208 results about "Posteriori probability" patented technology

The invention provides a crustal stress prediction method and device, electronic equipment and a storage medium, and relates to the technical field of seismic survey. The method comprises the following steps: obtaining observation seismic data of seismic wavelets in a strong VTI medium, and constructing a to-be-inverted parameter matrix based on a PP wave reflection coefficient corresponding to the strong VTI medium; constructing a posterior probability function obeyed by an inversion parameter matrix corresponding to the to-be-inverted parameter matrix based on a Bayesian inversion theory and observation seismic data, and determining a target functional based on a prior probability function and a likelihood function corresponding to the posterior probability function; determining medium density and each stiffness matrix coefficient based on an inversion parameter matrix solving result of the target functional, and determining a flexibility matrix of the strong VTI medium based on each stiffness matrix coefficient; and predicting the ground stress distribution of the target profile in the strong VTI medium based on the medium density and the positive strain matrix and the flexibility matrix corresponding to the strong VTI. Therefore, the crustal stress prediction accuracy under the strong VTI medium is improved.

The invention discloses a bare concrete appearance quality evaluation method, which belongs to the field of concrete quality evaluation, and adopts the technical scheme that process data are synchronously acquired in a bare concrete pouring process; obtaining concrete surface appearance index data after demolding; performing time synchronization and spatial inverse registration on the acquired data, and performing gridding association on the appearance indexes and the process data in the corresponding time windows; for each grid unit, extracting vibration effective energy, pumping pressure fluctuation characteristics, template deformation characteristics and environment correction factors from the process data in the corresponding time window and space range, and combining the vibration effective energy, the pumping pressure fluctuation characteristics, the template deformation characteristics and the environment correction factors into a joint characteristic vector; constructing a structural causal model, calculating the cause posterior probability of the appearance defect for each grid, and outputting cause probability distribution of under-vibration, over-vibration, unstable pumping, insufficient template rigidity and environmental sensitization; and generating an appearance quality score according to the cause probability distribution. The method has the beneficial effect that the method for evaluating the appearance quality of the fair-faced concrete is provided.

The invention relates to the technical field of industrial process prediction, in particular to a tundish erosion prediction method based on a hierarchical hybrid expert framework. The method comprises the following steps: collecting time sequence physical field data of the tundish, and screening features to construct a unified feature space; shunting the feature space to obtain a time sequence feature and a statistical aggregation feature; a statistical aggregation feature training classifier is utilized to generate a calibration posterior probability, and a gating network is constructed; generating an initial mode subset based on a posterior probability and training a corresponding expert model; the confidence of data to be measured is obtained by the gating network, and a single expert model is selected for prediction or multiple expert models are fused through a self-adaptive strategy for weighted prediction according to whether the confidence exceeds a threshold value or not; and finally, reconstructing the predicted value into an erosion thickness absolute value through inverse transformation. According to the method, the accuracy and adaptability of tundish erosion prediction are effectively improved.

The invention discloses a crane fault diagnosis method and system based on data driving, and the method comprises the steps: collecting the data of a PLC and a multi-source sensor of a crane, carrying out the preprocessing, and inputting a prediction model with the fusion of multi-scale causal convolution and an attention mechanism, so as to obtain a feature value prediction sequence; then calculating a residual error between a prediction sequence and an actual measurement sequence, modeling by using a first-class support vector machine, and triggering third-class early warning; based on the constructed Bayesian network, inputting the early warning evidence and updating the posterior probability, and outputting a Top-N fault reason; and finally, a risk score is calculated by integrating the posterior probability, the residual amplitude and the abnormal frequency, grading is carried out, and a diagnosis result and a disposal suggestion are pushed to a user terminal. According to the scheme, accurate diagnosis of complex coupling faults can be realized, 'beforehand 'early warning is realized, and unplanned shutdown and even safety accidents caused by fault expansion are effectively avoided.

The invention discloses an underground safety level intelligent evaluation response method and system fusing fuzzy credibility weighting and Bayesian reasoning, and relates to the technical field of underground safety monitoring. According to the method, a multi-source heterogeneous sensor is deployed to collect data, a fuzzy membership matrix is obtained through fuzzy membership function normalization, a credibility factor matrix is constructed by combining time sequence stability, spatial neighborhood consistency and node long-term health weight, and the two are subjected to weighted fusion to obtain a fusion index vector. And inputting the Bayesian network fused with the credible nodes, reasoning to obtain security level posterior probability distribution, and triggering a linkage response. The system correspondingly comprises a plurality of modules for implementing the steps. According to the method, the dynamic evaluation and credibility quantification of the data quality of the underground multi-source sensor are realized, and an intelligent evaluation mechanism for deeply integrating the credibility of the data source into a reasoning structure is constructed, so that the risk evaluation robustness and accuracy are improved, the closed-loop management from evaluation to response is realized, and the underground safety is guaranteed.

