KR20140096924A - The design method to optimize an impeller - Google Patents

Abstract

The present invention relates to a method for optimally designing an impeller shape and, more particularly, to a method for optimally designing an impeller shape using artificial intelligence with a method to obtain a global result and a solution quickly. The purpose of the present invention is to provide a method for optically designing the impeller shape capable of obtaining global improvement by setting a target function obtained by connecting a pressure ratio to efficiency receiving and changing a flow path and blade to improve the pressure ratio and the efficiency at an impeller design point.

Description

TECHNICAL FIELD [0001] The present invention relates to a method of optimizing an impeller shape,

The present invention relates to a method of optimizing an impeller shape, and more particularly, to an impeller shape optimal design method using an artificial intelligence in a way that can achieve a global result and solve it in a short time.

Centrifugal compressors are widely applied to relatively small engines and turbochargers because they exhibit good operating performance in low-speed regions. Compared with an axial flow type compressor, the compressor has a high compression ratio and a wide variation range of the air flow rate compared with the axial flow type at the same rotation speed.

The centrifugal compressor consists of an impeller and a diffuser. The rotational motion of the impeller applies kinetic energy to the intake air and the kinetic energy in the diffuser is converted into pressure. The high pressure air from the diffuser is used to deliver high pressure air to other devices through scrolling. If there is a large change in operating conditions, the diffuser will not be equipped with a vane.

A study on impeller design that directly affects compressor efficiency and operational stability has been conducted by many researchers for a long time. The design restrictions and design techniques are described in detail in the references. However, the reason why the shape of the impeller being used is different from that of the same type is that the shape design is performed by the experience and intuition of the designer in most designs, but there are also many influences of the adjustable design parameters in the impeller design.

In designing the impeller, it is designed to reach the design point, but ultimately it is aimed at obtaining a high efficiency impeller. Therefore, in order to increase the efficiency, the flow path and the blade must be formed to reduce the loss in the impeller. For this, a method of designing a shape by evaluating the influence of changes of these variables by selecting parameters related to the shape has been studied.

In this process, various optimization techniques are used. The characteristics of these studies are that the influence of the output due to the change of the shape parameters in the impeller does not change linearly, so the algorithm that can detect the global change is applied.

Such a designing method has a problem that depends on the experience and sense of the designer, and there is a problem that it is necessary to perform a rule analysis or a performance test that is designed as a complementary means.

In addition, the development of a new impeller due to a separate supplementary means is costly, and there is a problem in compatibility because the impeller for each maker can not be produced.

In addition, there is no reference point for improving the performance in the existing method, and the designer's experience becomes an important factor, and it takes a lot of time to develop, which makes it difficult to cope with the change.

Korean Patent Laid-Open Publication No. 10-2012-0013501 United States Patent Application Publication No. US 7047167

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above-mentioned problems, and an object of the present invention is to provide a method and apparatus for improving the pressure ratio and efficiency at an impeller design point by setting an objective function that relates pressure ratio and efficiency, In the impeller shape optimization method.

In order to search the objective function globally and shorten the time for it, an impeller shape using a combination of application of PSO (Particle Swarm Optimization) technique and genetic algorithm (ANN: Artificial Neural net) And provides an optimal design method.

The solution to the problem of the present invention is not limited to those mentioned above, and other solutions not mentioned can be clearly understood by those skilled in the art from the following description.

In order to achieve the above object, the present invention provides a design method for shortening a calculation time while obtaining global results for an impeller shape design, comprising the steps of: S1 obtaining design variables related to an objective function; Setting an area for forming an artificial intelligence network as an experimental design method in the design parameters obtained in step S1; Forming an artificial intelligence network based on the area set in step S2; A step S4 of applying a genetic algorithm to improve the artificial intelligence network and increasing the generation of the genetic algorithm to an optimal artificial intelligence network; And a step S5 of deriving a result of global optimization using the artificial intelligence network formed in step S4.

In step S2, the artificial intelligence network applies a back propagation algorithm, and the hidden layer and the output layer neurons are 10 and 2, respectively.

In step S2, the experimental design method is characterized by applying a partial factor arrangement (Fractional Factorial Design) and applying a resolution VI (resolution VI) arrangement method.

In addition, the experimental design method is used only in the initial artificial intelligent network forming step, and only the artificial intelligent network formed in step S4 is used for global results.

In step S4, PSO (Particle Swarm Optimization), which is a non-gradient based method, is applied to improve the artificial intelligence network.

