JP2003281194A - Optimum design support method for device, optimum design support system for the device and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide optimum design support technique, enabling efficient, quick and highly precise optimization of a device. <P>SOLUTION: Using the test results on the device and the numerical value simulation result on the device, optimizing technique and data analysis method are used to determine a parameter for optimizing the device. Preferably one or more steps of obtaining the test result on the device using the value of the parameter for optimizing the device are further included in some case. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for supporting optimum design of a device. More specifically, the present invention relates to an optimal design support technology for an apparatus using an experimental result for the apparatus and a numerical simulation result for the apparatus.

[0002]

2. Description of the Related Art Conventionally, when it is desired to optimize the design of a certain device, the device itself (actual device), a similar actual device or a test device simulating an actual device (in the present specification, these are also collectively referred to as actual device) is used. A method based on an experiment or a method based on a numerical simulation based on the model of the response of the device has been adopted.

[0003]

The physical quantity required for design optimization may be obtained by experiments in an actual machine or the like, or may be obtained as an estimated value by calculation in a numerical simulation.

Since it is rare that all physical quantities required for design optimization can be obtained by experiments in an actual machine or the like, it is useful to calculate by numerical simulation.

On the contrary, there are some physical quantities that cannot be calculated by numerical simulation, and the method of acquiring the physical quantity by experiments in an actual machine is also valuable.

In this case, if the experiment and the numerical simulation in the actual machine were individually optimized,
It was difficult to find the optimal solution for the purpose.

There, it has been believed that a simple combination of the two does not give a meaningful result, and that experiential means such as filling the groove is essential.

Further, when the optimized parameters obtained from the experimental apparatus are applied to the actual apparatus, or when the optimized parameters of other similar actual apparatus are applied to another actual apparatus, the scale difference between the experimental apparatus and the actual apparatus is , Even if it is a similar real machine,
It was not uncommon that it took a lot of time to find the optimum conditions at the actual machine test site, such as the optimum conditions deviated due to the difference in operating environment and the individuality of each device, and the constraint conditions were deviated. .

In such a case, it is desirable to be able to perform online optimization while updating the obtained optimization model while reflecting the experimental results of the actual machine experiment site. On-line optimization was difficult with the method of performing simulation and simulation separately.

The present invention overcomes such problems,
It is an object of the present invention to provide a technique for organically linking experimental data with a numerical simulation data with an actual machine or the like to assist the optimum design of the device more efficiently, quickly, and more accurately than ever before.

Still another object and advantage of the present invention are:
It will be apparent from the following description.

[0012]

According to the present invention, optimization of an apparatus is performed by using an experimental result of the apparatus and a numerical simulation result of the apparatus and an optimization method and a data analysis method. There is provided a method for supporting optimum design of a device, which includes a step of determining a value of a parameter for achieving the above.

Specifically, the experimental data analysis device causes an experiment on the device to be conducted, and the experimental result is transmitted,
The numerical simulation device performs a numerical simulation on the device, transmits the obtained numerical simulation result, the optimum design support device receives the experimental result and the numerical simulation result, and the optimum design support device A method of determining the value of the parameter for optimizing the apparatus by using the experimental result and the numerical simulation result and using the optimization method and the data analysis method is preferable.

According to the present invention, further, an experimental data analysis device for transmitting an experimental result for the device, a numerical simulation device for performing a numerical simulation on the device, and transmitting the obtained numerical simulation result, and the experiment. The result and the numerical simulation result are received, and the value of the parameter for the optimization of the device is determined by using the experimental result and the numerical simulation result and using the optimization method and the data analysis method. An optimum design support system for an apparatus is provided that includes an optimum design support apparatus for performing the above, and a communication unit that connects the experimental data analysis apparatus, the numerical simulation apparatus, and the optimum design support apparatus.

According to the invention of the present application, the computer further causes the optimum design support device to receive the experimental result of the device transmitted from the experimental data analysis device,
Transmitted from the numerical simulation device, the numerical simulation result obtained by performing a numerical simulation for the device is received by the optimal design support device, the optimal design support device, using the experimental result and the numerical simulation result, A program is provided for using the optimization method and the data analysis method to determine the values of the parameters for optimizing the device.

According to the above-mentioned invention, the experimental data and the numerical simulation data by an actual machine or the like are organically linked to each other, so that the optimum design of the device can be performed more efficiently, speedily and with higher accuracy than the conventional one.

