An underwater combat effectiveness evaluation method based on complex sound propagation environment

Abstract

This application discloses a method for evaluating underwater combat effectiveness under complex sound propagation environments, comprising: acquiring multi-source marine environmental parameters; calculating the time-varying three-dimensional sound field propagation loss field using a hybrid algorithm combining parabolic equations and coupled normal mode wave models; constructing the sound channel state matrix; establishing time-varying sonar equations to generate sonar performance surfaces; constructing a joint state representation and a partially observable Markov decision process model, outputting the confidence distribution of enemy tactical intentions; establishing an environmental tactical coupling operator to update the joint state representation; extracting bottom-level and mid-level indicators, calculating the environmental complexity factor and determining the time-varying weight coefficients; establishing a hierarchical effectiveness index mapping function to output a time-varying effectiveness evaluation sequence, and constructing a digital twin to generate statistical distributions and confidence intervals; quantifying uncertainty and establishing closed-loop correction parameters for the hybrid algorithm. This application achieves high-confidence dynamic evaluation of underwater combat effectiveness under complex sound propagation environments.

Description

Technical Field

[0001] This application relates to the field of underwater combat effectiveness assessment technology, and in particular to an underwater combat effectiveness assessment method based on complex acoustic propagation environments. Background Technology

[0002] Underwater combat effectiveness assessment is a core support for underwater combat decision-making, and its accuracy directly affects operational deployment, tactical selection, and mission completion efficiency. Existing underwater combat effectiveness assessment methods suffer from the following technical shortcomings: First, most existing methods employ a single sound propagation model, which cannot adapt to complex sound propagation scenarios across all frequency bands and terrains, leading to significant errors in propagation loss calculations and consequently affecting the accuracy of effectiveness assessment. Second, the uncertainty of the assessment results is not effectively quantified, resulting in unclear assessment confidence levels and difficulty in supporting precise operational decision-making.

[0003] Therefore, there is an urgent need for an underwater combat effectiveness assessment method that can adapt to complex sound propagation environments and achieve dynamic coupling between the environment and tactics. Summary of the Invention

[0004] The main objective of this application is to provide a method for evaluating underwater combat effectiveness under complex acoustic propagation environments, which solves the technical problems of poor adaptability of acoustic propagation modeling and lack of uncertainty quantification in existing underwater combat effectiveness evaluation methods under complex acoustic propagation environments, and achieves high-confidence evaluation of underwater combat effectiveness.

[0005] This application provides a method for evaluating underwater combat effectiveness in complex acoustic propagation environments, employing the following techniques: A method for evaluating underwater combat effectiveness in complex acoustic propagation environments includes: S1. Obtain multi-source marine environmental parameters of the target sea area to construct an initial parameter set, use a hybrid algorithm to calculate the time-varying three-dimensional sound field propagation loss field, and establish a time-varying impulse response function to construct a spatiotemporally related acoustic channel state matrix. S2. Based on the acoustic channel state matrix and the propagation loss field, calculate the time-varying detection threshold, construct a signal processing gain model, calculate the processing gain, establish and solve the time-varying sonar equation, and generate the sonar performance surface. S3. Construct a joint state representation based on the sonar performance surface and the acoustic channel state matrix, build a tactical action library and a partially observable Markov decision process model, and output the confidence distribution of the enemy's tactical intentions. S4. Establish an environmental tactical coupling operator based on the tactical intent confidence distribution and the tactical action library, and output the updated joint state representation. S5. Extract the bottom-level indicators from the updated joint state representation, calculate the middle-level indicators according to the tactical intent confidence distribution, and calculate the environmental complexity factor according to the bottom-level indicators, the middle-level indicators and the acoustic channel state matrix, and allocate weights to determine the time-varying weight coefficients. S6. Establish a hierarchical effectiveness index mapping function based on the time-varying weight coefficients, output a time-varying effectiveness evaluation sequence, and construct an underwater combat digital twin based on the updated joint state representation to generate the statistical distribution and confidence interval of the time-varying effectiveness evaluation sequence. S7. Quantify the uncertainty of the statistical distribution and establish a closed loop. Update the parameters of the hybrid algorithm based on the deviation between the measured and evaluated results.

[0006] Optionally, step S1 includes: Obtain multi-source marine environmental parameters of the target sea area to construct an initial parameter set; Based on the initial parameter set, a hybrid algorithm combining parabolic equations and coupled normal mode wave models is used to calculate the time-varying three-dimensional sound field propagation loss field. Based on the time-varying three-dimensional sound field propagation loss field, the random phase screen theory is introduced to parameterize the multipath delay spread and coherence bandwidth of the sound signal, and a time-varying impulse response function is established. Based on the time-varying impulse response function, a spatiotemporally correlated acoustic channel state matrix is ​​constructed.

[0007] Optionally, the model is dynamically selected based on the joint criterion of frequency gradient, specifically including: When the signal frequency At that time, a coupled normal mode model was used; When the signal frequency And slope Time, or frequency A parabolic equation model is adopted; For frequencies located at And seabed slope In the transition region, the weighted average of the calculation results of the coupled normal mode model and the parabolic equation model is used as the final sound field propagation loss value.