Further, in the step S4, the artificial intelligence network is supplemented by numerically analyzing the results obtained using the artificial intelligence network in a new area obtained by the genetic algorithm.

The method of optimizing the shape of the impeller according to the present invention can eliminate the designer's experience and sense, and can directly utilize the design parameters directly affecting the performance.

In addition, since the numerical analysis method is used as a complementary means, it is advantageous from the viewpoint of cost, and it is easy to grasp the factors influencing the performance, so that it is possible to provide quick feedback to the needs of the developer in designing.

In addition, engineers who are inexperienced have the advantage of being capable of design work, and there is no need to spend a lot of computation time to obtain global results.

The effects of the present invention are not limited to those mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the following description.

Fig. 1 shows the shape of an impeller designed according to the conventional design method.
Figure 2 shows the initial shape of the hub and shroud.
Figure 3 shows the distribution of the blade airfoil angles of the initial model.
4 shows the blade thickness distribution.
FIG. 5 shows a flowchart of the optimum design method.
FIG. 6 shows a minimum objective function value for each generation.
7 shows the shape of an optimized compressor designed by applying the optimum design method of the present invention.
Figure 8 shows the relationship between efficiency and pressure ratio for the entire set of design variables.
Figure 9 shows the results for stress and displacement in the shape of the compressor optimized in Figure 7.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of an impeller shape optimum designing method according to the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to designate the same or similar components throughout the drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

In this detailed description, the optimum design method of the shape of the impeller of the centrifugal compressor is described. However, the present invention can be applied not only to the centrifugal compressor but also to the optimum design of the impeller shape of the centrifugal and axial compressor, centrifugal and axial turbine.

FIG. 1 shows the shape of the impeller being used because it is designed in the conventional designing method, and the outer diameter of the impeller is 52.4 mm. The diameters at the shroud 20 and hub 10 at the inlet are 37.5 mm and 11.4 mm, respectively. Table 1 shows the design point specifications of the compressor equipped with this impeller. However, the results in the compressor include bearing losses and losses in scrolling. Therefore, it was obtained numerically at the same mass flow rate and number of revolutions to obtain specifications on the impeller.

<Table 1>

CFX-13, a commercial program for the analysis of three-dimensional compressive turbulent flow, was used for the numerical analysis, and the κ-ε model was applied to the turbulence model. The high-resolution method based on the second-order upwind method is applied to the convection term. The calculation area is divided into three areas, namely, inlet area, impeller area, and outlet area. The computation domain was performed in one channel, and the boundary conditions in the direction of rotation in the computation domain were given as periodic conditions. At the inlet, the conditions of the voltage force and the mass flow rate at the outlet were given. Tests were performed on the number of gratings and 390,000 samples were applied. Compared with the Eckardt impeller experimental results, the results were fairly well predicted. In the same manner, the pressure ratio and isentropic efficiency were 2.46 and 86.1% based on the impeller inlet and diffuser outlet, respectively, under the same operating conditions as in Table 1 for the impeller using the optimum design method according to the present invention.

In order to improve the performance of the impeller, it is necessary to select appropriate shape parameters to adjust the shape of the impeller. There are various methods of expressing the shape of the impeller, but in the present invention, a Bezier curve is used. In this case, the adjustment point of the Bezier curve can be selected as a design variable. In the present invention, five adjustment points are used in order to minimize the number of design variables. In the present invention, five adjustment points are used to minimize the number of design variables.

Fig. 2 shows the initial shape of the hub and shroud, and shows the area that changes in each shape when only the middle adjustment point is selected as the design variable among the five adjustment points. However, since the selected adjustment point can be moved in the Z-direction (axial direction) and the R-direction (radial direction), two design variables (φ1, φ2) are obtained from one adjustment point Pt1. Therefore, four design parameters were selected to adjust the shape of hub and shroud.

Movement of the selected adjustment point by the design variable should be limited. Otherwise, the problem of not being assembled to the casing and the problem that the area change in the internal flow path is not uniform is caused. Table 2 shows the adjustment points and constraints used in the optimization process in the hub and shroud. The change in the area shown in FIG. 2 is a result obtained when the object moves within the constraint.

<Table 2>

In order to express the 3D blade, the blade airfoil angle (β) is expressed by equation (1) using the angle (θ) of the camber line rotation and the curve length ( m ) of the airfoil angle on the meridional plane.

