diff --git a/docs/index.html b/docs/index.html index e4b4df0..158f36a 100644 --- a/docs/index.html +++ b/docs/index.html @@ -154,7 +154,7 @@

OPL – Optimisation problem library

BSO-toolbox C++ - {'40 seconds', '1 second'} + {'1 second', '40 seconds'} https://github.com/TUe-excellent-buildings/BSO-toolbox Building Spatial Design toolbox (TU/e) @@ -165,6 +165,47 @@

OPL – Optimisation problem library

+ + fn_cfd_based_non_premixed_burner_design + CFD-Based Non-Premixed Burner Design + Problem + continuous + + + multi-fidelity + box | unknown + 2 + none + none + no + no + [1, 2] + CFD-Based Non-Premixed Burner Design + Change the length and width of a combustor optimizing the combustion efficiency but keeping NOx emissions constrained. + + P. S. Lima, P. R. S. dos Reis, A. Á. B. Santos, E. A. del R. Chanona, and I. B. dos R. Nogueira, “Multi-fidelity Bayesian Optimization Framework for CFD-Based Non-Premixed Burner Design,” Nov. 28, 2025, arXiv: arXiv:2511.23140. doi: 10.48550/arXiv.2511.23140. + impl_cfd_based_non_premixed_burner_design + + 35-40 minutes + + real-world + + + + + CFD-Based Non-Premixed Burner Design + Python, Authors disclose using optimizer in Python, but the physical model happens in a simulation pipeline in FLUENT + {'35-40 minutes'} + + Change the length and width of a combustor optimizing the combustion efficiency but keeping NOx emissions constrained. + + + 1 + + + + + fn_convex_dtlz2 Convex DTLZ2 @@ -452,6 +493,129 @@

OPL – Optimisation problem library

+ + fn_layout_of_wind_farms + Layout of Wind Farms + Problem + continuous + + + multi-fidelity + box | unknown + >=2 + none + none + yes + + [1, 2] + + Optimizing a Wind Farm Layout: Input variables are coordinates of the turbines (box-constraints). Constraint that they don't overlap (hard constraint). Optimize the overall power output from all wind turbines and with different wind states. Use a simulation to evaluate a layout. It is expensive approx. 70 CPU hours. + + Layout of Wind Farms, https://arxiv.org/abs/2302.01071 + impl_layout_of_wind_farms + multimodal + 70 CPU hours + + real-world + + + + + Layout of Wind Farms + + {'70 CPU hours'} + + Optimizing a Wind Farm Layout: Input variables are coordinates of the turbines (box-constraints). Constraint that they don't overlap (hard constraint). Optimize the overall power output from all wind turbines and with different wind states. Use a simulation to evaluate a layout. It is expensive approx. 70 CPU hours. + + >=1 + + + + + + + + fn_materials_design + Materials Design + Problem + categorical | continuous + + + + + + + none + unknown + unknown + 1 + + Materials Design Optimization: Select appropriate material and microstructure optimization for optimizing the light absorbing of a quasi-random solar cell. Maximize light absorption by balancing the competing processes of light reflection and scattering.Choose from 5 materials. Another material design problem can be found here: https://github.com/clancyLab/NCM2018 + + Materials Design, https://www.nature.com/articles/s41598-020-60652-9 + impl_materials_design + + expensive simulation + + real-world + + + + + Materials Design + + {'expensive simulation'} + + Materials Design Optimization: Select appropriate material and microstructure optimization for optimizing the light absorbing of a quasi-random solar cell. Maximize light absorption by balancing the competing processes of light reflection and scattering.Choose from 5 materials. Another material design problem can be found here: https://github.com/clancyLab/NCM2018 + + + + + + + + + + fn_miprowa_millireactor_geometry_optimization + Miprowa millireactor geometry optimization + Problem + continuous + + + + box + >=1 + none + + no + unknown + + Miprowa millireactor design optimization + Optimizing the design of the mulireactor to achieve minimal stagnating flow and mixing variance + + Begall, M. J., Schweidtmann, A. M., Mhamdi, A., & Mitsos, A. (2023). Geometry optimization of a continuous millireactor via CFD and Bayesian optimization. Computers & chemical engineering, 171, 108140. + impl_miprowa_millireactor_geometry_optimization + + 6.5 hours + + real-world + + + + + Miprowa millireactor geometry optimization + + {'6.5 hours'} + + Optimizing the design of the mulireactor to achieve minimal stagnating flow and mixing variance + + >=1 + >=1 + + + + + fn_onemax_sphere_deceptive_rotell Onemax+Sphere / DeceptiveTrap+RotatedEllipsoid @@ -534,6 +698,88 @@

OPL – Optimisation problem library

+ + fn_optimal_sensor_placement_concrete_frame + Optimal Sensor Placement - concrete frame + Problem + binary + + + + unknown + 1 + none + + no + no + 1 + Optimal sensor placement for structural damage segmentation + Select sensor locations from a set of options. Here, the goal is to optimize structural damage prediction (but can be arbitrary) + form-submission + Sajedi, S., & Liang, X. (2022). Deep generative Bayesian optimization for sensor placement in structural health monitoring. Computer‐Aided Civil and Infrastructure Engineering, 37(9), 1109-1127. + impl_optimal_sensor_placement_concrete_frame + + minutes to hours + + artificial + + + + + Optimal Sensor Placement - concrete frame + + {'minutes to hours'} + + Select sensor locations from a set of options. Here, the goal is to optimize structural damage prediction (but can be arbitrary) + + + + + + + + + + fn_optimization_of_milireactor + Optimization of milireactor + Problem + continuous + + + + + + none + none + no + no + + Optimization of milireactor + Polymerization processes in the (specialty) chemical industry increasingly rely on continuous flow processing. Continuous polymerization reactors are promising, because their large surface to volume ratio allows for faster heat removal and better temperature control than batch reactors. Furthermore, this ratio can be kept constant during scaleup, reducing the need for pilot stages making it easier to go from lab- to production scale. However, continuous polymerization reactors, especially milli-scale reactors with internal mixing elements, can be highly complex devices, which can be difficult to characterize experimentally, as many of their properties are not readily available via measurements. Additionally, their small size renders them especially prone to fouling + + M. J. Begall, A. M. Schweidtmann, A. Mhamdi, and A. Mitsos, “Geometry optimization of a continuous millireactor via CFD and Bayesian optimization,” Computers & Chemical Engineering, vol. 171, p. 108140, Mar. 2023, doi: 10.1016/j.compchemeng.2023.108140. + impl_optimization_of_milireactor + + 6.6 hours + + real-world + + + + + Optimization of milireactor + COMSOL (It's suspected the authors used MATLAB) + {'6.6 hours'} + + Polymerization processes in the (specialty) chemical industry increasingly rely on continuous flow processing. Continuous polymerization reactors are promising, because their large surface to volume ratio allows for faster heat removal and better temperature control than batch reactors. Furthermore, this ratio can be kept constant during scaleup, reducing the need for pilot stages making it easier to go from lab- to production scale. However, continuous polymerization reactors, especially milli-scale reactors with internal mixing elements, can be highly complex devices, which can be difficult to characterize experimentally, as many of their properties are not readily available via measurements. Additionally, their small size renders them especially prone to fouling + + + + + + + + fn_radar_waveform Radar waveform @@ -575,6 +821,129 @@

OPL – Optimisation problem library

+ + fn_shape_optimization_of_a_lid_driven_cavity + Shape optimization of a lid-driven cavity + Problem + continuous + + + multi-fidelity | noisy + unknown + 1 + none + ["none", "though potentially one shall revise the meshing algorithm."] + no + no + [1, 2] + + "Consider an incompressible two-dimensional cavity flow, where all walls are at rest except the upper wall (lid) which moves with the constant velocity U0 in the positive horizontal direction, x. The aim is to optimize the shape of the left and right walls so that the energy dissipation is either minimized or maximized with the constraint of keeping the total volume of the fluid confined in the cavity fixed and also retain the length and position of the upper and lower walls. Therefore, the objective is to find the wall shape which either minimizes or maximizes" + form-submission + + impl_shape_optimization_of_a_lid_driven_cavity + + [] + + real-world + + + + + Shape optimization of a lid-driven cavity + Physical Model: Nek500; CFD Solver: OpenFOAM (Not disclosed the coupling, but could be Python) + + + "Consider an incompressible two-dimensional cavity flow, where all walls are at rest except the upper wall (lid) which moves with the constant velocity U0 in the positive horizontal direction, x. The aim is to optimize the shape of the left and right walls so that the energy dissipation is either minimized or maximized with the constraint of keeping the total volume of the fluid confined in the cavity fixed and also retain the length and position of the upper and lower walls. Therefore, the objective is to find the wall shape which either minimizes or maximizes" + + + + + + + + + + fn_simulation_driven_antenna_design + Simulation-Driven Antenna Design + Problem + continuous + + + + box + >=1 + + + no + unknown + 1 + + Simulation-Driven Antenna Design Optimization: One scalar merit function encoding the design specifications (gain, voltage standing wave ratio,..) calculated by using a high-fidelity EM simulator of the antenna under design. The decision variables are designable parameters like geometry parameters and they are box-constrained. + + Simulation-Driven Antenna Design, https://ieeexplore.ieee.org/abstract/document/9298949 + impl_simulation_driven_antenna_design + + expensive + + real-world + + + + + Simulation-Driven Antenna Design + + {'expensive'} + + Simulation-Driven Antenna Design Optimization: One scalar merit function encoding the design specifications (gain, voltage standing wave ratio,..) calculated by using a high-fidelity EM simulator of the antenna under design. The decision variables are designable parameters like geometry parameters and they are box-constrained. + + >=1 + >=1 + + + + + + + fn_transverse_bending_under_pole_impact + Transverse bending under pole impact + Problem + continuous + + + + box | unknown + 2 + none + none + no + + 1 + + A beam, simply supported at both ends, is struck laterally by a rigid pole, mimicking side-pole crash tests. The objective is to minimize intrusion, defined as the maximal dynamic displacement of the impactor before rebound, subject to a volume constraint of at most 50% of the design domain. The design is parameterized by a level-set function with 15 variables, mapped onto an LS-Dyna mesh of 1,600 eight-node solid elements with piecewise linear plasticity. Void elements are deleted from the mesh, and the out-of-plane (z) direction is fixed. + + Transverse bending under pole impact, http://www.sciencedirect.com/science/article/pii/S0045782519300726 + impl_transverse_bending_under_pole_impact + + 2 minutes + + real-world + + + + + Transverse bending under pole impact + Python + {'2 minutes'} + + A beam, simply supported at both ends, is struck laterally by a rigid pole, mimicking side-pole crash tests. The objective is to minimize intrusion, defined as the maximal dynamic displacement of the impactor before rebound, subject to a volume constraint of at most 50% of the design domain. The design is parameterized by a level-set function with 15 variables, mapped onto an LS-Dyna mesh of 1,600 eight-node solid elements with piecewise linear plasticity. Void elements are deleted from the mesh, and the out-of-plane (z) direction is fixed. + + 1 + + + + + + gen_beacon BEACON @@ -1518,6 +1887,47 @@

