Multi-source sensing fusion slope intelligent monitoring method and system

Abstract

The invention discloses a multi-source sensing fusion slope intelligent monitoring method and system, and relates to the technical field of slope monitoring. A multi-source sensing fusion slope intelligent monitoring system comprises a data acquisition module, a data processing module, a model construction module, a state evaluation module and a scheme distribution module. Based on the prediction result and the displacement distance of the side slope in the latest sampling period, the stability score of the side slope is obtained through calculation, the instant deformation condition of the side slope can be reflected by means of the current actual displacement, and the possible displacement change in the future can be considered in advance by means of the predicted displacement; therefore, the real-time state and the potential development trend of the side slope are comprehensively considered, the potential side slope instability risk is better predicted, the stability evaluation is more comprehensive, meanwhile, the comprehensive consideration mode can improve the evaluation accuracy and enhance the pre-judgment capability of the side slope state change, and the stability evaluation is more comprehensive. Therefore, more reliable data support is provided for maintenance, management and monitoring of the side slope.

Description

Technical Field

[0001] This invention relates to the field of slope monitoring technology, specifically to a multi-source sensor fusion intelligent slope monitoring method and system. Background Technology

[0002] A slope is a sloping surface with a certain gradient, either on the ground surface or after artificial excavation. It is commonly found on mountainsides, along roadsides, and in building foundations. Slopes are prone to deformation and instability under the influence of natural factors and human activities, leading to serious geological disasters such as collapses, landslides, and debris flows, threatening human lives and the integrity of infrastructure. Real-time monitoring of slopes using professional equipment and technology can accurately grasp key indicators such as deformation, stress changes, and groundwater level dynamics, allowing for early detection of potential risks and providing valuable time for disaster warning and prevention, thus protecting the lives and property of residents. Simultaneously, monitoring data can be fed back into the slope design and construction stages, helping to optimize support schemes and reinforcement measures, improving slope stability and reliability, saving engineering costs, extending the service life of projects, and ensuring the long-term stable operation of infrastructure. Furthermore, this also has positive implications for the rational development and utilization of land resources and ecological environmental protection.

[0003] Existing intelligent slope monitoring methods and systems typically monitor various parameters of the slope using different sensors, then assess the slope's condition based on the monitoring results and assign corresponding monitoring plans. However, this condition assessment method and plan allocation strategy has a high lag, making it impossible to predict the slope's condition and allocate suitable plans in advance. This results in low rationality in the allocation of monitoring plans by these intelligent slope monitoring methods and systems. Furthermore, existing intelligent slope monitoring methods and systems struggle to effectively fuse multi-source sensor data acquired from various sensors, leading to low accuracy in assessing the slope's condition.

[0004] Based on the above, this invention proposes a multi-source sensor fusion-based intelligent slope monitoring method and system with high evaluation accuracy and reasonable monitoring scheme allocation. Summary of the Invention

[0005] To overcome the shortcomings of existing intelligent slope monitoring methods and systems, such as high lag in slope condition assessment and monitoring scheme allocation, inability to predict slope condition and pre-allocate suitable schemes, resulting in low rationality of monitoring scheme allocation and difficulty in effectively fusing multi-source sensor data from various sensors, leading to low accuracy in slope condition assessment, this invention proposes a multi-source sensor fusion-based intelligent slope monitoring method and system with high assessment accuracy and reasonable monitoring scheme allocation.

[0006] A multi-source sensor fusion-based intelligent slope monitoring method includes the following steps:

[0007] Surface monitoring data of the slope is obtained through GNSS monitoring stations and MEMS tilt sensors; underground monitoring data of the slope is obtained through pore water pressure gauges and soil moisture sensors; and environmental monitoring data of the area where the slope is located is obtained through weighing rain gauges and temperature sensors.

[0008] The acquired surface monitoring data, subsurface monitoring data, and environmental monitoring data are preprocessed to obtain preprocessed surface monitoring data, subsurface monitoring data, and environmental monitoring data;

[0009] A slope displacement prediction model based on an optimized bidirectional LSTM model is constructed. The slope displacement prediction model is trained, and the hyperparameters of the slope displacement prediction model are optimized during the model training period using the Saurus Aquatic Optimization (SSA) algorithm.

