CN114742289B - Production process parameter-oriented Gaussian process robust optimization method - Google Patents
Production process parameter-oriented Gaussian process robust optimization method Download PDFInfo
- Publication number
- CN114742289B CN114742289B CN202210328327.6A CN202210328327A CN114742289B CN 114742289 B CN114742289 B CN 114742289B CN 202210328327 A CN202210328327 A CN 202210328327A CN 114742289 B CN114742289 B CN 114742289B
- Authority
- CN
- China
- Prior art keywords
- data
- parameters
- uncertainty
- functions
- sample
- 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 94
- 238000005457 optimization Methods 0.000 title claims abstract description 48
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 22
- 238000009826 distribution Methods 0.000 claims abstract description 14
- 238000012216 screening Methods 0.000 claims abstract description 11
- 238000005070 sampling Methods 0.000 claims abstract description 10
- 230000000694 effects Effects 0.000 claims description 15
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 13
- 239000013598 vector Substances 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 5
- 238000011002 quantification Methods 0.000 claims description 5
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 2
- 238000001914 filtration Methods 0.000 claims 1
- 230000007547 defect Effects 0.000 abstract description 17
- 238000013461 design Methods 0.000 abstract description 2
- 230000015556 catabolic process Effects 0.000 abstract 1
- 238000006731 degradation reaction Methods 0.000 abstract 1
- 230000001105 regulatory effect Effects 0.000 abstract 1
- 229910000831 Steel Inorganic materials 0.000 description 13
- 239000010959 steel Substances 0.000 description 13
- 230000004913 activation Effects 0.000 description 4
- 238000013139 quantization Methods 0.000 description 4
- 238000009776 industrial production Methods 0.000 description 3
- 238000004886 process control Methods 0.000 description 2
- 238000003908 quality control method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000013386 optimize process Methods 0.000 description 1
- 238000009628 steelmaking Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- Educational Administration (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Factory Administration (AREA)
- Feedback Control In General (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A Gaussian process steady optimization method for production process parameters belongs to the technical field of industrial design. In the method, firstly, uncertainty in production data is quantified, uncertainty screening of the data is carried out, and then a Gaussian process optimization model facing process parameters is constructed. The method comprises the following steps: 1) Generating a same-dimension sample based on original sample distribution by using a Gan network, and obtaining an uncertainty confidence interval by using RASTRIGIN or Griewank functions after MCMC sampling; 2) Screening out samples outside the uncertainty confidence interval according to the output of the original samples in RASTRIGIN or Griewank functions; 3) And constructing a Gaussian process optimization model oriented to the technological parameters based on the screened samples. The invention can reduce the influence of uncertainty factors such as equipment degradation and the like on the regulation and control of the process parameters, establish an intelligent regulation and control mechanism of the process parameters, quickly obtain the value of the regulating and optimizing parameters and reduce the occurrence rate of product defects.
Description
Technical Field
The invention belongs to the technical field of industrial design, and relates to a Gaussian process steady optimization method for production process parameters.
Background
In an industrial process, the quality of the product is affected by many factors in the production process. For example, in the steelmaking process, the quality of steel is affected by adjustable factors such as temperature, stress, transmission speed and the like, and the selection of proper process parameters is important for quality control and defect control of products. The product quality is ensured mainly by post control, and after the production is finished, an expert performs inspection and judgment on the product quality according to experience, so that the following conditions often occur: firstly, when a product with poor quality is found, the correction is usually carried out later; secondly, the cost of expert experience is high, and the adjustment of the technological parameters is easy to make mistakes without reference range. The method has the advantages that a steady process parameter online optimization model is established, the optimized process parameters can be rapidly given through product data, the quality control time is greatly shortened, the occurrence rate of product defects is reduced, and an expert is guided to make judgment. Meanwhile, the process production flow is complicated, uncertainty factors are more, and data with high uncertainty are very easy to cause a model to make wrong judgment, so that screening out the data with high uncertainty before parameter optimization is important to the performance and robustness of the model. The processing of uncertainty data is not considered in most of the current online optimization methods of process parameters, and the result is unreliable.
