CN110262233B - Optimization method for technological parameters of magnetic control film plating instrument - Google Patents
Optimization method for technological parameters of magnetic control film plating instrument Download PDFInfo
- Publication number
- CN110262233B CN110262233B CN201910466571.7A CN201910466571A CN110262233B CN 110262233 B CN110262233 B CN 110262233B CN 201910466571 A CN201910466571 A CN 201910466571A CN 110262233 B CN110262233 B CN 110262233B
- Authority
- CN
- China
- Prior art keywords
- neural network
- threshold
- film
- layer
- weight
- 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
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Physical Vapour Deposition (AREA)
Abstract
The invention provides a method for optimizing process parameters of a magnetic control film plating instrument based on a genetic algorithm and a BP neural network, which comprises the following steps: the method comprises the following steps: selecting technological parameters of a magnetic control coating instrument for testing to obtain the resistivity, transmittance and thickness of the film; step two: taking the technological parameters of a magnetic control coating instrument as input, and taking the resistivity, transmittance and thickness of a film as output, and constructing a BP neural network model containing a plurality of hidden layers; step three: optimizing initial weight and threshold of the BP neural network by using a genetic algorithm to obtain optimal individual weight and threshold; step four: assigning the weight and the threshold of the optimal individual obtained in the third step to a BP neural network model, and training the BP neural network by using the sample obtained in the first step; step five: optimizing the technological parameters of the magnetron sputtering coating instrument by using the genetic algorithm again, wherein a fitness function is constructed by using the predicted output of the neural network obtained by training in the step four; the invention can quickly and accurately optimize the parameters of the coating process of the magnetron sputtering method.
Description
Technical Field
The invention relates to a film preparation technology, in particular to a method for optimizing process parameters of a magnetic control film coating instrument.
Background
The magnetron sputtering technology is a new film preparation technology developed in the aspect of vacuum coating in the seventies, is a high-speed and low-temperature sputtering technology, can be used for preparing multi-materials such as metal, semiconductor, insulator and the like by a general sputtering method, and has the advantages of simple equipment, easiness in control, large coating area, strong adhesive force and the like.
The artificial neural network is used as an intelligent information processing system which simulates the structure and the function of the human brain and is formed by connecting a large number of simple computing units, has the advantages of large-scale parallel processing, self-learning, self-adaptability and the like, and is suitable for constructing a complex nonlinear correlation model. The BP artificial neural network is a multilayer feedforward network trained according to an error back propagation algorithm, is one of neural network models which are most widely applied at present, but has some defects, such as low convergence rate, easiness in trapping at local minimum points and the like, which can cause large deviation in a parameter optimization process, so that the optimization result is not ideal. In recent years, the application of BP neural network models in the industrial field is increasing, including identification, classification and grading, simulation and control of the processing process, prediction of a single index value and the like, and certain effects are achieved, but the selection of magnetron sputtering coating process parameters depends on the experience of workers, a related optimization algorithm is lacked to optimize the magnetron sputtering coating process parameters, the time spent for manually selecting the parameters is long, the experimental process is long, and the material waste is caused due to the low precision of manually selecting the parameters.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides an optimization method of the process parameters of a magnetron coating instrument, which can quickly and accurately optimize the coating process parameters of a magnetron sputtering method based on a genetic algorithm and a BP neural network.
The present invention achieves the above-described object by the following technical means.
