38 results about "Sparse regularization" patented technology

The invention relates to the technical field of electric power task adjustment, and provides an electric power task model fine tuning method and system, terminal equipment and a storage medium, and the method comprises the steps: inserting a shared low-rank expert module in a pre-training model trunk, dynamically calculating the activation weight of each expert according to the input characteristics through a gating network, and carrying out the fine tuning of an electric power task model; and a sparse regularization constraint activation number is introduced. During training, expert output is subjected to weighted fusion to generate fine adjustment parameters, multi-task joint training is adopted, balance optimization is achieved by dynamically adjusting task weights, and during reasoning, only part of experts with the highest weights are activated, and sparse calculation is achieved. According to the method, a plurality of low-rank expert modules are inserted into a large model trunk, a trainable gating routing mechanism is introduced, the optimal expert combination weight is dynamically calculated according to semantic features of input task samples, and knowledge fusion, parameter multiplexing and reasoning sparse activation among multiple tasks are achieved.

The invention discloses a hyperspectral image waveband selection method and system based on a three-dimensional convolution auto-encoder. The method comprises the following steps: carrying out block processing on an original hyperspectral image; designing a three-dimensional space-spectrum reconstruction network based on a global spectrum sensing module; constructing a combined loss function of a reconstruction error term and a sparse regularization constraint term, and performing iterative updating on the network by using an Adam optimizer; constructing a wave band comprehensive evaluation criterion based on triple constraints and a dynamic fusion mechanism; and successively selecting an optimal wave band by adopting a greedy algorithm according to a comprehensive evaluation criterion, and outputting a wave band selection result of the hyperspectral image. Through fusion of a three-dimensional spatial spectrum reconstruction module and a multi-head self-attention mechanism, local and global association of hyperspectral data is comprehensively mined, and deep joint representation of spectrum and spatial information is realized. Meanwhile, sparse regularization constraint is introduced into a loss function, and the network is guided to focus the most valuable key wave band.

The invention discloses a causal graph inference method and a causal graph inference system which are oriented to strong nonlinear continuous variable distribution and based on RealNVP and micro NOTEARS constraints. According to the method, a structure parameter matrix is constructed to represent candidate causal connection, parent variable condition input is constructed for each variable based on structure parameters, and accurate likelihood modeling is performed on the condition density of each variable by adopting a condition RealNVP normalization flow model. By constructing an objective function containing a conditional log-likelihood item, a sparse regular item and a NOTEARS style differentiable acyclic constraint item, a sparse causal structure meeting acyclic constraint is obtained by utilizing gradient optimization, and a causal graph is output. Further, based on the learned causal graph and a conditional RealNVP model, anti-fact inference is executed under a given intervention condition, and an anti-fact result is output. The method is suitable for strong nonlinear relation and complex continuous distribution scenes, and has the advantages of being stable in structure inference, high in interpretability and capable of supporting anti-fact analysis.

The application discloses a deep learning assisted accelerated fracturing construction parameter intelligent real-time optimization method, and belongs to the field of unconventional oil and gas field development, and the steps are as follows: a structured sample database covering geological properties, engineering parameters and fracturing effect responses is constructed, and unified pretreatment is carried out on multi-source data; a multi-modal space-time collaborative prediction model is constructed, combined with multi-expansion time convolution network, three-dimensional residual network and cross attention mechanism; supervised learning training is carried out based on the database, sparse regularization and learning rate scheduling are introduced, and the model precision and generalization ability are improved; a hierarchical collaborative optimization framework of outer layer Bayesian search-inner layer CMA-ES refinement is proposed, reservoir reconstruction volume maximization and construction feasibility are realized under complex constraints; compared with the existing method, the optimization parameters of the application not only include cluster distance and section distance, but also include construction parameters such as displacement, sand concentration and fracturing fluid type, and the application can realize real-time optimization of fracturing construction parameters by using a machine learning agent model.

The application discloses a hyperspectral anomaly detection method and device based on factor group sparse regularization and belongs to the technical field of remote sensing image processing. The method reduces three-dimensional hyperspectral data into a two-dimensional matrix form, decomposes the matrix into a background part and an anomaly part based on a background dictionary; utilizes a Schatten-p norm to regularize and constrain the low-rank characteristics of the background part, utilizes a 2,1 norm to regularize and constrain the column sparse characteristics of the anomaly part; converts the Schatten-p norm into a factor group sparse regularization form, introduces an auxiliary matrix to replace variables after performing singular value decomposition on a coefficient matrix; iteratively solves the replaced variables by using an alternating direction multiplier method to obtain an anomaly matrix and calculates an anomaly detection value of each pixel. The application avoids a complex calculation process, guarantees detection accuracy and significantly improves detection speed.

