CN112950572B - Data processing and iterative correction method, system, medium and equipment - Google Patents
Data processing and iterative correction method, system, medium and equipment Download PDFInfo
- Publication number
- CN112950572B CN112950572B CN202110218654.1A CN202110218654A CN112950572B CN 112950572 B CN112950572 B CN 112950572B CN 202110218654 A CN202110218654 A CN 202110218654A CN 112950572 B CN112950572 B CN 112950572B
- Authority
- CN
- China
- Prior art keywords
- calculating
- line spectrum
- data
- input
- layer
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 81
- 238000012545 processing Methods 0.000 title claims abstract description 47
- 238000012937 correction Methods 0.000 title claims abstract description 41
- 238000001228 spectrum Methods 0.000 claims abstract description 132
- 238000009826 distribution Methods 0.000 claims abstract description 49
- 238000004364 calculation method Methods 0.000 claims abstract description 33
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 32
- 238000005530 etching Methods 0.000 claims abstract description 29
- 238000003860 storage Methods 0.000 claims abstract description 18
- 230000003287 optical effect Effects 0.000 claims abstract description 14
- 230000005855 radiation Effects 0.000 claims description 58
- 210000002569 neuron Anatomy 0.000 claims description 45
- 238000012549 training Methods 0.000 claims description 29
- 230000006870 function Effects 0.000 claims description 17
- 238000010521 absorption reaction Methods 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 13
- 230000000694 effects Effects 0.000 claims description 12
- 238000010801 machine learning Methods 0.000 claims description 12
- 230000008859 change Effects 0.000 claims description 11
- 210000004205 output neuron Anatomy 0.000 claims description 8
- 238000010183 spectrum analysis Methods 0.000 claims description 8
- 230000009466 transformation Effects 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 7
- 239000013598 vector Substances 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000003628 erosive effect Effects 0.000 claims 1
- 238000012821 model calculation Methods 0.000 claims 1
- 230000003595 spectral effect Effects 0.000 abstract description 28
- 238000005259 measurement Methods 0.000 abstract description 10
- 230000008569 process Effects 0.000 abstract description 10
- 238000013528 artificial neural network Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 13
- 239000002245 particle Substances 0.000 description 12
- 238000004590 computer program Methods 0.000 description 10
- 238000004458 analytical method Methods 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 230000005461 Bremsstrahlung Effects 0.000 description 2
- 239000003795 chemical substances by application Substances 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000002405 diagnostic procedure Methods 0.000 description 2
- 238000007429 general method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 150000002500 ions Chemical class 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 238000004381 surface treatment Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Drying Of Semiconductors (AREA)
Abstract
The invention discloses a data processing and iterative correction method, a system, a medium and equipment, which firstly processes complex background noise and continuous spectrum components in a self-etched line spectrum instead of being regarded as constants by an error back propagation algorithm of a multilayer feedforward neural network, reduces errors in the processing process, improves the accuracy of calculation, calculates line spectrum data processed by the error back propagation algorithm by the iterative correction method, considers the influence of optical thickness which is usually ignored, solves calculation deviation caused by plasma self-etching, and improves the accuracy of calculation. By the calculation program, the temperature distribution parameters at each spatial position in the observation direction can be deduced and calculated by the line spectrum data obtained by the experimental measurement of the spectrometer, and meanwhile, compared with other methods, the accuracy and precision are improved. In addition, by the storage medium and the device, a system for processing a large number of spectral lines can be constructed.
Description
Technical Field
The invention belongs to the technical field of plasma diagnosis spectroscopy, and particularly relates to a data processing and iterative correction method, a system, a medium and equipment.
Background
The plasma is a quasi-neutral gas composed of charged particles and neutral particles and showing concentrated behaviors, and is widely applied to the fields of gas discharge, controlled thermonuclear fusion, ETC emission, material surface treatment and the like. The plasma characteristic parameters mainly comprise temperature, density and components, and are an important basis for controlling and optimizing the capillary plasma discharge process in various application fields. For example, in ETC emission applications, the agent burn rate actually increases with increasing initial temperature, and the agent increases the light-off temperature by absorbing the heat of the plasma radiation, so the plasma jet temperature entering the combustion chamber typically needs to reach around 20,000K. Experimental diagnostic methods and data processing methods are therefore particularly important.
In general, there are a variety of diagnostic methods for temperature parameters, including spectroscopic measurements, thomson scattering, thermocouples, and the like. Thomson scattering is complex to build and debug, and thermocouples are difficult to develop, so that characteristic parameters of plasma can be judged through self-luminous spectrums of the plasma obtained through experimental measurement. Under the condition that the wavelength corresponding to each spectral line is known, the particle component composition of the plasma can be obtained through table lookup, and meanwhile, the corresponding particle density can be obtained through spectral line broadening calculation; on the other hand, the temperature parameter of the plasma can be calculated from the radiation intensity corresponding to the spectral line wavelength by, for example, boltzmann's oblique line method, bartels method, or the like.
