Numerical simulation method of wall function based on improved DP algorithm

Through the improved DP algorithm and SED theory, the wall function in the form of multi-level power function is automatically constructed, which solves the problem of capture difficulties and manual calibration dependence of existing wall functions in the transition areas of the viscosity bottom layer and the logarithmic law layer, and realizes the Efficient and unified numerical simulation and accuracy improvement.

CN118627323BActive Publication Date: 2024-11-19CALCULATION AERODYNAMICS INST CHINA AERODYNAMICS RES & DEV CENT
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411116811.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-11-19
Estimated Expiration
2044-08-15

AI Technical Summary

Technical Problem

The existing wall function is difficult to capture in the transition area between the viscous bottom layer and the logarithmic law layer, and relies on manual calibration, resulting in low numerical simulation efficiency.

Method used

Using improved DP algorithm and SED theory, the model parameters of the wall function are automatically obtained through curve fitting and nonlinear optimization methods, and the wall function in the form of multi-level power function is constructed to achieve consistency from the viscous bottom layer to the logarithmic law region. Validity.

Benefits of technology

Fast and efficient numerical simulation without manual calibration is achieved, the unity of wall functions and flow field continuity is maintained, the simulation efficiency and accuracy are improved, and subjectivity is reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118627323B_ABST
    Figure CN118627323B_ABST
Patent Text Reader

Abstract

The present invention discloses a numerical simulation method of wall function based on an improved DP algorithm, comprising: selecting dimensionless parameters involved in curve fitting; obtaining a number of trajectory points and coordinates, obtaining a number of broken lines based on the number of trajectory points and the coordinates of each trajectory point, and performing thinning processing on the number of broken lines using the DP algorithm to obtain M The power function required M trajectory points, based on the obtained M trajectory points, obtain the intersection points and exponents between the power functions of the multi-stage continuous power functions; based on M The wall function is obtained by using the intersection points and exponents of the segment power functions; a curve fitting is performed based on the wall function to obtain a fitting result; based on the fitting result, the wall function is optimized using an optimization method to obtain an optimized wall function; a numerical simulation is performed based on the optimized wall function to obtain a numerical simulation result; the present invention can quickly and efficiently obtain all the model parameters of the wall function without manual calibration, thereby efficiently realizing numerical simulation based on the wall function.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of numerical simulation, in particular to a numerical simulation method of a wall function based on an improved DP algorithm. Background Art

[0002] Combining wall functions with turbulence models can significantly reduce the dependence of wall turbulence simulation on grid resolution. However, existing wall functions such as the classical wall function proposed by Von Karman et al. cannot capture the transition region between the viscous bottom layer and the logarithmic law layer. The segmented wall function constructed by Castro-Orgaz et al. has a non-smooth velocity distribution.

[0003] In order to achieve a smooth and continuous representation of the turbulent boundary layer along the normal section of the wall, Spalding proposed a unified expression formula for the wall function, which is consistent and effective in both the viscous bottom layer and the logarithmic law layer. Its definition is as follows:

[0004] ;

[0005] in, y + is the dimensionless vertical distance to the wall, u + is the dimensionless tangential velocity, κ is the Karman constant, which is usually approximated to be around 0.41, B is a constant that depends on the specific fluid conditions and the degree of wall roughness, and is usually taken as 5.5. However, the Spalding wall function relies on physical cognition and is an implicit expression, which requires an iterative method to solve when applied to turbulence simulation.

[0006] Allmaras analytically derived the SA wall function (Spalart-Allmaras wall function) based on the SA turbulence model, which is applicable to incompressible turbulence and is defined as follows:

[0007] ;

[0008] in y + is the dimensionless vertical distance to the wall, u + is the dimensionless tangential velocity, and other parameters are defined as follows: , a1, b1, a2, b2, c1, c2, c3, and c4 are all constants.

