Pith. sign in

REVIEW 3 major objections 5 minor 91 references

TOBACO: Topology Optimization via Band-limited Coordinate Networks for Compositionally Graded Alloys

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read By tuning a coordinate network's frequency band, topology optimization can enforce composition-gradation limits implicitly, without explicit constraint equations.

desk verdict A genuinely nice use of Bernstein's inequality to map gradation limits to network bandwidth, but the paper overstates the implicit guarantee because it never keeps the input frequencies inside their band or bounds the field pointwise. read the letter →

arxiv 2508.10320 v1 pith:5HC3IZTJ submitted 2025-08-14 cs.CE cs.NAmath.NA

classification cs.CEcs.NAmath.NA
keywords topologyoptimizationcompositionallygradedalloysband-limitedcoordinatenetworksgradationconstraintadditivemanufacturingthermo-elasticdesignFourierbandwidthmultiplicativefilter
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that the limiting manufacturing constraint for compositionally graded alloys—the maximum allowed spatial rate of composition change—can be enforced automatically by representing the composition field with a band-limited coordinate network. The authors link the manufacturing limit $L_{\max}$ to a Fourier bandwidth $B=1/(2\pi L_{\max})$ via a classical signal-processing bound: a field whose spectrum vanishes outside $[-B,B]$ cannot have a spatial gradient exceeding $2\pi B$. They build a topology-optimization framework around a multiplicative-filter coordinate network whose input-layer frequencies set the bandwidth, so the gradation constraint is satisfied by construction rather than by adding constraint equations. Numerical experiments on 2D thermo-elastic problems, anisotropic gradation limits, and a 3D turbine blade show optimized compositions staying within the imposed band and performance improving as the band is widened. A sympathetic reader would care because the representation is continuous, mesh-independent, and differentiable, removing the standard discretization problems of element-based gradation constraints.

What carries the argument

The central object is a band-limited coordinate network: a multiplicative filter network in which spatial coordinates pass through sine layers with tunable frequencies $\omega_i \in (-B_i, B_i)$ and phases $\phi_i\in(-\pi,\pi)$, with $\sum_i B_i=B$. Hadamard products of these sinusoids generate an exponential number of sine bases from a polynomial number of parameters, and the output layer maps the last hidden state to component fractions. The network carries the argument because its output is band-limited by construction, so Bernstein's inequality converts the chosen bandwidth directly into a bound on the spatial derivative of composition—no explicit gradation constraint is needed.

What would settle it

Take the paper's 2D validation problem and run it to convergence; then query the final network on a fine grid and compute both the maximum input-layer frequency and the maximum value of $|\partial\rho/\partial x|$. If any converged frequency lies outside its initialized interval, or if the maximum gradient exceeds $1/L_{\max}$ anywhere, then the trained network is not band-limited to the intended $B$ and the implicit-constraint claim does not hold for the actual optimized design.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a topology-optimized composition field can be made gradation-constrained by construction. For each component fraction $\rho^{(j)}$ bounded in $[0,1]$, Bernstein's inequality gives $|d\rho^{(j)}/dx|\le 2\pi B$ whenever the field is band-limited to $B$; choosing $B=1/(2\pi L_{\max})$ therefore guarantees the desired limit $|\partial\rho^{(j)}/\partial x|\le 1/L_{\max}$. The realization is a coordinate network with sinusoidal input layers whose frequencies are initialized inside intervals that sum to the target bandwidth, making the network output band-limited. Physical bounds on the fractions are enforced by explicit log-barrier constraints rathe

Load-bearing premise

The gradation guarantee holds only if the network's input-layer frequencies stay inside the prescribed intervals $(-B_i,B_i)$ throughout optimization; the paper initializes them there but never states that they are fixed or projected back after each gradient update.

Editorial extensions

If this is right

  • Gradation limits become a knob on the network rather than a constraint set: changing $B$ directly changes the maximum allowed spatial rate of composition change.
  • Because the design is a continuous field, it can be evaluated at any resolution, decoupling design complexity from the finite-element mesh and enabling coarse-mesh optimization with high-resolution design extraction.
  • Anisotropic manufacturing limits, as in processes that allow faster gradation along the build direction than in-plane, are handled by setting different bandwidths per axis.
  • Relaxing the bandwidth monotonically enlarges the design space, and the paper reports compliance decreasing as $B$ increases (37.3, 28.6, 25.2 for $B=2,5,10$; relative compliances 1, 0.76, 0.62 in the turbine-blade study).
  • The framework is end-to-end differentiable, so sensitivities of objectives and constraints are obtained by automatic differentiation rather than hand-derived adjoint equations.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the same bandwidth-to-gradient mechanism applies to any field represented by the network, so the method could serve as a generic smoothness or length-scale constraint in topology optimization, not only for alloy composition.
  • Testable extension: track the input-layer frequencies during training and, at convergence, evaluate the maximum composition gradient on a fine grid; this checks whether the trained network (not just the initialized architecture) still satisfies the limit.
  • The paper's own conclusion flags that the material model assumes an idealized composition–property relationship and that experimental fabrication and testing remain outstanding; the gradation guarantee is therefore a property of the design representation, and its practical payoff depends on closing that gap.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes TOBACO, a topology-optimization framework for compositionally graded alloys in which the composition field is represented by a band-limited multiplicative-filter coordinate network. The maximum composition gradation is formulated in Fourier space: by Bernstein's inequality, a bandlimit B = 1/(2πLmax) implies |∇ρ| ≤ 1/Lmax if also |ρ| ≤ 1. The network's input-layer frequencies are initialized inside intervals (−Bi, Bi), and the sum of the intervals is set to B, so the representation is argued to satisfy the gradation constraint implicitly. The optimization minimizes compliance (including thermo-elastic problems) with mass, bound, and partition-of-unity constraints handled by log-barrier penalties; automatic differentiation and Adam update all network weights. Experiments include 2D isotropic and anisotropic bandlimits and a 3D turbine-blade example, showing that larger bandwidths lead to more detailed designs and lower compliance. The paper claims this removes the need for explicit gradation constraints and yields mesh-independent, end-to-end differentiable design.