It is to be noted that a step of obtaining an experimental result for the apparatus by using the values of the parameters for the optimization of the apparatus, which are determined by using the optimization method and the data analysis method, is further included. It is preferable to include it more than once.

This is because the optimum design of the device can be performed more quickly and efficiently.

When applying such results to different devices of the same type, the values of the parameters for optimizing the device, which are determined by using the optimization method and the data analysis method. It is preferable to perform an experiment on an apparatus of the same type as that of a different apparatus using the above, obtain an experimental result, include this experimental result in an experimental result obtained previously, and execute the optimal design support method again. Often. This is because even if the devices are of the same type and look almost the same, the optimum conditions may differ greatly. In this case, the previously obtained experimental results or the like will be used as the experimental results or the like for different devices of the same type.

Further, a concrete target of the apparatus is plasma CVD (chemical vapor depos).
It is preferable that the experimental result includes data that cannot be obtained by the numerical simulation, and the numerical simulation result includes data that cannot be obtained by the experiment.

This is because the effect is remarkable in the case of this device.

Further features of the present invention will be clarified in the following embodiments and drawings of the invention.

[0023]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings, examples, formulas and the like. It should be noted that these drawings, examples, formulas, and explanations exemplify the present invention, and do not limit the scope of the present invention. It goes without saying that other embodiments may belong to the scope of the present invention as long as they match the gist of the present invention.

The optimization method used in the optimum design support technology according to the present invention can include the functions of experimental design, mathematical programming, and response surface approximation model creation. However, similar publicly known functions can be used as long as they do not violate the gist of the present invention. For example, JP-A-10-207926 and JP-A-20001-125.
Japanese Patent No. 933 discloses such a design support method.

Further, the data analysis method can include a function of analysis of variance. Even in this case, the same known function can be used as long as it does not violate the gist of the present invention.

The design of experiments was used to obtain the response data required for optimization and to infer the optimal parameters, the mathematical programming was used to obtain the optimal parameters, and the response surface approximation model was obtained. This is a mathematical model for obtaining an optimized solution from response data.

In the experimental design method, the degree of influence of the factor on the characteristic value is analyzed by using an orthogonal table (also called an orthogonal array table).

As the orthogonal table, L16 (2 ¹⁵ ), L18
(2 ¹ × 3 ⁷ ), L16 (4 ⁵ ) and other general examples,
As a multilevel example, L32 (2 ¹ × 4 ⁹ ), L64 (4 ²¹ ).
There are various factors such as L36 (2 ¹¹ × 3 ¹² ), L
54 (2 ¹ × 3 ²⁵ ) and the like are known. An appropriate orthogonal table is selected depending on the number of design factors, the setting of interaction, and the number of levels. It is necessary to determine the number of levels in consideration of the characteristic (trend) of the influence of the design factor on the characteristic value.

Mathematical programming is a technique for optimizing an objective function that depends on some decision variables under a set of constraints.

Although many methods are known for mathematical programming, the sequential quadratic programming (SQP) is an efficient method for nonlinear optimization problems with constraints, which occupy most of practical problems. Method: Sequential Quad
(ratic programming) can be given as an example.

In the SQP method, a quadratic programming problem that is successively approximated to the optimization problem is created as a subproblem, and the quadratic programming problem is repeatedly solved.

Also. In a multimodal problem in which a plurality of local solutions exist, for example, GA (genetic algorithm) or SA (simulated annealing) capable of searching for a global optimum solution can be used.

The response surface approximation model is a mathematical model for obtaining an optimized solution from the obtained response data, and can be created using, for example, an orthogonal polynomial which is a Chebyshev orthogonal function.

The analysis of variance is used, for example, in order to know the degree of contribution of the parameter to the objective function, the positive / negative correlation and the like like the correlation analysis.

In the analysis of variance, the degree of influence of the design factor on the characteristic value is orthogonally decomposed into polynomial components and evaluated.

The degree of influence can be obtained for the characteristics (propensity) decomposed into primary or secondary.

For example, the variation S _i of each design factor and the order component i is first obtained using the obtained characteristic value data, and the variation V _i is obtained by dividing the variation by the degree of freedom n _i .