The spectral line radiated outward from the emitter is absorbed by its own atoms, i.e. photons radiated by the process of transition from the upper to lower energy levels of the particle are re-absorbed by nearby lower energy level particles, and the phenomenon of weakening or even reversing the central intensity of the spectral line is called self-etching. The self-etching phenomenon has the characteristics of thicker optical thickness and incomplete conformity with the local thermodynamic equilibrium assumption, so that the boltzmann diagonal method requiring the optical thickness to be thin and the local thermodynamic equilibrium assumption is not suitable for self-etching spectral line analysis. On the other hand, the Bartels method, the Cowan-Dieke-karabournitis method and the like realize the calculation analysis of self-etching spectral lines from theoretical analysis and experimental measurement, but under certain plasma temperature ranges or energy level states, the methods have larger deviation in accuracy and precision, and a better processing method is not proposed for continuous spectrum components overlapped in a line spectrum. In addition, deriving the temperature distribution from the spatial distribution of the emission coefficient generally requires that the density be known, the density being a function of temperature, air pressure; however, in some practical measurements, the pressure at the plasma is difficult to directly measure only by estimation. Therefore, the functional relation of deriving the temperature distribution from the spatial distribution of the emission coefficient is not clear in practical applications.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a data processing and iterative correction method, a system, a medium and equipment aiming at the defects in the prior art, so as to solve the defects of inapplicability or lower accuracy and precision existing in the prior general methods such as the Boltzmann oblique line method, the Bartels method and the like, improve the accuracy by processing data through machine learning, improve the accuracy by the iterative correction method, and be suitable for high density of the self-etching effect>10 20 /m 3 ) The temperature of the plasma diagnoses the calculated range.
The invention adopts the following technical scheme:
a method of data processing and iterative correction comprising the steps of:
s1, acquiring line spectrum image data with spatial resolution at a plasma;
s2, processing the line spectrum image data obtained in the step S1 by adopting an error back propagation method in machine learning, wherein an input layer of the error back propagation method is the line spectrum image data stored in the step S1, and an output layer is the line spectrum data with background noise and continuous spectrum components removed;
s3, carrying out iterative correction on the line spectrum image data with the background noise removed and the continuous spectrum components removed in the step S2, and correcting the radiation intensity change caused by self-etching through a spectrum analysis principle.
Specifically, in step S1, the line spectrum image data is derived into a matrix form containing 1024×1024 elements, and the spatial positions of different observation points in the slit light-entering direction and the corresponding radiation intensities at different wavelengths are described.
Specifically, step S2 specifically includes:
s201, measuring, collecting and storing spectrogram image data;
s202, taking the data stored in the step S201 as a sample space D, and dividing the sample space D into a training set S and a test set T according to a reservation method; after the model with the background noise and continuous spectrum removed is trained on the training set S, the test error is estimated by using the T test set to be used as the estimation of the generalization error;
s203, processing the background noise and continuous spectrum components in the line spectrum by adopting an error back propagation algorithm.
Further, step S203 specifically includes:
s2031, setting a learning rate eta and a training set S;
s2032, randomly initializing all connection weights and thresholds in the range of (0, 1), and determining by compressing the nature of the activation function in the range of (0, 1);
s2033, calculating the input of the corresponding hidden layer neurons and the input of the corresponding output neurons according to the connection weights and the threshold values;
s2034, importing the calculation result in the step S2033, and obtaining the output of the neuron structure under the current connection weight and the threshold value;
s2035, judging whether the output under the current connection weight and the threshold meets the condition;
s2036, calculating the gradient of the output layer;
s2037, calculating gradient of the hidden layer;
and S2038, calculating the negative gradient direction connection weight and the value of which the threshold value is to be updated.
Further, the neuron basic structure of the error back propagation algorithm includes:
input layerFor d-dimensional real value vectors, i=1, 2..m, described by d attributes, m being the number of training samples;
the hidden layer contains q neurons, the input layer and the output layer are connected through the connection weight and the threshold value, and the threshold value of the h neuron of the hidden layer is gamma h The connection weight between the input layer and the ith neuron hidden layer is v ih ;
Output layerFor l-dimensional real value vectorsI=1, 2..m, described by l attributes, m is the number of training samples, and the threshold value of the j-th neuron of the output layer is θ j The connection weight between the hidden layer and the h f neuron is w hj ;
Setting the activation function of neurons of an implicit layer and an output layer as a SIGMOID function, namely SIGMOID (x) =1/(1+e) -x ) Squeezing the input value changing in the real number range into the (0, 1) output range;
the input of the h hidden layer neuron isThe input of the jth output neuron is +.>
Specifically, the step S3 specifically includes:
s301, program input, iteration number record initialization i=0,output data calculated by the learning model is obtained; />The radiation intensity at different wavelengths at different observation positions is represented and is processed line spectrum data;
s302, calculating and judging the difference value between the radiation intensity data under the current iteration and the previous iteration: if true, outputting the temperature distribution of the previous iteration; if yes, the step S303 is carried out; step S302 is skipped when i=0, and the process proceeds directly to step S303;
s303, calculating emission coefficients at different spatial positions in the observation direction according to the radiation intensities at a specific wavelength at different observation positions through Abelian inverse transformation;
s304, calculating the temperature distribution of the corresponding space position by the emission coefficient;
s305, calculating the distribution of the emission coefficients in the full wave range at the spatial positions in the overall observation directions through the inverse Abbe transform;
s306, calculating the absorption coefficient distribution in the full wave range at the spatial position in each observation direction according to the function relation of the emission coefficient, the absorption coefficient and the Planckian equation;
s307, calculating the optical thickness distribution in the full wave range at the spatial position in each observation direction based on the temperature distribution calculated in the step S304 and the absorption coefficient distribution calculated in the step S306 by Abbe transformation;
s308, calculating corrected radiation intensity line spectrum data considering the self-etching effect according to the Lambert-Beer law;
s309, recording the next iteration times, finally entering step S302 to judge whether the current radiation intensity meets the condition, repeating the steps until the judging condition is true, and outputting the temperature distribution in the plasma observation direction.
Further, in step S304, a model relationship between the training plasma temperature and the radiation intensity at each position in space is calculated using an error back propagation algorithm.