OPL – Optimisation problem library

+ + suite_biological_sequence_design + Biological sequence design + Suite + categorical + + + multi-fidelity | noisy + unknown + >=1 + none + observational + no + unknown + [1, 2] + + Biological sequence design + form-submission + Biological sequence design, https://arxiv.org/pdf/2203.12742 + impl_biological_sequence_design + multimodal + expensive + + real-world + + + + + Biological sequence design + Python + {'expensive'} + https://github.com/samuelstanton/lambo + Biological sequence design + + + + + + + + suite_bp BP @@ -1710,11 +2120,11 @@

OPL – Optimisation problem library

>=1 - CEC2013 reference code | IOHexperimenter + IOHexperimenter | CEC2013 reference code C++/Python - https://github.com/P-N-Suganthan/CEC2013 | https://github.com/IOHprofiler/IOHexperimenter - Suganthan's reference implementation | IOHprofiler experimenter framework + https://github.com/IOHprofiler/IOHexperimenter | https://github.com/P-N-Suganthan/CEC2013 + IOHprofiler experimenter framework | Suganthan's reference implementation @@ -2338,6 +2748,47 @@

OPL – Optimisation problem library

+ + suite_guacamol + GuacaMol + Suite + unknown + + + multi-fidelity + unknown + >=1 + none + none + unknown + + [1, 2] + GuacaMol: Benchmarking Models for de Novo Molecular Design + De novo molecular design is a computational technique to generate novel compounds with desirable property profiles from scratch. It complements virtual screening, where large virtual compound libraries are pregenerated, stored, and then subsequently ranked on demand. Virtual screening is particularly useful when the virtual compounds are readily available (commercially, or in in-house screening collections), or easy to synthesize. Data sets sized on the order of 10^13 can be routinely screened under current computational constraints. However, this is only a tiny fraction of drug-like chemical space, which has an estimated size anywhere between 10^24 and 10^60 possible structures. This number might be even larger considering new modalities such as PROTACs. Efficient approaches to query larger combinatorial spaces without full enumeration via similarity to given structures have been reported. However, it is not straightforward to perform more complex, multiobjective queries without enumerating substantial parts of the space.\n\nIn contrast, in de novo design, only a relatively small number of molecules is explicitly generated, and chemical space is explored via search or optimization procedures. Thus, by focusing on the most relevant areas of chemical space for the current multiobjective query, de novo design algorithms can in principle explore the full chemical space, given any ranking or scoring function. + + GuacaMol, https://doi.org/10.1021/acs.jcim.8b00839 + impl_guacamol + + + + real-world + + + + + GuacaMol + Python + + https://github.com/BenevolentAI/guacamol + De novo molecular design is a computational technique to generate novel compounds with desirable property profiles from scratch. It complements virtual screening, where large virtual compound libraries are pregenerated, stored, and then subsequently ranked on demand. Virtual screening is particularly useful when the virtual compounds are readily available (commercially, or in in-house screening collections), or easy to synthesize. Data sets sized on the order of 10^13 can be routinely screened under current computational constraints. However, this is only a tiny fraction of drug-like chemical space, which has an estimated size anywhere between 10^24 and 10^60 possible structures. This number might be even larger considering new modalities such as PROTACs. Efficient approaches to query larger combinatorial spaces without full enumeration via similarity to given structures have been reported. However, it is not straightforward to perform more complex, multiobjective queries without enumerating substantial parts of the space.\n\nIn contrast, in de novo design, only a relatively small number of molecules is explicitly generated, and chemical space is explored via search or optimization procedures. Thus, by focusing on the most relevant areas of chemical space for the current multiobjective query, de novo design algorithms can in principle explore the full chemical space, given any ranking or scoring function. + + + + + + + + suite_iohclustering IOHClustering @@ -2696,7 +3147,7 @@

OPL – Optimisation problem library

MECHBench Python - {'7 minutes', '1 minute'} + {'1 minute', '7 minutes'} https://github.com/BayesOptApp/MECHBench Structural mechanics design optimization benchmark @@ -2953,6 +3404,47 @@

OPL – Optimisation problem library

+ + suite_optimization_of_a_spoiler_ice_airfoil_model + Optimization of a spoiler-ice airfoil model + Suite + continuous + + + multi-fidelity + box + 2 + none + none + no + no + [1, 2] + Optimization of a spoiler-ice airfoil model + When operating in cold climate, the performance of the wind turbines can be reduced by icing. In an extreme condition, heavy icing can even lead to a complete stop of the turbine. Icing is also a risky condition for airplane wings since it may induce flow separation at a small angles of attack, which then might lead to stall and consequently loss of lift. Ice accretion is a complex process which depends on both aerodynamics and thermodynamics. The process is affected by many parameters, for instance, ambient temperature, surface properties, relative speed of the airfoil, and the time span over which the icing event takes place. As a result, numerous possibilities for an ice profile exist. Bragg et al. categorized the leading-edge ice on the airfoils into four types: roughness, horn, streamwise, and spanwise-ridge ice. The main characteristic of the horn ice is the presence of the large protuberances which induce a large flow separation downstream of the ice, and hence dramatically affect the aerodynamic performance, including the lift, drag and moment coefficients. On the other hand, streamwise ice is smoothly formed along the streamlines and is less dangerous in the sense of having adverse effects on the aerodynamic performance. + + Y. Morita, S. Rezaeiravesh, N. Tabatabaei, R. Vinuesa, K. Fukagata, and P. Schlatter, “Applying Bayesian Optimization with Gaussian Process Regression to Computational Fluid Dynamics Problems,” Journal of Computational Physics, vol. 449, p. 110788, Jan. 2022, doi: 10.1016/j.jcp.2021.110788. + impl_optimization_of_a_spoiler_ice_airfoil_model + + The authors disclose it takes 1 hour. + + real-world + + + + + Optimization of a spoiler-ice airfoil model + Besides Python (as the authors use GPy) for optimizing the problem, there is the need of a commercial mesher (ANSYS ICEM) + {'The authors disclose it takes 1 hour.'} + + When operating in cold climate, the performance of the wind turbines can be reduced by icing. In an extreme condition, heavy icing can even lead to a complete stop of the turbine. Icing is also a risky condition for airplane wings since it may induce flow separation at a small angles of attack, which then might lead to stall and consequently loss of lift. Ice accretion is a complex process which depends on both aerodynamics and thermodynamics. The process is affected by many parameters, for instance, ambient temperature, surface properties, relative speed of the airfoil, and the time span over which the icing event takes place. As a result, numerous possibilities for an ice profile exist. Bragg et al. categorized the leading-edge ice on the airfoils into four types: roughness, horn, streamwise, and spanwise-ridge ice. The main characteristic of the horn ice is the presence of the large protuberances which induce a large flow separation downstream of the ice, and hence dramatically affect the aerodynamic performance, including the lift, drag and moment coefficients. On the other hand, streamwise ice is smoothly formed along the streamlines and is less dangerous in the sense of having adverse effects on the aerodynamic performance. + + 2 + 2 + + + + + suite_pbo PBO @@ -3199,6 +3691,47 @@

OPL – Optimisation problem library

+ + suite_shape_optimization_of_a_diverging_channel + Shape optimization of a diverging channel + Suite + continuous + + + + + + none + + no + + + + The objective of this section is to efficiently obtain the optimal shape of the upper wall of a channel so that a target β along the edge of the Turbulent Boundary Layer over the lower wall is achieved. + + + + + + + real-world + + + + + + + + + + + + + + + + + suite_submodular Submodular Optimization @@ -3270,7 +3803,7 @@

OPL – Optimisation problem library

TulipaEnergyModel.jl Julia / JuMP - {'hours', 'minutes'} + {'minutes', 'hours'} https://tulipaenergy.github.io/TulipaEnergyModel.jl/stable/ | https://github.com/TulipaEnergy/Tulipa-OBZ-CaseStudy Large linear program for optimal investment and operation of energy systems @@ -3281,6 +3814,47 @@