The difference between the total variation S _T and the sum ΣS _i of the variations S _i of each design factor / order component becomes the variation of the error term, which is divided by the degree of freedom n _e of the error to obtain the error variance.

Next, the variance of the design factor / order component is divided by the error variance to obtain the F value.

The evaluation of the analysis is performed by, for example, the risk factor 1 of the F distribution.
The F value of each design factor / order component is compared with the values of 5% and 5%, and the test of significant difference can be performed.

When the value of the parameter for optimizing the device is determined using the optimization method and the data analysis method using the experimental result of the device and the numerical simulation result of the device, It has been found that the optimum design of is much easier.

On the other hand, conventionally, it was general to repeat the trial-and-error experiment to reflect the conditions satisfying the experimenter in the design of the actual machine.

Even when the optimization is performed, if the numerical simulation reveals that a certain condition is good for optimizing a certain characteristic value, the condition is different from that predicted from the experimental result. In some cases, it was far from the purpose of optimizing the characteristic value of.

In such a case, after all, it was always the case that the numerical simulation was used only as a reference, and various experiments were accumulated and optimized based on experience.

FIG. 1 is a diagram for explaining a conventional optimum design support technique, and FIG. 2 is a flow chart showing a flow of the conventional optimum design support technique.

In the following, referring to FIGS. 1 and 2, first, a mathematical model of the apparatus for numerical simulation is performed according to step S1, and numerical simulation is performed using the numerical simulation apparatus 2 according to step S2. Perform a simulation.

Then, according to step S3, the optimization process is performed by the optimum design support device 3 using the calculation result of the numerical simulation. In the optimization processing, it is possible to use the optimization method used in the above-described optimum design support technology.

A personal computer is often used as the numerical simulation device 2 and the optimum design support device 3, and a single personal computer can also serve as the personal computer.

Then, according to step S4, it is confirmed whether or not the objective function converges as a result of the optimization process by the optimal design support device 3, and if it does not converge (N), step S2
Then, the conditions of the numerical simulation are changed based on the calculation result of the optimization process, the numerical simulation is executed again, and then the calculation result of the numerical simulation is used according to step S3 to perform optimization by the optimum design support device 3. Perform processing.

If it converges (Y), an experiment with an actual machine or the like is conducted to confirm the optimization effect.

That is, in the conventional method, the experimental result was not used for the optimization process.

The adjustment between the experimental result and the numerical simulation result is performed through mathematical modeling work and parameter adjustment, but the actual situation is that the work is performed by a person who needs experience.

However, if the experiment result of the apparatus and the numerical simulation result of the apparatus are used together or appropriately selected and used, the optimization method and the data analysis method are used to optimize the apparatus. It has been shown that the device can be optimally designed with higher efficiency, speed, and accuracy than ever before.

That is, by applying only the numerical simulation result to the optimum design support technology without using the experimental result, the experimental result and the numerical simulation result are directly applied to the optimum design support technology. The effect of the present invention has come to be obtained.

In this case, the step in which the experimental data analysis device performs an experiment on the device and sends the experimental result, and the numerical simulation device performs the numerical simulation for the device and sends the obtained numerical simulation result. And a step in which the optimum design support device receives the experimental result and the numerical simulation result, and the optimum design support device uses the experimental result and the numerical simulation result to perform an optimization method and data analysis. Using the method and the step of determining the value of the parameter for achieving the optimization of the apparatus, it is possible to perform the optimal design of the apparatus more quickly and efficiently with less labor. Become.

In particular, if the experimental data analysis device, the numerical simulation device, and the optimum design support device are connected by communication means, the device can be designed more quickly and efficiently, and You can modify the data to get better results in less time.

Here, the experimental apparatus includes an apparatus having a size different from that of the actual apparatus and an apparatus used as an actual apparatus for other purposes only partially different from the actual apparatus. If necessary, an apparatus capable of conducting an experiment on the apparatus belongs to the category of the experimental apparatus referred to in the present invention.

Further, in the above, it is necessary to select the parameters prior to the experiment or the numerical simulation, but this action may be performed by a person. You can also select historical data below.

When a parameter is used, it may be a part or all of the parameter. This is because it is not always necessary to use all the selected parameters for experiments and numerical simulations.

Further, the values of the above-mentioned parameters may be in a certain range.