The other technical scheme of the invention is that the data processing and iterative correction system for calculating the plasma temperature by self-etching spectral lines comprises:
the acquisition module acquires line spectrum image data with spatial resolution at the plasma;
the processing module is used for processing the line spectrum image data obtained by the acquisition module by adopting an error back propagation method in machine learning, wherein an input layer of the error back propagation method is the line spectrum image data stored by the acquisition module, and an output layer is the line spectrum data from which background noise and continuous spectrum components are removed;
and the correction module is used for carrying out iterative correction on the line spectrum image data from which the background noise and continuous spectrum components are removed by the processing module, and correcting the radiation intensity change caused by self-etching through a spectrum analysis principle.
Another aspect of the invention is a computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computing device, cause the computing device to perform any of the methods.
Another aspect of the present invention is a computing device, including:
one or more processors, memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing any of the methods.
Compared with the prior art, the invention has at least the following beneficial effects:
the invention relates to a data processing and iterative correction method, which is used for carrying out interference-free optical diagnosis on plasma by collecting line spectrum image data with spatial resolution at the plasma. In addition, complex background noise and continuous spectrum components can be removed through an error back propagation method, instead of simply removing the continuous spectrum components regarded as constants by a general method, and the accuracy of spectrum line calculation is improved. On the other hand, the radiation intensity change caused by self-etching can be corrected by an iterative correction method instead of directly adopting original radiation intensity to calculate, so that the accuracy of calculation is improved.
Further, step S1 is to record and acquire line spectrum image data with spatial resolution at the plasma.
Further, in the step S2, complex background noise and continuous spectrum components can be calculated by using an error back propagation method, so that the line spectrum radiation intensity distribution is better restored, the accuracy of a subsequent program algorithm for calculating the temperature based on the line spectrum radiation intensity distribution can be improved, and calculation errors caused by the background noise and the continuous spectrum components are avoided.
Further, in step S203, from the sample space divided into the training set S and the test set T, a model relationship between the input "original radiation intensity distribution data" and the output "radiation intensity distribution data with noise removed and continuous spectrum" may be established through a training algorithm.
Furthermore, the neuron basic structure setting of the error back propagation algorithm can build a specific model relation between input and output through training by training refreshing weights and thresholds, namely by a data analysis mode.
Further, step S3 considers the influence of the self-etching effect caused by the thicker plasma thickness on the spectral line intensity, and calculates the current radiation intensity by counting the optical thickness of each time, and calculates the radiation intensity difference value after the correction of the front and back two times in an iterative manner within a reasonable setting range.
Further, step S304 also uses an error back propagation algorithm to calculate the relationship between the temperature of the training plasma and the radiation intensity at each spatial position, so as to avoid the estimation assumption condition of other methods on the particle density, and the model relationship between the temperature and the radiation intensity is reliably given by data analysis.
In summary, the invention adopts the error back propagation method to calculate the background noise and continuous spectrum components, and adopts the iterative correction method to calculate the influence of the self-etching effect, thereby improving the accuracy and precision of calculation.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
Drawings
FIG. 1 is a block diagram of the overall computing architecture of the present invention;
FIG. 2 is a block diagram of a machine learning/BP algorithm of the present invention;
FIG. 3 is a block diagram of the basic structure of neurons of the BP algorithm of the present invention;
FIG. 4 is a block diagram of a specific calculation flow of the learning model of the present invention;
FIG. 5 is a flow chart of an iterative correction calculation method of the present invention;
FIG. 6 is a graph of the spectrum after experimental measurement and algorithm processing in an application example of the present invention;
fig. 7 is a graph of a temperature profile calculated from an algorithmically processed spectral line graph.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be understood that the terms "comprises" and "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It is also to be understood that the terminology used in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.
Various structural schematic diagrams according to the disclosed embodiments of the present invention are shown in the accompanying drawings. The figures are not drawn to scale, wherein certain details are exaggerated for clarity of presentation and may have been omitted. The shapes of the various regions, layers and their relative sizes, positional relationships shown in the drawings are merely exemplary, may in practice deviate due to manufacturing tolerances or technical limitations, and one skilled in the art may additionally design regions/layers having different shapes, sizes, relative positions as actually required.
The invention provides a data processing and iterative correction method, a system, a medium and equipment, wherein firstly, a complex background noise and continuous spectrum components in a self-etched line spectrum are processed by an error counter propagation algorithm of a multilayer feedforward neural network instead of being regarded as constants, so that errors in the corresponding processing process are reduced, and the accuracy of calculation is improved; on the other hand, line spectrum data processed by an error back propagation algorithm is calculated through an iterative correction method, the influence of the optical thickness which is usually ignored is considered, calculation deviation caused by plasma self-etching is solved, and the calculation accuracy is improved. By the calculation program, the temperature distribution parameters at each spatial position in the observation direction can be deduced and calculated by the line spectrum data obtained by the experimental measurement of the spectrometer, and meanwhile, compared with other methods, the accuracy and precision are improved.
Referring to fig. 1, in the data processing and iterative correction method of the present invention, a data set of self-etched spectral lines is processed by adopting an error back propagation (error BackPropagation, abbreviated as BP) algorithm of a multi-layer feedforward neural network, so as to solve the problems of noise floor and continuous spectrum components contained in the line spectrum, and improve the accuracy of calculation;
and (3) calculating line spectrum data processed by the BP algorithm by an iterative correction method, solving calculation deviation caused by plasma self-etching and improving calculation accuracy. The method comprises the following specific steps:
s1, experimental measurement and data acquisition;
the experiment records the line spectrum image with spatial resolution at the plasma through an optical diagnostic instrument spectrometer and other matched equipment. The line spectrum image can be derived into a matrix form containing 1024 x 1024 elements, and describes the spatial positions of different observation points in the slit light entering direction and the corresponding radiation intensities at different wavelengths. Generally, the line spectrum data contains background noise and continuous spectrum components, which have great influence on subsequent calculation and need to be separated and removed.