The invention discloses a personalized teaching course recommendation method and system based on artificial intelligence, and relates to the technical field of artificial intelligence and education recommendation, and the method comprises the steps: collecting text data for preprocessing, and extracting a standard target set, a teaching target candidate set and a knowledge point candidate set; calculating cosine similarity and weight based on the standard target set and the teaching target candidate set, calculating knowledge point mastery degree and weight based on the knowledge point candidate set, and splicing the knowledge point mastery degree and weight to generate a learning feature vector; based on the learning feature vector, clustering is carried out by using a k-means + + algorithm, a clustering label and a clustering center set are output, after the clustering center is updated in combination with a Thompson Sampling algorithm and Monte Carlo, the posterior probability is recalculated, and a final recommendation result is output; the robustness and recommendation accuracy of personalized teaching course recommendation are effectively improved.

The invention belongs to the field of seat control, and relates to a dynamic optimization method for seat adjustment parameters of a commercial vehicle. The method comprises the following steps: collecting and processing a motor signal to obtain a current vector sequence in an angle domain; extracting a composite feature matrix capable of reflecting the current state of the seat system through time-frequency analysis; taking the composite feature matrix sequence as an observation value of a factorial hidden Markov model, and diagnosing and distinguishing posterior probabilities of parallel hidden state chains corresponding to passenger loads and mechanical resistance; when the deviation between the diagnosed system state and the ideal state is too large, a multi-objective optimization program is started, and a plurality of performance indexes such as the adjustment time, the driving torque fluctuation rate and the predicted temperature rise are subjected to collaborative optimization through a non-dominated sorting genetic algorithm; and the optimal motor control parameters of the next period, which can balance the adjustment time, the riding comfort and the energy consumption, can be efficiently solved. The comprehensive requirements of a modern commercial vehicle for efficient seat adjustment, comfortable experience and reliable operation can be met.

The invention relates to the technical field of concrete detection, in particular to a concrete structure crack damage evaluation system based on acoustic emission sensing, which comprises a signal acquisition module, a characteristic discrete evolution module, a probability inference module, an energy gradient analysis module and a boundary defining module. According to the method, a multi-dimensional evolution set is constructed by extracting the amplitude and energy standard deviation of an acoustic emission signal in a continuous time window, the synchronous change trend of the multi-parameter standard deviation is analyzed by using a Bayesian probability model, and the non-uniform expansion posterior probability is calculated to lock a key signal set of dominant damage expansion. An energy fluctuation coefficient is calculated and a space sequence is generated by combining sensor space coordinates, and a fluctuation coefficient stable interval is identified according to a gradient attenuation rule of energy along with a distance, so that a physical boundary of a crack damage dynamic active region is quantitatively defined, and accurate evaluation of a non-uniform expansion state and an active range of a concrete crack is realized.

The invention discloses an intelligent inquiry method and device and electronic equipment, and relates to the technical field of intelligent inquiry and the technical field of data processing.The intelligent inquiry method comprises the steps that symptom information provided by a patient in the current round of inquiry is obtained, and the symptom information serves as latest symptom information; generating a current candidate disease set based on the latest symptom information; for each candidate disease in the current candidate disease set, calculating the posterior probability of the candidate disease after the current round of inquiry based on the latest symptom information by adopting a Bayesian algorithm; based on the posterior probability of each candidate disease in the current candidate disease set, the uncertainty entropy of the current candidate disease set is calculated, and the uncertainty entropy represents the uncertainty of the disease diagnosis result of the current candidate disease set; and determining whether to end the intelligent inquiry or not based on a size relationship between the uncertainty entropy and a preset entropy threshold value. By adopting the scheme, the inquiry efficiency and the inquiry integrity in intelligent inquiry are effectively balanced.

