CN114861538B - Method and system for estimating retired radiation field of post-treatment plant based on random forest algorithm - Google Patents

Abstract

The invention discloses a method and a system for estimating retired radiation fields of a post-treatment plant based on a random forest algorithm, which are implemented by acquiring the position and the distribution of source item data; generating a plurality of different radiation field data based on a random forest algorithm according to the position and the distribution of the source item data to obtain sample data; dividing the sample data into a training set and a testing set; training a random forest algorithm by using the training set to obtain a plurality of decision trees; generating a plurality of classification results by using the test set and the plurality of decision trees, and taking the average value of the plurality of classification results as a final classification result; and performing visual reduction on the final classification result to obtain an estimation result of the retired radiation field of the post-treatment plant. The invention can reduce the deviation between the simulated radiation field data and the field measurement level of the retired nuclear facilities, thereby solving the key problem that the deviation between the simulated radiation field data and the field measurement level of the retired nuclear facilities is larger and restricts the practical application of the virtual simulation technology.

Description

Method and system for estimating retired radiation field of post-treatment plant based on random forest algorithm

Technical Field

The invention relates to the technical field of data processing, in particular to a method and a system for estimating retired radiation fields of a post-treatment plant based on a random forest algorithm.

Background

The nuclear facility retirement operation process has high radioactivity and high complexity, and a large amount of related scientific research work needs to be carried out in the early stage. In the process of making a retirement engineering plan with safety, reliability and economy, a large amount of reliability calculation and data acquisition of a three-dimensional radiation field in a nuclear facility retirement environment are important. In consideration of the fact that the irradiation of operators is kept at the lowest possible level and the decommissioning cost is saved, the virtual reality technology is applied to the calculation and estimation of the three-dimensional radiation field in the decommissioning process of the nuclear facility, and the method has enough guarantee on the safety of the operators and outstanding advantages in economy and reliability.

The nuclear facility retirement engineering is very complex, the design is high in radiation and pollution, the task links are many, and the related data are huge. Foreign nuclear facilities are recently beginning to face retirement, and related digitization technology is not enough in storage and still needs to be quickened. In addition, the retirement industry currently lacks analog simulation tools for digital retirement, and the efficiency of nuclear facility retirement is low, so that time and cost control is very difficult.

Two technical problems need to be solved in the process of calculating the three-dimensional radiation field: firstly, calculating the radiation field of the whole space in the whole factory building in space; secondly, it is necessary to give a new dose distribution with the removal of the device in as short a time as possible. Conventional methods for calculating a three-dimensional radiation field include: discrete longitudinal label method, monte Carlo method and kernel method. The discrete longitudinal mark method is suitable for solving the deep penetration problem, but is not suitable for calculating complex source items and geometric areas and is not suitable for rapidly calculating a radiation field; the Monte Carlo method can be used for calculating complex source items and geometric areas, but cannot give reliable results in reasonable time for shielding from large space; the radiation shielding problem calculated by the point-kernel method is suitable for the deep penetration problem, the calculation speed is high, the method is not suitable for the calculation of complex source items and geometric areas, and meanwhile, larger calculation errors can be brought to the estimation of accumulation factors, and the reliability is low. Therefore, it is difficult to solve both problems simultaneously with the conventional single estimation method for the retired radiation field of the post-treatment plant.

In view of this, the present application has been made.

Disclosure of Invention

The technical problems to be solved by the invention are as follows: how to reduce the deviation between the simulated radiation field data and the retired nuclear facility field measurement level, the purpose is to provide a post-treatment plant retired radiation field estimation method and system based on a random forest algorithm, aiming at the complex source item distribution characteristics of the nuclear facility field, the rapid and accurate reconstruction of the radiation field based on a machine learning algorithm is developed, thereby solving the problem of larger deviation between the simulated radiation field data and the retired nuclear facility field measurement level, promoting the practical application of a virtual simulation technology, and providing a technical support means for the optimization and decision of a nuclear facility retired scheme.