[0010] The preprocessed surface monitoring data, underground monitoring data, and environmental monitoring data are input into the trained slope displacement prediction model to obtain the prediction results. The prediction results are the predicted displacement distance of the slope in the next sampling period. Based on the prediction results and the displacement distance of the slope in the latest sampling period, the stability score of the slope is calculated.

[0011] Based on stability scores and according to preset grading rules, slopes are classified into corresponding risk levels, and corresponding monitoring schemes are assigned to slopes based on risk levels.

[0012] As a preferred aspect of the invention, the surface monitoring data includes slope displacement change data and slope angle change data; while the underground monitoring data includes slope pore water pressure change data and soil moisture content change data; and the environmental monitoring data includes rainfall change data and temperature change data in the area where the slope is located.

[0013] As a preferred aspect of the invention, the specific steps for preprocessing the acquired surface monitoring data, underground monitoring data, and environmental monitoring data are as follows:

[0014] Data transformation and time alignment convert all acquired data into time series format, unify the starting reference time point of all time series, unify the time frequency of all time series, and fill the missing values ​​of time series with low time frequency at new time points through linear interpolation or polynomial interpolation methods to ensure that all time series data have values ​​at the same time points.

[0015] Missing and outlier handling: Linear or polynomial interpolation methods are used to fill in null and missing values ​​in the time series. Outliers that do not conform to the expected pattern in the time series are identified by Z-Score or IQR methods. Outliers are removed and replaced with the mean or median of the time series or repaired by interpolation methods.

[0016] Data normalization, through min-max normalization, normalizes data of different dimensions. The formula for min-max normalization is as follows: ,in Represents the original data value. This represents the minimum value in a time series. This represents the maximum value in a time series. This represents the normalized data value;

[0017] Kalman filtering noise reduction sets corresponding state transition models based on the changing characteristics of different types of data, and then performs personalized noise reduction processing on each data item through Kalman filtering iteration.

[0018] As a preferred aspect of the invention, the slope displacement prediction model based on the optimized bidirectional LSTM model includes an input layer, a bidirectional LSTM layer, a temporal attention layer, a spatial attention layer, and an output layer. The input layer receives input data. The bidirectional LSTM layer receives the data from the input layer and simultaneously utilizes forward and backward information from the time series to capture long-term dependencies within the time series. The temporal attention layer applies an attention mechanism to the temporal dimension of the sequence, i.e., it calculates the weight of each time step and performs a weighted summation of the outputs of the bidirectional LSTM layer. The spatial attention layer applies an attention mechanism to the spatial dimension of the sequence, i.e., it weights different feature channels of the sequence output by the temporal attention layer, highlighting important features and suppressing irrelevant features. The output layer receives the output of the spatial attention layer and generates the final prediction result.

[0019] As a preferred aspect of the invention, the specific steps for training the slope displacement prediction model and optimizing the hyperparameters of the slope displacement prediction model during model training using the Slug-Sea-Shy optimization algorithm (SSA) are as follows:

[0020] Historical surface monitoring data, historical underground monitoring data, and historical environmental monitoring data of the slope were acquired. All acquired data were preprocessed, and the preprocessed data were then proportionally analyzed. The slope displacement prediction model was initialized by dividing it into training, validation, and test sets.

[0021] The SSA optimization algorithm for salps is initialized, and the population size is set to 0. The maximum number of iterations is The search results range is The initial population position is randomly generated, and each individual represents a set of hyperparameter combinations, including learning rate, number of hidden layer nodes and temporal window size;

[0022] Using root mean square error As the fitness function, the error of the slope displacement prediction model on the validation set is calculated using the following formula:

[0023]

[0024] in Indicates the first in the verification set The actual value of the slope displacement Then it means the first Predicted values ​​of slope displacement. This represents the total number of data points in the validation set;

[0025] The leader and follower positions are updated according to the formula until the maximum number of iterations is reached or the fitness converges. At this point, the optimal individual position, i.e., the optimal hyperparameter combination, is output. The specific calculation formula is as follows:

[0026]

[0027]

[0028] in Indicates the updated leader position. This indicates the position of the individual with the best fitness value in the current population. Indicates the current iteration number. This indicates the generation of a value in the interval The random number within, on the left side of the formula Indicates the first The updated position of each follower, on the right side of the formula. Indicates the first The current location of each follower. Indicates the first The position preceding the individual follower;

[0029] The slope displacement prediction model is re-initialized using the optimal hyperparameter combination, then trained using the training set and tuned using the validation set, and finally evaluated using the test set.

