Energy material processing parameter optimization method and system based on force-thermal coupling model

Abstract

The application discloses an energetic material processing parameter optimization method and system based on a force-heat coupling model, and relates to the technical field of intelligent simulation optimization based on specific calculation models. In view of the problems that 3D printing and mechanical processing data are disconnected, and common force-heat coupling models are not suitable for energetic materials in the prior art, the method of the application first collects 3D printing operation and initial state data of the grain, then constructs a force-heat two-way coupling intelligent calculation model special for energetic materials, adopts a Bayesian algorithm to multi-objective optimize processing parameters based on the model, and carries out model-driven closed-loop dynamic processing control. The method of the application realizes full data connection and adaptation to material characteristics, takes into account safety, precision and efficiency, and improves the processing stability and consistency of the energetic material.

Description

Technical Field

[0001] This application relates to the field of intelligent simulation optimization technology based on specific computational models, and in particular to a method for optimizing the machining parameters of energetic materials after 3D printing based on a force-thermal coupling model. Background Technology

[0002] Energetic materials, as core functional materials in the aerospace field, encompass key categories such as composite solid propellants and polymer-bonded explosives (PBX). Their processing quality directly determines the range, power, and operational safety of equipment. Energetic material processing is divided into two methods: traditional machining and the emerging additive manufacturing (3D printing). While traditional turning and milling processes offer controllable precision, they struggle to fabricate complex combustion profiles. 3D printing, on the other hand, achieves near-net-shape forming, solving the challenge of manufacturing complex structures. However, printed parts exhibit defects such as dimensional deviations, surface roughness, uneven initial internal stress, and micropores, necessitating subsequent precision machining (turning, milling, and combustion profile reshaping) to meet engineering application requirements.

[0003] In the field of 3D printing, existing technologies have developed mature data optimization solutions that enable multi-sensor data acquisition, data fusion analysis, and dynamic control of safety thresholds during the printing process, such as CN119871893A. These solutions effectively ensure the basic forming quality and production safety of energetic material printed parts. However, this technology only focuses on parameter control during the printing process and does not address post-printing machining procedures, thus failing to solve the problems of parameter adaptation and safety control in subsequent precision machining.

[0004] In the field of machining, parameter optimization methods based on force-thermal coupling models have been developed for common metals and polymer materials. These methods simulate the stress and temperature fields during the cutting process to achieve a reasonable match of machining parameters. However, energetic materials possess unique properties such as strong viscoelasticity, high mechanical sensitivity, low thermal stability, and low thermal conductivity. Existing force-thermal coupling models for common materials have significant limitations in their applicability. This is mainly because the properties of energetic materials differ significantly from those of common materials. The force-thermal coupling mechanism for common materials does not consider the dynamic influence of temperature on the material's mechanical properties or the reverse heating effect of stress deformation. If this mechanism is directly used in the machining of energetic materials, it can easily lead to safety risks such as hot spots, microcracks, and interface debonding during the machining process. Therefore, it cannot be directly applied to the machining of energetic materials after 3D printing.

[0005] Therefore, there is an urgent need to provide a method for optimizing the machining parameters after 3D printing of energetic materials that adapts to the special properties of energetic materials, connects with 3D printing, and achieves bidirectional force-thermal coupling. Summary of the Invention

[0006] This application provides a method and system for optimizing processing parameters of energetic materials based on a force-thermal coupling model, in order to solve the problem that the force-thermal coupling model for ordinary materials in the prior art is not suitable for the fine processing of energetic materials.

[0007] On the one hand, embodiments of this application provide a method for optimizing processing parameters of energetic materials based on a force-thermal coupling model, including the following steps: S1. Collect the operating parameters and initial state data of energetic material propellant during the 3D printing process, extract key features, and establish a 3D printing parameter feature set and a propellant initial state set. S2. Establish a force-thermal coupling model, which includes a viscoelastic force field sub-model, a thermal field sub-model, and a two-way coupling correlation module. S3. A composite loss function is used to train the force-thermal coupling model based on the 3D printing parameter feature set and the initial state set of the drug column. S4. Select optimization variables. Based on the trained force-thermal coupling model, with the safety of processing energetic materials as a constraint, and with the objectives of minimizing temperature rise, minimizing stress, optimizing accuracy and maximizing efficiency, use the Bayesian optimization algorithm to perform global optimal screening of the machining parameters in the optimization variables to obtain the optimal parameters. S5. Process energetic material propellant columns using optimal parameters, collect working condition data in real time during the processing, and use the working condition data to dynamically correct the model parameters and boundary conditions of the force-thermal coupling model. Based on the prediction results of the dynamically corrected force-thermal coupling model, combined with the working condition data, use the PID (proportional integral derivative) control algorithm to adaptively fine-tune the processing parameters. S6. Detect the performance data of the processed energetic material propellant column and feed the performance data back to the force-thermal coupling model to achieve iterative data updates.

[0008] On the other hand, embodiments of this application also provide a system for optimizing processing parameters of energetic materials based on a force-thermal coupling model, including: The feature extraction module is used to collect the operating parameters and initial state data of energetic material propellant during the 3D printing process, extract key features, and establish a 3D printing parameter feature set and a propellant initial state set. The model building module is used to build a force-thermal coupling model, which includes a viscoelastic force field sub-model, a thermal field sub-model, and a two-way coupling correlation module. The model training module is used to train the force-thermal coupling model based on the 3D printing parameter feature set and the initial state set of the drug grains using a composite loss function; The parameter selection module is used to select optimization variables. Based on the trained force-thermal coupling model, with the safety of processing energetic materials as a constraint, and with the objectives of minimizing temperature rise, minimizing stress, optimizing accuracy and maximizing efficiency, the Bayesian optimization algorithm is used to perform global optimal selection of the machining parameters in the optimization variables to obtain the optimal parameters. The parameter fine-tuning module is used to process energetic material propellant columns using optimal parameters, collect working condition data in real time during the processing, and dynamically correct the model parameters and boundary conditions of the force-thermal coupling model using the working condition data. Based on the prediction results of the dynamically corrected force-thermal coupling model, combined with the working condition data, the PID control algorithm is used to adaptively fine-tune the processing parameters. The data update module is used to detect the performance data of the processed energetic material propellant column and feed the performance data back to the force-thermal coupling model to achieve iterative data updates.