The experimental data analysis device, the numerical simulation device, and the optimum design support device are respectively
As long as it has an experimental data analysis function, a numerical simulation function, and a function for executing the optimization method and the data analysis method described above, any function can be used.
It is practical to use a personal computer.
A single personal computer can also serve as multiple devices.

These devices should have a data communication interface for the measurement data communication of the device under consideration.

The data communication interface is an interface for taking measurement data into a computer.
It is equipped with standard bus interfaces such as GP-IB, Ethernet (trademark), PCI, and serial, and mediates data acquisition and data transmission.

As the communication means, any known means including a network can be used as long as it does not violate the gist of the present invention.

However, when using the Internet, encryption processing should be performed to prevent leakage of secrets. In this respect, it can be said that an intranet separated from other networks is preferable.

FIGS. 3 and 8 are diagrams for explaining the optimum design support technique according to the present invention, and FIG. 4 is a flow chart showing the flow.

This will be described below with reference to FIGS. 3, 4 and 8. First, in accordance with step S101, the actual device 1 (FIG.
Represents the CVD apparatus) and obtains the experimental results.

Further, according to step S102, a mathematical modeling work of the apparatus is performed, and a numerical simulation is performed using the numerical simulation apparatus 2.

The experiment result and the numerical simulation result are provided to the optimization process performed by using the optimum design support device 3 according to step S103. This situation is briefly shown in FIG.

In the optimization process, the optimization method used in the above-mentioned optimum design support technique can be used.

Then, according to step S104, it is confirmed whether or not the objective function converges as a result of the optimization processing by the optimal design support device 3. If the objective function does not converge (N), the process returns to step S102 and the numerical simulation device 2 is operated. As shown in FIG. 8, a numerical simulation is performed using the simulation conditions based on the calculation result of the optimization process, and then the optimization process is performed using the optimum design support device 3 according to S103. .

The experimental result and the numerical simulation result are also used for the optimum processing in this case. Normally, the numerical simulation result is a newly obtained one, and the experimental result is the previously used one. It will be used as is.

If it converges (Y), the process may be terminated here.

Further, when it is desired to confirm the actual device,
According to step S105, as shown in FIG. 8, an experiment with an actual machine or the like is performed to confirm the optimization effect.

As a result, according to step S106, for example, when a problem such as deviation from the experimental constraint condition occurs (N), optimization is performed again using the optimum design support device 3 according to step S103. Return to processing. Then, when no inconvenience such as deviation from the experimental constraint condition occurs (Y), the process ends.

Although the experimental result and the numerical simulation result are also used for the optimum processing in this case, the experimental result is usually the newly obtained one, and the latest numerical simulation result is used. Will be used.

As exemplified in this, both the experimental result and the numerical simulation result may be used for the optimization process later, so that the optimum design support apparatus
It is desirable to have the function of storing that information.

As shown in FIG. 3, if the experimental data analysis device 4, the numerical simulation device 2, and the optimum design support device 3 are connected to each other via, for example, the intranet 5, the above steps are used for the experiment and the numerical simulation of the device. Can be used to the maximum extent to perform the optimization process.

Even if the experimental device 1, the numerical simulation device 2, and the optimum design support device 3 are separated from each other,
Optimum device design can be performed quickly and efficiently.

Further, the above will be described by taking an electronic device manufacturing using plasma CVD as an example, for example, as follows. In the following, the symbols in the sentence or in the parentheses at the end of the sentence mean the step symbols in FIG.

The objective function for optimization is determined. For example, maximizing the electrical conductivity of the device and minimizing the particle density in the plasma CVD vessel.

As parameters to be used next, for example, gas flow rate, gas pressure, power, and substrate temperature are selected.

At this time, the above-mentioned electrical conductivity data is measurement data that cannot be obtained by numerical simulation and can be obtained only as an experimental result by an actual machine, and the particle density data cannot be obtained by an experiment. It is a physical quantity obtained only from numerical simulations. In such a case, the effect of the present invention is particularly great.

The electric conductivity data is the experimental result according to the present invention, and the particle density data is the numerical simulation calculation result (see FIG. 8).

First, based on the experimental design method used as the optimum design support technique, an experiment (S101) and a numerical simulation (S102) by an actual machine or the like are performed, and an experiment on the above objective function when the parameters are changed is performed. The results and the numerical simulation calculation results are collected.