식(1)

Equation (1)

Therefore, the Bezier curves were used to show the change of the airfoil angle along the camber line in the hub and shroud, and the adjustment points of the Bezier curve were each 5. FIG. 3 shows the distribution of the blade airfoil angles of the initial model.

In the distribution of the airfoil angle, the adjustment points at the inlet, the middle point and the outlet are set as design points (φ7, φ8, φ9) in the shroud, and the variation range of the airfoil angle is widened. Point (phi 5, phi 6). Therefore, a total of nine design variables are selected in the present invention. Table 3 shows the adjustment points and constraints of the airfoil angle. The airfoil angle is defined as the airfoil shape.

<Table 3>

Even if the design range of the design point is set, the airfoil angle at the impeller is structurally seriously distorted even within the limiting condition, so that the lean angle at the outlet is designed in the hub to be obtained as the initial shape Adjustment points other than points were automatically changed. In addition, the thickness of the airfoil is the same as that of the initial model, and the thickness distribution is shown in FIG.

In the optimization of the centrifugal compressor, the efficiency or the pressure ratio, which is usually applied as an objective function, does not show a linear result with respect to the change of design variables, so a global method should be selected. However, the global optimization method requires a lot of computation time. Therefore, it is important to obtain accurate results while shortening the calculation time. In the present invention, artificial intelligence is basically used as a method for optimization. Artificial intelligence applied a back propagation algorithm and 10 and 2 neurons in the hidden layer and the output layer, respectively. The transfer function used the hyperbolic tangent function in the hidden layer and the linear function in the output layer. Weights and biases were obtained by applying the Levernberg-Marquardt algorithm in order to make the artificial intelligence network coincide with the given results, and the artificial intelligence network was formed using MATLAB's nntool.

The input of the artificial intelligence network to form the initial artificial intelligence network was obtained by applying the experimental design method from nine design variables. This is because the application of the experimental design method is an effective way of knowing the information on the main effects or interactions of design variables and reducing the number of experiments. In the present invention, Fractional Factorial Design was applied and the resolution VI method was applied. Therefore, if the number of center points is included, a space for a total of 65 design parameter sets is formed. Therefore, numerical analysis results for 65 design parameter sets are needed. Experimental design can be used to optimize. However, when the result is nonlinear, it is applied only to the initial artificial intelligence network formation because there may be a difference between the final result obtained by the experimental design method and the actual result.

Once the initial artificial intelligence network is formed, an objective function can be found for a new set of design variables. To find the optimal objective function, a set of effective design variables must be applied in a set of infinite design parameters.

For this, PSO (Particle Swarm Optimization), which is a non-slope based method, is applied. In the PSO application, the number of each generation is set to 100. In order to find the next generation, the initial weight variable is set to 1.4 and the confidence variables C1 and C2 are set to 1.5 and 2.5.

The set of design parameters obtained from the PSO is applied to the artificial intelligence network to find the objective function and then the dominant result is obtained again by numerical analysis. A set of these results are added to the input of the artificial intelligence network to improve the accuracy of the artificial intelligence network.

As shown in Fig. 5, the flowchart of the optimal design method finds a set of design variables for the next generation in the same way and uses the advanced artificial intelligence network previously. Finally, the optimal value is found by iterating until the accuracy of the artificial intelligence network is sufficiently improved.

The objective function of the compressor optimization may vary depending on the user. That is, it can be aimed only at high efficiency or at high compression ratio. In the present invention, an objective function (Obj) which simultaneously considers the efficiency (Pη) and the compression ratio (PΠ) is set as shown in equation (2).

식(2)

Equation (2)

)과 이상적인 최대효율(100%)과의 차이를 식(3)과 같이 효율에 대한 목적함수로 설정하였다. 따라서 효율의 증대를 얻기 위하여서는 Pη 은 최소화되어야 한다.In equation (2), ω is a weighting function. Pη is the total efficiency obtained from the calculation (

) And the ideal maximum efficiency (100%) as an objective function for efficiency as shown in Equation (3). Therefore, Pη should be minimized to obtain an increase in efficiency.

식(3)

Equation (3)

The compression ratio is also an important factor in representing the performance of a compressor. The compression ratio is initially set to meet the purpose of use, but it is problematic if it is lower than the compression ratio required for the purpose of use. Also, the power loss is increased even if it is higher than the required compression ratio. Therefore, the compression ratio is increased as compared with the required compression ratio (Πrq), while the importance is limited as compared with the efficiency. The objective function of the compression ratio is set as shown in Equation (4), the weight is set to 0.3, and the efficiency weight is set to 1.0. Minimizing the objective function maximizes the compression ratio and efficiency.

