Composite solid propellant time-temperature equivalent prediction method and system based on deep learning
Through a hybrid deep learning network model combining CNN and LSTM, the problem of time-consuming and insufficient accuracy of time-temperature equivalent prediction of traditional solid propellants is solved, efficient and accurate prediction is achieved, the design and storage of propellants are optimized, and research efficiency and system applicability are improved.
Patent Information
- Application Number
- CN202510591813.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-09
- Publication Date
- 2025-08-29
AI Technical Summary
The traditional time-temperature equivalent prediction method of solid propellant is time-consuming, costly and insufficient prediction accuracy. The existing deep learning models have challenges in model structure selection and parameter optimization, which affects the practicality and accuracy of prediction.
A hybrid deep learning network model combining convolutional neural network (CNN) and long and short-term memory network (LSTM) is used to build a prediction system with high nonlinear mapping capabilities through data collection, preprocessing, model training and prediction, and optimize model parameters to improve accuracy and generalization capabilities.
It significantly improves the accuracy of time temperature equivalent prediction of composite solid propellants, shortens the research cycle, reduces R&D costs, optimizes the design and storage stability of propellants, and improves the research efficiency and the ease of realization and scalability of the system.
Smart Images
Figure CN120564893A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of solid propellants, and specifically relates to a method and system for predicting the time-temperature equivalence of composite solid propellants based on deep learning. Background Art
[0002] In the field of solid propellants, long-term storage stability is a key performance indicator. The Time-Temperature Superposition (TTS) principle is an effective method for predicting propellant performance changes during long-term storage. This principle converts short-term experimental data at different temperatures to the same reference temperature, thereby predicting the propellant's performance under long-term storage. This principle is crucial for the design, manufacture, and quality control of solid propellants.
[0003] However, traditional time-temperature equivalence principle prediction methods rely primarily on experimental testing, which often requires long-term storage and testing of propellant samples at different temperatures. This process is not only time-consuming (possibly requiring months or even years) but also costly. Furthermore, operational errors and environmental variations that may occur during the experiment can also affect the accuracy of the prediction results.
[0004] In recent years, the rapid development of artificial intelligence (AI), particularly the widespread application of deep learning in image recognition, natural language processing, and speech recognition, has opened up new possibilities for traditional prediction methods. Deep learning, a powerful machine learning method, builds multi-layered neural network models to learn the inherent patterns and characteristics of data and has demonstrated remarkable predictive capabilities in many fields.
[0005] In the field of solid propellants, the use of deep learning technology to predict the time-temperature equivalence principle has become a new research hotspot. Deep learning models can process large amounts of experimental data and identify complex data relationships, thereby improving the accuracy and efficiency of predictions while reducing the number of experiments and shortening the research cycle. Despite this, current deep learning-based time-temperature equivalence principle prediction models still face several challenges, such as model structure selection, parameter optimization, and overfitting prevention. Addressing these issues is crucial to improving the practicality and scalability of prediction models. Therefore, developing a more efficient and accurate deep learning-based method for predicting the time-temperature equivalence principle of composite solid propellants has important practical significance and application value for technological advancement in the field of solid propellants. Summary of the Invention
[0006] This invention addresses the existing challenges of predicting the time-temperature equivalence principle of composite solid propellants, such as long experimental cycles, high costs, and insufficient prediction accuracy. By proposing a deep learning-based method and system for predicting the time-temperature equivalence principle of composite solid propellants, this method utilizes deep learning techniques such as convolutional neural networks (CNNs) and long short-term memory networks (LSTMs) to construct a hybrid deep learning network model with highly nonlinear mapping capabilities. This model effectively learns from a large amount of historical experimental data the performance patterns of composite solid propellants under different temperature and time conditions, thereby accurately predicting the time-temperature equivalence principle of composite solid propellants. This invention not only significantly improves prediction accuracy but also significantly shortens research cycles and reduces R&D costs. Furthermore, this method and system provide reliable technical support for solid propellant research and development, and have important practical value and theoretical significance for optimizing solid propellant design and improving its storage stability and service life. This invention has broad application prospects in the solid propellant industry and will play a positive role in promoting the development of solid propellant technology in my country.
