evaluated based on the vibration source data, the structural data, the environmental data, and the target vibration propagation diagram, resulting in a vibration reduction effect evaluation result, including: The vibration reduction effect of the vibration isolation trench is evaluated using a multimodal large model based on the vibration source data, the structural data, the environmental data, and the target vibration propagation diagram. The vibration reduction effect evaluation result includes the overall vibration reduction efficiency of the vibration isolation trench, the confidence level of the overall vibration reduction efficiency, the probability of vibration exceeding the standard at the location of the vibration sensor and the specific protected location in the vibration reduction protection zone, and the risk interpretation information of the probability of vibration exceeding the standard.

Images