Skip to main content
SpringerLink
Log in

A novel global optimization algorithm and data-mining methods for turbomachinery design

  • Research Paper
  • Published:
Structural and Multidisciplinary Optimization Aims and scope Submit manuscript

Abstract

A new multi-objective, multi-disciplinary global optimization strategy is proposed to address the high-dimensional, computationally expensive black box problem (HEB) in turbomachinery design. The strategy consists of an adaptive sampling hybrid optimization algorithm (ASHOA), two data-mining techniques, a 3D blade parameterization method, and the aerodynamic/mechanical codes. Firstly, the ASHOA is established by integrating a novel adaptive sampling Kriging metamodel and a new hybrid optimization method. Secondly, two data-mining methods (analysis of variance (ANOVA) and self-organizing map (SOM)) are applied to set the initial design space and optimization objectives of the transonic centrifugal compressor. A refined design space and objective parameters of the optimization problem are eventually obtained. Finally, the optimization process of a transonic centrifugal compressor is carried out based on the refined design space and objectives using ASHOA. The results show that the search efficiency of the optimization strategy is 2–10 times higher when compared to other excellent optimization algorithms. For the optimized compressor, both isentropic efficiency and total pressure ratio at design condition are improved by 1.61% and 4.13%, respectively, and the maximum stress decreases by 9.68%.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11
Fig. 12
Fig. 13
Fig. 14
Fig. 15
Fig. 16
Fig. 17
Fig. 18
Fig. 19
Fig. 20
Fig. 21
Fig. 22
Fig. 23
Fig. 24
Fig. 25

Similar content being viewed by others

References

  • Aslimani N, Ellaia R (2017) A new hybrid algorithm combining a new chaos optimization approach with gradient descent for high dimensional optimization problems. Comput Appl Math 4:1–29

    MATH  Google Scholar 

  • Barsi D, Perrone A, Ratto L, Simoni D, Zunino P (2015) Radial inflow turbine design through multi-disciplinary optimisation technique. ASME Turbo Expo 2015:GT2015–G42702

    Google Scholar 

  • Barsi D, Perrone A, Qu Y, Ratto L, Ricci G, Sergeev V, Zunino P (2018) Compressor and turbine multidisciplinary design for highly efficient micro-gas turbine. J Therm Sci 27(3):259–269

    Article  Google Scholar 

  • Blondet G, Boudaoud N, Duigou J (2015) Simulation data management for adaptive design of experiments. Mech Ind 16(6):611

    Article  Google Scholar 

  • Chen L, Chen J (2015) Aerodynamic optimization design of multi-stage turbine using the continuous adjoint method. Int J Turbo Jet-Engines 32(2):199–211

    Google Scholar 

  • Deb K, Pratap A, Agarwal S, Meyarivan T (2002) A fast and elitist multi objective genetic algorithm: NSGA-II. IEEE Trans Evol Comput 6(2):182–197

    Article  Google Scholar 

  • Deshmukh A, Allison J (2016) Multidisciplinary dynamic optimization of horizontal axis wind turbine design. Struct Multidiscip Optim 53(1):15–27

  • Dinh C, Ma S, Kim K (2017) Aerodynamic optimization of a single-stage axial compressor with stator shroud air injection. AIAA J 55(8):2739–2754

    Article  Google Scholar 

  • Elfert M, Weber A, Wittrock D, Peters A, Voss C, Nicke E (2017) Experimental and numerical verification of an optimization of a fast rotating high-performance radial compressor impeller. J Turbomach 139(10):101007

    Article  Google Scholar 

  • Gowda K, Prasad S, Nagarajaiah V (2016). Design optimization of t-root geometry of a gas engine HP compressor rotor blade for lifing the blade against fretting failure. ASME 2016, Power 2016–59331

  • Grosso A, Jamali A, Locatelli M (2009) Finding maximin Latin hypercube designs by iterated local search heuristics. Eur J Oper Res 197(2):541–547

    Article  MATH  Google Scholar 

  • Hehn A, Mosdzien M, Grates D, Jeschke P (2018) Aerodynamic optimization of a transonic centrifugal compressor by using arbitrary blade surfaces. J Turbomach 140(5):051011

    Article  Google Scholar 

  • Heinrich M, Schwarze R (2016) Genetic algorithm optimization of the volute shape of a centrifugal compressor. Int J Rotating Mach 2016:13