The invention is realized by the following technical scheme:

in one aspect, the invention provides a method for estimating retired radiation fields of a post-treatment plant based on a random forest algorithm, which comprises the following steps:

S1: acquiring the position and distribution of source item data;

s2: generating a plurality of different radiation field data based on a random forest algorithm according to the position and the distribution of the source item data to obtain sample data;

S3: dividing the sample data into a training set and a testing set;

s4: training a random forest algorithm by using the training set to obtain a plurality of decision trees;

S5: generating a plurality of classification results by using the test set and the plurality of decision trees, and taking the average value of the plurality of classification results as a final classification result;

s6: and performing visual reduction on the final classification result to obtain an estimation result of the retired radiation field of the post-treatment plant.

As a further description of the present invention,

The step S3 is preceded by preprocessing the sample data, and comprises the following steps:

Carrying out normalization processing on the sample data in a range scaling mode;

and carrying out augmentation expansion on the sample data.

As a further description of the present invention,

Before the step S5, the method comprises the following steps: the complexity of the model is reduced by adjusting parameters, so that the generalization error is minimized.

As a further description of the present invention,

The step S2 comprises the following steps:

S21: constructing a feature matrix according to the position and the distribution of the source item data, and establishing a label aiming at the missing value in the feature matrix;

S22: filling all positions with labels in the feature matrix with a number 0 to obtain the sample data.

As a further description of the present invention,

The visual reduction comprises the following steps:

Filling the missing values in the feature matrix by using a random forest regression algorithm;

Returning the filled missing value to the feature matrix corresponding to the original data sample;

Saving the filled data sample to a two-dimensional array, and assigning the scene data analyzed in the data sample to corresponding map coordinates;

and reading and analyzing the data in the two-dimensional array, and generating a map according to the analysis result.

As a further description of the present invention,

Filling the missing values in the feature matrix by using a random forest regression algorithm comprises the following steps:

circularly executing A1 to A4 until the leaf node is accessed, and returning the predicted value of the leaf node;

a1: judging whether the current node is a leaf node or not from the root node of the binary decision tree;

A2: if yes, returning the predicted value of the leaf node;

a3: comparing the value of the corresponding variable in the sample data with the segmentation value of the current node according to the segmentation value of the segmentation variable of the current node;

A4: if the value of the corresponding variable in the sample data is less than the segmentation value of the current node, accessing the left child node of the current node; if the value of the corresponding variable in the sample data is greater than the segmentation value of the current node, accessing the right child node of the current node.

In another aspect, the present invention provides a system for estimating a retired radiation field of a post-treatment plant based on a random forest algorithm, comprising:

the data acquisition module is used for acquiring the position and distribution of the source item data;

The sample generation module is used for generating a plurality of different radiation field data based on a random forest algorithm according to the position and the distribution of the source item data to obtain sample data;

the data splitting module is used for dividing the sample data into a training set and a testing set;

The model training module is used for training a random forest algorithm by utilizing the training set to obtain a plurality of decision trees;

The model test module is used for generating a plurality of classification results by utilizing the test set and the plurality of decision trees, and taking the average value of the plurality of classification results as a final classification result;

And the visual reduction module is used for performing visual reduction on the final classification result to obtain an estimation result of the retired radiation field of the post-treatment plant.

As a further description of the present invention,

The system further comprises:

The preprocessing module is used for preprocessing the sample data;

and the error adjusting module is used for adjusting parameters, reducing the complexity of the model and enabling the generalization error to be minimum.

As a further description of the present invention,

The sample generation module includes:

The matrix and label construction unit is used for constructing a feature matrix according to the position and the distribution of the source item data and constructing labels aiming at the missing values in the feature matrix;

And the matrix filling unit is used for filling all the positions with the labels in the feature matrix with the number 0 to obtain the sample data.

