CN110276497A - A point prediction method for air tightness index of granary building - Google Patents

Abstract

The invention relates to the technical field of grain and oil storage, and discloses a point prediction method for the air tightness index of a granary building. The result of the point prediction is a certain value. Since the number of collected samples is small, the support vector machine is used in this study as the The regression tool includes the following steps. First, the original data is preprocessed. Second, the data is divided into training set and test set. The third step is to train the support vector machine with the training set. The fourth step is to test the trained support vector machine with the test set. Model, the fifth step is to use support vector machine to make predictions. The invention preprocesses the data on the basis of obtaining the airtightness data of the granary building. The preprocessing method includes normalizing the numerical variables and performing onehot coding on the type variables, thereby effectively improving the calculation accuracy and generalization of the model. According to the prediction results, the air tightness prediction of the granary can be realized by using small sample data.

Description

A point prediction method for air tightness index of granary building

technical field

The invention relates to the technical field of grain and oil storage, in particular to a point prediction method for an air-tightness index of a granary building.

Background technique

Grain is an important material, but in the storage stage, food loss is very serious. Taking technical measures to reduce the loss of grain in the storage stage is of great significance for ensuring food security. Studies have shown that the main causes of food loss during the storage stage include insect pests and mildew, and factors such as temperature and humidity will affect the quality of grain during storage. For insect pests in the grain storage stage, currently available measures include insecticide fumigation and carbon dioxide fumigation. Ozone fumigation is an effective method to avoid grain mildew during storage and to control the temperature and humidity of grain. By adjusting the air environment for grain storage, reducing the oxygen content in the air, and increasing the carbon dioxide content, it has also been proved to be effective in controlling food pests. It can be seen that grain storage has high requirements for the air environment, and the grain storage facilities must have good air tightness. For grain storage facilities, China's national standard for air-tightness of bungalows adopts pressure decay to test the air-tightness of the granary. The specific method is: use a fan to press the air into the granary, so that the pressure inside and outside the granary reaches the specified pressure difference (500Pa ) and then shut down, and judge the air tightness of the granary according to the half-life of the pressure difference.

The granary building needs to form a closed space to ensure that it meets the air tightness requirements, and the air tightness of the granary building depends to a large extent on the architectural design plan, the selection of the silo type in the architectural design plan and the measures taken to meet the air tightness requirements. Various structural measures will affect the air tightness of the granary. The current results mainly focus on studying the influence of specific architectural design measures and building structures on air tightness, and there is a lack of methods to evaluate and predict the air tightness of granary from an overall perspective. Predicting the airtightness of the design scheme during the design stage of the granary can provide technical reference for designers, discover the defects of the design scheme in time, and take remedial measures to ensure that the granary can meet the airtightness requirements after completion. At the same time, the airtightness prediction of the granary can avoid unnecessary cost expenditure caused by excessive airtightness technical measures, resulting in excessive airtightness redundancy of the granary.

However, there are many factors that affect the air tightness of the granary, and there is a complex nonlinear relationship between the influencing factors and the air tightness index. This patent uses a machine learning method to try to reveal the quantitative relationship between the design scheme of the granary and the air tightness. , so as to predict the air tightness of the granary design scheme.

SUMMARY OF THE INVENTION

(1) Technical problems solved

In view of the deficiencies of the prior art, the present invention provides a point prediction method for the airtightness index of a granary building, which has the advantages of high prediction effect, simple and uncomplicated operation, etc. more complex issues.

(2) Technical solutions

In order to realize the above-mentioned purpose of high prediction effect, simple and uncomplicated operation, the present invention provides the following technical scheme: a point prediction method for airtightness index of granary buildings, characterized in that the result of point prediction is a definite value, and due to collecting The number of samples obtained is small, this study uses support vector machine as a regression tool, including the following steps:

Step1: Randomly divide the original data set into training set and test set;

Step2: Data preprocessing, normalize the numerical variables in the cause variables, and perform onehot coding 3 for the category variables in the cause variables;

Step3: Use the training set for SVM training to obtain a regression model f( ), and the training parameters of SVM are optimized by grid;

Step4: Substitute the test set into the trained regression model f( ) to obtain the predicted value {t ₁ ^p ,...,t _m ^p } of the test set, and the predicted set includes m pieces of data;

Step5: Calculate the average relative error of the model, the calculation formula is Determine whether the average relative error meets the requirements. If it meets the requirements, go to Step 6. If not, adjust the parameters of the regression model and repeat Step 3 to Step 5;

Step6: Substitute the data x ^* of the granary design scheme to be evaluated into the support vector machine model f(·), get the air tightness prediction result t ^* , check whether the air tightness requirements can be met, and if so, the design scheme can be determined , if it cannot be satisfied, the design scheme needs to be adjusted and the air tightness is re-evaluated.