Significance. Implicit constraint enforcement via bandlimited coordinate networks is a potentially valuable idea for AM-aware topology optimization. The continuous representation, automatic differentiation, and ability to prescribe anisotropic limits are attractive, and the numerical examples show expected compliance trends. However, the force of the contribution rests on two invariants—the input-layer frequencies staying within their prescribed bands and the composition being pointwise bounded—and the current manuscript does not secure either. The experimental validation is also largely a check of the construction rather than a test of the mechanism. These are fixable in a revision by freezing/projecting frequencies, adding pointwise bounds, and reporting frequency/gradation diagnostics. If fixed, the method would be a solid contribution to the field.

major comments (3)
  1. [§3.3, §3.8, §3.9] The central bandlimited invariant is not preserved by the implemented optimization. The ω_i are introduced as tunable input-layer parameters and are included in 'the weights of the network' updated by Adam (§3.8–§3.9). The paper nowhere fixes them after initialization, nor projects/clamps them into (−B_i, B_i) after a gradient step. Adam moves every parameter generically; once any |ω_i| leaves its interval, the Hadamard-product construction can generate output frequencies above the nominal sum B, so Theorem 1 no longer bounds the gradient. This defeats the claimed 'implicit compliance'. Please either treat ω_i as non-trainable constants, reparameterize as ω_i = B_i tanh(θ_i), or clamp after each update, and separately report the actual frequency support/trajectory.
  2. [§3.7 and Theorem 1 (§3.3)] Theorem 1 requires the pointwise bound |ρ(x)| ≤ 1 on all of R, but the implementation only enforces ρ ∈ [0,1] at element centers through the soft LSE/log-barrier constraints (Eqs. 9–11, 14–15). The network output is unbounded (§3.3). Even if the sampled constraints hold at convergence, a bandlimited interpolant can overshoot between sample points; the derivative bound then need not hold, and the phrase 'physically valid' in §3.7 is too strong. Use an architecture or hard parameterization that gives pointwise bounds while preserving bandlimitedness, or provide dense numerical evidence and clearly restrict the claim to the sampled mesh.
  3. [§4.1] The validation of the central hypothesis is close to a consistency check rather than an external test. The network is constructed to be bandlimited, so the FFT and gradient checks confirm the construction rather than demonstrating that gradation control is achieved under optimization. This would be convincing if the design variables that control bandwidth are monitored and reported, and if the maximum gradation is checked on an unsampled dense grid and compared with an independent explicit-constraint or filtered baseline. Please add such diagnostics, especially the evolution and final values of ω_i, to substantiate 'ensures implicit compliance'.
minor comments (5)
  1. [§3.3] Notation inconsistency: the text says the design must be bandlimited with 'a maximum frequency of 2πB' after defining B = 1/(2πLmax). Clarify whether B is a radian frequency or a cyclic frequency, and state the corresponding Fourier-support interval precisely.
  2. [§3.3, Eq. (4)] The output layer is described as having S−1 neurons, but W_out is written as R^{dh×S} and the composition vector has S components. This dimension mismatch should be corrected, especially since the partition-of-unity constraint is then enforced separately.
  3. [§3.8, Eqs. (14)–(15)] The penalty function is written as ψ but used as ψ_τ in the text; the iteration update τ = τ0 μ^k appears only in prose. Define these symbols explicitly and consistently.
  4. [§4.2] For anisotropic bandlimits (Bx, By) = (2,8) and (8,2), the paper does not define how these limits relate to Lmax in each direction or how they are encoded in the input-layer frequencies ω_i. A precise mapping would help readers implement the method.
  5. [Replication of Results] The statement 'implementation will be made available upon reasonable request' is weak for a computational methods paper. A public repository with the code and data would substantially improve reproducibility and allow verification of the frequency-handling issue raised above.

Circularity Check

1 steps flagged · score 2.0 of 10

No load-bearing circularity: the gradation guarantee is an external theorem plus a bandlimited-network construction; the Sec. 4.1 gradient check is a self-fulfilling sanity check, not an independent prediction, and the missing frequency-bound invariant is a soundness gap, not a circular reduction.

  1. self definitional [Section 4.1 ("Validation of Hypothesis"), building on Sections 3.3 and 3.9]
    "To evaluate the validity of the imposed gradation constraint, we compute the gradations by backpropagating through the network. The computed gradations, presented in Figure 8(d) and (e), clearly adhere to the imposed constraint, thereby confirming our central hypothesis."