As a further description of the present invention,

The visual reduction module comprises:

The missing value feedback unit is used for filling missing values in the feature matrix by using a random forest regression algorithm;

the missing value feedback unit is used for returning the filled missing value to the feature matrix corresponding to the original data sample;

The coordinate assignment unit is used for storing the filled data samples into a two-dimensional array and assigning the scene data analyzed in the data samples to corresponding map coordinates;

and the mapping generation unit is used for reading and analyzing the data in the two-dimensional array and generating a mapping according to the analysis result.

Compared with the prior art, the invention has the following advantages and beneficial effects:

According to the method and the system for estimating the retired radiation field of the post-treatment plant based on the random forest algorithm, aiming at the complex source item distribution characteristics of the nuclear facility site, the rapid and accurate reconstruction of the radiation field based on the machine learning algorithm is carried out, and the deviation between simulated radiation field data and the retired nuclear facility site measurement level can be reduced, so that the key problem that the actual application of a virtual simulation technology is restricted due to the large deviation between the simulated radiation field data and the retired nuclear facility site measurement level is solved, and a technical support means is provided for optimizing and deciding a nuclear facility retired scheme.

Drawings

In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the drawings that are needed in the examples will be briefly described below, it being understood that the following drawings only illustrate some examples of the present invention and therefore should not be considered as limiting the scope, and that other related drawings may be obtained from these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a logic architecture diagram of a method for estimating a retired radiation field of a post-treatment plant based on a random forest algorithm according to an embodiment of the present invention;

FIG. 2 is a flowchart of a method for estimating a retired radiation field of a post-treatment plant based on a random forest algorithm according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of a relationship between generalization error and model complexity according to an embodiment of the present invention.

Detailed Description

For the purpose of making apparent the objects, technical solutions and advantages of the present invention, the present invention will be further described in detail with reference to the following examples and the accompanying drawings, wherein the exemplary embodiments of the present invention and the descriptions thereof are for illustrating the present invention only and are not to be construed as limiting the present invention.

Example 1

Aiming at the problem that deviation between simulated radiation field data and field measurement level of retired nuclear facilities is large, so that practical application of a virtual simulation technology is restricted, the embodiment provides a retired radiation field estimation method of a post-treatment plant based on a random forest algorithm, and aiming at the complex source item distribution characteristics of the nuclear facilities, gamma radiation field rapid and accurate reconstruction based on a machine learning algorithm is developed, and the method logic is shown in figure 1.

The problem of three-dimensional radiation field reconstruction in retired scenes is essentially the problem of analysis and prediction of unknown radiation field data, and the attribute of the problem is very consistent with the mechanism of machine learning. Therefore, further research on key core technologies of virtual simulation of retired nuclear facilities by means of advanced algorithm theory of machine learning has become a trend to effectively solve important problems in the field.

As can be seen from fig. 1, the object of the present invention is to establish a technical support for the simulation of retired radiation fields in a post-treatment plant, and to achieve this object, precise reconstruction of gamma radiation fields in a scene containing unknown source items is required; aiming at the problem of how to realize accurate reconstruction, the embodiment adopts a random forest algorithm to carry out frame design, standard evaluation and optimization, and further needs to generate a training data set aiming at a specific retired scene, wherein the generation of the data set can be realized by a radiation source item investigation and measurement mode.

Considering the problem of three-dimensional radiation field reconstruction in retired scenes, the problem of analysis and prediction of unknown radiation field data is essentially solved, and the attribute of the problem is very consistent with the mathematical mechanism of front-edge deep learning, so that the system is carried out in a random forest mode, and the method is realized by the following steps:

Step 1: and acquiring the position and distribution of the source item data.

The source data structure is shown in table 1 below:

TABLE 1 Source data Structure Table

x	y	z	α	β	γ
						1	2	1	0.02	0.21	5.12
1	3	1	0.02		4.3

Wherein x, y and z represent three-dimensional coordinates of source item data, and alpha, beta and gamma represent radiation field quantity labels.

Step 2: and generating a plurality of different radiation field data based on a random forest algorithm according to the position and the distribution of the source item data to obtain sample data. The method comprises the following steps:

s21: and constructing a feature matrix according to the position and the distribution of the source item data, and establishing a label aiming at the missing value in the feature matrix.