The process parameter optimization method adopts a Bayesian optimization method based on a Gaussian process, and compared with other optimization methods: the Bayesian parameter adjustment adopts a Gaussian process, the prior parameter information is considered, the prior is continuously updated, and the prior parameter information is not considered in the grid searching method; the Bayesian parameter adjustment iteration times are small, the speed is high, the grid searching method and the genetic algorithm are slow, and the parameters are more, so that dimension explosion is easy to cause; bayesian tuning is still robust against non-convex problems, while grid search is prone to get a locally optimal solution against non-convex problems. Therefore, a robust model with less constraints and high speed can be constructed by adopting Gaussian process parameter optimization.
The uncertainty quantization in the invention uses Gan network, MCMC sampling and function for uncertainty quantization, the Gan network is composed of a generator and a discriminator, the distribution condition of original data can be fully extracted, random samples are generated according to the original distribution of the data, and then the uncertainty confidence interval is obtained by using the function for uncertainty quantization.
Disclosure of Invention
The invention mainly aims at solving the problems of more uncertainty factors, more process control targets, conflict, high optimization difficulty and the like in the industrial production process, and provides a Gaussian process steady optimization method based on uncertainty quantification, which is characterized in that the uncertainty in production data is quantified, the data with higher uncertainty is screened out, and then a Gaussian process optimization model facing the process parameters is constructed, so that the optimization parameters with better effects are obtained quickly and efficiently, the process optimization is guided, the product quality is improved, and the occurrence rate of product defects is reduced.
In order to achieve the above purpose, the invention adopts the following technical scheme:
a Gaussian process robust optimization method for production process parameters comprises the following steps:
first, a Gan network is used to generate a co-dimensional sample based on the original sample distribution, and after MCMC sampling, an uncertainty confidence interval is obtained using RASTRIGIN or Griewank functions. The method comprises the following steps:
First, let the original data set X have N samples, where the dimension of the samples is d, and only one tag with a value range of [0,1] is included (if the tag range is not [0,1], normalization is performed on the tag). Firstly, constructing a Gan network to generate a same-dimension sample based on original sample distribution, wherein a generator G and a discriminator D of the Gan network are both multi-layer neural networks. The initial vector of the dimension t (t < d) designated by the (program) is taken as the input of the generator G, and the data with dimension d and the same dimension as the original sample is output. The discriminator D takes the generator output G (X) and the original sample X as inputs, outputting a probability value between 0 and 1. The step of generating countermeasure training by alternately training the discriminator and the generator, and finishing model training when the appointed iteration times are reached, wherein the generator G is the required target network. For any random t-dimensional initial vector z, G (z) is obtained by constructing a Gan network, wherein G (z) is one random sample based on the original data distribution generated by a model, and therefore a plurality of random sample sets based on the original data distribution can be obtained.
Then, MCMC sampling is employed. And performing MCMC sampling on a random sample set generated by the Gan network to obtain m d-dimensional vectors, and recording the vector set as a data set Z.
Finally, adopting RASTRIGIN functions shown in a formula (1) or Griewank functions shown in a formula (2) for quantifying uncertainty;
Where x represents one sample in the dataset Z, the dimension of sample x is d, i.e., x= (x 1,x2,…xd). Since the data set Z contains m samples, inputting the data set Z into any of the functions above can result in m one-dimensional values, the magnitude of which can measure the uncertainty of each data sample. The data set Z is ordered according to the uncertainty, and the middle 95% of data is taken as trusted data. And taking the maximum value of uncertainty in the trusted data as an upper limit b and the minimum value as a lower limit a, and obtaining uncertainty confidence intervals [ a, b ] of the data set Z.
Second, samples outside the uncertainty confidence interval are screened out based on the output of the original samples in RASTRIGIN or Griewank functions. Specific:
Taking the original sample X as an input of RASTRIGIN functions or Griewank functions, wherein the selection of the functions is consistent with the first step, and if the output of the functions is within an uncertainty confidence interval [ a, b ], reserving data; and otherwise deleting the data. The step can realize the quantification and screening of the uncertainty of the original sample, and the screened data set is marked as P.
Thirdly, constructing a Gaussian process optimization model facing the process parameters based on the screened samples (data set P).