[0009] The method proposed in this application has the following advantages: 1. Achieving seamless integration between 3D printing and machining processes. This application uses key state parameters such as the operating parameters of the 3D printing process, the initial internal stress of the printed part, porosity, and the initial modulus of the material as the initial boundary conditions of the force-thermal coupling model. This solves the problems of existing machining models neglecting the initial printing state and having large deviations between the model and actual working conditions, significantly improving the simulation accuracy of the model and providing a reliable data foundation for optimizing machining parameters.

[0010] 2. A force-thermal coupling model specific to energetic materials was constructed to adapt to the material's unique properties. Addressing the strong viscoelasticity, high sensitivity, and low thermal stability characteristics of energetic materials, this application abandons the linear elastic constitutive assumption frequently used in existing technologies and introduces a viscoelastic constitutive equation to characterize the material's deformation characteristics. A two-way dynamic coupling mechanism between the force and thermal fields was established to characterize the effects of cutting stress-plastic deformation heat generation-temperature rise-material mechanical property decay-stress redistribution-further heat generation. Specific safety constraint thresholds for energetic materials were set (i.e., maximum temperature rise less than 35℃, maximum stress less than the material's yield strength), effectively avoiding safety risks such as hot spots, microcracks, and interface debonding during processing, thus solving the problem that existing models are not suitable for energetic materials.

[0011] 3. A model-driven multi-objective intelligent optimization algorithm is adopted to balance processing safety, forming accuracy, and processing efficiency. This application takes minimizing processing temperature rise, lowest residual stress, optimal dimensional accuracy, and highest processing efficiency as optimization objectives, and uses the safety threshold of energetic materials as a constraint. A Bayesian optimization algorithm is introduced to achieve global optimal selection of machining parameters, avoiding the blindness of traditional empirical trial-and-error methods. Under the premise of strictly ensuring processing safety, the dimensional accuracy and surface quality of the propellant grain are significantly improved, balancing the contradiction between safety, accuracy, and efficiency.

[0012] 4. A dynamic control system for the machining process is established, enabling adaptive real-time parameter adjustment. This application collects cutting force, machining temperature, vibration, and other operating condition data in real time during machining, feeding this data back to the force-thermal coupling model to dynamically correct model parameters and boundary conditions. This allows for real-time fine-tuning of machining parameters such as spindle speed and feed rate, achieving adaptive control of the machining process. Simultaneously, a multi-level safety warning mechanism is set up. When real-time operating parameters approach a safety threshold, an automatic warning is issued and parameter downgrade adjustments are triggered, forming a control system of simulation optimization, actual machining, data feedback, model correction, and parameter re-optimization, significantly improving the stability and safety of the machining process.

[0013] 5. Establish a data iterative optimization mechanism to improve processing consistency and stability. This application feeds back the performance data of the processed propellant grains, such as dimensional accuracy, surface quality, internal integrity, and combustion performance, to the force-thermal coupling model, enabling iterative updates to the model structure, parameter weights, and optimization strategies. This gradually improves the database of energetic material processing parameters, solves the problem of poor processing consistency caused by individual differences in 3D printed parts, and provides technical support for the large-scale and standardized production of energetic materials after 3D printing and machining. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A flowchart of a method for optimizing processing parameters of energetic materials based on a force-thermal coupling model, provided in an embodiment of this application. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] Figure 1 A flowchart illustrating a method for optimizing processing parameters of energetic materials based on a force-thermal coupling model, provided in this application embodiment. This application embodiment provides a method for optimizing processing parameters of energetic materials based on a force-thermal coupling model, including: S1. Collect the operating parameters and initial state data of energetic material propellant during the 3D printing process, extract key features, and establish a 3D printing parameter feature set and a propellant initial state set.

[0018] The goal of this step is to collect the operating parameters during the 3D printing process of energetic materials and the initial state data of the propellant grains after printing. By preprocessing to remove abnormal data and extracting key features, a standardized dataset is constructed to provide accurate initial input for the subsequent construction of the force-thermal coupling model.

[0019] Specifically, step S1 includes the following sub-contents: S11. For 3D printed parts of energetic materials such as composite solid propellants and PBX explosives, the data acquisition objects are divided into two categories: 3D printing operation parameters and initial state data of the propellant. Industrial-grade high-precision sensors and detection equipment are selected to ensure that the data acquisition accuracy meets the modeling requirements.

[0020] The operating parameters of 3D printing include material delivery flow rate, nozzle temperature, material viscosity, printing platform stress, printing chamber ambient temperature, printing chamber ambient humidity, and printing speed. These parameters are collected by turbine flow sensor, miniature thermocouple temperature sensor, vibration viscosity sensor, piezoelectric stress sensor, and integrated temperature and humidity sensor, respectively.

[0021] The initial state data of the propellant column includes dimensional accuracy, internal porosity, material homogeneity, initial internal stress, initial elastic modulus of the material, and initial shear modulus of the material. These data are detected by high-precision 3D scanner, micro-CT (computed tomography) equipment, electron microscope, blind hole stress tester, and dynamic mechanical analyzer, respectively.

[0022] S12 3D printing operating parameters are collected throughout the entire printing process. Sensors are installed at corresponding positions on the printing equipment to collect data in real time and transmit it to the local storage terminal via a high-speed data acquisition card. The sampling frequency is set to 10Hz to ensure that dynamic changes in parameters during the printing process are captured. Initial state data of the propellant column is collected after printing is completed and the propellant column has naturally cooled to room temperature (25℃±1℃). Each data point is measured three times, and the arithmetic mean is taken as the final test result to avoid the impact of single test errors on data reliability.

