Dam landslide disease identification method and device based on artificial intelligence

Abstract

The invention relates to a dam landslide disease identification method and device based on artificial intelligence, and belongs to the field of dam landslide disease identification. The method comprises the following steps: collecting dam landslide data of an actual dam monitoring point, and carrying out manual marking to obtain a marked dam landslide data set; inputting the marked dam landslide data set into the trained generative adversarial network based on hierarchical reinforcement to obtain an expanded dam landslide data set; inputting the expanded dam landslide data set into a trained neural network based on adaptive convolution kernel optimization for feature extraction to obtain extracted dam landslide data features; and inputting the extracted dam landslide data features into a trained support vector machine algorithm based on high-order feature interpolation to obtain a dam landslide disease identification result. According to the invention, the precision and efficiency of feature extraction can be improved, so that different landslide levels can be accurately identified.

Description

Technical Field

[0001] The present invention belongs to the field of dam landslide disease identification, and particularly relates to an artificial intelligence-based dam landslide disease identification method and device. Background Art

[0002] With the frequent occurrence of global climate change and extreme weather events, the safety of dams has gradually become an important issue in the management of water conservancy infrastructure. Dam landslide is a relatively common and serious type of dam disease. If not identified and treated in time, it may pose a major threat to the surrounding ecological environment, the safety of residents' lives and property, and the stability of the dam itself. Existing methods for dam landslide disease identification technology include: Landslide identification method based on physically constrained machine learning: First, collect slope data and save it. Embed physical constraints such as the dynamic equation of erosive landslide fluidity, boundary conditions, and initial conditions into neural network learning to obtain a prediction model. Then, collect data of the slope to be measured and use the model to predict its landslide dynamic process, so that the model can better understand the physical meaning of landslides for accurate prediction. Risk assessment method for bedding landslide disasters based on machine learning: Obtain regional topographic and geomorphic image data to identify potential bedding landslide points and construct a sample library. Use machine learning to identify potential hazards in unknown areas to obtain geological disaster data, and solve the problem of low existing assessment accuracy. Landslide susceptibility assessment method based on CNN - DF: Conduct a multicollinearity test and frequency ratio calculation on landslide influencing factors, screen disaster - inducing environmental factors to construct a similar environment and data set, build a model based on CNN module and DF module for landslide susceptibility prediction to generate a susceptibility map, and finally evaluate the model performance.

[0003] In the task of dam landslide disease identification, the above technical solutions still have the following problems to be further solved: 1. Dam landslide disease identification often relies on limited historical data or a single data source, resulting in insufficient training samples and affecting the generalization ability and accuracy of the model. 2. Traditional feature extraction methods ignore the temporal, multi - scale, and complex spatial distribution relationships of dam landslide data, easily causing information loss or redundancy, and requiring a large amount of artificial feature engineering when dealing with high - dimensional, non - linear, and noisy data. 3. The classification accuracy of dam landslide identification is not high, especially it is difficult to distinguish different landslide grades such as minor and moderate landslides. 4. Some models perform poorly on different data sets. Facing the complex and changeable dam environment, traditional models are difficult to adapt to the new data distribution. Summary of the Invention

[0004] In order to solve the above problems, the present invention provides an artificial intelligence - based dam landslide disease identification method and device.

[0005] To achieve the above object, the present invention is realized through the following technical solutions: The present invention provides an artificial intelligence - based dam landslide disease identification method, including the following steps: S1. Collect the dam landslide data of the actual dam monitoring points and perform manual annotation to obtain the labeled dam landslide data set; S2. Input the labeled dam landslide data set into the trained hierarchical reinforcement-based generative adversarial network for data augmentation to obtain the augmented dam landslide data set; the hierarchical reinforcement-based generative adversarial network includes a generator and a discriminator; a hierarchical reinforcement learning mechanism is adopted in the generator, and the quality of the generated samples is optimized through a multi-layer decision-making process; S3. Input the augmented dam landslide data set into the trained neural network optimized based on adaptive convolution kernels for feature extraction to obtain the extracted dam landslide data features; the neural network optimized based on adaptive convolution kernels includes an input layer, a convolution layer, a pooling layer, and a fully connected layer; the convolution layer performs residual connections to form residual blocks, and each residual block performs skip connections, and the skip connections are the conventional residual neural network connection methods in the art; S4. Input the extracted dam landslide data features into the trained support vector machine algorithm based on high-order feature interpolation for dam landslide disease identification and classification to obtain the dam landslide disease identification result.

[0006] Further, step S1 specifically includes: The dam landslide data includes the geographical location Ra, the current temperature Da, the humidity value Ia, the precipitation Pa around the dam, the water level Ha, the displacement Va, the seepage pressure Sa, the crack size Fa occurring on the dam surface, the cumulative rainfall La, and the historical disease record Ca; The annotation categories of the dam landslide data are four categories: "normal", "slight landslide", "moderate landslide", and "severe landslide".

[0007] Further, step S2 specifically includes: The training process of the hierarchical reinforcement-based generative adversarial network is as follows: S21. Initialize the network parameters of the generator and the discriminator. The initialization strategy is initialized by using a normal distribution with a mean of 0 and a variance of the identity matrix. The parameter initialization formula is expressed as follows: , , where, represents the parameters of the generator, represents the parameters of the discriminator, represents being subject to a specific distribution, represents a normal distribution with a mean of 0 and a variance of the identity matrix, represents the identity matrix; S22. Pre-train the discriminator. The loss function formula of the discriminator in the pre-training stage is expressed as follows: , Among them, represents the real dam landslide data, is the loss function of the discriminator, represents the input noise sampled from the noise distribution, represents the noise distribution, represents the expectation, represents the discriminator function, represents the generator function, represents the distribution of the real dam landslide data, represents the natural logarithm; S23. In the adversarial training stage, first, fix the generator parameters, and input the real dam landslide data and the generated dam landslide data output by the generator into the discriminator for processing, and update the discriminator parameters. The generated dam landslide data output by the generator is the data obtained by inputting the noise sampled from the noise distribution into the generator; then, fix the discriminator parameters and update the generator parameters. The generator adjusts its generation strategy through a hierarchical reinforcement learning strategy; the loss function formula of the generator is expressed as follows: Among them, represents the loss function of the generator, represents the weight of the combined embedding loss, represents the combined embedding loss; the formula of the combined embedding loss is expressed as follows: , , , Among them, represents the number of generated samples in the current batch, represents the feature extraction function of the neural roll network layer of the generator, represents the non-linear transformation function of the feature extraction function for the th feature dimension, represents the feature dimension of the generated dam landslide data; represents the first weight parameter, represents the second weight parameter, represents the boundary parameter for controlling the balance of the loss; represents the L2 norm, represents taking the maximum value, represents the th real dam landslide data, represents the th input noise; At each layer of the generator, it is independently adjusted through reinforcement learning during training. During the training process of the generator, the reward for each adjustment is calculated based on the feedback of the discriminator and the embedding loss, and the formula is as follows: , , where, represents the Sigmoid function, represents the hierarchical reinforcement learning rate parameter, represents the reward for the adjustment of the neural network layer of the generator; S25. Repeat steps S22 - S24 until the preset stop iteration condition is met, complete the training, and obtain the trained hierarchical reinforcement-based generative adversarial network.