[0007] The technical solutions of the present invention are as follows:
[0008] A deep learning-based time-temperature equivalent prediction method for composite solid propellants includes the following steps: data collection, data preprocessing, construction of a hybrid deep learning network model, model training, and prediction. The model combines the advantages of convolutional neural networks (CNNs) and long short-term memory networks (LSTMs) to learn and predict the performance changes of composite solid propellants under different temperature and time conditions.
[0009] Furthermore, the data preprocessing step includes data cleaning, normalization, and feature selection and dimensionality reduction based on data characteristics to improve the training efficiency and prediction performance of the hybrid deep learning network model.
[0010] Furthermore, the construction module of the hybrid deep learning network model further includes: adjusting the number of network layers, the number of neurons, and optimizing algorithm parameters to minimize prediction errors and ensure high accuracy of the model.
[0011] Furthermore, the model training step includes adjusting model parameters using the back-propagation algorithm and gradient descent method, while preventing overfitting and improving the generalization ability of the model through learning rate scheduling, batch normalization and dropout technology.
[0012] A deep learning-based time-temperature equivalent prediction system for composite solid propellants includes: a data acquisition module, a data preprocessing module, a hybrid deep learning network model construction module, a model training module, and a prediction module, which is used to predict the performance of composite solid propellants under different conditions. The data acquisition module can automatically collect performance parameters of composite solid propellants under various temperature and time conditions, including but not limited to viscoelasticity, mechanical properties, thermal stability, and combustion performance, and ensure the accuracy and real-time nature of the data.
[0013] Furthermore, the hybrid deep learning network model building module can adaptively adjust the model structure, including the hierarchical configuration of convolutional neural networks (CNNs) and long short-term memory networks (LSTMs), to improve the accuracy of the model's prediction of composite solid propellant performance.
[0014] Furthermore, the prediction module can receive new input data, use the trained hybrid deep learning network model to predict the time-temperature equivalence principle of composite solid propellant, and output accurate prediction results of propellant performance under future conditions.
[0015] Furthermore, it also includes a model optimization module for monitoring the performance of the hybrid deep learning network model and optimizing the model by adjusting parameters, adding regularization terms or changing the network architecture to improve its prediction accuracy and stability.
[0016] The advantages and beneficial effects of the present invention are as follows:
[0017] The present invention has the following advantages:
[0018] The prediction module's high precision provides accurate and reliable predictions, which are crucial for guiding solid propellant research and development. This high accuracy means researchers and engineers can more confidently rely on the model's performance predictions, enabling them to make more scientific and rational decisions regarding the design, synthesis, storage, and use of solid propellants. This precise prediction capability helps optimize propellant formulations and improve performance, while also effectively reducing R&D costs and risks, accelerating the pace of innovation and practical application of solid propellant technology.
[0019] Deep learning models with high predictive accuracy can significantly reduce experimental time and costs, thereby improving research efficiency. By accurately predicting the performance of solid propellants under different conditions, researchers can reduce their reliance on traditional experiments and avoid repeated and redundant testing. This not only accelerates research progress but also significantly reduces the material and human resources required for experiments. Furthermore, through computer simulation and prediction, researchers can quickly evaluate the performance of multiple propellant formulations and quickly screen the most promising candidates, further accelerating the solid propellant R&D cycle and improving overall research efficiency.