The application discloses a sparse regularization DOA estimation method based on risk minimization principle, and belongs to the technical field of array signal processing and underwater acoustic signal processing. The application receives array signals and establishes an observation model; constructs sparse representation and an overcomplete dictionary; establishes and initializes a regularization optimization model; adaptively updates weights and selects parameters in a risk-driven manner; after determining the regularization parameter, a fast iterative shrinkage threshold algorithm (FISTA) is used for optimization solving, and after each iteration convergence, dictionary refinement is performed on the detected direction to reduce off-grid errors; after a small amount of outer iteration is repeated, the final DOA estimation result and the corresponding power are output. The application can realize adaptive selection of the regularization parameter and the noise level, effectively overcome the precision degradation problem under complex conditions such as low signal-to-noise ratio, limited snapshot number, signal source correlation and power imbalance without manual parameter adjustment, and thus significantly improve the robustness and practicability of estimation.

A structural damage identification method based on random sparse regularization comprises the following steps: S1, establishing a finite element model of a to-be-identified damage structure, dispersing the structure into a finite number of units, setting a damage working condition, and simulating the damage working condition; s2, performing modal analysis, calculating a stiffness matrix and a mass matrix of the structure, adding boundary conditions, and solving feature values and feature vectors; s3, constructing an objective function of damage identification; s4, introducing an l1 / 2 regularization optimization method, and establishing a structural damage identification equation; s5, introducing a random class algorithm, optimizing and solving the structural damage identification target equation, and obtaining a solution vector; and S6, determining the damage position and the damage degree in the structure according to the solution vector of the damage identification equation. According to the method, the damage position can be quickly positioned, the damage degree can be accurately quantified, the accuracy of a damage identification result is improved, and the identification time of damage identification is shortened.

The application particularly relates to a UAV target main body and micro-motion component echo signal separation method based on factor decomposition group sparse regularization, which comprises the following steps: firstly, constructing a radar echo time-frequency representation into a Hankel matrix; secondly, using the characteristic that the echo Hankel matrix of the target main body uniform motion has low rank, establishing a signal separation model containing a low rank term, a sparse term and a noise term; thirdly, using a factor decomposition group sparse regularization method to effectively relax the rank function in the model, so as to improve the robustness of the algorithm; finally, using a linearized alternating direction multiplier method to iteratively solve the optimization model, so as to realize effective separation of the target main body echo and the micro-motion component echo in the time-frequency domain. The application is particularly suitable for narrow-band radar and short coherent accumulation time detection conditions, can effectively overcome the problems of model mismatch and low calculation efficiency existing in traditional methods, and simulation and measured data verify that the method has good separation precision and robustness in a noise environment.

The invention relates to a method and device for predicting the residual life of a frequency converter and computer equipment. The method comprises the following steps: acquiring state data of a to-be-tested frequency converter; inputting the working condition data into a generative adversarial network model for processing to obtain a health index of the to-be-tested frequency converter; reasoning the generative adversarial network model based on a sparse regularization algorithm; performing trend filtering on the health index of the to-be-tested frequency converter, and determining an initial degradation time point of the to-be-tested frequency converter; screening out state data after the initial degradation time point from the state data of the frequency converter as degradation stage data; the degradation stage data are input into a feature extraction model for feature extraction, and related features of the to-be-tested frequency converter are obtained; and inputting the related characteristics into a life prediction model for life prediction to obtain the residual life of the to-be-tested frequency converter. By setting the sparse regularization algorithm and combining the bidirectional gating cycle unit and the domain invariant model, the prediction accuracy of the residual life of the frequency converter is improved.