[0008] Furthermore, step S3 specifically includes: The training process of the neural network based on the optimized adaptive convolution kernel is as follows: S31. In the initial stage of training, hierarchical processing is performed on the data in the augmented dam landslide dataset by grouping the dam landslide data according to different categories and allocating them to different sub-networks. The formula is as follows: , , where, represents the output of the dam landslide data processed by the th layer of the neural network, represents the weight parameter of the th layer of the neural network, represents the feature matrix of the input dam landslide data of the neural network, represents the activation function of the neural network, represents the weight of the th feature in the th layer of the neural network, represents the feature dimension of the dam landslide data input to the neural network, represents the th feature of the dam landslide data input to the neural network, represents the bias of the th layer of the neural network, represents the Sigmoid activation function; the adaptive update of the neural network parameters is performed using the gradient descent method, and the formula for parameter update is as follows: , , where, represents the learning rate of the neural network, Represents the gradient of the loss function of the neural network with respect to the weights of the th layer, Represents the gradient of the loss function of the neural network with respect to the bias of the th layer, Represents the loss function of the neural network, Represents the parameter update operation; S32. In the initialization stage of the adaptive convolution kernel, the convolutional layer automatically selects the most suitable convolution kernel size and stride according to the input dam landslide data, and optimizes the parameters of the convolution kernel through a regularization strategy. The formula is as follows: , where, Represents the output features of the convolutional layer, Represents the weights of the adaptive convolution kernel, Represents the convolution operation, Represents the bias of the convolutional layer; Parameter solving is carried out according to the regularization strategy of minimizing the objective error to complete the adaptive optimization of the initial value of the convolution kernel. The formula is as follows: , where, Is the parameter update operation, Represents the regularization coefficient, Represents the Frobenius norm; The gradient descent method is used to update the convolution kernel so that the convolution kernel can adapt to the dam landslide data distribution in real time during training. The formula is as follows: , where, Represents the loss function of the weights of the adaptive convolution kernel; S33. Adopt a local attention mechanism to let the model focus on the local regions with key discriminability in the input dam landslide data, suppress redundancy and noise interference by weighting the features, and achieve effective feature extraction. The formula is as follows: , where, Represents the feature importance score, Represents the attention weight, Represents the attention bias; The local region is weighted according to the Softmax function to convert the feature importance score into an attention weight. The formula is as follows: , where, Represents the local attention weight, Represents the th feature corresponding feature importance score, Indicates the total number of samples input to the neural network in the current batch; S34. In the multi-scale feature fusion stage, according to the local attention weights obtained in step S33, the features extracted by different convolutional kernels are weighted and superimposed, and the formula is as follows: , , , Among them, represents the fused multi-scale features, represents the th weight coefficient of the convolutional scale, represents the th convolutional kernel weight, represents the set number of multi-scales, represents the th importance coefficient of the scale, represents the th complexity adjustment coefficient of the scale; S35. In the optimization process of the residual connection, by building skip connections between different layers, the vanishing gradient is alleviated and the transfer efficiency of deep features is enhanced. The update method of the convolutional kernel weights in the residual connection is as follows: , Among them, represents the convolutional kernel weights in the residual connection, represents the learning rate of the residual connection, represents the gradient of the loss function of the neural network with respect to the convolutional kernel weights in the residual connection; Incorporate the input features into the residual path and implement non-linear transformation, and complete the residual output according to the skip addition method. The formula is as follows: , Among them, represents the overall output features of the neural network, represents the bias of the residual connection; S36. Repeat iterations of S32 - S35 until the preset stop iteration condition is satisfied, and obtain the trained neural network optimized based on the adaptive convolutional kernel.

[0009] Furthermore, step S4 specifically includes: The training process of the support vector machine algorithm based on high-order feature interpolation is as follows: S41. Initialize the phase space mapping parameters, define the initial phase space mapping, adapt to the dam landslide data features after different feature extractions through a dynamically adjusted strategy, and define the kernel function of the support vector machine. The formula is as follows: , , Among them, represents the kernel function of the extreme learning machine, represents the th data point in the input space, represents the th data point in the input space, represents the width parameter of the kernel function of the extreme learning machine, represents the L2 norm, represents the adaptive adjustment coefficient, is the variance of the data samples input to the extreme learning machine, representing the global variance; the adaptive adjustment coefficient is automatically adjusted according to the dispersion degree of the extracted dam landslide data features, and the formula is as follows: , Among them, represents the adaptive adjustment coefficient, represents the number of samples input to the extreme learning machine in the current batch, represents the th data point in the input space, represents the mean value of all training samples input to the extreme learning machine in the current batch; S42. Transform the original feature space into a higher-dimensional feature space through the high-order feature interpolation method. The interpolation of high-order features is achieved through the following polynomial mapping: , Among them, represents the function that maps the data input to the extreme learning machine to the high-dimensional space, represents the th data point in the input space; represents the highest order of the polynomial; S43. Perform phase space dynamic adjustment, automatically adjust the structure of the mapping function according to the distribution characteristics of the training data, and then optimize the classification boundary. Dynamically adjust the structure of the mapping function based on the distribution of the training data, and use the loss function to optimize the mapping parameters. The formula is as follows: , Among them, represents the loss function of the extreme learning machine, represents the class label of the th data point, represents the kernel function of the extreme learning machine, represents the width parameter value of the kernel function of the extreme learning machine in the previous iteration, represents the maximum value function; S44. Optimization training of the support vector machine After obtaining a suitable high-dimensional feature space, by adjusting the weight and bias parameters of the support vector machine, the objective function is minimized, and the formula is as follows: , wherein, is the objective function of the support vector machine, represents making the parameters and achieve the minimization constraint; The constraint conditions of the objective function are: wherein, represents the weight vector of the support vector machine, represents the bias term of the support vector machine; represents the regularization parameter of the support vector machine, controlling the trade-off between the error term and the model complexity; represents the parameter that the objective function affects, used to adjust the influence of the kernel function in the optimization process; represents the th slack variable, represents the th slack variable; S45. The stability of the algorithm is evaluated through cross-validation and stability indicators, and the formula is as follows: , wherein, represents the stability indicator, represents the number of folds of cross-validation, represents the accuracy of the th fold of cross-validation.