A method for optimizing technological parameters of a magnetic control coating instrument comprises the following steps:
the method comprises the following steps: selecting technological parameters of a magnetron coating instrument for testing to obtain the resistivity, the light transmittance and the film thickness of the film, carrying out normalization processing on the obtained resistivity, the light transmittance and the film thickness data, and then selecting a training sample and a detection sample from the data, wherein the technological parameters of the magnetron coating instrument comprise sputtering power, sputtering pressure, substrate temperature, vacuum degree, sputtering time and argon flow;
step two: taking the technological parameters of the magnetic control coating instrument as input, taking the resistivity, the light transmittance and the thickness of the film as output, constructing a BP neural network model containing a plurality of hidden layers, setting the number of the hidden layers as l layers, setting the number of nodes of each hidden layer as H, and selecting proper excitation functions for each hidden layer and each output layer;
step three: optimizing the initial weight and threshold of the BP neural network by using a genetic algorithm to obtain the optimal weight and threshold of an individual;
step four: assigning the weight and the threshold of the optimal individual obtained in the third step to the BP neural network model, training the BP neural network by using the sample in the first step, and updating the weight and the threshold of each hidden layer by using an error inverse propagation algorithm in the training process until the cost function J is less than the set precision or the training of the maximum iteration number is finished;
step five: and (3) optimizing the technological parameters of the magnetic control coating instrument by using the genetic algorithm again, wherein a fitness function F2 is constructed by using the prediction output of the neural network obtained by training in the step four, and the fitness function F2 is obtained by:
wherein F2 is a fitness function, rho 2 is the predicted film resistivity of the neural network obtained in the step four, eta 2 is the predicted film transmittance of the neural network obtained in the step four, the predicted film thickness of the neural network obtained in the step four and rho 2 0 Is a film resistivity target value, η 0 Is a target value of the transmittance, delta 0 Is a target film thickness value.
Preferably, the relationship between the resistivity, the light transmittance and the thickness of the film in the first step and the process parameters of the magnetron coating apparatus is represented as follows:
ρ＝f 1 (P,P a ,T,V,t,F)，η＝f 2 (P,P a ,T,V,t,F)，δ＝f 3 (P,P a ,T,V,t,F)
wherein rho is the resistivity of the film, eta is the light transmittance of the film, delta is the thickness of the film, P is the sputtering power of the magnetron coating instrument, and P is a T is the sputtering pressure, T is the substrate temperature, V is the vacuum, T is the sputtering time, and F is the argon flow.
Preferably, in the second step, the excitation function of each hidden layer selects a logistic function
The excitation function of the output layer selects a linear function g (x) = x.
Preferably, the third step is specifically:
3.1 firstly, determining the weight value and the number of threshold values of the neural network according to the topological graph of the BP neural network model, and following the following formula:
where Num is the total number of weight and threshold, H i The node number of the L layer neuron;
3.2, coding the weight threshold value of the neural network by adopting a real number coding mode, initializing a population, randomly taking a value of the initial weight threshold value between (-1, 1), and setting a fitness function of the population as F1;
3.3 calculate the fitness value of all individuals in the population and use roulette algorithm to select the individuals with high fitness from the parent to generate the next generation of individuals, the probability of each individual being selected following the formula:
wherein p is k Probability of being selected for the kth individual, F k Is the fitness value of the kth individual, and K is the total number of individuals in the population;
3.4 Cross-operating individuals in the population, and setting the cross probability as p c If a random number is generated and is smaller than the crossing probability, carrying out crossing operation, randomly selecting two individuals and randomly selecting a crossing position during crossing, and carrying out the crossing operation according to the following formula:
wherein, a kj Is the real number of the kth individual in the j position, a lj Is the real number of the l-th individual at the j position, and b is a random number between (0, 1);
3.5 performing mutation operation on individuals in the population, and setting the mutation probability as p m If a random number is generated and is smaller than the mutation probability, carrying out mutation operation, randomly selecting an individual and randomly selecting a mutation site during mutation, and carrying out mutation operation according to the following formula:
wherein, a ij Is the real number of the ith chromosome at position j, G is the current iteration number, G max To the maximum number of iterations, a max Is a ij Upper limit of value, a min Is a ij The lower limit of the value, r and r' are random numbers between (0, 1);
and 3.5, circulating the steps 3.3-3.5 until a satisfactory fitness value is obtained or a limited iteration number is reached, and outputting the optimal individual, namely the individual with the maximum fitness value.
