Abstract
In this study, we develop a design methodology with a basis in gradient-based topology optimization and a geometrical reduced-order thermal/hydraulic model for actively cooled microvascular composite panels. The proposed method is computationally very efficient owing to the suggested simplifications while preserving the required accuracy. The analytical sensitivity for the topology optimization scheme is derived. Several numerical examples are solved to demonstrate the applicability of the proposed method for active-cooling applications. Using topology optimization, the maximum temperature of the composite panel is reduced by up to 59% compared to a benchmark design. The optimization framework is compared to hybrid topology/shape (HyTopS) and shape optimization (SO) methods based on several measures such as maximum and average temperatures, temperature uniformity, network redundancy, and manufacturability. The solution obtained from the proposed TO scheme outperforms the other approaches in terms of the aforementioned measures.
Similar content being viewed by others
Notes
In the IGFEM framework, the velocity field is defined as the partial derivative of the nodal coordinates with respect to the design variables.
Note that for this problem, it is not possible to have a uniform design which simultaneously satisfies both the volume fraction and pressure drop constraints.
References
Achtziger W, Stolpe M (2009) Comput Optim Appl 44(2):315
Alexandersen J, Aage N, Andreasen CS, Sigmund O (2014) Int J Numer Methods Fluids 76(10):699
Aragón AM, Saksena R, Kozola BD, Geubelle PH, Christensen KT, White SR (2013) J Comput Phys 233:132
Asadpoure A, Tootkaboni M, Guest JK (2011) Comput Struct 89(11-12):1131
Asmussen J, Alexandersen J, Sigmund O, Andreasen CS (2019) Struct Multidiscip Optim 59(4):1105
Bahrami M, Yovanovich M, Culham J (2005) In: ASME 3rd international conference on microchannels and minichannels. American Society of Mechanical Engineers, pp 269–280
Bejan A (1997) Int J Heat Mass Transfer 40(4):799
Borrvall T, Petersson J (2003) Int J Numer Methods Fluids 41(1):77
Brooks AN, Hughes TJ (1982) Comput Methods Appl Mech Eng 32(1-3):199
Coppola A, Griffin A, Sottos N, White S (2015) Compos Part A Appl Sci Manuf 78:412
Dilgen SB, Dilgen CB, Fuhrman DR, Sigmund O, Lazarov BS (2018) Struct Multidiscip Optim 57(5):1905
Dong X, Liu X (2020) Numer Heat Transfer Part A Appl 77(1):90
Dong X, Liu X (2021) Chem Eng J 412:128594
Dunavant D (1985) Int J Numer Methods Eng 21(6):1129
Esser-Kahn AP, Thakre PR, Dong H, Patrick JF, Vlasko-Vlasov VK, Sottos NR, Moore JS, White SR (2011) Adv Mater 23(32):3654
Gersborg-Hansen A, Bendsøe MP, Sigmund O (2006) Struct Multidiscip Optim 31(4):251
Guest JK, Prévost JH (2006) Int J Numer Methods Eng 66(3):461
Hadad Y, Pejman R, Ramakrishnan B, Chiarot PR, Sammakia BG (2018) In: 2018 17th IEEE Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems (ITherm) IEEE, pp 673–682
Hadad Y, Ramakrishnan B, Pejman R, Rangarajan S, Chiarot PR, Pattamatta A, Sammakia B (2019) Appl Therm Eng 150:720
Hadad Y, Rangararajan S, Nemati K, Ramakrishnann B, Pejman R, Chiarot PR, Sammakia B (2020) Int J Therm Sci 148:106145
Jarrett A, Kim IY (2011) J Power Sources 196(23):10359