[0009] Based on the theory of structural ensemble dynamics (SED), She et al. proposed a multi-level power function multiplication form to express the distribution of stress-length function. This work provides a unique way to explain and understand the complex relationship between empirical parameters and the physical phenomena they represent, and provides ideas for the construction of high-fidelity wall functions. However, some empirical parameters of SED theory need to be manually calibrated, which makes the existing numerical simulation methods based on wall functions less efficient. Summary of the invention

[0010] The purpose of the present invention is to obtain all parameters of the wall function model quickly and efficiently without manual calibration, and then efficiently realize numerical simulation based on the wall function, and the method can be consistently effective from the viscous bottom layer to the logarithmic law region.

[0011] To achieve the above-mentioned object of the invention, the present invention provides a numerical simulation method of wall function based on an improved DP algorithm, characterized in that the method comprises:

[0012] Step 1: Select dimensionless parameters involved in curve fitting;

[0013] Step 2: Obtain several trajectory points and the coordinates of each trajectory point, obtain several broken lines based on the several trajectory points and the coordinates of each trajectory point, and use the DP algorithm to perform thinning on the several broken lines to obtain M The power function required M trajectory points, based on the obtained M trajectory points, obtain the intersection points and exponents between the power functions of the multi-stage continuous power functions; based on M The intersection points and exponents between the segment power functions obtain the wall function;

[0014] Step 3: performing curve fitting based on the dimensionless parameters involved in curve fitting and the wall function to obtain a fitting result;

[0015] Step 4: Based on the fitting results, the wall function is optimized using an optimization method to obtain an optimized wall function;

[0016] Step 5: Perform numerical simulation based on the optimized wall function to obtain numerical simulation results.

[0017] The principle of the present invention is based on the advantages of SED theory and DP algorithm. SED theory is a theory for quantifying the multi-layer structure of complex systems, which believes that the multi-layer structure can be described by a set of polynomial multiplications. The necessary condition for using SED is to find a similarity law in which a certain physical quantity has good consistency with another independent variable. On the other hand, as far as the flow itself is concerned, in incompressible wall turbulence, the traditional y+ and u + It has been proved that it has a good similarity law, which is a sufficient condition for constructing the wall function model. However, the wall function in the past is usually in the form of a transcendental function, which is not convenient for numerical solution. Through simple transformation, it can be considered that R c equal u + Multiply y + Likewise with y + has a good similarity law, so we use SED theory to R c The DP algorithm uses simplified polylines to replace the original excessive polyline data points, extracts important feature points and reconstructs the trajectory to make it close to the original trajectory. Therefore, it can effectively simplify the curve and reduce the amount of storage and calculation while maintaining the basic shape of the curve.

[0018] Among them, in some embodiments of the present invention, the M The way to obtain the trajectory points is:

[0019] N The trajectory points are P 1, P 2, ..., P N , N is the number of trajectory points, i The coordinates of the trajectory points are marked as , point P 1 and P N Connect to obtain a line, and use this line as a distance reference line. N Points in the trajectory except P 1 and P N The remaining trajectory points outside the reference line are perpendicular to the distance line, and the distances are recorded as , record the distance value from each trajectory point to the distance reference line, sort all the distance values ​​from large to small to obtain the sorting result, and then sort the distance values ​​based on the top of the sorting result. M The trajectory points corresponding to the distance values ​​are obtained. M points.

[0020] In some embodiments of the present invention, the intersection points of the power functions of each segment are obtained as follows:

[0021] ;

[0022] in, is the horizontal coordinate set of the intersection points of the power function, P j( x j , y j ) is the j Track points, x j For the j The horizontal coordinate of the trajectory point, y j For the j The vertical coordinate of each trajectory point;

[0023] The exponent of each power function is obtained as follows:

[0024] ;

[0025] in, r j The power function exponential set, x j-1 For the j-1 The horizontal coordinate of the trajectory point, y j-1 For the j-1 The vertical coordinate of the trajectory point.