Preferably, the fitness function F1 takes:
wherein F1 is a fitness function, ρ 1 is the resistivity of the film predicted by the neural network using the initial weight and threshold, η 1 is the transmittance predicted by the neural network using the initial weight and threshold, δ 1 is the film thickness predicted by the neural network using the initial weight and threshold, ρ 1 is the fitness function 0 Is a film resistivity target value, η 0 Is a target value of the transmittance, delta 0 Is a target film thickness value.
Preferably, the cost function J in step four is set as:
wherein m is the number of data, n is the number of nodes of the input layer,
y p is the actual output of the output node p, target p Is the desired output of the output node p.
Preferably, the specific process of the step four is as follows:
4.1 assigning the weight and the threshold of the optimal individual obtained in the third step to the BP neural network model, and calculating the input and output values of each layer of the neural network, wherein the calculation formula is as follows:
net (l) ＝w (l) *y (l-1) +b (l)
y (l) ＝f l (net (l) )
therein, net (l) Is the input to the first layer neurons of the neural network, w (l) Is the weight of layer l-1 neurons to layer l neurons, b (l) Is the threshold value of layer I neurons, y (l) Is the output of layer I neurons, f l Is an activation function for layer I neurons;
4.2, by using a gradient descent method, obtaining a weight threshold after change:
wherein J is the cost function of the neural network, alpha is the learning rate of the neural network, and w (l) Is the weight of layer l-1 neurons to layer l neurons, b (l) Is the threshold for layer I neurons.
The invention has the beneficial effects that:
1) According to the invention, a complex mapping relation between process parameters (including sputtering power, sputtering pressure, substrate temperature, vacuum degree, sputtering time and argon flow) of the magnetron sputtering coating instrument and performance indexes (including resistivity, light transmittance and film thickness) of a produced film is established by adopting a multi-hidden layer BP neural network, the initial weight and threshold of the neural network are optimized by adopting a genetic algorithm, the global search capability of the neural network is enhanced, the training speed and the prediction precision of the neural network are improved, then the process parameters of the magnetron sputtering coating instrument are optimized by using the genetic algorithm again according to the performance requirements of a target product, and the optimization of the coating process parameters of the magnetron sputtering coating instrument is realized, so that the process parameters of the magnetron sputtering coating instrument can be predicted according to the parameters of the required product, the process of a pre-experiment is reduced, and the waste of materials and energy is reduced.
2) The invention is independent of a fixed film plating instrument system, is suitable for magnetron sputtering film plating machines of different models, and has high algorithm transportability.
Drawings
FIG. 1 is a flow chart of the optimization method of the process parameters of the magnetron coating apparatus of the invention.
FIG. 2 is a diagram of a topology structure of a multi-hidden layer BP neural network of the present invention.
Detailed Description
The invention will be further described with reference to the following figures and specific examples, but the scope of the invention is not limited thereto.
As shown in fig. 1, a method for optimizing process parameters of a magnetron coating apparatus according to an embodiment of the present invention includes the following steps:
the method comprises the following steps: selecting technological parameters of a magnetron coating instrument for testing to obtain the resistivity, the light transmittance and the film thickness of the film, carrying out normalization processing on the obtained resistivity, the light transmittance and the film thickness data, then selecting a data sample of a neural network from the data sample, selecting 80% of the data sample as a training sample, and using the rest data sample as a detection sample, wherein the technological parameters of the magnetron coating instrument comprise sputtering power, sputtering pressure, substrate temperature, vacuum degree, sputtering time and argon flow;
the relationship among the resistivity, the light transmittance and the thickness of the film and the process parameters of the magnetic control coating instrument is expressed as follows:
ρ＝f 1 (P,P a ,T,V,t,F)，η＝f 2 (P,P a ,T,V,t,F)，δ＝f 3 (P,P a ,T,V,t,F)
wherein rho is the resistivity of the film, eta is the light transmittance of the film, delta is the thickness of the film, P is the sputtering power of the magnetron coating instrument, and P is a The sputtering pressure is T, the substrate temperature is T, the vacuum degree is V, the sputtering time is T, and the argon flow is F;
step two: taking the technological parameters of the magnetic control coating instrument as input, taking the resistivity, the light transmittance and the thickness of the film as output, constructing a BP neural network model containing a plurality of hidden layers, setting the number of the hidden layers as l layers, setting the number of nodes of each hidden layer as H, and selecting proper excitation functions for each hidden layer and each output layer;