Preferably, the original data set is Φ, Φ={(x ₁ ,t ₁ ),...,(x _l ,t _l )}, where x _i ∈R ⁿ is the cause vector, and t _i ∈ R ¹ is the cause vector The resulting value, l∈N ⁺ .

Preferably, the original data set is in the form of a matrix, each row of the matrix corresponds to a piece of data, and each column corresponds to a variable.

Preferably, the original data set is a matrix with 104 rows and 19 columns.

(3) Beneficial effects

Compared with the prior art, the present invention provides a point prediction method for the airtightness index of a granary building, which has the following beneficial effects: the present invention preprocesses the data on the basis of obtaining the airtightness data of the granary building, The processing methods include normalizing the numerical variables and onehot encoding the category variables, thereby effectively improving the computational accuracy and generalization ability of the model. This patent uses the support vector machine regression model to establish the mapping relationship between the influencing factors and the air tightness of the granary building, and then realizes the prediction of the air tightness of the granary by using small sample data according to the prediction results.

Description of drawings

Fig. 1 is the point prediction flow chart of the air-tightness index of granary building of the present invention;

FIG. 2 is a point prediction result of the air tightness index of the granary building of the present invention and a comparison diagram with the actual value.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

The invention provides a point prediction method for the air tightness index of a granary building. First, data is preprocessed, and the preprocessing of the data can improve the calculation efficiency, the accuracy of the calculation and the generalization ability. The processing mainly includes the normalization of numerical variables and the onehot coding of category variables. The so-called numerical variables refer to variables whose values can be sorted and compared, and category variables refer to variables whose values cannot be sorted and compared.

(1) Use the following formula to normalize the numerical variables

The input data can be mapped to the interval [0,1] by the above formula;

(2) For the category variable, onehot encoding is required, that is, the category number is rewritten into the form of a binary vector.

Divide the data into training sets and test sets. In the specific implementation process, the data used in this study come from the survey results of granaries in South China, Central China and Northeast China. After sorting out, there are 104 pieces of data. According to the opinions of granary design and operation management experts , the data set includes 18 cause variables and 1 result variable, the variables are described as follows:

Table 1. Characteristics of air-tightness variables of granary buildings

property name unit Kind of variable *500Pa pressure half-life sec. Numerical value Wall structure layer thickness mm Numerical value warehouse type — type The quantity of grain intake (unloading) in a single warehouse number Numerical value Single warehouse volume m³ Numerical value Single warehouse building area m² Numerical value Single warehouse floor height m Numerical value Number of Axial Fan Ports in a Single Chamber number Numerical value Door and window sealing measures — type Wall Structure Layer Type — type Wall moisture barrier height m Numerical value Number of Circulation Fumigation Holes in Single Warehouse number Numerical value Number of mechanical vents in a single compartment number Numerical value Single warehouse door and window area m² Numerical value Number of natural ventilation openings in a single warehouse number Numerical value Warehouse ventilation form — type roof structure — type Wall moisture barrier — type Granary ground practice — type

Note: Attributes with * are result attributes.

The data set is in the form of a matrix, each row of the matrix corresponds to a piece of data, and each column corresponds to a variable. The data set used in the present invention is a matrix with 104 rows and 19 columns.

Use the above training set to train the support vector machine. The support vector machine has advantages in processing small sample data. The support vector machine can complete the tasks of classification, regression and distribution estimation. It is assumed that the data format is {(x ₁ ,t ₁ ),… ,(x _l ,t _l )}, where x _i ∈ R ⁿ is the cause vector, and t _i ∈ R ¹ is the target output. The above data can be used to train the support vector machine.

The result of point prediction is a definite value. Due to the small number of samples collected, this study uses support vector machine as a regression tool. The specific methods are as follows:

Input: original data set Φ,Φ={(x ₁ ,t ₁ ),…,(x _l ,t _l )}, where x _i ∈R ⁿ is the cause vector, t _i ∈ R ¹ is the result value, l∈ N ⁺ .