The electric conductivity data and the fine particle density data obtained from each are centrally managed in a database together with the parameters.

Based on this data, the optimum design process (S1
It is also possible to check the correlation between parameters and predict the optimum parameter based on the analysis of variance as the technique of (03). The most powerful method is to create a response surface approximation model.

For this response surface approximation model, a mathematical programming method such as a nonlinear programming method or an annealing method (Simula) is used.
Ted Annealing (SA), etc.,
Optimal parameters can be obtained under constraint conditions.

Generally, in order to obtain the optimum parameters by using the mathematical programming, it is necessary to obtain the values of the objective function for a large number of parameter values, and it is often difficult to carry out this by numerical simulation or experiment.

However, if there is a response surface approximation model, it may be possible to quickly obtain optimum parameters without conducting numerical simulations or experiments.

Further, if there is a response test surface approximation model of a simulated test apparatus or similar real machine, it is possible to utilize it for optimization. If there is a problem with the model accuracy, it is possible to collect the actual machine data and update the approximate model for optimization.

In the above, the response surface approximation model (Response Surface Method) is used.
is the n predictor variables x _i (i =
1, ..., N) is an approximation of the relational expression of the response y predicted, and is given below.

Y = f (x ₁ , ..., X _n ) + ε Here, ε is called an error. In the design / manufacturing field, x
_i is often used as a design and operation variable, and y is often used as a characteristic value of a product.

In the response surface method, any function shape may be used for the function form f, but if a linear function or a linearizable function is used, it is easy to apply the least squares method to The coefficient can be statistically estimated. Therefore, the most commonly used functional form is a polynomial. That is, usually, a plurality of experiments or numerical simulations are performed to find the relationship between x _i and y, and solve the problem that the square of the error ε is minimized under the assumed functional form (least squares method), The coefficient of each term can be obtained.

Further, the annealing method is a method modeled from the physical phenomenon of metal annealing, and its characteristic is that while looking for a design variable that improves the objective function, the solution may be deteriorated with an appropriate probability. Is also forgiven. This method can be applied to a multimodal problem, which is a problem where there are multiple local optimal solutions. Since the solution space can be searched globally, there is less risk of falling into local optimal solutions and it is easy to find global optimal solutions. There are features.

When performing an experiment or a numerical simulation in accordance with step S104 of FIG. 4, the optimum parameters can be transmitted as a control signal to an actual machine or the like by a communication means such as an intranet and can be used for an optimization experiment of the apparatus. It can also be used for numerical simulation.

Unless the conditions of the apparatus are changed, it is useful to carry out an experiment on the actual machine without performing a numerical simulation because the amount of work can be reduced, and the experiment on the actual machine gives highly reliable data. Often there is.

In this way, in many cases, the optimum results can be obtained by re-optimizing the obtained results of the experiment of the actual machine and repeating the optimization process and the experiment of the actual machine.

When the optimization is carried out at the experimental site such as the actual machine or at the place where the optimum design support device is present, the above-mentioned optimization processing or the experiment of the actual machine can be remotely operated through the communication means. .

In this way, since the experimental device, the numerical simulation device, and the optimum design support device, which are located apart from each other, can be used through the communication means, the numerical simulation device and the optimum design support device can be connected to each other. It can be used efficiently for the purpose of optimizing the device while being used for the purpose.

The calculation programs used in the numerical simulation device and the optimum design support device are often sophisticated and special, and often exist only in a specific department.

Therefore, not only in the case of being remote from each other, but in the case of being close to each other,
In many cases, the effects of the present invention can be enjoyed.

The optimization target apparatus to which the invention of the present application can be applied may be any apparatus as long as the object of the invention of the present application is satisfied. Other than the following examples, a ship, a high-speed boat, a vehicle body, an aircraft, an airship, an artificial satellite, a wave generator, a riser pipe guiding device, a thermal power plant, and a nuclear power plant can be exemplified.

Data may be obtained by both the experiment and the numerical simulation. In such a case, which data should be selected and which data should be used in combination is determined by the difficulty level of each data acquisition. However, it is practical to decide in consideration of accuracy.

[0105]

EXAMPLES Next, examples of the present invention will be described in detail.

[Embodiment 1] This is a concrete example of assistance for optimizing process conditions in a thin film forming process using plasma CVD.