[0030] As a preferred aspect of the invention, the specific steps for calculating the slope stability score based on the prediction results and the slope displacement distance in the latest sampling period are as follows:

[0031] The displacement distance of the slope in the latest sampling period was obtained from the preprocessed surface monitoring data. ;

[0032] Based on the slope displacement distance in the latest sampling period and the predicted displacement distance in the next sampling period The stability score of the slope was obtained through calculation. The specific calculation formula is as follows:

[0033]

[0034] in and Displacement distance and predicted displacement distance The weighting coefficients, and ,in This indicates the preset slope displacement risk threshold.

[0035] A multi-source sensor fusion intelligent slope monitoring system includes:

[0036] The data acquisition module is used to acquire surface monitoring data of the slope through GNSS monitoring station and MEMS tilt sensor, underground monitoring data of the slope through pore water pressure gauge and soil moisture sensor, and environmental monitoring data of the area where the slope is located through weighing rain gauge and temperature sensor.

[0037] The data processing module is used to preprocess the acquired surface monitoring data, underground monitoring data and environmental monitoring data to obtain preprocessed surface monitoring data, underground monitoring data and environmental monitoring data;

[0038] The model building module is used to build a slope displacement prediction model based on an optimized bidirectional LSTM model, train the slope displacement prediction model, and optimize the hyperparameters of the slope displacement prediction model through the Saurus Elixir Algorithm (SSA) during model training.

[0039] The state assessment module is used to input preprocessed surface monitoring data, underground monitoring data and environmental monitoring data into the trained slope displacement prediction model to obtain prediction results. The prediction results are the predicted displacement distance of the slope in the next sampling period. Based on the prediction results and the displacement distance of the slope in the latest sampling period, the stability score of the slope is calculated.

[0040] The scheme allocation module is used to classify slopes into corresponding risk levels based on stability scores and preset grading rules, and to allocate corresponding monitoring schemes to slopes based on risk levels.

[0041] The present invention has the following advantages:

[0042] 1. This invention calculates the slope stability score based on prediction results and the slope's displacement distance within the latest sampling period. It not only reflects the slope's immediate deformation using the current actual displacement but also anticipates potential future displacement changes using predicted displacement. This comprehensive approach considers both the slope's real-time state and potential development trends, enabling better prediction of potential slope instability risks and a more comprehensive stability assessment. Furthermore, this integrated approach not only improves the accuracy of the assessment but also enhances the ability to predict slope state changes, providing more reliable data support for slope maintenance, management, and monitoring. This improves the accuracy of the intelligent slope monitoring method and system and the rationality of monitoring scheme allocation.

[0043] 2. This invention combines a bidirectional LSTM model with a temporal and spatial dual attention mechanism and applies it to slope displacement prediction. This not only fully mines and utilizes the spatiotemporal features in the data, but also enables the slope displacement prediction model to focus on the time step in historical data that is most valuable for current prediction, and to highlight the features with the strongest correlation to displacement in multi-source data. This effectively improves the slope displacement prediction model's learning ability on complex spatiotemporal data, thereby improving the model's prediction accuracy and enhancing its ability to capture slope displacement change trends. This provides stronger support for slope stability assessment and disaster prevention, and indirectly improves the assessment accuracy and the rationality of monitoring scheme allocation of this intelligent slope monitoring method and system.

[0044] 3. During model training, this invention optimizes the hyperparameters of the slope displacement prediction model using the Salicylia Optimization Algorithm (SSA). This not only efficiently searches for the optimal hyperparameter combination, thereby improving the convergence speed and generalization performance of the slope displacement prediction model and shortening its training time, but also effectively explores and develops globally by simulating the group foraging behavior of salicylia, avoiding local optima. This improves the prediction accuracy and performance of the slope displacement prediction model, indirectly enhancing the evaluation accuracy and the rationality of the monitoring scheme allocation of this intelligent slope monitoring method and system. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating a multi-source sensor fusion-based intelligent slope monitoring method used in an embodiment of the present invention.