[0010] Furthermore, the weight of the combined embedding loss in step S23 is set to 0.3, and the boundary parameter is set to 0.2; the hierarchical reinforcement learning rate parameter in step S24 is set to 0.3; the learning rate of the neural network in step S31 is set to 0.03; the regularization coefficient in step S32 is set to 0.3; the attention bias in step S33 is set to 0.3; the learning rate of the residual connection in step S35 is set to 0.03; the highest order of the polynomial in step S42 is set to 6; the regularization parameter of the support vector machine in step S44 is set to 1; the number of folds of cross-validation in step S45 is set to 10.

[0011] Furthermore, the preset stop iteration condition is to reach the preset maximum number of iterations, and the preset maximum number of iterations is set to 1000 times.

[0012] The present invention also provides an artificial intelligence-based dike landslide disease identification device, which executes the artificial intelligence-based dike landslide disease identification method, including: Sample collection module: It is used to collect the dam landslide data of the actual dam monitoring points and perform manual annotation to obtain the labeled dam landslide data set; Sample augmentation module: It is used to input the labeled dam landslide data set into the trained hierarchical reinforcement-based generative adversarial network for data augmentation to obtain the augmented dam landslide data set; The hierarchical reinforcement-based generative adversarial network includes a generator and a discriminator; A hierarchical reinforcement learning mechanism is adopted in the generator, and the quality of the generated samples is optimized through a multi-layer decision-making process; Feature extraction module: It is used to input the augmented dam landslide data set into the trained neural network optimized based on adaptive convolution kernels for feature extraction to obtain the extracted dam landslide data features; The neural network optimized based on adaptive convolution kernels includes an input layer, a convolutional layer, a pooling layer, and a fully connected layer; The convolutional layer forms residual blocks, and each residual block performs skip connections, and the skip connections are the conventional residual neural network connection methods in the art; Dam landslide disease identification module: It is used to input the extracted dam landslide data features into the trained support vector machine algorithm based on high-order feature interpolation for dam landslide disease identification and classification to obtain the dam landslide disease identification result.

[0013] The advantages of the present invention are as follows: The present invention adopts a data augmentation method based on a hierarchical reinforcement generative adversarial network. The generator uses a reinforcement learning mechanism to optimize the quality of the generated samples. The discriminator effectively improves the diversity and authenticity of the generated data through adversarial training. The parameters of the generator and the discriminator are dynamically optimized, making the generated data more practical, improving the diversity of samples and the generalization ability of the training model; Adopt a neural network algorithm optimized based on adaptive convolution kernels for feature extraction. By selecting the optimal convolution kernel size and stride, the ability to capture multi-scale features is enhanced. At the same time, a local attention mechanism is adopted to further focus on the key areas in the dam landslide data, suppress noise interference, and improve the accuracy and efficiency of feature extraction; Adopt a support vector machine algorithm based on high-order feature interpolation to map the dam landslide data to a higher-dimensional feature space, thereby enhancing the model's ability to express non-linear relationships. At the same time, by dynamically adjusting the phase space mapping structure, the classification boundary is optimized, making the model more accurate when identifying different landslide levels; Adopt a strengthened training strategy and model evaluation mechanism. An adaptive strategy is introduced during the training process, and technical means such as combined embedding loss, margin loss, and residual connection are used to optimize the feature learning process. At the same time, cross-validation and stability indicators are used to evaluate the performance of the model on different data subsets to ensure the effectiveness and stability of the algorithm in practical applications Description of the Drawings The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention, but do not constitute a limitation to the present invention.

[0014] Figure 1 is the flowchart of the steps of the present invention; Figure 2 is the comparison chart of the discriminator recognition accuracy between the method of the present invention and the conventional method; Figure 3 is the comparison chart of the computational complexity between using the local attention mechanism and without local attention; Figure 4 is the comparison of the classification accuracy between the method of the present invention and other methods. Detailed implementation manners

[0015] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0016] Embodiment 1 In this embodiment, as Figure 1 shown, the present invention provides an artificial intelligence-based method for identifying dam landslide diseases, and the specific steps include: S1. Collect the dam landslide data of the actual dam monitoring points and perform manual annotation to obtain the labeled dam landslide data set; Specifically, the dam landslide data includes geographical location Ra, current temperature Da, humidity value Ia, precipitation Pa around the dam, water level Ha, displacement Va, seepage pressure Sa, crack size Fa occurring on the dam surface, cumulative rainfall La, and historical disease records Ca; the data collection method is mainly realized by automatic collection of sensors and manual calibration. The data of each monitoring point is regularly collected by sensors or remote sensing devices and uploaded to the cloud storage platform at high frequency to ensure the integrity and real-time nature of the data; the storage format of the data uniformly adopts a standardized data table format, including CSV or JSON format.

[0017] It should be noted that this embodiment is only to illustrate a data format and type of the present invention. In actual applications, the attributes of the data are usually more than 10 attributes, and the number of data attributes may reach dozens or even hundreds.

[0018] The annotation categories of the dam landslide data are four categories: "normal", "slight landslide", "moderate landslide", and "severe landslide".