    The network is constructed so that its input-layer frequencies lie within (-B_i,B_i) with the sum of B_i equal to B (Section 3.3), which makes the output bandlimited. By Bernstein's inequality (Theorem 1), this construction already forces |grad rho| <= 2*pi*B = 1/L_max. The Section 4.1 validation recomputes the gradients and observes the bound, so it is a numerical restatement of the construction rather than an independent empirical test. This is a self-fulfilling check, but it is not load-bearing for the paper's actual results: the optimized compliance values, the bandwidth-vs-performance trends, and the mass/bound constraints are all independent of this redundant verification.

full rationale

The central claim—that a bandlimited coordinate network can enforce a maximum composition gradation—is derived from Bernstein's inequality (an external textbook result) plus the multiplicative-filter-network construction. The design is explicitly built so that the sum of input-layer bandwidths equals B = 1/(2*pi*L_max), so the gradient bound follows mathematically. This is not a case where a fitted parameter is renamed as a prediction, nor where a self-citation supplies the load-bearing uniqueness or ansatz. The paper's self-citations ([23,51,63,88]) are used for the general neural-network TO representation and automatic differentiation, not to establish the gradation theorem. The only circular-flavored passage is the Section 4.1 sanity check, which confirms the gradient bound that was already guaranteed by construction; this is a presentational redundancy, not a logical circularity in the optimization. A more serious concern is non-circular: the input-layer frequencies omega_i are described as 'tunable' and all network weights are updated by Adam (Sections 3.3, 3.8, 3.9), but the paper never states that omega_i are fixed or projected into (-B_i,B_i). If the optimizer moves a frequency outside that interval, the output is no longer bandlimited and the implicit gradation guarantee fails. This is a missing invariant/omitted proof, not a circular reduction, so it is weighed here as a soundness caveat rather than an additional circularity. Overall score 2: one self-fulfilling validation and multiple non-load-bearing self-citations, but the core derivation is independent and externally grounded.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on Bernstein's inequality and on the bandlimited-network analysis inherited from BACON. The additional hand-chosen parameters are optimizer hyperparameters. The main unstated assumption is that element-center bound enforcement gives pointwise validity, which is needed for the Bernstein bound to apply with M=1.

free parameters (4)
  • Bandwidth split per layer B_i = not given; user-chosen
    The total bandwidth B is derived from Lmax, but how B is split into B_i across layers is a free choice that affects the design but not the gradation bound.
  • LSE scaling factor t = 10
    Chosen by hand in Eq. (11) to approximate max/min constraints.
  • Log-barrier parameters tau0 and mu = tau0=3, mu=1.04
    Chosen by hand in Section 3.8 to control the penalty continuation schedule.
  • Network size and optimizer hyperparameters = 3 hidden layers, 100 neurons, lr=1e-2, gradient clip 1
    These affect convergence and design quality but not the central constraint mechanism.
assumptions (5)
  • standard math Bernstein's inequality: a bandlimited function with magnitude bound M has derivative bounded by 2*pi*B*M.
    Invoked in Section 3.3 to link bandwidth to maximum gradient; standard result cited as [22].
  • domain assumption The multiplicative filter network output is bandlimited to the sum of the input-layer bandwidths.
    The paper relies on the spectral analysis of MFNs from [72,24] without proving it; this is the foundation of the gradation-control guarantee.
  • domain assumption Material properties at arbitrary compositions can be accurately obtained via RBF interpolation from a discrete set of property data.
    Used in Section 3.4; the paper defines properties at sample compositions and interpolates everywhere, an idealization acknowledged in the conclusion.
  • standard math The multigrid FE solver (PyAMG/JAX-FEM) accurately solves the thermal and structural PDEs on the chosen mesh.
    Used in Section 3.5; assumes standard FEM convergence and sufficient mesh resolution.
  • domain assumption Bounds and partition-of-unity constraints enforced at element centers imply physical validity everywhere in the domain.
    Section 3.7 aggregates constraints at element centers only; the claim that the design is physically valid pointwise is not proven, and Bernstein's M=1 bound requires pointwise |rho|<=1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TOBACO: Topology Optimization via Band-limited Coordinate Networks for Compositionally Graded Alloys." pith.science (2026). https://pith.science/paper/5HC3IZTJ

@misc{pith2026250810320,
  author       = {Pith},
  title        = {Pith review of: TOBACO: Topology Optimization via Band-limited Coordinate Networks for Compositionally Graded Alloys},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5HC3IZTJ}},
  note         = {Machine review of arXiv:2508.10320}
}
read the original abstract

Compositionally Graded Alloys (CGAs) offer unprecedented design flexibility by enabling spatial variations in composition; tailoring material properties to local loading conditions. This flexibility leads to components that are stronger, lighter, and more cost-effective than traditional monolithic counterparts. The fabrication of CGAs have become increasingly feasible owing to recent advancements in additive manufacturing (AM), particularly in multi-material printing and improved precision in material deposition. However, AM of CGAs requires imposition of manufacturing constraints; in particular limits on the maximum spatial gradation of composition. This paper introduces a topology optimization (TO) based framework for designing optimized CGA components with controlled compositional gradation. In particular, we represent the constrained composition distribution using a band-limited coordinate neural network. By regulating the network's bandwidth, we ensure implicit compliance with gradation limits, eliminating the need for explicit constraints. The proposed approach also benefits from the inherent advantages of TO using coordinate networks, including mesh independence, high-resolution design extraction, and end-to-end differentiability. The effectiveness of our framework is demonstrated through various elastic and thermo-elastic TO examples.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

91 extracted references · 73 canonical work pages

  1. [2]

    Novel designs of turbine blades for additive manufacturing

    Liubov Magerramova, Boris Vasilyev, and Vladimir Kinzburskiy. Novel designs of turbine blades for additive manufacturing. In Turbo Expo: Power for Land, Sea, and Air , volume 49804, page V05CT18A001. American Society of Mechanical Engineers, 2016

  2. [3]

    Thermal barrier coatings overview: Design, manufacturing, and applications in high-temperature industries

    Kunal Mondal, Luis Nunez III, Calvin M Downey, and Isabella J Van Rooyen. Thermal barrier coatings overview: Design, manufacturing, and applications in high-temperature industries. Industrial & Engineering Chemistry Research, 60(17):6061– 6077, 2021

  3. [4]

    Simultaneous optimization of the material properties and the topology of functionally graded structures

    Qi Xia and Michael Yu Wang. Simultaneous optimization of the material properties and the topology of functionally graded structures. Computer-Aided Design, 40(6):660–675, 2008

  4. [5]

    Simultaneous optimisation of structural topology and material grading using level set method

    PD Dunning, CJ Brampton, and HA Kim. Simultaneous optimisation of structural topology and material grading using level set method. Materials Science and Technology, 31(8):884–894, 2015

  5. [6]

    Optimal design of compliant mechanisms using functionally graded materials

    Cian Conlan-Smith, Anurag Bhattacharyya, and Kai A James. Optimal design of compliant mechanisms using functionally graded materials. Structural and Multidisciplinary Optimization, 57:197–212, 2018

  6. [7]

    Jikai Liu, Andrew T Gaynor, Shikui Chen, Zhan Kang, Krishnan Suresh, Akihiro Takezawa, Lei Li, Junji Kato, Jinyuan Tang, Charlie CL Wang, Lin Cheng, Xuan Liang, and Albert. C. To. Current and future trends in topology optimization for additive manufacturing. Structural and multidisciplinary optimization, 57(6):2457–2483, 2018

  7. [8]

    Metaset: Exploring shape and property spaces for data-driven metamaterials design

    Yu-Chin Chan, Faez Ahmed, Liwei Wang, and Wei Chen. Metaset: Exploring shape and property spaces for data-driven metamaterials design. Journal of Mechanical Design, 143(3):031707, 2021

  8. [9]

    Remixing functionally graded structures: data-driven topology optimization with multiclass shape blending

    Yu-Chin Chan, Daicong Da, Liwei Wang, and Wei Chen. Remixing functionally graded structures: data-driven topology optimization with multiclass shape blending. Structural and Multidisciplinary Optimization, 65(5):135, 2022

Show all 91 references
  1. [10]

    A review on functionally graded materials and structures via additive manufacturing: from multi-scale design to versatile functional properties

    Yan Li, Zuying Feng, Liang Hao, Lijing Huang, Chenxing Xin, Yushen Wang, Emiliano Bilotti, Khamis Essa, Han Zhang, Zheng Li, et al. A review on functionally graded materials and structures via additive manufacturing: from multi-scale design to versatile functional properties. ...

  2. [11]

    On-the-fly path planning for the design of compositional gradients in high dimensions

    Samuel Price, Zhaoxi Cao, and Ian McCue. On-the-fly path planning for the design of compositional gradients in high dimensions. Materials & Design, page 113983, 2025

  3. [12]

    Functionally graded materials manufactured by direct energy deposition: a review

    D Dev Singh, Suresh Arjula, and A Raji Reddy. Functionally graded materials manufactured by direct energy deposition: a review. Materials Today: Proceedings, 47:2450–2456, 2021

  4. [13]

    Simulation and validation of three dimension functionally graded materials by material jetting

    Eduardo Salcedo, Dongcheon Baek, Aaron Berndt, and Jong Eun Ryu. Simulation and validation of three dimension functionally graded materials by material jetting. Additive Manufacturing, 22:351–359, 2018

  5. [14]

    Additive manufacturing of heat exchangers in aerospace applications: a review

    Francesco Careri, Raja HU Khan, Catherine Todd, and Moataz M Attallah. Additive manufacturing of heat exchangers in aerospace applications: a review. Applied Thermal Engineering, page 121387, 2023. xvi

  6. [15]

    Innovative design, structural optimization, and additive manufacturing of new-generation turbine blades

    Lorenzo Pinelli, Andrea Amedei, Enrico Meli, Federico Vanti, Benedetta Romani, Gianluigi Benvenuti, Marco Fabbrini, Nicolò Morganti, Andrea Rindi, and Andrea Arnone. Innovative design, structural optimization, and additive manufacturing of new-generation turbine blades. Journa...

  7. [16]

    Additive manufacturing of liquid rocket engine combustion devices: a summary of process developments and hot-fire testing results

    Paul R Gradl, Sandy E Greene, Christopher Protz, Brad Bullard, James Buzzell, Chance Garcia, Jessica Wood, Robin Osborne, James Hulka, and Kenneth G Cooper. Additive manufacturing of liquid rocket engine combustion devices: a summary of process developments and hot-fire testin...

  8. [17]

    In-situ blended 316l-si and ph48s via laser directed energy deposition for functionally graded applications

    Rujing Zha, Nhung Thi-Cam Nguyen, Gregory B Olson, and Jian Cao. In-situ blended 316l-si and ph48s via laser directed energy deposition for functionally graded applications. CIRP Annals, 73(1):137–140, 2024

  9. [19]

    Design and fabrication of functionally graded material from ti toγ-tial by laser metal deposition

    Xueyang Chen, Lei Yan, Joseph William Newkirk, and Frank W Liou. Design and fabrication of functionally graded material from ti toγ-tial by laser metal deposition. 2017

  10. [20]

    Yi Su, Bo Chen, Caiwang Tan, Xiaoguo Song, and Jicai Feng. Influence of composition gradient variation on the microstruc- ture and mechanical properties of 316 l/inconel718 functionally graded material fabricated by laser additive manufacturing. Journal of Materials Processing...

  11. [21]

    Design methodology for functionally graded materials: Framework for considering cracking

    Zhening Yang, Hui Sun, Zi-Kui Liu, and Allison M Beese. Design methodology for functionally graded materials: Framework for considering cracking. Additive Manufacturing, 73:103672, 2023

  12. [22]

    A foundation in digital communication

    Amos Lapidoth. A foundation in digital communication. Cambridge University Press, 2017

  13. [23]

    Tounn: Topology optimization using neural networks

    Aaditya Chandrasekhar and Krishnan Suresh. Tounn: Topology optimization using neural networks. Structural and Multidisciplinary Optimization, 63(3):1135–1149, 2021

  14. [24]

    Bacon: Band-limited coordinate networks for multiscale scene representation

    David B Lindell, Dave Van Veen, Jeong Joon Park, and Gordon Wetzstein. Bacon: Band-limited coordinate networks for multiscale scene representation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16252–16262, 2022

  15. [25]

    Topology optimization additive manufacturing-oriented for a biomedical application

    F Cucinotta, E Guglielmino, G Longo, G Risitano, D Santonocito, and F Sfravara. Topology optimization additive manufacturing-oriented for a biomedical application. In Advances on Mechanics, Design Engineering and Manufacturing II: Proceedings of the International Joint Confere...

  16. [26]

    A review of topology optimization for additive manufacturing: Status and challenges

    Zhu Jihong, Zhou Han, Wang Chuang, Zhou Lu, Yuan Shangqin, and Weihong Zhang. A review of topology optimization for additive manufacturing: Status and challenges. Chinese Journal of Aeronautics, 34(1):91–110, 2021

  17. [27]

    Bridging topology optimization and additive manufacturing

    Tomás Zegard and Glaucio H Paulino. Bridging topology optimization and additive manufacturing. Structural and Multidisciplinary Optimization, 53:175–192, 2016

  18. [28]

    Topology optimization approaches: A comparative review

    Ole Sigmund and Kurt Maute. Topology optimization approaches: A comparative review. Structural and multidisciplinary optimization, 48(6):1031–1055, 2013

  19. [29]

    A level set method for structural topology optimization

    Michael Yu Wang, Xiaoming Wang, and Dongming Guo. A level set method for structural topology optimization. Computer methods in applied mechanics and engineering, 192(1-2):227–246, 2003

  20. [30]

    A geometry projection method for continuum-based topology optimization with discrete elements

    Julian Norato, BK Bell, and Daniel A Tortorelli. A geometry projection method for continuum-based topology optimization with discrete elements. Computer Methods in Applied Mechanics and Engineering, 293:306–327, 2015

  21. [31]

    Multi-material topology optimization using ordered simp interpolation

    Wenjie Zuo and Kazuhiro Saitou. Multi-material topology optimization using ordered simp interpolation. Structural and Multidisciplinary Optimization, 55:477–491, 2017

  22. [32]

    Topology optimization of thermo-elastic structures with multiple materials under mass constraint

    Tong Gao, Pengli Xu, and Weihong Zhang. Topology optimization of thermo-elastic structures with multiple materials under mass constraint. Computers & Structures, 173:150–160, 2016

  23. [33]

    Thermo-elastic optimization of material distribution of functionally graded structures by an isogeometrical approach

    A Hassanzadeh Taheri, Behrooz Hassani, and NZ Moghaddam. Thermo-elastic optimization of material distribution of functionally graded structures by an isogeometrical approach. International Journal of Solids and Structures, 51(2):416–429, 2014

  24. [34]

    Additive man- ufacturing of porous structures for unmanned aerial vehicles applications

    Helge Klippstein, Hany Hassanin, Alejandro Diaz De Cerio Sanchez, Yahya Zweiri, and Lakmal Seneviratne. Additive man- ufacturing of porous structures for unmanned aerial vehicles applications. Advanced Engineering Materials, 20(9):1800290, 2018

  25. [35]

    Controllable synthesis of a robust sucrose-derived bio-carbon foam with 3d hierarchical porous structure for thermal insulation, flame retardancy and oil absorption

    Shiqi Wu, Daming Chen, Guangdong Zhao, Yuan Cheng, Boqian Sun, Xiaojie Yan, Wenbo Han, Guiqing Chen, and Xinghong Zhang. Controllable synthesis of a robust sucrose-derived bio-carbon foam with 3d hierarchical porous structure for thermal insulation, flame retardancy and oil ab...

  26. [36]

    V oroto: multiscale topology optimization of voronoi structures using surrogate neural networks

    Rahul Kumar Padhy, Krishnan Suresh, and Aaditya Chandrasekhar. V oroto: multiscale topology optimization of voronoi structures using surrogate neural networks. Engineering with Computers, 41(2):1123–1139, 2025

  27. [37]

    Porous sca ffold design for tissue engineering

    Scott J Hollister. Porous sca ffold design for tissue engineering. Nature materials, 4(7):518–524, 2005. xvii

  28. [38]

    Design of graded lattice structures in turbine blades using topology optimization

    Ebrahim Ahmed Ali Alkebsi, Hacene Ameddah, Toufik Outtas, and Abdallah Almutawakel. Design of graded lattice structures in turbine blades using topology optimization. International Journal of Computer Integrated Manufacturing , 34(4):370–384, 2021

  29. [39]

    30 years of functionally graded materials: An overview of manufacturing methods, applications and future challenges

    Bassiouny Saleh, Jinghua Jiang, Reham Fathi, Tareq Al-Hababi, Qiong Xu, Lisha Wang, Dan Song, and Aibin Ma. 30 years of functionally graded materials: An overview of manufacturing methods, applications and future challenges. Composites Part B: Engineering, 201:108376, 2020

  30. [40]

    Multi-material additive manufacturing: A systematic review of design, properties, applications, challenges, and 3d printing of materials and cellular metamaterials

    Aamer Nazir, Ozkan Gokcekaya, Kazi Md Masum Billah, Onur Ertugrul, Jingchao Jiang, Jiayu Sun, and Sajjad Hussain. Multi-material additive manufacturing: A systematic review of design, properties, applications, challenges, and 3d printing of materials and cellular metamaterials...

  31. [41]

    Laser powder bed fusion of compositionally graded cocrmo-inconel 718

    Yaojie Wen, Baicheng Zhang, Ramasubramanian Lakshmi Narayan, Pei Wang, Xu Song, Hao Zhao, Upadrasta Ramamurty, and Xuanhui Qu. Laser powder bed fusion of compositionally graded cocrmo-inconel 718. Additive Manufacturing, 40:101926, 2021

  32. [42]

    Computational design of compositionally graded alloys for property monotonicity

    Tanner Kirk, Richard Malak, and Raymundo Arroyave. Computational design of compositionally graded alloys for property monotonicity. Journal of Mechanical Design, 143(3):031704, 2021

  33. [43]

    A graph algorithm for the design of functionally graded alloy components

    Marshall D Allen, Raymundo Arróyave, and Richard Malak. A graph algorithm for the design of functionally graded alloy components. In International Design Engineering Technical Conferences and Computers and Information in Engineering Conference, volume 88360, page V03AT03A033. ...

  34. [44]

    Additively manufactured ni-20cr to v functionally graded material: Computational predictions and experimental verification of phase formations

    Beril Tonyali, Hui Sun, Brandon Bocklund, John Paul Borgonia, Richard A Otis, Shun-Li Shang, Zi-Kui Liu, and Allison M Beese. Additively manufactured ni-20cr to v functionally graded material: Computational predictions and experimental verification of phase formations. Journal...

  35. [45]

    Topology optimization with manufacturing constraints: A unified projection-based approach

    Sandro L Vatanabe, Tiago N Lippi, Cícero R de Lima, Glaucio H Paulino, and Emílio CN Silva. Topology optimization with manufacturing constraints: A unified projection-based approach. Advances in Engineering Software, 100:97–112, 2016

  36. [46]

    Topology optimization subject to additive manufacturing constraints

    Moritz Ebeling-Rump, Dietmar Hömberg, Robert Lasarzik, and Thomas Petzold. Topology optimization subject to additive manufacturing constraints. Journal of Mathematics in Industry, 11:1–19, 2021

  37. [47]

    Topology optimization of 3d self-supporting structures for additive manufacturing

    Matthijs Langelaar. Topology optimization of 3d self-supporting structures for additive manufacturing. Additive manufactur- ing, 12:60–70, 2016

  38. [48]

    An explicit length scale control approach in simp-based topology optimization

    Weisheng Zhang, Wenliang Zhong, and Xu Guo. An explicit length scale control approach in simp-based topology optimization. Computer Methods in Applied Mechanics and Engineering, 282:71–86, 2014

  39. [49]

    Minimum length scale in topology optimization by geometric constraints

    Mingdong Zhou, Boyan S Lazarov, Fengwen Wang, and Ole Sigmund. Minimum length scale in topology optimization by geometric constraints. Computer Methods in Applied Mechanics and Engineering, 293:266–282, 2015

  40. [50]

    On the design of compliant mechanisms using topology optimization

    Ole Sigmund. On the design of compliant mechanisms using topology optimization. Journal of Structural Mechanics, 25(4):493–524, 1997

  41. [51]

    Approximate length scale filter in topology optimization using fourier enhanced neural networks

    Aaditya Chandrasekhar and Krishnan Suresh. Approximate length scale filter in topology optimization using fourier enhanced neural networks. Computer-Aided Design, 150:103277, 2022

  42. [52]

    Additive manufacturing of functionally graded transition joints between ferritic and austenitic alloys

    JS Zuback, TA Palmer, and T DebRoy. Additive manufacturing of functionally graded transition joints between ferritic and austenitic alloys. Journal of Alloys and Compounds, 770:995–1003, 2019

  43. [53]

    Directed energy deposition additive manufacturing of functionally graded al-w composites

    JPl Kelly, JW Elmer, FJ Ryerson, JRI Lee, and JJ Haslam. Directed energy deposition additive manufacturing of functionally graded al-w composites. Additive Manufacturing, 39:101845, 2021

  44. [54]

    Seong-Won Yoo, Choon-Man Lee, and Dong-Hyeon Kim. Effect of functionally graded material (fgm) interlayer in metal additive manufacturing of inconel-stainless bimetallic structure by laser melting deposition (lmd) and wire arc additive manufacturing (waam). Materials, 16(2):535, 2023

  45. [55]

    Qualify-as-you- go: sensor fusion of optical and acoustic signatures with contrastive deep learning for multi-material composition monitoring in laser powder bed fusion process

    Vigneashwara Pandiyan, Antonios Baganis, Roland Axel Richter, Rafał Wróbel, and Christian Leinenbach. Qualify-as-you- go: sensor fusion of optical and acoustic signatures with contrastive deep learning for multi-material composition monitoring in laser powder bed fusion proces...