in this embodiment, the number g of hidden layers and the number H of nodes in each hidden layer are set by selecting the number of different hidden layers and the number of nodes in the hidden layers to train for the same number of times, comparing the prediction result with the measured data, and selecting the number of hidden layers and the number of nodes in the hidden layers with the minimum cost function J by comparing the cost function J;
the excitation function of each hidden layer selects a logistic function
The excitation function of the output layer selects a linear function g (x) = x;
step three: optimizing the initial weight and threshold of the BP neural network by using a genetic algorithm to obtain the optimal weight and threshold of the individual, wherein the method specifically comprises the following steps:
3.1 firstly, determining the weight value and the number of threshold values of the neural network according to the topological graph of the BP neural network model, and following the following formula:
where Num is the total number of weight values and threshold values, H i The number of nodes of the layer I neurons;
3.2, coding the weight threshold of the neural network by adopting a real number coding mode, initializing a population, wherein the length of a chromosome gene is equal to the sum of all weights and the number of thresholds in the network, the initial weight threshold randomly takes values within (-1, 1), and the fitness function of the population is set as F1:
where F1 is the fitness function and ρ 1 is the neural network predicted thinness using the initial weight and thresholdResistivity of the film,. Eta.1 is the transmittance predicted by the neural network using the initial weight and threshold, and. Delta.1 is the film thickness predicted by the neural network using the initial weight and threshold, ρ 0 Is a film resistivity target value, eta 0 Is a target value of transmittance, δ 0 Is a target film thickness value;
3.3 calculate the fitness value of all individuals in the population and use roulette algorithm to select the individuals with high fitness from the parent to generate the next generation of individuals, the probability of each individual being selected following the formula:
wherein p is k Probability of being selected for the kth individual, F k Is the fitness value of the kth individual, and K is the total number of individuals in the population;
3.4 Cross-operating individuals in the population, and setting the cross probability as p c If a random number is generated and is smaller than the crossing probability, carrying out crossing operation, randomly selecting two individuals and randomly selecting a crossing position during crossing, and carrying out the crossing operation according to the following formula:
wherein, a kj Is the real number of the kth individual in the j position, a lj Is the real number of the l-th individual at the j position, and b is a random number between (0, 1);
3.5 mutation operation is carried out on individuals in the population, and the mutation probability is set as p m If a random number is generated and is smaller than the mutation probability, carrying out mutation operation, randomly selecting an individual and randomly selecting a mutation site during mutation, and carrying out mutation operation according to the following formula:
wherein, a ij Is the real number of the ith chromosome at position j, G is the current iteration number, G max To the maximum number of iterations, a max Is a ij Upper limit of value, a min Is a ij The lower limit of the value, r and r' are random numbers between (0, 1);
and 3.5, circulating the steps 3.3-3.5 until a satisfactory fitness value is obtained or a limited iteration number is reached, and outputting the optimal individual, namely the individual with the maximum fitness value.
Step four: assigning the weight and the threshold of the optimal individual obtained in the third step to the BP neural network model, training the BP neural network by using the sample of the first step, and updating the weight and the threshold of each hidden layer by using an error inverse propagation algorithm in the training process until the cost function J of the neural network is less than the set precision or the training of the maximum iteration number is finished, wherein the specific process is as follows:
4.1, assigning the weight and the threshold of the optimal individual obtained in the third step to the BP neural network model, and calculating the input and output values of each layer of the neural network, wherein the calculation formula is as follows:
net (l) ＝w (l) *y (l-1) +b (l)
y (l) ＝f l (net (l) )
therein, net (l) Is the input to the first layer neurons of the neural network, w (l) Is the weight of layer I-1 neurons to layer I neurons, b (l) Is the threshold value of layer I neurons, y (l) Is the output of layer I neurons, f l Is an activation function for layer I neurons;
4.2, by using a gradient descent method, obtaining a weight threshold after change:
where α is the learning rate of the neural network, w (l) Is the weight of layer l-1 neurons to layer l neurons, b (l) Is the threshold of the l layer neuron, and J is the cost function of the neural network, and the formula is as follows:
where m is the number of data, n is the number of nodes in the input layer, y p Is the actual output of the output node p, target p Is the desired output of the output node p.
