CN112733439A - Method for calculating shielding material accumulation factor based on BP neural network - Google Patents
Method for calculating shielding material accumulation factor based on BP neural network Download PDFInfo
- Publication number
- CN112733439A CN112733439A CN202011602928.9A CN202011602928A CN112733439A CN 112733439 A CN112733439 A CN 112733439A CN 202011602928 A CN202011602928 A CN 202011602928A CN 112733439 A CN112733439 A CN 112733439A
- Authority
- CN
- China
- Prior art keywords
- layer
- output
- neural network
- weight
- calculation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 27
- 238000009825 accumulation Methods 0.000 title claims abstract description 23
- 239000000463 material Substances 0.000 title claims abstract description 16
- 238000004364 calculation method Methods 0.000 claims abstract description 30
- 210000002569 neuron Anatomy 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 20
- 230000008569 process Effects 0.000 claims abstract description 13
- 230000006870 function Effects 0.000 claims description 21
- 238000004422 calculation algorithm Methods 0.000 claims description 13
- 230000004913 activation Effects 0.000 claims description 12
- 238000003062 neural network model Methods 0.000 claims description 8
- 230000005251 gamma ray Effects 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 claims description 3
- 210000004205 output neuron Anatomy 0.000 claims description 3
- 238000012360 testing method Methods 0.000 abstract description 8
- 230000005855 radiation Effects 0.000 abstract description 4
- 238000010606 normalization Methods 0.000 abstract description 2
- 238000012545 processing Methods 0.000 abstract description 2
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 229910052742 iron Inorganic materials 0.000 description 2
- 230000000694 effects Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for calculating an accumulation factor of a shielding material based on a BP neural network. The invention relates to the technical field of radiation shielding calculation for determinism; carrying out forward propagation, wherein the forward ship process is that an input signal is transmitted to an output layer from an input layer through a hidden layer neuron, an output signal is generated at an output end, and when the output signal meets a given output requirement, the calculation is finished; when the output signal does not meet the given output requirement, the signal is shifted to the reverse propagation. The invention divides the data into a training set and a testing set according to a reasonable proportion, and carries out normalization processing on the data. The network is trained using a training set and validated using a test set. The optimal hidden layer and number of neurons per layer can be determined by trial and error.
Description
Technical Field
The invention relates to the technical field of radiation shielding calculation for determinism, in particular to a method for calculating an accumulation factor of a shielding material based on a BP neural network.
Background
The point-kernel integration method is a method for simplifying the calculation of radiation dose field distribution in three-dimensional space. Point-kernel integration simplifies the radiation dose field distribution calculation in a complex three-dimensional space into two parts, the first part is the calculation of the direct-through term which does not interact with the shielding material, and the second part is the calculation of the accumulation factor of the interaction between the particles and the material. For the calculation of the accumulation factor, the fitting taylor formula and G-P formula are mainly used in the conventional calculation method. For the fitting formula, the expression is complex, and the calculation parameters need to be subjected to table lookup for linear interpolation calculation in practical application, so that certain errors are brought to the calculation of the accumulation factors.
Disclosure of Invention
The invention provides a method for training the accumulation factors in the ANSI6.4.3 standard database by adopting a BP neural network to obtain a neural network model for calculating and predicting the accumulation factors under different energies and different mean free paths. The model can calculate the accumulation factor of the material by only inputting two parameters of energy (E) and mean free path (mfp), thereby not only reducing the complicated parameter input and calculation, but also greatly improving the calculation precision and being more beneficial to programming. The invention provides a method for calculating an accumulation factor of a shielding material based on a BP neural network, which provides the following technical scheme:
a method for calculating shielding material accumulation factor based on BP neural network carries out forward propagation, the forward propagation process is that input signals are transmitted to an output layer from an input layer through hidden layer neurons, output signals are generated at an output end, and when the output signals meet given output requirements, calculation is finished; when the output signal does not meet the given output requirement, the signal is shifted to the reverse propagation.