  46. [56]

    Layout and material gradation in topology optimization of functionally graded structures: a global–local approach

    Sylvia RM Almeida, Glaucio H Paulino, and Emilio CN Silva. Layout and material gradation in topology optimization of functionally graded structures: a global–local approach. Structural and Multidisciplinary Optimization, 42:855–868, 2010

  47. [57]

    Topology optimization of thermoelastic structures with single and functionally graded materials exploring energy and stress-based formulations

    Rui F Silva, Pedro G Coelho, Fábio M Conde, Cláudia J Almeida, and Ana L Custódio. Topology optimization of thermoelastic structures with single and functionally graded materials exploring energy and stress-based formulations. Structural and Multidisciplinary Optimization, 68(...

  48. [58]

    Performance- driven computational design of multi-terminal compositionally graded alloy structures using graphs

    Marshall D Allen, Vahid Attari, Brent Vela, James Hanagan, Richard Malak, and Raymundo Arróyave. Performance- driven computational design of multi-terminal compositionally graded alloy structures using graphs. arXiv preprint arXiv:2412.03674, 2024. xviii

  49. [59]

    Deepsdf: Learning continuous signed distance functions for shape representation

    Jeong Joon Park, Peter Florence, Julian Straub, Richard Newcombe, and Steven Lovegrove. Deepsdf: Learning continuous signed distance functions for shape representation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 165–174, 2019

  50. [60]

    Nerf: Representing scenes as neural radiance fields for view synthesis

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis. Communications of the ACM, 65(1):99–106, 2021

  51. [61]

    On the use of artificial neural networks in topology optimisation

    Rebekka V Woldseth, Niels Aage, J Andreas Bærentzen, and Ole Sigmund. On the use of artificial neural networks in topology optimisation. Structural and Multidisciplinary Optimization, 65(10):294, 2022

  52. [62]

    Neural networks for topology optimization

    Ivan Sosnovik and Ivan Oseledets. Neural networks for topology optimization. Russian Journal of Numerical Analysis and Mathematical Modelling, 34(4):215–223, 2019

  53. [63]

    Multi-material topology optimization using neural networks

    Aaditya Chandrasekhar and Krishnan Suresh. Multi-material topology optimization using neural networks. Computer-Aided Design, 136:103017, 2021

  54. [64]

    Integrating material selection with design optimization via neural networks

    Aaditya Chandrasekhar, Saketh Sridhara, and Krishnan Suresh. Integrating material selection with design optimization via neural networks. Engineering with Computers, 38(5):4715–4730, 2022

  55. [65]

    Photos: topology optimization of photonic components using a shape library

    Rahul Kumar Padhy and Aaditya Chandrasekhar. Photos: topology optimization of photonic components using a shape library. Engineering with Computers, pages 1–13, 2024

  56. [66]

    Tomas: topology optimization of multiscale fluid flow devices using variational auto-encoders and super-shapes

    Rahul Kumar Padhy, Krishnan Suresh, and Aaditya Chandrasekhar. Tomas: topology optimization of multiscale fluid flow devices using variational auto-encoders and super-shapes. Structural and Multidisciplinary Optimization, 67(7):119, 2024

  57. [67]

    Implicit neural representations with periodic activation functions

    Vincent Sitzmann, Julien Martel, Alexander Bergman, David Lindell, and Gordon Wetzstein. Implicit neural representations with periodic activation functions. Advances in neural information processing systems, 33:7462–7473, 2020

  58. [68]

    Toplogical optimization of structures using fourier representations

    Daniel A White, Mark L Stowell, and Daniel A Tortorelli. Toplogical optimization of structures using fourier representations. Structural and Multidisciplinary Optimization, 58:1205–1220, 2018

  59. [69]

    Computational modeling of effects of alloying elements on elastic coefficients

    ZK Liu, H Zhang, S Ganeshan, Y Wang, and SN Mathaudhu. Computational modeling of effects of alloying elements on elastic coefficients. Scripta Materialia, 63(7):686–691, 2010

  60. [70]

    Application of the calphad method to predict the thermal conductivity in dielectric and semiconductor crystals

    Aïmen E Gheribi and Patrice Chartrand. Application of the calphad method to predict the thermal conductivity in dielectric and semiconductor crystals. Calphad, 39:70–79, 2012

  61. [71]

    The development of phase-based property data using the calphad method and infrastructure needs

    Carelyn E Campbell, Ursula R Kattner, and Zi-Kui Liu. The development of phase-based property data using the calphad method and infrastructure needs. In Zentropy, pages 621–660. Jenny Stanford Publishing, 2024

  62. [72]

    Multiplicative filter networks

    Rizal Fathony, Anit Kumar Sahu, Devin Willmott, and J Zico Kolter. Multiplicative filter networks. In International Conference on Learning Representations, 2020

  63. [73]

    Numerical instabilities in topology optimization: a survey on procedures dealing with checkerboards, mesh-dependencies and local minima

    Ole Sigmund and Joakim Petersson. Numerical instabilities in topology optimization: a survey on procedures dealing with checkerboards, mesh-dependencies and local minima. Structural optimization, 16:68–75, 1998

  64. [74]

    Residual multiplicative filter networks for multiscale reconstruction

    Shayan Shekarforoush, David Lindell, David J Fleet, and Marcus A Brubaker. Residual multiplicative filter networks for multiscale reconstruction. Advances in Neural Information Processing Systems, 35:8550–8563, 2022

  65. [75]

    High-throughput rapid experimental alloy development (ht-read)

    Kenneth S Vecchio, Olivia F Dippo, Kevin R Kaufmann, and Xiao Liu. High-throughput rapid experimental alloy development (ht-read). Acta Materialia, 221:117352, 2021

  66. [76]

    High-throughput alloy development using advanced characterization techniques during directed energy deposition additive manufacturing

    Niklas Sommer, André Bauer, Martin Kahlmeyer, Thomas Wegener, Sebastian Degener, Alexander Liehr, Artjom Bolender, Malte V ollmer, Hendrik Holz, Stefan Zeiler, et al. High-throughput alloy development using advanced characterization techniques during directed energy deposition...

  67. [77]

    High-throughput design of high-performance lightweight high-entropy alloys

    Rui Feng, Chuan Zhang, Michael C Gao, Zongrui Pei, Fan Zhang, Yan Chen, Dong Ma, Ke An, Jonathan D Poplawsky, Lizhi Ouyang, et al. High-throughput design of high-performance lightweight high-entropy alloys. Nature Communications, 12(1):4329, 2021

  68. [78]

    Computational thermodynamics: the Calphad method

    Hans Lukas, Suzana G Fries, and Bo Sundman. Computational thermodynamics: the Calphad method. Cambridge university press, 2007

  69. [79]

    Numerical recipes 3rd edition: The art of scientific computing

    William H Press. Numerical recipes 3rd edition: The art of scientific computing. Cambridge university press, 2007

  70. [80]

    Olson, and Jacob Schroder

    Nathan Bell, Luke N. Olson, and Jacob Schroder. Pyamg: Algebraic multigrid solvers in python. Journal of Open Source Software, 7(72):4142, 2022

  71. [81]

    A two-material topology optimization method for structures under steady thermo-mechanical loading

    Pierre F Thurier, George A Lesieutre, Mary I Frecker, and James H Adair. A two-material topology optimization method for structures under steady thermo-mechanical loading. Journal of Intelligent Material Systems and Structures, 30(11):1717– 1726, 2019

  72. [82]

    Multi–materials topology optimization using deep neural network for coupled thermo–mechanical problems

    Md Imrul Reza Shishir and Alireza Tabarraei. Multi–materials topology optimization using deep neural network for coupled thermo–mechanical problems. Computers & Structures, 291:107218, 2024. xix

  73. [83]

    Fourier features let networks learn high frequency functions in low dimensional domains

    Matthew Tancik, Pratul Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoor- thi, Jonathan Barron, and Ren Ng. Fourier features let networks learn high frequency functions in low dimensional domains. Advances in neural information pr...

  74. [84]

    Dive into deep learning

    Aston Zhang, Zachary C Lipton, Mu Li, and Alexander J Smola. Dive into deep learning. arXiv preprint arXiv:2106.11342, 2021

  75. [85]

    Constrained deep networks: Lagrangian optimization via log-barrier extensions

    Hoel Kervadec, Jose Dolz, Jing Yuan, Christian Desrosiers, Eric Granger, and Ismail Ben Ayed. Constrained deep networks: Lagrangian optimization via log-barrier extensions. In 2022 30th European Signal Processing Conference (EUSIPCO), pages 962–966. IEEE, 2022

  76. [86]

    Numerical optimization

    Jorge Nocedal and Stephen J Wright. Numerical optimization. Springer, 1999

  77. [87]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014

  78. [88]

    Auto: a framework for automatic differentiation in topology optimization

    Aaditya Chandrasekhar, Saketh Sridhara, and Krishnan Suresh. Auto: a framework for automatic differentiation in topology optimization. Structural and Multidisciplinary Optimization, 64(6):4355–4365, 2021

  79. [89]

    Jax-fem: A differentiable gpu-accelerated 3d finite element solver for automatic inverse design and mechanistic data science

    Tianju Xue, Shuheng Liao, Zhengtao Gan, Chanwook Park, Xiaoyu Xie, Wing Kam Liu, and Jian Cao. Jax-fem: A differentiable gpu-accelerated 3d finite element solver for automatic inverse design and mechanistic data science. Computer Physics Communications, 291:108802, 2023

  80. [90]

    JAX: composable transformations of Python+NumPy programs, 2018

    James Bradbury, Roy Frostig, Peter Hawkins, Matthew James Johnson, Chris Leary, Dougal Maclaurin, George Necula, Adam Paszke, Jake VanderPlas, Skye Wanderman-Milne, and Qiao Zhang. JAX: composable transformations of Python+NumPy programs, 2018

  81. [91]

    Automatic penalty continuation in structural topology optimization

    Susana Rojas-Labanda and Mathias Stolpe. Automatic penalty continuation in structural topology optimization. Structural and Multidisciplinary Optimization, 52(6):1205–1221, 2015

  82. [92]

    A material based model for topology optimization of thermoelastic structures

    H Rodrigues and P Fernandes. A material based model for topology optimization of thermoelastic structures. International journal for numerical methods in engineering, 38(12):1951–1965, 1995

  83. [93]

    Co-design of geometry and thermal-elastic gradient alloy distribution with temperature-dependent material properties

    Stefan Knapik, Shiguang Deng, Liwei Wang, Aaditya Chandrasekhar, Wing Kam Liu, Jian Cao, and Wei Chen. Co-design of geometry and thermal-elastic gradient alloy distribution with temperature-dependent material properties. Structural and Multidisciplinary Optimization, 68(7), Ju...

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.