As the process conditions of the plasma CVD apparatus, five parameters of electric power, power supply frequency, gas pressure, gas flow rate, gas flow rate ratio (mixing ratio of two kinds of gas) were used as parameters.

The objective function is a multi-objective function for maximizing the film formation rate and minimizing the plasma potential.

Generally, in the case of a multi-objective function, a weighting coefficient is assigned to each of them and a one-objective function obtained by linear combination is used.

Here, f = −1.0 × deposition rate + 1.0 × plasma potential (2), and it is defined as a minimization problem as a whole, that is, a problem of minimizing f.

The film formation rate is an actual measurement value of an experiment using an actual model machine, and the plasma potential is a value obtained as a calculation result of plasma CVD numerical simulation.

First, the experimental design method was executed to grasp the correlation between each process condition and the objective function, and to prepare the response surface approximation model.

In the design of experiments method, an orthogonal table was used, and L18 having three levels of each parameter was applied.

1 assigned based on the orthogonal table of L18
Eighteen film forming experiments and numerical simulations were carried out based on combinations of eight parameters, and the values of the film forming rate and the plasma potential were recorded as data together with the parameters in a computer.

Based on the obtained data, analysis of variance was carried out to examine main effects and contribution rates.

FIG. 5 is a main effect graph. In the main effect graph, the parameters other than the individual parameters of interest are fixed at average values, and the characteristic values when the parameters change, that is, the changes in the film formation rate are shown here. It is possible to see how the film formation rate is roughly related to each parameter.

Here, the vertical axis of the main effect graph of FIG. 5 is the film formation rate, and the horizontal axis is each parameter.

FIG. 6 is a contribution rate graph. This is the percentage of each parameter that contributes to the change in the characteristic value, here, the deposition rate. The darker one has a positive correlation and the lighter one has a negative correlation. This makes it possible to know which parameter has an effect on the film formation rate.

In FIG. 6, the vertical axis represents the factors such as each parameter, the squared value of the parameter, and the interaction between the two parameters, and the horizontal axis represents the contribution rate of these factors.

On the other hand, in a quadratic polynomial model with five parameters, there are a maximum of 22 coefficient terms.

Therefore, from the result of analysis of variance, only significant variables were selected by using the t-test, and 13 coefficient terms were narrowed down.

As a result, the response surface approximation model obtained by the quadratic polynomial model had a correlation coefficient of 0.9986 and a degree of freedom adjusted correlation coefficient of 0.9952.

Surfac using the obtained response surface approximation model of the objective function with the power and gas pressure as parameters
The e display is shown in FIG. In FIG. 7, Press (To
The axis labeled rr) represents the gas pressure, PW (W)
The axis marked with represents electric power, and the axis marked obj represents the objective function of equation (2).

Using this response surface approximation model, a sequential quadratic programming method, which is a kind of mathematical programming method, was applied to solve the problem of minimizing the above objective function (2).

At this time, the obtained parameter values are as follows: power = 40 W frequency = 84.26 MHz mixed gas pressure = 277.7 Pa mixed gas flow rate = 311.7 cm ³ / min (where 0
C., 1 atm conversion value) Gas flow rate ratio = 1.5 / 98.5 (using two kinds of gas, hydrogen gas and monosilane gas). The film forming rate at this time was 2 nm / s and the plasma potential was 25.1 V.

That is, this film forming rate and the plasma potential are the best combinations that can be expected, and the above parameter values are the optimum parameter values.

After that, an experiment was carried out using an actual model machine using these parameter values, and it was confirmed that there was no inconvenience such as deviation from the constraint conditions.

[0128]

[Comparative Example 1] Using only the numerical simulation result regarding the plasma potential, without using the experimental result regarding the film formation rate, and setting f = 1.0 × plasma potential (2) ′, the maximization problem as a whole, that is, f Is defined as the problem of maximizing.

At this time, the obtained parameter values are: power = 10 W frequency = 72.16 MHz mixed gas pressure = 320.8 Pa mixed gas flow rate = 589.8 cm ³ / min (however, 0
C., 1 atm conversion value) The gas flow rate ratio was 1.5 / 98.5. The plasma potential at this time is 15.4.
It was V.

The film forming rate obtained by applying the above parameter values to the experimental results was 0.6 nm / s.