Kays WM, Crawford ME, Weigand B (2005) Convective heat and mass transfer. McGraw-Hill
Koga AA, Lopes ECC, Nova HFV, De Lima CR, Silva ECN (2013) Int J Heat Mass Transfer 64:759
Lee JS, Yoon SY, Kim B, Lee H, Ha MY, Min JK (2021) Int J Heat Mass Transfer 172:121172
Li P, Liu Y, Zou T, Huang J (2017) Adv Mech Eng 9(7):1687814017708175
Marck G, Nemer M, Harion JL (2013) Numer Heat Transfer Part B Fundamen 63(6):508
MATLAB (2015) version 8.5 (R2015a) The MathWorks Inc., Natick Massachusetts
Matsumori T, Kondoh T, Kawamoto A, Nomura T (2013) Struct Multidiscip Optim 47 (4):571
Najafi AR, Safdari M, Tortorelli DA, Geubelle PH (2015) Comput Methods Appl Mech Eng 296:1
Najafi AR, Safdari M, Tortorelli DA, Geubelle PH (2017) Int J Numer Methods Eng 111(10):927
Najafi AR, Safdari M, Tortorelli DA, Geubelle PH (2020) Int J Numer Methods Eng
Olesen LH, Okkels F, Bruus H (2006) Int J Numer Methods Eng 65(7):975
Ozguc S, Pan L, Weibel JA (2021) Int J Heat Mass Transfer 169:120896
Pejman R, Najafi AR (2020) In: ASME 2020 international design engineering technical conferences and computers and information in engineering conference. American Society of Mechanical Engineers Digital Collection
Pejman R, Aboubakr SH, Martin WH, Devi U, Tan MHY, Patrick JF, Najafi AR (2019) Int J Heat Mass Transfer 144:118606
Pejman R, Maghami E, Najafi AR (2020) Appl Therm Eng 181:115916
Pejman R, Keshavarzzadeh V, Najafi AR (2021) Comput Methods Appl Mech Eng 375:113624
Petersson J, Sigmund O (1998) Int J Numer Methods Eng 41(8):1417
Pety SJ, Tan MHY, Najafi AR, Barnett PR, Geubelle PH, White SR (2017) Int J Heat Mass Transfer 115:513
Pety SJ, Tan MHY, Najafi AR, Gendusa AC, Barnett PR, Geubelle PH, White SR (2018) Appl Therm Eng 131:965
Rodrigues H, Fernandes P (1995) Int J Numer Methods Eng 38(12):1951
Safdari M, Najafi AR, Sottos NR, Geubelle PH (2016) J Comput Phys 318:373
Shah R, London A (1978) Laminar flow forced convection in ducts (a sourcebook for compact heat transfer exchange analytical data)
Soghrati S, Najafi AR, Lin JH, Hughes KM, White SR, Sottos NR, Geubelle PH (2013) Int J Heat Mass Transfer 65:153
Tan MHY, Safdari M, Najafi AR, Geubelle PH (2015) Comput Methods Appl Mech Eng 283(1):1382
Tan MHY, Najafi AR, Pety SJ, White SR, Geubelle PH (2016) Int J Heat Mass Transfer 103:594
Tan MHY, Najafi AR, Pety SJ, White SR, Geubelle PH (2018) Appl Therm Eng 135:145
Wiker N, Klarbring A, Borrvall T (2007) Int J Numer Methods Eng 69(7):1374
Yaji K, Ogino M, Chen C, Fujita K (2018) Struct Multidiscip Optim 58(2):817
Yan S, Wang F, Sigmund O (2018) Int J Heat Mass Transfer 122: 660
Yan S, Wang F, Hong J, Sigmund O (2019) Int J Heat Mass Transfer 143:118462
Yoon GH (2010) J Mech Sci Technol 24(6):1225
Zhao X, Zhou M, Sigmund O, Andreasen CS (2018) Int J Heat Mass Transfer 116:1108
Zhou M, Lian H, Sigmund O, Aage N (2018) Int J Numer Methods Fluids 88(6):296
Acknowledgements
This work has been supported by Drexel University Career Development and Summer Research Awards. The authors acknowledge the high-performance computing resources (PROTEUS: the Drexel Cluster) and support at the Drexel University. The authors also acknowledge the support from the Villum Foundation through the VILLUM Investigator Project InnoTop.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare no competing interests.