The invention discloses an image recognition method based on multidirectional structure maintenance and adaptive sparse regularization, and belongs to the technical field of image processing and mode recognition. Comprising the steps of obtaining an image sample data set; decomposing each image in the image sample data set into gradient images by using local gradient image decomposition; constructing an objective function, wherein the objective function comprises a nuclear norm regression loss item, a local preserving loss item and an adaptive sparse regularization loss item; and solving the objective function model, wherein a coefficient vector obtained by solving is used for a final image classification decision. According to the method, the local preserving item is introduced into the target function, so that the manifold structure between samples is effectively utilized, and the discrimination capability of the model is enhanced; meanwhile, an adaptive sparse regular term is introduced, interference in the noise direction is effectively suppressed, and the stability of the model in complex scenes such as shielding and illumination variation is improved. The model can be automatically focused in the gradient direction with rich information, and more intelligent feature utilization is realized.

This invention discloses an adaptive detection method and apparatus for range-extended targets based on scattering center estimation. A binary hypothesis testing model for range-extended targets is established, and the probability density functions under two hypotheses are determined. Maximum likelihood calculation is performed on the unknown parameters under both hypotheses to obtain the adaptive matched filter test statistic of the composite Gaussian model. The texture components of the detection unit are estimated using the correlation of texture components in the range dimension, yielding the test statistic after texture component estimation. A sparse regularized optimization model is constructed and solved to obtain a range-extended target detector with adaptive scattering center estimation. This method adaptively estimates the scattering center without requiring prior information about the target, reducing interference from clutter in units without a target scattering center. Furthermore, this method estimates the texture components based on the range-dimensional correlation of clutter, resulting in superior detection performance compared to traditional adaptive detection methods.

The invention provides an electric power system line admittance leakage threat assessment method and system based on a graph neural network and a related device, and the method comprises the following steps: processing to-be-processed data through employing a pre-constructed graph neural network model, and obtaining an electric power system line admittance leakage threat assessment result, wherein the pre-constructed graph neural network model comprises a feature extraction module, and the output of the feature extraction module is connected with the input of a data inference module through a domain discriminator; the multi-objective loss function of the pre-constructed graph neural network model comprises inference error loss, physical constraint loss, topology loss and sparse regularization loss; according to the method, sensitive information leakage threats possibly brought by public data are simulated and evaluated through a real scene, the existence and potential influence of the threats are verified, a scientific basis is provided for an operator to identify risks and formulate a targeted defense strategy, and the practical value is high.

The application relates to a space target radar image data generation method based on low-light enhancement correction. The method comprises the following steps: designing a content transfer decomposition network, decomposing low-light features into illumination-independent reflection components and adaptive illumination components through Hadamard product constraints in a latent space, proposing a learnable intensity compression function, dynamically generating a latent space mask through gradient consistency constraints and sparse regularization, embedding the mask into a reverse denoising process to realize directional repair of degradation areas, combining the generation ability of a diffusion model and the guidance of a physical prior, reconstructing a diffusion path based on a bidirectional diffusion principle, generating the physical consistency of the process through endpoint binding constraints, adaptively adjusting the diffusion step, introducing a cyclic implicit iteration mechanism, gradually refining the generation result through a recurrent neural network module, and realizing accurate optical-ISAR image translation.

The invention belongs to the technical field of image processing, and particularly relates to a hyper-spectral image classification method for semantic-guided kernel low-rank sparse preserving projection, which comprises the following steps: firstly, acquiring a training sample set and a test sample set of a hyper-spectral image, and calculating a kernel matrix and a cross kernel matrix of the hyper-spectral image; further constructing a kernel method dimension reduction model objective function fusing a sparse reconstruction error term, a sparse regularization term, a semantic guidance low-rank constraint term and a spatial adaptive manifold regularization term; the complex objective function is efficiently solved by adopting an alternating direction multiplier method, and an optimal projection matrix is finally obtained by introducing an auxiliary variable and alternately updating a projection matrix coefficient, a sparse coding matrix and other variables; and mapping the test sample to a low-dimensional feature space by using the projection matrix so as to finish classification. According to the method, through semantic guidance and spatial-spectral information adaptive fusion, the discrimination ability and classification precision of dimension reduction features are significantly improved.

The invention discloses a method for recognizing an early brain fatigue state through a two-dimensional attention model, and belongs to the technical field of electroencephalogram signal processing. Stable electroencephalogram characteristics are obtained through self-adaptive electrode channel selection and data enhancement; in the model part, a multi-branch convolution structure (3 * 3, 5 * 5 and 7 * 7) is combined with cavity convolution to extract multi-scale features, and a space and channel two-dimensional attention mechanism is introduced to enhance key electrode and channel response; adaptive weighted fusion of multi-scale features is realized through an improved SKNet structure, and sparse regularization and pruning strategies are added in the training process to reduce the complexity of the model. The brain fatigue level can be accurately recognized in the early stage, and the method has the advantages of being high in recognition precision, small in calculation overhead and high in generalization ability and is suitable for driving safety monitoring, aviation work, brain health assessment and other scenes.