[0019] S2. The labeled dam landslide dataset is input into the trained hierarchical reinforcement-based generative adversarial network for data augmentation to obtain the augmented dam landslide dataset; the hierarchical reinforcement-based generative adversarial network includes a generator and a discriminator; a hierarchical reinforcement learning mechanism is adopted in the generator to optimize the quality of the generated samples through a multi-layer decision-making process; Specifically, the training process of the hierarchical reinforcement-based generative adversarial network is as follows: S21. Initialize the network parameters of the generator and the discriminator. The generated dam landslide data must have a certain degree of diversity and realism. The initialization strategy is to initialize using a normal distribution with a mean of 0 and a variance of the identity matrix. The generator can generate diverse fake data at the beginning stage and provide it to the discriminator for discrimination, so as to gradually improve the quality of the generated data through adversarial training. For the generator and the discriminator, the parameter initialization formula is as follows: , , where, represents the parameters of the generator, represents the parameters of the discriminator, represents being subject to a specific distribution, represents a normal distribution with a mean of 0 and a variance of the identity matrix, represents the identity matrix; S22. To alleviate the imbalance between the generator and the discriminator in the initial stage of training, pre-train the discriminator so that it can distinguish between real dam landslide data and generated dam landslide data to a certain extent. In the dam landslide data augmentation task, the discriminator needs to be sensitive to different attributes (such as water level, displacement, crack size, etc.). Especially when a dam landslide or disease occurs, the changes in these data attributes before and after the landslide are usually significant. For example, before a landslide, phenomena such as changes in water level, increase in precipitation, and expansion of cracks may be precursors. Through pre-training, the discriminator can better identify these precursor features. In the pre-training stage, the discriminator not only learns how to distinguish between real dam landslide data and generated data, but also needs to be able to identify the evolution laws of attributes such as water level, displacement, and precipitation over time. Through pre-training, the discriminator can make accurate judgments on the key features (such as the correlation between cumulative rainfall and disease records) in the dam landslide data at the initial stage, and then provide more effective feedback to the generator to ensure that the generated data is more realistic. The loss function formula of the discriminator in the pre-training stage is as follows: , where, represents real dam landslide data, is the loss function of the discriminator, represents the input noise sampled from the noise distribution, represents the noise distribution, represents the expectation, represents the discriminator function, represents the generator function, represents the distribution of real dam landslide data, represents the natural logarithm; S23. In the adversarial training phase, first, fix the generator parameters, input the real dam landslide data and the generated dam landslide data output by the generator into the discriminator for processing, and update the discriminator parameters. The generated dam landslide data output by the generator is the data obtained by inputting the noise sampled from the noise distribution into the generator; the optimization objective of this process is to maximize the probability that the real dam landslide data is correctly classified, while minimizing the probability that the generated dam landslide data is misclassified. The generator not only needs to learn how to generate the basic state data of the dam (such as temperature, humidity, water level, displacement, etc.), but also needs to understand the internal relationships between various data attributes. For example, the water level and displacement may affect each other, and the increase in precipitation will directly affect the crack size and seepage pressure on the dam surface. In this case, the generator needs to be able to generate dam data with reasonable correlations so that the discriminator cannot distinguish between real and fake data; Then, fix the discriminator parameters and update the generator parameters. The generator adjusts its generation strategy through a hierarchical reinforcement learning strategy, making the discriminator misclassify the generated dam landslide data as real dam landslide data, while minimizing the combined embedding loss to improve the quality and diversity of the generated dam landslide data. In the dam landslide data augmentation task, in this process, the objective of the discriminator is to maximize the recognition accuracy of real data, especially in the relationship between the historical disease records of the dam (such as whether there are diseases) and important features such as displacement and crack size. The discriminator will learn how to accurately identify the complex dependencies between these data through training, and then help the generator adjust the generation strategy to make the generated data more realistic; the loss function formula of the generator is expressed as follows: , where, represents the loss function of the generator, represents the weight of the combined embedding loss, is set to 0.3, Denote the combined embedding loss; in the data augmentation of dam landslides, the introduction of the combined embedding loss can help the generator maintain the diversity of the generated data, while avoiding the over-concentration or overfitting of the generated dam data. For dam landslide data, features such as water level (Ha), displacement amount (Va), and crack size (Fa) have temporal and spatial distribution characteristics. The combined embedding loss can help the generator maintain the diversity and rationality of these data attributes when generating data. Through the combined embedding loss, the generator can learn how to balance the relationships between these attributes, so that the generated data not only maintains consistency in the feature space but also adapts to different actual situations. The formula of the combined embedding loss is expressed as follows: , where, denotes the number of generated samples in the current batch, denotes the feature extraction function of the neural rolling network layer of the generator, denotes the first weight parameter, denotes the second weight parameter, denotes the boundary parameter, which is used to control the balance of the loss; denotes the L2 norm, denotes taking the maximum value, denotes the th real dam landslide data, denotes the th input noise; Specifically, the calculation method of the squared loss term in the combined embedding loss can be expressed as: , where, denotes the non-linear transformation function of the feature extraction function for the th feature dimension, denotes the feature dimension of the generated dam landslide data; Specifically, the margin loss encourages the distance between the generated features and the real features to be at least by setting the boundary, so as to avoid the model overfitting the small batch of dam landslide training data. For dam landslide data, especially the historical disease records of dams and some extreme values (such as abnormal large displacement amounts or crack sizes) are of special importance. The margin loss can ensure that the data generated by the generator avoids extreme values that deviate too much from the real distribution. For example, there may be a certain boundary relationship between the historical disease record (Ca) of the dam and the precipitation (Pa) or the cumulative rainfall (La). The margin loss can ensure that the generator does not generate too many or too few disease record data and avoid generating unreasonable disease labels, thereby improving the authenticity of the dataset. The calculation method of the margin loss term is expressed as: , Among them, is the boundary parameter. Preferably, is set to 0.2; Each layer of the generator is independently adjusted through reinforcement learning during training. During the training process of the generator, the reward for each adjustment is calculated based on the feedback of the discriminator and the embedding loss. The formula is as follows: , , Among them, represents the Sigmoid function, represents the hierarchical reinforcement learning rate parameter, is set to 0.3, represents the reward for the adjustment of the neural network layer of the generator; S25. Repeat steps S22 - S24 until the preset stop iteration condition is met, complete the training, and obtain the trained generative adversarial network based on hierarchical reinforcement. The preset stop iteration condition is to reach the preset maximum number of iterations. Preferably, the preset maximum number of iterations is set to 1000 times.

[0020] S3. Input the expanded dam landslide dataset into the trained neural network optimized based on adaptive convolution kernels for feature extraction to obtain the dam landslide data features after extraction; the neural network optimized based on adaptive convolution kernels includes an input layer, a convolutional layer, a pooling layer, and a fully connected layer; the convolutional layer performs residual connections to form residual blocks, and each residual block performs skip connections, and the skip connection is a conventional residual neural network connection method in the art. The construction method of the residual block is composed of 3 convolutional layers, and each convolutional layer is followed by an activation function (such as ReLU) to extract features and introduce non-linearity. Each convolutional layer processes the input data to extract specific feature information, and the activation function enhances the expressive ability of the model; in each residual block, the input data is directly passed to the output part of the residual block, rather than obtaining the output only through the calculation result of the convolutional layer. Specifically, after the input data is processed by multiple convolutional layers, it is added to the original input data to form a "residual" output. That is, the network not only learns the mapping from the input to the output, but also learns the difference (residual) between the input and the output. The skip connection method between multiple residual blocks is that if there are multiple residual blocks in the network, these residual blocks are connected by skip connections. For example, the output of the first residual block can be directly connected to the input of the second residual block, rather than only passing the output after convolutional calculation. Each residual block uses skip connections to make information flow more smoothly from the bottom layer to the top layer of the network, thereby avoiding the problem of performance degradation during the training of deep neural networks. To address the imbalance problem caused by different types of features (such as high-dimensional, non-linear, and noise) in complex dam landslide data, a neural network algorithm optimized based on adaptive convolution kernels is used for feature extraction. The optimal convolution kernel size and stride are selected through a regularization strategy to enhance the ability to capture multi-scale features. In addition, to address the problem of redundant information interference, a local attention mechanism is used to focus on key regions through feature weighting and suppress noise interference to improve the efficiency of feature extraction.