OPL – Optimisation problem library

+ + suite_vae_molecule_generation + VAE Molecule Generation + Suite + unknown + + + multi-fidelity + unknown + >=1 + none + none + unknown + + [1, 2] + + The goal of drug and material design is to identify novel molecules that have certain desirable properties. We view this as an optimization problem, in which we are searching for the molecules that maximize our quantitative desiderata.\n\nMaking and testing new compounds are costly and time-consuming, and the number of potential candidates is overwhelming. Only about 10^8 substances have ever been synthesized, whereas the range of potential drug-like molecules is estimated to be between 10^23 and 10^60. Virtual screening can be used to speed up this search. Virtual libraries containing thousands to hundreds of millions of candidates can be assayed with first-principles simulations or statistical predictions based on learned proxy models, and only\nthe most promising leads are selected and tested experimentally.\n\nWays to assess the generated molecules are the water−octanol partition coefficient (logP), the synthetic accessibility score (SAS), and Quantitative Estimation of Drug-likeness (QED), which ranges in value between 0 and 1, with higher values indicating that the molecule is more drug-like. + form-submission + VAE Molecule Generation, https://doi.org/10.1039/c9sc04026a + impl_vae_molecule_generation + + + + real-world + + + + + VAE Molecule Generation + Python + + https://github.com/Ryan-Rhys/Constrained-Bayesian-Optimisation-for-Automatic-Chemical-Design + The goal of drug and material design is to identify novel molecules that have certain desirable properties. We view this as an optimization problem, in which we are searching for the molecules that maximize our quantitative desiderata.\n\nMaking and testing new compounds are costly and time-consuming, and the number of potential candidates is overwhelming. Only about 10^8 substances have ever been synthesized, whereas the range of potential drug-like molecules is estimated to be between 10^23 and 10^60. Virtual screening can be used to speed up this search. Virtual libraries containing thousands to hundreds of millions of candidates can be assayed with first-principles simulations or statistical predictions based on learned proxy models, and only\nthe most promising leads are selected and tested experimentally.\n\nWays to assess the generated molecules are the water−octanol partition coefficient (logP), the synthetic accessibility score (SAS), and Quantitative Estimation of Drug-likeness (QED), which ranges in value between 0 and 1, with higher values indicating that the molecule is more drug-like. + + + + + + + + suite_vehicle_dynamics VehicleDynamics diff --git a/docs/problems.html b/docs/problems.html index 92dc2f8..03c2f01 100644 --- a/docs/problems.html +++ b/docs/problems.html @@ -128,7 +128,7 @@ BSO-toolbox C++ - {'40 seconds', '1 second'} + {'1 second', '40 seconds'} https://github.com/TUe-excellent-buildings/BSO-toolbox Building Spatial Design toolbox (TU/e) @@ -139,6 +139,47 @@ + + fn_cfd_based_non_premixed_burner_design + CFD-Based Non-Premixed Burner Design + Problem + continuous + + + multi-fidelity + box | unknown + 2 + none + none + no + no + [1, 2] + CFD-Based Non-Premixed Burner Design + Change the length and width of a combustor optimizing the combustion efficiency but keeping NOx emissions constrained. + + P. S. Lima, P. R. S. dos Reis, A. Á. B. Santos, E. A. del R. Chanona, and I. B. dos R. Nogueira, “Multi-fidelity Bayesian Optimization Framework for CFD-Based Non-Premixed Burner Design,” Nov. 28, 2025, arXiv: arXiv:2511.23140. doi: 10.48550/arXiv.2511.23140. + impl_cfd_based_non_premixed_burner_design + + 35-40 minutes + + real-world + + + + + CFD-Based Non-Premixed Burner Design + Python, Authors disclose using optimizer in Python, but the physical model happens in a simulation pipeline in FLUENT + {'35-40 minutes'} + + Change the length and width of a combustor optimizing the combustion efficiency but keeping NOx emissions constrained. + + + 1 + + + + + fn_convex_dtlz2 Convex DTLZ2 @@ -426,6 +467,129 @@ + + fn_layout_of_wind_farms + Layout of Wind Farms + Problem + continuous + + + multi-fidelity + box | unknown + >=2 + none + none + yes + + [1, 2] + + Optimizing a Wind Farm Layout: Input variables are coordinates of the turbines (box-constraints). Constraint that they don't overlap (hard constraint). Optimize the overall power output from all wind turbines and with different wind states. Use a simulation to evaluate a layout. It is expensive approx. 70 CPU hours. + + Layout of Wind Farms, https://arxiv.org/abs/2302.01071 + impl_layout_of_wind_farms + multimodal + 70 CPU hours + + real-world + + + + + Layout of Wind Farms + + {'70 CPU hours'} + + Optimizing a Wind Farm Layout: Input variables are coordinates of the turbines (box-constraints). Constraint that they don't overlap (hard constraint). Optimize the overall power output from all wind turbines and with different wind states. Use a simulation to evaluate a layout. It is expensive approx. 70 CPU hours. + + >=1 + + + + + + + + fn_materials_design + Materials Design + Problem + categorical | continuous + + + + + + + none + unknown + unknown + 1 + + Materials Design Optimization: Select appropriate material and microstructure optimization for optimizing the light absorbing of a quasi-random solar cell. Maximize light absorption by balancing the competing processes of light reflection and scattering.Choose from 5 materials. Another material design problem can be found here: https://github.com/clancyLab/NCM2018 + + Materials Design, https://www.nature.com/articles/s41598-020-60652-9 + impl_materials_design + + expensive simulation + + real-world + + + + + Materials Design + + {'expensive simulation'} + + Materials Design Optimization: Select appropriate material and microstructure optimization for optimizing the light absorbing of a quasi-random solar cell. Maximize light absorption by balancing the competing processes of light reflection and scattering.Choose from 5 materials. Another material design problem can be found here: https://github.com/clancyLab/NCM2018 + + + + + + + + + + fn_miprowa_millireactor_geometry_optimization + Miprowa millireactor geometry optimization + Problem + continuous + + + + box + >=1 + none + + no + unknown + + Miprowa millireactor design optimization + Optimizing the design of the mulireactor to achieve minimal stagnating flow and mixing variance + + Begall, M. J., Schweidtmann, A. M., Mhamdi, A., & Mitsos, A. (2023). Geometry optimization of a continuous millireactor via CFD and Bayesian optimization. Computers & chemical engineering, 171, 108140. + impl_miprowa_millireactor_geometry_optimization + + 6.5 hours + + real-world + + + + + Miprowa millireactor geometry optimization + + {'6.5 hours'} + + Optimizing the design of the mulireactor to achieve minimal stagnating flow and mixing variance + + >=1 + >=1 + + + + + fn_onemax_sphere_deceptive_rotell Onemax+Sphere / DeceptiveTrap+RotatedEllipsoid @@ -508,6 +672,88 @@ + + fn_optimal_sensor_placement_concrete_frame + Optimal Sensor Placement - concrete frame + Problem + binary + + + + unknown + 1 + none + + no + no + 1 + Optimal sensor placement for structural damage segmentation + Select sensor locations from a set of options. Here, the goal is to optimize structural damage prediction (but can be arbitrary) + form-submission + Sajedi, S., & Liang, X. (2022). Deep generative Bayesian optimization for sensor placement in structural health monitoring. Computer‐Aided Civil and Infrastructure Engineering, 37(9), 1109-1127. + impl_optimal_sensor_placement_concrete_frame + + minutes to hours + + artificial + + + + + Optimal Sensor Placement - concrete frame + + {'minutes to hours'} + + Select sensor locations from a set of options. Here, the goal is to optimize structural damage prediction (but can be arbitrary) + + + + + + + + + + fn_optimization_of_milireactor + Optimization of milireactor + Problem + continuous + + + + + + none + none + no + no + + Optimization of milireactor + Polymerization processes in the (specialty) chemical industry increasingly rely on continuous flow processing. Continuous polymerization reactors are promising, because their large surface to volume ratio allows for faster heat removal and better temperature control than batch reactors. Furthermore, this ratio can be kept constant during scaleup, reducing the need for pilot stages making it easier to go from lab- to production scale. However, continuous polymerization reactors, especially milli-scale reactors with internal mixing elements, can be highly complex devices, which can be difficult to characterize experimentally, as many of their properties are not readily available via measurements. Additionally, their small size renders them especially prone to fouling + + M. J. Begall, A. M. Schweidtmann, A. Mhamdi, and A. Mitsos, “Geometry optimization of a continuous millireactor via CFD and Bayesian optimization,” Computers & Chemical Engineering, vol. 171, p. 108140, Mar. 2023, doi: 10.1016/j.compchemeng.2023.108140. + impl_optimization_of_milireactor + + 6.6 hours + + real-world + + + + + Optimization of milireactor + COMSOL (It's suspected the authors used MATLAB) + {'6.6 hours'} + + Polymerization processes in the (specialty) chemical industry increasingly rely on continuous flow processing. Continuous polymerization reactors are promising, because their large surface to volume ratio allows for faster heat removal and better temperature control than batch reactors. Furthermore, this ratio can be kept constant during scaleup, reducing the need for pilot stages making it easier to go from lab- to production scale. However, continuous polymerization reactors, especially milli-scale reactors with internal mixing elements, can be highly complex devices, which can be difficult to characterize experimentally, as many of their properties are not readily available via measurements. Additionally, their small size renders them especially prone to fouling + + + + + + + + fn_radar_waveform Radar waveform @@ -549,6 +795,129 @@ + + fn_shape_optimization_of_a_lid_driven_cavity + Shape optimization of a lid-driven cavity + Problem + continuous + + + multi-fidelity | noisy + unknown + 1 + none + ["none", "though potentially one shall revise the meshing algorithm."] + no + no + [1, 2] + + "Consider an incompressible two-dimensional cavity flow, where all walls are at rest except the upper wall (lid) which moves with the constant velocity U0 in the positive horizontal direction, x. The aim is to optimize the shape of the left and right walls so that the energy dissipation is either minimized or maximized with the constraint of keeping the total volume of the fluid confined in the cavity fixed and also retain the length and position of the upper and lower walls. Therefore, the objective is to find the wall shape which either minimizes or maximizes" + form-submission + + impl_shape_optimization_of_a_lid_driven_cavity + + [] + + real-world + + + + + Shape optimization of a lid-driven cavity + Physical Model: Nek500; CFD Solver: OpenFOAM (Not disclosed the coupling, but could be Python) + + + "Consider an incompressible two-dimensional cavity flow, where all walls are at rest except the upper wall (lid) which moves with the constant velocity U0 in the positive horizontal direction, x. The aim is to optimize the shape of the left and right walls so that the energy dissipation is either minimized or maximized with the constraint of keeping the total volume of the fluid confined in the cavity fixed and also retain the length and position of the upper and lower walls. Therefore, the objective is to find the wall shape which either minimizes or maximizes" + + + + + + + + + + fn_simulation_driven_antenna_design + Simulation-Driven Antenna Design + Problem + continuous + + + + box + >=1 + + + no + unknown + 1 + + Simulation-Driven Antenna Design Optimization: One scalar merit function encoding the design specifications (gain, voltage standing wave ratio,..) calculated by using a high-fidelity EM simulator of the antenna under design. The decision variables are designable parameters like geometry parameters and they are box-constrained. + + Simulation-Driven Antenna Design, https://ieeexplore.ieee.org/abstract/document/9298949 + impl_simulation_driven_antenna_design + + expensive + + real-world + + + + + Simulation-Driven Antenna Design + + {'expensive'} + + Simulation-Driven Antenna Design Optimization: One scalar merit function encoding the design specifications (gain, voltage standing wave ratio,..) calculated by using a high-fidelity EM simulator of the antenna under design. The decision variables are designable parameters like geometry parameters and they are box-constrained. + + >=1 + >=1 + + + + + + + fn_transverse_bending_under_pole_impact + Transverse bending under pole impact + Problem + continuous + + + + box | unknown + 2 + none + none + no + + 1 + + A beam, simply supported at both ends, is struck laterally by a rigid pole, mimicking side-pole crash tests. The objective is to minimize intrusion, defined as the maximal dynamic displacement of the impactor before rebound, subject to a volume constraint of at most 50% of the design domain. The design is parameterized by a level-set function with 15 variables, mapped onto an LS-Dyna mesh of 1,600 eight-node solid elements with piecewise linear plasticity. Void elements are deleted from the mesh, and the out-of-plane (z) direction is fixed. + + Transverse bending under pole impact, http://www.sciencedirect.com/science/article/pii/S0045782519300726 + impl_transverse_bending_under_pole_impact + + 2 minutes + + real-world + + + + + Transverse bending under pole impact + Python + {'2 minutes'} + + A beam, simply supported at both ends, is struck laterally by a rigid pole, mimicking side-pole crash tests. The objective is to minimize intrusion, defined as the maximal dynamic displacement of the impactor before rebound, subject to a volume constraint of at most 50% of the design domain. The design is parameterized by a level-set function with 15 variables, mapped onto an LS-Dyna mesh of 1,600 eight-node solid elements with piecewise linear plasticity. Void elements are deleted from the mesh, and the out-of-plane (z) direction is fixed. + + 1 + + + + + + gen_beacon BEACON @@ -1492,6 +1861,47 @@ + + suite_biological_sequence_design + Biological sequence design + Suite + categorical + + + multi-fidelity | noisy + unknown + >=1 + none + observational + no + unknown + [1, 2] + + Biological sequence design + form-submission + Biological sequence design, https://arxiv.org/pdf/2203.12742 + impl_biological_sequence_design + multimodal + expensive + + real-world + + + + + Biological sequence design + Python + {'expensive'} + https://github.com/samuelstanton/lambo + Biological sequence design + + + + + + + + suite_bp BP @@ -1684,11 +2094,11 @@ >=1 - CEC2013 reference code | IOHexperimenter + IOHexperimenter | CEC2013 reference code C++/Python - https://github.com/P-N-Suganthan/CEC2013 | https://github.com/IOHprofiler/IOHexperimenter - Suganthan's reference implementation | IOHprofiler experimenter framework + https://github.com/IOHprofiler/IOHexperimenter | https://github.com/P-N-Suganthan/CEC2013 + IOHprofiler experimenter framework | Suganthan's reference implementation @@ -2312,6 +2722,47 @@ + + suite_guacamol + GuacaMol + Suite + unknown + + + multi-fidelity + unknown + >=1 + none + none + unknown + + [1, 2] + GuacaMol: Benchmarking Models for de Novo Molecular Design + De novo molecular design is a computational technique to generate novel compounds with desirable property profiles from scratch. It complements virtual screening, where large virtual compound libraries are pregenerated, stored, and then subsequently ranked on demand. Virtual screening is particularly useful when the virtual compounds are readily available (commercially, or in in-house screening collections), or easy to synthesize. Data sets sized on the order of 10^13 can be routinely screened under current computational constraints. However, this is only a tiny fraction of drug-like chemical space, which has an estimated size anywhere between 10^24 and 10^60 possible structures. This number might be even larger considering new modalities such as PROTACs. Efficient approaches to query larger combinatorial spaces without full enumeration via similarity to given structures have been reported. However, it is not straightforward to perform more complex, multiobjective queries without enumerating substantial parts of the space.\n\nIn contrast, in de novo design, only a relatively small number of molecules is explicitly generated, and chemical space is explored via search or optimization procedures. Thus, by focusing on the most relevant areas of chemical space for the current multiobjective query, de novo design algorithms can in principle explore the full chemical space, given any ranking or scoring function. + + GuacaMol, https://doi.org/10.1021/acs.jcim.8b00839 + impl_guacamol + + + + real-world + + + + + GuacaMol + Python + + https://github.com/BenevolentAI/guacamol + De novo molecular design is a computational technique to generate novel compounds with desirable property profiles from scratch. It complements virtual screening, where large virtual compound libraries are pregenerated, stored, and then subsequently ranked on demand. Virtual screening is particularly useful when the virtual compounds are readily available (commercially, or in in-house screening collections), or easy to synthesize. Data sets sized on the order of 10^13 can be routinely screened under current computational constraints. However, this is only a tiny fraction of drug-like chemical space, which has an estimated size anywhere between 10^24 and 10^60 possible structures. This number might be even larger considering new modalities such as PROTACs. Efficient approaches to query larger combinatorial spaces without full enumeration via similarity to given structures have been reported. However, it is not straightforward to perform more complex, multiobjective queries without enumerating substantial parts of the space.\n\nIn contrast, in de novo design, only a relatively small number of molecules is explicitly generated, and chemical space is explored via search or optimization procedures. Thus, by focusing on the most relevant areas of chemical space for the current multiobjective query, de novo design algorithms can in principle explore the full chemical space, given any ranking or scoring function. + + + + + + + + suite_iohclustering IOHClustering @@ -2670,7 +3121,7 @@ MECHBench Python - {'7 minutes', '1 minute'} + {'1 minute', '7 minutes'} https://github.com/BayesOptApp/MECHBench Structural mechanics design optimization benchmark @@ -2927,6 +3378,47 @@ + + suite_optimization_of_a_spoiler_ice_airfoil_model + Optimization of a spoiler-ice airfoil model + Suite + continuous + + + multi-fidelity + box + 2 + none + none + no + no + [1, 2] + Optimization of a spoiler-ice airfoil model + When operating in cold climate, the performance of the wind turbines can be reduced by icing. In an extreme condition, heavy icing can even lead to a complete stop of the turbine. Icing is also a risky condition for airplane wings since it may induce flow separation at a small angles of attack, which then might lead to stall and consequently loss of lift. Ice accretion is a complex process which depends on both aerodynamics and thermodynamics. The process is affected by many parameters, for instance, ambient temperature, surface properties, relative speed of the airfoil, and the time span over which the icing event takes place. As a result, numerous possibilities for an ice profile exist. Bragg et al. categorized the leading-edge ice on the airfoils into four types: roughness, horn, streamwise, and spanwise-ridge ice. The main characteristic of the horn ice is the presence of the large protuberances which induce a large flow separation downstream of the ice, and hence dramatically affect the aerodynamic performance, including the lift, drag and moment coefficients. On the other hand, streamwise ice is smoothly formed along the streamlines and is less dangerous in the sense of having adverse effects on the aerodynamic performance. + + Y. Morita, S. Rezaeiravesh, N. Tabatabaei, R. Vinuesa, K. Fukagata, and P. Schlatter, “Applying Bayesian Optimization with Gaussian Process Regression to Computational Fluid Dynamics Problems,” Journal of Computational Physics, vol. 449, p. 110788, Jan. 2022, doi: 10.1016/j.jcp.2021.110788. + impl_optimization_of_a_spoiler_ice_airfoil_model + + The authors disclose it takes 1 hour. + + real-world + + + + + Optimization of a spoiler-ice airfoil model + Besides Python (as the authors use GPy) for optimizing the problem, there is the need of a commercial mesher (ANSYS ICEM) + {'The authors disclose it takes 1 hour.'