[0046] Figure 2This is a schematic diagram of the structure of a multi-source sensor fusion intelligent slope monitoring system used in an embodiment of the present invention. Detailed Implementation

[0047] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0048] Example 1: A multi-source sensor fusion intelligent slope monitoring method, such as... Figure 1 As shown, it includes the following steps:

[0049] Surface monitoring data of the slope is obtained through GNSS monitoring stations and MEMS tilt sensors; underground monitoring data of the slope is obtained through pore water pressure gauges and soil moisture sensors; and environmental monitoring data of the area where the slope is located is obtained through weighing rain gauges and temperature sensors.

[0050] The acquired surface monitoring data, subsurface monitoring data, and environmental monitoring data are preprocessed to obtain preprocessed surface monitoring data, subsurface monitoring data, and environmental monitoring data;

[0051] A slope displacement prediction model based on an optimized bidirectional LSTM model is constructed. The slope displacement prediction model is trained, and the hyperparameters of the slope displacement prediction model are optimized during the model training period using the Saurus Aquatic Optimization (SSA) algorithm.

[0052] The preprocessed surface monitoring data, underground monitoring data, and environmental monitoring data are input into the trained slope displacement prediction model to obtain the prediction results. The prediction results are the predicted displacement distance of the slope in the next sampling period. Based on the prediction results and the displacement distance of the slope in the latest sampling period, the stability score of the slope is calculated.

[0053] Based on stability scores and according to preset grading rules, slopes are classified into corresponding risk levels, and corresponding monitoring schemes are assigned to slopes based on risk levels.

[0054] It should be noted that the preset classification rules need to be set based on the properties of the slope soil. For example, the classification rules for ordinary soil slopes will be more lenient than those for rock slopes. The different monitoring schemes based on risk levels should include, but are not limited to, adding or removing slope monitoring points and increasing or decreasing the monitoring frequency. Furthermore, after classifying the slope into the corresponding risk level, the corresponding level of early warning can be issued and corresponding protective measures can be taken.

[0055] The surface monitoring data includes slope displacement and dip angle changes; the underground monitoring data includes slope pore water pressure and soil moisture content changes; and the environmental monitoring data includes rainfall and temperature changes in the area where the slope is located.

[0056] The specific steps for preprocessing the acquired surface monitoring data, subsurface monitoring data, and environmental monitoring data are as follows:

[0057] Data transformation and time alignment convert all acquired data into time series format, unify the starting reference time point of all time series, unify the time frequency of all time series, and fill the missing values ​​of time series with low time frequency at new time points through linear interpolation or polynomial interpolation methods to ensure that all time series data have values ​​at the same time points.

[0058] Missing and outlier handling: Linear or polynomial interpolation methods are used to fill in null and missing values ​​in the time series. Outliers that do not conform to the expected pattern in the time series are identified by Z-Score or IQR methods. Outliers are removed and replaced with the mean or median of the time series or repaired by interpolation methods.

[0059] Data normalization, through min-max normalization, normalizes data of different dimensions. The formula for min-max normalization is as follows: ,in Represents the original data value. This represents the minimum value in a time series. This represents the maximum value in a time series. This represents the normalized data value;

[0060] Kalman filtering noise reduction sets corresponding state transition models based on the changing characteristics of different types of data, and then performs personalized noise reduction processing on each data item through Kalman filtering iteration.

[0061] It should be noted that Kalman filtering for data noise reduction is a mature existing technology, so it will not be discussed in detail here.

[0062] The slope displacement prediction model based on the optimized bidirectional LSTM model includes an input layer, a bidirectional LSTM layer, a temporal attention layer, a spatial attention layer, and an output layer. The input layer receives the input data. The bidirectional LSTM layer receives the data from the input layer and simultaneously uses forward and backward information from the time series to capture long-term dependencies in the time series. The temporal attention layer applies an attention mechanism to the temporal dimension of the sequence, i.e., it calculates the weight of each time step and performs a weighted summation of the output of the bidirectional LSTM layer. The spatial attention layer applies an attention mechanism to the spatial dimension of the sequence, i.e., it weights different feature channels of the sequence output by the temporal attention layer, highlighting important features and suppressing irrelevant features. The output layer receives the output of the spatial attention layer and generates the final prediction result.