[0021] Specifically, the training process of the neural network optimized based on adaptive convolution kernels is as follows: S31. In the initial stage of training, in the task of dam landslide data feature extraction, the dam landslide data involves multiple different types of features. Hierarchical processing is performed on the data in the expanded dam landslide dataset to address the imbalance problem caused by different types of features (including high-dimensional, non-linear, and noise, etc.) in the dam landslide data. By grouping the dam landslide data according to different categories and allocating them to different sub-networks, the focusing ability on the features of various types of dam landslide data is strengthened, and the redundant interference caused by feature mixing is reduced. The formula is as follows: , where, represents the Output of the embankment landslide data after layer processing represents the weight parameter of the th layer of the neural network, represents the feature matrix of the input embankment landslide data of the neural network, represents the activation function of the neural network.

[0022] Specifically, in order to enable the processing of each layer of the neural network to adaptively optimize different embankment landslide data features, some features change drastically in specific time periods or regions, while other features change less. To handle this imbalance, a hierarchical processing method is adopted. The data is grouped according to different categories of its features and assigned to different sub-networks for processing. For example, for the strong correlation between water level (Ha) and displacement (Va), a sub-network is assigned in the network to specifically process the relationship between these two features, while another sub-network can process the relationship between humidity (Ia) and temperature (Da). This can reduce feature mixing and redundant interference, improve the network's focusing ability on each category of features, and thus improve efficiency and accuracy in feature extraction. During this process, the calculation method of the activation function of the neural network is expressed as , where, represents the weight of the th feature in the th layer of the neural network, represents the feature dimension of the embankment landslide data input into the neural network, represents the th feature of the embankment landslide data input into the neural network, represents the bias of the th layer of the neural network, represents the Sigmoid activation function; to achieve dynamic optimization of the parameters between layers of the neural network, the gradient descent method is used for adaptive update of the neural network parameters. For embankment landslide data, the non-linear relationship between features is particularly significant. For example, the historical disease records (Ca) of the embankment may have a complex non-linear relationship with the surrounding precipitation (Pa) and cumulative rainfall (La). Through continuous iteration and the optimization of the adaptive activation function, it can help the neural network better capture these non-linear features, improve the model's expressive ability, enable the model to more accurately learn the deep relationships between different features, and enhance the effect of feature extraction. During this process, the formula for updating the neural network parameters is as follows: , , where, represents the learning rate of the neural network, represents the loss function of the neural network with respect to the The gradient of the weights of the layer represents the loss function of the neural network with respect to the bias of the layer represents the loss function of the neural network represents the parameter update operation; preferably, is set to 0.03

[0023] S32. In the initialization stage of the adaptive convolution kernel, the convolutional layer automatically selects the most suitable convolution kernel size and stride according to the input dam landslide data to enhance the ability to capture multi-scale information. By adaptively selecting the size and stride of the convolution kernel, the ability to capture multi-scale information in the dam landslide data can be enhanced. Dam landslide data often has multi-scale characteristics. For example, the change in water level (Ha) may be a small short-term fluctuation or a long-term gradual change, and the displacement (Va) may also show different scale patterns from local to global. By adaptively optimizing the convolution kernel, the most suitable data scale can be automatically selected, avoiding the problem that a fixed convolution kernel cannot adapt to the different scale characteristics of dam data. In addition, by using a regularization strategy to optimize the parameters of the convolution kernel, it can be ensured that the convolution operation adapts to the distribution of dam landslide data adaptively during the learning process, improving the quality of feature extraction. The formula is as follows: , where represents the output features of the convolutional layer represents the weights of the adaptive convolution kernel represents the convolution operation represents the bias of the convolutional layer; parameter solving is performed according to the regularization strategy of minimizing the objective error to complete the adaptive optimization of the initial value of the convolution kernel. The formula is as follows: , where is the parameter update operation represents the regularization coefficient represents the Frobenius norm; preferably, is set to 0.3; the gradient descent method is used to update the convolution kernel so that the convolution kernel adapts to the dam landslide data distribution in real time during training. The formula is as follows: , where represents the loss function of the weights of the adaptive convolution kernel S33. Adopt a local attention mechanism to enable the model to focus on the local regions with key discriminative power in the input dam landslide data. By weighting the features, redundant and noise interferences are suppressed to achieve effective feature extraction. In the dam landslide data processing task, features such as water level, temperature, and humidity are often interfered by external factors, such as sudden weather changes or equipment errors. These redundant or noisy data will affect the learning process of the model. Through the local attention mechanism, the neural network can focus on the feature regions that play a key role during the occurrence of dam landslides. For example, before a landslide occurs, some features such as a sharp increase in precipitation or a rapid rise in water level may be key warning signals. By weighting these regions, the network can suppress the noise in irrelevant regions, thereby improving the accuracy of feature extraction. The formula is as follows: , where, represents the feature importance score, represents the attention weight, represents the attention bias, is set to 0.3; According to the Softmax function, the weighting of the local region is realized, and the feature importance score is converted into the attention weight. The formula is as follows: , where, represents the local attention weight, represents the th feature importance score corresponding to the feature, represents the total number of samples input to the neural network in the current batch; S34. In the multi-scale feature fusion stage, according to the local attention weight obtained in step S33, the features extracted by different convolutional kernels are weighted and superimposed to simultaneously obtain local details and global structure information, so as to address the problem of uneven distribution of dam landslide data. The occurrence of dam landslides not only depends on the change of a certain feature, but is the combined effect of multiple factors. For example, a sharp change in water level may be accompanied by a significant expansion of cracks, and at the same time, it may also be accompanied by an increase in seepage pressure. Through multi-scale feature fusion, the model can comprehensively consider data changes at different scales, helping the model to understand the occurrence mechanism of dam landslides from both the global and local levels, thereby improving the integrity and accuracy of feature extraction. The formula is as follows: , where, represents the fused multi-scale feature, represents the weight coefficient of the th convolutional scale, represents the weight of the th convolutional kernel, Represents the set number of multi-scales; Specifically, in order to make the multi-scale fusion process more adaptive, the importance of each scale is dynamically allocated according to the training situation of the convolutional kernel, and the formula is expressed as follows: , where, represents the importance coefficient of the th scale; Specifically, the complexity of the convolutional kernel is characterized by the Frobenius norm, and then the importance coefficient of each scale is calculated. The formula is expressed as follows: , where, represents the complexity adjustment coefficient of the th scale; S35. In the optimization process of the residual connection, by building skip connections between different layers, the vanishing gradient is alleviated and the transmission efficiency of deep features is enhanced. In the process of feature extraction of dam landslide data, the problem of vanishing gradient in the deep network may be encountered, especially when dealing with high-dimensional features with complex dependencies. Through the residual connection, the neural network can directly transmit information between different layers, so as to ensure that key information will not be lost during the transmission process. For example, the long-distance dependence relationship between the historical disease records (Ca) and geographical location (Ra) of the dam can be effectively processed through the deep residual connection, ensuring that in the multi-layer network structure, information can be fully transmitted and fused, thereby improving the depth and breadth of the model in extracting the features of dam landslide data. The update method of the convolutional kernel weights in the residual connection is as follows: , where, represents the convolutional kernel weights in the residual connection, represents the learning rate of the residual connection, is set to 0.03, represents the gradient of the loss function of the neural network with respect to the convolutional kernel weights in the residual connection; Incorporate the input features into the residual path and implement non-linear transformation, and complete the residual output according to the skip addition method. The formula is expressed as follows: , where, represents the overall output features of the neural network, represents the bias of the residual connection; S36. Repeat iterations of S32 - S35 until the preset stop iteration condition is satisfied, and obtain the trained neural network optimized based on the adaptive convolutional kernel. The preset stop iteration condition is to reach the preset maximum number of iterations. Preferably, the preset maximum number of iterations is set to 1000 times.