Output: point predicted value t ^* .

Step1: Randomly divide the original data set into training set and test set;

Step2: Data preprocessing, normalize the numerical variables in the cause variables, and perform onehot coding on the type variables in the cause variables;

Step3: Use the training set for SVM training to obtain a regression model f( ), and the training parameters of SVM are optimized by grid;

Step4: Substitute the test set into the trained regression model f( ) to obtain the predicted value {t ₁ ^p ,...,t _m ^p } of the test set, and the predicted set includes m pieces of data;

Step5: Calculate the average relative error of the model, the calculation formula is as follows

Determine whether the average relative error meets the requirements. If it meets the requirements, go to Step 6. If not, adjust the parameters of the regression model and repeat Step 3 to Step 5;

Step6: Substitute the data x ^* of the granary design scheme to be evaluated into the support vector machine model f(·), get the air tightness prediction result t ^* , check whether the air tightness requirements can be met, and if so, the design scheme can be determined , if it cannot be satisfied, the design scheme needs to be adjusted and the air tightness is re-evaluated.

The point prediction process of the air tightness index of the granary building is shown in Figure 1.

Among the 104 pieces of original data, 20 pieces of data were randomly selected to form the test set, and the other 84 pieces of data formed the training set.

Table 2 The point prediction results of the air tightness index of the granary building and the comparison with the actual value

data number 1 2 3 4 5 actual value 134 300 60 40 300 Predictive value 139 300 60 40 300 data number 6 7 8 9 10 actual value 40 137 325 40 650 Predictive value 40 139 300 40 900 data number 11 12 13 14 15 actual value 98 40 40 40 124 Predictive value 95 40 40 40 139 data number 16 17 18 19 20 actual value 40 40 40 100 40 Predictive value 40 40 40 110 40

The beneficial effects of the present invention are as follows: the present invention preprocesses the data on the basis of obtaining the airtightness data of the granary building, and the preprocessing method includes normalizing the numerical variables and performing onehot coding on the type variables, thereby effectively improving the Computational accuracy and generalization ability of the model. This patent uses the support vector machine regression model to establish the mapping relationship between the influencing factors and the air tightness of the granary building, and then realizes the prediction of the air tightness of the granary by using small sample data according to the prediction results.

Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, and substitutions can be made in these embodiments without departing from the principle and spirit of the invention and modifications, the scope of the invention is defined by the appended claims and their equivalents.

Claims (4)

1. a kind of point prediction method of barn building air-tightness index, which is characterized in that point prediction the result is that one determining Value, since the sample size collected is less, this research use support vector machines as recurrence tool the following steps are included:

Step1: being that training set and test are gathered by original data set random division；

Step2: the numerical variable in causal variable is normalized in data prediction, and the type in causal variable is become Amount carries out onehot coding 3；

Step3: training set is shared into the training in support vector machines and obtains regression model f (), the training ginseng of support vector machines Number uses grid optimizing；

Step4: the regression model f () after substituting into training is gathered into test, obtains the predicted value { t of test set₁ ^p,…,t_m ^p, It altogether include m data in prediction sets；

Step5: the average relative error of computation model, calculation formula are

Judge whether average relative error meets the requirements, if met the requirements, into Step6, if conditions are not met, carrying out recurrence mould Shape parameter adjustment, repeats Step3~Step5；

Step6: by the data x of practice and text scheme to be evaluated^*It substitutes into supporting vector machine model f (), obtains air-tightness prediction As a result t^*, can inspection meet air-tightness requirement, if can satisfy, can determine design scheme, if be not able to satisfy, need Adjusted design scheme is wanted, and reappraises air-tightness.

2. a kind of point prediction method of barn building air-tightness index according to claim 1, which is characterized in that described Raw data set is combined into Φ, Φ={ (x₁,t₁),…,(x_l,t_l), wherein x_i∈RⁿIt is reason vector, t_i∈R¹It is end value, l ∈N⁺。

3. a kind of point prediction method of barn building air-tightness index according to claim 1, which is characterized in that the original Beginning data acquisition system be matrix form, each one data of behavior of matrix, it is each column it is corresponding with a variable.

4. a kind of point prediction method of barn building air-tightness index according to claim 1, which is characterized in that described It is 104 rows, 19 matrixes arranged that raw data set, which is combined into,.