[0063] The above steps, by combining a bidirectional LSTM model with a temporal and spatial dual attention mechanism and applying it to slope displacement prediction, not only fully explore and utilize the spatiotemporal features in the data, but also enable the slope displacement prediction model to focus on the time steps in historical data that are most valuable for current prediction, and highlight the features with the strongest correlation to displacement in multi-source data. This effectively improves the slope displacement prediction model's learning ability on complex spatiotemporal data, thereby improving the model's prediction accuracy and enhancing its ability to capture slope displacement change trends. This provides stronger support for slope stability assessment and disaster prevention, and indirectly improves the assessment accuracy and the rationality of monitoring scheme allocation of this intelligent slope monitoring method and system.

[0064] The specific steps for training the slope displacement prediction model and optimizing its hyperparameters using the Slug-Schefflera Optimization (SSA) algorithm during model training are as follows:

[0065] Historical surface monitoring data, historical underground monitoring data, and historical environmental monitoring data of the slope were acquired. All acquired data were preprocessed, and the preprocessed data were then proportionally analyzed. The slope displacement prediction model was initialized by dividing it into training, validation, and test sets.

[0066] The SSA optimization algorithm for salps is initialized, and the population size is set to 0. The maximum number of iterations is The search results range is The initial population position is randomly generated, and each individual represents a set of hyperparameter combinations, including learning rate, number of hidden layer nodes and temporal window size;

[0067] Using root mean square error As the fitness function, the error of the slope displacement prediction model on the validation set is calculated using the following formula:

[0068]

[0069] in Indicates the first in the verification set The actual value of the slope displacement Then it means the first Predicted values ​​of slope displacement. This represents the total number of data points in the validation set;

[0070] The leader and follower positions are updated according to the formula until the maximum number of iterations is reached or the fitness converges. At this point, the optimal individual position, i.e., the optimal hyperparameter combination, is output. The specific calculation formula is as follows:

[0071]

[0072]

[0073] in Indicates the updated leader position. This indicates the position of the individual with the best fitness value in the current population. Indicates the current iteration number. This indicates the generation of a value in the interval The random number within, on the left side of the formula Indicates the first The updated position of each follower, on the right side of the formula. Indicates the first The current location of each follower. Indicates the first The position preceding the individual follower;

[0074] The slope displacement prediction model is re-initialized using the optimal hyperparameter combination, then trained using the training set and tuned using the validation set, and finally evaluated using the test set.

[0075] The above steps optimize the hyperparameters of the slope displacement prediction model during model training using the Salicylia Optimization Algorithm (SSA). This not only efficiently searches for the optimal hyperparameter combination, thereby improving the convergence speed and generalization performance of the slope displacement prediction model and shortening its training time, but also effectively explores and develops globally by simulating the collective foraging behavior of salicylia, avoiding local optima. This improves the prediction accuracy and performance of the slope displacement prediction model, indirectly enhancing the evaluation accuracy and rationality of the intelligent slope monitoring method and system, as well as the allocation of monitoring schemes.

[0076] The specific steps for calculating the slope stability score based on the prediction results and the slope displacement distance in the latest sampling period are as follows:

[0077] The displacement distance of the slope in the latest sampling period was obtained from the preprocessed surface monitoring data. ;

[0078] Based on the slope displacement distance in the latest sampling period and the predicted displacement distance in the next sampling period The stability score of the slope was obtained through calculation. The specific calculation formula is as follows:

[0079]

[0080] in and Displacement distance and predicted displacement distance The weighting coefficients, and ,in This indicates the preset slope displacement risk threshold.

[0081] The above steps, based on the prediction results and the slope's displacement distance within the latest sampling period, calculate the slope's stability score. This not only reflects the slope's immediate deformation using the current actual displacement but also anticipates potential future displacement changes using predicted displacement. This comprehensive approach considers both the slope's real-time state and potential development trends, enabling better prediction of potential slope instability risks and a more comprehensive stability assessment. Furthermore, this integrated approach not only improves the accuracy of the assessment but also enhances the ability to predict changes in slope condition, providing more reliable data support for slope maintenance, management, and monitoring. This improves the accuracy of the intelligent slope monitoring method and system's assessment and the rationality of monitoring scheme allocation.