Step five: and optimizing the technological parameters of the magnetron sputtering coating instrument by using the genetic algorithm again, wherein the method comprises the following steps:
5.1, coding process parameters of the magnetron sputtering coating instrument by adopting a real number coding mode, initializing a population, predicting the output value of each individual by using the neural network obtained in the fourth step, and constructing a fitness function of the population as F2:
wherein F2 is a fitness function, rho 2 is the predicted film resistivity of the neural network obtained in the step four, eta 2 is the predicted film transmittance of the neural network obtained in the step four, the predicted film thickness of the neural network obtained in the step four and rho 2 0 Is a film resistivity target value, eta 0 Is a target value of the transmittance, delta 0 Is a target film thickness value;
5.2 calculating the fitness value of all individuals in the population, and performing selection operation by using a roulette algorithm;
5.3, performing cross operation on individuals in the population;
5.4 carrying out mutation operation on individuals in the population;
and (4) the specific process of the crossing in the step 5.3 and the mutation in the step 5.4 is similar to that in the step (3), then the steps 5.2-5.4 are circulated, the operations of selection, crossing and mutation are repeatedly carried out until a satisfactory fitness value is obtained or a limited number of iterations is reached, and the optimal individual, namely the optimized process parameters of the magnetron sputtering coating instrument, is output.
The present invention is not limited to the above-described embodiments, and any obvious improvements, substitutions or modifications can be made by those skilled in the art without departing from the spirit of the present invention.
Claims (7)
1. A method for optimizing technological parameters of a magnetic control coating instrument is characterized by comprising the following steps:
the method comprises the following steps: selecting technological parameters of a magnetron coating instrument for testing to obtain the resistivity, the light transmittance and the film thickness of the film, carrying out normalization processing on the obtained resistivity, the light transmittance and the film thickness data, and then selecting a training sample and a detection sample from the data, wherein the technological parameters of the magnetron coating instrument comprise sputtering power, sputtering pressure, substrate temperature, vacuum degree, sputtering time and argon flow;
step two: taking the technological parameters of the magnetic control coating instrument as input, taking the resistivity, the light transmittance and the thickness of the film as output, constructing a BP neural network model containing a plurality of hidden layers, setting the number of the hidden layers as l layers, setting the number of nodes of each hidden layer as H, and selecting proper excitation functions for each hidden layer and each output layer;
step three: optimizing the initial weight and threshold of the BP neural network model by using a genetic algorithm to obtain the optimal weight and threshold of the individual;
step four: assigning the weight and the threshold of the optimal individual obtained in the third step to the BP neural network model, training the BP neural network model by using the sample obtained in the first step, and updating the weight and the threshold of each hidden layer by using an error inverse propagation algorithm in the training process until the cost function J is less than the set precision or the training of the maximum iteration times is finished;
step five: and (3) optimizing the technological parameters of the magnetic control coating instrument by using the genetic algorithm again, wherein a fitness function F2 is constructed by using the prediction output of the neural network model obtained by training in the step four, and the fitness function F2 is obtained by:
wherein F2 is a fitness function, rho 2 is the film resistivity predicted by the neural network model obtained in the step four, eta 2 is the film light transmittance predicted by the neural network model obtained in the step four, delta 2 is the film thickness predicted by the neural network model obtained in the step four, and rho 0 Is a film resistivity target value, η 0 Is a target value of the transmittance, delta 0 Is a target film thickness value.