} + + When operating in cold climate, the performance of the wind turbines can be reduced by icing. In an extreme condition, heavy icing can even lead to a complete stop of the turbine. Icing is also a risky condition for airplane wings since it may induce flow separation at a small angles of attack, which then might lead to stall and consequently loss of lift. Ice accretion is a complex process which depends on both aerodynamics and thermodynamics. The process is affected by many parameters, for instance, ambient temperature, surface properties, relative speed of the airfoil, and the time span over which the icing event takes place. As a result, numerous possibilities for an ice profile exist. Bragg et al. categorized the leading-edge ice on the airfoils into four types: roughness, horn, streamwise, and spanwise-ridge ice. The main characteristic of the horn ice is the presence of the large protuberances which induce a large flow separation downstream of the ice, and hence dramatically affect the aerodynamic performance, including the lift, drag and moment coefficients. On the other hand, streamwise ice is smoothly formed along the streamlines and is less dangerous in the sense of having adverse effects on the aerodynamic performance. + + 2 + 2 + + + + + suite_pbo PBO @@ -3173,6 +3665,47 @@ + + suite_shape_optimization_of_a_diverging_channel + Shape optimization of a diverging channel + Suite + continuous + + + + + + none + + no + + + + The objective of this section is to efficiently obtain the optimal shape of the upper wall of a channel so that a target β along the edge of the Turbulent Boundary Layer over the lower wall is achieved. + + + + + + + real-world + + + + + + + + + + + + + + + + + suite_submodular Submodular Optimization @@ -3244,7 +3777,7 @@ TulipaEnergyModel.jl Julia / JuMP - {'hours', 'minutes'} + {'minutes', 'hours'} https://tulipaenergy.github.io/TulipaEnergyModel.jl/stable/ | https://github.com/TulipaEnergy/Tulipa-OBZ-CaseStudy Large linear program for optimal investment and operation of energy systems @@ -3255,6 +3788,47 @@ + + suite_vae_molecule_generation + VAE Molecule Generation + Suite + unknown + + + multi-fidelity + unknown + >=1 + none + none + unknown + + [1, 2] + + The goal of drug and material design is to identify novel molecules that have certain desirable properties. We view this as an optimization problem, in which we are searching for the molecules that maximize our quantitative desiderata.\n\nMaking and testing new compounds are costly and time-consuming, and the number of potential candidates is overwhelming. Only about 10^8 substances have ever been synthesized, whereas the range of potential drug-like molecules is estimated to be between 10^23 and 10^60. Virtual screening can be used to speed up this search. Virtual libraries containing thousands to hundreds of millions of candidates can be assayed with first-principles simulations or statistical predictions based on learned proxy models, and only\nthe most promising leads are selected and tested experimentally.\n\nWays to assess the generated molecules are the water−octanol partition coefficient (logP), the synthetic accessibility score (SAS), and Quantitative Estimation of Drug-likeness (QED), which ranges in value between 0 and 1, with higher values indicating that the molecule is more drug-like. + form-submission + VAE Molecule Generation, https://doi.org/10.1039/c9sc04026a + impl_vae_molecule_generation + + + + real-world + + + + + VAE Molecule Generation + Python + + https://github.com/Ryan-Rhys/Constrained-Bayesian-Optimisation-for-Automatic-Chemical-Design + The goal of drug and material design is to identify novel molecules that have certain desirable properties. We view this as an optimization problem, in which we are searching for the molecules that maximize our quantitative desiderata.\n\nMaking and testing new compounds are costly and time-consuming, and the number of potential candidates is overwhelming. Only about 10^8 substances have ever been synthesized, whereas the range of potential drug-like molecules is estimated to be between 10^23 and 10^60. Virtual screening can be used to speed up this search. Virtual libraries containing thousands to hundreds of millions of candidates can be assayed with first-principles simulations or statistical predictions based on learned proxy models, and only\nthe most promising leads are selected and tested experimentally.\n\nWays to assess the generated molecules are the water−octanol partition coefficient (logP), the synthetic accessibility score (SAS), and Quantitative Estimation of Drug-likeness (QED), which ranges in value between 0 and 1, with higher values indicating that the molecule is more drug-like. + + + + + + + + suite_vehicle_dynamics VehicleDynamics diff --git a/problems.yaml b/problems.yaml index c32939e..0736d62 100644 --- a/problems.yaml +++ b/problems.yaml @@ -37,7 +37,8 @@ fn_ato: link: type: null url: https://doi.org/10.1145/3795101.3805368 - title: Multi-Objective Pipeline Optimisation and Configuration of Automatic Train Operation Trajectories + title: Multi-Objective Pipeline Optimisation and Configuration of Automatic Train + Operation Trajectories type: null source: - real-world @@ -54,18 +55,18 @@ fn_building_spatial: - equality: null hard: 'yes' number: null - type: unknown + type: box - equality: null hard: 'yes' number: null - type: box + type: unknown description: Optimise the spatial layout of a building to minimise energy consumption for climate control and minimise the strain on the structure. Many hard constraints; mixed-variable (continuous+binary); expensive evaluations. dynamic_type: null evaluation_time: - - 1 second - 40 seconds + - 1 second fidelity_levels: null implementations: - impl_bso_toolbox @@ -91,11 +92,57 @@ fn_building_spatial: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 - type: binary + type: continuous +fn_cfd_based_non_premixed_burner_design: + allows_partial_evaluation: 'no' + can_evaluate_objectives_independently: 'no' + code_examples: null + constraints: + - equality: null + hard: 'no' + number: 1 + type: box + - equality: null + hard: 'no' + number: 1 + type: unknown + description: Change the length and width of a combustor optimizing the combustion + efficiency but keeping NOx emissions constrained. + dynamic_type: + - none + evaluation_time: + - 35-40 minutes + fidelity_levels: + - 1 + - 2 + implementations: + - impl_cfd_based_non_premixed_burner_design + instances: null + long_name: CFD-Based Non-Premixed Burner Design + modality: null + name: CFD-Based Non-Premixed Burner Design + noise_type: + - none + objectives: null + references: + - authors: [] + link: null + title: 'P. S. Lima, P. R. S. dos Reis, A. Á. B. Santos, E. A. del R. Chanona, + and I. B. dos R. Nogueira, “Multi-fidelity Bayesian Optimization Framework for + CFD-Based Non-Premixed Burner Design,” Nov. 28, 2025, arXiv: arXiv:2511.23140. + doi: 10.48550/arXiv.2511.23140.' + type: null + source: + - real-world + tags: null + type: problem + variables: + - dim: null + type: continuous fn_convex_dtlz2: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -143,16 +190,16 @@ fn_emdo: constraints: - equality: null hard: 'yes' - number: null - type: box + number: 12 + type: unknown - equality: null hard: some number: null type: unknown - equality: null hard: 'yes' - number: 12 - type: unknown + number: null + type: box description: '# Goal Find a design of a synchronous electric motor for power steering systems that @@ -324,11 +371,11 @@ fn_invdeceptive_deceptive_rotell: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 - type: binary + type: continuous fn_inverted_dtlz1: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -410,6 +457,133 @@ fn_jsec2019: variables: - dim: 32 type: continuous +fn_layout_of_wind_farms: + allows_partial_evaluation: 'yes' + can_evaluate_objectives_independently: null + code_examples: null + constraints: + - equality: null + hard: 'yes' + number: null + type: box + - equality: null + hard: 'yes' + number: null + type: unknown + description: 'Optimizing a Wind Farm Layout: Input variables are coordinates of + the turbines (box-constraints). Constraint that they don''t overlap (hard constraint). + Optimize the overall power output from all wind turbines and with different wind + states. Use a simulation to evaluate a layout. It is expensive approx. 70 CPU + hours.' + dynamic_type: + - none + evaluation_time: + - 70 CPU hours + fidelity_levels: + - 1 + - 2 + implementations: + - impl_layout_of_wind_farms + instances: null + long_name: null + modality: + - multimodal + name: Layout of Wind Farms + noise_type: + - none + objectives: null + references: + - authors: [] + link: + type: null + url: https://arxiv.org/abs/2302.01071 + title: Layout of Wind Farms + type: null + source: + - real-world + tags: null + type: problem + variables: + - dim: null + type: continuous +fn_materials_design: + allows_partial_evaluation: unknown + can_evaluate_objectives_independently: unknown + code_examples: null + constraints: [] + description: 'Materials Design Optimization: Select appropriate material and microstructure + optimization for optimizing the light absorbing of a quasi-random solar cell. + Maximize light absorption by balancing the competing processes of light reflection + and scattering.Choose from 5 materials. Another material design problem can be + found here: https://github.com/clancyLab/NCM2018' + dynamic_type: null + evaluation_time: + - expensive simulation + fidelity_levels: + - 1 + implementations: + - impl_materials_design + instances: null + long_name: null + modality: null + name: Materials Design + noise_type: + - none + objectives: null + references: + - authors: [] + link: + type: null + url: https://www.nature.com/articles/s41598-020-60652-9 + title: Materials Design + type: null + source: + - real-world + tags: null + type: problem + variables: + - dim: null + type: categorical + - dim: null + type: continuous +fn_miprowa_millireactor_geometry_optimization: + allows_partial_evaluation: 'no' + can_evaluate_objectives_independently: unknown + code_examples: null + constraints: + - equality: null + hard: null + number: null + type: box + description: Optimizing the design of the mulireactor to achieve minimal stagnating + flow and mixing variance + dynamic_type: + - none + evaluation_time: + - 6.5 hours + fidelity_levels: null + implementations: + - impl_miprowa_millireactor_geometry_optimization + instances: null + long_name: Miprowa millireactor design optimization + modality: null + name: Miprowa millireactor geometry optimization + noise_type: null + objectives: null + references: + - authors: [] + link: null + title: Begall, M. J., Schweidtmann, A. M., Mhamdi, A., & Mitsos, A. (2023). Geometry + optimization of a continuous millireactor via CFD and Bayesian optimization. + Computers & chemical engineering, 171, 108140. + type: null + source: + - real-world + tags: null + type: problem + variables: + - dim: null + type: continuous fn_onemax_sphere_deceptive_rotell: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -442,11 +616,11 @@ fn_onemax_sphere_deceptive_rotell: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 - type: binary + type: continuous fn_onemax_sphere_zeromax_sphere: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -479,11 +653,94 @@ fn_onemax_sphere_zeromax_sphere: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 + type: continuous +fn_optimal_sensor_placement_concrete_frame: + allows_partial_evaluation: 'no' + can_evaluate_objectives_independently: 'no' + code_examples: null + constraints: + - equality: null + hard: 'yes' + number: 1 + type: unknown + description: Select sensor locations from a set of options. Here, the goal is to + optimize structural damage prediction (but can be arbitrary) + dynamic_type: + - none + evaluation_time: + - minutes to hours + fidelity_levels: + - 1 + implementations: + - impl_optimal_sensor_placement_concrete_frame + instances: null + long_name: Optimal sensor placement for structural damage segmentation + modality: null + name: Optimal Sensor Placement - concrete frame + noise_type: null + objectives: null + references: + - authors: [] + link: null + title: Sajedi, S., & Liang, X. (2022). Deep generative Bayesian optimization for + sensor placement in structural health monitoring. Computer‐Aided Civil and Infrastructure + Engineering, 37(9), 1109-1127. + type: null + source: + - artificial + tags: + - form-submission + type: problem + variables: + - dim: null type: binary +fn_optimization_of_milireactor: + allows_partial_evaluation: 'no' + can_evaluate_objectives_independently: 'no' + code_examples: null + constraints: [] + description: Polymerization processes in the (specialty) chemical industry increasingly + rely on continuous flow processing. Continuous polymerization reactors are promising, + because their large surface to volume ratio allows for faster heat removal and + better temperature control than batch reactors. Furthermore, this ratio can be + kept constant during scaleup, reducing the need for pilot stages making it easier + to go from lab- to production scale. However, continuous polymerization reactors, + especially milli-scale reactors with internal mixing elements, can be highly complex + devices, which can be difficult to characterize experimentally, as many of their + properties are not readily available via measurements. Additionally, their small + size renders them especially prone to fouling + dynamic_type: + - none + evaluation_time: + - 6.6 hours + fidelity_levels: null + implementations: + - impl_optimization_of_milireactor + instances: null + long_name: Optimization of milireactor + modality: null + name: Optimization of milireactor + noise_type: + - none + objectives: null + references: + - authors: [] + link: null + title: 'M. J. Begall, A. M. Schweidtmann, A. Mhamdi, and A. Mitsos, “Geometry + optimization of a continuous millireactor via CFD and Bayesian optimization,” + Computers & Chemical Engineering, vol. 171, p. 108140, Mar. 2023, doi: 10.1016/j.compchemeng.2023.108140.' + type: null + source: + - real-world + tags: null + type: problem + variables: + - dim: null + type: continuous fn_radar_waveform: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -522,6 +779,137 @@ fn_radar_waveform: max: 12 min: 4 type: integer +fn_shape_optimization_of_a_lid_driven_cavity: + allows_partial_evaluation: 'no' + can_evaluate_objectives_independently: 'no' + code_examples: null + constraints: + - equality: null + hard: 'yes' + number: 1 + type: unknown + description: '"Consider an incompressible two-dimensional cavity flow, where all + walls are at rest except the upper wall (lid) which moves with the constant velocity + U0 in the positive horizontal direction, x. The aim is to optimize the shape of + the left and right walls so that the energy dissipation is either minimized or + maximized with the constraint of keeping the total volume of the fluid confined + in the cavity fixed and also retain the length and position of the upper and lower + walls. Therefore, the objective is to find the wall shape which either minimizes + or maximizes"' + dynamic_type: + - none + evaluation_time: [] + fidelity_levels: + - 1 + - 2 + implementations: + - impl_shape_optimization_of_a_lid_driven_cavity + instances: null + long_name: null + modality: null + name: Shape optimization of a lid-driven cavity + noise_type: + - none + - though potentially one shall revise the meshing algorithm. + objectives: null + references: null + source: + - real-world + tags: + - form-submission + type: problem + variables: + - dim: null + type: continuous +fn_simulation_driven_antenna_design: + allows_partial_evaluation: 'no' + can_evaluate_objectives_independently: unknown + code_examples: null + constraints: + - equality: null + hard: null + number: null + type: box + description: 'Simulation-Driven Antenna Design Optimization: One scalar merit function + encoding the design specifications (gain, voltage standing wave ratio,..) calculated + by using a high-fidelity EM simulator of the antenna under design. The decision + variables are designable parameters like geometry parameters and they are box-constrained.' + dynamic_type: null + evaluation_time: + - expensive + fidelity_levels: + - 1 + implementations: + - impl_simulation_driven_antenna_design + instances: null + long_name: null + modality: null + name: Simulation-Driven Antenna Design + noise_type: null + objectives: null + references: + - authors: [] + link: + type: null + url: https://ieeexplore.ieee.org/abstract/document/9298949 + title: Simulation-Driven Antenna Design + type: null + source: + - real-world + tags: null + type: problem + variables: + - dim: null + type: continuous +fn_transverse_bending_under_pole_impact: + allows_partial_evaluation: 'no' + can_evaluate_objectives_independently: null + code_examples: null + constraints: + - equality: null + hard: 'yes' + number: 1 + type: box + - equality: null + hard: 'yes' + number: 1 + type: unknown + description: A beam, simply supported at both ends, is struck laterally by a rigid + pole, mimicking side-pole crash tests. The objective is to minimize intrusion, + defined as the maximal dynamic displacement of the impactor before rebound, subject + to a volume constraint of at most 50% of the design domain. The design is parameterized + by a level-set function with 15 variables, mapped onto an LS-Dyna mesh of 1,600 + eight-node solid elements with piecewise linear plasticity. Void elements are + deleted from the mesh, and the out-of-plane (z) direction is fixed. + dynamic_type: + - none + evaluation_time: + - 2 minutes + fidelity_levels: + - 1 + implementations: + - impl_transverse_bending_under_pole_impact + instances: null + long_name: null + modality: null + name: Transverse bending under pole impact + noise_type: + - none + objectives: null + references: + - authors: [] + link: + type: null + url: http://www.sciencedirect.com/science/article/pii/S0045782519300726 + title: Transverse bending under pole impact + type: null + source: + - real-world + tags: null + type: problem + variables: + - dim: null + type: continuous gen_beacon: allows_partial_evaluation: 'no' can_evaluate_objectives_independently: null @@ -648,7 +1036,7 @@ gen_ealain: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 @@ -656,7 +1044,7 @@ gen_ealain: - dim: max: null min: 1 - type: binary + type: continuous gen_gnbg: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -892,8 +1280,8 @@ gen_mubqp: instances: null long_name: null modality: - - quadratic - multimodal + - quadratic name: mUBQP noise_type: null objectives: @@ -996,7 +1384,7 @@ gen_randoptgen: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 @@ -1004,7 +1392,7 @@ gen_randoptgen: - dim: max: null min: 1 - type: binary + type: continuous gen_rho_mnk_landscapes: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -1064,8 +1452,8 @@ gen_rho_mtsp: instances: null long_name: null modality: - - quadratic - multimodal + - quadratic name: ρmTSP noise_type: null objectives: @@ -1150,6 +1538,17 @@ impl_beacon: name: BEACON requirements: null type: implementation +impl_biological_sequence_design: + description: Biological sequence design + evaluation_time: + - expensive + language: Python + links: + - type: repository + url: https://github.com/samuelstanton/lambo + name: Biological sequence design + requirements: null + type: implementation impl_bonobench: description: Bi-objective Numerical Optimization Benchmark (BONO-Bench) evaluation_time: null @@ -1163,8 +1562,8 @@ impl_bonobench: impl_bso_toolbox: description: Building Spatial Design toolbox (TU/e) evaluation_time: - - 1 second - 40 seconds + - 1 second language: C++ links: - type: repository @@ -1214,6 +1613,17 @@ impl_cfd: name: CFD test problem suite requirements: null type: implementation +impl_cfd_based_non_premixed_burner_design: + description: Change the length and width of a combustor optimizing the combustion + efficiency but keeping NOx emissions constrained. + evaluation_time: + - 35-40 minutes + language: Python, Authors disclose using optimizer in Python, but the physical model + happens in a simulation pipeline in FLUENT + links: null + name: CFD-Based Non-Premixed Burner Design + requirements: null + type: implementation impl_coco: description: 'Comparing Continuous Optimizers: black-box optimization benchmarking platform' @@ -1278,8 +1688,8 @@ impl_expobench: description: EXPensive Optimization benchmark library (wind farm layout, gas filter design, pipe shape, hyperparameter tuning, hospital simulation) evaluation_time: - - 2 seconds - 80 seconds + - 2 seconds language: Python links: - type: repository @@ -1329,6 +1739,35 @@ impl_gnbg_ii: name: GNBG-II requirements: null type: implementation +impl_guacamol: + description: 'De novo molecular design is a computational technique to generate + novel compounds with desirable property profiles from scratch. It complements + virtual screening, where large virtual compound libraries are pregenerated, stored, + and then subsequently ranked on demand. Virtual screening is particularly useful + when the virtual compounds are readily available (commercially, or in in-house + screening collections), or easy to synthesize. Data sets sized on the order of + 10^13 can be routinely screened under current computational constraints. However, + this is only a tiny fraction of drug-like chemical space, which has an estimated + size anywhere between 10^24 and 10^60 possible structures. This number might be + even larger considering new modalities such as PROTACs. Efficient approaches to + query larger combinatorial spaces without full enumeration via similarity to given + structures have been reported. However, it is not straightforward to perform more + complex, multiobjective queries without enumerating substantial parts of the space. + + + In contrast, in de novo design, only a relatively small number of molecules is + explicitly generated, and chemical space is explored via search or optimization + procedures. Thus, by focusing on the most relevant areas of chemical space for + the current multiobjective query, de novo design algorithms can in principle explore + the full chemical space, given any ranking or scoring function.' + evaluation_time: null + language: Python + links: + - type: repository + url: https://github.com/BenevolentAI/guacamol + name: GuacaMol + requirements: null + type: implementation impl_iohclustering: description: Clustering-based optimization benchmark built on ML datasets evaluation_time: null @@ -1370,6 +1809,19 @@ impl_jsec2019: name: JSEC 2019 competition requirements: null type: implementation +impl_layout_of_wind_farms: + description: 'Optimizing a Wind Farm Layout: Input variables are coordinates of + the turbines (box-constraints). Constraint that they don''t overlap (hard constraint). + Optimize the overall power output from all wind turbines and with different wind + states. Use a simulation to evaluate a layout. It is expensive approx. 70 CPU + hours.' + evaluation_time: + - 70 CPU hours + language: null + links: null + name: Layout of Wind Farms + requirements: null + type: implementation impl_ma_bbob: description: Example notebook for MA-BBOB in IOHexperimenter evaluation_time: null @@ -1380,11 +1832,24 @@ impl_ma_bbob: name: MA-BBOB (IOHexperimenter) requirements: null type: implementation +impl_materials_design: + description: 'Materials Design Optimization: Select appropriate material and microstructure + optimization for optimizing the light absorbing of a quasi-random solar cell. + Maximize light absorption by balancing the competing processes of light reflection + and scattering.Choose from 5 materials. Another material design problem can be + found here: https://github.com/clancyLab/NCM2018' + evaluation_time: + - expensive simulation + language: null + links: null + name: Materials Design + requirements: null + type: implementation impl_mechbench: description: Structural mechanics design optimization benchmark evaluation_time: - - 1 minute - 7 minutes + - 1 minute language: Python links: - type: repository @@ -1402,6 +1867,16 @@ impl_mf2: name: mf2 requirements: null type: implementation +impl_miprowa_millireactor_geometry_optimization: + description: Optimizing the design of the mulireactor to achieve minimal stagnating + flow and mixing variance + evaluation_time: + - 6.5 hours + language: null + links: null + name: Miprowa millireactor geometry optimization + requirements: null + type: implementation impl_mmopp: description: ECI lab distribution page for MMOPP evaluation_time: null @@ -1453,6 +1928,59 @@ impl_mpm2: name: MPM2 (smoof) requirements: null type: implementation +impl_optimal_sensor_placement_concrete_frame: + description: Select sensor locations from a set of options. Here, the goal is to + optimize structural damage prediction (but can be arbitrary) + evaluation_time: + - minutes to hours + language: null + links: null + name: Optimal Sensor Placement - concrete frame + requirements: null + type: implementation +impl_optimization_of_a_spoiler_ice_airfoil_model: + description: 'When operating in cold climate, the performance of the wind turbines + can be reduced by icing. In an extreme condition, heavy icing can even lead to + a complete stop of the turbine. Icing is also a risky condition for airplane wings + since it may induce flow separation at a small angles of attack, which then might + lead to stall and consequently loss of lift. Ice accretion is a complex process + which depends on both aerodynamics and thermodynamics. The process is affected + by many parameters, for instance, ambient temperature, surface properties, relative + speed of the airfoil, and the time span over which the icing event takes place. + As a result, numerous possibilities for an ice profile exist. Bragg et al. categorized + the leading-edge ice on the airfoils into four types: roughness, horn, streamwise, + and spanwise-ridge ice. The main characteristic of the horn ice is the presence + of the large protuberances which induce a large flow separation downstream of + the ice, and hence dramatically affect the aerodynamic performance, including + the lift, drag and moment coefficients. On the other hand, streamwise ice is smoothly + formed along the streamlines and is less dangerous in the sense of having adverse + effects on the aerodynamic performance.' + evaluation_time: + - The authors disclose it takes 1 hour. + language: Besides Python (as the authors use GPy) for optimizing the problem, there + is the need of a commercial mesher (ANSYS ICEM) + links: null + name: Optimization of a spoiler-ice airfoil model + requirements: null + type: implementation +impl_optimization_of_milireactor: + description: Polymerization processes in the (specialty) chemical industry increasingly + rely on continuous flow processing. Continuous polymerization reactors are promising, + because their large surface to volume ratio allows for faster heat removal and + better temperature control than batch reactors. Furthermore, this ratio can be + kept constant during scaleup, reducing the need for pilot stages making it easier + to go from lab- to production scale. However, continuous polymerization reactors, + especially milli-scale reactors with internal mixing elements, can be highly complex + devices, which can be difficult to characterize experimentally, as many of their + properties are not readily available via measurements. Additionally, their small + size renders them especially prone to fouling + evaluation_time: + - 6.6 hours + language: COMSOL (It's suspected the authors used MATLAB) + links: null + name: Optimization of milireactor + requirements: null + type: implementation impl_puboi: description: A benchmark in which variable importance is tunable, based on the Walsh function @@ -1518,6 +2046,34 @@ impl_reproblems: name: reproblems requirements: null type: implementation +impl_shape_optimization_of_a_lid_driven_cavity: + description: '"Consider an incompressible two-dimensional cavity flow, where all + walls are at rest except the upper wall (lid) which moves with the constant velocity + U0 in the positive horizontal direction, x. The aim is to optimize the shape of + the left and right walls so that the energy dissipation is either minimized or + maximized with the constraint of keeping the total volume of the fluid confined + in the cavity fixed and also retain the length and position of the upper and lower + walls. Therefore, the objective is to find the wall shape which either minimizes + or maximizes"' + evaluation_time: null + language: 'Physical Model: Nek500; CFD Solver: OpenFOAM (Not disclosed the coupling, + but could be Python)' + links: null + name: Shape optimization of a lid-driven cavity + requirements: null + type: implementation +impl_simulation_driven_antenna_design: + description: 'Simulation-Driven Antenna Design Optimization: One scalar merit function + encoding the design specifications (gain, voltage standing wave ratio,..) calculated + by using a high-fidelity EM simulator of the antenna under design. The decision + variables are designable parameters like geometry parameters and they are box-constrained.' + evaluation_time: + - expensive + language: null + links: null + name: Simulation-Driven Antenna Design + requirements: null + type: implementation impl_transfer_rf_bbob_rw: description: Real-world BBOB-like problem implementations (Porkchop, KinematicsRobotArm) evaluation_time: null @@ -1528,12 +2084,27 @@ impl_transfer_rf_bbob_rw: name: Transfer Random Forests BBOB Real-world requirements: null type: implementation +impl_transverse_bending_under_pole_impact: + description: A beam, simply supported at both ends, is struck laterally by a rigid + pole, mimicking side-pole crash tests. The objective is to minimize intrusion, + defined as the maximal dynamic displacement of the impactor before rebound, subject + to a volume constraint of at most 50% of the design domain. The design is parameterized + by a level-set function with 15 variables, mapped onto an LS-Dyna mesh of 1,600 + eight-node solid elements with piecewise linear plasticity. Void elements are + deleted from the mesh, and the out-of-plane (z) direction is fixed. + evaluation_time: + - 2 minutes + language: Python + links: null + name: Transverse bending under pole impact + requirements: null + type: implementation impl_tulipa: description: Large linear program for optimal investment and operation of energy systems evaluation_time: - - minutes - hours + - minutes language: Julia / JuMP links: - type: website @@ -1543,6 +2114,35 @@ impl_tulipa: name: TulipaEnergyModel.jl requirements: null type: implementation +impl_vae_molecule_generation: + description: 'The goal of drug and material design is to identify novel molecules + that have certain desirable properties. We view this as an optimization problem, + in which we are searching for the molecules that maximize our quantitative desiderata. + + + Making and testing new compounds are costly and time-consuming, and the number + of potential candidates is overwhelming. Only about 10^8 substances have ever + been synthesized, whereas the range of potential drug-like molecules is estimated + to be between 10^23 and 10^60. Virtual screening can be used to speed up this + search. Virtual libraries containing thousands to hundreds of millions of candidates + can be assayed with first-principles simulations or statistical predictions based + on learned proxy models, and only + + the most promising leads are selected and tested experimentally. + + + Ways to assess the generated molecules are the water−octanol partition coefficient + (logP), the synthetic accessibility score (SAS), and Quantitative Estimation of + Drug-likeness (QED), which ranges in value between 0 and 1, with higher values + indicating that the molecule is more drug-like.' + evaluation_time: null + language: Python + links: + - type: repository + url: https://github.com/Ryan-Rhys/Constrained-Bayesian-Optimisation-for-Automatic-Chemical-Design + name: VAE Molecule Generation + requirements: null + type: implementation impl_vehicle_dynamics: description: Zenodo archive for the vehicle dynamics benchmark evaluation_time: null @@ -1595,7 +2195,7 @@ suite_amvop: - dim: max: null min: 1 - type: continuous + type: categorical - dim: max: null min: 1 @@ -1603,7 +2203,7 @@ suite_amvop: - dim: max: null min: 1 - type: categorical + type: continuous suite_bbob: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -1705,11 +2305,11 @@ suite_bbob_biobj_mixint: - dim: max: 160 min: 5 - type: continuous + type: integer - dim: max: 160 min: 5 - type: integer + type: continuous suite_bbob_constrained: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -1815,11 +2415,11 @@ suite_bbob_mixint: - dim: max: 160 min: 5 - type: continuous + type: integer - dim: max: 160 min: 5 - type: integer + type: continuous suite_bbob_noisy: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -1855,6 +2455,48 @@ suite_bbob_noisy: max: null min: 1 type: continuous +suite_biological_sequence_design: + allows_partial_evaluation: 'no' + can_evaluate_objectives_independently: unknown + code_examples: null + constraints: + - equality: null + hard: null + number: null + type: unknown + description: Biological sequence design + dynamic_type: + - none + evaluation_time: + - expensive + fidelity_levels: + - 1 + - 2 + implementations: + - impl_biological_sequence_design + long_name: null + modality: + - multimodal + name: Biological sequence design + noise_type: + - observational + objectives: null + problems: null + references: + - authors: [] + link: + type: null + url: https://arxiv.org/pdf/2203.12742 + title: Biological sequence design + type: null + source: + - real-world + tags: + - form-submission + type: suite + variables: + - dim: null + type: categorical suite_bp: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -2248,11 +2890,11 @@ suite_cre: - dim: max: 7 min: 3 - type: continuous + type: integer - dim: max: 7 min: 3 - type: integer + type: continuous suite_cuter: allows_partial_evaluation: 'no' can_evaluate_objectives_independently: null @@ -2289,7 +2931,7 @@ suite_cuter: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 @@ -2297,22 +2939,22 @@ suite_cuter: - dim: max: null min: 1 - type: binary + type: continuous suite_cutest: allows_partial_evaluation: 'no' can_evaluate_objectives_independently: null code_examples: null constraints: - - equality: null - hard: 'yes' - number: null - type: box - equality: null hard: some number: max: null min: 1 type: unknown + - equality: null + hard: 'yes' + number: null + type: box description: CUTEst for optimization software dynamic_type: null evaluation_time: null @@ -2342,7 +2984,7 @@ suite_cutest: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 @@ -2350,7 +2992,7 @@ suite_cutest: - dim: max: null min: 1 - type: binary + type: continuous suite_dtlz: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -2552,14 +3194,14 @@ suite_expobench: can_evaluate_objectives_independently: null code_examples: null constraints: - - equality: null - hard: 'yes' - number: null - type: box - equality: null hard: some number: null type: unknown + - equality: null + hard: 'yes' + number: null + type: box description: Wind farm layout optimization, gas filter design, pipe shape optimization, hyperparameter tuning, and hospital simulation dynamic_type: null @@ -2591,15 +3233,15 @@ suite_expobench: - dim: max: 135 min: 10 - type: integer + type: categorical - dim: max: 135 min: 10 - type: categorical + type: continuous - dim: max: 135 min: 10 - type: continuous + type: integer suite_gbea: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -2708,6 +3350,64 @@ suite_gnbg_ii: max: null min: 1 type: continuous +suite_guacamol: + allows_partial_evaluation: unknown + can_evaluate_objectives_independently: null + code_examples: null + constraints: + - equality: null + hard: 'yes' + number: null + type: unknown + description: 'De novo molecular design is a computational technique to generate + novel compounds with desirable property profiles from scratch. It complements + virtual screening, where large virtual compound libraries are pregenerated, stored, + and then subsequently ranked on demand. Virtual screening is particularly useful + when the virtual compounds are readily available (commercially, or in in-house + screening collections), or easy to synthesize. Data sets sized on the order of + 10^13 can be routinely screened under current computational constraints. However, + this is only a tiny fraction of drug-like chemical space, which has an estimated + size anywhere between 10^24 and 10^60 possible structures. This number might be + even larger considering new modalities such as PROTACs. Efficient approaches to + query larger combinatorial spaces without full enumeration via similarity to given + structures have been reported. However, it is not straightforward to perform more + complex, multiobjective queries without enumerating substantial parts of the space. + + + In contrast, in de novo design, only a relatively small number of molecules is + explicitly generated, and chemical space is explored via search or optimization + procedures. Thus, by focusing on the most relevant areas of chemical space for + the current multiobjective query, de novo design algorithms can in principle explore + the full chemical space, given any ranking or scoring function.' + dynamic_type: + - none + evaluation_time: null + fidelity_levels: + - 1 + - 2 + implementations: + - impl_guacamol + long_name: 'GuacaMol: Benchmarking Models for de Novo Molecular Design' + modality: null + name: GuacaMol + noise_type: + - none + objectives: null + problems: null + references: + - authors: [] + link: + type: null + url: https://doi.org/10.1021/acs.jcim.8b00839 + title: GuacaMol + type: null + source: + - real-world + tags: null + type: suite + variables: + - dim: null + type: unknown suite_iohclustering: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -2808,11 +3508,11 @@ suite_l1_zdt: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 - type: binary + type: continuous suite_l2_dtlz: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -2884,11 +3584,11 @@ suite_l2_zdt: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 - type: binary + type: continuous suite_l3_dtlz: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -2960,11 +3660,11 @@ suite_l3_zdt: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 - type: binary + type: continuous suite_maop: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -3220,8 +3920,8 @@ suite_modact: evaluation_time: null fidelity_levels: null implementations: - - impl_pymoo - impl_modact + - impl_pymoo long_name: multiobjective design of actuators modality: null name: MODAct @@ -3244,10 +3944,10 @@ suite_modact: tags: null type: suite variables: - - dim: 20 - type: integer - dim: 20 type: continuous + - dim: 20 + type: integer suite_morepo: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -3279,6 +3979,62 @@ suite_morepo: variables: - dim: 0 type: unknown +suite_optimization_of_a_spoiler_ice_airfoil_model: + allows_partial_evaluation: 'no' + can_evaluate_objectives_independently: 'no' + code_examples: null + constraints: + - equality: null + hard: null + number: 2 + type: box + description: 'When operating in cold climate, the performance of the wind turbines + can be reduced by icing. In an extreme condition, heavy icing can even lead to + a complete stop of the turbine. Icing is also a risky condition for airplane wings + since it may induce flow separation at a small angles of attack, which then might + lead to stall and consequently loss of lift. Ice accretion is a complex process + which depends on both aerodynamics and thermodynamics. The process is affected + by many parameters, for instance, ambient temperature, surface properties, relative + speed of the airfoil, and the time span over which the icing event takes place. + As a result, numerous possibilities for an ice profile exist. Bragg et al. categorized + the leading-edge ice on the airfoils into four types: roughness, horn, streamwise, + and spanwise-ridge ice. The main characteristic of the horn ice is the presence + of the large protuberances which induce a large flow separation downstream of + the ice, and hence dramatically affect the aerodynamic performance, including + the lift, drag and moment coefficients. On the other hand, streamwise ice is smoothly + formed along the streamlines and is less dangerous in the sense of having adverse + effects on the aerodynamic performance.' + dynamic_type: + - none + evaluation_time: + - The authors disclose it takes 1 hour. + fidelity_levels: + - 1 + - 2 + implementations: + - impl_optimization_of_a_spoiler_ice_airfoil_model + long_name: Optimization of a spoiler-ice airfoil model + modality: null + name: Optimization of a spoiler-ice airfoil model + noise_type: + - none + objectives: null + problems: null + references: + - authors: [] + link: null + title: 'Y. Morita, S. Rezaeiravesh, N. Tabatabaei, R. Vinuesa, K. Fukagata, and + P. Schlatter, “Applying Bayesian Optimization with Gaussian Process Regression + to Computational Fluid Dynamics Problems,” Journal of Computational Physics, + vol. 449, p. 110788, Jan. 2022, doi: 10.1016/j.jcp.2021.110788.' + type: null + source: + - real-world + tags: null + type: suite + variables: + - dim: null + type: continuous suite_pbo: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -3388,11 +4144,11 @@ suite_re: - dim: max: 7 min: 2 - type: integer + type: continuous - dim: max: 7 min: 2 - type: continuous + type: integer suite_rwmvop: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -3429,7 +4185,7 @@ suite_rwmvop: - dim: max: null min: 1 - type: continuous + type: categorical - dim: max: null min: 1 @@ -3437,7 +4193,7 @@ suite_rwmvop: - dim: max: null min: 1 - type: categorical + type: continuous suite_sbox_cost: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -3515,6 +4271,33 @@ suite_sdp: max: null min: 1 type: continuous +suite_shape_optimization_of_a_diverging_channel: + allows_partial_evaluation: 'no' + can_evaluate_objectives_independently: null + code_examples: null + constraints: [] + description: The objective of this section is to efficiently obtain the optimal + shape of the upper wall of a channel so that a target β along the edge of the + Turbulent Boundary Layer over the lower wall is achieved. + dynamic_type: + - none + evaluation_time: null + fidelity_levels: null + implementations: null + long_name: null + modality: null + name: Shape optimization of a diverging channel + noise_type: null + objectives: null + problems: null + references: null + source: + - real-world + tags: null + type: suite + variables: + - dim: null + type: continuous suite_submodular: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -3556,11 +4339,11 @@ suite_tulipa_energy: code_examples: null constraints: - equality: null - hard: 'yes' + hard: some number: null type: unknown - equality: null - hard: some + hard: 'yes' number: null type: unknown description: Determine the optimal investment and operation decisions for different @@ -3599,6 +4382,65 @@ suite_tulipa_energy: max: null min: 1 type: continuous +suite_vae_molecule_generation: + allows_partial_evaluation: unknown + can_evaluate_objectives_independently: null + code_examples: null + constraints: + - equality: null + hard: 'yes' + number: null + type: unknown + description: 'The goal of drug and material design is to identify novel molecules + that have certain desirable properties. We view this as an optimization problem, + in which we are searching for the molecules that maximize our quantitative desiderata. + + + Making and testing new compounds are costly and time-consuming, and the number + of potential candidates is overwhelming. Only about 10^8 substances have ever + been synthesized, whereas the range of potential drug-like molecules is estimated + to be between 10^23 and 10^60. Virtual screening can be used to speed up this + search. Virtual libraries containing thousands to hundreds of millions of candidates + can be assayed with first-principles simulations or statistical predictions based + on learned proxy models, and only + + the most promising leads are selected and tested experimentally. + + + Ways to assess the generated molecules are the water−octanol partition coefficient + (logP), the synthetic accessibility score (SAS), and Quantitative Estimation of + Drug-likeness (QED), which ranges in value between 0 and 1, with higher values + indicating that the molecule is more drug-like.' + dynamic_type: + - none + evaluation_time: null + fidelity_levels: + - 1 + - 2 + implementations: + - impl_vae_molecule_generation + long_name: null + modality: null + name: VAE Molecule Generation + noise_type: + - none + objectives: null + problems: null + references: + - authors: [] + link: + type: null + url: https://doi.org/10.1039/c9sc04026a + title: VAE Molecule Generation + type: null + source: + - real-world + tags: + - form-submission + type: suite + variables: + - dim: null + type: unknown suite_vehicle_dynamics: allows_partial_evaluation: null can_evaluate_objectives_independently: null @@ -3712,8 +4554,8 @@ suite_zdt: - dim: max: null min: 1 - type: continuous + type: binary - dim: max: null min: 1 - type: binary + type: continuous