Through the line spectrum data, the particle type information of the plasma component can be obtained through the line spectrum wavelength; on the other hand, in order to satisfy the plasma with thin optical thickness and local thermodynamic equilibrium, parameters such as plasma temperature, density and the like can be directly calculated from the radiation intensity and broadening information of spectral lines. However, in practical application, the conditions of thick optical thickness and incomplete local thermodynamic equilibrium are often generated, and plasma density parameters are difficult to directly measure, so that great difficulty is brought to the diagnosis and calculation of plasma temperature.
S2, processing the stored data obtained in the step S1 by adopting an error back propagation (error BackPropagation, BP for short) algorithm in machine learning, and separating and removing background noise and continuous spectrum components;
generally, the background noise is independent of wavelength and position, and the change of the background noise is negligible in weak influence, so that the background noise can be regarded as a constant, and the background noise can be recognized by the intensity of a spatial edge position spectral line (the plasma density is small and can be regarded as background noise) through machine learning training.
On the other hand, the continuum component is complex in composition and is generally caused by interactions between charged particles, and mainly includes bremsstrahlung caused by changes in the velocity of movement of charged particles and complex radiation caused by electrons being bound by ions. Because of the complex composition, continuous spectral components in the line spectrum are mostly regarded as constant separation and removal in practical application, however, such processing may cause large errors in the line spectrum and introduce errors into subsequent temperature calculation.
S2, removing continuous spectrum components by adopting a BP algorithm, wherein an input layer is line spectrum data stored in the step S1, and an output layer is line spectrum data from which background noise and continuous spectrum components are removed. The judging condition is introduced by spectral line broadening and line spectrum overall contour contrast, namely, the method is considered to be as follows: other regions are considered to be caused by continuum components and background noise, except for the apparent radiation intensity in the Lorentz spread range of the spectral line corresponding to a particular particle. And continuous spectrum components in a specific spectral line range are trained by a large number of samples through a BP algorithm, and the input and output relation is fitted from the data angle.
Assume the condition:
1. because the bremsstrahlung is continuous and smooth, and the composite radiation energy level transition is determined by particles, the upper limit value exists between the continuous spectrum component in the spectral line range and the continuous spectrum intensity difference value at the adjacent wavelength;
2. after removing the background noise and the continuous spectrum components, the Lorentzian line type needs to be satisfied in the spectrum line range.
Referring to fig. 2, fig. 2 shows a block diagram of the machine learning/BP algorithm shown in fig. 1.
S201, acquiring spectral line data, which comprises experimental measurement by a spectrometer and data storage, wherein the spectral line data are derived by Andor software into a matrix form containing 1024 x 1024 elements and comprise three pieces of information including space positions of observation points, radiation intensity and corresponding wavelengths; measuring, collecting and storing line spectrum data, namely original line spectrum data;
s202, taking the stored data in the step S201 as a sample space D, and dividing the sample space into a training set S and a test set T according to a reservation method; after the model with the background noise and continuous spectrum removed is trained on the S, the test error of the model is evaluated by using T and is used as the estimation of the generalization error; the sample space D is divided by adopting a reserving method, 3/4 samples are used for training, and the rest samples are used for testing;
s203, processing the background noise and continuous spectrum components in the line spectrum by adopting a BP algorithm, wherein the BP algorithm can be used for a multilayer feedforward neural network.
Referring to fig. 3, a basic structural block diagram of a neuron adopting a BP algorithm is shown as follows:
input layerD is d dimension real value vector, described by d attributes, m is training sample number;
the hidden layer contains q neurons, the input layer and the output layer are connected through the connection weight and the threshold value, and the threshold value of the h neuron of the hidden layer is gamma h The connection weight between the input layer and the ith neuron hidden layer is v ih ;
Output layerFor the l-dimensional real value vector, the value is described by l attributes, m is the number of training samples, and the threshold value of the j-th neuron of the output layer is theta j The connection weight between the hidden layer and the h neuron is w hj 。
In addition, the activation function of neurons of an implicit layer and an output layer is set as a SIGMOID function, namely SIGMOID (x) =1/(1+e) -x ) The input values varying in the real range are squeezed into the (0, 1) output range.
In addition, there are: the input of the h hidden layer neuron isThe input of the j-th output neuron is
Referring to fig. 4, the specific calculation process in step 203 is as follows:
S2031、setting a learning rate eta and a training set S, wherein +.>For inputting value +.>M is the number of training samples and is the output value;
learning rate eta epsilon (0, 1) is an adjustable value, influences the magnitude of gradient change, and a proper value 0.5 is selected here; training setDivided by the aforementioned sample space,/->Here, n x 1024 data are input, n (1-1024) is optionally determined by the spatial position to be calculated, and 1024 represents 1024 points divided in the full wavelength range. That is, the input is the radiation intensity over the full wavelength range at n spatial locations, including line spectrum, continuous spectrum, background noise. While the output is again n x 1024 data, but only the line spectrum intensity.
S2032, randomly initializing all connection weights and thresholds in the range of (0, 1), and determining by compressing the nature of the activation function in the range of (0, 1);
in a subsequent step, the adjustment parameters are updated by the training set.