2. The optimization method of process parameters of the magnetron coating device according to claim 1,
the relationship among the resistivity, the light transmittance and the thickness of the film in the first step and the process parameters of the magnetron coating instrument is shown as follows:
ρ＝f 1 (P,P a ,T,V,t,F)，η＝f 2 (P,P a ,T,V,t,F)，δ＝f 3 (P,P a ,T,V,t,F)
wherein rho is the resistivity of the film, eta is the light transmittance of the film, delta is the thickness of the film, P is the sputtering power of the magnetron coating instrument, and P is a T is the sputtering pressure, T is the substrate temperature, V is the vacuum, T is the sputtering time, and F is the argon flow.
3. The method according to claim 1, wherein in step two, the excitation function of each hidden layer is selected from logistic functions
Excitation of the output layerThe function selects a linear function g (x) = x.
4. The optimization method of the technological parameters of the magnetron coating instrument according to claim 1, wherein the third step is specifically:
3.1 firstly, determining the weight value and the number of threshold values of the neural network according to the topological graph of the BP neural network model, and following the following formula:
where Num is the total number of weight and threshold, H l The node number of the l layer neuron;
3.2, coding the weight threshold value of the neural network by adopting a real number coding mode, initializing a population, randomly taking a value of the initial weight threshold value between (-1, 1), and setting a fitness function of the population as F1;
3.3 calculate the fitness value of all individuals in the population and use roulette algorithm to select the individuals with high fitness from the parent to generate the next generation of individuals, the probability of each individual being selected following the formula:
wherein p is k Probability of being selected for the kth individual, F k Is the fitness value of the kth individual, and K is the total number of individuals in the population;
3.4 Cross-operating individuals in the population, and setting the cross probability as p c If a random number is generated and is smaller than the crossing probability, carrying out crossing operation, randomly selecting two individuals and randomly selecting a crossing position during crossing, and carrying out the crossing operation according to the following formula:
wherein, a kj Is the real number of the kth individual at position j, a lj Is the real number of the l-th individual at the j position, and b is a random number between (0, 1);
3.5 mutation operation is carried out on individuals in the population, and the mutation probability is set as p m If a random number is generated and is smaller than the mutation probability, carrying out mutation operation, randomly selecting an individual and randomly selecting a mutation site during mutation, and carrying out mutation operation according to the following formula:
wherein, a ij Is the real number of the ith chromosome at position j, G is the current iteration number, G max As maximum number of iterations, a max Is a ij Upper limit of value, a min Is a ij The lower limit of the value, r and r' are random numbers between (0, 1);
and 3.6, circulating the steps 3.3-3.5 until a satisfactory fitness value is obtained or a limited iteration number is reached, and outputting the optimal individual, namely the individual with the maximum fitness value.
5. The optimization method of the technological parameters of the magnetron coating instrument according to claim 4, characterized in that: the fitness function F1 takes:
wherein F1 is a fitness function, ρ 1 is the resistivity of the film predicted by the neural network using the initial weight and the threshold, η 1 is the transmittance predicted by the neural network using the initial weight and the threshold, δ 1 is the thickness of the film predicted by the neural network using the initial weight and the threshold, ρ 1 0 Is a film resistivity target value, η 0 Is a target value of transmittance, δ 0 Is a target film thickness value.
6. The optimization method of the technological parameters of the magnetron coating instrument according to claim 1, characterized in that: the cost function J of the neural network in step four is set as:
wherein m is the number of data, n is the number of nodes of the input layer,
y p is the actual output of the output node p, target p Is the desired output of the output node p.
