CN118153459A - Solid rocket engine ignition process model correction method, device and equipment - Google Patents

Abstract

The invention relates to a method, a device and equipment for correcting a model in the ignition process of a solid rocket engine, which are used for obtaining a model dataset by normalizing acquired pressure sequence data and dividing the model dataset into a training set and a testing set; then constructing a pre-training correction model of the solid rocket engine ignition process; training the pre-training correction model through a training set and a pre-constructed loss function to obtain a trained correction model of the ignition process of the solid rocket engine; and finally, inputting the test set into a solid rocket engine ignition process correction model to obtain a solid rocket engine ignition process prediction result. The solid rocket engine ignition process correction model provided by the invention has better applicability and generalization capability, can effectively improve the precision of the solid rocket engine ignition process correction model, improves the design efficiency of the solid rocket engine, accelerates development and reduces cost.

Description

Solid rocket engine ignition process model correction method, device and equipment

Technical Field

The present invention relates to the technical field of solid rocket engines, and in particular to a method, device and equipment for correcting a solid rocket engine ignition process model.

Background technique

The ignition transient process of a solid rocket engine is a crucial part of the work of a solid rocket engine. It refers to the time interval from the ignition of the fuel by the igniter to the gradual entry of the fuel combustion flow into a quasi-steady-state flow. In this process, the success or failure and quality of ignition are crucial to the normal operation of the solid rocket engine. The ignition transient process directly affects the performance, efficiency, reliability and safety of the solid rocket engine. The key to the ignition transient process is whether the igniter can effectively ignite the fuel and stably propagate the combustion flow throughout the engine. The success or failure of ignition will directly determine whether the solid rocket engine can start smoothly and operate normally. Once the ignition fails or the combustion is insufficient, the engine will not be able to generate enough thrust, provide the required power, or even work properly.

Therefore, the factors that affect the performance, efficiency, reliability and safety of the engine mainly include the degree of complete combustion of the fuel during the ignition process, the combustion time, the stability of the combustion process, etc. If the ignition transient process is unstable or the combustion is incomplete, it will lead to a decrease in engine performance, low efficiency, and even cause unsafe phenomena. In order to ensure the normal operation of the solid rocket engine, a series of measures need to be taken during the ignition transient process, such as optimizing the igniter design, improving the fuel combustion quality, optimizing the combustion chamber structure, etc., to ensure a good ignition effect and a stable combustion process.

Summary of the invention

Based on this, it is necessary to provide a solid rocket engine ignition process model correction method, device and equipment to address the above-mentioned technical problems. By constructing a solid rocket engine ignition process simulation model, it can adaptively learn and adjust model parameters, and perform model correction on the transient simulation of the solid rocket engine ignition process, so as to improve the design efficiency of the solid rocket engine, speed up development and reduce costs.

A solid rocket engine ignition process model correction method, the method comprising:

Acquire pressure sequence data, perform normalization processing on the pressure sequence data to obtain a model data set, and divide the model data set into a training set and a test set;

Construct a pre-trained correction model for the solid rocket motor ignition process;

The pre-trained correction model is trained by the training set and the pre-constructed loss function to obtain a trained solid rocket engine ignition process correction model; wherein the pre-trained correction model performs deep convolution feature sequence extraction on the input training set to obtain a deep convolution feature sequence; after the deep convolution feature sequence is activated, feature extraction is performed through positive order input and reverse order input respectively and splicing and fusion are performed to obtain a new feature sequence; the new feature sequence is used to generate an attention vector through an attention mechanism, and the attention vector is regularized; the regularized attention vector is integrated to generate a final prediction output; at the same time, the generated prediction result process is guided and constrained by the pre-constructed loss function;

The test set is input into the solid rocket engine ignition process correction model to obtain a solid rocket engine ignition process prediction result.

In one embodiment, the pre-trained correction model includes a deep convolution module, a bidirectional LSTM module, an attention module, a regularization module and a fully connected layer;

Performing deep convolution feature sequence extraction on the input training set through the deep convolution module to obtain a deep convolution feature sequence, and activating the deep convolution feature sequence;

The activated deep convolution feature sequence is input in forward order and reverse order respectively through the bidirectional LSTM module, and concatenated and fused after feature extraction to obtain a new feature sequence;

The new feature sequence is calculated by the attention module to obtain an attention vector;

Regularizing the attention vector by the regularization module;

The regularized attention vector is integrated through the fully connected layer to generate the final prediction output.

