CN117150877A - Method for predicting optimal pressing process of press-loading mixed explosive based on Bagging algorithm - Google Patents

Abstract

The invention belongs to the field of explosive press-fitting, and discloses a Bagging algorithm-based optimal press-fitting process prediction method for press-fitting mixed explosive; randomly collecting samples with the same number as 21 samples of the training set, wherein the content of the samples is different, carrying out 1000 times of random sampling on the training set with 21 samples, and the 1000 sampling sets are different due to randomness; carrying out arithmetic average on regression results obtained by 1000 weak learners by using an aggregation strategy of a simple average method to obtain final model output; gradually decreasing from the theoretical density according to the output model, and searching the corresponding optimal process condition under the maximum density; the method for predicting the optimal pressing process of the press-fit mixed explosive based on the Bagging algorithm has the advantages that the optimal process condition is accurately predicted, and the difficult problem that the charging condition is difficult to determine is solved. Providing reference for selecting proper technological conditions and optimizing the technological design of the pressed mixed explosive.

Description

A prediction of the optimal charging process of mixed explosives based on bagging algorithm method

Technical field

The invention relates to a method for predicting the optimal pressing process of mixed explosives based on bagging algorithm, and belongs to the field of explosive pressing.

Background technique

In the production process of press-loaded mixed explosives, the higher the percentage of the press density to the theoretical density, the better the performance of the explosive. The density of explosives is closely related to process conditions. If the process conditions change, the density of the explosive will also change accordingly. How to determine the optimal process conditions so that the density of the pressed explosive is infinitely close to the theoretical density is a major problem in the design and research of press-loaded mixed explosives.

In order to solve the above problems, a method for predicting the optimal charging process of mixed explosives based on Bagging algorithm is proposed. This method has the characteristics of high accuracy, high safety and good economy, and has been used in a CL- 20 base pressure-packing mixed explosive design has been applied.

The Bagging algorithm (Bootstrap aggregating), also known as the bagging algorithm, is a group learning algorithm in the field of machine learning. The Bagging algorithm can be combined with other classification and regression algorithms to improve its accuracy and stability by reducing The variance of the results avoids the occurrence of overfitting.

In this patent, process parameters such as charge specific pressure, charge temperature, and holding time are used as sample features, and charge density is used as sample label. Supervised regression learning training is performed through the bagging algorithm to obtain a charge density prediction model. Using the trained model, starting from the theoretical density, the density is gradually reduced according to the specified step size to obtain the best process conditions.

Contents of the invention

The purpose of this invention is to solve the problem of difficulty in determining the process parameters of press-packed mixed explosives and provide a method for predicting the optimal press-packed mixed explosive process based on the Bagging algorithm. This method creatively introduces machine learning algorithms into the mixed explosive process-performance relationship. During the study, the idea and solution of bidirectional prediction of process parameters and charge density were proposed, which provided the possibility to accelerate the design of mixed explosive formulas.

The object of the present invention is achieved through the following technical solutions.

A method for predicting the optimal charging process of mixed explosives based on Bagging algorithm. The specific steps are as follows:

Step 1. Organize and collect relevant data, and divide the data into training sets x_train, y_train and test sets x_test, y_test. 80% of it is the training set and 20% is the test set.

Step 2: Call the BaggingRegressor module in Python's sklearn library. Set parameters: base_estimator=decision tree, n_estimators=1000. The Bagging algorithm process is as follows:

1) Randomly collect 21 samples from the training set in step 1. As sampling set D, a total of 1000 sampling sets are collected.

2) Use the sampling set D to train 1000 weak learners.

3) The arithmetic average of the regression results obtained by 1000 weak learners is the final model output.

Step 3: Use the test set in Step 1 to evaluate the Bagging training model, and perform error analysis by comparing the true value and the predicted value. Further draw the learning curve to evaluate the generalization effect of model fitting.

Step 4: Write the process-density program, and set the pressing pressure x1 range to 2000-3600kg/cm ² and the step length 300, the pressing temperature x2 range to 20-90°C, the step length 5, and the holding time x3 range 20-75min, step size 5, input the maximum charge density value y and error limit a=1.03*10 ^-5 , loop through x1, x2, x3 in sequence, take one group as input each time, and put it into the third step of training Training is carried out in the prediction model, and the predicted value of charge density is compared with the set value. This method uses the mean square error. If the error is greater than the set error limit a, it does not meet the requirements. On the contrary, if it meets the requirements, it will be output.

Step 5: Select the best process conditions from the output results and verify them through experiments.

beneficial effects

1. The present invention’s method for predicting the optimal charging process of mixed explosives based on the bagging algorithm can determine the optimal process conditions with extremely high accuracy.

2. A method for predicting the optimal charging process of mixed explosives based on the bagging algorithm of the present invention. The bagging algorithm will generally randomly collect the same number of samples as the number m of samples in the training set. The number of samples in the sampling set and the training set obtained in this way is the same, but the content of the samples is different. If we perform T random sampling on a training set with m samples, then the T sampling sets will be different due to randomness. The set strategy of bagging is also relatively simple. For regression problems, the simple average method is usually used, and the regression results obtained by T weak learners are arithmetic averaged to obtain the final model output. This provides a new idea for predicting charge density and has guiding significance for the production and process design of pressure-packed mixed explosives.