7. The optimization method of the technological parameters of the magnetron coating instrument according to claim 1, characterized in that: the specific process of the step four is as follows:
4.1 assigning the weight and the threshold of the optimal individual obtained in the third step to the BP neural network model, and calculating the input and output values of each layer of the neural network, wherein the calculation formula is as follows:
net (l) ＝w (l) *y (l-1) +b (l)
y (l) ＝f l (net (l) )
therein, net (l) Is the input to the first layer neurons of the neural network, w (l) Is the weight of layer I-1 neurons to layer I neurons, b (l) Is the threshold for layer I neurons, y (l) Is the output of layer I neurons, f l Is an activation function for layer I neurons;
4.2, by using a gradient descent method, obtaining a weight threshold after change:
wherein J is the cost function of the neural network, alpha is the learning rate of the neural network, and w (l) Is the weight of layer I-1 neurons to layer I neurons, b (l) Is the threshold for layer I neurons.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910466571.7A CN110262233B (en) | 2019-05-31 | 2019-05-31 | Optimization method for technological parameters of magnetic control film plating instrument |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910466571.7A CN110262233B (en) | 2019-05-31 | 2019-05-31 | Optimization method for technological parameters of magnetic control film plating instrument |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110262233A CN110262233A (en) | 2019-09-20 |
| CN110262233B true CN110262233B (en) | 2022-12-16 |
Family
ID=67916158
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910466571.7A Active CN110262233B (en) | 2019-05-31 | 2019-05-31 | Optimization method for technological parameters of magnetic control film plating instrument |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110262233B (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111126560A (en) * | 2019-11-07 | 2020-05-08 | 云南民族大学 | Method for optimizing BP neural network based on cloud genetic algorithm |
| CN111753886A (en) * | 2020-06-09 | 2020-10-09 | 上海电气集团股份有限公司 | Equipment fault processing method and device and terminal |
| CN111880402B (en) * | 2020-07-30 | 2023-01-24 | 广州大学 | Method and device for controlling parameters of fluorescent powder layer product and storage medium |
| CN112084596B (en) * | 2020-09-08 | 2021-06-15 | 北京航空航天大学 | Intelligent recommendation system and recommendation method for structural parameters of floor |
| CN112446157A (en) * | 2020-12-14 | 2021-03-05 | 广东省科学院智能制造研究所 | Method and device for predicting residual life of traffic equipment |
| CN112685504B (en) * | 2021-01-06 | 2021-10-08 | 广东工业大学 | Production process-oriented distributed migration chart learning method |
| CN113779887A (en) * | 2021-09-22 | 2021-12-10 | 上海上实龙创智能科技股份有限公司 | MF impregnating adhesive production process parameter optimization control method |
| CN114091848A (en) * | 2021-11-04 | 2022-02-25 | 北京北方华创微电子装备有限公司 | Method and system for automatically acquiring semiconductor process formula and semiconductor process equipment |
| CN114525497A (en) * | 2022-02-16 | 2022-05-24 | 东北大学 | Multi-parameter optimization method for PECVD (plasma enhanced chemical vapor deposition) coating process based on LSTM (localized surface plasmon resonance) and NSGA-II (non-uniform gas chromatography-tandem chemical vapor deposition) |
| CN116128102B (en) * | 2022-12-09 | 2024-01-23 | 常熟中佳新材料有限公司 | Production optimization method and system for special-shaped copper pipe for refrigeration equipment |
| CN115689063A (en) * | 2022-12-30 | 2023-02-03 | 工业富联(杭州)数据科技有限公司 | Gold film thickness prediction method and device, electronic device and storage medium |
| CN117408052A (en) * | 2023-10-18 | 2024-01-16 | 南栖仙策(南京)高新技术有限公司 | Coating control optimization method, device and equipment for evaporator and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0375204A (en) * | 1989-08-18 | 1991-03-29 | Sumitomo Cement Co Ltd | Production of oxide superconductive film pattern |
| CN105608494A (en) * | 2015-12-21 | 2016-05-25 | 中国科学院长春光学精密机械与物理研究所 | Magnetron sputtering velocity space distribution inversion method based on genetic algorithm |