This invention discloses a training method for an intrusion detection model robust to adversarial examples, comprising: Step 1, introducing perturbations into the original feature representation of the model to generate adversarial feature representations; Step 2, designing feature consistency loss constraints and information distribution consistency regularization constraints based on the feature activation representations within the intermediate layer, and combining them with the original discriminative loss to form a joint loss function; Step 3, applying sparse weight vectors to the feature activation representations within the key intermediate layer, and constructing sparse regularization terms based on the sparse weight vectors; Step 4, calculating adversarial perturbation projections using the manifold of normal training samples in the feature space of the key intermediate layer to form effective perturbations and obtain adversarial feature samples; Step 5, assigning training weight coefficients to the adversarial feature samples, and applying the training weight coefficients to the joint loss function during the training update process. This invention can improve the robustness and generalization ability of the model in adversarial attack environments.

The invention discloses a sparse constraint-based adaptive scheduling method for cloud edge cooperative industrial visual inspection, which comprises the following steps of: quantifying a visual inspection task, performing multi-dimensional feature extraction, and constructing a task feature vector; multi-dimensional modeling is carried out on available resources of the computational nodes, and node resource state vectors are constructed; a distributed reinforcement learning scheduling model based on sparse constraint is constructed, the distributed reinforcement learning scheduling model based on sparse constraint comprises a plurality of intelligent agents deployed at each edge node, the intelligent agents construct a deep neural network, and the deep neural network outputs Q values corresponding to all possible actions according to an input state vector, a dynamic structured sparse regularization item capable of sensing the system state and the decision history is introduced into the reward function, and a scheduling strategy is obtained through optimization. The method can guide a scheduling strategy to not only tend to process tasks at an edge end, but also can form a stable and grouped decision mode, thereby avoiding decision jitter and resource fragmentation.

The application discloses a kind of based on compact manifold mapping linearization processing method of high-dimensional data, it is related to data processing technical field.The application is adaptively preprocessed to high-dimensional time series data, through improved order-preserving compression transformation, combined with the scale and attenuation factor of dynamic adjustment, each dimension non-negative infinite domain data is mapped to one-dimensional torus compact manifold angle coordinate;Linear state vector is constructed by angle coordinate, and linear state space model containing sparse regularization and adaptive noise suppression is established;Rolling window multi-step prediction combined with attention mechanism is used, window and model parameters are dynamically updated, and spatial prediction value is obtained;The spatial prediction value is restored to original data domain by accurate inverse transformation;The rationality of prediction is judged by triple check, and related parameters are dynamically adjusted according to the type of anomaly.The application maps infinite domain data to compact manifold, realizes that the computational complexity is reduced from exponential level to linear level, and has wide application value.

The application provides a repeated fracturing well production time series data decomposition method based on minimum absolute deviation loss, comprising the following steps: (S1) collecting and arranging original data; (S2) using bilateral filtering to denoise input data; (S3) robustly extracting a trend by solving minimum absolute deviation regression with sparse regularization; (S4) using non-local seasonal filtering to obtain a seasonal component; (S5) adjusting the trend and the season; (S6) repeating steps 2-5 until the calculation result converges. The application can stably extract a seasonality and a trend from a long seasonal cycle and high-noise data, and can provide a solid foundation for subsequent fitting and prediction of repeated fracturing production data.

The application provides a kind of urban solid waste incineration process multi-objective integrated random configuration network furnace temperature prediction method, relating to urban solid waste incineration process furnace temperature intelligent prediction technical field, the method is to the urban solid waste incineration process data is pretreated, by self-service sampling strategy, obtain S training data set;Using random configuration network algorithm, base model construction is carried out, and the base model corresponding to each training data set is obtained;Using global negative correlation integration strategy based, S base model integrated model is trained in parallel, and the learning ability of integrated model to multi-region furnace temperature is realized by introducing group sparse regularization technology, and the trained integrated model is obtained;Using the trained integrated model, the urban solid waste incineration process data is analyzed, and the furnace temperature prediction result is obtained, to complete the prediction of furnace temperature.The application solves the problem of low accuracy and stability of the existing furnace temperature prediction method.