First, the data processing is performed again on the screened data set P. Specific: the adjustable parameters and fixed state parameters to be optimized in the data set P are manually set, and each data in the data set P is composed of an input-output pair (x p,yp), wherein x p is a process parameter (including the adjustable parameters and the state parameters), and y p is a process effect value (such as the defect number) when the process parameter is set to x p. Screening out data with y p as an abnormal value (the value of y p is far greater or far smaller than most data, namely, the process effect is good or poor), and carrying out normalization processing on all the data.
Next, a multi-layer neural network net is constructed using the normalized data (x p,yp) as a training set, where x p is the input and y p is the output. And a group of process parameters are transmitted in, the multi-layer neural network net can realize the prediction of the process effect value y under the current parameter x, and the better the predicted value is, the better the parameter is explained. Then, when new production process parameters are input, fixing state parameters in the input parameters, and testing the network net through the adjustable parameters in the traversal history data, thereby obtaining the following steps: in the current production state, each of the historically adjustable parameters is capable of producing a process effect value. And sorting all the historical adjustable parameters according to the process effect values, and selecting the first K historical adjustable parameter values with the best output process effect values, wherein the K adjustable parameter values are the first K most likely adopted historical process parameter regulation values according with the current state.
Finally, the value range of each parameter value can be known according to expert experience, and data outside the value range are filtered out from the K historical adjustable parameter values, so that L effective parameter values (L is less than or equal to K) are finally obtained. And constructing a Gaussian process regression model according to the L technological parameters (effective parameter values) and the output of the corresponding multi-layer neural network net, and optimizing the adjustable parameters by using Bayesian optimization iteration to obtain a final optimization model, wherein the optimization model is the established Gaussian process optimization model facing the technological parameters.
And fourthly, stopping optimizing when the Gaussian process optimizing model reaches the iteration times or the optimizing result is worse than the last optimizing result, and taking the last optimizing result as the optimizing value of the final adjustable parameter.
The beneficial effects of the invention are as follows:
Aiming at the problems of more uncertainty factors, more process control targets, conflict, high optimization difficulty and the like in the industrial production process, the invention provides a Gaussian process robust optimization method based on uncertainty quantification, which is characterized in that the uncertainty in production data is quantified, the data with higher uncertainty is screened out, and then a Gaussian process optimization model facing the process parameters is constructed, so that the optimization parameters with better effects are obtained quickly and efficiently, the process optimization is guided, the product quality is improved, and the occurrence rate of product defects is reduced. The method provided by the invention is simple and convenient to operate, meets the process realizability, is quick and efficient, and is convenient to integrate in the industrial production process.
Drawings
Fig. 1 is a flowchart of a method for implementing robust optimization of gaussian process for production process parameters.
Detailed Description
In order to make the solution to the problems of the method, the method scheme adopted and the effect of the method achieved by the invention more clear, the invention is further described in detail below with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting thereof. It should be further noted that, for convenience of description, only some, but not all of the matters related to the present invention are shown in the accompanying drawings.
Fig. 1 is a flowchart of a robust optimization method of gaussian process for production process parameters according to an embodiment of the present invention. The invention provides the implementation cases: in the embodiment, the steel production process is adopted, the data comprise 3000 samples of each technological parameter data and the final steel defect number, wherein 306 technological parameters are available, the optimized technological parameters are the first 6 technological parameters, and the rest 300 technological parameters are all non-adjustable state parameters. We will optimize these 6 optimizable parameters by the present invention, optimizing the number of steel defects produced. As shown in fig. 1, the gaussian process robust optimization method for production process parameters provided by the embodiment of the invention comprises the following steps:
First, a Gan network is used to generate a co-dimensional sample based on the original sample distribution, and after MCMC sampling, a Griewank function is used to obtain an uncertainty confidence interval. The method comprises the following steps:
Firstly, steel data have 306 technological parameters in total, the labels are defect numbers, normalization processing is needed to be carried out on the labels, and all the labels are divided by the maximum value in the labels, so that the value range of the labels is changed into [0,1].