    Article  Google Scholar 

  • Hu J, Zhou Q, Jiang P, Shao X, Xie T (2017) An adaptive sampling method for variable-fidelity surrogate models using improved hierarchical kriging. Eng Optim 50(1):145–163

    Article  Google Scholar 

  • Huang D, Allen T, Notz W, Zeng N (2006) Global optimization of stochastic black-box systems via sequential kriging meta-models. J Glob Optim 34(3):441–466

    Article  MathSciNet  MATH  Google Scholar 

  • Javed A, Pecnik R, Buijtenen J (2013) Optimization of a centrifugal compressor impeller design for robustness to manufacturing uncertainties. J Eng Gas Turbines Power 138(11):43

    Google Scholar 

  • Jones D, Schonlau M, Welch W (1998) Efficient global optimization of expensive black-box functions. J Glob Optim 13(4):455–492

    Article  MathSciNet  MATH  Google Scholar 

  • Kato H, Funazaki K (2014) POD-driven adaptive sampling for efficient surrogate modeling and its application to supersonic turbine optimization. ASME Turbo Expo 2014:GT2014–G27229

    Google Scholar 

  • Kobrunov A, Priezzhev I (2016) Hybrid combination genetic algorithm and controlled gradient method to train a neural network. Geophysics 81(4):IM35–IM43

    Article  Google Scholar 

  • Kristensen J, Ling Y, Asher I, Wang L (2016) Expected-improvement-based methods for adaptive sampling in multi-objective optimization problems. ASME 2016:DETC2016–DET59266

    Google Scholar 

  • Leborgne M, Lonfils T, Lepot I (2015) Development and application of a multi-disciplinary multi-regime design methodology of a low-noise contra-rotating open-rotor. ASME Turbo Expo 2015:GT2015–G43432

    Google Scholar 

  • Li X, Zhao Y, Liu Z, Chen H (2016) The optimization of a centrifugal impeller based on a new multi-objective evolutionary strategy. ASME Turbo Expo 2016:GT2016–G56592

    Google Scholar 

  • Li X, Liu Z, Lin Y (2017) Multi point and multi objective optimization of a centrifugal compressor impeller based on genetic algorithm. Math Probl Eng 2017(1):1–18

    Article  Google Scholar 

  • Liu H, Xu S, Ma Y, Chen X, Wang X (2015) An adaptive Bayesian sequential sampling approach for global metamodeling. J Mech Des 138(1):011404

    Article  Google Scholar 

  • Liu Y, Shi Y, Zhou Q, Xiu R (2016) A sequential sampling strategy to improve the global fidelity of metamodels in multi-level system design. Struct Multidiscip Optim 53(6):1295–1313

    Article  MathSciNet  Google Scholar 

  • Liu B, Grout V, Nikolaeva A (2018) Efficient global optimization of actuator based on a surrogate model assisted hybrid algorithm. IEEE Trans Ind Electron 65(7):5712–5721

    Article  Google Scholar 

  • Long Q, Wu C (2014) A hybrid method combining genetic algorithm and Hooke-Jeeves method for constrained global optimization. J Ind Manag Optim 10(4):1279–1296

    MathSciNet  MATH  Google Scholar 

  • Lophaven S, Nielsen H, Sondergaard J (2002) Dace—a MATLAB kriging toolbox (version 2) informatics and mathematical modeling. Technical University of Denmark, Copenhagen

    Google Scholar 

  • Lu H, Li Q, Pan T (2016) Optimization of cantilevered stators in an industrial multistage compressor to improve efficiency. Energy 106:590–601

    Article  Google Scholar 

  • Luo L, Hou X, Zhong J, Cai W, Ma J (2017) Sampling-based adaptive bounding evolutionary algorithm for continuous optimization problems. Inf Sci 382–383:216–233

    Article  Google Scholar 

  • Ma C, Su X, Yuan X (2017) An efficient unsteady adjoint optimization system for multistage turbomachinery. J Turbomach 139(1):011003

    Article  Google Scholar 

  • Ma S, Afzal A, Kim K (2018) Optimization of ring cavity in a centrifugal compressor based on comparative analysis of optimization algorithms. Appl Therm Eng 138:633–647

    Article  Google Scholar 

  • Mahmood G, Acharya S (2006) Experimental investigation of secondary flow structure in a blade passage with and without leading edge fillets. J Fluids Eng 129(3):253–262