Preferably, the forward propagation of a neuron in the neural network is represented by:
wherein,representing the input of the ith neuron of the jth layer, and when j is the first layer, the input parameters are gamma ray energy E and mean free radicalStage mfp;represents the input and weight w of the ith neuron of the j layeriThe sum of the products of (a) and (b) plus an offsetj(ii) a (z) represents an activation function,representing the output of the j-th layer and simultaneously being used as the input of the j + 1-th layer, when the j-th layer is the output layerFor the final signal output value, when the network output value is not equal to the actual value, there is a loss value L, which is expressed by the following equation:
where y' is the expected value of the neural network output and y is the actual value.
Preferably, the neural network model suggests using relu activation functions, i.e., σ (z) ═ max { z,0}, and other activation functions may also implement the method, and sigmoid and tanh activation functions may also be used.
Preferably, the BP training algorithm employed optimizes the weights by attempting to minimize the sum of squared differences between the expected and actual values of the output neurons, the computation ending when L < ═ e, where e is the desired accuracy of the computation, or going on to a pre-set number of learning times, otherwise the back propagation computation is done.
Preferably, the BP algorithm calculates the weight w and the offset b so that L (w, b) is minimized.
Preferably, the principle of adjusting the weight value is to make the error decrease continuously, so that the adjustment should be performed along the negative gradient direction of the weight value, and the adjustment amount of the weight value is proportional to the decrease of the gradient of the error by the following formula:
wherein, alpha represents the learning rate and is a given constant, 0 < alpha < 1;representing the partial derivative of the loss function L to the ith input weight of the jth layer;representing the partial derivative of the loss function L to the offset b of the j layer;
iteratively updating the neuron connection weight and the offset used for the next round of network learning and training according to the obtained change increment of the neuron connection weight and the offset of each layer, and updating by the following formula:
bj=bj+Δbj
after solving new weight and offset of each layer, carrying out iterative calculation in the forward propagation process of steering;
the above is to complete a BP iteration loop, and when the calculated loss value L < ═ e is the set desired precision, and the smaller the setting, the higher the learning precision, the whole machine learning process is completed.
The invention has the following beneficial effects:
the invention divides the data into a training set and a testing set according to a reasonable proportion, and carries out normalization processing on the data. The network is trained using a training set and validated using a test set. The optimal hidden layer and number of neurons per layer can be determined by trial and error. In the neural network model provided by the algorithm, the activation functions of the hidden layer and the output layer are recommended to be relu functions, and the loss value L is recommended to be set to be less than 1e-4, so that the precision is fully improved. In addition, the learning rate is also crucial to the configuration of the BP neural network, and in the algorithm, the initial learning rate is set to 0.001, and the learning rate is gradually reduced in the training process.
Drawings
FIG. 1 is a schematic diagram of a neural network architecture;
FIG. 2 is a flow chart of a method for calculating a shielding material accumulation factor based on a BP neural network;
FIG. 3 is a diagram illustrating the variation of mean square error with training times;
FIG. 4 is a graph comparing data prediction and results.
Detailed Description
The present invention will be described in detail with reference to specific examples.
The first embodiment is as follows:
according to fig. 1 to 4, the invention provides a method for calculating a shielding material accumulation factor based on a BP neural network, which comprises the following steps:
the method provides that a neural network algorithm is used for realizing the rapid and accurate calculation of the accumulation factor of the material, an ANSI6.4.3 standard database is used as basic training data to train a neural network model, and the input parameters are x respectively1(gamma ray energy E), x2(mean free path mfp) penetration depth of up to 40mfp can be calculated. The computation process includes two parts, forward propagation computation and backward propagation computation, respectively.