[0024] S4. Input the extracted dam landslide data features into the trained support vector machine algorithm based on high - order feature interpolation for dam landslide disease identification and classification, and obtain the dam landslide disease identification result.

[0025] Specifically, the training process of the support vector machine algorithm based on high - order feature interpolation is as follows: S41. Initialize the phase - space mapping parameters, define the initial phase - space mapping, adapt to the dam landslide data features after different feature extractions through a dynamic adjustment strategy, and define the kernel function of the support vector machine. The formula is as follows: , , Among them, represents the kernel function of the extreme learning machine, represents the - th data point in the input space, represents the - th data point in the input space, represents the width parameter of the kernel function of the extreme learning machine, represents the L2 norm, represents the adaptive adjustment coefficient, is the variance of the data samples input into the extreme learning machine, representing the global variance; the adaptive adjustment coefficient is automatically adjusted according to the dispersion degree of the extracted dam landslide data features. The formula is as follows: , Among them, represents the adaptive adjustment coefficient, represents the number of samples input into the extreme learning machine in the current batch, represents the - th data point in the input space, represents the mean value of all training samples input into the extreme learning machine in the current batch; S42. Transform the original feature space into a higher - dimensional feature space through the high - order feature interpolation method, which can not only enhance the model's ability to capture non - linear relationships, but also retain the important information of the dam landslide data after the original feature extraction. The interpolation of high - order features is achieved through the following polynomial mapping: , Among them, represents the function that maps the data input into the extreme learning machine to the high - dimensional space, represents the - th data point in the input space; represents the highest order of the polynomial, Set to 6; The specific implementation method of the function that maps data to a high-dimensional space includes the combination of cross terms and high-order terms, and the calculation method is expressed as: , In the formula, is the th data point in the th input space, is the th data point in the th input space.

[0026] S43. Perform phase space dynamic adjustment, automatically adjust the structure of the mapping function according to the distribution characteristics of the training data, and then optimize the classification boundary. Dynamically adjust the structure of the mapping function based on the distribution of the training data, and use the loss function to optimize the mapping parameters. The formula is expressed as follows: , where, represents the loss function of the extreme learning machine, represents the class label of the th data point. For example, 0, 1, 2, and 3 are respectively used to correspond to the four categories of "normal", "slight landslide", "moderate landslide", and "severe landslide"; represents the kernel function of the extreme learning machine, represents the width parameter value of the kernel function of the extreme learning machine in the previous iteration, represents the maximum value function; S44. Optimization training of the support vector machine After obtaining a suitable high-dimensional feature space, by adjusting the weight and bias parameters of the support vector machine, minimize the objective function. The formula is expressed as follows: , where, is the objective function of the support vector machine, represents making the parameters and achieve the minimum constraint; The constraint condition of the objective function is: where, represents the weight vector of the support vector machine, represents the bias term of the support vector machine; represents the regularization parameter of the support vector machine, which controls the trade-off between the error term and the model complexity; represents the objective function influence parameter, which is used to adjust the influence of the kernel function in the optimization process; represents the th slack variable, represents the th slack variable; Preferably, Set to 1.

[0027] S45. Evaluate the stability of the algorithm through cross - validation and stability indicators, and the formula is as follows: , Where, represents the stability indicator, represents the number of folds of cross - validation, represents the accuracy of the -th fold of cross - validation, which is used to evaluate the performance of the model on different subsets of dam landslide data. Preferably, is set to 10.

[0028] Example 2 In this example, as Figure 2 shown, we conduct an experimental comparison between the generative adversarial network based on hierarchical reinforcement in the method of the present invention and the generative adversarial network based on the existing method. By analyzing the recognition accuracy of the discriminator, the improvement trend of the discriminator in the recognition accuracy of real and generated data is shown. It can be seen from the figure that the present invention has an accelerating effect on the improvement of accuracy during the training process.

[0029] Example 3 In this example, we divide the constructed dataset into four subsets: dataset A, dataset B, dataset C, and dataset D. As Figure 3 shown, we conduct an experimental comparison before and after the present invention method adopts the local attention mechanism. It can be seen from the figure that the method of the present invention is significantly lower in computational complexity than the model without the attention mechanism. The method of the present invention can focus on the key - area features and reduce redundant calculations.

[0030] Example 4 In this example, as shown in 4, we conduct an experimental comparison between the method of the present invention and the support vector machine, random forest, and traditional CNN network. It can be seen from the figure that in terms of the comparison of the accuracy of dam landslide disease classification, the performance of the present invention is significantly better than other methods, reflecting better processing ability for complex data.