    Article  Google Scholar 

  • Ning F, Xu L (2001) Numerical investigation of transonic compressor rotor flow using an implicit 3D flow solver with one-equation Spalart-Allmaras turbulence model, ASME Turbo Expo 2001-GT-0359

  • Pellegrini R, Iemma U, Leotardi C, Campana E, Diez M (2016) Multi-fidelity adaptive global metamodel of expensive computer simulations. IEEE World Congress Comput Intell https://doi.org/10.1109/CEC.2016.7744355

  • Pholdee N, Bureerat S (2015) An efficient optimum Latin hypercube sampling technique based on sequencing optimisation using simulated annealing. Int J Syst Sci 46(10):1780–1789

    Article  MATH  Google Scholar 

  • Picheny V, Ginsbourger D, Roustant O, Haftka R, Kim N (2010) Adaptive designs of experiments for accurate approximation of a target region. J Mech Des 132(7):071008

    Article  Google Scholar 

  • Salnikov A, Danilov M (2017) A centrifugal compressor impeller: a multidisciplinary optimization to improve its mass, strength, and gas-dynamic characteristics. ASME Turbo Expo 2017:GT2017–G64123

    Google Scholar 

  • Sevastyanov V (2010) Hybrid multi-gradient explorer algorithm for global multi-objective optimization. AIAA/ISSMO Multidisciplinary Analysis Optimization Conference

  • Song L, Guo Z, Li J, Feng Z (2016) Research on meta-model based global design optimization and data mining methods. J Eng Gas Turbines Power 138(9):092604

    Article  Google Scholar 

  • Verstraete T, Mueller L, Mueller J (2017) Multidisciplinary adjoint optimization of turbomachinery components including aerodynamic and stress performance. 35th AIAA, Applied Aerodynamics Conference, https://doi.org/10.2514/6.2017-4083

  • Viana F, Venter G, Balabanov V (2010) An algorithm for fast optimal Latin hypercube design of experiments. Int J Numer Methods Eng 82(2):135–156

    MathSciNet  MATH  Google Scholar 

  • Viana F, Haftka R, Watson L (2013) Efficient global optimization algorithm assisted by multiple surrogate techniques. J Glob Optim 56(2):669–689

    Article  MATH  Google Scholar 

  • Walther B, Nadarajah S (2015) Optimum shape design for multi-row turbomachinery configurations using a discrete adjoint approach and an efficient RBF deformation scheme for complex multi-block grids. J Turbomach 137(8):33

    Article  Google Scholar 

  • Wang L, Han R, Wang T, Ke S (2018) Uniform decomposition and positive-gradient differential evolution for multi-objective design of wind turbine blade. Energies 11(5):1–19

    Google Scholar 

  • Wu J, Xia J, Chen J, Cui Z (2011) Moving object classification method based on SOM and K-means. J Comput 6(8):1654–1661

    Article  Google Scholar 

  • Wu G, Qiu D, Yu Y, Pedrycz W, Ma M, Li H (2014) Superior solution guided particle swarm optimization combined with local search techniques. Expert Syst Appl 41(16):7536–7548

    Article  Google Scholar 

  • Xu S, Liu H, Wang X, Jiang X (2014) A robust error-pursuing sequential sampling approach for global metamodeling based on Voronoi diagram and cross validation. J Mech Des 136(7):69–74

    Article  Google Scholar 

  • Zhu H, Liu L, Long T, Peng L (2012) A novel algorithm of maximin Latin hypercube design using successive local enumeration. Eng Optim 44(5):551–564

    Article  Google Scholar 

Download references

Acknowledgments

The authors would like to acknowledge National Laboratory of Engine Turbocharging Technology, North China Engine Research Institute, for providing experimental data.

Funding

This work was supported by the National Natural Science Foundation of China under Grant number 11672206.

Author information

Authors and Affiliations

  1. Department of Mechanics, Tianjin University, Tianjin, 300072, China

    Xiaojian Li, Yijia Zhao & Zhengxian Liu

Authors
  1. Xiaojian Li
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Yijia Zhao
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Zhengxian Liu
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Zhengxian Liu.

Ethics declarations

Disclosure statement

The authors declare that they have no conflicts of interest.

Additional information

Responsible Editor: Gengdong Cheng

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published mapsand institutional affiliations.