In the forward propagation process, an input signal is transmitted from an input layer to an output layer through a hidden layer neuron, an output signal is generated at an output end, and if the output signal meets a given output requirement, the calculation is finished; if the output signal does not meet the given output requirement, the signal is shifted to the reverse propagation. The forward propagation of a neuron in a neural network is calculated as follows:
in the formula,representing the input of the ith neuron of the jth layer, if j is the first layer, the input parameters are gamma ray energy E and mean free path mfp;represents the input and weight w of the ith neuron of the j layeriThe sum of the products of (a) and (b) plus an offsetj(ii) a f (z) represents an activation function, commonly used activation functions are sigmoid, tanh, relu and the like, and the neural network model proposed by the algorithm suggests adopting a relu activation function, namely sigma (z) max { z,0 }.Represents the output of the j-th layer and also serves as the input of the j + 1-th layer. If the jth layer is an output layer, thenOutputting the value for the final signal. When the network output value is not equal to the actual value, there is a loss value L, and the loss function is defined as follows:
where y' is the expected value of the neural network output and y is the actual value. The BP training algorithm employed in the present algorithm optimizes the weights by attempting to minimize the sum of squared differences between the expected and actual values of the output neurons. When L < ═ e (e is the desired accuracy of calculation) or the preset learning times are reached, the calculation is ended, otherwise, the back propagation calculation is carried out.
The purpose of the BP algorithm is to compute the weights w and offsets b to minimize L (w, b). The principle of adjusting the weight is to make the error decrease continuously, so the adjustment should be performed along the negative gradient direction of the weight, that is, the adjustment amount of the weight is proportional to the gradient decrease of the error, that is:
wherein, alpha represents the learning rate and is a given constant, and is usually 0 < alpha < 1;representing the partial derivative of the loss function L to the ith input weight of the jth layer;representing the partial derivative of the penalty function L to the layer j offset b. Iteratively updating the neuron connection weight and the offset used for the next round of network learning and training according to the obtained change increment of the neuron connection weight and the offset of each layer, wherein the updating formula is as follows:
bj=bj+Δbj
and after new weight values and offset values of each layer are solved, carrying out iterative calculation in a forward propagation process. The above description is that a BP iteration loop is completed, and when the calculated loss value L < ═ e (e is the set desired precision, and the smaller the value is, the higher the learning precision is), the whole machine learning process is completed.
The data is divided into a training set and a test set according to the ratio of 8:2, and the data is normalized. The network is trained using a training set and validated using a test set. The optimal hidden layer and number of neurons per layer can be determined by trial and error. In the neural network model provided by the algorithm, the activation functions of the hidden layer and the output layer are recommended to be relu functions, and the loss value L is recommended to be set to be less than 1e-4, so that the precision is fully improved. In addition, the learning rate is also important for the configuration of the BP neural network, and in the algorithm, the initial learning rate is set to be 0.001, and the learning rate is gradually reduced in the training process.
Taking the calculation of the iron accumulation factor as an example:
after training, the data of the training set and the data of the test set both converge, and the mean square error calculated by the training set and the test set is less than 0.5. And selecting gamma-ray energy of 1MeV, and verifying the calculation result, wherein the verification result is shown in the following tables 1 and 2.