In one embodiment, the calculation formula of the depth convolution module is expressed as:

;

In the formula, Represents a deep convolutional feature sequence; /> Indicates weight; /> Represents the input sequence data; /> represents the bias; where, /> ,/> .

In one of the embodiments, the deep convolution module includes a 5-layer convolution structure.

In one embodiment, the calculation formula of the bidirectional LSTM module is expressed as:

;

In the formula, Represents the hidden layer state; /> Indicates positive sequence input; /> Indicates reverse order input; /> and/> Respectively represent the weight parameters of the reverse hidden layer and the reverse hidden layer; /> Represents the bias parameter of the hidden layer; /> A sequence of features representing the activation function.

In one embodiment, the calculation formula of the attention module is expressed as:

;

In the formula, Represents the attention vector; /> Represents the hidden layer state; /> Represents the weight of the input feature; /> Represents the bias of the input feature.

In one embodiment, the pre-constructed loss functions include mean absolute error, mean square error, root mean square error, coefficient of determination, and mean absolute percentage error.

A solid rocket engine ignition process model correction device, the device comprising:

A data processing module, used for acquiring pressure sequence data, performing normalization processing on the pressure sequence data to obtain a model data set, and dividing the model data set into a training set and a test set;

A model building module, which is used to build a pre-trained correction model of the solid rocket motor ignition process;

A model pre-training module is used to train the pre-trained correction model through the training set and the pre-constructed loss function to obtain a trained solid rocket engine ignition process correction model; wherein the pre-trained correction model performs deep convolution feature sequence extraction on the input training set to obtain a deep convolution feature sequence; after the deep convolution feature sequence is activated, feature extraction is performed through positive order input and reverse order input respectively and splicing and fusion are performed to obtain a new feature sequence; the new feature sequence is used to generate an attention vector through an attention mechanism, and the attention vector is regularized; the regularized attention vector is integrated to generate a final prediction output; at the same time, the generated prediction result process is guided and constrained by the pre-constructed loss function;

The prediction result generation module is used to input the test set into the solid rocket engine ignition process correction model to obtain the solid rocket engine ignition process prediction result.

A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above-mentioned methods for correcting the solid rocket engine ignition process model when executing the computer program.

The above-mentioned solid rocket engine ignition process model correction method, device and equipment obtain a model data set by normalizing the acquired pressure sequence data, and divide the model data set into a training set and a test set; then construct a pre-trained correction model of the solid rocket engine ignition process; train the pre-trained correction model through the training set and a pre-constructed loss function to obtain a trained solid rocket engine ignition process correction model; wherein, the pre-trained correction model performs deep convolution feature sequence extraction on the input training set to obtain a deep convolution feature sequence; after the deep convolution feature sequence is activated, feature extraction is performed through positive order input and reverse order input respectively and splicing and fusion are performed to obtain a new feature sequence; the new feature sequence is used to generate an attention vector through an attention mechanism, and the attention vector is regularized; the regularized attention vector is integrated to generate a final prediction output; at the same time, the generated prediction result process is guided and constrained by a pre-constructed loss function; finally, the test set is input into the solid rocket engine ignition process correction model to obtain a solid rocket engine ignition process prediction result. The present invention constructs a solid rocket engine ignition process correction model to learn patterns and laws from a large amount of actual pressure sequence data, without relying on precise physical models, and has better applicability and generalization ability; on the other hand, the constructed solid rocket engine ignition process correction model can effectively improve the accuracy of the solid rocket engine ignition process correction model through deep convolution feature sequence extraction combined with positive sequence input and reverse sequence input, and through regularization to deal with uncertain factors in the training process, the accuracy of the model is further improved, the design efficiency of the solid rocket engine is improved, the development is accelerated, and the cost is reduced; the output prediction results can more accurately predict the pressure changes of the solid rocket engine at different time points, thereby providing support for the design of solid rocket engines.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic flow chart of a method for correcting a solid rocket engine ignition process model in one embodiment;

FIG2 is a schematic diagram of the module structure of a solid rocket engine ignition process correction model in one embodiment;

FIG3 is a schematic diagram of a loss function in the training of a correction model for the ignition process of a solid rocket engine in one embodiment;

FIG4 is a schematic diagram of a curve showing predicted values and original values of a correction model for a solid rocket engine ignition process in one embodiment;

FIG5 is a schematic diagram of the structural framework of a solid rocket engine ignition process model correction device in one embodiment;

FIG. 6 is a schematic diagram of the internal structure of a computer device in one embodiment.