[0026] Wherein, in some embodiments of the present invention, the expression of the wall function is:

[0027] ;

[0028] in, All trajectory points output by the wall function P i The vertical coordinate value of C 1 is the coefficient of the first power function, p is the shape factor, For the j The power function and j-1 The intersection of the segment power function, r j-1 For the j-1 The exponent of the segment power function, x i For trajectory points P i The horizontal axis value of is the fixed expression of the first power function.

[0029] In some embodiments of the present invention, step 3 is specifically: curve fitting is performed based on the dimensionless parameters involved in curve fitting and the wall function to obtain each trajectory point P i The vertical coordinate value of ; Step 4 is specifically as follows: since the intersection points and exponents of the power functions of each segment obtained in step 2 are all initial values ​​obtained by the DP algorithm, if the initial values ​​are directly used for curve fitting, the effect is not good, so all the initial values ​​obtained are nonlinearly optimized until they are consistent with the ordinate value The iteration stops when the relevant relative loss function is the smallest. The initial values ​​include: the horizontal coordinates of the intersection points of each power function segment, the exponents of each power function segment, and the shape factor p .

[0030] Among them, in some embodiments of the present invention, the loss function is , y i For the i The vertical coordinate of the trajectory point.

[0031] Among them, in some embodiments of the present invention, the quasi-Newton gradient method BFGS is used to perform nonlinear optimization on all the initial values ​​obtained. The advantage of BFGS is that it converges quickly, does not require second-order derivatives, and has good numerical stability. Other commonly used optimization methods include the Gradient Descent method, which is suitable for some simple problems but has a slow convergence speed; the Conjugate Gradient method has low numerical stability and sometimes cannot guarantee convergence; the Stochastic Gradient Descent method (SGD) may converge slowly and may converge to a local minimum due to its randomness, which is not the global optimum.

[0032] In some embodiments of the present invention, the method further comprises: comparing the calculated value with y i The data are compared to obtain a comparison result, and an accuracy evaluation result of the wall function fitting is obtained based on the comparison result.

[0033] One or more technical solutions provided by the present invention have at least the following technical effects or advantages:

[0034] This method can be consistently satisfied from the viscous bottom layer to the logarithmic law region, maintaining the uniformity of the wall function and not causing discontinuity in the basic quantities of the flow field.

[0035] This method can automatically iterate the optimal parameter set using the initial values ​​obtained by the DP algorithm.

[0036] This method gives a loss function and can obtain the parameter set when the loss function is the smallest.

[0037] The method of the present invention is used to construct the R cThe wall function is implanted into the CFD fluid dynamics simulation solver, which does not require iterative solution and has strong applicability and high efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation on the embodiments of the present invention;

[0039] Figure 1 It is a flow chart of the numerical simulation method of the wall function based on the improved DP algorithm;

[0040] Figure 2-Figure 5 It is a schematic diagram of the specific steps of implementing the DP algorithm in the present invention;

[0041] Figure 6 The present invention is based on y + and u + Schematic diagram of the comparison between the wall function fitting results and DNS data, SA wall function and Spalding wall function;

[0042] Figure 7 The present invention is based on y + and u + Schematic diagram of the error between the wall function, SA wall function and Spalding wall function and DNS data;

[0043] Figure 8 The present invention is based on R c and y + Schematic diagram of the comparison between the wall function fitting results and DNS data, SA wall function and Spalding wall function;

[0044] Fig. 9 The present invention is based on R c and y + Schematic diagram of the error between the wall function, SA wall function and Spalding wall function and DNS data. DETAILED DESCRIPTION

[0045] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0046] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those within the scope of this description. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0047] Embodiment 1;

[0048] Please refer to Figure 1 , Figure 1 The present invention is a flow chart of a numerical simulation method of a wall function based on an improved DP algorithm. The present invention provides a numerical simulation method of a wall function based on an improved DP algorithm, and the method comprises:

[0049] Step 1: Select dimensionless parameters involved in curve fitting;

[0050] Step 2: Obtain several trajectory points and the coordinates of each trajectory point, obtain several broken lines based on the several trajectory points and the coordinates of each trajectory point, and use the DP algorithm to perform thinning on the several broken lines to obtain M The power function required M trajectory points, based on the obtained M trajectory points, obtain the intersection points and exponents between the power functions of the multi-stage continuous power functions; based on M The intersection points and exponents between the segment power functions obtain the wall function;

[0051] Step 3: performing curve fitting based on the dimensionless parameters involved in curve fitting and the wall function to obtain a fitting result;

[0052] Step 4: Based on the fitting results, the wall function is optimized using an optimization method to obtain an optimized wall function;

[0053] Step 5: Perform numerical simulation based on the optimized wall function to obtain numerical simulation results. For specific numerical simulation methods, please refer to Luo Canyan, Bi Lin, Xu Jinglei, et al. Study on wall function method of different turbulence models under Cartesian grid [J / OL]. Engineering Mechanics, 2024, (08): 11-22. The embodiments of the present invention will not be described in detail.

[0054] The present invention can fit a wall function expression in the form of multi-level power function multiplication according to arbitrarily selected dimensionless parameters, without relying on manual calibration. The method is simple and efficient, and the constructed wall function has high accuracy and strong robustness compared with traditional wall functions.

[0055] To achieve automatic calibration, the Douglas-Peucker algorithm (DP algorithm for short) is introduced in the embodiment of the present invention. The DP algorithm plays an important role in the field of vector data simplification and map generalization. The classic DP algorithm is a classic simplification algorithm based on the recursive and thinning method of retaining the farthest vertex to effectively compress the trajectory while retaining the main spatial features of the trajectory. The specific implementation process is as follows. Assume that there are points in the trajectory P 1, P 2, ..., P N , whose trajectory point coordinates are marked as First, point P 1 and P N Connected, take this line as the distance reference line, draw perpendicular lines through the remaining points to the line segment, and record the distances as .according to d i The maximum distance of the current trajectory point can be obtained d max =max( d i ), the classic DP algorithm pre-sets the required distance threshold d threshold ,like d max ≥d threshold , then keep the trajectory point and use this point as the boundary to search for distance on both sides; if d max <d threshold , then delete the track point until all track points are traversed. The resulting thinned line graph can effectively simplify the curve and reduce storage and calculation while maintaining the basic shape of the curve. Figure 2-Figure 5 shown.

[0056] The present invention proposes for the first time a wall function based on SED and improved DP method to construct a multi-level power function multiplication form, which ensures the uniformity from the viscous bottom layer to the logarithmic law region, and does not require manual calibration of empirical parameters. Users can also select two appropriate dimensionless parameters according to actual needs, and use the construction method of the present invention to obtain a wall function expression suitable for a specific problem, thereby realizing numerical simulation based on the wall function.

[0057] The wall function constructed by the method provided by the embodiment of the present invention realizes numerical simulation and can be used for multi-level power function fitting of any two suitable dimensionless parameters, and the fitting curve is highly consistent with the original data curve.

[0058] The specific implementation steps of the present invention are:

[0059] First, two dimensionless parameters are selected to participate in curve fitting.y + and u + ,or R c and y + . y + is the dimensionless vertical distance to the wall, u + is the dimensionless tangential velocity, R c for u + Multiply y + .

[0060] The improved DP algorithm is used for line thinning. The improved DP algorithm deletes the threshold of the classic DP algorithm, which is equivalent to setting it to 0. In the calculation process, it is assumed that there are N Points P 1, P 2, ..., P N , whose trajectory point coordinates are marked as First, point P 1 and P N Connected, take this line as the distance reference line, draw perpendicular lines through the remaining points to the line segment, and record the distances as Record the distance from each trajectory point to the reference line, and finally sort all the distances from large to small, keeping the first number of power function segments required by the method of the present invention. M Track points.