[0031] Example 5 This example provides an artificial - intelligence - based dam landslide disease recognition device, which executes the artificial - intelligence - based dam landslide disease recognition method described in Example 1, including: Sample acquisition module: used to collect the dam landslide data of the actual dam monitoring points and perform manual annotation to obtain the labeled dam landslide dataset; Sample augmentation module: used to input the labeled dam landslide dataset into the trained hierarchical reinforcement-based generative adversarial network for data augmentation to obtain the augmented dam landslide dataset; the hierarchical reinforcement-based generative adversarial network includes a generator and a discriminator; a hierarchical reinforcement learning mechanism is adopted in the generator, and the quality of the generated samples is optimized through a multi-layer decision-making process; Feature extraction module: used to input the augmented dam landslide dataset into the trained neural network optimized based on adaptive convolution kernels for feature extraction to obtain the extracted dam landslide data features; the neural network optimized based on adaptive convolution kernels includes an input layer, a convolutional layer, a pooling layer, and a fully connected layer; the convolutional layer forms residual blocks, and each residual block performs skip connections, and the skip connection is a conventional residual neural network connection method in the art; Dam landslide disease identification module: used to input the extracted dam landslide data features into the trained support vector machine algorithm based on high-order feature interpolation for dam landslide disease identification and classification to obtain the dam landslide disease identification result.

[0032] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for identifying dam landslide hazards based on artificial intelligence, characterized in that: The following steps are involved: S1. Collect the dam landslide data of the actual dam monitoring points and manually annotate them to obtain annotated dam landslide data set; S2. The annotated dam landslide data set is input into a trained hierarchical enhancement-based generative adversarial network for data expansion to obtain an expanded dam landslide data set; the hierarchical enhancement-based generative adversarial network includes a generator and a discriminator; The generator uses a hierarchical reinforcement learning mechanism to optimize the quality of generated samples through a multi-layer decision-making process; S3. Inputting the expanded dam landslide data set into a trained neural network based on adaptive convolution kernel optimization for feature extraction to obtain extracted dam landslide data features; The neural network based on adaptive convolution kernel optimization includes an input layer, a convolution layer, a pooling layer and a fully connected layer; the convolution layer is residually connected to form a residual block, and each residual block is skipped, and the skip connection is a conventional residual neural network connection method in the art; S4. Input the extracted dam landslide data features into the trained support vector machine algorithm based on high-order feature interpolation to perform dam landslide disease identification and classification, and obtain the dam landslide disease identification results.

2. The method for identifying dam landslide hazards based on artificial intelligence according to claim 1 is characterized in that: Step S1 specifically includes: The dam landslide data include geographical location Ra, current temperature Da, humidity value Ia, precipitation Pa around the dam, water level Ha, displacement Va, seepage pressure Sa, crack size Fa on the dam surface, accumulated rainfall La and historical disease records Ca; The labeling categories of dam landslide data are "normal", "minor landslide", "moderate landslide" and "severe landslide".

3. The method for identifying dam landslide hazards based on artificial intelligence according to claim 2 is characterized in that: Step S2 specifically includes: The training process of the generative adversarial network based on hierarchical reinforcement is as follows: S21. Initialize the network parameters of the generator and discriminator. The initialization strategy is to use a normal distribution with a mean of 0 and a variance of the unit matrix. The parameter initialization formula is as follows: ， ， in, represents the parameters of the generator, represents the parameters of the discriminator, It means that it obeys a specific distribution. represents a normal distribution with mean 0 and variance as the identity matrix, represents the identity matrix; S22. Pre-train the discriminator. The loss function formula of the discriminator in the pre-training stage is expressed as follows: ， in, represents the real dam landslide data, is the loss function of the discriminator, represents the input noise sampled from the noise distribution, represents the noise distribution, Express expectations, represents the discriminator function, represents a generator function, represents the distribution of real dam landslide data, represents the natural logarithm; S23. In the adversarial training stage, first, the generator parameters are fixed, the real dam landslide data and the generated dam landslide data output by the generator are input into the discriminator for processing, and the discriminator parameters are updated. The generated dam landslide data output by the generator is the data obtained by inputting the noise sampled from the noise distribution into the generator; then, the discriminator parameters are fixed, the generator parameters are updated, and the generator adjusts its generation strategy through the hierarchical reinforcement learning strategy; the loss function formula of the generator is expressed as follows: in, represents the loss function of the generator, represents the weight of the combined embedding loss, represents the combined embedding loss; the formula of the combined embedding loss is as follows: ， ， ， in, Indicates the number of samples generated in the current batch, represents the feature extraction function of the neural rolling network layer of the generator, Represents the feature extraction function for the The nonlinear transformation function of feature dimensions, It represents the characteristic dimension of generating embankment landslide data; represents the first weight parameter, represents the second weight parameter, represents the boundary parameter, which is used to control the balance of losses; represents the L2 norm, Indicates taking the maximum value, Indicates Real dam landslide data, Indicates Input noise; S24. Each layer of the generator is independently adjusted through reinforcement learning during training. During the training process of the generator, the reward for each adjustment is calculated based on the feedback of the discriminator and the embedding loss. The formula is as follows: ， ， in, represents the Sigmoid function, represents the hierarchical reinforcement learning rate parameter, Rewards representing adjustments to the generator’s neural network layers; S25. Repeat steps S22-S24 until the preset stop iteration condition is met, the training is completed, and a trained hierarchical reinforcement-based generative adversarial network is obtained.