Detailed ways

In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

As is known to all, the ignition process of solid rocket engines is often affected by many uncertainties and environmental changes, such as fuel changes, temperature changes, etc. In the prior art, the ignition process of solid rocket engines is generally corrected by fluid transient simulation. However, there are problems such as complex parameter modification and long simulation time in the fluid transient simulation of the ignition process of solid rocket engines. Therefore, some people have proposed to simplify the ignition transient process of the engine appropriately, such as the proposed two-dimensional unsteady calculation model. However, this model is suitable for describing the ignition transient process of axisymmetric solid rocket engines.

In the process of implementing the technology of this solution, the inventor found that deep learning can learn a more accurate ignition model and correct the existing model through training based on a large amount of data. Compared with traditional physical models, it can better capture nonlinear relationships and complex dynamic behaviors. At present, deep learning methods have been applied in the field of solid rocket engines, such as reconstructing the internal structure of the engine under dual perspectives based on deep learning models, and performing real-time defect detection on the propellant column. The deep learning method can adapt to different working conditions and environmental changes through adaptive learning and adjustment of model parameters. Therefore, the inventor proposed a model correction for the transient simulation of the ignition process of the solid rocket engine based on deep learning. By constructing a correction model for the ignition process of the solid rocket engine, and integrating the attention mechanism and the long short-term memory network in the model, the time-pressure curve of the solid rocket engine is predicted, and the sequence data obtained by the transient simulation is input for training, and the error between the predicted value and the simulated value is calculated. Finally, the predicted curve, the simulation curve and the error value are automatically output and the trained deep learning model is saved. The output result can more accurately predict the pressure change of the solid rocket engine at different time points, thereby providing support for the design of the solid rocket engine. It is worth noting that, for the convenience of description and naming, the inventor named the constructed solid rocket engine ignition process correction model as ADCBiLSTM model.

The following will describe the implementation of the present invention in detail with reference to the accompanying drawings in the embodiment diagram of the present invention.

In one embodiment, as shown in FIG1 , a solid rocket engine ignition process model correction method is provided, comprising the following steps:

Step 202, obtaining pressure sequence data, normalizing the pressure sequence data to obtain a model data set, and dividing the model data set into a training set and a test set.

Specifically, the time-pressure sequence data of different mass flow rates are obtained. When training, the sequence data is read, and its data size is 691×9. The title of the data set is time, pressure0.3~pressure1.0. The data is normalized by the MinMaxScaler function, and then the data is divided into a training set and a test set. The training set includes Train_X (pressure0.3~0.5) and Trian_Y (pressure0.6), and the test set is divided into Test_X (pressure0.7~0.9) and Test_Y (pressure1.0). Normalization can improve the convergence speed of the model and improve the accuracy.

Step 204, construct a pre-trained correction model for the solid rocket motor ignition process.

Specifically, the constructed pre-trained correction model includes a deep convolution module, a bidirectional LSTM module, an attention module, a regularization module and a fully connected layer, where:

The deep convolution feature sequence is extracted from the input training set through the deep convolution module to obtain a deep convolution feature sequence, and the deep convolution feature sequence is activated through the activation layer.

The activated deep convolution feature sequence is input in forward and reverse order through the bidirectional LSTM module, and is concatenated and fused after feature extraction to obtain a new feature sequence.

The new feature sequence is calculated through the attention module to obtain the attention vector.

The attention vector is regularized through the regularization module.

The regularized attention vector is integrated through a fully connected layer to generate the final prediction output.

Specifically, the deep convolution module is generally a 5-layer convolution structure, and can also be set to more layers according to the situation. Information will be convolutionally calculated through the convolution layer to obtain the feature representation of the input information. Therefore, convolution is a core component. For data with local features, it has excellent feature extraction capabilities. A deep convolutional neural network layer with a 5-layer convolution structure is established to extract features. The sequence data is represented as /> , let the convolution kernel size be ,/> It is represented by the number of layers where the convolution kernel is located. /> Indicates the index value of the convolution kernel at the current layer, /> Represents the index value of the convolution kernel in the previous layer, and the deep convolution feature sequence is obtained through the forward propagation of the deep convolution module .