TABLE 1 iron accumulation factor (E. 1MeV)
E=1MeV | Raw data | Calculating data | Relative error |
mfp=0.5 | 1.53 | 1.5126 | -1.1373% |
mfp=1.0 | 2.14 | 2.1279 | -0.5654% |
mfp=2.0 | 3.50 | 3.4700 | -0.8571% |
mfp=3.0 | 5.04 | 5.0388 | -0.0238% |
mfp=4.0 | 6.79 | 6.7649 | -0.3697% |
mfp=5.0 | 8.74 | 8.7235 | -0.1888% |
mfp=6.0 | 10.90 | 10.8850 | -0.1376% |
mfp=7.0 | 13.20 | 13.1871 | -0.0977% |
mfp=8.0 | 15.70 | 15.6891 | -0.0694% |
mfp=10.0 | 21.10 | 21.0803 | -0.0934% |
mfp=15.0 | 37.10 | 37.0874 | -0.0340% |
mfp=20.0 | 56.20 | 56.1902 | -0.0174% |
mfp=25.0 | 77.90 | 77.8861 | -0.0178% |
mfp=30.0 | 102.00 | 102.0074 | 0.0073% |
mfp=35.0 | 128.00 | 127.9324 | -0.0528% |
mfp=40.0 | 156.00 | 156.0267 | 0.0171% |
Table 2 data prediction vs. results (Fe, E ═ 1MeV)
mfp | BP neural network | G-P formula | Taylor formula | MCNP |
0.5 | 1.51265 | 1.52806 | 1.41874 | 1.57773 |
1 | 2.12789 | 2.13000 | 1.85522 | 2.20260 |
1.5 | 2.77177 | 2.78954 | 2.31006 | 2.86442 |
2 | 3.46995 | 3.50194 | 2.78388 | 3.47787 |
2.5 | 4.13720 | 4.26462 | 3.27732 | 4.35071 |
3 | 5.03882 | 5.07588 | 3.79105 | 5.16554 |
3.5 | 5.83232 | 5.93447 | 4.32577 | 6.03475 |
4 | 6.76491 | 6.83941 | 4.88219 | 6.57329 |
4.5 | 7.45890 | 7.78988 | 5.46104 | 7.93790 |
5 | 8.72351 | 8.78514 | 6.06307 | 8.95766 |
5.5 | 9.61206 | 9.82453 | 6.68908 | 10.02863 |
6 | 10.88496 | 10.90739 | 7.33986 | 11.11221 |
6.33 | 11.73176 | 11.64560 | 7.78335 | 11.83507 |
6.66 | 12.46271 | 12.40231 | 8.23825 | 12.60586 |
7 | 13.18708 | 13.20112 | 8.71912 | 13.41480 |
7.33 | 13.94953 | 13.99485 | 9.19794 | 14.24595 |
7.66 | 14.63634 | 14.80655 | 9.68894 | 15.08462 |
8 | 15.68905 | 15.66144 | 10.20784 | 15.97168 |
8.6 | 17.60833 | 17.21557 | 11.15668 | 17.52485 |
9.3 | 19.37477 | 19.10084 | 12.31910 | 19.43345 |
10 | 21.08035 | 21.06210 | 13.54388 | 21.46756 |
11 | 24.23921 | 23.99214 | 15.40756 | 24.38852 |
12.5 | 28.73396 | 28.65931 | 18.47271 | 28.96302 |
14 | 33.90510 | 33.63653 | 21.89078 | 33.93299 |
15 | 37.08741 | 37.11776 | 24.38320 | 37.35614 |
17.5 | 47.63629 | 46.35269 | 31.44756 | 46.46070 |
20 | 56.19024 | 56.28650 | 41.73355 | 55.79451 |
21.5 | 61.68744 | 62.55533 | 45.66534 | 62.39607 |
23.5 | 70.65006 | 71.25219 | 54.38566 | 71.14275 |
25 | 77.88609 | 78.01985 | 61.75071 | 77.70180 |
26.5 | 86.72389 | 84.99326 | 69.90130 | 84.62409 |
28.5 | 96.05846 | 94.60897 | 82.12687 | 93.82347 |
30 | 102.00735 | 102.05905 | 92.42754 | 100.39644 |
31.5 | 109.74730 | 109.713047 | 103.80504 | 107.45932 |
33.5 | 119.66223 | 120.23089 | 120.83548 | 116.14194 |
35 | 127.93239 | 128.34454 | 135.15732 | 124.48145 |
36.5 | 136.33310 | 136.63656 | 150.95244 | 131.17084 |
38.5 | 147.36690 | 147.93264 | 174.55657 | 141.23308 |
40 | 156.02669 | 156.54549 | 194.37665 | 148.79000 |
The prediction result shows that the BP neural network can correctly predict the value of the accumulation factor, the overall prediction effect is superior to the result calculated by the Taylor formula, and in a higher mfp interval, the result obtained by the BP neural network prediction is closer to the result of an ANSI database, and is better than the result obtained by MCNP simulation. The accumulation factor result predicted by the BP neural network can be used for determining the theoretical mask calculation.