Secondly, constructing a Gan network according to the processed data. The generator G of the Gan network is set to be a five-layer neural network, the activation function of the hidden layer is relu functions, and the activation function of the output layer is a sigmoid function; the discriminator D is set as a three-layer neural network, the hidden layer activation function is relu functions, and the output layer activation function is a sigmoid function. The 20-dimensional initial vector is randomly generated as input to the generator G, outputting 306-dimensional data of the same dimension as the original sample. The discriminator D takes the output of the generator and the original sample as inputs, and outputs a probability value between 0 and 1. The method comprises the steps of alternately training a discriminator and a generator to generate countermeasure training, and finishing model training when the appointed iteration number reaches 3000, wherein the generator G is the required target network. 3000 generated random data based on the original data distribution were obtained based on the trained generator.
Then, MCMC sampling is employed. MCMC sampling is performed on 3000 random samples generated to obtain 2000 data of 306 dimensions, which have the same distribution as the original steel data.
Finally, the 2000 data with 306 dimensions obtained above are input into Griewank functions, so that uncertainty quantized values of the 2000 data are obtained, the 2000 data are ordered by the uncertainty quantized values, and the middle 95% of the data are taken as trusted data. Uncertainty confidence intervals are obtained with a maximum of 0.039 and a minimum of 0.021 for the trusted data [0.021,0.039].
And secondly, screening out samples positioned outside the uncertainty confidence interval according to the output of the original samples in the Griewank functions. The method comprises the following steps:
And taking 3000 pieces of original steel data as input of Griewank functions to obtain an uncertainty quantized value of each piece of data, obtaining 230 pieces of data with the uncertainty quantized value outside a confidence interval [0.021,0.039], and deleting the 230 pieces of data. The method finishes the quantification and screening of the uncertainty of the steel data, and finally obtains 2770 steel data with lower uncertainty.
Thirdly, constructing a Gaussian process optimization model facing the process parameters based on the screened samples.
First, 2770 steel data after screening are subjected to data processing again. The first 6 adjustable parameters and the rest 300 fixed parameters in the steel data to be optimized. The labels in the data set are defect numbers, the values of the defect numbers are mostly between 0 and 500, the few defect numbers reach 2000 or 3000, the defect numbers are regarded as abnormal values, samples containing abnormal defect data are deleted, all the data are normalized, and 2753 pieces of normalized steel data are obtained after the normalization.
Secondly, constructing a multi-layer neural network net by using normalized steel data, setting the defect number in the data as q, and outputting 1-q as training data of the network, wherein the probability value of the network output as [0,1] is smaller, and the probability is smaller, so that the defects are more; the larger the probability, the fewer the defects, thereby maximizing the network output as a training target for the network. When new technological parameters are input, the state parameters of the input data are fixed, namely 300 parameters after the state parameters are fixed, 6 adjustable parameters in the traversal history data are tested on the network net, and the first 10 historical adjustable parameter values with the smallest output probability are selected, wherein the 10 adjustable parameter values are the first 10 most likely adopted historical technological parameter regulation values which accord with the current state.
Finally, the 10 adjustable parameters are in the normal range according to expert experience, and are judged to be effective. And constructing a Gaussian process regression model according to the 10 adjustable parameter values and the output probability of the corresponding neural network, and optimizing the adjustable parameters by using Bayesian optimization iteration, wherein the optimization model is the established Gaussian process optimization model facing the process parameters.
And fourthly, stopping optimizing when the Gaussian process optimizing model reaches the iteration times or the optimizing result is worse than the last optimizing result, and taking the last optimizing result as the optimizing value of the final adjustable parameter. Table 1 shows the parameter optimization results of this example, table 1 is a schematic diagram of gaussian process robust optimization results based on uncertain quantization. From the table contents, it can be seen that: along with the increasing of the iteration number iter, the adjustable parameters x1, x2, x3, x4, x5 and x6 are continuously optimized, the output value target of the neural network reaches 0.8211 from the initial 0.4647, and the parameter optimization effect is good. Through optimization, the parameters when the iteration number iter is 24 are taken as the final optimization result.
TABLE 1
Finally, it should be noted that: the above examples are only intended to illustrate the method aspects of the invention, not to limit it; although the invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will appreciate that: which modifies the method aspects described in the foregoing embodiments or provides equivalent replacement of some or all of the method features therein without departing from the spirit and scope of the method aspects of the embodiments of the invention.