[0020] ● The system was designed with ease of implementation and scalability in mind, enabling it to excel in practical applications. Ease of implementation means the system features a clear architecture and well-designed modularity, making it easy for technical personnel to quickly deploy and operate. This design philosophy reduces the complexity of the system, making it easy for even non-professionals to get started, significantly shortening the time it takes to get the system up and running. At the same time, the system is highly scalable and can adapt to the needs of future technological developments. As scientific research needs continue to change and technology advances, the system can easily add new functional modules, integrate more advanced algorithms, or process larger data sets. This flexibility ensures that the system can keep pace with the times, meet long-term research and development needs, and provide a solid support platform for research in solid propellants and even other fields of materials science. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 This is a flow chart of the time-temperature equivalence principle method of composite solid propellant using a hybrid deep learning network in a preferred embodiment provided by the present invention. DETAILED DESCRIPTION
[0022] The following will describe the technical solutions in the embodiments of the present invention in detail with reference to the accompanying drawings. The described embodiments are only a part of the embodiments of the present invention.
[0023] The technical solution of the present invention to solve the above technical problems is:
[0024] A method for predicting the time-temperature equivalence principle of composite solid propellants based on a hybrid deep learning network includes the following steps:
[0025] (1) Data collection: In the initial stage of the invention, we focused on systematically collecting and organizing experimental data of composite solid propellants, which is the basis for building an efficient prediction model. We recorded in detail the performance parameters of the propellant under different temperature and storage time conditions, including but not limited to key indicators such as viscoelasticity, mechanical properties, thermal stability, and combustion performance. In order to ensure the comprehensiveness and accuracy of the data, we took the following measures: First, we covered a wide temperature range from low temperature to high temperature, as well as multiple time points from short term to long term, in order to accurately capture the changes in propellant performance. Second, we considered the effects of different storage environments (such as humidity and pressure) to ensure the diversity of experimental conditions. In addition, we conducted multiple tests on the same sample to ensure the repeatability and consistency of the data. This data collection method makes the dataset we obtained both comprehensive and reliable, providing solid data support for the subsequent training and prediction of the hybrid deep learning network model.
[0026] (2) Data preprocessing: After successfully collecting the experimental data of composite solid propellants, the next step is to strictly preprocess the data. Data preprocessing is a necessary step before model training, which includes but is not limited to data cleaning, outlier processing, normalization or standardization, etc. First, we thoroughly cleaned the data and removed abnormal data points caused by experimental errors or recording errors to eliminate the interference of noise on model training. Then, in order to enable performance parameters of different dimensions and ranges to be effectively processed by the model, we normalized the data and converted all parameters to the same scale, thereby improving the efficiency and stability of model training. In addition, we also performed appropriate transformation and segmentation on the data to ensure the reasonable distribution of the data set during training, validation and testing, which helps to evaluate the generalization ability of the model in practical applications. Through these preprocessing steps, we have made full preparations for the subsequent deep learning model training and ensured that the model can effectively learn and predict based on high-quality data.
[0027] (3) Constructing a hybrid deep learning network model: In order to fully utilize the potential of deep learning technology in the prediction of the time-temperature equivalence principle of composite solid propellants, we designed and constructed a hybrid deep learning network model that combines the advantages of convolutional neural networks (CNNs) and recurrent neural networks (RNNs). In this model, CNN is used to capture the spatial features and local patterns in the input data. It can identify the key structures and correlations in the data, which is crucial for understanding the intrinsic connection between the propellant performance parameters. At the same time, RNN, especially its variant long short-term memory network (LSTM), is responsible for processing time series data and can effectively capture the time dependence and long-term trends in the data, which is particularly critical for predicting the performance changes of propellants during long-term storage. By combining these two network structures, our hybrid deep learning network model can not only understand the complexity of the data from the spatial dimension, but also predict the evolution of performance from the temporal dimension, thereby providing a powerful learning framework for the prediction of the time-temperature equivalence principle of composite solid propellants. Such a model design makes the prediction results more accurate and reliable, providing advanced technical support for the research and development of solid propellants.