The invention provides a Bayesian network-based charging pile load prediction method and system, and belongs to the technical field of electric vehicles and load prediction. The method comprises the steps of obtaining real-time state data of a target charging pile, the real-time state data comprises current occupation state information of the target charging pile, real-time voltage data of a power grid node where the target charging pile is located and line loss data of the power grid node where the target charging pile is located; inputting the real-time state data into a preset charging pile load prediction model, so that the charging pile load prediction model takes the real-time state data as an input node of a Bayesian network, and calculating a posterior probability of a target node of the Bayesian network through a variable elimination method or a sampling approximate reasoning method, therefore, by implementing the method and the device, the problem that the accuracy and the efficiency of predicting the load of the charging pile are not high in the prior art can be solved.

The invention relates to the technical field of data intelligence and industrial internet, in particular to a multi-source data mining method, which comprises the following steps of: receiving multi-source business data, uniformly modeling and aligning according to a semantic contract, generating an event stream and a business graph and forming windowed data; extracting a multi-scale structure feature and a time sequence feature, and implementing cross-environment alignment to obtain a robust representation vector; constructing a structural causal model, calculating a signal sequential logic satisfaction degree, obtaining a violation relaxation vector, a shortest violation path and a minimum responsibility set, and generating an action sequence and an anti-factual benefit; and fusing multi-source evidences to obtain a posterior probability, combining a cost matrix and a time risk model to determine a trigger condition and an intervention time point, and outputting a disposal suggestion and an evidence collection package.

The invention discloses a new unit equipment risk management and control and spare part demand optimization method and system, and relates to the technical field of intelligent operation and maintenance, and the method comprises the steps: collecting operation condition signals in real time, constructing a digital twin model of each part of unit equipment, and forming a unified discrete state vector; performing feature extraction on the operation condition signal according to a fixed window, and constructing a Bayesian network of a hierarchical causal structure in combination with process attributes; performing posterior reasoning on the Bayesian network through belief propagation to obtain a fault posterior probability of a causal node, and calculating a risk score according to a weight and a consequence cost; and life parameter estimation is carried out, Monte Carlo simulation is used to predict the demand quantity, and the optimal spare part order quantity is calculated in combination with inventory constraints. According to the method, potential risks can be found in time, non-planned shutdown is reduced, inventory redundancy and capital occupation are reduced, and the safety, reliability and economical efficiency of operation of a new unit are improved.

The invention provides a hydraulic floating bridge ontology-free data reliability analysis method based on a time sequence decoupling network and a Bayesian network, and relates to the technical field of hydraulic floating bridges. Comprising the steps of determining source domain equipment according to a bill of material and a structure diagram of a target hydraulic floating bridge; determining a standardized time sequence monitoring data set of the source domain equipment; training a time sequence decoupling network by using the standardized time sequence monitoring data set to obtain a health probability data set of each key component of the target hydraulic floating bridge; inputting the health probability data set into a four-layer Bayesian network, and recursively calculating the posterior failure probability of each layer from bottom to top by adopting a belief propagation algorithm to obtain failure probability data and posterior probability distribution data; and performing risk assessment according to the failure probability data and the posterior probability distribution data to obtain a risk assessment result. The technical problems that in the prior art, data scarcity and model island exist in operation and maintenance of a hydraulic floating bridge and similar equipment, and reliability evaluation cannot be carried out are solved.

The invention discloses a transformer winding vibration abnormity diagnosis and early warning method based on lightweight edge calculation, and relates to the technical field of transformer monitoring. According to the method, vibration signals and magnetic flux signals of a transformer winding are collected in real time at edge nodes, and a basic data set is obtained after preprocessing; vibration and magnetic flux features are extracted to construct a basic diagnosis feature vector, a three-dimensional tensor is constructed through directional coherence analysis, a boundary enhancement structure map is generated, and a structure enhancement fusion feature vector is formed after fusion. The method comprises the following steps of: realizing abnormal grade and trend prediction by a lightweight Transform diagnosis model based on knowledge distillation and parameter compression optimization, and triggering model adaptive updating by combining risk index modeling and a dynamic risk threshold value. And finally, calculating an abnormal level posterior probability through a Bayesian updating rule, and outputting multi-level early warning. The method has the advantages of high precision, low calculation overhead and self-adaptive capability, and is suitable for online monitoring and intelligent operation and maintenance of the state of the transformer.