This invention relates to the field of geophysical exploration technology and provides a method and system for compressed sensing reconstruction of magnetotelluric data. The method includes the following steps: acquiring raw magnetotelluric time-series data, performing segmentation and standardization preprocessing to obtain multiple standardized signal segments; transforming each standardized signal segment to obtain a sparse representation in the transform domain, and compressing and sampling the sparse representation to obtain a compressed measurement vector; extracting multidimensional feature vectors from each signal segment, predicting the corresponding regularization parameter scaling factor based on a neural network model, and calculating an adaptive regularization parameter; using the adaptive regularization parameter as the regularization intensity, constructing a convex optimization model containing L2 data fidelity terms and L1 sparse regularization terms, and reconstructing the magnetotelluric time-series data based on a greedy, accelerated iterative threshold shrinkage algorithm. This invention can reduce bandwidth and facilitate real-time monitoring in environments with high transmission requirements, such as field environments.

This invention discloses a prompt-based dynamic multi-scale coding source load prediction method and system. The method constructs a three-branch encoder (short window / seasonal window / long window), utilizes prompt-based fine-tuning, and employs a fine-tuned encoder feedforward network to achieve three different branches. It also incorporates lightweight gate output branch weights and sparse regularization to achieve dynamic context selection and interpretability display in different scenarios. By using a Gaussian kernel to learn parameterized lag kernels for exogenous channels, and aligning exogenous data according to lag distribution to multiple views via one-dimensional convolution, the method feeds these views into a cross-attention mechanism. This explicitly models the influence strength of exogenous data on endogenous data at different lags, improving causal alignment capabilities. The aim is to improve the accuracy and interpretability of source load prediction through prompt-driven multi-scale routing and lag influence kernels.

The application discloses a lightweight target detection method, and particularly relates to the technical field of target detection. Firstly, the input image is preprocessed, and low-computational and rich-semantic features are extracted through a parallel convolution branch. Then, an ICAneck lightweight backbone is adopted, and a separable convolution and a coordinate attention are combined to extract semantics. A strategy selection model is constructed, and according to the target scale, the definition and the small target density, switching is carried out between the fixed and dynamic resolution-channel strategies, so that the resolution and the number of channels are finely controlled. In stage 2-4, an incremental ICAblock is used for deepening abstraction. In the fusion stage, a proportional attention and a channel alignment module are introduced to integrate scale features, and a lightweight detection head is used to output the category and the frame. Through joint training of sparse regularization and multi-task loss, the calculation and the memory consumption are reduced, the accuracy of small targets is ensured, and the edge real-time deployment is adapted.

This invention belongs to the field of power battery life prediction technology, and particularly relates to a power battery life prediction method based on mechanism-aware sparse coding and WTA prediction. First, multi-source operating data is collected and clustered into typical operating conditions. A sparse autoencoder with dynamic sparse masking, two-order sparse regularization, and physical consistency constraints is constructed to extract low-dimensional core aging features. Then, a multi-branch prediction structure adapted to the operating conditions is established, and a WTA competitive training mechanism is used to update only the optimal branch parameters to achieve specialized probabilistic prediction. Finally, the sliding window is dynamically adjusted based on feature sparse entropy, and the multi-branch results are fused through Kalman filtering and Bayesian rules to output the final RUL with confidence intervals, triggering multi-condition maintenance decisions. This invention solves the problem in existing technologies where the lack of deep integration of time-series sparse modeling, probabilistic prediction, and power battery degradation mechanisms leads to poor power battery life prediction accuracy.

The invention discloses an asset standing book intelligent checking method based on learnable fuzzy measure, comprising the following steps: collecting asset standing book field information and performing standardization and vectorization representation to obtain standing book field vectors; constructing a learnable fuzzy measure, and endowing any field subset with a capacity parameter; on the basis of the capacity parameters, realizing micro-calculation of Schka integrals by adopting Lowitz expansion, and obtaining the similarity of the candidate asset pairs; constructing a comparative learning objective function by taking the similarity as a unique measure, and respectively generating a positive sample pair and a difficult negative sample pair; monotonic projection and interactive sparse regularization are executed in the training process, and an optimized similarity matrix is obtained; and outputting the checking difference, and generating a processing suggestion. According to the method, calculus can be realized through learnable fuzzy measure and LoWitz extension to complete field soft matching and anomaly recognition, so that the checking accuracy and the intelligent level are improved.