[0082] Example 2: A multi-source sensor fusion intelligent slope monitoring system, such as... Figure 2 As shown, it includes:

[0083] The data acquisition module is used to acquire surface monitoring data of the slope through GNSS monitoring station and MEMS tilt sensor, underground monitoring data of the slope through pore water pressure gauge and soil moisture sensor, and environmental monitoring data of the area where the slope is located through weighing rain gauge and temperature sensor.

[0084] The data processing module is used to preprocess the acquired surface monitoring data, underground monitoring data and environmental monitoring data to obtain preprocessed surface monitoring data, underground monitoring data and environmental monitoring data;

[0085] The model building module is used to build a slope displacement prediction model based on an optimized bidirectional LSTM model, train the slope displacement prediction model, and optimize the hyperparameters of the slope displacement prediction model through the Saurus Elixir Algorithm (SSA) during model training.

[0086] The state assessment module is used to input preprocessed surface monitoring data, underground monitoring data and environmental monitoring data into the trained slope displacement prediction model to obtain prediction results. The prediction results are the predicted displacement distance of the slope in the next sampling period. Based on the prediction results and the displacement distance of the slope in the latest sampling period, the stability score of the slope is calculated.

[0087] The scheme allocation module is used to classify slopes into corresponding risk levels based on stability scores and preset grading rules, and to allocate corresponding monitoring schemes to slopes based on risk levels.

[0088] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A multi-source sensor fusion intelligent monitoring method for slopes, characterized in that, Includes the following steps: Surface monitoring data of the slope is obtained through GNSS monitoring stations and MEMS tilt sensors; underground monitoring data of the slope is obtained through pore water pressure gauges and soil moisture sensors; and environmental monitoring data of the area where the slope is located is obtained through weighing rain gauges and temperature sensors. The acquired surface monitoring data, subsurface monitoring data, and environmental monitoring data are preprocessed to obtain preprocessed surface monitoring data, subsurface monitoring data, and environmental monitoring data; A slope displacement prediction model based on an optimized bidirectional LSTM model is constructed. The slope displacement prediction model is trained, and the hyperparameters of the slope displacement prediction model are optimized during the model training period using the Saurus Aquatic Optimization (SSA) algorithm. The preprocessed surface monitoring data, underground monitoring data, and environmental monitoring data are input into the trained slope displacement prediction model to obtain the prediction results. The prediction results are the predicted displacement distance of the slope in the next sampling period. Based on the prediction results and the displacement distance of the slope in the latest sampling period, the stability score of the slope is calculated. Based on stability scores and according to preset grading rules, slopes are classified into corresponding risk levels, and corresponding monitoring schemes are assigned to slopes based on risk levels.

2. The intelligent slope monitoring method based on multi-source sensor fusion according to claim 1, characterized in that, The surface monitoring data includes slope displacement and dip angle changes; the underground monitoring data includes slope pore water pressure and soil moisture content changes; and the environmental monitoring data includes rainfall and temperature changes in the area where the slope is located.

3. The intelligent slope monitoring method based on multi-source sensor fusion according to claim 2, characterized in that, The specific steps for preprocessing the acquired surface monitoring data, subsurface monitoring data, and environmental monitoring data are as follows: Data transformation and time alignment convert all acquired data into time series format, unify the starting reference time point of all time series, unify the time frequency of all time series, and fill the missing values ​​of time series with low time frequency at new time points through linear interpolation or polynomial interpolation methods to ensure that all time series data have values ​​at the same time points. Missing and outlier handling: Linear or polynomial interpolation methods are used to fill in null and missing values ​​in the time series. Outliers that do not conform to the expected pattern in the time series are identified by Z-Score or IQR methods. Outliers are removed and replaced with the mean or median of the time series or repaired by interpolation methods. Data normalization, through min-max normalization, normalizes data of different dimensions. The formula for min-max normalization is as follows: ,in Represents the original data value. This represents the minimum value in a time series. This represents the maximum value in a time series. This represents the normalized data value; Kalman filtering noise reduction sets corresponding state transition models based on the changing characteristics of different types of data, and then performs personalized noise reduction processing on each data item through Kalman filtering iteration.