4. The method for identifying dam landslide hazards based on artificial intelligence according to claim 3 is characterized in that: Step S3 specifically includes: The training process of the neural network based on adaptive convolution kernel optimization is as follows: S31. In the initial stage of training, the data in the expanded dam landslide dataset is hierarchically processed by grouping the dam landslide data into different categories and assigning them to different sub-networks. The formula is as follows: ， ， in, Represents the neural network The dam landslide data output after layer processing, Represents the neural network The weight parameters of the layer, represents the input dam landslide data feature matrix of the neural network, represents the activation function of the neural network, Represents the neural network Layer The weight of the feature, represents the feature dimension of the dam landslide data input to the neural network, The first Features, Represents the neural network The bias of the layer, Represents the Sigmoid activation function; the gradient descent method is used to adaptively update the neural network parameters, and the parameter update formula is expressed as follows: ， ， in, represents the learning rate of the neural network, Represents the loss function of the neural network for the The gradient of the layer's weights, Represents the loss function of the neural network for the The gradient of the layer's bias, represents the loss function of the neural network, Indicates parameter update operation; S32. In the initialization stage of the adaptive convolution kernel, the convolution layer automatically selects the most suitable convolution kernel size and step size according to the input dam landslide data, and optimizes the parameters of the convolution kernel through the regularization strategy. The formula is as follows: ， in, represents the output features of the convolutional layer, represents the adaptive convolution kernel weight, represents the convolution operation, Represents the bias of the convolution layer; the parameters are solved according to the regularization strategy of minimizing the target error, and the adaptive optimization of the initial value of the convolution kernel is completed. The formula is as follows: ， in, is the parameter update operation, represents the regularization coefficient, represents the Frobenius norm; the convolution kernel is updated by the gradient descent method so that the convolution kernel can adapt to the distribution of embankment landslide data in real time during training. The formula is as follows: ， in, Represents the loss function of the adaptive convolution kernel weights; S33. The local attention mechanism is used to let the model focus on the local area with key differentiation in the input dam landslide data. By weighting the features, redundancy and noise interference are suppressed to achieve effective feature extraction. The formula is as follows: ， in, represents the feature importance score, represents the attention weight, Represents attention bias; weighting of local areas is realized according to the Softmax function, and the feature importance score is converted into attention weight. The formula is as follows: ， in, represents the local attention weight, Indicates The feature importance scores corresponding to the features, Indicates the total number of samples input to the neural network in the current batch; S34. In the multi-scale feature fusion stage, according to the local attention weights obtained in step S33, the features extracted by different convolution kernels are weighted and superimposed, and the formula is expressed as follows: ， ， ， in, represents the fused multi-scale features, Indicates The weight coefficient of the convolution scale, Indicates convolution kernel weights, Indicates the number of multi-scales set, Indicates The importance coefficient of the scale, Indicates The complexity adjustment coefficient of each scale; S35. In the process of optimizing the residual connection, by building jump connections between different layers, the gradient vanishing problem is alleviated and the transmission efficiency of deep features is enhanced. The convolution kernel weights in the residual connection are updated as follows: ， in, represents the convolution kernel weight in the residual connection, represents the learning rate of the residual connection, Represents the gradient of the loss function of the neural network to the convolution kernel weight in the residual connection; the input features are integrated into the residual path and nonlinear transformation is realized, and the residual output is completed according to the jump summation method. The formula is as follows: ， in, represents the overall output characteristics of the neural network, Represents the bias of the residual connection; S36. Repeat iterative steps S32-S35 until the preset stop iteration condition is met to obtain a trained neural network based on adaptive convolution kernel optimization.

5. The method for identifying dam landslide hazards based on artificial intelligence according to claim 4 is characterized in that: Step S4 specifically includes: The training process of the support vector machine algorithm based on high-order feature interpolation is as follows: S41. Initialize the phase space mapping parameters, define the initial phase space mapping, adapt to the different characteristics of the dam landslide data after feature extraction through a dynamic adjustment strategy, and define the kernel function of the support vector machine. The formula is as follows: ， ， in, represents the kernel function of the extreme learning machine, Indicates data points in the input space, Indicates data points in the input space, represents the width parameter of the kernel function of the extreme learning machine, represents the L2 norm, represents the adaptive adjustment coefficient, is the data sample variance input to the extreme learning machine, representing the global variance; the adaptive adjustment coefficient is automatically adjusted according to the dispersion of the extracted dam landslide data features, and the formula is as follows: ， in, represents the adaptive adjustment coefficient, Indicates the number of samples input to the extreme learning machine in the current batch, Indicates data points in the input space, Represents the mean of all training samples input to the extreme learning machine in the current batch; S42. The original feature space is transformed into a higher-dimensional feature space by a high-order feature interpolation method. The interpolation of high-order features is achieved by the following polynomial mapping: ， in, represents the function that maps the data input to the extreme learning machine into a high-dimensional space, Indicates data points in the input space; Represents the highest order of the polynomial; S43. Perform dynamic adjustment of the phase space, automatically adjust the structure of the mapping function according to the distribution characteristics of the training data, and then optimize the classification boundary. Dynamically adjust the structure of the mapping function based on the distribution of the training data, and use the loss function to optimize the mapping parameters. The formula is as follows: ， in, represents the loss function of the extreme learning machine, Indicates The class labels of the data points, represents the kernel function of the extreme learning machine, represents the width parameter value of the kernel function of the extreme learning machine in the previous iteration, It represents the maximum value function; S44. After obtaining a suitable high-dimensional feature space, the optimization training of the support vector machine minimizes the objective function by adjusting the weight and bias parameters of the support vector machine. The formula is as follows: ， in, is the objective function of the support vector machine, Indicates that the parameters and Implement minimization constraints; The constraints of the objective function are: ,in, represents the weight vector of the support vector machine, Represents the bias term of the support vector machine; represents the regularization parameter of the support vector machine, which controls the trade-off between the error term and the model complexity; Represents the influencing parameter of the objective function, which is used to adjust the influence of the kernel function in the optimization process; Indicates slack variables, Indicates slack variables; S45. The stability of the algorithm is evaluated by cross-validation and stability index. The formula is as follows: ， in, It represents the stability index. represents the number of cross validation folds, Represents the cross validation The accuracy of folding.

6. The method for identifying dam landslide hazards based on artificial intelligence according to claim 5 is characterized in that: The weight of the combined embedding loss in step S23 is set to 0.3, and the boundary parameter is set to 0.2; the hierarchical reinforcement learning rate parameter in step S24 is set to 0.3; the learning rate of the neural network in step S31 is set to 0.03; the regularization coefficient in step S32 is set to 0.3; the attention bias in step S33 is set to 0.3; the learning rate of the residual connection in step S35 is set to 0.03; the highest order of the polynomial in step S42 is set to 6; the regularization parameter of the support vector machine in step S44 is set to 1; the number of cross-validation folds in step S45 is set to 10.

7. The method for identifying dam landslide hazards based on artificial intelligence according to claim 6 is characterized in that: The preset condition for stopping iteration is reaching a preset maximum number of iterations, and the preset maximum number of iterations is set to 1000 times.

8. An artificial intelligence-based dam landslide disease identification device, which executes the artificial intelligence-based dam landslide disease identification method according to claim 1, characterized in that: include: Sample collection module: used to collect dam landslide data from actual dam monitoring points and manually annotate them to obtain annotated dam landslide data sets; Sample expansion module: used to input the labeled dam landslide data set into the trained hierarchical reinforcement-based generative adversarial network for data expansion, and obtain the expanded dam landslide data set; the hierarchical reinforcement-based generative adversarial network includes a generator and a discriminator; The generator uses a hierarchical reinforcement learning mechanism to optimize the quality of generated samples through a multi-layer decision-making process; Feature extraction module: used to input the expanded dam landslide data set into the trained neural network based on adaptive convolution kernel optimization for feature extraction, and obtain the extracted dam landslide data features; The neural network based on adaptive convolution kernel optimization includes an input layer, a convolution layer, a pooling layer and a fully connected layer; the convolution layer is composed of residual blocks, and each residual block is jump-connected, and the jump connection is a conventional residual neural network connection method in the art; Dam landslide disease identification module: used to input the extracted dam landslide data features into the trained support vector machine algorithm based on high-order feature interpolation to perform dam landslide disease identification and classification, and obtain dam landslide disease identification results.

Images