Claims (3)
1. A Gaussian process robust optimization method for production process parameters is characterized by comprising the following steps:
Firstly, using a Gan network to generate a co-dimensional sample based on original sample distribution, and using RASTRIGIN or Griewank functions to obtain an uncertainty confidence interval after MCMC sampling; the method comprises the following steps:
1.1 Setting the original data set X to have N samples, wherein the dimension of the samples is d, and only one label with the value range of [0,1] is contained; firstly, constructing a Gan network to generate a homonymous sample based on original sample distribution, wherein a generator G and a discriminator D of the Gan network are both multi-layer neural networks; taking the designated t (t < d) dimension initial vector as the input of the generator G, and outputting data with the dimension d, which is in the same dimension as the original sample; the discriminator D takes the output G (X) of the generator and the original sample X as inputs, and outputs a probability value between 0 and 1; generating countermeasure training by alternately training a discriminator and a generator, and finishing model training when the number of iterations reaches a specified number, wherein the generator G is the required target network; for any random t-dimensional initial vector z, G (z) is obtained by constructing a Gan network, wherein G (z) is a random sample based on original data distribution generated by a model, so that a plurality of random sample sets based on the original data distribution can be obtained;
1.2 MCMC sampling is carried out on a random sample set generated by the Gan network, m vectors with d dimensions are obtained, and the vector set is recorded as a data set Z;
1.3 Using RASTRIGIN functions as shown in equation (1) or Griewank functions as shown in equation (2) to quantify uncertainty;
where x represents one sample in the dataset Z, the dimension of sample x is d, i.e., x= (x 1,x2,…xd); since the data set Z contains m samples, inputting the data set Z into any of the above functions can result in m one-dimensional values, the magnitude of which can measure the uncertainty of each data sample; ordering the data set Z according to the uncertainty, and taking the intermediate data as trusted data; taking the maximum value and the minimum value of uncertainty in the trusted data as an upper limit b and a lower limit a to obtain uncertainty confidence intervals [ a, b ] of a data set Z;
A second step, taking the original sample X as an input of RASTRIGIN functions or Griewank functions, wherein the selection of the functions is consistent with the first step, and if the output of the functions is within an uncertainty confidence interval [ a, b ], the data is reserved; otherwise, deleting the data, namely screening out samples positioned outside the uncertainty confidence interval; the step can realize the quantification and screening of the uncertainty of the original sample, and the screened data set is marked as P;
Thirdly, constructing a Gaussian process optimization model facing the process parameters based on the screened data set P;
3.1 Data processing is carried out on the screened data set P again; specific: manually setting adjustable parameters and fixed state parameters to be optimized in a data set P, wherein each data in the data set P is composed of an input-output pair (x p,yp), wherein x p is a process parameter, and y p is a process effect value when the process parameter is set as x p; screening out data with y p as an abnormal value, and carrying out normalization processing on all the data;
3.2 Using the normalized data (x p,yp) as a training set to construct a multi-layer neural network net, wherein x p is input and y p is output; a group of process parameters are transmitted, and the multi-layer neural network net can realize the prediction of the process effect value y under the current parameter x; then, when new production process parameters are input, fixing state parameters in the input parameters, and testing the network net through the adjustable parameters in the traversal history data, thereby obtaining the following steps: in the current production state, each history adjustable parameter can generate a process effect value; sorting all the historical adjustable parameters according to the process effect values, and selecting the first K historical adjustable parameter values with the best output process effect values, wherein the K adjustable parameter values are the first K most likely to be adopted historical process parameter regulation values according with the current state;
3.3 Filtering and removing unsuitable technological parameters from K historical adjustable parameter values according to expert experience to finally obtain L effective parameter values, wherein L is less than or equal to K; constructing a Gaussian process regression model according to the L effective process parameters and the output of the corresponding multi-layer neural network net, and optimizing the adjustable parameters by using Bayesian optimization iteration to obtain a final optimization model, wherein the optimization model is the established Gaussian process optimization model facing the process parameters;
And fourthly, stopping optimizing when the Gaussian process optimizing model reaches the iteration times or the optimizing result is worse than the last optimizing result, and taking the last optimizing result as the optimizing value of the final adjustable parameter.