[0061] The previous determined M By using trajectory points, we can get the intersection points and exponents of each segment of the multi-stage power function.

[0062] The intersection points of each power function segment are obtained by the following formula:

[0063] ;

[0064] in, is the horizontal coordinate set of the intersection points of the power function, P j ( x j , y j ) is the j The coordinates of the trajectory points, x j For the j The horizontal coordinate of the trajectory point, y j For thej The vertical coordinate of the trajectory point.

[0065] The exponent of each power function is obtained by the following formula:

[0066] ;

[0067] in, r j The power function exponential set is calculated by M Adjacent track points P j ( x j , y j ) is obtained by the gradient between .

[0068] The transition function between the two power functions can be expressed as the following formula. Determine the shape factor p to obtain the first distribution function:

[0069] ;

[0070] in, represents the intersection of the first power function and the second power function, C 1 is the coefficient of the first power function, p is the shape factor, which can be used to adjust the curvature of the curve to better match the experimental data.

[0071] Combining the first distribution function, we can get the entire distribution function, that is, the wall function, expressed as:

[0072] ;

[0073] Use the wall function fitting formula based on the improved DP algorithm proposed above to curve fit each trajectory point P i The vertical coordinate value of ;

[0074] Use the quasi-Newton gradient method BFGS to perform nonlinear optimization on all the initial values ​​until the relative loss function When it is the smallest, the iteration stops. x i Track Points P i The horizontal axis value of is the trajectory point fitted using the method proposed in this paper P i The vertical coordinate value of .

[0075] The calculated Value and yj By comparing the data, the fitting accuracy of the expression can be obtained.

[0076] According to the wall function construction method based on the improved DP algorithm proposed in this paper, an experimental comparison is carried out on the direct numerical simulation (DNS) data, and the results are as follows:

[0077] The independent variable is y + , the dependent variable is u + ;

[0078] ;

[0079] in, for u + The predicted value of

[0080] The fitting results are as follows Figure 6-Figure 7 As shown:

[0081] Under the double logarithmic scale, it can be clearly seen that the formula fitting curve is highly consistent with the DNS data curve. Figure 6 As shown. The error values ​​of each data point of the formula proposed by the present invention are almost all below 0.02%, showing a good fitting effect. The error of the SA formula is below 6.3%. The error of the formula proposed by Spalding et al. is as high as 12.2%. Figure 7 As shown, by comparing the SA formula and the Spalding formula, the fitting accuracy of the method proposed in the present invention is higher.

[0082] The independent variable is R c , the dependent variable is y + . It is the predicted value obtained by fitting the formula of this method.

[0083] ;

[0084] The fitting results are as follows: Figure 8-Figure 9 As shown:

[0085] Under the double logarithmic scale, the formula fitting curve is highly matched with the original data curve, such as Figure 8 As shown. The error values ​​of each data point of the formula proposed by the present invention are almost all below 0.06%, the curve error of the SA formula is below 6.6%, and the curve error of the Spalding formula is below 12.3%. The formula proposed by the method of the present invention shows a better fitting effect. By comparison, as Fig. 9 As shown, the formula proposed in the present invention has higher fitting accuracy.

[0086] This method constructs a unified wall function in the form of a continuous power exponent constructed based on SED, and uses an improved DP algorithm and BFGS nonlinear optimization algorithm to automatically select the optimal parameters, thereby improving the accuracy and adaptability of the model parameters and making the model better fit the actual flow field data. A loss function is also constructed to obtain the optimal parameter set while minimizing the error. The construction method of the present invention can improve the performance and prediction accuracy of the wall function, while reducing subjectivity and manual intervention.