S2033、Calculating the input of corresponding hidden layer neurons and the input of corresponding output neurons from the connection weights and the threshold values, wherein alpha h Input to the h hidden layer neuron, v ih The h neuron of the hidden layer and the i god of the input layerConnection right, x between meta hidden layers i Input for the ith neuron of the input layer, b h Output of the h neuron of hidden layer, gamma h The threshold value of the h neuron of the hidden layer, beta is the input of the j output neuron, and w hj For the connection between the jth neuron of the output layer and the h neuron of the hidden layer,/->For outputting the j-th neuron of the output layer, θ j A j-th neuron threshold for the output layer;
s2034, importing the calculation result in the step S2033, and obtaining the output of the neuron structure under the current connection weight and the threshold value;
s2035, judging whether the output under the current connection weight and the threshold meets the condition;
the judgment conditions here are:
1. selecting whether the Lorentzian line type is satisfied in the spectral line range;
2. selecting whether the difference value between the continuous spectrum component in the spectral line range and the continuous spectrum component with the adjacent wavelength is in the maximum energy level difference value range; if true, outputting the current connection weight and the threshold value, and determining a learning model, namely the relation between input and output; if false, the flow advances to step S2036.
S2036、Calculating the gradient of the output layer g j Gradient for the j-th neuron of the output layer, < >>Is a standard value;
S2037、calculating gradient of hidden layer e h A gradient of the hidden h-th neuron;
S2038、Δw hj =ηg j b h /Δθ j =-ηg j 、Δv ih =ηe h x i /Δγ h =-ηe h calculating the value of the negative gradient direction connection weight and the value of the threshold value to be updated, and deltaw hj 、Δv ih And updating the values for the corresponding connection weights respectively.
The BP algorithm is based on a gradient descent strategy, so that finally, the step S2038 calculates the value of the negative gradient direction connection weight and the value of the threshold to be updated; and outputting the updated connection weight and threshold value, determining a learning model, namely the relation between input and output, and repeating the steps until the step S2035 is met and the training program is jumped out.
S3, carrying out iterative correction, and correcting radiation intensity change caused by self-etching through a spectrum analysis principle.
Referring to fig. 5, the iterative correction calculation process specifically includes:
s301, program input, iteration number record initialization i=0, in additionOutput data (non-sample space data, other experimental measurement data) calculated by the learning model; />The radiation intensities at different wavelengths at different observation positions are shown as processed line spectrum data.
S302、i represents the iteration times, θ is a set value, and the difference between the radiation intensity data under the current iteration and the previous iteration is calculated and judged: if true, outputting the temperature distribution of the previous iteration; if yes, the step S303 is carried out; step S302 is skipped when i=0, and the process proceeds directly to step S303;
S303、by inverse Abbe transformation, the radiation intensity L at a specific wavelength from different observation positions (i) (y) calculating the emissivity epsilon at different spatial positions in the viewing direction (i) (r)。
S304, calculating the temperature distribution of the corresponding space position by the emission coefficient through a program flow similar to the step S203;
the emission coefficient is related to the temperature distribution function:
at density n g The temperature T can be calculated from the emission coefficient epsilon, as is known.
However, in some practical applications, the pressure at the plasma location cannot be measured directly, but can be carried into the equation by way of estimation, thereby introducing errors into the final temperature result.
Thus, step S304 is presented herein: the corresponding model relation is directly trained by the data through BP algorithm, and the calculation flow is similar to the steps of processing the background noise and continuous spectrum components. The difference is that here the input is the distribution of the emission coefficients at different spatial positions in the direction of view and the output is the temperature distribution at different spatial positions.
The judging conditions are as follows:
1. the same sample, namely the same spectrum, and the temperature calculation errors of different spectrum lines are in the allowable range;
2. for each sample measured for multiple times under the same experimental condition, the temperature calculation errors of different line spectrums are within the allowable range.
Step S304 is used for training a relation model of the emission coefficient and the temperature distribution, and the relation model is used for calculating the temperature distribution under other experiments by the emission coefficient;
S305、radiation intensity from different observation positions in the full wave range by inverse Abelian transformation>Calculating a full wave range at spatial positions in the general directions of viewDistribution of emission coefficients within a circleThe calculation formulas of the inverse Abbe's transformation and the Abbe's transformation are respectively as follows:
the Abbe's (inverse) transform expresses the relationship between the radiation intensity accumulated in the observation direction and the radiation intensity at each spatial position in the observation direction, without taking the self-etching effect into account;
S306、by the emission coefficient->Absorption coefficient kappa (i) (r, lambda) Planck equation B λ (T (i) (r), λ) calculating an absorption coefficient distribution in the full-wave range at the spatial position in each viewing direction;
S307、the optical thickness distribution τ in the full wave range at the spatial position in each observation direction is calculated based on the temperature distribution calculated in step S304 and the absorption coefficient distribution calculated in step S306 by the abbe transform (i) (y,λ)。
S308、Corrected radiation intensity line spectrum data taking into account the self-etching effect is calculated according to Lambert-Beer law.
The formula of Lambert-Beer law isThe expression is: the self-etching effect, i.e. the influence of the iso-optical thickness or absorption effect, is taken into account, and the line intensity at the central position of the plasma in the viewing direction is based on the corrective action made by the self-etching effect.
And S309, i=i+1, recording the next iteration times, finally, proceeding to step S302, judging whether the current radiation intensity meets the condition or not, repeating the steps until the judging condition is true, and outputting the temperature distribution in the plasma observation direction.