In one embodiment, the deep convolution module can be expressed in the form of a matrix as follows:

;

In the formula, Represents a deep convolutional feature sequence; /> Indicates weight; /> Represents the input sequence data; /> represents the bias; where, /> ,/> .

In order to effectively obtain information in more dimensions, it is necessary to introduce nonlinear mapping in the modeling process to enhance the model's language expression ability in more complex spaces. This nonlinear mapping is called an activation function.

In one embodiment, the calculation formula of the activation layer is as follows:

;

The deep convolution feature sequence Input into the activation function to get the feature sequence of the activated activation function/> It can be understood that the activation function can be used to solve the problem of gradient disappearance.

In one embodiment, the characteristic sequence of the activation function is Input into the bidirectional LSTM module, and the calculation formula for each level of hidden layer state h _t is as follows:

;

In the formula, Represents the hidden layer state; /> Indicates positive sequence input; /> Indicates reverse order input; /> and/> Respectively represent the weight parameters of the reverse hidden layer and the reverse hidden layer; /> represents the bias parameter of the hidden layer.

It can be understood that there are three "gates" inside the long short-term memory network (LSTM) to control the discarding and retention of historical information and current information, namely the forget gate, input gate and output gate. The specific calculation process is that the input sequence data first passes through the forget gate, and the forget gate uses the long-term memory input of the previous moment to and short-term memory input/> The information that needs to be forgotten after calculation is recorded as/> , the information that needs to be retained after calculating the input gate is recorded as/> , and also save the current memory unit as/> , and finally calculate the output value of the output gate as/> . Combining the forget gate and the input gate can get a new long-term memory unit/> ,/> The output gate is passed to the tanh function, and the hidden layer output value is calculated through sigmoid .

The bidirectional long short-term memory network structure (bidirectional LSTM) can be divided into a forward long short-term memory network and a backward long short-term memory network. The input sequence is input into the two networks in forward and reverse order for feature extraction, and finally concatenated to obtain a new feature sequence.

In one embodiment, the calculation formula of the attention module is expressed as:

;

In the formula, Represents the attention vector; /> Represents the hidden layer state; /> Represents the weight of the input feature; /> Represents the bias of the input feature.

It can be understood that the attention mechanism allows the model to focus on different parts of the input data when making predictions or generating outputs. It can directly obtain the connection between the global and the local, set fewer parameters, and reduce the model complexity.

The output of the bidirectional LSTM module Input to the attention module, and use the softmax function to calculate the attention vector/> .

In one of the embodiments, the establishment of the regularization module is mainly used to deal with uncertainties in the training process, reduce the overfitting problem in the neural network, and further improve the accuracy of the ADCBiLSTM model. It is achieved by randomly discarding (shielding) the output of a part of the neurons during the training process. Regularization technology sets the output of certain neurons to 0 with a certain probability during the forward propagation of each training sample, thereby "discarding" them. The discarded neurons will not affect the subsequent neurons in this forward propagation. This integrated approach helps to reduce overfitting and improve the generalization ability of the model.

In one embodiment, the fully connected layer acts as a "classifier" in the entire neural network. The fully connected layer integrates the attention vector features together, so that the features finally seen by the neural network are global features, and expands the calculation results of the ADCBiLSTM model into a one-dimensional vector to generate the final prediction output, which is convenient for result output.

Step 206, training the pre-trained correction model through the training set and the pre-constructed loss function to obtain a trained solid rocket engine ignition process correction model; wherein, the pre-trained correction model performs deep convolution feature sequence extraction on the input training set to obtain a deep convolution feature sequence; after the deep convolution feature sequence is activated, feature extraction is performed through the forward order input and the reverse order input and the splicing and fusion are performed to obtain a new feature sequence; the new feature sequence is used to generate an attention vector through the attention mechanism, and the attention vector is regularized; the regularized attention vector is integrated to generate the final prediction output; at the same time, the generated prediction result process is guided and constrained by the pre-constructed loss function.

Specifically, when training the ADCBiLSTM model, the initial parameters are first set as follows:

in_channels: the dimension of the vector; out_channels: the number of channels generated by the convolution; kernel_size: the size of the convolution kernel; stride: the step size; padding: the number of padding layers; input size: the size of the last dimension of the input tensor; output_size: the size of the output tensor; hidden_size: the size of the hidden state and memory unit; num_layers: the number of layers of the LSTM stack; batch_first: whether the input and output tensors use the batch dimension as the first dimension; bidirectional: set bidirectional LSTM. learning rate: learning rate; number epochs: the number of training rounds; Dropout: the probability of discarding neurons.

Then start training, input the training set into the ADCBiLSTM model for forward propagation, and calculate the loss function. Then use the backpropagation algorithm to train the ADCBiLSTM model and optimize the parameters of the ADCBiLSTM model. The training process repeats the above process until the number of cycles reaches the number of cycles initially set for the model.

During the training process, the generated prediction results are guided and constrained by the pre-built loss function. In order to measure the accuracy of model prediction from different angles, the pre-built loss functions include mean absolute error, mean square error, root mean square error, coefficient of determination and mean absolute percentage error, where:

The mean absolute error (MAE) represents the average of the absolute errors between the predicted and observed values. MAE is a linear score where all individual differences are equally weighted on the mean. The mean absolute error is expressed as:

;

In the formula, Indicates the first/> The predicted value of samples, /> Indicates the first/> The true value of the samples.

The mean square error (MSE) is an indicator used to measure the difference between the model's predicted value and the actual observed value, and is used to evaluate the degree of fit of the model on the given data. The mean square error is expressed as:

;

The root mean square error (RMSE) is an indicator used to measure the difference between the model's predicted value and the actual observed value. It is used to evaluate the degree of fit of the model on the given data. The root mean square error is expressed as:

;

decisive factor A statistical indicator used to evaluate the goodness of fit of a regression model. It indicates the proportion of the variability of the dependent variable that can be explained by the model, that is, the degree of fit of the model to the data, expressed as:

;

In the formula, Represents the sample mean.

The mean absolute percentage error (MAPE) is a statistical indicator to measure the accuracy of predictions and is expressed as:

;

Through these five pre-built loss functions, the accuracy of model prediction can be evaluated from different angles, making the prediction results generated by the model more accurate.

Step 208, input the test set into the solid rocket engine ignition process correction model to obtain the solid rocket engine ignition process prediction result.

Specifically, the trained solid rocket engine ignition process correction model is changed to test mode through the model.eval() function, and the test set is input for testing.

The predicted solution result obtained is a predicted pressure curve. The predicted pressure curve and the simulated pressure curve are plotted through the predicted pressure curve, and the predicted error value is calculated.

The above-mentioned solid rocket engine ignition process model correction method obtains a model data set by normalizing the acquired pressure sequence data, and divides the model data set into a training set and a test set; then constructs a pre-trained correction model of the solid rocket engine ignition process; trains the pre-trained correction model through the training set and a pre-constructed loss function to obtain a trained solid rocket engine ignition process correction model; wherein, the pre-trained correction model extracts a deep convolution feature sequence from the input training set to obtain a deep convolution feature sequence; after the deep convolution feature sequence is activated, feature extraction is performed through positive order input and reverse order input and splicing and fusion are performed to obtain a new feature sequence; the new feature sequence is used to generate an attention vector through an attention mechanism, and the attention vector is regularized; the regularized attention vector is integrated to generate a final prediction output; at the same time, the generated prediction result process is guided and constrained by a pre-constructed loss function; finally, the test set is input into the solid rocket engine ignition process correction model to obtain a solid rocket engine ignition process prediction result. The present invention constructs a solid rocket engine ignition process correction model to learn patterns and laws from a large amount of actual pressure sequence data, without relying on an accurate physical model, and has better applicability and generalization ability; on the other hand, the constructed solid rocket engine ignition process correction model can effectively improve the accuracy of the solid rocket engine ignition process correction model by extracting deep convolution feature sequences combined with positive input and reverse input, and by regularization to deal with uncertain factors in the training process, reduce overfitting problems, further improve the accuracy of the model, improve the design efficiency of solid rocket engines, speed up development, and reduce costs; the output prediction results can more accurately predict the pressure changes of solid rocket engines at different time points, thereby providing support for the design of solid rocket engines. Compared with fluid simulation, the ADCBiLSTM model provided by the present invention is more convenient to modify parameters, and there is no need to perform operations such as grid division.

In one of the embodiments, in order to verify the effectiveness of the ADCBiLSTM model provided by the present invention, it is verified.

First, based on the established ADCBiLSTM model, make the following settings:

The deep convolution module has a total of 5 layers. The dimension of the vector is set to 1. The number of channels generated by the output convolution is set to 64, 32, 16, 8, and 4 respectively. The convolution kernel size of all layers is 3, the convolution step is 1, and the padding is 1. After extracting features from the sequence data, it is input into the bidirectional LSTM module. The input tensor size of the network is 4, the output tensor size is 1, the hidden layer is 1, and the number of stacking layers is 1. Then enter the attention module, which calculates the attention vector. The regularization module is used to reduce overfitting during training. The drop rate is set to 0.1. Finally, the training results are output in the fully connected layer. The learning rate of training is set to 0.001, and the number of training rounds is set to 200 rounds.

Then, the pressure sequence data is normalized and divided into a training set and a test set. The pre-trained correction model is trained using the training set and the pre-constructed loss function to obtain a trained solid rocket engine ignition process correction model. Adam is selected as the optimizer for training, and MAE is selected as the loss function. As shown in Figure 3, this is the plotted loss function image.

Finally, the test set is input into the solid rocket engine ignition process correction model to obtain the solid rocket engine ignition process prediction result, as shown in Figure 4, and the predicted pressure curve is output.

The experiment was repeated 5 times and the average error of the prediction results was calculated to reduce randomness, as shown in Table 1.

Table 1

The average values of the 5 prediction errors on the test samples are MAE=0.0063, MSE=0.00009, RMSE=0.0095, R ² =0.9991, and MAPE=3.05%, indicating that the solid rocket engine ignition process model correction method, device and equipment proposed in the present invention have high accuracy.

It should be understood that, although the various steps in the flowchart of FIG. 1 are shown in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a portion of the steps in FIG. 1 may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

In one embodiment, as shown in FIG5 , a solid rocket engine ignition process model correction device is provided, comprising: a data processing module 402, a model building module 404, a model pre-training module 406 and a prediction result generation module 408, wherein:

The data processing module 402 is used to obtain pressure sequence data, normalize the pressure sequence data, obtain a model data set, and divide the model data set into a training set and a test set.

The model building module 404 is used to build a pre-trained correction model of the solid rocket engine ignition process.

The model pre-training module 406 is used to train the pre-trained correction model through the training set and the pre-constructed loss function to obtain a trained solid rocket engine ignition process correction model; wherein, the pre-trained correction model performs deep convolution feature sequence extraction on the input training set to obtain a deep convolution feature sequence; after the deep convolution feature sequence is activated, feature extraction is performed through the forward input and the reverse input respectively and the splicing and fusion are performed to obtain a new feature sequence; the new feature sequence is used to generate an attention vector through the attention mechanism, and the attention vector is regularized; the regularized attention vector is integrated to generate the final prediction output; at the same time, the generated prediction result process is guided and constrained by the pre-constructed loss function.

The prediction result generation module 408 is used to input the test set into the solid rocket engine ignition process correction model to obtain the solid rocket engine ignition process prediction result.

The specific definition of the solid rocket engine ignition process model correction device can be found in the above definition of the solid rocket engine ignition process model correction method, which will not be repeated here. The various modules in the above-mentioned solid rocket engine ignition process model correction device can be implemented in whole or in part through software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG6 . The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store solid rocket engine ignition process model correction data. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a solid rocket engine ignition process model correction method is implemented.

Those skilled in the art will understand that the structure shown in FIG. 6 is merely a block diagram of a partial structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

Step 202, obtaining pressure sequence data, normalizing the pressure sequence data to obtain a model data set, and dividing the model data set into a training set and a test set.

Step 204, construct a pre-trained correction model for the solid rocket motor ignition process.

Step 206, training the pre-trained correction model through the training set and the pre-constructed loss function to obtain a trained solid rocket engine ignition process correction model; wherein, the pre-trained correction model performs deep convolution feature sequence extraction on the input training set to obtain a deep convolution feature sequence; after the deep convolution feature sequence is activated, feature extraction is performed through the forward order input and the reverse order input and the splicing and fusion are performed to obtain a new feature sequence; the new feature sequence is used to generate an attention vector through the attention mechanism, and the attention vector is regularized; the regularized attention vector is integrated to generate the final prediction output; at the same time, the generated prediction result process is guided and constrained by the pre-constructed loss function.

Step 208: input the test set into the solid rocket engine ignition process correction model to obtain a solid rocket engine ignition process prediction result.

A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and/or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

The above-described embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims (9)

1. A solid rocket engine ignition process model correction method, characterized in that the method comprises: Acquire pressure sequence data, perform normalization processing on the pressure sequence data to obtain a model data set, and divide the model data set into a training set and a test set; Construct a pre-trained correction model for the solid rocket motor ignition process; The pre-trained correction model is trained by the training set and the pre-constructed loss function to obtain a trained solid rocket engine ignition process correction model; wherein the pre-trained correction model performs deep convolution feature sequence extraction on the input training set to obtain a deep convolution feature sequence; after the deep convolution feature sequence is activated, feature extraction is performed through positive order input and reverse order input respectively and splicing and fusion are performed to obtain a new feature sequence; the new feature sequence is used to generate an attention vector through an attention mechanism, and the attention vector is regularized; the regularized attention vector is integrated to generate a final prediction output; at the same time, the generated prediction result process is guided and constrained by the pre-constructed loss function; The test set is input into the solid rocket engine ignition process correction model to obtain a solid rocket engine ignition process prediction result. 2. The solid rocket engine ignition process model correction method according to claim 1, characterized in that the pre-training correction model includes a deep convolution module, a bidirectional LSTM module, an attention module, a regularization module and a fully connected layer; Performing deep convolution feature sequence extraction on the input training set through the deep convolution module to obtain a deep convolution feature sequence, and activating the deep convolution feature sequence; The activated deep convolution feature sequence is input in forward order and reverse order respectively through the bidirectional LSTM module, and concatenated and fused after feature extraction to obtain a new feature sequence; The new feature sequence is calculated by the attention module to obtain an attention vector; Regularizing the attention vector by the regularization module; The regularized attention vector is integrated through the fully connected layer to generate the final prediction output. 3. The solid rocket engine ignition process model correction method according to claim 2 is characterized in that the calculation formula of the deep convolution module is expressed as: ; In the formula, Represents a deep convolutional feature sequence; /> Indicates weight; /> Represents the input sequence data; /> represents the bias; where, /> ,/> . 4. The solid rocket engine ignition process model correction method according to claim 3 is characterized in that the deep convolution module includes a 5-layer convolution structure. 5. The solid rocket engine ignition process model correction method according to claim 2, characterized in that the calculation formula of the bidirectional LSTM module is expressed as: ; In the formula, Represents the hidden layer state; /> Indicates positive sequence input; /> Indicates reverse order input; /> and/> Respectively represent the weight parameters of the reverse hidden layer and the reverse hidden layer; /> Represents the bias parameter of the hidden layer; /> A sequence of features representing the activation function. 6. The solid rocket engine ignition process model correction method according to claim 2, characterized in that the calculation formula of the attention module is expressed as: ; In the formula, Represents the attention vector; /> Represents the hidden layer state; /> Represents the weight of the input feature; /> Represents the bias of the input feature. 7. The solid rocket engine ignition process model correction method according to claim 1 or 2 is characterized in that the pre-constructed loss function includes mean absolute error, mean square error, root mean square error, determination coefficient and mean absolute percentage error. 8. A solid rocket engine ignition process model correction device, characterized in that the device comprises: A data processing module, used for acquiring pressure sequence data, performing normalization processing on the pressure sequence data to obtain a model data set, and dividing the model data set into a training set and a test set; A model building module, which is used to build a pre-trained correction model of the solid rocket motor ignition process; A model pre-training module is used to train the pre-trained correction model through the training set and the pre-constructed loss function to obtain a trained solid rocket engine ignition process correction model; wherein the pre-trained correction model performs deep convolution feature sequence extraction on the input training set to obtain a deep convolution feature sequence; after the deep convolution feature sequence is activated, feature extraction is performed through positive order input and reverse order input respectively and splicing and fusion are performed to obtain a new feature sequence; the new feature sequence is used to generate an attention vector through an attention mechanism, and the attention vector is regularized; the regularized attention vector is integrated to generate a final prediction output; at the same time, the generated prediction result process is guided and constrained by the pre-constructed loss function; A prediction result generation module is used to input the test set into the solid rocket engine ignition process correction model to obtain a solid rocket engine ignition process prediction result. 9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein the processor implements the steps of the method according to any one of claims 1 to 7 when executing the computer program.