식(4)

Equation (4)

Since the airfoil angle of the impeller airfoil changes in accordance with the change of the design variables in the optimization process, the impeller airfoil is severely bent and the stress exceeds the allowable stress. In this case, the limiting condition that the stress applied to the impeller is not to exceed the allowable stress is to be set as in (5).

식(5)

Equation (5)

In the artificial intelligence, we obtained the output by setting two efficiency and pressure ratio for the input of the design parameter set. From the obtained efficiency and pressure ratio, we obtain the objective function as in Eq. (2) and get the set of next generation design variables by PSO. The set of obtained design variables was applied to the artificial intelligence network and the objective function improved more than a certain amount was recalculated by numerical analysis. The recalculated numbers for each household are shown in Table 4.

<Table 4>

In Table 4, the pressure ratio and efficiency obtained from the artificial intelligence network are recalculated to show the accuracy of the obtained results. Equation (6) shows the method of calculating the efficiency prediction accuracy by reference. The accuracy of pressure ratio and efficiency prediction in 7th generation is less than 0.6%. The objective function shows an accuracy of about 6% because the actual pressure and efficiency scales are different. However, since the pressure ratio and the accuracy of efficiency to be obtained are improved by generation, it is appropriate to optimize using artificial intelligence network.

식(6)

Equation (6)

For a set of design variables over 1,200 times, FIG. 6 shows the minimum objective function value for each generation. It can be seen that the objective function has not been improved since the seventh generation. Table 5 shows the set of design variables from which the minimum objective function is obtained.

<Table 5>

The pressure ratio and efficiency of the designed impeller were 2.69 and 87.1%, respectively. Figure 7 shows the shape of the optimized compressor and improved results can be obtained without changing the inlet and outlet radii and outlet width of the impeller as compared to the initial impeller.

Figure 8 shows the relationship between efficiency and pressure ratio for the entire set of design variables. The efficiency and the pressure ratio are in opposition, and the design parameters in the front line of the pallet can be selected and used according to the user. The results of the stress and displacement in the optimized impeller are shown in FIG. 9, where the maximum displacement is about 30% of the tip gap and the safety factor 2.1 is shown for the maximum stress.

The present invention finds a method of improving the efficiency or the pressure ratio by changing only the shape of the impeller and the shroud of the casing without changing the components of the impeller of the compressor. In order to apply the global optimization and to access the objective function quickly, AI algorithm was applied, and the experimental design method and the PSO method were mixed. In this way, the actual numerical analysis is carried out 150 times, and the calculation is performed for a total of 1,200 design parameters. As a result, the efficiency is improved by 1% and the pressure ratio is improved by 9.3% compared with the initial model.

The foregoing description is merely illustrative of the technical idea of the present invention and various changes and modifications may be made by those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are intended to illustrate rather than limit the scope of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents should be construed as falling within the scope of the present invention.

10: Hub
20: Shroud

Claims (6)

A design method for shortening calculation time while obtaining global results for impeller shape design,
Obtaining a design parameter related to the objective function;
Setting an area for forming an artificial intelligence network as an experimental design method in the design parameters obtained in step S1;
Forming an artificial intelligence network based on the area set in step S2;
A step S4 of applying a genetic algorithm to improve the artificial intelligence network and increasing the generation of the genetic algorithm to an optimal artificial intelligence network; And
And a step S5 of deriving a result of global optimization using the artificial intelligence network formed in step S4.
The method according to claim 1,
Wherein the artificial intelligence network applies a back propagation algorithm and the neurons of the hidden layer and the output layer are 10 and 2, respectively.
The method according to claim 1,
Wherein the experimental design method is a Fractional Factorial Design and the resolution VI method is applied in the step S2.
The method according to claim 1,
Wherein the experimental design method is used only in the initial artificial intelligent network forming step and only the artificial intelligent network formed in step S4 is used for global results.
The method according to claim 1,
Wherein the PSO (Particle Swarm Optimization) method is applied to improve the artificial intelligence network in step S4.
The method according to claim 1,
Wherein the artificial intelligence network is supplemented by numerically analyzing the results obtained using the artificial intelligence network in a new area obtained by the genetic algorithm in step S4.

Images