[0023] S13. Due to factors such as sensor noise, environmental interference, and equipment fluctuations, the original collected data may have problems such as outliers, noise interference, and inconsistent dimensions. Data quality needs to be improved through three preprocessing steps: filtering, outlier removal, and normalization.

[0024] The first step is noise filtering. A moving average filtering algorithm is used to reduce noise in the original data, with a filter window length set to 5. The filtered data is calculated using the following formula to eliminate high-frequency noise interference:

[0025] Where, x f (k) represents the filtered data at the k-th sampling point, N w The value represents the length of the filtering window, which is 5. x(t) represents the original data of the t-th sampling point.

[0026] The second step is outlier removal. The 3σ criterion is used to identify and remove outlier data. The mean and standard deviation of the data are calculated using the following two formulas. Data exceeding the range [μ−3σ, μ+3σ] are identified as outliers and removed:

[0027]

[0028] Where μ represents the mean of the filtered data, and N s σ represents the number of valid sampling points, and σ represents the standard deviation of the filtered data.

[0029] The third step is data normalization. Due to significant differences in the dimensions and numerical ranges of the parameters, to eliminate the influence of dimensions and improve the efficiency of subsequent model training, a linear normalization method is used to map all data to the [0, 1] interval. The calculation formula is as follows:

[0030] Where, x n x represents the normalized data. f The original data after filtering, x min x represents the minimum value of the filtered data. max This represents the maximum value of the filtered data.

[0031] S14. To reduce data dimensionality and highlight core influencing factors, key features are extracted from the preprocessed data to construct a 3D printing parameter feature set and a drug column initial state feature set.

[0032] The 3D printing parameter feature set includes temperature-viscosity change rate, stress distribution standard deviation, and flow rate-printing speed ratio. The temperature-viscosity change rate reflects the changes in material rheological properties during the printing process, the stress distribution standard deviation reflects the stress uniformity of the printed part, and the flow rate-printing speed ratio reflects the material delivery stability. The initial state feature set of the propellant grain includes the mean porosity, mean initial internal stress, mean initial elastic modulus, and mean initial shear modulus. These data characterize the initial level of internal defects and mechanical properties of the printed part.

[0033] The extracted key features were organized in the "sample-feature" format and divided into training set (70%), validation set (20%), and test set (10%). The three datasets have no overlap, ensuring the independence and reliability of subsequent model training, validation, and testing.

[0034] S2. Establish a force-thermal coupling model, which includes a viscoelastic force field sub-model, a thermal field sub-model, and a two-way coupling correlation module.

[0035] The goal of this step is to construct a dedicated model for bidirectional dynamic coupling of force field, thermal field, and material properties based on preprocessed data and the special properties of energetic materials, which are characterized by strong viscoelasticity and high sensitivity. The model consists of three parts: a viscoelastic force field sub-model, a thermal field sub-model, and a bidirectional coupling correlation module.

[0036] This step specifically includes the following: S21. Energetic materials are typical viscoelastic materials, possessing both the elastic deformation of elastic solids and the viscous flow characteristics of viscous fluids. They cannot be characterized using traditional linear elastic constitutive models. This application uses the generalized Maxwell viscoelastic constitutive equation to construct a viscoelastic force field sub-model to characterize the stress-strain evolution of the propellant during machining. The model inputs are machining parameters and initial mechanical parameters of the propellant, and the outputs are the stress field distribution, strain field distribution, and residual stress during the machining process.

[0037] The generalized Maxwell viscoelastic constitutive equation consists of a single elastic spring element and multiple Maxwell elements (spring-viscose pot series) connected in parallel. It can characterize the instantaneous elastic deformation, delayed elastic deformation, and viscous flow deformation of energetic materials. The constitutive equation is shown below:

[0038] Where σ(τ) represents the stress at time τ, in MPa; E0 represents the instantaneous elastic modulus, in MPa; and ε(τ) represents the strain at time τ, in N. M This indicates the number of Maxwell elements, with a value of 3, balancing the accuracy of viscoelastic characterization with computational efficiency. m is the index of the Maxwell element, and E... m This represents the elastic modulus of the m-th Maxwell element, in MPa. τ represents the strain rate at time s. m τ represents the relaxation time of the m-th Maxwell element in seconds, and τ represents the processing time in seconds.

[0039] During machining, the cutting tool comes into contact with the propellant charge, generating a cutting force. This cutting force is the core factor inducing stress deformation and frictional heat generation. The cutting force can be decomposed into three orthogonal components: the main cutting force, the feed resistance, and the radial force. The calculation formula is as follows:

[0040] Among them, F c This represents the main cutting force, in N or K. c A represents the cutting force coefficient per unit cutting area, which is related to material properties and tool parameters. c This represents the cutting cross-sectional area, in mm. 2 F f This represents the feed resistance, expressed in N or K. f F represents the feed resistance proportionality coefficient. r This represents radial force, measured in N, and Kr represents the radial force proportionality coefficient.

[0041] Based on the actual machining conditions after 3D printing of energetic materials, the boundary conditions of the viscoelastic force field sub-model are set as follows: Fixed boundary: The clamping end of the propellant grain adopts fully constrained boundary conditions to restrict displacement and rotation in all directions; Load boundary: A dynamic cutting force load is applied to the area where the tool contacts the propellant, and the magnitude of the load changes dynamically with the tool's movement trajectory; Initial stress boundary: The initial internal stress field of the propellant column is used as the initial stress condition of the model to truly reflect the distribution of residual stress after printing.

[0042] S22. During machining, the temperature rise of energetic materials mainly comes from frictional heat generation at the tool-propellant interface, heat generation from plastic deformation of the material, and heat generation from shear slip. Due to the low thermal conductivity of energetic materials, heat is easily accumulated to form hot spots. This application constructs a thermal field sub-model based on Fourier's law of heat conduction to characterize the temperature field distribution and hot spot evolution law during the machining process. The input of the model is cutting force, machining parameters, initial temperature of the propellant, and material thermophysical parameters. The output is the temperature field distribution, maximum temperature rise, and hot spot region during the machining process.

[0043] During machining, heat transfer inside the propellant grain follows Fourier's law of heat conduction. Considering transient heat conduction, heat generation from internal heat sources, and heat dissipation via convection, the governing equation for heat conduction is as follows:

[0044] Where ρ represents the density of the energetic material, in kg / m³. 3 c represents the specific heat capacity of the material, in J / (kg⋅K); ∂T / ∂τ represents the partial derivative of temperature with respect to time, where T represents temperature in K; λ represents the thermal conductivity of the material, in W / (m⋅K); x, y, z represent coordinates in a Cartesian coordinate system; qf This represents the frictional heat generation rate per unit volume, expressed in W / m³. 3 q p This represents the heat generation rate per unit volume of plastic deformation, expressed in W / m³. 3 q c This represents the heat dissipation rate per unit volume, expressed in W / m². 3 .

[0045] The heat generation models include: Frictional heat generation: The work done by friction at the interface between the cutting tool and the propellant is converted into heat. The formula for calculating the frictional heat generation rate is:

[0046] Where, η f V represents the frictional heat generation efficiency, with a value of 0.8, and v represents the cutting linear velocity in m / s. c This represents the volume of material cut per unit time, expressed in meters (m). 3 / s.

[0047] Heat generation from plastic deformation: The work done by plastic deformation during material cutting is converted into heat. The formula for calculating the heat generation rate from plastic deformation is:

[0048] Where, η p This represents the heat generation efficiency during plastic deformation, with a value of 0.9, σ. e This represents the equivalent stress, with units of MPa. Equivalent rate of change, which is the change in equivalent change per unit time, is expressed in seconds (s). -1 .

[0049] Based on the safety requirements for processing energetic materials, the boundary conditions for the thermal field sub-model are set as follows: Initial temperature boundary: The initial temperature of the propellant column is set to room temperature (25℃), consistent with the actual processing environment; Thermal convection boundary: The unmachined surface of the propellant grain is in contact with the air, and a natural convection heat dissipation boundary is set, with a convection heat transfer coefficient of 10 W / (m²). 2 ⋅K); Thermal radiation boundary: Ignoring heat dissipation from the surface of the propellant grain, and considering the safety constraints of energetic material processing (maximum safe temperature ≤35℃), the proportion of heat dissipation from thermal radiation in low-temperature environments is <1%, and its impact can be ignored; Safety temperature constraint: Set the upper limit of the model temperature to 35℃. Temperatures exceeding this limit are considered hotspot risk areas.

[0050] S23. The force field and thermal field do not exist independently, but rather have a bidirectional dynamic coupling relationship. On the one hand, the stress deformation generated by machining induces plastic heat generation, altering the temperature field distribution; on the other hand, the increase in temperature leads to a decrease in the elastic modulus and shear modulus of energetic materials, softening the material's mechanical properties, inducing stress field redistribution, and exacerbating deformation and heat generation. This application constructs a bidirectional coupling correlation module to realize real-time data interaction and dynamic correction between the force field and thermal field. The coupling mechanism is as follows: 1. Force field → thermal field coupling: The stress field, strain field and strain rate data calculated by the viscoelastic force field sub-model are input into the heat generation equation of the thermal field sub-model to update the friction heat generation and plastic deformation heat generation rates in real time, driving the evolution of the temperature field.

[0051] 2. Thermal Field → Force Field Coupling: The temperature field data calculated by the thermal field sub-model is input into the viscoelastic constitutive equation of the viscoelastic force field sub-model to dynamically correct the material's elastic modulus and shear modulus, and update the distribution of the stress and strain fields.

[0052] Where E(T) represents the elastic modulus at temperature T, in MPa, E0 represents the initial elastic modulus at room temperature T0, in MPa, and α E The temperature coefficient of elastic modulus is represented by T0, which represents room temperature (298 K, or 25 °C). G(T) represents the shear modulus at temperature T (in MPa), and G0 represents the initial shear modulus at room temperature T0 (in MPa). α G This represents the temperature coefficient of shear modulus.

[0053] The S24 force-thermal coupling model adopts a finite element-neural network hybrid architecture. The finite element module is responsible for solving the force and thermal field control equations and outputting stress and temperature field distribution data. The neural network module is responsible for fitting the nonlinear coupling relationship between the force and thermal fields to improve the model's computational efficiency. The input layer of the model includes machining parameters, initial state parameters of the propellant charge, and material thermophysical property parameters. Five fully connected layers are set in the hidden layer, and the ReLU (linear rectified) function is used as the activation function. The output layer includes stress field distribution, strain field distribution, temperature field distribution, maximum temperature rise, and residual stress. The key parameters of the model are summarized in Table 1.

[0054] Table 1 Key parameters of the force-thermal coupling model

[0055] S3. A composite loss function is used to train the force-thermal coupling model based on the 3D printing parameter feature set and the initial state set of the drug column.

[0056] The goal of this step is to use the preprocessed dataset, including the 3D printing parameter feature set and the initial state set of the propellant, to complete the training, validation, and testing of the force-thermal coupling model, so as to optimize the model's weight parameters, improve the model's simulation accuracy and generalization ability, and ensure that the model can accurately predict the stress field and temperature field distribution under different processing parameters, providing reliable support for subsequent parameter optimization.

[0057] This step specifically includes the following: S31. The training set (70%) partitioned in step S1 is used as the model training data. Each training sample contains two parts: input features and output labels. The input features include machining parameters, 3D printing parameters, and initial state features of the propellant charge. The output labels include the stress field distribution, strain field distribution, temperature field distribution, maximum temperature rise, and residual stress under the corresponding working conditions, obtained through finite element simulation calculations. The output labels are normalized to ensure that the data dimensions are consistent with the input features.