The application discloses a SAR interference self-elimination method based on weak background prior, constructs an interference suppression optimization model based on weak background prior by detecting homologous weak background prior interference to be suppressed, protects useful signals by sparse regularization with hyperparameters, constructs an equivalent unconstrained optimization model by using a Lagrange, obtains an iterative relationship by an alternating direction multiplier method, and calculates a closed-form solution of a low-rank component of the interference suppression model to the useful signals by a soft threshold operator. The method can well utilize information of homologous interference to complete interference suppression on a specified area in SAR data polluted by interference, and has a certain energy protection for useful signals.

The invention provides an industrial process fault diagnosis method based on a Kolmogorov-Arnold network piecewise linear spline edge function, and belongs to the technical field of industrial artificial intelligence and process safety monitoring. The method comprises the following steps: performing data preprocessing on original industrial sensing data through standardization and a sliding window to form a fixed-length sample sequence, and inputting the fixed-length sample sequence into a KAN edge function layer; the method comprises the following steps: adding a linear transformation item and a first-order piecewise linear spline item to obtain a low-dimensional implicit representation; then, outputting a fault category through a lightweight classification network; sparse regularization is introduced in the model training process to promote part of edge function coefficients to converge to zero, after training is completed, the importance ranking of all variables can be obtained through edge function coefficient analysis, and interpretable analysis of a fault judgment basis is achieved in combination with a representative edge function response curve. The method is high in diagnosis precision, the model is light and efficient, the result is transparent and explainable, and variable-level and interval-level accurate contribution degree tracing can be achieved.

The invention discloses a continuous causal discovery and causal intensity estimation method and system based on variational potential space representation. According to the method, continuous observation variables are mapped to submerged space variables through a variational auto-encoder, reconstruction learning is carried out, learnable causal structure parameters are constructed in a submerged space, acyclic constraints are applied, and the submerged variables meet a structure equation generation mechanism under the condition of parent variables. The method comprises the following steps: constructing a joint objective function comprising an evidence lower bound term, a structure sparse regular term and an acyclic constraint term, performing joint or alternate optimization to obtain a submerged space causal diagram, and mapping a causal structure and causal intensity to an observation space by combining the sensitivity of a decoder; meanwhile, causal effect estimation can be realized based on intervention generation. The scheme is suitable for a strong nonlinear continuous variable scene, can output a quantifiable causal intensity result while improving the structural interpretability, and provides support for decision analysis and causal reasoning.

The application relates to the technical field of wind farm cable insulation fault diagnosis, and discloses a high-precision fault diagnosis method and system for the insulation state of a wind farm cable, which comprises the following steps: acquiring radial multi-measurement-point electric-thermal response data, using a cross-correlation algorithm to calculate a time delay difference to generate a radial time delay gradient vector; acquiring axial multi-measurement-point electrical measurement values to generate an axial spatial distribution measurement vector; constructing a three-dimensional grid discretization space to generate a three-dimensional parameter tensor; based on a multi-physical field coupling simulator, using the radial time delay gradient vector and the axial spatial distribution measurement vector to inversely calculate radial and axial parameter distributions respectively; using a sparse regularization tensor completion algorithm to fuse bidirectional constraints to complete a complete three-dimensional parameter tensor; identifying abnormal grid units and extracting a defect area to generate a three-dimensional defect geometric body; and calculating characteristic parameters and outputting a hazard degree evaluation. The application solves the technical problem that the prior art cannot accurately identify the three-dimensional spatial position and geometric morphology of internal defects of insulation.

The invention provides an aviation structure impact load sparse algorithm expansion network identification method. The method comprises the following steps: collecting a monitoring object and making a data set; constructing an impact load sparse identification model based on an L1 norm; establishing an iterative algorithm for solving the impact load sparse recognition model; constructing a depth algorithm expansion network based on the iterative algorithm; training the depth algorithm expansion network by using the data set until an optimal network model is obtained; and acquiring a response signal of a monitored object under the action of a to-be-identified impact load, identifying the impact load by using the optimal network model, and outputting the to-be-identified impact load. The method is suitable for identifying the impact load of the aviation structure, and compared with a traditional impact load identification method based on a sparse regularization method, the method has the advantages that the calculation speed is high, and parameters do not need to be manually set; compared with a traditional artificial neural network, the neural network provided by the invention has better interpretability in structural design.