[0028] (4) Model training: After data preprocessing, we use these cleaned and normalized data to conduct detailed training on the constructed hybrid deep learning network model. During the training process, we use backpropagation and gradient descent algorithms to continuously adjust the network parameters to reduce the prediction error. In order to improve the generalization ability of the model and prevent overfitting, we divide the data into training sets and validation sets, and use techniques such as learning rate adjustment, batch normalization and dropout to optimize the training. After multiple iterations and performance evaluations, we gradually refine the model parameters until a satisfactory accuracy is achieved on the validation set. This step is time-consuming and resource-intensive, but the model finally trained can accurately predict the long-term storage performance of composite solid propellants, providing strong technical support for the research and application of propellants.
[0029] (5) Prediction: After successfully completing the training and optimization of the hybrid deep learning network model, we applied it to predict the time-temperature equivalence principle of composite solid propellants under unknown conditions. Using the trained model, we can input new data that represents the performance parameters of the propellant under different storage conditions. The model will output a prediction of the performance of the propellant under specific time-temperature conditions based on the learned complex patterns and relationships. These predictions are important for understanding the storage stability of propellants, evaluating their service life, and guiding the design and manufacture of propellants. In addition, in this way, we can provide researchers with a powerful tool to help them quickly and accurately predict the behavior of propellants under new or extreme conditions, thereby accelerating the development and application of solid propellant technology.
[0030] Example:
[0031] According to the above method, an embodiment based on the following system:
[0032] A time-temperature equivalence principle prediction system for composite solid propellants based on a hybrid deep learning network includes: (a) a data acquisition module: This module is the foundation of the entire system, and its main function is to collect experimental data on composite solid propellants. The design takes into account the diversity and accuracy of experimental data. The module can automatically collect data from different experimental settings, including but not limited to key indicators such as the viscoelasticity, mechanical properties, thermal stability, and combustion performance of the propellant. Through high-precision sensors and data processing equipment, the data acquisition module ensures the reliability and real-time nature of the collected data, providing a solid foundation for subsequent data analysis and model training. In addition, the module can also adjust the acquisition frequency and parameters as needed to adapt to different experimental requirements and conditions, thereby ensuring the flexibility and adaptability of the data acquisition process.
[0033] (b) Data Preprocessing Module: Immediately following data collection, the data preprocessing module is responsible for performing key processing tasks such as cleaning and normalizing the collected data. This module is designed to improve data quality and lay a solid foundation for model training. During the data cleaning phase, the module automatically identifies and corrects or removes outliers, missing values, and inconsistent data points to ensure data consistency and accuracy. Normalization converts performance parameters of different dimensions and ranges to a unified scale, enabling the model to learn and predict more effectively. In addition, the data preprocessing module may also include steps such as data standardization, feature selection, and dimensionality reduction to improve model training efficiency and predictive performance. Through these preprocessing steps, the data preprocessing module provides high-quality, consistently formatted, and easy-to-process data input for subsequent deep learning models. (c) Hybrid Deep Learning Network Model Construction Module: This module is responsible for building and optimizing a hybrid deep learning network model that combines a convolutional neural network (CNN) and a recurrent neural network (RNN). In this module, users can design a network structure tailored to the characteristics of propellant performance data and automatically adjust model parameters to minimize prediction errors. It not only supports model visualization and debugging, but also provides model evaluation tools to help monitor and improve model performance, thereby building a powerful model that can simultaneously capture spatial characteristics and dynamic changes in time series, providing technical support for the accurate prediction of propellant performance.
[0034] (d) Prediction Module: This module is the final step in the system and is specifically designed to perform prediction tasks based on the time-temperature equivalence principle for composite solid propellants. By loading a trained hybrid deep learning network model, the prediction module is able to receive new input data representing the propellant performance parameters under different storage conditions. The module will use the model to analyze this data and output accurate predictions of the propellant's performance under future storage conditions based on the complex patterns and relationships previously learned. This function is crucial for evaluating the long-term storage stability and safe use of propellants, providing researchers and engineers with fast and reliable decision-making support, further promoting the advancement and application of solid propellant technology.