The above is only a preferred embodiment of the method for calculating the accumulation factor of the shielding material based on the BP neural network, and the protection scope of the method for calculating the accumulation factor of the shielding material based on the BP neural network is not limited to the above embodiments, and all technical solutions belonging to the idea belong to the protection scope of the present invention. It should be noted that modifications and variations which do not depart from the gist of the invention will be those skilled in the art to which the invention pertains and which are intended to be within the scope of the invention.
Claims (7)
1. A method for calculating an accumulation factor of a shielding material based on a BP neural network is characterized by comprising the following steps: carrying out forward propagation, wherein the forward propagation process is that an input signal is transmitted to an output layer from an input layer through a hidden layer neuron, an output signal is generated at an output end, and when the output signal meets a given output requirement, the calculation is finished; when the output signal does not meet the given output requirement, the signal is shifted to the reverse propagation.
2. The method of claim 1, wherein the method comprises the following steps: the forward propagation of a neuron in a neural network is represented by:
wherein,representing the input of the ith neuron of the jth layer, when j is the first layer, the input parameters are gamma ray energy E and mean free path mfp;represents the input and weight w of the ith neuron of the j layeriThe sum of the products of (a) and (b) plus an offsetj(ii) a (z) represents an activation function,representing the output of the j-th layer and simultaneously being used as the input of the j + 1-th layer, when the j-th layer is the output layerFor the final signal output value, when the network output value is not equal to the actual value, there is a loss value L, which is expressed by the following equation:
where y' is the expected value of the neural network output and y is the actual value.
3. The method of claim 2, wherein the method comprises the following steps: the neural network model uses the relu activation function, i.e., σ (z) ═ max { z,0 }.
4. The method of claim 2, wherein the method comprises the following steps: the neural network model may also employ sigmoid and tanh activation functions.
5. The method of claim 3, wherein the method comprises the following steps:
the BP training algorithm is adopted to optimize the weight by trying to minimize the sum of squared differences between the expected value and the actual value of the output neuron, and when L < ═ e, wherein e is the expected accuracy of calculation, or the learning times are preset, the calculation is finished, otherwise, the back propagation calculation is carried out.
6. The method of claim 5, wherein the method comprises the following steps: the BP algorithm calculates the weight w and the offset b to minimize L (w, b).
7. The method of claim 6, wherein the calculating the accumulation factor of the shielding material based on the BP neural network comprises: the principle of adjusting the weight is to reduce the error continuously, so the adjustment should be performed along the negative gradient direction of the weight, and the adjustment amount of the weight is proportional to the gradient decrease of the error by the following formula:
wherein, alpha represents the learning rate and is a given constant, 0 < alpha < 1;representing the partial derivative of the loss function L to the ith input weight of the jth layer;representing the partial derivative of the loss function L to the offset b of the j layer;
iteratively updating the neuron connection weight and the offset used for the next round of network learning and training according to the obtained change increment of the neuron connection weight and the offset of each layer, and updating by the following formula:
bj=bj+Δbj
after solving new weight and offset of each layer, carrying out iterative calculation in the forward propagation process of steering;
the above is to complete a BP iteration loop, and when the calculated loss value L < ═ e is the set desired precision, and the smaller the setting, the higher the learning precision, the whole machine learning process is completed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011602928.9A CN112733439A (en) | 2020-12-29 | 2020-12-29 | Method for calculating shielding material accumulation factor based on BP neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011602928.9A CN112733439A (en) | 2020-12-29 | 2020-12-29 | Method for calculating shielding material accumulation factor based on BP neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112733439A true CN112733439A (en) | 2021-04-30 |
Family
ID=75610559