[0087] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0088] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. Numerical simulation method of wall function based on improved DP algorithm, It is characterized in that The method comprises: Step 1: Select dimensionless parameters involved in curve fitting; Step 2: Obtain several trajectory points and the coordinates of each trajectory point, obtain several broken lines based on the several trajectory points and the coordinates of each trajectory point, and use the DP algorithm to perform thinning on the several broken lines to obtain M The power function required M trajectory points, based on the obtained M trajectory points, obtain the intersection points and exponents between the power functions of the multi-stage continuous power functions; based on M The intersection points and exponents between the segment power functions obtain the wall function; Step 3: performing curve fitting based on the dimensionless parameters involved in curve fitting and the wall function to obtain a fitting result; Step 4: Based on the fitting results, the wall function is optimized using an optimization method to obtain an optimized wall function; Step 5: Based on the optimized wall function, perform turbulence numerical simulation on the flat plate, convex hull and airfoil to obtain turbulence numerical simulation results.

2. The numerical simulation method of wall function based on the improved DP algorithm according to claim 1, It is characterized in that Said M The way to obtain the trajectory points is: N The trajectory points are: P 1 , P 2 , …, P N , N is the number of trajectory points, i The coordinates of the trajectory points are: , point P 1 and P N Connect to obtain a line, and use this line as a distance reference line. N Points in the trajectory P 1 and P N The remaining trajectory points outside the reference line are perpendicular to the distance line, and the distances are recorded as , record the distance value from each trajectory point to the distance reference line, sort all the distance values ​​from large to small to obtain the sorting result, and then sort the distance values ​​based on the top of the sorting result. M The trajectory points corresponding to the distance values ​​are obtained. M points.

3. The numerical simulation method of wall function based on the improved DP algorithm according to claim 1, It is characterized in that The intersection points of each power function are obtained as follows: ; in, is the horizontal coordinate set of the intersection points of the power function, P j ( x j , y j ) is the j Track points, x j For the j The horizontal coordinate of the trajectory point, y j For the j The vertical coordinate of each trajectory point; The exponent of each power function is obtained as follows: ; in, r j The power function exponential set, x j-1 For the j -1 horizontal coordinate of the trajectory point, y j-1 For the j -1 The vertical coordinate of the trajectory point.

4. The numerical simulation method of wall function based on the improved DP algorithm according to claim 3, It is characterized in that The expression of the wall function is: ; in, All trajectory points output by the wall function P i The vertical coordinate value of C 1 is the coefficient of the first power function, p is the shape factor, For the j The power function and j-1 The intersection of the segment power function, r j-1 For the j-1 The exponent of the segment power function, x i For trajectory points P i The horizontal axis value of is the fixed expression of the first power function.

5. The numerical simulation method of wall function based on the improved DP algorithm according to claim 1, It is characterized in that The step 3 is specifically as follows: curve fitting is performed based on the dimensionless parameters involved in the curve fitting and the wall function to obtain each trajectory point P i The vertical coordinate value of ; The specific step 4 is: perform nonlinear optimization on all the initial values ​​obtained until they are consistent with the ordinate value When the relevant relative loss function is the minimum, the iteration is stopped. The initial values ​​include: the horizontal coordinates of the intersection points of each power function segment, the exponents of each power function segment, and the shape factor p .

6. The numerical simulation method of wall function based on the improved DP algorithm according to claim 5, It is characterized in that The loss function is , y i For the i The ordinate of the trajectory point, N is the number of trajectory points.

7. The numerical simulation method of wall function based on the improved DP algorithm according to claim 5, It is characterized in that The quasi-Newton gradient method BFGS is used to perform nonlinear optimization on all the initial values ​​obtained.

8. The numerical simulation method of wall function based on the improved DP algorithm according to claim 5, It is characterized in that The method further comprises: Value and y i The comparison is performed to obtain the comparison result, and the accuracy evaluation result of the wall function fitting is obtained based on the comparison result. y i For the i The vertical coordinate of the trajectory point.