The invention discloses an SRAF placement method based on a Bayesian model. The method comprises the steps that historical SRAF configuration parameters and photoetching simulation data are collected and preprocessed; based on the preprocessed data, using kernel density estimation to construct prior probability distribution; constructing a likelihood function of multiple photoetching result indexes in combination with a Hopins photoetching model; calculating approximate distribution of a posterior probability through variation inference by using the Bayesian theorem; and finally selecting an optimal SRAF configuration parameter according to the posterior distribution, and verifying the manufacturing feasibility of the process window and the mask. According to the method, historical prior knowledge and real-time photoetching data are organically integrated through the Bayesian model, and high-precision, high-efficiency and high-robustness optimization of the SRAF layout is realized by combining specific technical characteristics such as preprocessing, kernel density estimation, a Hopkinson physical model and variation approximation; the problems of complex rule base, time-consuming calculation and insufficient adaptability in the prior art are effectively solved.

The invention belongs to the technical field of geotechnical engineering numerical simulation and parameter identification, and discloses a parameter probability inversion method and system based on SPH water and soil coupling scouring erosion, and the method constructs a probability inversion framework combining smoothed particle hydrodynamics (SPH) and a Bayesian-Markov chain Monte Carlo (MCMC) method. Large deformation and dynamic evolution of a water-soil interface in a water-sand two-phase flow scouring erosion process are simulated through an SPH method, priori knowledge and observation data are fused by adopting a Bayesian theory, and posterior probability distribution estimation of key soil parameters (such as an initial internal friction angle and an erosion rate) is realized by utilizing an MCMC algorithm. According to the method, the probability prediction from parameter uncertainty quantification to the scouring erosion model is realized, the reliability of scouring erosion numerical simulation is improved, the dependence on empirical parameters is reduced, and the technical problem that the existing method is difficult to consider complex water-soil coupling process simulation and parameter uncertainty quantification inversion is effectively solved.

The application relates to an intelligent education management method and system based on an Internet of Things. The method comprises the following steps: inputting a multi-channel physiological feature sequence into a hierarchical probability state space model to obtain a short-term cognitive state and a short-term cognitive state posterior probability; performing stability statistical testing on the short-term cognitive state posterior probability, and marking the short-term cognitive state as a phase-established state when the stability condition is met; constructing a cost function based on the phase-established state and a probability evolution track, and performing rolling time domain optimization based on the minimum cost function and a dynamic refractory period constraint to obtain an optimal learning action; calculating a decision oscillation index according to a current executed learning action sequence, and increasing the weight of an action change amplitude penalty and prolonging the locking time of the dynamic refractory period constraint by a set step length when the decision oscillation index exceeds a preset first oscillation threshold. The method can guarantee the continuity of a learning process and the stability of a cognitive state.

This invention discloses a bearing remaining life prediction method based on BN-RBM and DA-BiLSTM. First, raw vibration signals from the bearing during operation are collected and preprocessed. Then, a bearing remaining life prediction model is constructed. This model includes a BN-RBM feature extraction network, a DA-BiLSTM feature fusion network, and a linear regression layer. The BN-RBM feature extraction network dynamically updates the prior weight distribution of the Restricted Boltzmann Machine (RBM) using posterior probabilities constructed from Bayesian BN, and extracts features using the trained RBM. The DA-BiLSTM feature fusion network performs bidirectional feature fusion, and the linear regression layer performs linear activation, outputting the predicted bearing remaining life. This invention effectively improves the accuracy of bearing remaining life prediction by providing timely and effective life prediction before failure, reducing the risk of downtime caused by failures.

This invention discloses a deep learning-based method and system for predicting oil product sales trends, comprising: collecting and preprocessing multi-source data to obtain a time-series feature sequence; constructing training inputs and prediction targets by slicing using a sliding window; inputting the training inputs into an improved TSMixer model to obtain sales predictions and the penultimate latent space representation; constructing change point observations based on the differences in latent space representations and prediction residuals between adjacent time steps; performing Bayesian online change point detection calculations on the change point observations to obtain the posterior probability of the change point and the run length distribution; generating dynamic gating signals based on the posterior probability of the change point and scaling and updating the weight matrix of the time mixing layer; setting an asymmetric penalty coefficient based on the run length distribution and performing gradient pruning to update the model parameters; and outputting online sales trend prediction results and warning confidence information. This invention achieves change point perception prediction and updating, supporting inventory replenishment and delivery scheduling.