Through the calculation flow and the steps, the temperature distribution parameters in the observing direction of the plasma can be deduced from the line spectrum data measured by the experimental spectrometer. The calculation program mainly relates to BP algorithm and iterative correction method.
In still another embodiment of the present invention, a data processing and iterative correction system is provided, which can be used to implement the above data processing and iterative correction method, and in particular, the data processing and iterative correction system includes an acquisition module, a processing module, and a correction module.
The acquisition module acquires line spectrum image data with spatial resolution at the plasma;
the processing module is used for processing the line spectrum image data obtained by the acquisition module by adopting an error back propagation method in machine learning, wherein an input layer of the error back propagation method is the line spectrum image data stored by the acquisition module, and an output layer is the line spectrum data from which background noise and continuous spectrum components are removed;
and the correction module is used for carrying out iterative correction on the line spectrum image data from which the background noise and continuous spectrum components are removed by the processing module, and correcting the radiation intensity change caused by self-etching through a spectrum analysis principle.
In yet another embodiment of the present invention, a terminal device is provided, the terminal device including a processor and a memory, the memory for storing a computer program, the computer program including program instructions, the processor for executing the program instructions stored by the computer storage medium. The processor may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf Programmable gate arrays (FPGAs) or other Programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., which are the computational core and control core of the terminal adapted to implement one or more instructions, in particular adapted to load and execute one or more instructions to implement a corresponding method flow or a corresponding function; the processor of the embodiment of the invention can be used for data processing and iterative correction methods, systems, media and equipment operation, and comprises the following steps:
acquiring line spectrum image data with spatial resolution at the plasma; the line spectrum image data is processed by adopting an error back propagation method in machine learning, wherein an input layer of the error back propagation method is the line spectrum image data, and an output layer is the line spectrum data with background noise and continuous spectrum components removed; and (3) carrying out iterative correction on the line spectrum image data from which the background noise and continuous spectrum components are removed, and correcting the radiation intensity change caused by self-etching through a spectrum analysis principle.
In a further embodiment of the present invention, the present invention also provides a storage medium, in particular, a computer readable storage medium (Memory), which is a Memory device in a terminal device, for storing programs and data. It will be appreciated that the computer readable storage medium herein may include both a built-in storage medium in the terminal device and an extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space storing an operating system of the terminal. Also stored in the memory space are one or more instructions, which may be one or more computer programs (including program code), adapted to be loaded and executed by the processor. The computer readable storage medium herein may be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory.
One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the corresponding steps of the methods, systems, media, and apparatus for data processing and iterative correction in the above embodiments; one or more instructions in a computer-readable storage medium are loaded by a processor and perform the steps of:
acquiring line spectrum image data with spatial resolution at the plasma; the line spectrum image data is processed by adopting an error back propagation method in machine learning, wherein an input layer of the error back propagation method is the line spectrum image data, and an output layer is the line spectrum data with background noise and continuous spectrum components removed; and (3) carrying out iterative correction on the line spectrum image data from which the background noise and continuous spectrum components are removed, and correcting the radiation intensity change caused by self-etching through a spectrum analysis principle.
Referring to fig. 6, the solid line shows the radiation intensity distribution curve of the plasma at a certain spatial position with a wavelength range of 500-553 nm, which is measured by experiment, and the dotted line shows the radiation intensity distribution curve after removing the background noise and continuous spectrum components by the error back propagation algorithm and calculating the spectral line self-etching effect by the iterative correction method. Through algorithm calculation, the spectral line intensity is closer to a theoretical radiation intensity value, so that the accuracy and precision of calculation are improved. Referring to fig. 7, a plasma temperature distribution curve calculated according to spectral line radiation intensity data processed by the algorithm is shown, which accords with the actual result.
In summary, the accuracy and precision of the method for calculating the plasma temperature by the self-etching spectral line can be effectively improved by the error back propagation data processing and iteration correction method, and a system for processing a large number of spectral lines can be formed by the storage medium and the equipment.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above is only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited by this, and any modification made on the basis of the technical scheme according to the technical idea of the present invention falls within the protection scope of the claims of the present invention.
Claims (7)
1. A method of data processing and iterative correction comprising the steps of:
s1, acquiring line spectrum image data with spatial resolution at a plasma;
s2, processing the line spectrum image data obtained in the step S1 by adopting an error back propagation method in machine learning, wherein an input layer of the error back propagation method is the line spectrum image data stored in the step S1, and an output layer is the line spectrum data with background noise and continuous spectrum components removed, and specifically comprises the following steps:
measuring, collecting and storing line spectrum image data;
taking the stored data as a sample space D, and dividing the sample space D into a training set S and a test set T according to a reservation method; after the model with the background noise and continuous spectrum removed is trained on the training set S, the test error is estimated by using the T test set to be used as the estimation of the generalization error;
the background noise and continuous spectrum components in the line spectrum are processed by adopting an error back propagation algorithm, and the method specifically comprises the following steps:
setting a learning rateA training set S; randomly initializing all connection weights and thresholds in the range of (0, 1), and determining the property of the activation function compressed in the range of (0, 1); calculating the input of the corresponding hidden layer neurons and the input of the corresponding output neurons according to the connection weights and the threshold values; the calculation result is carried in, and the output of the neuron structure under the current connection weight and the threshold value can be obtained; judging whether the output under the current connection weight and the threshold meets the condition or not; calculating the gradient of the output layer; calculating gradient of the hidden layer; calculating the value of the negative gradient direction connection weight and the value of which the threshold value is to be updated;
s3, carrying out iterative correction on line spectrum image data with background noise removed and continuous spectrum components removed in the step S2, and correcting radiation intensity change caused by self-etching through a spectrum analysis principle, wherein the method specifically comprises the following steps:
program input, iteration number record initializationi=0,Is learned by the aboveOutput data after model calculation;the radiation intensity at different wavelengths at different observation positions is represented and is processed line spectrum data; calculating and judging the difference value between the radiation intensity data under the current iteration and the previous iteration: if true, outputting the temperature distribution of the previous iteration; if the judgment is false, the next step is carried out; at the position ofiWhen the value is=0, skipping the step, and directly entering the next step; calculating the emission coefficients at different spatial positions in the observation direction from the radiation intensities at a specific wavelength at the different observation positions by means of an inverse Abbe transform; calculating the temperature distribution of the corresponding space position by the emission coefficient; calculating a distribution of emission coefficients in a full wave range at spatial positions in the total of the respective observation directions by inverse Abbe transform; calculating the absorption coefficient distribution in the full wave range at the spatial position in each observation direction by the functional relation of the emission coefficient, the absorption coefficient and the Planckian equation; calculating an optical thickness distribution in a full wave range at a spatial position in each observation direction based on the calculated temperature distribution and the calculated absorption coefficient distribution by Abbe transformation; according to Lambert-Beer law, calculating corrected radiation intensity line spectrum data considering self-etching effect; and recording the next iteration times, finally judging whether the current radiation intensity meets the condition, repeating the steps until the judging condition is true, and outputting the temperature distribution in the plasma observation direction.