[0035] advantage:
[0036] The present invention has the following advantages:
[0037] The prediction module's high precision provides accurate and reliable predictions, which are crucial for guiding solid propellant research and development. This high accuracy means researchers and engineers can more confidently rely on the model's performance predictions, enabling them to make more scientific and rational decisions regarding the design, synthesis, storage, and use of solid propellants. This precise prediction capability helps optimize propellant formulations and improve performance, while also effectively reducing R&D costs and risks, accelerating the pace of innovation and practical application of solid propellant technology.
[0038] Deep learning models with high predictive accuracy can significantly reduce experimental time and costs, thereby improving research efficiency. By accurately predicting the performance of solid propellants under different conditions, researchers can reduce their reliance on traditional experiments and avoid repeated and redundant testing. This not only accelerates research progress but also significantly reduces the material and human resources required for experiments. Furthermore, through computer simulation and prediction, researchers can quickly evaluate the performance of multiple propellant formulations and quickly screen the most promising candidates, further accelerating the solid propellant R&D cycle and improving overall research efficiency.
[0039] ● The system was designed with ease of implementation and scalability in mind, enabling it to excel in practical applications. Ease of implementation means the system features a clear architecture and well-designed modularity, making it easy for technical personnel to quickly deploy and operate. This design philosophy reduces the complexity of the system, making it easy for even non-professionals to get started, significantly shortening the time it takes to get the system up and running. At the same time, the system is highly scalable and can adapt to the needs of future technological developments. As scientific research needs continue to change and technology advances, the system can easily add new functional modules, integrate more advanced algorithms, or process larger data sets. This flexibility ensures that the system can keep pace with the times, meet long-term research and development needs, and provide a solid support platform for research in solid propellants and even other fields of materials science.
[0040] In a specific embodiment, in order to further enrich and expand the application scope and functions of the composite solid propellant time-temperature equivalence principle prediction system based on the hybrid deep learning network, we can consider the following enhancements and innovations:
[0041] 1. Multimodal Data Fusion
[0042] In the data acquisition module, in addition to traditional propellant performance parameters, multimodal data such as hyperspectral imaging, microstructure images, and chemical composition analysis can be considered to provide more comprehensive input features. This multimodal data fusion can capture more dimensional information about propellant performance changes, further improving the accuracy and generalization capabilities of the prediction model.
[0043] 2. Adaptive Model Optimization
[0044] Within the hybrid deep learning network model building module, an adaptive model optimization algorithm is implemented, enabling dynamic adjustments to the network structure and parameters during model training. For example, dynamic learning rate adjustment, weight initialization strategies, and regularization techniques are used to address the complexity and variability of data from different propellant types. Furthermore, by incorporating advanced deep learning techniques such as transfer learning or reinforcement learning, the model can learn from existing related tasks, enabling it to more quickly achieve optimal performance when processing new composite solid propellants.
[0045] 3. Process automation and intelligent decision-making
[0046] The prediction module can be further upgraded to enable fully automated predictions, from data input to output, reducing human intervention and improving prediction efficiency. Furthermore, combined with expert systems or decision tree algorithms, the prediction module can provide users with intelligent decision-making recommendations based on prediction results, such as recommending optimal storage conditions and predicting service life. This enables researchers and engineers to make decisions more quickly and optimize the design and management of solid propellants.
[0047] 4. Integration of high-performance computing and cloud computing
[0048] Given the computationally intensive nature of deep learning model training and prediction, the system can be integrated with high-performance computing clusters or cloud computing platforms to leverage distributed computing resources to accelerate model training and data processing. This integration can significantly reduce computing time and improve model training efficiency. It also opens the door to processing larger datasets and more complex models, further enhancing the system's predictive capabilities and scientific research value.
[0049] 5. Dynamic updates and online learning
[0050] The system can be designed with dynamic updating and online learning capabilities, meaning it can receive new experimental data in real time and automatically update the model to adapt to dynamic changes in propellant properties. This design enables the system to continuously optimize its prediction performance, promptly reflect the latest progress in propellant research, and maintain the timeliness and accuracy of prediction results.
[0051] 6. Visualization and interactive interface
[0052] To improve the system's user-friendliness, a visualization interface can be developed to display propellant performance trends, model training status, prediction results, and their credibility. This interface allows users to intuitively monitor the entire prediction process, understand the model's working principles, and query and analyze prediction results, enhancing the system's transparency and operability.
[0053] 7. Forecast Error Evaluation and Feedback Mechanism
[0054] The prediction module incorporates a prediction error assessment and feedback mechanism that automatically analyzes discrepancies between predictions and actual data. This information is fed back into the model training phase for further optimization of model parameters and structure. This closed-loop optimization strategy continuously improves prediction accuracy, ensuring the system maintains stable and reliable performance in the complex and ever-changing propellant research environment.
[0055] Through the above expansion and innovation, the composite solid propellant time-temperature equivalence principle prediction system based on hybrid deep learning network will be able to better meet scientific research needs, accelerate technological innovation and application promotion in the field of solid propellants, and provide strong technical support for the development of aerospace, military, energy and other fields in my country and even the world.
[0056] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
[0057] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.
[0058] The above embodiments should be understood as merely illustrating the present invention and not as limiting the scope of protection of the present invention. After reading the contents of the present invention, technicians may make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.
Claims
1. A time-temperature equivalent prediction method for composite solid propellant based on deep learning, characterized in that: The following steps are involved: Data collection, data preprocessing, construction of a hybrid deep learning network model, model training, and prediction, wherein the model combines the advantages of convolutional neural networks (CNN) and long short-term memory networks (LSTM) to learn and predict the performance changes of composite solid propellants under different temperature and time conditions.
2. The prediction method according to claim 1, characterized in that The data preprocessing steps include data cleaning, normalization, feature selection and dimensionality reduction based on data characteristics to improve the training efficiency and prediction performance of the hybrid deep learning network model.
3. The prediction method according to claim 1, wherein: The building blocks of the hybrid deep learning network model further include: adjusting the number of network layers, the number of neurons, and optimizing algorithm parameters to minimize prediction errors and ensure high accuracy of the model.
4. The prediction method according to claim 1, wherein: The model training step includes adjusting model parameters using the back-propagation algorithm and gradient descent method, while preventing overfitting and improving the generalization ability of the model through learning rate scheduling, batch normalization and dropout technology.
5. A time-temperature equivalent prediction system for composite solid propellants based on deep learning, characterized in that: include: The data acquisition module, data preprocessing module, hybrid deep learning network model construction module, model training module and prediction module are used to realize the performance prediction of composite solid propellant under different conditions; the data acquisition module can automatically collect the performance parameters of composite solid propellant under various temperature and time conditions, including but not limited to viscoelasticity, mechanical properties, thermal stability and combustion performance, and ensure the accuracy and real-time nature of the data.
6. The prediction system according to claim 5, characterized in that The hybrid deep learning network model building module can adaptively adjust the model structure, including the hierarchical configuration of convolutional neural networks (CNNs) and long short-term memory networks (LSTMs), to improve the accuracy of the model's prediction of composite solid propellant performance.
7. The prediction system according to claim 5, characterized in that The prediction module can receive new input data, use the trained hybrid deep learning network model to predict the time-temperature equivalence principle of composite solid propellant, and output accurate prediction results of propellant performance under future conditions.
8. The prediction system according to claim 5, characterized in that It also includes a model optimization module for monitoring the performance of the hybrid deep learning network model and optimizing the model by adjusting parameters, adding regularization terms, or changing the network architecture to improve its prediction accuracy and stability.