4. The intelligent slope monitoring method based on multi-source sensor fusion according to claim 3, characterized in that, The slope displacement prediction model based on the optimized bidirectional LSTM model includes an input layer, a bidirectional LSTM layer, a temporal attention layer, a spatial attention layer, and an output layer. The input layer is used to receive input data. The bidirectional LSTM layer is used to receive data from the input layer and simultaneously use the forward and backward information of the time series to capture long-term dependencies in the time series. The temporal attention layer is used to apply the attention mechanism in the temporal dimension of the sequence, that is, to calculate the weight of each time step and perform a weighted summation of the output of the bidirectional LSTM layer; The spatial attention layer is used to apply the attention mechanism in the spatial dimension of the sequence, that is, to weight the different feature channels of the sequence output by the temporal attention layer, highlighting important features and suppressing irrelevant features; The output layer receives the output of the spatial attention layer and generates the final prediction result.

5. The intelligent slope monitoring method based on multi-source sensor fusion according to claim 4, characterized in that, The specific steps for training the slope displacement prediction model and optimizing its hyperparameters using the Slug-Schefflera Optimization (SSA) algorithm during model training are as follows: Historical surface monitoring data, historical underground monitoring data, and historical environmental monitoring data of the slope were acquired. All acquired data were preprocessed, and the preprocessed data were then proportionally analyzed. The slope displacement prediction model was initialized by dividing it into training, validation, and test sets. The SSA optimization algorithm for salps is initialized, and the population size is set to 0. The maximum number of iterations is The search results range is The initial population position is randomly generated, and each individual represents a set of hyperparameter combinations, including learning rate, number of hidden layer nodes and temporal window size; Using root mean square error As the fitness function, the error of the slope displacement prediction model on the validation set is calculated using the following formula: ； in Indicates the first in the verification set The actual value of the slope displacement Then it means the first Predicted values ​​of slope displacement. This represents the total number of data points in the validation set; The leader and follower positions are updated according to the formula until the maximum number of iterations is reached or the fitness converges. At this point, the optimal individual position, i.e., the optimal hyperparameter combination, is output. The specific calculation formula is as follows: ； ； in Indicates the updated leader position. This indicates the position of the individual with the best fitness value in the current population. Indicates the current iteration number. This indicates the generation of a value in the interval The random number within, on the left side of the formula Indicates the first The updated position of each follower, on the right side of the formula. Indicates the first The current location of each follower. Indicates the first The position preceding the individual follower; The slope displacement prediction model is re-initialized using the optimal hyperparameter combination, then trained using the training set and tuned using the validation set, and finally evaluated using the test set.

6. The intelligent slope monitoring method based on multi-source sensor fusion according to claim 5, characterized in that, The specific steps for calculating the slope stability score based on the prediction results and the slope displacement distance in the latest sampling period are as follows: The displacement distance of the slope in the latest sampling period was obtained from the preprocessed surface monitoring data. ; Based on the slope displacement distance in the latest sampling period and the predicted displacement distance in the next sampling period The stability score of the slope was obtained through calculation. The specific calculation formula is as follows: ； in and Displacement distance and predicted displacement distance The weighting coefficients, and ,in This indicates the preset slope displacement risk threshold.

7. A multi-source sensor fusion-based intelligent slope monitoring system, applied to the multi-source sensor fusion-based intelligent slope monitoring method described in any one of claims 1-6, characterized in that, Including: The data acquisition module is used to acquire surface monitoring data of the slope through GNSS monitoring station and MEMS tilt sensor, underground monitoring data of the slope through pore water pressure gauge and soil moisture sensor, and environmental monitoring data of the area where the slope is located through weighing rain gauge and temperature sensor. The data processing module is used to preprocess the acquired surface monitoring data, underground monitoring data and environmental monitoring data to obtain preprocessed surface monitoring data, underground monitoring data and environmental monitoring data; The model building module is used to build a slope displacement prediction model based on an optimized bidirectional LSTM model, train the slope displacement prediction model, and optimize the hyperparameters of the slope displacement prediction model through the Saurus Elixir Algorithm (SSA) during model training. The state assessment module is used to input preprocessed surface monitoring data, underground monitoring data and environmental monitoring data into the trained slope displacement prediction model to obtain prediction results. The prediction results are the predicted displacement distance of the slope in the next sampling period. Based on the prediction results and the displacement distance of the slope in the latest sampling period, the stability score of the slope is calculated. The scheme allocation module is used to classify slopes into corresponding risk levels based on stability scores and preset grading rules, and to allocate corresponding monitoring schemes to slopes based on risk levels.

Images