| FR3063675A1 (en) * | 2017-03-10 | 2018-09-14 | Lvmh Swiss Manufactures Sa | MATERIAL COMPRISING A THIN LAYER OF AN ALLOY COMPRISING TITANIUM AND GOLD, AND PROCESS FOR OBTAINING SUCH MATERIAL |
| CN208008884U (en) * | 2018-03-27 | 2018-10-26 | 合肥科晶材料技术有限公司 | A kind of five target magnetic control sputtering plated film instrument for high-throughput coating process |
-
2019
- 2019-05-31 CN CN201910466571.7A patent/CN110262233B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0375204A (en) * | 1989-08-18 | 1991-03-29 | Sumitomo Cement Co Ltd | Production of oxide superconductive film pattern |
| CN105608494A (en) * | 2015-12-21 | 2016-05-25 | 中国科学院长春光学精密机械与物理研究所 | Magnetron sputtering velocity space distribution inversion method based on genetic algorithm |
| FR3063675A1 (en) * | 2017-03-10 | 2018-09-14 | Lvmh Swiss Manufactures Sa | MATERIAL COMPRISING A THIN LAYER OF AN ALLOY COMPRISING TITANIUM AND GOLD, AND PROCESS FOR OBTAINING SUCH MATERIAL |
| CN208008884U (en) * | 2018-03-27 | 2018-10-26 | 合肥科晶材料技术有限公司 | A kind of five target magnetic control sputtering plated film instrument for high-throughput coating process |
Non-Patent Citations (1)
| Title |
|---|
| K575DX溅射镀膜仪反应溅射制备ZnO薄膜的研究;付晶晶;《中国优秀硕士学位论文全文数据库(工程科技Ⅰ辑)》;20120815(第8期);第B020-138页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110262233A (en) | 2019-09-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110262233B (en) | Optimization method for technological parameters of magnetic control film plating instrument | |
| CN111340273B (en) | Short-term load prediction method for power system based on GEP parameter optimization XGboost | |
| CN111260030B (en) | A-TCN-based power load prediction method and device, computer equipment and storage medium | |
| Grimaldi et al. | PSO as an effective learning algorithm for neural network applications | |
| CN112581263A (en) | Credit evaluation method for optimizing generalized regression neural network based on wolf algorithm | |
| CN103105246A (en) | Greenhouse environment forecasting feedback method of back propagation (BP) neural network based on improvement of genetic algorithm | |
| Alfred | A genetic-based backpropagation neural network for forecasting in time-series data | |
| Liu et al. | A fault diagnosis intelligent algorithm based on improved BP neural network | |
| JP2005521167A (en) | Improving the performance of an artificial neural network model in the presence of mechanical noise and measurement errors | |
| CN110111606A (en) | A kind of vessel traffic flow prediction technique based on EEMD-IAGA-BP neural network | |
| CN116542382A (en) | Sewage treatment dissolved oxygen concentration prediction method based on mixed optimization algorithm | |
| Karlov et al. | Application of high-speed algorithms for training neural networks for forecasting financial markets | |
| CN114004008B (en) | Airplane assembly line resource configuration optimization method based on neural network and genetic algorithm | |
| CN111126560A (en) | Method for optimizing BP neural network based on cloud genetic algorithm | |
| CN112766548A (en) | Order completion time prediction method based on GASA-BP neural network | |
| CN117335425A (en) | Tidal current calculation method based on GA-BP neural network | |
| CN111415265A (en) | Social relationship data generation method of generative confrontation network | |
| CN111415051A (en) | Power material demand plan prediction method based on BP neural network | |
| CN115879412A (en) | Layout level circuit diagram size parameter optimization method based on transfer learning | |
| CN115345303A (en) | Convolutional neural network weight tuning method, device, storage medium and electronic equipment | |
| Taik et al. | Hybrid particle swarm and neural network approach for streamflow forecasting | |
| Mane et al. | Classification problem solving using multi-objective optimization approach and local search | |
| CN114202106A (en) | Air conditioning system load prediction method based on deep learning | |
| CN113837474A (en) | Regional soil heavy metal pollution index prediction method and device | |
| Zainun et al. | Forecasting low-cost housing demand in Pahang, Malaysia using Artificial Neural Networks |
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 |