Additional information
Responsible Editor: Hyunsun Alicia Kim
Replication of results
We provided comprehensive implementation details, as a result, we are confident that the design methodology is reproducible. For this reason, we have decided not to publish the code. If the information provided in the paper is not sufficient, the readers are welcome to contact the authors for further explanations.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix 1
In this study, we aim to minimize the maximum temperature of the domain. However, the maximum temperature is not differentiable and as is mentioned in Section 3, we replace it with a differentiable alternative, i.e., p-mean temperature. The p-mean temperature converges to the maximum temperature when p approaches infinity, i.e., \(\lim _{p\rightarrow \infty }\vert \vert T \vert \vert _{p}=T_{max}\). However, using large values for p can be problematic. First of all, since we use the Gauss-Dunavant quadrature (Dunavant 1985) to precisely calculate the integration of p-mean temperature, the larger values of p require more quadrature points which increase the computational cost. Moreover, the large values of p make the p-mean temperature ill-conditioned or less smooth. Small values of p can also be problematic. Choosing a small value for p may not enable ||T||p to capture reliably the trend in Tmax. This issue may lead to local areas of high temperature similar to the stress concentration regions in the structural optimization problems.
In this section, we consider the reference design of CHF1 problem in Section 4.1. The maximum temperature of the panel in this design is 76.1°C. We evaluate the value of p-mean temperature using p equal to 2, 5, 8, 10, 12, 17, and 19. The numbers of quadrature points that we need to consider based on Gauss-Dunavant quadrature (Dunavant 1985) are 3, 7, 16, 33, 61, and 73 for p equal to 2, 5, 8, 10, 12, 17, and 19, respectively. To compute the percent of the difference between the maximum temperature and p-mean temperature, we define a variable Td = (||Tp||− Tmax)/Tmax. Figure 17 compares the values of percent of the difference between the p-mean and maximum temperatures(Td) and the number of quadrature points (NQ) for different values of p. Based on the results, we decided to select p = 10 in this study. Note that we did not observe the problems of smoothness and being ill-conditioned for ||Tp|| with p = 10 in the problems solved in this study.
Appendix 2
As explained in the main text, the major role of Dmin in (10) is preventing numerical issues. But we need to make sure that the value that we choose for Dmin would not adversely impact the objective value. Thus, to find a suitable value for Dmin, we considered the reference design of each problem and we computed the amount of error produced in the value of the objective function when we consider a nonzero value for minimum diameter. We define the error as \(e=(|\theta ^{(@ D_{min}=0)}-\theta ^{(@ D_{min}\neq 0)}|)/\theta ^{(@ D_{min}=0)}\). Obviously, if the minimum diameter is 0, the amount of error will be 0 and as the minimum diameter increases, the amount of error rises. Figure 18 shows the results for the reference designs of UHF1 and CHF1. Based on the presented results, we select 60 μm and 40 μm for the minimum diameter of the problems UHF1 and CHF1, respectively. The amount of error (e) until these two values are less than 0.01%.
Appendix 3
In this section, we aim to investigate the effect of removing the microchannels with the flow rates lower than a certain threshold on the objective function for the optimization problem UHF1 solved in Section 4.1. For that problem, we select a mass flow rate threshold of 3 ⋅ 10− 3 gr/s. We reevaluate the p-mean temperature value for all of the designs obtained in each iteration of the optimization problem UHF1 before removing the microchannels. Figure 19 compares the p-mean temperature values obtained from the cases where we remove the low mass flow rate microchannels with the cases where we do not remove any microchannel. The results are very close to each other. Thus, we can conclude that removing the microchannels with the mass flow rates lower than the prescribed threshold has a negligible effect on the p-mean temperature value.
Appendix 4
A verification study of the analytic adjoint sensitivity analysis is performed by comparing the computed sensitivity with the central finite difference method for the reference design of the problem UHF1 in Section 4.1. The error between the adjoint and finite difference sensitivity analysis is given by \(\epsilon =\mid \frac {((d\theta /d\alpha )^{Adj}-(d\theta /d\alpha )^{FD})}{(d\theta /d\alpha )^{Adj}}\mid \), where ∣ ∗∣ indicates the absolute value of *, and superscripts of Adj and FD indicate adjoint method and finite difference approach, respectively. The error is plotted in Fig. 20 for a sequence of perturbations from △α = 10− 2 to 10− 12. The amount error is very small (≈ 10− 11) and we can conclude that the sensitivity analysis is correctly derived and implemented in this study.
Rights and permissions
About this article
Cite this article
Pejman, R., Sigmund, O. & Najafi, A.R. Topology optimization of microvascular composites for active-cooling applications using a geometrical reduced-order model. Struct Multidisc Optim 64, 563–583 (2021). https://doi.org/10.1007/s00158-021-02951-x
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00158-021-02951-x