In order to explore the effectiveness of the functional data analysis method in the hyperspectral image processing, the application proposes a classification method combining the Gaussian regression mixture model and the MRF hyperspectral function data; first, the polynomial regression is used to fit the hyperspectral image pixel spectrum curve, so as to express the pixel spectrum information in the form of function; then, the neighborhood relationship is introduced to establish the Markov random field model, and the neighborhood Gaussian regression mixture model is established in combination with the Gaussian regression mixture model; finally, according to the maximum posterior probability criterion, the final hyperspectral image classification result is obtained. Since the spatial-spectral information of the hyperspectral image is fully combined, the algorithm has high-precision classification result, and effectively improves the classification performance of the hyperspectral image.

The invention provides a method for establishing a high-dimensional feature posterior probability estimator based on a sparse Bayesian method. The method comprises the following steps: (1) constructing a sparse Bayesian posterior probability estimation model; (2) model parameter representation based on the Bayesian theorem; (3) performing model parameter solving by using the sample data; and (4) realizing model sparsification based on automatic correlation confirmation. The invention designs a high-dimensional feature posterior probability estimator by using a sparse Bayesian method in order to solve the problem that the feature category affiliation probability is difficult to solve due to the fact that the high-dimensional feature probability distribution is difficult to estimate.

The invention discloses a few-sample readability evaluation method and system, electronic equipment and a storage medium, and belongs to the technical field of natural language processing. The method comprises the steps of obtaining a to-be-evaluated text, and constructing a global prompt with an independent structure and at least one local prompt for the to-be-evaluated text; combining the text with each prompt, inputting the combined text and prompt into a pre-training language model, and extracting global and local prompt feature representations; respectively calculating similarity distribution between each prompt feature representation and a plurality of preset static category prototypes, wherein the static category prototypes are kept fixed in model training; and performing joint modeling on the similarity distribution from different prompts based on a Bayesian probability fusion mechanism, generating fused posterior probability distribution, and determining the readability level of the text according to the fused posterior probability distribution. According to the method, multi-dimensional language features can be effectively modeled, tag semantic fuzziness is properly processed, and high-precision and high-robustness readability evaluation is kept in a few-sample scene.

The invention discloses a bearing fault diagnosis method and device based on an optimization algorithm, and a medium, and relates to the technical field of bearing fault diagnosis. The method comprises the following steps: collecting a vibration signal sequence of a target bearing in a current monitoring time period; inputting the fault signals into a pre-calibrated fault signal generation type model library; based on a Bayesian inference framework, taking the vibration signal sequence as evidence, and constructing posterior probability estimation taking a fault physical parameter set as an unknown quantity; carrying out iterative solution on posterior probability estimation by adopting particle filtering in a sequence Monte Carlo optimization algorithm, and generating and updating parameter particles carrying weights so as to approach probability distribution of a fault physical parameter set at the current moment; and based on the probability distribution, calculating the marginal probability of the fault type identifier, determining the type corresponding to the maximum marginal probability as the current fault, and performing extrapolation calculation to obtain residual service life probability distribution. According to the method, the probabilization and quantification of the diagnosis result are realized instead of a single label.

The invention discloses a micro-earthquake dry and wet event separation method and device based on a Gaussian mixture model, and the method comprises the steps: S1, building a corresponding Gaussian mixture model based on a micro-earthquake detection result, carrying out the random initialization of parameters of the Gaussian mixture model, and setting the number of mixed components; s2, calculating the posterior probability of each microseismic event point relative to each Gaussian component; s3, updating Gaussian mixture distribution parameters based on the posterior probability of each microseismic event point relative to each Gaussian component; s4, based on the updated Gaussian mixture distribution parameters, calculating a maximum likelihood function and evaluating whether the maximum likelihood function accords with a threshold value, if not, returning to S2, and if yes, executing S5; and S5, outputting the Gaussian mixture model after parameter updating and a dry and wet event classification result. According to the method, effective transformation events corresponding to fracturing construction and fracture activity events caused by ground stress changes can be rapidly distinguished.

The purpose is to appropriately manage the imbalance in the amount of signal processing. [Solution] The abnormality management device 1 includes a learning unit 12 that uses the characteristic directions of the normal data as training data to learn, by maximum likelihood estimation, parameters of a probability model that outputs the posterior probability that the number of signals allocated to the opposing device 3 for each time period corresponding to each of the characteristic directions of the normal data is normal; a derivation unit 13 that derives a probability distribution for the characteristic directions of the abnormal data that indicates the number of abnormal signals allocated to the opposing device 3 for each time period that deviates from the range of normal signal numbers, based on the posterior probability estimated by the probability model learned by the learning unit 12, the probability distribution for the characteristic directions of the normal data, and the prior probability of normality; and a calculation unit 14 that calculates a first index value that indicates the degree of spatial agreement formed by the probability distribution for the characteristic directions of the abnormal data derived by the derivation unit 13 and the probability distribution for the characteristic directions of the normal data.