[0008] Optionally, step S2 includes: Based on the time-varying three-dimensional sound field propagation loss field and the sound channel state matrix, the propagation loss value and channel coherence characteristic parameters at the target location are extracted; Calculate the time-varying detection threshold of the sonar based on the propagation loss value; Based on the channel coherence characteristic parameters, a signal processing gain model is constructed, subspace projection filtering is used to suppress reverberation, and the effective processing gain is calculated. Based on the time-varying detection threshold and the effective processing gain, a time-varying sonar equation is established and the detection probability, maximum effective range, and standard deviation of ranging error are solved to generate a sonar performance surface.

[0009] Optionally, step S3 includes: A joint state representation is constructed based on the sonar performance surface and the acoustic channel state matrix; Establish a tactical action library and construct a partially observable Markov decision process model based on the joint state representation; Based on the observation space of the partially observable Markov decision process model, the intention recognition network is used to make probabilistic inferences about the enemy's tactical behavior and output the tactical intention confidence distribution.

[0010] Optionally, step S4 includes: Analyze the impact of tactical actions on the sound propagation environment and sound channel characteristics, and quantify the changes of each tactical action on each element in the sound channel state matrix. Based on the quantification results of the influence of the tactical intent confidence distribution and tactical actions, an environmental tactical coupling operator is established. The environmental tactical coupling operator is applied to the original acoustic channel state matrix to obtain the updated acoustic channel state matrix. Combined with the real-time update result of the sonar performance surface, the updated joint state representation is output.

[0011] Optionally, step S5 includes: Based on the updated joint state characterization, the detection probability, maximum effective range and the standard deviation of ranging error are extracted from the sonar performance surface as underlying indicators. Based on the confidence distribution of the tactical intent, the enemy's tactical threat level is calculated, and based on the strength of the environmental tactical coupling operator, the degree of influence of tactical actions on the environment is calculated. The enemy's tactical threat level, the degree of influence of the tactical action environment, and the degree of completion of the combat mission are used as mid-level indicators; Calculate the environmental complexity factor based on the acoustic channel state matrix; Based on the underlying and intermediate indicators, evidence theory and analytic hierarchy process are used to perform mixed weight allocation, and the mixed weights are dynamically adjusted according to the environmental complexity factor to determine the final time-varying weight coefficients.

[0012] Optionally, step S6 includes: A hierarchical performance index mapping function is established based on the bottom-level indicators, the middle-level indicators, and the time-varying weight coefficients. The efficiency index at different times is calculated by the hierarchical efficiency index mapping function, and the efficiency index at each time is arranged in time series to output the time-varying efficiency evaluation sequence. Construct an underwater combat digital twin and integrate it with the time-varying performance evaluation sequence and the updated joint state representation; Construct a multi-agent reinforcement learning model in an underwater combat digital twin; Using the time-varying performance evaluation sequence as the reward function, multiple simulations were performed using the Monte Carlo simulation method. Statistical analysis is performed on the time-varying performance evaluation sequences obtained from multiple simulations to generate the statistical distribution and confidence intervals of the performance evaluation values.

[0013] Optionally, step S7 includes: A Bayesian neural network model is constructed, with the statistical distribution of the time-varying performance evaluation sequence as input and the evaluation uncertainty as output, and the model is trained. The statistical distribution obtained from the Monte Carlo derivation is input into the trained Bayesian neural network, the uncertainty is quantified, and the evaluation confidence is output. Parameters that affect the performance evaluation results are extracted by using neural networks to form a set of key sensitive factors; The measured data of parameters in the set of key sensitive factors are compared with the corresponding parameter evaluation values ​​in the acoustic channel state matrix to perform deviation analysis and calculate the deviation rate of each parameter. A deviation threshold is set. When the deviation rate of any parameter exceeds the threshold, or the average deviation rate of multiple parameters exceeds the threshold, the online parameter update mechanism is triggered to adjust the model parameters of the hybrid algorithm of the parabolic equation sound propagation model and the coupled normal mode model. The updated hybrid algorithm was reused to calculate the time-varying 3D sound field propagation loss field.

[0014] Optionally, the online parameter update mechanism adopts a tiered triggering method: A deviation of >3dB between measured and evaluated values ​​for 10 minutes is defined as a slight deviation, and only the depth layer parameter subset of the sound velocity profile interpolation node is updated. If the deviation between the measured and evaluated values ​​is greater than 6 dB, it is considered a severe deviation, and the sound velocity profile interpolation nodes need to be recalculated.

[0015] The advantages of this application are as follows: The underwater combat effectiveness evaluation method based on complex sound propagation environment proposed in the embodiments of this application adopts a hybrid algorithm of parabolic equation and coupled normal mode model to achieve accurate adaptation to complex sound propagation scenario and reduce propagation loss calculation error; it realizes bidirectional dynamic coupling between tactical actions and sound propagation environment through environmental tactical coupling operator, and the evaluation results are more in line with actual combat scenario; it quantifies the evaluation uncertainty through Bayesian neural network and outputs evaluation confidence and confidence interval, providing quantitative confidence basis for combat decision making. Attached Figure Description

[0016] Figure 1A flowchart illustrating an overall process for evaluating underwater combat effectiveness in a complex acoustic propagation environment, as provided in an embodiment of this application. Figure 2 A flowchart illustrating the construction process of the acoustic channel state matrix provided in one embodiment of this application; Figure 3 A flowchart illustrating the generation of a sonar performance surface provided in one embodiment of this application; Figure 4 A flowchart illustrating the confidence distribution of enemy tactical intentions according to an embodiment of this application; Figure 5 This is a flowchart illustrating the interaction between environmental tactical coupling operators and the joint state representation update process, as provided in an embodiment of this application. Figure 6 A flowchart illustrating the allocation of time-varying weight coefficients according to an embodiment of this application; Figure 7 A flowchart of underwater combat digital twin and Monte Carlo statistical simulation provided for an embodiment of this application.

[0017] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] The main solution in this application's embodiments is: A method for evaluating underwater combat effectiveness in complex acoustic propagation environments includes: S1. Obtain multi-source marine environmental parameters of the target sea area to construct an initial parameter set, use a hybrid algorithm to calculate the time-varying three-dimensional sound field propagation loss field, and establish a time-varying impulse response function to construct a spatiotemporally related acoustic channel state matrix. S2. Based on the acoustic channel state matrix and propagation loss field, calculate the time-varying detection threshold, construct a processing gain correction model to effectively process the gain, establish the time-varying sonar equation, solve it, and generate the sonar performance surface. S3. Construct a joint state representation based on the sonar performance surface and the acoustic channel state matrix, build a tactical action library and a partially observable Markov decision process model, and output the confidence distribution of the enemy's tactical intentions. S4. Establish an environment-tactical coupling operator based on the tactical intent confidence distribution and the tactical action library, and output the updated joint state representation. S5. Extract the bottom-level indicators from the updated joint state representation, calculate the middle-level indicators based on the tactical intent confidence distribution, and calculate the environmental complexity factor based on the bottom-level indicators, middle-level indicators and the acoustic channel state matrix, and allocate weights to determine the time-varying weight coefficients. S6. Establish a hierarchical effectiveness index mapping function based on the time-varying weight coefficient, output the time-varying effectiveness evaluation sequence, and construct an underwater combat digital twin based on the updated joint state representation to generate the statistical distribution and confidence interval of the time-varying effectiveness evaluation sequence. S7. Quantify the uncertainty of the statistical distribution and establish a closed loop. Update the parameters of the hybrid algorithm based on the deviation between the measured and evaluated results.

[0019] This application provides a method for evaluating underwater combat effectiveness under complex acoustic propagation environments. By introducing a hybrid algorithm combining parabolic equations and coupled normal mode wave models, and integrating stochastic phase screen theory, it can accurately calculate the time-varying three-dimensional sound field propagation loss field and time-varying impulse response function, constructing a spatiotemporally correlated acoustic channel state matrix, thereby achieving high-precision dynamic modeling of complex acoustic propagation environments. By constructing an environmental tactical coupling operator, the influence of enemy tactical intent confidence distribution and tactical actions on the acoustic channel state is quantitatively coupled, achieving dynamic closed-loop feedback between tactical decision-making and environmental perception, overcoming the shortcomings of traditional evaluation methods that separate environmental and tactical factors. A hierarchical effectiveness index mapping function is adopted, combined with environmental complexity factors to dynamically adjust the weight coefficients of bottom-level and mid-level indicators, enabling the evaluation results to adapt to the operational mission requirements under different environmental conditions. By constructing an underwater combat digital twin and integrating multi-agent reinforcement learning and Monte Carlo simulation, the statistical distribution and confidence interval of the effectiveness evaluation values ​​are generated, realizing the quantification of the uncertainty of the evaluation results. A closed-loop correction mechanism based on the deviation between measured and evaluated data was established. Uncertainty was quantified using a Bayesian neural network, and a hierarchical online parameter update mechanism was adopted to achieve continuous optimization and adaptive calibration of the evaluation model.

[0020] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can understand it.

[0021] refer to Figure 1 This application provides an embodiment of a method for evaluating underwater combat effectiveness in complex acoustic propagation environments. The following will describe in detail each step of the method provided in this embodiment of the invention, in conjunction with specific scenarios.

[0022] See Figure 2 The present invention provides a method for evaluating underwater combat effectiveness in complex acoustic propagation environments, wherein step S1 specifically includes the following sub-steps: S101. Obtain multi-source marine environmental parameters of the target sea area and construct an initial parameter set.

[0023] The sound propagation characteristics of the target sea area are affected by multiple factors, including temperature, salinity, depth, seabed sediment, and marine environmental noise. To accurately characterize the spatiotemporal changes of the sound field, this application obtains environmental parameters through various means: acquiring temperature and salinity profile data of the target sea area using marine observation stations, unmanned underwater vehicles, or satellite remote sensing; obtaining seabed sediment parameters through seabed survey databases; and acquiring local environmental noise spectral levels through real-time measurements using sonar systems. These parameters constitute the initial environmental parameter set.

[0024] S102. Based on the initial parameter set, a hybrid algorithm combining parabolic equations and coupled normal mode wave models is used to calculate the time-varying three-dimensional sound field propagation loss field.

[0025] Hybrid algorithms refer to acoustic field calculation methods that dynamically select or weight and fuse parabolic equation models and coupled normal mode wave models based on a joint criterion of frequency and seabed slope. When the signal frequency... And slope Time, or frequency A parabolic equation model is adopted. The mathematical form of the parabolic equation is: in, Sound pressure level; Local wavenumber; Using the reference wavenumber, the parabolic equation is numerically solved using the split-step Fourier transform method to calculate the sound propagation loss in this region.

[0026] When the signal frequency At that time, the coupled normal mode model was used to calculate the sound field distribution. The mathematical form of the coupled normal mode model is: in, For the first The mode function of the first normal mode; The corresponding horizontal wavenumber; is the coupling coefficient.

[0027] For frequencies located at And seabed slope In the transition region, a weighted average of the results from the two models is calculated. The formula for the weighted average in the transition region is: Among them, the weighting coefficient , This refers to the local seabed slope.

[0028] S103. Based on the time-varying three-dimensional sound field propagation loss field, the random phase screen theory is introduced to parameterize the multipath delay spread and coherence bandwidth of the sound signal, and a time-varying impulse response function is established.

[0029] Random phase screen theory is an approximation method that discretizes the sound propagation path into multiple phase screens, each applying a random phase perturbation to the sound signal related to the inhomogeneity of the medium. When a sound signal propagates in a randomly inhomogeneous medium, it is modulated by random perturbations of the medium, causing random fluctuations in the signal phase. This embodiment introduces random phase screen theory to model this: the propagation path is discretized into multiple phase screens, each applying a random phase perturbation to the sound signal related to the inhomogeneity of the medium. The multipath arrival structure of the sound channel is extracted based on the propagation loss field. The delay spread parameter is obtained through statistical analysis of the multipath arrival delay, and the coherence bandwidth parameter is obtained through statistical analysis of the coherence bandwidth of different frequency components. Delay spread parameter The calculation formula is: in, The time-varying impulse response function is derived from the propagation loss field. Determined in conjunction with phase fluctuations: in, For the first The amplitude of each vocal syllable; This corresponds to the propagation delay; This represents the total number of vocal timbres.

[0030] Coherence bandwidth parameters The calculation formula is: S104. Based on the time-varying impulse response function, construct the acoustic channel state matrix with spatiotemporal correlation.

[0031] Based on the time-varying impulse response function, acoustic channel feature parameters at different spatiotemporal points in the target sea area are extracted to form an acoustic channel feature parameter set. According to the spatiotemporal correlation of the marine environment, an acoustic channel state matrix is ​​constructed with time and space as two-dimensional row / column dimensions and acoustic channel feature parameters as matrix elements.

[0032] See Figure 3 The present invention provides a method for evaluating underwater combat effectiveness in complex acoustic propagation environments. Step S2 specifically includes the following sub-steps: S201. Based on the time-varying three-dimensional sound field propagation loss field and the sound channel state matrix, extract the propagation loss value and channel coherence characteristic parameters at the target location.

[0033] In the acoustic channel state matrix In this process, spatial interpolation is performed based on the target's spatial coordinates at the current moment to extract the target's position. Corresponding propagation loss value Simultaneously, the coherence bandwidth and delay spread parameters at that location are extracted from the channel characteristic axis as channel coherence characteristic parameters. The propagation loss value is used to calculate the sonar received signal strength, and the channel coherence characteristic parameters are used to evaluate the correction amount for the signal processing gain.

[0034] S202. Calculate the time-varying detection threshold of the sonar based on the propagation loss value, the local noise field measurement value, and the sonar system parameters.

[0035] Time-varying detection threshold The calculation formula is: in, For detection index; The center frequency; For signal bandwidth; This represents the probability of a false alarm. This represents the detection probability.

[0036] S203. Based on the channel coherence characteristic parameters, construct a signal processing gain model, use subspace projection filtering to suppress reverberation, and calculate the processing gain.

[0037] Traditional sonar signal processing assumes an ideal channel, but real underwater acoustic channels suffer from multipath effects and time-varying characteristics, leading to matched filter mismatch and array gain degradation. This embodiment constructs a signal processing gain model and employs subspace projection filtering to project the received signal into a signal subspace, suppressing the reverberation subspace orthogonal to the signal. Subspace projection filtering is a reverberation suppression algorithm based on the orthogonality of signal and noise subspaces. By performing feature decomposition on the received signal, the reverberant signal is projected into the noise subspace, while the target signal remains in the target subspace, thus achieving reverberation suppression. (Projection matrix) Determined by the eigenvalue decomposition of the channel covariance matrix. Coherence gain loss factor. Bandwidth mismatch loss factor The calculation formula is: in, The signal pulse width; This refers to the signal bandwidth.

[0038] Processing gain The calculation formula is: in, For ideal processing gain; This represents the subspace projection gain.

[0039] S204. Based on the time-varying detection threshold and processing gain, establish the time-varying sonar equation and solve for the detection probability, maximum effective distance and ranging error standard deviation to generate the sonar performance surface.

[0040] The standard form of the time-varying sonar equation is: in, It can be active or passive sonar; To receive the directional index.

[0041] Processing gain Substituting into the time-varying sonar equations, for a given detection threshold, we can solve the time-varying sonar equations to obtain the detection probability at each grid point. The formula for calculating the detection probability is: in, For Markum's Q function; The signal-to-noise ratio is time-varying.

[0042] Maximum effective range By solving The distance was obtained using a binary search algorithm with a precision of 100 meters. The standard deviation of the distance measurement error was... The calculation formula is: The detection probability, maximum effective range, and standard deviation of ranging error are organized according to time frames and spatial coordinates to generate a four-dimensional tensor structure sonar performance surface. .

[0043] See Figure 4 The present invention provides a method for evaluating underwater combat effectiveness in complex acoustic propagation environments. Step S3 specifically includes the following sub-steps: S301. Construct a joint state representation based on the sonar performance surface and the acoustic channel state matrix.

[0044] The sonar performance surface reflects the temporal and spatial distribution of detection capability, while the acoustic channel state matrix reflects the physical characteristics of the acoustic propagation environment. Joint state representation integrates these two: at each grid point of the acoustic channel state matrix, the corresponding sonar performance index is appended to form the joint state representation. The calculation formula is: in, This represents the platform's self-noise level. For feature splicing operators.

[0045] S302. Establish a tactical action library and construct a partially observable Markov decision process model based on joint state representation.

[0046] The tactical maneuver library contains typical tactical maneuvers for underwater combat platforms, including: depth maneuvers, speed maneuvers, turning maneuvers, and acoustic stealth measures. A partially observable Markov decision process (POMDP) ​​model is used to describe sequential decision problems in uncertain environments, where the state space... Joint state characterization Action space Select from the tactical action library; state transition probability. Determined by the environmental tactical coupling operator; observation space Detection probability distribution and range estimation extracted from sonar performance surface; observation probability Determined by the statistical characteristics of sonar detection; reward function Determined by a weighted average of mission completion rate and platform survival probability; discount factor The configuration is from 0.9 to 0.99.

[0047] S303. Based on the observation space of the POMDP model, the intention recognition network is used to make probabilistic inferences about the enemy's tactical behavior and output the tactical intention confidence distribution.

[0048] The intent recognition network employs a spatiotemporal graph neural network (ST-GNN) architecture. The node embedding layer encodes enemy movement trajectories, sonar activation timing, and communication behaviors as graph node features. A temporal attention mechanism captures the temporal dependencies of tactical actions, with 4 to 8 attention heads configured. The graph attention layer infers multi-target collaborative relationships and outputs the probability distribution of intent categories. (Tactical intent confidence distribution) The calculation formula is: in, This represents the pooled feature vector output by the spatiotemporal graph neural network. For the first Classification weights for class intents; This represents the total number of intent categories.

[0049] See Figure 5 The present invention provides a method for evaluating underwater combat effectiveness in complex acoustic propagation environments. Step S4 specifically includes the following sub-steps: S401. Analyze the impact of tactical actions on the sound propagation environment and sound channel characteristics, and quantify the changes of each tactical action on each element in the sound channel state matrix.

[0050] The environmental impacts of tactical maneuvers include self-noise radiation, depth variations, and changes in sonar operating modes. The quantification formula for these impacts is: in, for; For depth; For sonar source level, Indicates action The resulting change in parameters.

[0051] S402. Based on the confidence distribution of tactical intentions and the quantitative results of the impact of tactical actions, establish an environmental tactical coupling operator.

[0052] Environmental tactical coupling operator The mathematical form is: The specific calculation process is as follows: in, To select a strategy based on the current state; For action The resulting change in state; For intentional threat modification; This is an indicator function.

[0053] S403. Apply the environmental tactical coupling operator to the original acoustic channel state matrix to obtain the updated acoustic channel state matrix. Based on the real-time update results of the sonar performance surface, output the updated joint state characterization.

[0054] The updated joint state characterization includes the updated sonar performance surface. Updated audio channel state matrix Temporal evolution of the confidence distribution of tactical intentions .

[0055] See Figure 6 The present invention provides a method for evaluating underwater combat effectiveness in complex acoustic propagation environments. Step S5 specifically includes the following sub-steps: S501. Based on the updated joint state characterization, the detection probability, maximum effective distance, and standard deviation of ranging error are extracted from the sonar performance surface as underlying indicators.

[0056] The underlying metrics directly reflect the current state of detection capability. These metrics include detection probability, maximum effective range, and standard deviation of ranging error.

[0057] S502. Calculate the enemy's tactical threat level based on the confidence distribution of tactical intent; calculate the degree of influence of tactical actions on the environment based on the strength of the environmental tactical coupling operator, and use the enemy's tactical threat level, the degree of environmental influence of tactical actions, and the degree of completion of combat missions as mid-level indicators.

[0058] Combat mission completion rate Determined by mission phase and objective achievement rate. Enemy tactical threat level. The calculation formula is: in, For the first Threat weight of class intent.

[0059] The degree of influence of the tactical action environment The calculation formula is: in, It is the Frobenius norm.

[0060] S503. Calculate the environment complexity factor based on the acoustic channel state matrix.

[0061] Environmental complexity factor The calculation formula is: in, To propagate the spatial gradient of the loss function; The signal coherence time; The variance of the noise field; Ocean front indicator function; weighting coefficients , , , Determined using the entropy weight method, satisfying... .

[0062] S504. Based on the bottom-level and middle-level indicators, the evidence theory and the analytic hierarchy process are used to allocate mixed weights, and the mixed weights are dynamically adjusted according to the environmental complexity factor to determine the final time-varying weight coefficients.

[0063] The mixed weight allocation adopts a combination of evidence theory and analytic hierarchy process (AHP). First, AHP is used to construct pairwise comparison judgment matrices of bottom-level and middle-level indicators based on expert experience to calculate subjective weights. Second, evidence theory is used to integrate the confidence levels of different indicators in historical data to calculate objective weights. Finally, the final mixed weights are obtained by dynamically adjusting the weights through linear weighting or multiplication.

[0064] The dynamic adjustment formula is: After normalization, the time-varying weighting coefficients are obtained, which satisfy... .

[0065] See Figure 7The present invention provides a method for evaluating underwater combat effectiveness in complex acoustic propagation environments. Step S6 specifically includes the following sub-steps: S601. Establish a hierarchical performance index mapping function based on the bottom-level indicators, middle-level indicators and time-varying weight coefficients.

[0066] Hierarchical efficiency index mapping function for: The efficiency index for each layer is calculated using Sigmoid normalization: in, This is the gain coefficient; The midpoint of the threshold.

[0067] S602. The effectiveness index at different times is calculated using a hierarchical effectiveness index mapping function. The effectiveness indices at each time point are arranged in a time series, and a time-varying effectiveness assessment sequence is output. An underwater combat digital twin is constructed, and the time-varying effectiveness assessment sequence and the updated joint state representation are integrated into it.

[0068] Time-varying performance evaluation sequence It reflects the dynamic evolution of combat effectiveness over time, covering the entire combat mission cycle, with a time resolution configured from 1 minute to 10 minutes, dynamically adjusted according to the rate of change in the tactical situation. The digital twin includes a physical entity mirror, an environmental mirror, and an agent model. The physical entity mirror includes sonar parameters, platform kinematics models, and weapon system states. The environmental mirror is a synchronized acoustic channel state matrix. The agent model is an autonomous decision-making model based on POMDP.

[0069] S603. Construct a multi-agent reinforcement learning model in an underwater combat digital twin. Use a time-varying performance evaluation sequence as the reward function and conduct multiple simulations using the Monte Carlo simulation method.

[0070] The multi-agent reinforcement learning model employs the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm. The red agent makes decisions using the platform's sonar performance surface and joint state representation, while the blue agent simulates behavior using tactical intent confidence distribution. The training process uses a centralized training and distributed execution framework: during training, agents share global information for policy optimization; during execution, agents rely only on local observations for decision-making. Monte Carlo simulations employ a Latin hypercube sampling strategy, performing hierarchical sampling of ocean environmental parameters, target motion parameters, and sensor measurement noise. Each simulation independently generates a set of random parameter combinations, and multiple simulations are performed on the same random parameter combination to obtain a statistical average. Each simulation generates a complete time-varying performance evaluation sequence. In this embodiment, the number of simulations is 1000.

[0071] S604. Perform statistical analysis on the time-varying performance evaluation sequence obtained from multiple simulations to generate the statistical distribution and confidence interval of the performance evaluation values.

[0072] Statistical analysis was performed on the time-varying performance evaluation sequences obtained from 1000 Monte Carlo simulations, and the mean performance index at each time point was calculated. with standard deviation : Calculate the confidence interval for the performance evaluation value at a 95% confidence level: The present invention provides a method for evaluating underwater combat effectiveness in complex acoustic propagation environments, wherein step S7 specifically includes the following sub-steps: S701. Construct a Bayesian neural network model, using the statistical distribution of the time-varying performance evaluation sequence as input and the evaluation uncertainty as output, and train the model.

[0073] Bayesian neural networks, building upon traditional neural networks, treat weight parameters as random variables rather than deterministic values, approximating the posterior distribution through variational inference or Monte Carlo dropout. This embodiment employs the Monte Carlo dropout method: maintaining dropout activation during inference, performing multiple forward propagations on the same input to obtain the output distribution. The model uses the statistical distribution of the time-varying performance evaluation sequence as input features and evaluation uncertainty as the output label, trained using historical simulation data.

[0074] S702. Input the statistical distribution obtained from the Monte Carlo derivation into the trained Bayesian neural network, perform uncertainty quantification, and output the evaluation confidence score. The formula for calculating the evaluation confidence score is: in, The standard deviation of the predictions output by the Bayesian neural network; This is for predicting the mean.

[0075] S703 uses a neural network to extract the parameters that have the greatest impact on the performance evaluation results, forming a set of key sensitive factors. The measured data of the parameters in the key sensitive factor set are then compared with the corresponding parameter evaluation values ​​in the acoustic channel state matrix to perform deviation analysis, and the deviation rate of each parameter is calculated.

[0076] Sensitive factor extraction employs the Layer-by-Layer Correlation Propagation (LRP) algorithm: in, , For connection weights, This represents the neuron activation value.

[0077] The formula for calculating the deviation rate is: in, These are measured values; These are the model's predicted values; To predict uncertainty.

[0078] S704. Set a deviation threshold. When the deviation rate of any parameter exceeds the threshold, or the average deviation rate of multiple parameters exceeds the threshold, trigger the online parameter update mechanism to adjust the model parameters of the hybrid algorithm of parabolic equation and coupled normal mode wave model.

[0079] A slight deviation is defined as a difference between measured and evaluated values ​​exceeding 3 dB for 10 minutes. This triggers an incremental update of the online adapter, employing a resilient weight consolidation algorithm to retain historical knowledge and update only a subset of parameters relevant to the current environment. For example, only the interpolation nodes for the sound velocity profile at depths between 50 and 200 meters are adjusted, while parameters for deeper nodes remain unchanged. The resilient weight consolidation algorithm measures the importance of each parameter using a Fisher information matrix, imposing greater constraints on important parameters during the update process to prevent catastrophic forgetting. A severe deviation is defined as a difference between measured and evaluated values ​​exceeding 6 dB. This triggers a hybrid algorithm grid parameter adjustment, recalculating the sound velocity profile interpolation nodes and appropriately densifying or sparsening the computational grid.

[0080] S705. The updated hybrid algorithm parameters are reused to calculate the time-varying three-dimensional sound field propagation loss field, completing the closed loop.

[0081] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0082] In the description of this application, it should be noted that the terms "first", "second", and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0083] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0084] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0085] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0086] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0087] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the technical scope disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

[0088] Furthermore, although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

Claims

1. A method for evaluating underwater combat effectiveness under complex acoustic propagation environments, characterized in that, include: S1. Obtain multi-source marine environmental parameters of the target sea area to construct an initial parameter set, use a hybrid algorithm to calculate the time-varying three-dimensional sound field propagation loss field, and establish a time-varying impulse response function to construct a spatiotemporally related acoustic channel state matrix. S2. Based on the acoustic channel state matrix and the propagation loss field, calculate the time-varying detection threshold, construct a signal processing gain model, calculate the processing gain, establish and solve the time-varying sonar equation, and generate the sonar performance surface. S3. Construct a joint state representation based on the sonar performance surface and the acoustic channel state matrix, build a tactical action library and a partially observable Markov decision process model, and output the confidence distribution of the enemy's tactical intentions. S4. Establish an environmental tactical coupling operator based on the tactical intent confidence distribution and the tactical action library, and output the updated joint state representation. S5. Extract the bottom-level indicators from the updated joint state representation, calculate the middle-level indicators according to the tactical intent confidence distribution, and calculate the environmental complexity factor according to the bottom-level indicators, the middle-level indicators and the acoustic channel state matrix, and allocate weights to determine the time-varying weight coefficients. S6. Establish a hierarchical effectiveness index mapping function based on the time-varying weight coefficients, output a time-varying effectiveness evaluation sequence, and construct an underwater combat digital twin based on the updated joint state representation to generate the statistical distribution and confidence interval of the time-varying effectiveness evaluation sequence. S7. Quantify the uncertainty of the statistical distribution and establish a closed loop. Update the parameters of the hybrid algorithm based on the deviation between the measured and evaluated results.

2. The underwater combat effectiveness evaluation method based on complex acoustic propagation environment according to claim 1, characterized in that, Step S1 includes: Obtain multi-source marine environmental parameters of the target sea area to construct an initial parameter set; Based on the initial parameter set, a hybrid algorithm combining parabolic equations and coupled normal mode wave models is used to calculate the time-varying three-dimensional sound field propagation loss field. Based on the time-varying three-dimensional sound field propagation loss field, the random phase screen theory is introduced to parameterize the multipath delay spread and coherence bandwidth of the sound signal, and a time-varying impulse response function is established. Based on the time-varying impulse response function, a spatiotemporally correlated acoustic channel state matrix is ​​constructed.

3. The underwater combat effectiveness evaluation method based on complex acoustic propagation environment according to claim 2, characterized in that, The model is dynamically selected based on the joint criterion of frequency gradient, specifically including: When the signal frequency At that time, a coupled normal mode model was used; When the signal frequency And slope Time, or frequency A parabolic equation model is adopted; For frequencies located at And seabed slope In the transition region, the weighted average of the calculation results of the coupled normal mode model and the parabolic equation model is used as the final sound field propagation loss value.

4. The underwater combat effectiveness evaluation method based on complex acoustic propagation environment according to claim 1, characterized in that, Step S2 includes: Based on the time-varying three-dimensional sound field propagation loss field and the sound channel state matrix, the propagation loss value and channel coherence characteristic parameters at the target location are extracted; Calculate the time-varying detection threshold of the sonar based on the propagation loss value; Based on the channel coherence characteristic parameters, a signal processing gain model is constructed, subspace projection filtering is used to suppress reverberation, and the effective processing gain is calculated. Based on the time-varying detection threshold and the effective processing gain, a time-varying sonar equation is established and the detection probability, maximum effective range, and standard deviation of ranging error are solved to generate a sonar performance surface.

5. The underwater combat effectiveness evaluation method based on complex acoustic propagation environment according to claim 1, characterized in that, Step S3 includes: A joint state representation is constructed based on the sonar performance surface and the acoustic channel state matrix; Establish a tactical action library and construct a partially observable Markov decision process model based on the joint state representation; Based on the observation space of the partially observable Markov decision process model, the intention recognition network is used to make probabilistic inferences about the enemy's tactical behavior and output the tactical intention confidence distribution.

6. The underwater combat effectiveness evaluation method based on complex acoustic propagation environment according to claim 1, characterized in that, Step S4 includes: Analyze the impact of tactical actions on the sound propagation environment and sound channel characteristics, and quantify the changes of each tactical action on each element in the sound channel state matrix. Based on the quantification results of the influence of the tactical intent confidence distribution and tactical actions, an environmental tactical coupling operator is established. The environmental tactical coupling operator is applied to the original acoustic channel state matrix to obtain the updated acoustic channel state matrix. Combined with the real-time update result of the sonar performance surface, the updated joint state representation is output.

7. The underwater combat effectiveness evaluation method based on complex acoustic propagation environment according to claim 1, characterized in that, Step S5 includes: Based on the updated joint state characterization, the detection probability, maximum effective range and the standard deviation of ranging error are extracted from the sonar performance surface as underlying indicators. Based on the confidence distribution of the tactical intent, the enemy's tactical threat level is calculated, and based on the strength of the environmental tactical coupling operator, the degree of influence of tactical actions on the environment is calculated. The enemy's tactical threat level, the degree of influence of the tactical action environment, and the degree of completion of the combat mission are used as mid-level indicators; Calculate the environmental complexity factor based on the acoustic channel state matrix; Based on the underlying and intermediate indicators, evidence theory and analytic hierarchy process are used to perform mixed weight allocation, and the mixed weights are dynamically adjusted according to the environmental complexity factor to determine the final time-varying weight coefficients.

8. The underwater combat effectiveness evaluation method based on complex acoustic propagation environment according to claim 7, characterized in that, Step S6 includes: A hierarchical performance index mapping function is established based on the bottom-level indicators, the middle-level indicators, and the time-varying weight coefficients. The efficiency index at different times is calculated by the hierarchical efficiency index mapping function, and the efficiency index at each time is arranged in time series to output the time-varying efficiency evaluation sequence. Construct an underwater combat digital twin and integrate it with the time-varying performance evaluation sequence and the updated joint state representation; Construct a multi-agent reinforcement learning model in an underwater combat digital twin; Using the time-varying performance evaluation sequence as the reward function, multiple simulations were performed using the Monte Carlo simulation method. Statistical analysis is performed on the time-varying performance evaluation sequences obtained from multiple simulations to generate the statistical distribution and confidence intervals of the performance evaluation values.

9. The underwater combat effectiveness evaluation method based on complex acoustic propagation environment according to claim 8, characterized in that, Step S7 includes: A Bayesian neural network model is constructed, with the statistical distribution of the time-varying performance evaluation sequence as input and the evaluation uncertainty as output, and the model is trained. The statistical distribution obtained from the Monte Carlo derivation is input into the trained Bayesian neural network, the uncertainty is quantified, and the evaluation confidence is output. Parameters that affect the performance evaluation results are extracted by using neural networks to form a set of key sensitive factors; The measured data of parameters in the set of key sensitive factors are compared with the corresponding parameter evaluation values ​​in the acoustic channel state matrix to perform deviation analysis and calculate the deviation rate of each parameter. A deviation threshold is set. When the deviation rate of any parameter exceeds the threshold, or the average deviation rate of multiple parameters exceeds the threshold, the online parameter update mechanism is triggered to adjust the model parameters of the hybrid algorithm of the parabolic equation sound propagation model and the coupled normal mode model. The updated hybrid algorithm was reused to calculate the time-varying 3D sound field propagation loss field.

10. The underwater combat effectiveness evaluation method based on complex acoustic propagation environment according to claim 9, characterized in that, The online parameter update mechanism adopts a tiered triggering mechanism: A deviation of >3dB between measured and evaluated values ​​for 10 minutes is defined as a slight deviation, and only the depth layer parameter subset of the sound velocity profile interpolation node is updated. If the deviation between the measured and evaluated values ​​is greater than 6 dB, it is considered a severe deviation, and the sound velocity profile interpolation nodes need to be recalculated.

Images