3. The present invention's method for predicting the optimal charging process of mixed explosives based on the bagging algorithm can predict the corresponding process conditions by continuously changing the specified density, thereby achieving the purpose of predicting the optimal process and solving the problem of process conditions. Difficult to pin down issues.

Description of the drawings

Figure 1 is a learning curve diagram of the Bagging model in the present invention;

Figure 2 is an evaluation score chart when the number of Bagging sub-trainers in the present invention is 10, 100, 500, 1000, 2000 and 5000;

Figure 3 is a comparison chart between the actual value and the predicted value of the charge density in the present invention;

Figure 4 is the schematic diagram of Bagging algorithm;

Detailed ways

The present invention will be further described below in conjunction with the embodiments and drawings.

Example 1. Training prediction model

First, the process parameters and charge density data set of a CL-20-based press-loaded mixed explosive were organized. The data set contains three characteristics: charge specific pressure, press temperature, and pressure holding time. The label is the numerical size of the charge density. After the data is loaded, the data set is divided into two groups, 80% of which is the training set and 20% of which is the test set, and the bagging regressor is started for training. Next, draw the learning curve of the Bagging regressor. From the Bagging regression learning curve (Figure 1), it can be seen that as the number of samples increases, the test curve gradually decreases, the training curve gradually increases, and the two gradually approach, proving the effect of Bagging regression. better.

Set the number of sub-trainers to 10, 100, 500, 1000, 2000, and 5000 respectively to evaluate the model, and obtain the model scores under different numbers of sub-trainers (Figure 2). The results are shown in the table below.

Table 1: Model scores under different numbers of sub-trainers

It can be seen from the results in Table 1 that for the data set in this example, as the number of sub-trainers increases, the model score first increases and then levels off. When the number of sub-trainers increases, the model training cost increases, so the present invention sets the number of sub-trainers to 1000. The principle of the Bagging algorithm is shown in Figure 4.

Example 2: Use the trained model to make predictions

After the model is trained, use the test set data to predict the model (Figure 3). The results are shown in Table 2.

Table 2: Comparison of actual values and predicted values of the test set

It can be seen from the results shown in Table 2 that the test set data is used to predict the model. The predicted value of the charge density of the method of the present invention is basically the same as the actual measured value, and has a smaller error range.

Example 3

In order to determine the optimal process conditions, the density is taken to be 2.05g·cm ^-3 and the reverse prediction method of the present invention is used to predict the above process parameters. There are no corresponding suitable process conditions. Using 0.005 as the step size, gradually reduce the input density, take the density to be 2.04g·cm ^-3 , and obtain the results shown in Table 3.

Table 3: Taking the density as 2.04g·cm ^-3 to predict the process parameters

result 1 2 Pressing temperature 85 85 Pressure specific pressure 3200 3500 Keeping time 65 60

From the results shown in Table 3, it can be seen that when the specified density is 2.04g·cm ^-3 , there are two kinds of process condition parameters for reverse prediction. After averaging, the optimal temperature is 85°C and the optimal pressure is 3350kg. /cm ² , the optimal holding time is 62.5min, which is close to the real experimental values of 85℃, 3300kg/cm ² and 60min. The prediction is very accurate.

The beneficial effect of using the optimal charging process prediction method for press-loading mixed explosives based on the Bagging algorithm is that the optimal process conditions are accurately predicted and the difficult problem of charging conditions being difficult to determine is solved. Provide a reference for selecting appropriate process conditions and optimizing the process design of press-loading mixed explosives.

Claims (2)

1. A method for predicting the optimal charging process of mixed explosives based on the Bagging algorithm, which is characterized by: the construction idea of the Bagging algorithm is: The bagging algorithm will randomly collect the same number of samples as the number of training set samples, 21. The number of samples in the sampling set and the training set obtained in this way is the same, but the content of the samples is different. If we randomly sample a training set of 21 samples 1000 times, the 1000 sampling sets will be different due to randomness. The set strategy of bagging is also relatively simple. For regression problems, the simple average method is usually used, and the regression results obtained by 1000 weak learners are arithmetic averaged to obtain the final model output. 2. A method for predicting the optimal charging process of mixed explosives based on Bagging algorithm, which is characterized by: the optimal process determination method is: The density is taken to be 2.05g·cm ^-3 and the reverse prediction method of the present invention is used to predict the above process parameters. There are no corresponding suitable process conditions. Using 0.005 as the step size, gradually reduce the input density and take the density to be 2.04g·cm ^-3 . There are 2 process condition parameters for reverse prediction. After averaging, the optimal temperature is 85°C and the optimal pressure is determined. It is 3350kg/cm ² and the optimal holding time is 62.5min, which is very close to the real experimental values of 85℃, 3300kg/ ^cm2 and 60min. The prediction is very accurate.