2. The method of claim 1, wherein in the first step 1.1), if the label range is not [0,1], the label is normalized.
3. The method for robust optimization of gaussian process for production process parameters according to claim 1, wherein in said first step 1.3), the data set Z is sorted according to the uncertainty and the middle 95% of the data is taken as the trusted data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210328327.6A CN114742289B (en) | 2022-03-31 | 2022-03-31 | Production process parameter-oriented Gaussian process robust optimization method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210328327.6A CN114742289B (en) | 2022-03-31 | 2022-03-31 | Production process parameter-oriented Gaussian process robust optimization method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114742289A CN114742289A (en) | 2022-07-12 |
| CN114742289B true CN114742289B (en) | 2024-08-16 |
Family
ID=82279116
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210328327.6A Active CN114742289B (en) | 2022-03-31 | 2022-03-31 | Production process parameter-oriented Gaussian process robust optimization method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114742289B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117453805B (en) * | 2023-12-22 | 2024-03-15 | 石家庄学院 | Visual analysis method for uncertainty data |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119778A (en) * | 2019-05-10 | 2019-08-13 | 辽宁大学 | A kind of equipment method for detecting health status improving chicken group's optimization RBF neural |
| CN113923104A (en) * | 2021-12-07 | 2022-01-11 | 南京信息工程大学 | Network fault diagnosis method, equipment and storage medium based on wavelet neural network |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021007812A1 (en) * | 2019-07-17 | 2021-01-21 | 深圳大学 | Deep neural network hyperparameter optimization method, electronic device and storage medium |
| CN113689055B (en) * | 2021-10-22 | 2022-01-18 | 西南石油大学 | Oil-gas drilling machinery drilling speed prediction and optimization method based on Bayesian optimization |
-
2022
- 2022-03-31 CN CN202210328327.6A patent/CN114742289B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119778A (en) * | 2019-05-10 | 2019-08-13 | 辽宁大学 | A kind of equipment method for detecting health status improving chicken group's optimization RBF neural |
| CN113923104A (en) * | 2021-12-07 | 2022-01-11 | 南京信息工程大学 | Network fault diagnosis method, equipment and storage medium based on wavelet neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114742289A (en) | 2022-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105786860B (en) | Data processing method and device in data modeling | |
| CN103853786B (en) | The optimization method and system of database parameter | |
| CN112987664A (en) | Flow shop scheduling method based on deep reinforcement learning | |
| CN110991495B (en) | Method, system, medium, and apparatus for predicting product quality in manufacturing process | |
| CN108805257A (en) | A kind of neural network quantization method based on parameter norm | |
| GB2600213A (en) | Plate performance prediction method and apparatus | |
| JP2008059601A (en) | Method for identifying bimodal data | |
| CN111583031A (en) | Application scoring card model building method based on ensemble learning | |
| CN111738477A (en) | Deep feature combination-based power grid new energy consumption capability prediction method | |
| CN111985825A (en) | Crystal face quality evaluation method for roller mill orientation instrument | |
| CN113869521A (en) | Method, device, computing equipment and storage medium for constructing prediction model | |
| CN111444963A (en) | Blast furnace molten iron silicon content prediction method based on SSA-SVR model | |
| CN114742289B (en) | Production process parameter-oriented Gaussian process robust optimization method | |
| CN113128124B (en) | Multi-grade C-Mn steel mechanical property prediction method based on improved neural network | |
| CN113032367A (en) | Dynamic load scene-oriented cross-layer configuration parameter collaborative tuning method and system for big data system | |
| CN115861246A (en) | Product quality abnormity detection method and system applied to industrial Internet | |
| CN112184412A (en) | Modeling method, device, medium and electronic equipment of credit rating card model | |
| CN111507824A (en) | Wind control model mold-entering variable minimum entropy box separation method | |
| CN118467992B (en) | Short-term power load prediction method, system and storage medium based on meta-heuristic optimization | |
| CN117421657B (en) | Method and system for screening and learning picture samples with noise labels based on oversampling strategy | |
| CN114385876B (en) | Model search space generation method, device and system | |
| CN115201394B (en) | Multi-component transformer oil chromatography online monitoring method and related device | |
| CN116402123A (en) | Pre-training model fine tuning method and system based on learning strategy | |
| CN116341768A (en) | Carbon emission prediction method based on deep learning | |
| CN113034316B (en) | Patent value conversion analysis method and system |
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 |