[0058] S32. The PyTorch deep learning framework is used for model training. The hardware configuration is an NVIDIA RTX 4090 graphics card and 64GB of RAM. The key training parameters are set as follows: Optimizer: The Adam optimizer is used to adaptively adjust the learning rate and improve the model convergence speed. Initial learning rate: set to 0.001. During training, a learning rate decay strategy is adopted, and the learning rate is multiplied by 0.5 every 100 iterations. Batch size: Set to 32 to balance training efficiency and gradient update stability; Maximum number of iterations (Epochs): Set to 500 epochs. An early stopping mechanism is set. If the loss function on the validation set does not decrease for 50 consecutive epochs, training is terminated early to avoid model overfitting. Regularization: Dropout regularization is used, with the Dropout probability set to 0.2, randomly discarding some neurons to prevent the model from overfitting.

[0059] S33. The goal of model training is to minimize the error between the model's predicted values ​​and the true labels. Considering the multiphysics data output by the force-thermal coupling model, this embodiment employs a composite loss function that is a weighted combination of the mean squared error loss function and the hotspot loss function. This balances overall prediction accuracy with the accuracy of hotspot risk prediction. The formula for calculating the composite loss function is as follows:

[0060] Among them, L total L represents the composite loss function. mse Let L represent the mean squared error loss function, ω represent the hotspot loss weighting coefficient with a value of 0.3, and L...hot This represents the hotspot loss function.

[0061] The mean squared error loss function is used to measure the overall error between the model's predicted stress field, temperature field, and other data and the true labels. The calculation formula is as follows:

[0062] Where, N train This represents the number of training samples, p is the index of the training sample, and y is the index of the training sample. p y represents the model's predicted value. t This represents the actual label value.

[0063] The hotspot loss function is used to enhance the model's prediction accuracy for hotspot areas (temperature > 35℃) and reduce the risk of missed hotspot detection. The calculation formula is as follows:

[0064] Where, N hot T represents the number of samples in the hotspot region, q is the index of the hotspot sample, and T is the index of the hotspot sample. p T represents the model's predicted temperature. t This indicates the actual temperature.

[0065] S34. The model training adopts an iterative training-validation feedback mode. After each round of iterative training, the model performance is evaluated using validation set data, the training set loss and validation set loss are recorded, and the model convergence status and overfitting risk are monitored in real time. The specific execution steps are as follows: 1. Initialize the model weight parameters using the Xavier initialization method to ensure a uniform initial weight distribution and avoid gradient vanishing or gradient exploding.

[0066] 2. Load the training set data, input it into the model in batches, and calculate the model predictions and composite loss function values ​​through forward propagation.

[0067] 3. Backpropagation is used to calculate the gradient of the loss function with respect to the model weight parameters, and the weight parameters are updated through the Adam optimizer.

[0068] 4. Repeat steps 2-3 to complete one round of iterative training.

[0069] 5. Load the validation set data, input the current model, calculate the validation set loss function value, and evaluate the model's generalization ability.

[0070] 6. Determine if the early stopping condition is met. If it is, terminate the training and save the optimal model. If not, repeat steps 2-5 until the maximum number of iterations is reached.

[0071] 7. After training is complete, save the weight parameters of the optimal model. The optimal model is defined as the model with the smallest loss function value on the validation set.

[0072] S35. Validate the performance of the optimal model using validation set data, selecting mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²). 2 As a verification indicator, the calculation formulas are as follows:

[0073]

[0074]

[0075] Where, N val This represents the number of validation samples, where r is the index of the validation sample. This represents the average value of the true labels of the validation samples.

[0076] The model was finally tested using a test set, with test metrics consistent with validation metrics. The model was required to have a MAE ≤ 0.5℃, RMSE ≤ 0.8℃, and R0.05 for the predicted maximum temperature rise. 2 ≥0.95; For the prediction of maximum stress, MAE≤2MPa, RMSE≤3MPa, R 2 A value ≥0.95 indicates that the model training is qualified and can be used for subsequent optimization of machining parameters.

[0077] S4. Select optimization variables. Based on the trained force-thermal coupling model, with the safety of processing energetic materials as a constraint, and with the goals of minimizing temperature rise, minimizing stress, optimizing accuracy, and maximizing efficiency, use the Bayesian optimization algorithm to perform global optimization screening on the machining parameters in the optimization variables to obtain the optimal parameters.

[0078] This step specifically includes the following: S41. Four key parameters that have the most significant impact on the force field and thermal field during machining are selected as optimization variables. The range of variable values ​​is set in combination with safety experience in processing energetic materials, as shown in Table 2.

[0079] Table 2 Optimization variables and their value ranges for machining.

[0080] Taking into account processing safety, forming accuracy, and processing efficiency, a four-objective optimization function is constructed, where each objective function is a minimum value function, specifically defined as follows: 1. Objective 1: Increase the processing temperature to the highest T. max Minimize risk and avoid hot topics; 2. Objective 2: Process the maximum equivalent stress σ max Minimize the amount of microcracks to prevent them from debonding from the interface; 3. Objective 3: Residual stress σ during processing resMinimize, improve the dimensional stability of the propellant column; 4. Objective 4: Processing time t m Minimize processing efficiency.

[0081] The multi-objective optimization problem is transformed into a single-objective optimization problem using the weighted normalization method. The formula for calculating the comprehensive objective function is as follows:

[0082] Among them, F obj Let T represent the comprehensive objective function. w1, w2, w3, and w4 represent the weights for temperature rise, stress, residual stress, and efficiency, respectively, with values ​​of 0.4, 0.3, 0.2, and 0.1. These weight coefficients satisfy the normalization principle, prioritizing processing safety. ref This represents the temperature rise reference value, taken as 35℃, σ ref This represents the stress reference value, which is the material's yield strength, σ. res，ref This represents the reference value for residual stress, taken as 10 MPa, t ref This represents a reference value for processing time, which is 10 minutes.

[0083] Based on the characteristics of high sensitivity and low thermal stability of energetic materials, two hard safety constraints are set. All optimized parameter combinations must meet the constraints; otherwise, they are considered invalid parameters. 1. Temperature rise constraint: T max ≤35℃, prevent hotspot formation; 2. Stress constraint: σ max ≤0.8σ s , where σ s This indicates the yield strength of energetic materials, with a 20% safety margin to prevent material yield deformation.

[0084] S42. The Bayesian optimization algorithm is suitable for optimization scenarios with black-box functions, high computational costs, and low parameter dimensionality. It can efficiently explore the parameter space and quickly converge to the global optimum. It is well-suited to the high computational cost of force-thermal coupling model simulation. The core of the algorithm includes three parts: surrogate model, acquisition function, and iterative optimization.

[0085] A Gaussian process (GP) is used as the surrogate model. A Gaussian process is a nonparametric Bayesian model that can fit the distribution of the objective function based on a small amount of sample data, outputting the mean and variance of the objective function to quantify prediction uncertainty. The Gaussian process model is defined as follows: f(x)∼GP(μ(x),k(x,x′))

[0086] Where f(x) represents the objective function, GP(·) represents the Gaussian process, x represents the combination of optimization parameters, μ(x) represents the mean function, initially set to 0, and k(x, x′) represents the kernel function, using the Matérn kernel function to balance the model's fitting ability and smoothness.

[0087] The acquisition function is used to balance exploration and utilization. The Expected Improvement (EI) is selected as the acquisition function, and its calculation formula is as follows:

[0088] Where EI(x) represents the expected improvement value, f best Let denot represent the current optimal objective function value, Φ(⋅) represent the cumulative distribution function of the standard normal distribution, ϕ(⋅) represent the probability density function of the standard normal distribution, and σ(x) represent the prediction standard deviation of the objective function.

[0089] The execution steps of iterative optimization are as follows: 1. Initial sample collection: Using the Latin hypercube sampling method, 20 sets of initial parameter combinations are generated within the range of optimization variable values. These are then input into the force-thermal coupling model, and the comprehensive objective function value corresponding to each set of parameters is calculated to construct the initial sample dataset.

[0090] 2. Proxy model training: Based on the initial sample dataset, train a Gaussian process proxy model to fit the mapping relationship between the optimization parameters and the comprehensive objective function.

[0091] 3. Optimal parameter search: Maximize the expected improvement of the acquisition function to obtain the next set of optimal parameter combinations to be evaluated.

[0092] 4. Model Evaluation and Update: Input the optimal parameter combination into the force-thermal coupling model, calculate the corresponding comprehensive objective function value, add the parameter combination and objective function value to the sample dataset, and update the Gaussian process surrogate model.

[0093] 5. Iteration termination judgment: Repeat steps 3-4 until the maximum number of iterations (50 rounds) is reached or the objective function value does not decrease for 10 consecutive rounds, then terminate the iteration.

[0094] 6. Optimal parameter output: From the final sample dataset, select the parameter combination that satisfies the safety constraints and minimizes the comprehensive objective function value, which is the optimal parameter for machining.

[0095] S43. Output the optimal combination of machining parameters, including spindle speed, feed rate, depth of cut, and tool rake angle. Input the optimal parameters into the force-thermal coupling model for simulation verification. Confirm that the simulation results meet the following requirements: maximum temperature rise ≤ 32℃, maximum stress ≤ 0.7 times yield strength, residual stress ≤ 5MPa, and machining time is reduced by more than 10% compared to empirical parameters. After verification, output the optimal parameters to the CNC machining system for actual machining.

[0096] S5. Process energetic material propellant columns using optimal parameters, collect working condition data in real time during the processing, and use the working condition data to dynamically correct the model parameters and boundary conditions of the force-thermal coupling model. Based on the prediction results of the dynamically corrected force-thermal coupling model, combined with the working condition data, use a PID control algorithm to adaptively fine-tune the processing parameters.

[0097] The goal of this step is to carry out actual machining based on optimal machining parameters. By real-time data acquisition, dynamic model correction, adaptive parameter adjustment, and safety early warning and protection, a closed-loop control system is constructed to cope with uncertainties such as individual material differences, environmental fluctuations, and tool wear during the machining process, ensuring the safety and stability of the machining process.

[0098] This step specifically includes the following: S51. Install the 3D-printed energetic material propellant grains onto the CNC machining tool, complete the tooling fixture calibration, tool installation and tool setting, input the optimal parameters, set the machining trajectory (turning the outer circle, milling the combustion surface, end face shaping), and debug the machine tool operation status to ensure that the equipment is normal.

[0099] S52. During machine tool processing, key operating condition data are collected in real time using a cutting force sensor, an infrared thermometer, and a vibration sensor. The sampling frequency is set to 20Hz. The collected data includes: Cutting force data: main cutting force, feed resistance, radial force; Temperature data: Temperature of the tool-propellant contact area, temperature of the propellant surface; Vibration data: machine tool spindle vibration acceleration, and dart clamping end vibration displacement.

[0100] The collected data is transmitted in real time to the local control terminal via a high-speed data acquisition card for real-time storage and preliminary preprocessing to remove noise interference and ensure data reliability.

[0101] S53. Input the real-time acquired working condition data into the force-thermal coupling model, and dynamically correct the model parameters and boundary conditions. The correction includes: 1. Material parameter correction: Based on real-time temperature data, update the thermophysical parameters such as elastic modulus, shear modulus, and thermal conductivity of energetic materials in the model to adapt to the material properties at actual processing temperatures.

[0102] 2. Load boundary correction: Based on real-time cutting force data, the magnitude and distribution of the tool-propellant contact load in the model are corrected to better match the actual cutting force fluctuations.

[0103] 3. Model weight correction: Based on real-time temperature and stress prediction errors, the weight parameters of the neural network module are fine-tuned to improve the real-time prediction accuracy of the model.

[0104] The model correction cycle is set to 1 second to ensure that the model tracks changes in processing conditions in real time and outputs accurate stress and temperature field prediction results.

[0105] S54. Based on the prediction results of the dynamically corrected force-thermal coupling model and combined with real-time operating data, a PID (Proportional Integral Derivative) control algorithm is used to adaptively fine-tune the machining parameters. The fine-tuning range is limited to ±10% of the optimal parameters. This range is determined based on the safety threshold for machining energetic materials and the servo response limit of the machine tool, taking into account both the flexibility of parameter adjustment and machining stability. The specific adjustment logic is as follows: 1. When the temperature rise is too high: If the real-time temperature is >30℃, reduce the spindle speed by 50-100 r / min and reduce the feed rate by 0.05-0.1 mm / r to reduce the rate of heat generation during cutting.

[0106] 2. When the stress is too high: If the real-time cutting force is greater than 1.1 times the cutting force corresponding to the optimal parameters, reduce the cutting depth by 0.1-0.2 mm and reduce the cutting load.

[0107] 3. Abnormal vibration: If the vibration acceleration is greater than 0.5g, reduce the spindle speed and adjust the tool rake angle to reduce vibration impact.

[0108] 4. When the working conditions are stable: If the temperature, stress and vibration are all within the safe range, the optimal parameters are maintained to ensure processing efficiency.

[0109] S55. To prevent safety accidents during processing, this application embodiment also sets up a three-level safety early warning mechanism to monitor working condition data and model prediction results in real time. The early warning levels and emergency measures are as follows: Level 1 Warning (Alert): When the real-time temperature is 28-32℃ and the maximum stress is 0.7-0.8 times the yield strength, the system will issue an audible and visual alert to remind the operator to pay attention to changes in the working conditions. Level 2 warning (warning level): When the real-time temperature is 32-35℃ and the maximum stress is 0.8-0.9 times the yield strength, the system will automatically trigger parameter downgrade adjustment and issue a continuous audible and visual alarm. Level 3 warning (danger level): When the real-time temperature is >35℃ and the maximum stress is ≥0.9 times the yield strength, the system will immediately control the machine tool to stop in an emergency, cut off the power source, and start the inert gas purging device to prevent hot spots from causing safety accidents.

[0110] S6. Detect the performance data of the processed energetic material propellant column and feed the performance data back to the force-thermal coupling model to achieve iterative data updates.

[0111] The goal of this step is to conduct comprehensive performance testing on the processed energetic material propellant grains, evaluate the processing quality, and feed the performance data back to the force-thermal coupling model to achieve iterative updates of the model and improvement of the parameter database, thereby enhancing the consistency and stability of subsequent processing.

[0112] This step specifically includes the following: S61. After processing, allow the propellant column to cool naturally to room temperature, and then conduct tests on four categories of indicators: dimensional accuracy, surface quality, internal integrity, and combustion performance. Each indicator is tested three times, and the average value is taken as the final test result. Dimensional accuracy inspection: A high-precision coordinate measuring machine is used to inspect the diameter, length, combustion profile, and end face flatness of the propellant charge. The dimensional tolerance is required to be ≤ ±0.05mm and the profile is required to be ≤ 0.03mm.

[0113] Surface quality inspection: A surface roughness meter is used to inspect the surface roughness of the propellant cartridge. The surface roughness is required to be Ra≤1.6μm, with no burrs, scratches, or cracks.

[0114] Internal integrity testing: Micro-CT equipment is used to test the porosity and number of microcracks inside the drug cartridge. The porosity is required to be ≤0.5% and there are no obvious microcracks.

[0115] Combustion performance testing: A static combustion test device is used to test the burning rate of the propellant, combustion stability, and thrust curve. The burning rate fluctuation is required to be ≤3% and the thrust eccentricity is required to be ≤1°.

[0116] S62. Correlate the performance data of the propellant charge, the operating condition data during processing, and the prediction data of the force-thermal coupling model to construct a processing-inspection-feedback dataset for model iterative optimization: 1. Model error analysis: Compare the dimensional accuracy, residual stress, and temperature field data predicted by the model with the actual test data to pinpoint the sources of model error.

[0117] 2. Model parameter update: Based on the error analysis results, the material parameters, neural network weight parameters and coupling correlation coefficients of the force-thermal coupling model are updated in reverse to correct the model bias.

[0118] 3. Optimize strategy iteration: Add the detection data to the parameter optimization sample dataset, update the surrogate model of the Bayesian optimization algorithm, optimize the weight coefficients of the objective function, and improve the accuracy of parameter optimization.

[0119] S63. Organize and archive the optimal parameters, working condition data and test data of each processing to build a database of machining parameters after 3D printing of energetic materials. The database is classified and stored according to material type, propellant size and 3D printing process. Subsequent processing can directly retrieve the appropriate parameters from the database and combine them with real-time working condition fine-tuning to greatly improve processing efficiency and consistency, and finally form a complete optimization system for machining parameters after 3D printing of energetic materials.

[0120] This application also provides an embodiment of an optimization system for processing parameters of energetic materials based on a force-thermal coupling model. The system includes: The feature extraction module is used to collect the operating parameters and initial state data of energetic material propellant during the 3D printing process, extract key features, and establish a 3D printing parameter feature set and a propellant initial state set. The model building module is used to build a force-thermal coupling model, which includes a viscoelastic force field sub-model, a thermal field sub-model, and a two-way coupling correlation module. The model training module is used to train the force-thermal coupling model based on the 3D printing parameter feature set and the initial state set of the drug grains using a composite loss function; The parameter selection module is used to select optimization variables. Based on the trained force-thermal coupling model, with the safety of processing energetic materials as a constraint, and with the objectives of minimizing temperature rise, minimizing stress, optimizing accuracy and maximizing efficiency, the Bayesian optimization algorithm is used to perform global optimal selection of the machining parameters in the optimization variables to obtain the optimal parameters. The parameter fine-tuning module is used to process energetic material propellant columns using optimal parameters, collect working condition data in real time during the processing, and dynamically correct the model parameters and boundary conditions of the force-thermal coupling model using the working condition data. Based on the prediction results of the dynamically corrected force-thermal coupling model, combined with the working condition data, the PID control algorithm is used to adaptively fine-tune the processing parameters. The data update module is used to detect the performance data of the processed energetic material propellant column and feed the performance data back to the force-thermal coupling model to achieve iterative data updates.

[0121] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0122] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for optimizing processing parameters of energetic materials based on a force-thermal coupling model, characterized in that, Including optimization steps based on a force-thermal coupling model: S1. Collect the operating parameters and initial state data of energetic material propellant during the 3D printing process, extract key features, and establish a 3D printing parameter feature set and a propellant initial state set. S2. Establish a force-thermal coupling model, which includes a viscoelastic force field sub-model, a thermal field sub-model, and a two-way coupling correlation module; S3. Using a composite loss function, train the force-thermal coupling model based on the 3D printing parameter feature set and the initial state set of the drug column; S4. Select optimization variables. Based on the trained force-thermal coupling model, with the safety of processing energetic materials as a constraint, and with the objectives of minimizing temperature rise, minimizing stress, optimizing accuracy and maximizing efficiency, use the Bayesian optimization algorithm to perform global optimal screening of the machining parameters in the optimization variables to obtain the optimal parameters. S5. Process the energetic material propellant column using the optimal parameters, collect the working condition data in real time during the processing, use the working condition data to dynamically correct the model parameters and boundary conditions of the force-thermal coupling model, and based on the prediction results of the dynamically corrected force-thermal coupling model, combine the working condition data and use a PID control algorithm to adaptively fine-tune the processing parameters. S6. Detect the performance data of the processed energetic material propellant column, and feed the performance data back to the force-thermal coupling model to achieve iterative data updates.

2. The method for optimizing processing parameters of energetic materials based on a force-thermal coupling model according to claim 1, characterized in that, In step S1, the operating parameters during the 3D printing process include material delivery flow rate, nozzle temperature, material viscosity, printing platform stress, printing chamber ambient temperature, printing chamber ambient humidity, and printing speed; the initial state data of the propellant grains include dimensional accuracy, internal porosity, material uniformity, initial internal stress, initial elastic modulus of the material, and initial shear modulus of the material.

3. The method for optimizing processing parameters of energetic materials based on a force-thermal coupling model according to claim 1, characterized in that, In step S1, after acquiring the operating parameters and the initial state data of the drug column, data preprocessing is performed, including moving average filtering, outlier removal, and linear normalization. Based on the preprocessed data, the key features are extracted, including temperature-viscosity change rate, stress distribution standard deviation, flow rate-printing speed ratio, mean porosity, mean initial internal stress, mean initial elastic modulus, and mean initial shear modulus.

4. The method for optimizing processing parameters of energetic materials based on a force-thermal coupling model according to claim 1, characterized in that, In step S2, the viscoelastic force field sub-model adopts the generalized Maxwell viscoelastic constitutive equation, the thermal field sub-model is based on Fourier's law of heat conduction, and the bidirectional coupling correlation module is used for real-time data interaction and dynamic correction of the force field and thermal field.

5. The method for optimizing processing parameters of energetic materials based on a force-thermal coupling model according to claim 1, characterized in that, In step S3, the composite loss function is a weighted combination of the mean squared error loss function and the hot spot loss function.

6. The method for optimizing processing parameters of energetic materials based on a force-thermal coupling model according to claim 1, characterized in that, In step S4, the optimization variables include spindle speed, feed rate, depth of cut, and tool rake angle.

7. The method for optimizing processing parameters of energetic materials based on a force-thermal coupling model according to claim 1, characterized in that, In step S4, the Bayesian optimization algorithm uses a Gaussian process surrogate model and a desired improved acquisition function to obtain the optimal parameters through iterative screening.

8. The method for optimizing processing parameters of energetic materials based on a force-thermal coupling model according to claim 1, characterized in that, In step S5, the real-time collected working condition data includes cutting force, temperature and vibration data, and the fine-tuning range of the machining parameters is ±10% of the optimal parameters.

9. The method for optimizing processing parameters of energetic materials based on a force-thermal coupling model according to claim 1, characterized in that, In step S6, the performance data includes dimensional accuracy, surface quality, internal integrity, and combustion performance.

10. A system applying the method for optimizing processing parameters of energetic materials based on a force-thermal coupling model as described in any one of claims 1-9, characterized in that, include: The feature extraction module is used to collect the operating parameters and initial state data of energetic material propellant during the 3D printing process, extract key features, and establish a 3D printing parameter feature set and a propellant initial state set. The model building module is used to establish a force-thermal coupling model, which includes a viscoelastic force field sub-model, a thermal field sub-model, and a two-way coupling correlation module. The model training module is used to train the force-thermal coupling model based on the 3D printing parameter feature set and the initial state set of the drug grain using a composite loss function; The parameter selection module is used to select optimization variables. Based on the trained force-thermal coupling model, with the safety of processing energetic materials as a constraint, and with the objectives of minimizing temperature rise, minimizing stress, optimizing accuracy and maximizing efficiency, the Bayesian optimization algorithm is used to perform global optimal selection of the machining parameters among the optimization variables to obtain the optimal parameters. The parameter fine-tuning module is used to process energetic material propellant columns using the optimal parameters, collect working condition data in real time during the processing, dynamically correct the model parameters and boundary conditions of the force-thermal coupling model using the working condition data, and adaptively fine-tune the processing parameters using a PID control algorithm based on the prediction results of the dynamically corrected force-thermal coupling model and the working condition data. The data update module is used to detect the performance data of the processed energetic material propellant column and feed the performance data back to the force-thermal coupling model to achieve iterative data updates.

Images