The plasma potential value and the film formation rate value were an insufficient combination in view of the conventional accumulated data.

[0132]

[Embodiment 2] Next, as a method of using the response surface approximation model, if a response surface approximation is obtained as shown in FIG. It will be explained that the response parameter approximation can be used for optimal parameter determination support without any numerical simulation without changing the device conditions.

In the case of the plasma CVD apparatus, when the optimized parameters obtained from the experimental apparatus are applied to the actual apparatus, or when the optimized parameters of other similar actual apparatus are applied to another actual apparatus, the experimental apparatus and the actual apparatus are combined. Even in the case of different scales and similar actual machines, the optimum conditions are often shifted due to the difference in use environment and the individuality of each apparatus, and the apparatus characteristics are often slightly different. Therefore, the previously determined optimum process conditions may deviate.

Therefore, in the case of applying the same kind to different actual machines, the optimum process conditions obtained in the first embodiment are not directly adopted in the design, but the actual machine experiments are performed using the optimum process conditions obtained in the first embodiment. It is often useful to carry out and measure the film formation rate.

In such a case, using the values of the parameters for optimizing the apparatus, which are determined by using the optimization method and the data analysis method for the previous apparatus described above, It may be preferable to perform an experiment on a new device, obtain an experiment result, and execute the same optimal design support method again including this experiment result.

Here, whether or not the same type is different is
It can be arbitrarily determined by experience or the like. If they are far apart, it can be judged that the obtained result is abnormal.

For example, similar devices belong to such cases, such as slightly different specifications, slightly different spatial arrangements of parts, and slightly different dimensional relationships of the devices. Here, “the same kind and different” typically means a device originally made for the same purpose or a device made at a different scale (generally small in general) from an actual device.

In this example, when applying to different actual machines of the same type, the optimum process conditions obtained in Example 1 were used to perform experiments on different actual machines of the same type to measure the film forming rate. To do.

If, as a result of the experiment, there are inconveniences that deviate from the constraint conditions, such as a low value of the film forming rate, a new response surface approximation model is recreated by reflecting the experiment result.
Optimization calculation can be performed based on this new response surface approximation model.

Based on the process conditions obtained as a result,
Further experiments are carried out, and if the above-mentioned inconvenience is eliminated, the above-mentioned work is terminated, and the finally obtained process condition becomes the optimum condition.

When this method is adopted, when the optimized parameters obtained from the experimental apparatus are applied to the actual apparatus or when the optimized parameters of other similar actual apparatus are applied to another actual apparatus, the experimental apparatus and the actual apparatus are Even in the case of similar scales, even in the case of similar actual machines, the optimum conditions may deviate from the constraints due to the difference in operating environment and the individuality of each device. It is possible to prevent a lot of time from being taken out.

Even in the case of the first embodiment, the result is reflected in the experiment using an actual machine, the experimental result is reflected in the response surface approximation model, and a new response surface approximation model is recreated. Optimization calculation is performed based on the response surface approximation model, further experiments are performed based on the process conditions obtained as a result, the film deposition rate is measured again, and there are inconveniences such as deviating from the constraint conditions. If it is not, it goes without saying that it may be useful to finish the above work and make the finally obtained process condition the optimum condition.

[Embodiment 3] An example of a position holding system for an offshore platform will be described.

FIG. 9 is a diagram for explaining the optimum design support technology according to the present invention for the position holding system for an offshore platform.

The offshore platform is required to hold its own position using a plurality of thrusters against external forces such as ocean currents, waves and winds.

Offshore platform (corresponds to actual machine 1)
The measurement data obtained in (1) are the position of the offshore platform by GPS, the yaw angle by the gyro, and the wind direction and speed by the wind direction / speed sensor.

In the Kalman filter model of the offshore platform in the numerical simulation device 2, the tidal force, the wave force and the unknown disturbance force are obtained by calculation based on the model based on the above measurement data, and the numerical value is returned to the optimization support system. .

Therefore, the optimum design support device 3 performs PID control and optimum distribution to each thruster based on the above measurement data and calculation results, and outputs the thruster output and the swing angle control signal to the thruster of the offshore platform. send.

In this way, the optimum design support device is used to optimize the device by using the experimental result of the device and the numerical simulation result of the device.

[Embodiment 4] An example of a motion control system for a submersible will be described.