This application relates to the fields of artificial intelligence and computer technology, specifically providing a risk decision-making method and system for the entire lifecycle of power data. The method includes: acquiring multi-source heterogeneous log data from the entire lifecycle of a power information system; using an unsupervised feature extraction module based on a variational autoencoder to extract latent feature vectors and determine the VAE reconstruction error; performing spatiotemporal feature extraction on the latent feature vectors to obtain the temporal behavior probability from the time-series analysis stream and the spatial graph embedding distance from the spatial analysis stream; and constructing a dynamic decision-making module based on a Bayesian network, using the VAE reconstruction error, temporal behavior probability, and spatial graph embedding distance as multi-source evidence nodes for the dynamic decision-making module to calculate the posterior probability of risk events. This application, through the collaborative fusion of spatiotemporal features and causal probabilistic inference, can significantly improve the safety risk perception and dynamic control capabilities of power data throughout the entire process of acquisition, transmission, and processing.

The application relates to the computer technical field and discloses a modeling method of a statistical mixed model in a big data distributed scene. The method comprises the following steps: distributing and storing data shards; initializing model parameters; iteratively performing an expectation step and a maximization step; the expectation step is scheduled to a GPU node to perform parallel calculation on posterior probability; the maximization step is scheduled to a CPU node to aggregate statistics and update parameters, perform component merging / deletion, and perform convergence judgment; meanwhile, a memory reuse mechanism based on reference counting and scope analysis is adopted to reduce redundant data transmission. Through heterogeneous task scheduling and memory collaborative optimization, the training speed, resource utilization rate, and model self-adaptation capability are improved.

The invention discloses a Bayesian model averaging-based power prediction model evaluation method and a related device, and belongs to the technical field of power system prediction. The method comprises the following steps: constructing a candidate model set comprising a plurality of power prediction models; secondly, historical data are used for training all the models, the posterior model probability of each model is calculated based on the Bayesian theorem, the posterior model probability serves as the scientific weight of the model, and the goodness of fit and complexity of the model are considered in the weight at the same time; and finally, for a new prediction input, performing weighted average on the prediction distribution of each candidate model by taking the posterior probability as the weight to generate comprehensive probability prediction distribution. According to the method, the advantages and disadvantages of each candidate model are scientifically evaluated through the posterior probability, and a comprehensive and probabilistic prediction result is finally generated, so that the robustness and reliability of prediction are improved, and richer decision information is provided for power grid dispatching.

The invention discloses a product key production parameter mining method and system based on user demand classification, and relates to the technical field of artificial intelligence. The method comprises the following steps: obtaining user preference data containing paired preference samples and a candidate parameter space formed by mass-producible parameter combinations; based on the data and the space, an improved self-adaptive direct preference optimization algorithm is adopted to train a strategy model, a dynamically calculated self-adaptive reward margin is introduced in the training process so as to adjust a model updating gradient according to a sample differentiation degree, and a Coubeck-Leibler divergence constraint is applied to limit parameter search in a candidate parameter space; and after the strategy model is converged, determining a group of product parameter combinations with the maximum posterior probability from the candidate parameter space according to the optimized model, and outputting the product parameter combinations as key production parameters.

The invention discloses an underwater acoustic anti-interference transmission method and system based on sparse time-frequency characteristic mapping, and the method comprises the steps: adaptively generating a fractional order linear frequency modulation waveform with a specific time-frequency shearing slope at a transmitting end according to the Doppler state of an underwater acoustic channel, and achieving the physical layer focusing of transmitted energy; at a receiving end, constructing a hybrid observation model comprising a wide block sparse channel component and a narrow block sparse burst noise component, and inferring posterior probability distribution of environmental burst noise in a fractional order domain through joint iteration by using a dual-channel variational Bayesian algorithm; and finally, recovering an original signal through soft threshold interference cancellation and fractional order channel equalization. Through active feature mapping of the waveform and heterogeneous sparse joint inference of the receiving end, the problem of communication failure caused by underwater acoustic high-dynamic Doppler diffusion and marine organism impulse noise interference is effectively solved, and the transmission reliability in a severe underwater acoustic environment is remarkably improved.