For data with n features, where feature T has a missing value, then feature T is taken as a tag, and the other n-1 features and the original tag form a new feature matrix. For T, the part that is not missing is y_test, which has both a tag and a feature, and the part that is missing has only a feature that is not a tag, which is the part that needs to be predicted.

The corresponding computer executable code is:

fillc＝df.iloc[:,i]

df＝pd.concat([df.iloc[:,df.columns！＝],pd.DataFrame(y_full)],axis＝1)

S22: filling all positions with labels in the feature matrix with a number 0 to obtain the sample data. As shown in table 2:

table 2 filled source item data structure table

x	y	z	α	β	γ
						1	2	1	0.02	0.21	5.12
1	3	1	0.02	0	4.3

The corresponding computer executable code is:

df_0＝SimpleImputer(missing_values＝np.nan,

strategy＝'constant',fill_value＝0).fit_transform(df)

step3: preprocessing the sample data, comprising the following steps:

S31: carrying out normalization processing on the sample data in a range scaling mode; the data is shifted by a minimum number of units and will be converged between 0, 1.

The corresponding computer executable code is:

x_MinMax＝preprocessing.MinMaxScaler()

df['x_scaler']＝x_MinMax.fit_transform(df[['x']])

s32: and carrying out augmentation expansion on the sample data.

Step 4: the sample data is divided into a training set and a testing set.

The corresponding computer executable code is:

Ytrain＝fillc[fillc.notnull()]

Ytest＝fillc[fillc.isnull()]

Xtrain＝df_0[Ytrain.index,:]

Xtest＝df_0[Ytest.index,:]

step 5: training a random forest algorithm by using the training set to obtain a plurality of decision trees.

The corresponding computer executable code is: rfc=rfc.fit (Xtrain, ytrain)

It should be noted that the number of the components,

(1) When the dependent variable of the data set is a continuity value, the mean value observed by the leaf nodes is used as a predicted value, and a weak classifier CART decision tree is used.

(2) Feature selection of the coefficient of foundation (GINI):

For CART decision tree, due to binary tree, reference can be made to the formula Gini (p) =2p (1-p);

when each division point of each feature is traversed, when the feature a=x is used, γ is divided into two parts, i.e., D1 (sample set satisfying a=x), D2 (sample set not satisfying a=x). The base index under the condition of characteristic a=x is:

Where Gini (D) represents the uncertainty of set D, gini (D, a): the uncertainty of the set D after a=a segmentation is represented.

(3) Each CART decision tree is obtained by continuously traversing all possible segmentation points of the feature subset of the tree, searching for the segmentation point of the feature with the smallest Gini coefficient, and dividing the data set into two subsets until the stopping condition is met.

(4) After t decision trees are generated, for each new test sample, the classification results of the decision trees are synthesized, and the average value of the t decision trees is taken as the classification result.

Step 6: the complexity of the model is reduced by adjusting parameters, so that the generalization error is minimized.

When the model is not good in unknown data (test set or out-of-bag data), the generalization degree of the model is insufficient, the generalization error is large, and the effect of the model is poor. The generalization error is affected by the structure (complexity) of the model. The generalization error versus model complexity is shown in figure 3,

When the model is too complex, the model is over-fitted, and the generalization capability is insufficient, so that the generalization error is large. When the model is too simple, the model is under-fitted, the fitting capacity is insufficient, and therefore the error is large. The goal of minimizing generalization errors can only be achieved if the complexity of the model is just good.

Therefore, the parameter adjustment mode adopted by the embodiment reduces the complexity of the model and minimizes the generalization error.

The adjusted parameters are shown in table 3:

Table 3 parameter adjustment list

The corresponding computer executable code is:

step 7: and generating a plurality of classification results by using the test set and the plurality of decision trees, and taking the average value of the plurality of classification results as a final classification result.

Step 8: and performing visual reduction on the final classification result to obtain an estimation result of the retired radiation field of the post-treatment plant.

Wherein, the visual reduction comprises the following steps:

(1) Random forest regression was used to fill in missing values.

(1.1) Starting from the root node of the binary decision tree, determining whether the current node is a leaf node, and if so, returning to the leaf node

-A predicted value of the point (i.e. the average value X of the sample target variable in the current leaf);

And (1.2) comparing the value of the corresponding variable in the sample with the segmentation value of the node according to the sum of the segmentation variables of the current node. If the sample variable value is less than or equal to the current node score value, accessing a left child node of the current node; if the sample variable value is greater than the current node score value, the right child node of the current node is accessed.

The corresponding computer executable code is:

(1.3) looping steps (1.1) - (1.2) above until a leaf node is accessed and a prediction value for the leaf node is returned.

(2) And returning the filled features to the original feature matrix.

The corresponding computer executable code is:

X_missing_reg.loc[X_missing_reg.iloc[:,i].isnull(),i]＝Ypredict

(3) The data is saved to the two-dimensional array and then read from the two-dimensional array.

(4) And saving the field quantity data analyzed according to the data set to the corresponding map coordinates.

(5) The radiation environment is visualized, data is read from a two-dimensional array, and then analyzed. The analyzed data generates a map and assigns a value to the Image.

The corresponding computer executable code is:

Example 2

The embodiment 1 of the method for estimating a retired radiation field of a post-treatment plant based on a random forest algorithm corresponds to the method for estimating a retired radiation field of a post-treatment plant based on a random forest algorithm, and the embodiment provides a retired radiation field estimating system of a post-treatment plant based on a random forest algorithm, which comprises the following steps:

the data acquisition module is used for acquiring the position and distribution of the source item data;

The sample generation module is used for generating a plurality of different radiation field data based on a random forest algorithm according to the position and the distribution of the source item data to obtain sample data;

the data splitting module is used for dividing the sample data into a training set and a testing set;

The model training module is used for training a random forest algorithm by utilizing the training set to obtain a plurality of decision trees;

The model test module is used for generating a plurality of classification results by utilizing the test set and the plurality of decision trees, and taking the average value of the plurality of classification results as a final classification result;

the visual reduction module is used for performing visual reduction on the final classification result to obtain an estimation result of the retired radiation field of the post-treatment plant;

The preprocessing module is used for preprocessing the sample data;

and the error adjusting module is used for adjusting parameters, reducing the complexity of the model and enabling the generalization error to be minimum.

Wherein,

The sample generation module includes:

The matrix and label construction unit is used for constructing a feature matrix according to the position and the distribution of the source item data and constructing labels aiming at the missing values in the feature matrix;

And the matrix filling unit is used for filling all the positions with the labels in the feature matrix with the number 0 to obtain the sample data.

The visual reduction module comprises:

The missing value feedback unit is used for filling missing values in the feature matrix by using a random forest regression algorithm;

the missing value feedback unit is used for returning the filled missing value to the feature matrix corresponding to the original data sample;

The coordinate assignment unit is used for storing the filled data samples into a two-dimensional array and assigning the scene data analyzed in the data samples to corresponding map coordinates;

and the mapping generation unit is used for reading and analyzing the data in the two-dimensional array and generating a mapping according to the analysis result.

The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the invention, and is not meant to limit the scope of the invention, but to limit the invention to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the invention are intended to be included within the scope of the invention.