Appendices

Appendix 1

1.1 ZDT1

ZDT1 has 30 design variables and multiple Pareto fronts. The mathematical expression is

$$ {\displaystyle \begin{array}{l}\operatorname{Minimize}\kern0.36em {f}_1={x}_1\kern0.36em \mathrm{and}\kern0.36em {f}_2=g\left(1-\sqrt{\frac{f_1}{g}}\right)\\ {}\mathrm{where},g=1+\frac{9}{n-1}\sum \limits_{i=2}^n{x}_i;\kern0.84em 0\le {x}_i\le 1,i=1,\cdots, n;n=30\end{array}} $$
(23)

The Pareto-optimal region for the case ZDT1 corresponds to x1 ∈ [0, 1], xi = 0, i = 2, ⋯, 30.

1.2 ZDT2

ZDT2 has 30 design variables and multiple Pareto fronts. The mathematical expression is

$$ {\displaystyle \begin{array}{l}\operatorname{Minimize}\kern0.36em {f}_1={x}_1\kern0.36em \mathrm{and}\kern0.36em {f}_2=g\left[1-{\left(\frac{f_1}{g}\right)}^2\right]\\ {}\mathrm{where},g=1+\frac{9}{n-1}\sum \limits_{i=2}^n{x}_i;\kern0.84em 0\le {x}_i\le 1,i=1,\cdots, n;n=30\end{array}} $$
(24)

The Pareto-optimal region for the case ZDT2 corresponds to x1 ∈ [0, 1], xi = 0, i = 2, ⋯, 30.

1.3 ZDT4

ZDT4 has 10 design variables and multiple local Pareto fronts. The mathematical expression is

$$ {\displaystyle \begin{array}{l}\operatorname{Minimize}\kern0.36em {f}_1={x}_1\kern0.36em \mathrm{and}\kern0.36em {f}_2=g\left(1-\sqrt{\frac{f_1}{g}}\right)\\ {}\mathrm{where},g=1+10\left(n-1\right)+\sum \limits_{i=2}^n\left({x}_i^2-10\cos \left(4\pi {x}_i\right)\right);\\ {}0\le {x}_1\le 1,\kern0.48em -5\le {x}_i\le 5,i=2,\cdots, n;n=10\end{array}} $$
(25)

The Pareto-optimal region for the case ZDT4 corresponds to x1 ∈ [0, 1], xi = 0, i = 2, ⋯, 10. There exist 219 local Pareto-optimal solutions and about 100 distinct Pareto fronts.

The basic optimization parameters of ASHOA for ZDT1, ZDT2, and ZDT4 are given in Table 13. Figure 26 shows the optimization results for test cases ZDT1, ZDT2, and ZDT4.

Table 13 Parameters of ASHOA for ZDT1, ZDT2, and ZDT4
Full size table
Fig. 26
figure 26

Results for test cases ZDT1, ZDT2, and ZDT4 optimized by ASHOA and five other state-of-the-art algorithms (AME, CV-Voronoi, EGO, ADE, and MBOE) at the same search time

Full size image

Appendix 2

The demonstrations of the procedure (see Section 2.3.3) to search the optimal λ for test cases ZDT3 and ZDT6 (Figs. 27 and 28).

Fig. 27
figure 27

The process of searching the optimal λ for ZDT3

Full size image
Fig. 28
figure 28

The process of searching the optimal λ for ZDT6

Full size image
Fig. 29
figure 29

Effect of different values of λ on optimization efficiency and solutions diversity for ZDT1, ZDT2, and ZDT4

Full size image

1.1 ZDT 3

(1) λ = 0, the growth rate Δλ = 0.05, λ1 = λ + Δλ = 0.05.

(2) λ1 < 1, go next.

(3) & (4)

Fig. 3a, the 1st iteration,

λ = 0,λ1 = 0.05,Nλ1 = 15 > Nλ = 1, then update λ = 0.05, λ1 = 0.1, go next.

Fig. 3b, the 2nd iteration,

λ = 0.05,λ1 = 0.1,Nλ1 = 6 > Nλ = 1, then update λ = 0.1,λ1 = 0.15, go next.

Fig. 3c, the 3rd iteration,

λ = 0.1,λ1 = 0.15,Nλ1 = 6 > Nλ = 2, then update λ = 0.15,λ1 = 0.2, go next.

Fig. 3d, the 4th iteration,

λ = 0.15,λ1 = 0.2,Nλ1 = 17 > Nλ = 1, then update λ = 0.2,λ1 = 0.25, go next.