FIG. 10 is a diagram for explaining an optimum design support technique according to the present invention for a motion control system of a submersible. The submersible is required to move according to the instructions of the mother ship (set depth, route, movement angle).

The measurement data obtained by the diving machine (corresponding to the actual machine 1) are the speed and depth of the diving machine, the pitch angle rate, the yaw angle rate, and the roll angle rate.

In the motion model of the submersible in the numerical simulation device 2, the submersible position,
The pitch angle, yaw angle, and roll angle are calculated, and the numerical values are returned to the optimization support system.

Therefore, the optimum design support device 3 performs PID control and optimum distribution to each thruster based on a function that combines measurement data and calculation data, and sends a control signal for thruster output to the submersible.

In this way, the optimum design support device is used to optimize the device by using the experimental result of the device and the numerical simulation result of the device.

[0156]

As described above, the apparatus can be optimized more efficiently, quickly, and with higher accuracy.

[Brief description of drawings]

FIG. 1 is a diagram illustrating a conventional optimum design support technique.

FIG. 2 is a flowchart showing a flow of a conventional optimum design support technique.

FIG. 3 is a diagram illustrating an optimum design support system according to the present invention.

FIG. 4 is a flowchart showing a flow of an optimum design support technique according to the present invention.

FIG. 5 shows an example of a main effect graph in analysis of variance.

FIG. 6 shows an example of a contribution rate graph in analysis of variance.

FIG. 7 shows an example of a response surface approximation model of an objective function.

FIG. 8 is a diagram illustrating an optimum design support technique according to the present invention.

FIG. 9 is a diagram illustrating an optimum design support technique according to the present invention for a position holding system for an offshore platform.

FIG. 10 is a diagram illustrating an optimum design support technique according to the present invention for a motion control system of a submersible.

[Explanation of symbols]

1 real machine 2 Numerical simulation device 3 Optimal design support device 4 Experimental data analyzer 5 Intranet

Claims (7)

[Claims] 1. A value of a parameter for optimizing the device is determined by using an experimental result of the device and a numerical simulation result of the device and using an optimization method and a data analysis method. An optimum design support method for an apparatus including steps. 2. A step of causing an experimental data analysis device to perform an experiment on the device and transmitting the experimental result, and a numerical simulation device performing a numerical simulation on the device and transmitting the obtained numerical simulation result. And a step in which the optimum design support device receives the experimental result and the numerical simulation result, and the optimum design support device uses the experimental result and the numerical simulation result to perform an optimization method and a data analysis method. And a step of determining the value of a parameter for optimizing the device, using the method and. 3. A step of obtaining an experimental result for the device by using a value of a parameter for the optimization of the device, which is determined by using the optimization method and the data analysis method. The optimal design support method for an apparatus according to claim 1 or 2, which is included once or more. 4. An experiment is performed on a device of the same kind as the device but different from the device, using the values of the parameters for optimizing the device, which are determined by using the optimization method and the data analysis method. An optimum design support method for an apparatus, which performs the optimum test support method according to any one of claims 1 to 3 again, including the experimental result. 5. The apparatus is an electronic device manufacturing apparatus using plasma CVD, the experimental result includes data that cannot be obtained by numerical simulation, and the numerical simulation result includes data that cannot be obtained by experiment. An optimum design support method for an apparatus according to any one of 4 to 4. 6. An experimental data analysis device for transmitting an experimental result of the device, a numerical simulation device for performing a numerical simulation for the device, and transmitting the obtained numerical simulation result, the experimental result and the numerical simulation. An optimal design support device that receives a result and uses the experimental result and the numerical simulation result to determine the value of a parameter for optimizing the device by using an optimization method and a data analysis method. And a communication means that connects the experimental data analysis device, the numerical simulation device, and the optimum design support device,
An optimum design support system for an apparatus for executing the optimum design support method for an apparatus according to claim 1. 7. A numerical value obtained by performing a numerical simulation on the device, which causes the optimum design support device to receive the experimental result about the device transmitted from the experimental data analysis device by the computer and transmitted from the numerical simulation device. The optimum design support device is made to receive the simulation result, and the optimum design support device uses the experimental result and the numerical simulation result to optimize the device by using the optimization method and the data analysis method. A program for executing the optimum design support method for an apparatus according to any one of claims 1 to 5, which determines the value of a parameter for the purpose.