Claims (6)

1. The method for estimating the retired radiation field of the post-treatment plant based on the random forest algorithm is characterized by comprising the following steps of:

S1: acquiring the position and distribution of source item data; the position of the source item data is represented by three-dimensional coordinates, and the distribution of the source item data is represented by a radiation field quantity label;

s2: generating a plurality of different radiation field data based on a random forest algorithm according to the position and the distribution of the source item data to obtain sample data;

The step S2 comprises the following steps:

S21: constructing a feature matrix according to the position and the distribution of the source item data, and establishing a label aiming at the missing value in the feature matrix;

S22: filling all positions with labels in the feature matrix with a number 0 to obtain the sample data;

S3: dividing the sample data into a training set and a testing set;

s4: training a random forest algorithm by using the training set to obtain a plurality of decision trees;

S5: generating a plurality of classification results by using the test set and the plurality of decision trees, and taking the average value of the plurality of classification results as a final classification result;

S6: performing visual reduction on the final classification result to obtain an estimation result of the retired radiation field of the post-treatment plant;

The visual reduction comprises the following steps:

Filling the missing values in the feature matrix by using a random forest regression algorithm;

Returning the filled missing value to the feature matrix corresponding to the original data sample;

Saving the filled data sample to a two-dimensional array, and assigning the scene data analyzed in the data sample to corresponding map coordinates;

and reading and analyzing the data in the two-dimensional array, and generating a map according to the analysis result.

2. A method for estimating a retired radiation field of a post-treatment plant based on a random forest algorithm according to claim 1, wherein said pre-processing of said sample data prior to S3 comprises the steps of:

Carrying out normalization processing on the sample data in a range scaling mode;

and carrying out augmentation expansion on the sample data.

3. A method for estimating a retired radiation field of a post-treatment plant based on a random forest algorithm according to claim 1, characterized in that before S5, it comprises the following steps: the complexity of the model is reduced by adjusting parameters, so that the generalization error is minimized.

4. A method for estimating a retired radiation field of a post-treatment plant based on a random forest algorithm according to claim 1, wherein said filling up missing values in said feature matrix using a random forest regression algorithm comprises the steps of:

circularly executing A1 to A4 until the leaf node is accessed, and returning the predicted value of the leaf node;

a1: judging whether the current node is a leaf node or not from the root node of the binary decision tree;

A2: if yes, returning the predicted value of the leaf node;

a3: comparing the value of the corresponding variable in the sample data with the segmentation value of the current node according to the segmentation value of the segmentation variable of the current node;

A4: if the value of the corresponding variable in the sample data is less than the segmentation value of the current node, accessing the left child node of the current node; if the value of the corresponding variable in the sample data is greater than the segmentation value of the current node, accessing the right child node of the current node.

5. A random forest algorithm-based post-treatment plant retired radiation field estimation system, comprising:

the data acquisition module is used for acquiring the position and distribution of the source item data;

The sample generation module is used for generating a plurality of different radiation field data based on a random forest algorithm according to the position and the distribution of the source item data to obtain sample data;

the data splitting module is used for dividing the sample data into a training set and a testing set;

The model training module is used for training a random forest algorithm by utilizing the training set to obtain a plurality of decision trees;

The model test module is used for generating a plurality of classification results by utilizing the test set and the plurality of decision trees, and taking the average value of the plurality of classification results as a final classification result;

the visual reduction module is used for performing visual reduction on the final classification result to obtain an estimation result of the retired radiation field of the post-treatment plant;

The sample generation module includes:

The matrix and label construction unit is used for constructing a feature matrix according to the position and the distribution of the source item data and constructing labels aiming at the missing values in the feature matrix;

The matrix filling unit is used for filling all positions with labels in the feature matrix with a number 0 to obtain the sample data;

The visual reduction module comprises:

The missing value feedback unit is used for filling missing values in the feature matrix by using a random forest regression algorithm;

the missing value feedback unit is used for returning the filled missing value to the feature matrix corresponding to the original data sample;

The coordinate assignment unit is used for storing the filled data samples into a two-dimensional array and assigning the scene data analyzed in the data samples to corresponding map coordinates;

and the mapping generation unit is used for reading and analyzing the data in the two-dimensional array and generating a mapping according to the analysis result.

6. A stochastic forest algorithm-based post-treatment plant retirement radiation field estimation system according to claim 5, further comprising:

The preprocessing module is used for preprocessing the sample data;

and the error adjusting module is used for adjusting parameters, reducing the complexity of the model and enabling the generalization error to be minimum.