Fig. 3e, the 5th iteration,

λ = 0.2,λ1 = 0.25,Nλ1 = 20 > Nλ = 1, then update λ = 0.25,λ1 = 0.3, go next.

Fig. 3f, the 6th iteration,

λ = 0.25,λ1 = 0.3,Nλ1 = 22 > Nλ = 0, then update λ = 0.3,λ1 = 0.35, go next.

Fig. 3g, the 7th iteration,

λ = 0.3,λ1 = 0.35,Nλ1 = 2 < Nλ = 24, end loop, go next.

(5) The current λ = 0.3 is the optimal value, and it is used in the optimization algorithm (ASHOA), and the procedure stops.

1.2 ZDT 6

(1) λ = 0, the growth rate Δλ = 0.05, λ1 = λ + Δλ = 0.05.

(2) λ1 < 1, go next.

(3) & (4)

Fig. 3a, the 1st iteration,

λ = 0,λ1 = 0.05,Nλ1 = 6 > Nλ = 0, then update λ = 0.05 λ1 = 0.1, go next.

Fig. 3b, the 2nd iteration,

λ = 0.05,λ1 = 0.1,Nλ1 = 10 > Nλ = 1, then update λ = 0.1,λ1 = 0.15, go next.

Fig. 3c, the 3rd iteration,

λ = 0.1,λ1 = 0.15,Nλ1 = 1 < Nλ = 8, end loop, go next.

(5) The current λ = 0.1 is the optimal value, and it is used in the optimization algorithm (ASHOA), and the procedure stops.

Appendix 3

The optimal λ found by the procedure (see Section 2.3.3) for test cases ZDT1, ZDT2, and ZDT4 are shown in Fig. 29.

Appendix 4

In Section 2.1, the proportion of individuals to carry out gradient mutation is specified to be 1/q × 100% in step (5) of the new gradient mutation algorithm. This threshold value is an important parameter which may influence the global optimization performance and convergence efficiency of the optimization algorithm. Therefore, sensitivity analysis of the proportion is investigated with the same test cases (ZDT1 to ZDT6, except for ZDT5).

In this paper, the proportion is 1/q × 100%, where q is the number of objectives in optimization. For multi-objective optimization problems, q usually equals 2–4. In the study, the five mathematical test cases all have two objectives (1/q = 0.50). Four types of proportion (0.25, 0.50, 0.75, 1.00) are imposed for the optimization of the five test cases. The Pareto front solutions are searched over the same period with different proportions, and the results are plotted in Fig. 30. Overall, the global optimization performance and convergence efficiency are not very sensitive to the proportion between 0.25 to 1.00 for all test cases, particularly for ZDT2, ZDT3, and ZDT5. If the Pareto front solutions are compared carefully, it is found that both convergence efficiency and global optimization performance decrease slightly under small or large proportions.

Fig. 30
figure 30

The comparison of Pareto front solutions searched with different individuals’ proportion over the same period

Full size image

This can be explained by the following facts. When a small proportion is imposed, few individuals are selected to carry out gradient mutation, and some individuals with large gradient values are ignored. Therefore, the convergence efficiency may suffer. On the other hand, if the proportion is large, more individuals are chosen to implement gradient mutation. Those individuals which have very small gradient (near zero) are also selected to conducted gradient mutation. It will increase the computational resources and further decrease convergence efficiency. Additionally, the gradient mutation of small gradient individuals has tiny contributions to convergence. By contrast, it cannot protect excellent individuals, leading to the poorer global optimization performance and convergence efficiency.

As shown in Section 2.3.3, the normalized gradient mutation step size λ is selected adaptively during the optimization process. When the proportion of individuals for gradient mutation is specified, the ASHOA will determine a suitable λ to balance the high convergence efficiency and good global optimization performance. Taking ZDT1 as an example, when the proportion is 0.25, the optimal λ = 0.9 while when the proportion is 0.50 the optimal λ = 0.8. The effect of the varying proportion on the optimization algorithm is counteracted by adaptive λ. Therefore, the specific proportion (1/q) for the proposed algorithm ASHOA is reasonable and robust.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Li, X., Zhao, Y. & Liu, Z. A novel global optimization algorithm and data-mining methods for turbomachinery design. Struct Multidisc Optim 60, 581–612 (2019). https://doi.org/10.1007/s00158-019-02227-5

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00158-019-02227-5

Keywords

Search

Navigation