2. The method according to claim 1, wherein in step S1, the line spectrum image data is derived in a matrix form comprising 1024 x 1024 elements, describing the spatial positions of different observation points in the slit light entering direction and the corresponding radiation intensities at different wavelengths.
3. The method according to claim 1, wherein in step S2, the neuron basic structure of the error back-propagation algorithm comprises:
input layerIs thatdDimension vector,/->By the following constitutiondA description of the attributes of the object,mthe number of training samples;
the hidden layer comprisesqThe neuron is connected with the input layer and the output layer through the connection weight and the threshold value, and the hidden layer is the first layerhThe neuron threshold isAnd input layer (th)iThe connection right between the hidden layers of the individual neurons is +.>;
Output layerIs thatlDimension vector,/->By the following constitutionlA description of the attributes of the object,mto train the number of samples, output layerjThe threshold of individual neurons is +.>And hidden layer (a)hPersonal (S)fThe connection right between neurons is +.>;
Setting the activation functions of neurons of an implicit layer and an output layer as Sigmoid functions, namelySqueezing the input value changing in the real number range into the (0, 1) output range;
first, thehThe input of the hidden layer neurons isFirst, thejThe input of the individual output neurons is +.>。
4. The method according to claim 1, wherein in step S3, a model relationship between the training plasma temperature and the radiation intensity at each position in space is calculated using an error back propagation algorithm.
5. A data processing and iterative correction system for calculating plasma temperature from an erosion spectrum, comprising:
the acquisition module acquires line spectrum image data with spatial resolution at the plasma;
the processing module is used for processing the line spectrum image data obtained by the acquisition module by adopting an error back propagation method in machine learning, wherein an input layer of the error back propagation method is the line spectrum image data stored by the acquisition module, and an output layer is the line spectrum data with background noise and continuous spectrum components removed, and specifically comprises the following steps:
measuring, collecting and storing line spectrum image data;
taking the stored data as a sample space D, and dividing the sample space D into a training set S and a test set T according to a reservation method; after the model with the background noise and continuous spectrum removed is trained on the training set S, the test error is estimated by using the T test set to be used as the estimation of the generalization error;
the background noise and continuous spectrum components in the line spectrum are processed by adopting an error back propagation algorithm, and the method specifically comprises the following steps:
setting a learning rateA training set S; randomly initializing all connection weights and thresholds in the range of (0, 1), and determining the property of the activation function compressed in the range of (0, 1); calculating the input of the corresponding hidden layer neurons and the input of the corresponding output neurons according to the connection weights and the threshold values; bringing in the calculation result to obtain the neuron under the current connection weight and the threshold valueOutputting a structure; judging whether the output under the current connection weight and the threshold meets the condition or not; calculating the gradient of the output layer; calculating gradient of the hidden layer; calculating the value of the negative gradient direction connection weight and the value of which the threshold value is to be updated;
the correction module is used for carrying out iterative correction on the line spectrum image data of which the processing module removes the background noise and continuous spectrum components, and correcting the radiation intensity change caused by self-etching through a spectrum analysis principle, and specifically comprises the following steps:
program input, iteration number record initializationi=0,Output data calculated by the learning model is obtained;the radiation intensity at different wavelengths at different observation positions is represented and is processed line spectrum data; calculating and judging the difference value between the radiation intensity data under the current iteration and the previous iteration: if true, outputting the temperature distribution of the previous iteration; if the judgment is false, the next step is carried out; at the position ofiWhen the value is=0, skipping the step, and directly entering the next step; calculating the emission coefficients at different spatial positions in the observation direction from the radiation intensities at a specific wavelength at the different observation positions by means of an inverse Abbe transform; calculating the temperature distribution of the corresponding space position by the emission coefficient; calculating a distribution of emission coefficients in a full wave range at spatial positions in the total of the respective observation directions by inverse Abbe transform; calculating the absorption coefficient distribution in the full wave range at the spatial position in each observation direction by the functional relation of the emission coefficient, the absorption coefficient and the Planckian equation; calculating an optical thickness distribution in a full wave range at a spatial position in each observation direction based on the calculated temperature distribution and the calculated absorption coefficient distribution by Abbe transformation; according to Lambert-Beer law, calculating corrected radiation intensity line spectrum data considering self-etching effect; and recording the next iteration times, finally judging whether the current radiation intensity meets the condition, repeating the steps until the judging condition is true, and outputting the temperature distribution in the plasma observation direction.
6. A computer readable storage medium storing one or more programs, wherein the one or more programs comprise instructions, which when executed by a computing device, cause the computing device to perform any of the methods of claims 1-4.
7. A computing device, comprising:
one or more processors, memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing any of the methods of claims 1-4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110218654.1A CN112950572B (en) | 2021-02-26 | 2021-02-26 | Data processing and iterative correction method, system, medium and equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110218654.1A CN112950572B (en) | 2021-02-26 | 2021-02-26 | Data processing and iterative correction method, system, medium and equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112950572A CN112950572A (en) | 2021-06-11 |
CN112950572B true CN112950572B (en) | 2024-04-05 |
Family
ID=76246527
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110218654.1A Active CN112950572B (en) | 2021-02-26 | 2021-02-26 | Data processing and iterative correction method, system, medium and equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112950572B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115113032B (en) * | 2022-06-29 | 2024-09-24 | 云南电网有限责任公司电力科学研究院 | Method, system and device for measuring plasma distribution parameters |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101814158A (en) * | 2009-02-20 | 2010-08-25 | 北京联合大学生物化学工程学院 | Method for analyzing and processing experimental data based on artificial neural network |
CN110958856A (en) * | 2017-07-21 | 2020-04-03 | 皇家飞利浦有限公司 | Creation of electron density datasets from spectral CT datasets |
KR102127597B1 (en) * | 2019-03-22 | 2020-06-29 | 스페클립스 주식회사 | Diagnosis method using laser induced breakdown spectroscopy and diagnosis device performing the same |
CN111415051A (en) * | 2020-05-18 | 2020-07-14 | 江苏电力信息技术有限公司 | Power material demand plan prediction method based on BP neural network |
-
2021
- 2021-02-26 CN CN202110218654.1A patent/CN112950572B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101814158A (en) * | 2009-02-20 | 2010-08-25 | 北京联合大学生物化学工程学院 | Method for analyzing and processing experimental data based on artificial neural network |
CN110958856A (en) * | 2017-07-21 | 2020-04-03 | 皇家飞利浦有限公司 | Creation of electron density datasets from spectral CT datasets |
KR102127597B1 (en) * | 2019-03-22 | 2020-06-29 | 스페클립스 주식회사 | Diagnosis method using laser induced breakdown spectroscopy and diagnosis device performing the same |
CN111415051A (en) * | 2020-05-18 | 2020-07-14 | 江苏电力信息技术有限公司 | Power material demand plan prediction method based on BP neural network |
Non-Patent Citations (1)
Title |
---|
卞国龙 ; 黄海松 ; 王安忆 ; 于凯华 ; .基于PSO-BP算法的无线传感器网络定位优化.电讯技术.2017,(第02期),全文. * |
Also Published As
Publication number | Publication date |
---|---|
CN112950572A (en) | 2021-06-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yan et al. | A novel hybrid feature selection strategy in quantitative analysis of laser-induced breakdown spectroscopy | |
US10346740B2 (en) | Systems and methods incorporating a neural network and a forward physical model for semiconductor applications | |
Okut | Bayesian regularized neural networks for small n big p data | |
TWI717737B (en) | Method and apparatus for evaluating an unknown effect of defects of an element of a photolithography process | |
Zhu et al. | Permittivity reconstruction in electrical capacitance tomography based on visual representation of deep neural network | |
JP4641619B2 (en) | Method and apparatus for predicting surface profile by plasma process | |
Gan et al. | Estimation of remaining fatigue life under two-step loading based on kernel-extreme learning machine | |
Yeratapally et al. | A digital twin feasibility study (Part I): Non-deterministic predictions of fatigue life in aluminum alloy 7075-T651 using a microstructure-based multi-scale model | |
Long et al. | Deep learning-based planar crack damage evaluation using convolutional neural networks | |
CN104483292B (en) | A kind of method that use multiline ratio method improves laser microprobe analysis accuracy | |
Bučar et al. | A neural network approach to describing the scatter of S–N curves | |
CN112816895B (en) | Analysis method, system, equipment and computer storage medium for electrochemical impedance spectrum | |
CN112950572B (en) | Data processing and iterative correction method, system, medium and equipment | |
Valente et al. | Classifying powder flowability for cold spray additive manufacturing using machine learning | |
JP2006250938A (en) | Device and method for correcting aberration | |
Peyvan et al. | RiemannONets: Interpretable neural operators for Riemann problems | |
Somehsaraei et al. | A novel approach based on artificial neural network for calibration of multi-hole pressure probes | |
Wang et al. | Probabilistic method for fatigue crack growth prediction with hybrid prior | |
Tertian | Mathematical matrix correction procedures for x‐ray fluorescence analysis. A critical survey | |
Xu et al. | A machine learning algorithm for retrieving the geometrical characteristic parameters of soot fractal aggregates from polarized light signal | |
Bučar et al. | An improved neural computing method for describing the scatter of S–N curves | |
CN113723471B (en) | Nanoparticle concentration and particle size estimation method and device | |
Anton et al. | Deterministic and statistical calibration of constitutive models from full-field data with parametric physics-informed neural networks | |
Luchnikov et al. | Simulating quantum circuits using the multi-scale entanglement renormalization ansatz | |
Bai et al. | FREmu: Power Spectrum Emulator for $ f (R) $ Gravity |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |