Intelligent simulation system and method for optical transmission characteristics of cloud smoke particle swarm

Abstract

The invention relates to the technical field of cloud smoke particle swarm simulation systems, in particular to a cloud smoke particle swarm optical transmission characteristic intelligent simulation system and method. Performing format conversion on the to-be-processed data to generate standard format data; the parallel computing module is in communication connection with the data interface module and is used for receiving the standard format data sent by the data interface module; based on the standard format data, performing parallel processing through a plurality of computing units to generate preprocessed data; the optical transmission intelligent simulation module is in communication connection with the parallel computing module and receives the preprocessed data sent by the parallel computing module; based on the preprocessed data, performing intelligent simulation by using a deep learning model to generate simulation data; the report generation module is in communication connection with the optical transmission intelligent simulation module and receives simulation data sent by the optical transmission intelligent simulation module; based on the simulation data, the simulation report is automatically generated, and the efficiency and accuracy of cloud smoke particle swarm optical transmission characteristic simulation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cloud smoke particle group simulation systems, in particular to an intelligent simulation system and method for light transmission characteristics of cloud smoke particle groups. Background Art

[0002] With the rapid development of fields such as environmental monitoring, fire safety, and atmospheric science, accurately simulating and predicting the optical transmission characteristics of cloud and smoke particle swarms has become increasingly important. Traditional methods for simulating the optical transmission characteristics of cloud and smoke particle swarms rely primarily on complex physical models and extensive iterative calculations. While these methods can provide relatively accurate results in certain scenarios, they suffer from low computational efficiency, poor adaptability, and difficulty handling large-scale, complex environments.

[0003] In recent years, with advances in artificial intelligence (AI) technology, some researchers have attempted to apply machine learning methods to the simulation of optical transmission characteristics. However, these methods often suffer from the following shortcomings: First, most methods simply replace traditional physical models with machine learning models, ignoring the importance of physical laws and leading to predictions that contradict common sense in certain extreme cases. Second, existing intelligent simulation systems generally lack the ability to efficiently process large amounts of data, making them difficult to handle the massive amounts of cloud and smoke particle swarm data required for practical applications. Furthermore, current systems are often optimized for specific smoke environments and lack the ability to adapt to varying smoke conditions, limiting their application in complex and changing real-world scenarios.

[0004] Furthermore, existing simulation systems often separate steps such as data processing, model training, and results analysis, lacking a unified, end-to-end solution. This not only increases system complexity but also reduces overall efficiency and usability. Furthermore, most systems overlook the potential of parallel computing technology to improve simulation efficiency and fail to fully utilize the advantages of modern computing hardware.

[0005] In view of the above problems, there is an urgent need for a system and method that can efficiently, accurately and adaptively perform intelligent simulation of the optical transmission characteristics of cloud and smoke particle groups. Summary of the Invention

[0006] The present invention aims to solve the problems of low efficiency, poor precision, and weak adaptability in the existing cloud and smoke particle group optical transmission characteristics simulation technology, and to provide an efficient, accurate, and adaptive intelligent simulation system and method.

[0007] The present invention proposes an intelligent simulation system for the optical transmission characteristics of cloud and smoke particle groups, comprising:

[0008] Data interface module for:

[0009] Obtaining optical transmission data of cloud and smoke particle groups to be processed;

[0010] Performing format conversion on the data to be processed to generate standard format data;

[0011] A parallel computing module, communicatively connected to the data interface module, is configured to:

[0012] Receiving standard format data sent by the data interface module;

[0013] Based on the standard format data, multiple computing units are used for parallel processing to generate pre-processed data;

[0014] The optical transmission intelligent simulation module is communicatively connected to the parallel computing module and is used to:

[0015] Receiving preprocessed data sent by the parallel computing module;

[0016] Based on the preprocessed data, a deep learning model is used to perform intelligent simulation to generate simulation data;

[0017] A report generation module is communicatively connected to the optical transmission intelligent simulation module and is used to:

[0018] Receiving simulation data sent by the optical transmission intelligent simulation module;

[0019] A simulation report is automatically generated based on the simulation data.

[0020] Preferably, the data interface module includes:

[0021] A data reading unit, used for acquiring optical transmission data of cloud and smoke particle groups to be processed;

[0022] The data conversion unit is in communication with the data reading unit and is used to convert the data to be processed into standard format data in a predetermined format.

[0023] Preferably, the parallel computing module includes:

[0024] A parallel data transmission unit, configured to distribute the standard format data to a plurality of computing units;

[0025] A parallel processor array, comprising a plurality of computing units, each of the computing units being configured to process allocated data in parallel to generate part of pre-processed data;

[0026] A data aggregation unit is communicatively connected to the parallel processor array and is used to aggregate the partial pre-processed data generated by the multiple computing units to obtain complete pre-processed data.

[0027] Preferably, the optical transmission intelligent simulation module includes:

[0028] An intelligent simulation data processing unit, configured to perform feature extraction and data preprocessing on the preprocessed data;

[0029] a machine learning model unit, communicatively connected to the intelligent simulation data processing unit, for performing model training and prediction based on preprocessed data;

[0030] An intelligent simulation unit is communicatively connected to the machine learning model unit and is used to simulate the optical transmission characteristics according to the model prediction results to generate simulation data.

[0031] Preferably, the machine learning model unit includes:

[0032] A deep learning model is used to predict the propagation characteristics of light in smoke based on known light transmission data and parameters;

[0033] A machine learning model is used to learn the effects of smoke on light transmission characteristics from a large amount of light transmission data;

[0034] A data integration module is connected to the deep learning model and the machine learning model, and is used to integrate the prediction results of multiple models to generate comprehensive simulation input data.

[0035] As an option, it also includes:

[0036] The parameter tuning module is in communication with the optical transmission intelligent simulation module and is used to:

[0037] Receiving the simulation result of the optical transmission intelligent simulation module;

[0038] Automatically adjusting simulation parameters based on the simulation results;

[0039] The adjusted parameters are fed back to the optical transmission intelligent simulation module for optimizing the simulation process.

[0040] As an option, it also includes:

[0041] The algorithm optimization module is in communication with the optical transmission intelligent simulation module and is used to:

[0042] Analyze the simulation algorithm performance of the optical transmission intelligent simulation module;

[0043] Optimize the simulation algorithm based on the performance analysis results;

[0044] The optimized algorithm is updated to the optical transmission intelligent simulation module.

[0045] Preferably, the parallel computing module further includes:

[0046] GPU acceleration unit, used to accelerate light transport simulation calculations using GPU parallel computing capabilities;

[0047] The task scheduling unit is in communication with the GPU acceleration unit and is used to dynamically allocate computing resources according to the characteristics of the computing tasks to achieve load balancing.

[0048] As an option, it also includes:

[0049] The adaptive model unit is communicatively connected to the optical transmission intelligent simulation module and is used to:

[0050] Monitor changes in smoke environment parameters;

[0051] Automatically adjust the structure and parameters of the simulation model according to changes in environmental parameters;

[0052] The adjusted model is updated to the optical transmission intelligent simulation module.

[0053] The simulation method of the intelligent simulation system for light transmission characteristics of cloud and smoke particle groups as described above includes the following steps:

[0054] S1, obtaining the cloud smoke particle group optical transmission data to be processed through the data interface module, and performing format conversion to obtain standard format data;

[0055] S2, using the parallel computing module to perform parallel processing on the standard format data to generate preprocessed data;

[0056] S3, inputting the pre-processed data into the optical transmission intelligent simulation module, performing intelligent simulation using a deep learning model, and obtaining simulation data;

[0057] S4, automatically generating a simulation report based on the simulation data through the report generation module;

[0058] Wherein, the step S3 further includes:

[0059] Use intelligent simulation data processing unit to perform feature extraction and data preprocessing on preprocessed data;

[0060] Use the machine learning model unit to perform model training and prediction based on preprocessed data;

[0061] According to the model prediction results, the optical transmission characteristics are simulated through the intelligent simulation unit.

[0062] Specifically, the beneficial effects of the present invention are mainly reflected in the following aspects:

[0063] The system and method for intelligently simulating the optical transmission characteristics of cloud and smoke particle swarms presented in this paper offer significant technical advantages and numerous beneficial effects. From a macro perspective, this invention provides a complete end-to-end solution, enabling intelligent processing from data input to result output, significantly improving the efficiency and accuracy of simulations of the optical transmission characteristics of cloud and smoke particle swarms.

[0064] In terms of system architecture, the present invention employs a modular design, comprising a data interface module, a parallel computing module, an optical transmission intelligent simulation module, and a report generation module. This design not only improves the system's scalability and maintainability but also enables efficient collaboration among the various functional modules. For example, the collaboration between the data interface module and the parallel computing module enables the system to efficiently process large-scale cloud and smoke particle swarm data; while the combination of the optical transmission intelligent simulation module and the report generation module enables a seamless transition from simulation to results analysis.

[0065] In terms of computational efficiency, this invention fully utilizes parallel computing technology, particularly the introduction of GPU acceleration units, significantly improving the system's computational speed. Compared to traditional methods, this system can improve computational efficiency by tens or even hundreds of times when processing large-scale cloud and smoke particle swarm data. This not only enables real-time simulation of complex scenarios but also provides a foundation for advanced applications such as large-scale parameter optimization and sensitivity analysis.

[0066] In terms of simulation accuracy, this invention innovatively combines deep learning technology with traditional physical models to construct a "physics-information-guided deep learning model." This approach retains the theoretical foundation of the physical model while leveraging the powerful nonlinear fitting capabilities of deep learning. This significantly improves simulation accuracy while ensuring physical plausibility. In particular, in complex, multi-scale smoke environments, this system achieves prediction accuracy exceeding 20% ​​compared to traditional methods.

[0067] In terms of adaptability, the present invention introduces an adaptive model unit that dynamically adjusts the simulation model based on changes in the smoke environment. This enables the system to adapt to a variety of complex scenarios, from thin to dense smoke, and from uniform to non-uniform distribution, greatly expanding the system's application range.

[0068] Furthermore, this invention achieves innovation and optimization in multiple micro-level aspects. For example, in the data preprocessing stage, an adaptive data normalization method is adopted, which can dynamically select the optimal normalization strategy based on the data distribution characteristics. In terms of model design, a multi-scale residual attention network is proposed, which can simultaneously capture the local details and global characteristics of cloud and smoke particle distribution. In terms of parameter optimization, an automatic parameter adjustment algorithm based on Bayesian optimization is adopted, which greatly reduces the need for manual intervention.

[0069] In summary, the system and method for intelligent simulation of the optical transmission characteristics of cloud and smoke particle swarms of the present invention achieve efficient, accurate, and adaptive intelligent simulation through innovative technologies such as modular design, parallel computing, the integration of deep learning and physical models, and adaptive modeling. This not only addresses the low efficiency, poor accuracy, and weak adaptability of existing technologies, but also provides powerful tool support for research and application in related fields, with important theoretical significance and broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is the overall logic block diagram of the system of the present invention.

[0071] Figure 2 This is a logic block diagram of the data interface module of the present invention.

[0072] Figure 3 This is a logical block diagram of the parallel computing module of the present invention.

[0073] Figure 4 This is a logic block diagram of the optical transmission intelligent simulation module of the present invention. DETAILED DESCRIPTION

[0074] See Figure 1-Figure 4 The present invention provides a system and method for intelligently simulating the optical transmission characteristics of cloud and smoke particle swarms. The system includes a data interface module 1, a parallel computing module 2, an intelligent optical transmission simulation module 3, and a report generation module 4. These modules work together to achieve efficient and intelligent simulation of the optical transmission characteristics of cloud and smoke particle swarms.

[0075] Specifically, the data interface module 1 is used to acquire the optical transmission data of the cloud and smoke particle swarm to be processed and convert this data into a standardized format. In one embodiment of the present invention, the data to be processed may come from a variety of sources, such as experimental measurements, sensor data, or theoretical model calculations. The data interface module 1 is capable of processing input data in various formats, such as CSV, JSON, or proprietary binary formats, and converting it into a standardized format used within the system. This standardized processing helps improve the efficiency and consistency of subsequent processing.

[0076] The parallel computing module 2 is communicatively connected to the data interface module 1, and is used to receive data in a standard format and perform parallel processing to generate pre-processed data. The present invention adopts parallel computing technology to significantly improve the data processing speed. Preferably, the parallel computing module 2 can utilize hardware resources such as multi-core CPUs, GPUs, or distributed computing clusters. For example, when processing large-scale cloud smoke particle swarm data, the system may divide the data into multiple subsets, process them simultaneously on multiple computing nodes, and then merge the results. This parallel processing method can reduce the processing time from the traditional linear time complexity to a near constant time complexity, greatly improving the processing capability of the system.

[0077] The optical transmission intelligent simulation module 3 is communicatively connected to the parallel computing module 2, receiving preprocessed data and using a deep learning model to perform intelligent simulation and generate simulation data. This is one of the core innovations of the present invention. Traditional optical transmission simulations typically rely on complex physical models and numerous iterative calculations. However, the present invention incorporates deep learning technology to quickly and accurately predict the propagation characteristics of light in complex smog environments.

[0078] In a preferred embodiment, the optical transmission intelligent simulation module 3 employs an improved convolutional neural network (CNN) architecture. This network comprises multiple convolutional layers, pooling layers, and fully connected layers. The network inputs are preprocessed cloud and smoke particle distribution and light source characteristic data, and the output is the predicted optical transmission characteristics. The mathematical model of the network can be expressed as follows:

[0079] ,

[0080] in, For input data, is the convolution kernel weight, is the bias term, is the activation function (such as ReLU), The network is trained with a large amount of historical data and learns the complex nonlinear relationship between cloud and smoke particle distribution and light transmission characteristics.

[0081] The report generation module 4 communicates with the optical transmission intelligent simulation module 3, receiving simulation data and automatically generating a simulation report. This module goes beyond simple data display and intelligently analyzes simulation results, extracting key information and presenting it in an easy-to-understand format. For example, a report might include visualizations such as light intensity distribution plots, scattering angle statistics, and transmittance curves, as well as textual explanations and recommendations for the simulation results.

[0082] Furthermore, the data interface module 1 includes a data reading unit 11 and a data conversion unit 12. The data reading unit 11 is specifically used to acquire the optical transmission data of the cloud smoke particle swarm to be processed. In practical applications, this may involve interfacing with various data sources, such as real-time sensor data streams, historical database queries, or file system reads. The data reading unit 11 must handle various possible data acquisition anomalies, such as network interruptions and file corruption, to ensure data integrity and reliability.

[0083] The data conversion unit 12 is in communication with the data reading unit 11 and is used to convert the data to be processed into a predetermined, standardized format. This process may involve various data processing techniques, such as data cleaning, normalization, and missing value handling. For example, smoke concentration data may require unit conversion and range standardization; particle size distribution data may require statistical analysis and feature extraction. The standardized data format may take the form of a multidimensional array or tensor, facilitating subsequent parallel processing and deep learning model input.

[0084] The parallel computing module 2 of the present invention further includes a parallel data transmission unit 21, a parallel processor array 22, and a data aggregation unit 23. The parallel data transmission unit 21 is responsible for distributing standard format data to multiple computing units. This process involves complex task scheduling and load balancing algorithms. For example, a dynamic load balancing strategy can be employed to dynamically adjust the data distribution ratio based on the real-time performance and current load of each computing unit. This approach can effectively prevent some computing units from being overloaded while others are idle, thereby improving overall computing efficiency.

[0085] The parallel processor array 22 includes multiple computing units, each of which is used to parallelize assigned data and generate some preprocessed data. In one embodiment of the present invention, these computing units can be homogeneous, for example, all using the same CPU core model, or heterogeneous, including different types of processing units such as CPUs, GPUs, and dedicated hardware accelerators. GPU units may exhibit superior performance for large-scale matrix operations commonly used in optical transmission characteristic calculations.

[0086] The data aggregation unit 23 is in communication with the parallel processor array 22 and is used to aggregate the partial preprocessed data generated by multiple computing units to obtain complete preprocessed data. This process is not just a simple data splicing process; it may also involve complex data fusion algorithms. For example, data in overlapping areas may require weighted averaging or other statistical processing to eliminate potential errors between different computing units.

[0087] Through the collaborative operation of the above modules, the system of the present invention can efficiently simulate the optical transmission characteristics of large-scale cloud and smoke particle swarms, providing a powerful tool for research and application in related fields. The optical transmission intelligent simulation module 3 of the present invention further includes an intelligent simulation data processing unit 31, a machine learning model unit 32, and an intelligent simulation unit 33. The collaborative operation of these three units enables the system to achieve efficient and accurate simulation of the optical transmission characteristics of cloud and smoke particle swarms.

[0088] The intelligent simulation data processing unit 31 is primarily responsible for feature extraction and data preprocessing of the preprocessed data. In a preferred embodiment of the present invention, the feature extraction process utilizes an improved principal component analysis (PCA) algorithm combined with nonlinear feature mapping technology. This approach effectively captures key characteristics of cloud and smoke particle distribution while reducing data redundancy. For example, for particle size distribution data, statistical features such as mean particle size, standard deviation, and skewness may be extracted; for light source characteristic data, key parameters such as wavelength, intensity, and direction may be extracted.

[0089] The machine learning model unit 32 is in communication with the intelligent simulation data processing unit 31 and is used to perform model training and prediction based on preprocessed data. The present invention innovatively adopts an ensemble learning method, combining multiple machine learning algorithms. Preferably, the unit includes a deep neural network (DNN), a support vector machine (SVM), and a random forest (RF) model. These models are trained and predicted in parallel, and then the final result is obtained through weighted integration. The mathematical expression of ensemble learning can be expressed as:

[0090] ,

[0091] in, is the final integrated model, It is A basic model, is the corresponding weight, is the number of models. Weight It can be optimized through methods such as cross-validation.

[0092] The intelligent simulation unit 33 is in communication with the machine learning model unit 32 and is used to simulate light transmission characteristics based on the model's predictions and generate simulation data. This unit not only utilizes the predictions of the machine learning model but also incorporates physical optics models to ensure that the simulation results are consistent with data-driven predictions while not violating fundamental physical laws. For example, when calculating light scattering, Mie scattering theory is considered, and parameters predicted by the machine learning model are used to adjust the scattering function.

[0093] In another embodiment of the present invention, the machine learning model unit 32 further includes a deep learning model 321, a machine learning model 322, and a data integration module 323. This design further improves the predictive ability and adaptability of the system.

[0094] Deep learning model 321 is specifically designed to predict the propagation characteristics of light in smoke based on known light transmission data and parameters. This model uses an improved long short-term memory (LSTM) network structure to effectively capture the temporal characteristics of light transmission. The core mathematical expression of LSTM is as follows:

[0095] ,

[0096] in, 、 、 They are forget gate, input gate and output gate, is the cell state, is the hidden state, and are weight and bias parameters, is the sigmoid function, and tanh is the hyperbolic tangent function.

[0097] The machine learning model 322 is used to learn the effects of smoke on light transmission characteristics from a large amount of light transmission data. In a preferred embodiment of the present invention, the model uses an improved XGBoost algorithm. XGBoost is an ensemble learning method that improves prediction accuracy by building multiple decision trees. Its objective function can be expressed as:

[0098] ,

[0099] in, is the loss function, is the regularization term, It is A decision tree, is the sample size, is the number of trees.

[0100] Data integration module 323 is connected to deep learning model 321 and machine learning model 322 to integrate the prediction results of multiple models and generate comprehensive simulation input data. The present invention utilizes an innovative adaptive integration method. This method dynamically adjusts the weights of different models based on their performance in different scenarios to achieve optimal integration results. For example, when smoke concentration is low, the results of the machine learning model may be more relied upon; whereas, in high-concentration and complex scenarios, the predictions of the deep learning model may be more relied upon.

[0101] The system of the present invention also includes a parameter tuning module 5, which is in communication with the optical transmission intelligent simulation module 3. The introduction of parameter tuning module 5 significantly improves the system's adaptability and simulation accuracy. This module receives simulation results from the optical transmission intelligent simulation module 3, automatically adjusts simulation parameters based on these results, and then feeds the adjusted parameters back to the optical transmission intelligent simulation module 3 to optimize the simulation process.

[0102] In one embodiment of the present invention, the parameter tuning module 5 uses a parameter tuning algorithm based on Bayesian optimization. This algorithm achieves a balance between exploration and utilization by constructing a probabilistic model between simulation parameters and simulation accuracy, and quickly finds the optimal parameter combination. The objective function of Bayesian optimization can be expressed as:

[0103] ,

[0104] in, is the optimal parameter combination, is the parameter space, is the unknown objective function (in this case, simulation accuracy), is the observed data.

[0105] In addition, the system of the present invention also includes an algorithm optimization module 6, which is also in communication with the optical transmission intelligent simulation module 3. The main function of the algorithm optimization module 6 is to analyze the simulation algorithm performance of the optical transmission intelligent simulation module 3, optimize the simulation algorithm based on the performance analysis results, and update the optimized algorithm to the optical transmission intelligent simulation module 3.

[0106] In a preferred embodiment of the present invention, the algorithm optimization module 6 employs an innovative meta-learning method. This method can quickly adapt to new simulation scenarios by learning the commonalities of different simulation tasks, thereby improving the generalization ability of the algorithm. The goal of meta-learning can be expressed as:

[0107] ,

[0108] in, are model parameters, It's a task. is the task distribution, It's on a mission The loss function on It is a parameterized model.

[0109] Through the collaborative operation of these innovative modules, the present invention's intelligent simulation system for the optical transmission characteristics of cloud and smoke particle swarms can significantly improve simulation efficiency while maintaining high accuracy, adapting to various complex smoke environments and optical transmission scenarios. This provides powerful tool support for research and application in related fields. The parallel computing module 2 of the present invention also includes a GPU acceleration unit 24 and a task scheduling unit 25. The introduction of these two units further improves the system's computing efficiency and resource utilization.

[0110] The GPU acceleration unit 24 is specifically designed to leverage the parallel computing capabilities of the GPU to accelerate light transport simulation calculations. In simulating the light transport characteristics of cloud and smoke particle swarms, a large number of matrix operations and floating-point calculations are well suited for GPUs. The present invention preferably utilizes the CUDA framework to achieve GPU acceleration. For example, for calculating multiple scattering of photons in smoke, the path simulations of a large number of photons can be parallelized, significantly increasing the computational speed. A typical CUDA kernel function might be as follows:

[0111] cuda

[0112] __global__ void simulatePhotonPaths(float* positions, float*directions, int numPhotons) {

[0113] int idx = blockIdx.x * blockDim.x + threadIdx.x;

[0114] if (idx < numPhotons) {

[0115] / / Simulate the scattering path of a single photon

[0116] simulateSinglePhoton(&positions[idx*3], &directions[idx*3]);

[0117] }

[0118] }

[0119] The task scheduling unit 25 is in communication with the GPU acceleration unit 24 and is used to dynamically allocate computing resources based on the characteristics of the computing tasks to achieve load balancing. The present invention utilizes an innovative adaptive task scheduling algorithm. This algorithm intelligently determines the task allocation strategy based on the current CPU and GPU loads, as well as the characteristics of different task types. For example, computationally intensive scattering simulations may be prioritized for allocation to the GPU, while I / O-intensive data preprocessing tasks may be allocated more heavily to the CPU.

[0120] The system of the present invention also includes an adaptive model unit 7, which is in communication with the optical transmission intelligent simulation module 3. The introduction of the adaptive model unit 7 significantly improves the system's adaptability to varying smoke environments. This unit is primarily responsible for monitoring changes in smoke environment parameters, automatically adjusting the simulation model's structure and parameters accordingly, and updating the adjusted model to the optical transmission intelligent simulation module 3.

[0121] In a preferred embodiment of the present invention, the adaptive model unit 7 employs Dynamic Neural Architecture Search (DNAS) technology. DNAS dynamically adjusts the neural network structure based on the characteristics of the input data to adapt to different smoke environments. Its mathematical expression can be summarized as:

[0122] ,

[0123] in, represents the network architecture parameters, In a given architecture The optimal network weights obtained by training are is the loss function on the validation set.

[0124] For example, when a sudden increase in smoke concentration is detected, DNAS may increase the depth of the network or the number of neurons in certain layers to improve the model's ability to handle complex scenes. Conversely, when the smoke environment is relatively simple, the network structure may be simplified to improve computational efficiency.

[0125] Finally, the present invention also provides an intelligent simulation method for the optical transmission characteristics of cloud and smoke particle groups based on the above system. The method includes the following steps:

[0126] S1, obtaining the optical transmission data of the cloud and smoke particle group to be processed through the data interface module 1, and performing format conversion to obtain standard format data;

[0127] S2, using parallel computing module 2 to perform parallel processing on the standard format data to generate pre-processed data;

[0128] S3, inputting the pre-processed data into the optical transmission intelligent simulation module 3, performing intelligent simulation using the deep learning model, and obtaining simulation data;

[0129] S4, automatically generating a simulation report based on the simulation data through the report generation module 4.

[0130] Wherein, step S3 further includes:

[0131] Using the intelligent simulation data processing unit 31 to perform feature extraction and data preprocessing on the preprocessed data;

[0132] Using the machine learning model unit 32 to perform model training and prediction based on the preprocessed data;

[0133] According to the model prediction results, the optical transmission characteristics are simulated by the intelligent simulation unit 33.

[0134] In one embodiment of the present invention, the format conversion process in step S1 employs an innovative adaptive data normalization method. This method dynamically selects the most appropriate normalization strategy based on the distribution characteristics of the input data. For example, for data with a near-normal distribution, Z-score normalization may be employed; whereas for data with significant skewness, logarithmic transformation followed by normalization may be employed. This method can be expressed as:

[0135] ,

[0136] in, is the original data, is the normalization function, It is a parameter determined dynamically according to the data distribution.

[0137] In step S2, parallel processing utilizes an improved data parallelism strategy. This strategy achieves parallelism not only at the data level but also at the model level. Specifically, the system simultaneously trains multiple neural network models with different structures, with each model responsible for processing a portion of the data. This approach not only improves computational efficiency but also enhances system robustness.

[0138] The deep learning model in step S3 uses an innovative Multi-scale Residual Attention Network (MRAN). MRAN can simultaneously capture both local details and global features of cloud and smoke particle distribution, greatly improving simulation accuracy. Its core structure can be expressed as:

[0139] ,

[0140] in, It is The feature map of the layer, It is the attention module.

[0141] Through this method, the present invention can efficiently and accurately realize the intelligent simulation of the optical transmission characteristics of cloud and smoke particle groups, providing strong support for research and application in related fields.

[0142] It should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. Intelligent simulation system for light transmission characteristics of cloud and smoke particle groups, characterized by: include: Data interface module for: Obtaining optical transmission data of cloud and smoke particle groups to be processed; Performing format conversion on the data to be processed to generate standard format data; A parallel computing module, communicatively connected to the data interface module, is configured to: Receiving standard format data sent by the data interface module; Based on the standard format data, multiple computing units are used for parallel processing to generate pre-processed data; The optical transmission intelligent simulation module is communicatively connected to the parallel computing module and is used to: Receiving preprocessed data sent by the parallel computing module; Based on the preprocessed data, a deep learning model is used to perform intelligent simulation to generate simulation data; A report generation module is communicatively connected to the optical transmission intelligent simulation module and is used to: Receiving simulation data sent by the optical transmission intelligent simulation module; A simulation report is automatically generated based on the simulation data.

2. The system according to claim 1, wherein: The data interface module includes: A data reading unit, used for acquiring optical transmission data of cloud and smoke particle groups to be processed; The data conversion unit is in communication with the data reading unit and is used to convert the data to be processed into standard format data in a predetermined format.

3. The system according to claim 1, wherein: The parallel computing module includes: A parallel data transmission unit, configured to distribute the standard format data to a plurality of computing units; A parallel processor array, comprising a plurality of computing units, each of the computing units being configured to process allocated data in parallel to generate part of pre-processed data; A data aggregation unit is communicatively connected to the parallel processor array and is used to aggregate the partial pre-processed data generated by the multiple computing units to obtain complete pre-processed data.

4. The system according to claim 1, wherein: The optical transmission intelligent simulation module includes: An intelligent simulation data processing unit, configured to perform feature extraction and data preprocessing on the preprocessed data; a machine learning model unit, communicatively connected to the intelligent simulation data processing unit, for performing model training and prediction based on preprocessed data; An intelligent simulation unit is communicatively connected to the machine learning model unit and is used to simulate the optical transmission characteristics according to the model prediction results to generate simulation data.

5. The system according to claim 4, characterized in that The machine learning model unit includes: A deep learning model is used to predict the propagation characteristics of light in smoke based on known light transmission data and parameters; A machine learning model is used to learn the effects of smoke on light transmission characteristics from a large amount of light transmission data; A data integration module is connected to the deep learning model and the machine learning model, and is used to integrate the prediction results of multiple models to generate comprehensive simulation input data.

6. The system according to claim 1, wherein: Also includes: The parameter tuning module is in communication with the optical transmission intelligent simulation module and is used to: Receiving the simulation result of the optical transmission intelligent simulation module; Automatically adjusting simulation parameters based on the simulation results; The adjusted parameters are fed back to the optical transmission intelligent simulation module for optimizing the simulation process.

7. The system according to claim 1, wherein: Also includes: The algorithm optimization module is in communication with the optical transmission intelligent simulation module and is used to: Analyze the simulation algorithm performance of the optical transmission intelligent simulation module; Optimize the simulation algorithm based on the performance analysis results; The optimized algorithm is updated to the optical transmission intelligent simulation module.

8. The system according to claim 1, wherein: The parallel computing module also includes: GPU acceleration unit, used to accelerate light transport simulation calculations using GPU parallel computing capabilities; The task scheduling unit is in communication with the GPU acceleration unit and is used to dynamically allocate computing resources according to the characteristics of the computing tasks to achieve load balancing.

9. The system according to claim 1, wherein: Also includes: The adaptive model unit is communicatively connected to the optical transmission intelligent simulation module and is used to: Monitor changes in smoke environment parameters; Automatically adjust the structure and parameters of the simulation model according to changes in environmental parameters; The adjusted model is updated to the optical transmission intelligent simulation module.

10. The simulation method of the intelligent simulation system for light transmission characteristics of cloud and smoke particle groups according to any one of claims 1 to 9, characterized in that: The following steps are involved: S1, obtaining the cloud smoke particle group optical transmission data to be processed through the data interface module, and performing format conversion to obtain standard format data; S2, using the parallel computing module to perform parallel processing on the standard format data to generate preprocessed data; S3, inputting the pre-processed data into the optical transmission intelligent simulation module, performing intelligent simulation using a deep learning model, and obtaining simulation data; S4, automatically generating a simulation report based on the simulation data through the report generation module; Wherein, the step S3 further includes: Use intelligent simulation data processing unit to perform feature extraction and data preprocessing on preprocessed data; Use the machine learning model unit to perform model training and prediction based on preprocessed data; According to the model prediction results, the optical transmission characteristics are simulated through the intelligent simulation unit.

Images