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011602928.9A Pending CN112733439A (en) | 2020-12-29 | 2020-12-29 | Method for calculating shielding material accumulation factor based on BP neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112733439A (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170091615A1 (en) * | 2015-09-28 | 2017-03-30 | Siemens Aktiengesellschaft | System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies |
CN109978778A (en) * | 2019-03-06 | 2019-07-05 | 浙江工业大学 | Convolutional neural networks medicine CT image denoising method based on residual error study |
CN110222844A (en) * | 2019-05-30 | 2019-09-10 | 西安交通大学 | A kind of compressor performance prediction technique based on artificial neural network |
CN111489046A (en) * | 2019-01-29 | 2020-08-04 | 广东省公共卫生研究院 | Regional food safety evaluation model based on supply chain and BP neural network |
CN111666719A (en) * | 2020-06-08 | 2020-09-15 | 南华大学 | Gamma radiation multilayer shielding accumulation factor calculation method, device, equipment and medium |
-
2020
- 2020-12-29 CN CN202011602928.9A patent/CN112733439A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170091615A1 (en) * | 2015-09-28 | 2017-03-30 | Siemens Aktiengesellschaft | System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies |
CN111489046A (en) * | 2019-01-29 | 2020-08-04 | 广东省公共卫生研究院 | Regional food safety evaluation model based on supply chain and BP neural network |
CN109978778A (en) * | 2019-03-06 | 2019-07-05 | 浙江工业大学 | Convolutional neural networks medicine CT image denoising method based on residual error study |
CN110222844A (en) * | 2019-05-30 | 2019-09-10 | 西安交通大学 | A kind of compressor performance prediction technique based on artificial neural network |
CN111666719A (en) * | 2020-06-08 | 2020-09-15 | 南华大学 | Gamma radiation multilayer shielding accumulation factor calculation method, device, equipment and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Gonsalves et al. | QCD radiative corrections to electroweak-boson production at large transverse momentum in hadron collisions | |
US11648419B2 (en) | Method, apparatus, and system for simulating a particle transport and determining human dose in a radiotherapy | |
CN108549753B (en) | Radiation shielding calculation method for coupling point kernel integration method and Monte Carlo method | |
CN111126591A (en) | Magnetotelluric deep neural network inversion method based on space constraint technology | |
CN112733449B (en) | CNN well-seismic joint inversion method, CNN well-seismic joint inversion system, CNN well-seismic joint inversion storage medium, CNN well-seismic joint inversion equipment and CNN well-seismic joint inversion application | |
CN111292525A (en) | Traffic flow prediction method based on neural network | |
CN104599302B (en) | The method for obtaining PET crystal energies peak value and setting energy frequency discriminator | |
CN107871155B (en) | Spectral overlapping peak decomposition method based on particle swarm optimization | |
CN113919221B (en) | BP neural network-based fan load prediction and analysis method, device and storage medium | |
CN112086172A (en) | Three-dimensional dose calculation method, computer equipment and readable medium | |
CN113536623A (en) | Topological optimization design method for robustness of material uncertainty structure | |
CN114707712A (en) | Method for predicting requirement of generator set spare parts | |
CN113178242B (en) | Automatic plan optimization system based on coupled generation countermeasure network | |
Becker et al. | A Hybrid Monte Carlo–Deterministic Method for Global Particle Transport Calculations | |
CN114647814A (en) | Nuclear signal correction method based on prediction model | |
CN112733439A (en) | Method for calculating shielding material accumulation factor based on BP neural network | |
CN112086173B (en) | Three-dimensional dose calculation method, three-dimensional dose calculation device, computer equipment and readable medium | |
CN115062757A (en) | Prediction model for optimizing BP neural network by simulated annealing algorithm | |
CN110782181A (en) | Low-voltage transformer area line loss rate calculation method and readable storage medium | |
CN108073074B (en) | Assembly quality control method based on motion characteristics of machine tool feeding system | |
Chiang et al. | Post-selection inference in three-dimensional panel data | |
Sood et al. | A hybrid ANN-BFOA approach for optimization of FDM process parameters | |
CN110532646B (en) | Lake and reservoir cyanobacteria bloom prediction method based on self-adaptive dynamic programming | |
CN113536685B (en) | Neural network extrapolation-based wind speed probability model modeling method | |
Jiang et al. | Deep learning based dosimetry evaluation at organs-at-risk in esophageal radiation treatment planning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |