Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Inverse Optimization”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 901 records · Page 50

Three-dimensional particle-in-cell simulations of laser-driven multiradiation sources based on double-layer targets

Double-layer targets (DLTs), made of a low-density foam on top of a solid substrate, can efficiently convert the energy of a high-intensity laser to provide sources of photons and protons. We investigate a 30-fs pulse with a peak intensity of I ~ 8.7×10 20 W/cm 2 and a peak power of ~120 TW interacting with a DLT using three-dimensional (3D) particle-in-cell simulations. We focus on providing quantitative results in full 3D geometry on the foam thickness dependence; on the competition between two photon-generating processes in DLTs, i.e., nonlinear inverse Compton scattering (NICS) and bremsstrahlung (BS); and on the acceleration of protons via enhanced target-normal sheath acceleration. Here, we discuss conversion efficiency, average energy, and angular distributions of such multiradiation sources. We find that NICS can prevail over BS if the DLT's substrate is thin enough (~µ⁢m) and that the optimal foam thickness that maximizes the conversion efficiency in NICS and BS photons and the proton cutoff energy, among those considered, is the same (15µ⁢m). These results show that DLTs constitute an excellent tool for developing relatively compact and optimized laser-driven multicomponent radiation sources.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Multiobjective and categorical global optimization of photonic structures based on ResNet generative neural networks

Abstract We show that deep generative neural networks, based on global optimization networks (GLOnets), can be configured to perform the multiobjective and categorical global optimization of photonic devices. A residual network scheme enables GLOnets to evolve from a deep architecture, which is required to properly search the full design space early in the optimization process, to a shallow network that generates a narrow distribution of globally optimal devices. As a proof-of-concept demonstration, we adapt our method to design thin-film stacks consisting of multiple material types. Benchmarks with known globally optimized antireflection structures indicate that GLOnets can find the global optimum with orders of magnitude faster speeds compared to conventional algorithms. We also demonstrate the utility of our method in complex design tasks with its application to incandescent light filters. These results indicate that advanced concepts in deep learning can push the capabilities of inverse design algorithms for photonics.

Jiang, Jiaqi↗

Estimating Jupiter’s Gravity Field Using Juno Measurements, Trajectory Estimation Analysis, and a Flow Model Optimization

The upcoming Juno spacecraft measurements have the potential of improving our knowledge of Jupiter’s gravity field. The analysis of the Juno Doppler data will provide a very accurate reconstruction of spatial gravity variations, but these measurements will be very accurate only over a limited latitudinal range. In order to deduce the full gravity field of Jupiter, additional information needs to be incorporated into the analysis, especially regarding the Jovian flow structure and its depth, which can influence the measured gravity field. In this study we propose a new iterative method for the estimation of the Jupiter gravity field, using a simulated Juno trajectory, a trajectory estimation model, and an adjoint-based inverse model for the flow dynamics. We test this method both for zonal harmonics only and with a full gravity field including tesseral harmonics. The results show that this method can fit some of the gravitational harmonics better to the “measured” harmonics, mainly because of the added information from the dynamical model, which includes the flow structure. Thus, it is suggested that the method presented here has the potential of improving the accuracy of the expected gravity harmonics estimated from the Juno and Cassini radio science experiments.

79 ASTRONOMY AND ASTROPHYSICS↗

Uncertainty quantification in scientific machine learning: Methods, metrics, and comparisons

Neural networks (NNs) are currently changing the computational paradigm on how to combine data with mathematical laws in physics and engineering in a profound way, tackling challenging inverse and ill-posed problems not solvable with traditional methods. However, quantifying errors and uncertainties in NN-based inference is more complicated than in traditional methods. This is because in addition to aleatoric uncertainty associated with noisy data, there is also uncertainty due to limited data, but also due to NN hyperparameters, overparametrization, optimization and sampling errors as well as model misspecification. Although there are some recent works on uncertainty quantification (UQ) in NNs, there is no systematic investigation of suitable methods towards quantifying the total uncertainty effectively and efficiently even for function approximation, and there is even less work on solving partial differential equations and learning operator mappings between infinite-dimensional function spaces using NNs. In this work, we present a comprehensive framework that includes uncertainty modeling, new and existing solution methods, as well as evaluation metrics and post-hoc improvement approaches. Further, to demonstrate the applicability and reliability of our framework, we present an extensive comparative study in which various methods are tested on prototype problems, including problems with mixed input-output data, and stochastic problems in high dimensions. In the Appendix, we include a comprehensive description of all the UQ methods employed. Further, to help facilitate the deployment of UQ in Scientific Machine Learning research and practice, we present and develop in [1] an open-source Python library (github.com/Crunch-UQ4MI/neuraluq), termed NeuralUQ, that is accompanied by an educational tutorial and additional computational experiments.

11 physics-informed neural networks↗

Evolution of Active Sites in Pt-Based Nanoalloy Catalysts for the Oxidation of Carbonaceous Species by Combined in Situ Infrared Spectroscopy and Total X-ray Scattering

Here, we present results from combined in situ infrared spectroscopy and total X-ray scattering studies on the evolution of catalytically active sites in exemplary binary and ternary Pt-based nanoalloys during a sequence of CO oxidation–reactivation–CO oxidation reactions. We find that when within a particular compositional range, the fresh nanoalloys may exhibit high catalytic activity for low-temperature CO oxidation. Using surface-specific atomic pair distribution functions (PDFs) extracted from the in situ total X-ray scattering data, we find that, regardless of their chemical composition and initial catalytic activity, the fresh nanoalloys suffer a significant surface structural disorder during CO oxidation. Upon reactivation in oxygen atmosphere, the surface of used nanoalloy catalysts both partially oxidizes and orders. Remarkably, it largely retains its structural state when the nanoalloys are reused as CO oxidation catalysts. The seemingly inverse structural changes of studied nanoalloy catalysts occurring under CO oxidation and reactivation conditions affect the active sites on their surface significantly. In particular, through different mechanisms, both appear to reduce the CO binding strength to the nanoalloy’s surface and thus increase the catalytic stability of the nanoalloys. The findings provide clues for further optimization of nanoalloy catalysts for the oxidation of carbonaceous species through optimizing their composition, activation, and reactivation. Besides, the findings demonstrate the usefulness of combined in situ infrared spectroscopy and total X-ray scattering coupled to surface-specific atomic PDF analysis to the ongoing effort to produce advanced catalysts for environmentally and technologically important applications.

36 MATERIALS SCIENCE↗

Orbital-dependent second-order scaled-opposite-spin correlation functionals in the optimized effective potential method

The performance of correlated optimized effective potential (OEP) functionals based on the spin-resolved second-order correlation energy is analysed. The relative importance of singly- and doubly- excited contributions as well as the effect of scaling the same- and opposite- spin components is investigated in detail comparing OEP results with Kohn–Sham (KS) quantities determined via an inversion procedure using accurate ab initio electronic densities. Special attention is dedicated in particular to the recently proposed scaled-opposite–spin OEP functional [I. Grabowski, E. Fabiano, and F. Della Sala, Phys. Rev. B 87, 075103 (2013)] which is the most advantageous from a computational point of view. We find that for high accuracy, a careful, system dependent, selection of the scaling coefficient is required. We analyse several size-extensive approaches for this selection. Finally, we find that a composite approach, named OEP2-SOSh, based on a post-SCF rescaling of the correlation energy can yield high accuracy for many properties, being comparable with the most accurate OEP procedures previously reported in the literature but at substantially reduced computational effort.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

MTUQ: a framework for estimating moment tensors, point forces, and their uncertainties

SUMMARY We introduce MTUQ, an open-source Python package for seismic source estimation and uncertainty quantification, emphasizing flexibility and operational scalability. MTUQ provides MPI-parallelized grid search and global optimization capabilities, compatibility with 1-D and 3-D Green’s function database formats, customizable data processing, C-accelerated waveform and first-motion polarity misfit functions, and utilities for plotting seismic waveforms and visualizing misfit and likelihood surfaces. Applicability to a range of full- and constrained-moment tensor, point force, and centroid inversion problems is possible via a documented application programming interface, accompanied by example scripts and integration tests. We demonstrate the software using three different types of seismic events: (1) a 2009 intraslab earthquake near Anchorage, Alaska; (2) an episode of the 2021 Barry Arm landslide in Alaska; and (3) the 2017 Democratic People’s Republic of Korea underground nuclear test. With these events, we illustrate the well-known complementary character of body waves, surface waves, and polarities for constraining source parameters. We also convey the distinct misfit patterns that arise from each individual data type, the importance of uncertainty quantification for detecting multimodal or otherwise poorly constrained solutions, and the software’s flexible, modular design.

58 GEOSCIENCES↗

SU-F-T-346: Dose Mimicking Inverse Planning Based On Helical Delivery Treatment Plans for Head and Neck Patients

Purpose: We aim to evaluate a new commercial dose mimicking inverse-planning application that was designed to provide cross-platform treatment planning, for its dosimetric quality and efficiency. The clinical benefit of this application allows patients treated on O-shaped linac to receive an equivalent plan on conventional L-shaped linac as needed for workflow or machine downtime. Methods: The dose mimicking optimization process seeks to create a similar DVH of an O-shaped linac-based plans with an alternative treatment technique (IMRT or VMAT), by maintaining target conformity, and penalizing dose falloff outside the target. Ten head and neck (HN) helical delivery plans, including simple and complex cases were selected for re-planning with the dose mimicking application. All plans were generated for a 6 MV beam model, using 7-field/ 9-field IMRT and VMAT techniques. PTV coverage (D1, D99 and homogeneity index [HI]), and OARs avoidance (Dmean / Dmax) were compared. Results: The resulting dose mimicked HN plans achieved acceptable PTV coverage for HI (VMAT 7.0±2.3, 7-fld 7.3±2.4, and 9-fld 7.0±2.4), D99 (98.0%±0.7%, 97.8%±0.7%, and 98.0%±0.7%), as well as D1 (106.4%±2.1%, 106.5%±2.2%, and 106.4%±2.1%), respectively. The OAR dose discrepancy varied: brainstem (2% to 4%), cord (3% to 6%), esophagus (−4% to −8%), larynx (−4% to 2%), and parotid (4% to 14%). Mimicked plans would typically be needed for 1–5 fractions of a treatment course, and we estimate <1% variance would be introduced in target coverage while maintaining comparable low dose to OARs. All mimicked plans were approved by independent physician and passed patient specific QA within our established tolerance. Conclusion: Dose mimicked plans provide a practical alternative for responding to clinical workflow issues, and provide reliability for patient treatment. The quality of dose mimicking for HN patients highly depends on the delivery technique, field numbers and angles, as well as user selection of structures.

60 APPLIED LIFE SCIENCES↗

Monitoring Offshore CO 2 Sequestration Using Marine CSEM Methods; Constraints Inferred from Field- and Laboratory-Based Gas Hydrate Studies

Offshore geological sequestration of CO 2 offers a viable approach for reducing greenhouse gas emissions into the atmosphere. Strategies include injection of CO 2 into the deep-ocean or ocean-floor sediments, whereby depending on pressure–temperature conditions, CO 2 can be trapped physically, gravitationally, or converted to CO 2 hydrate. Energy-driven research continues to also advance CO 2 -for-CH 4 replacement strategies in the gas hydrate stability zone (GHSZ), producing methane for natural gas needs while sequestering CO 2 . In all cases, safe storage of CO 2 requires reliable monitoring of the targeted CO 2 injection sites and the integrity of the repository over time, including possible leakage. Electromagnetic technologies used for oil and gas exploration, sensitive to electrical conductivity, have long been considered an optimal monitoring method, as CO 2 , similar to hydrocarbons, typically exhibits lower conductivity than the surrounding medium. We apply 3D controlled-source electromagnetic (CSEM) forward modeling code to simulate an evolving CO 2 reservoir in deep-ocean sediments, demonstrating sufficient sensitivity and resolution of CSEM data to detect reservoir changes even before sophisticated inversion of data. Laboratory measurements place further constraints on evaluating certain systems within the GHSZ; notably, CO 2 hydrate is measurably weaker than methane hydrate, and >1 order of magnitude more conductive, properties that may affect site selection, stability, and modeling considerations.

58 GEOSCIENCES↗

Higher-order triangular spectral element method with optimized cubature points for seismic wavefield modeling

The mass-lumped method avoids the cost of inverting the mass matrix and simultaneously maintains spatial accuracy by adopting additional interior integration points, known as cubature points. To date, such points are only known analytically in tensor domains, such as quadrilateral or hexahedral elements. Thus, the diagonal-mass-matrix spectral element method (SEM) in non-tensor domains always relies on numerically computed interpolation points or quadrature points. However, only the cubature points for degrees 1 to 6 are known, which is the reason that we have developed a p-norm-based optimization algorithm to obtain higher-order cubature points. In this way, we obtain and tabulate new cubature points with all positive integration weights for degrees 7 to 9. The dispersion analysis illustrates that the dispersion relation determined from the new optimized cubature points is comparable to that of the mass and stiffness matrices obtained by exact integration. Simultaneously, the Lebesgue constant for the new optimized cubature points indicates its surprisingly good interpolation properties. As a result, such points provide both good interpolation properties and integration accuracy. The Courant–Friedrichs–Lewy (CFL) numbers are tabulated for the conventional Fekete-based triangular spectral element (TSEM), the TSEM with exact integration, and the optimized cubature-based TSEM (OTSEM). A complementary study demonstrates the spectral convergence of the OTSEM. A numerical example conducted on a half-space model demonstrates that the OTSEM improves the accuracy by approximately one order of magnitude compared to the conventional Fekete-based TSEM. In particular, the accuracy of the 7th-order OTSEM is even higher than that of the 14th-order Fekete-based TSEM. Furthermore, the OTSEM produces a result that can compete in accuracy with the quadrilateral SEM (QSEM). The high accuracy of the OTSEM is also tested with a non-flat topography model. In terms of computational efficiency, the OTSEM is more efficient than the Fekete-based TSEM, although it is slightly costlier than the QSEM when a comparable numerical accuracy is required. - Highlights: • Higher-order cubature points for degrees 7 to 9 are developed. • The effects of quadrature rule on the mass and stiffness matrices has been conducted. • The cubature points have always positive integration weights. • Freeing from the inversion of a wide bandwidth mass matrix. • The accuracy of the TSEM has been improved in about one order of magnitude.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Feasibility of using intermediate x-ray energies for highly conformal extracranial radiotherapy

Purpose: To investigate the feasibility of using intermediate energy 2 MV x-rays for extracranial robotic intensity modulated radiation therapy. Methods: Two megavolts flattening filter free x-rays were simulated using the Monte Carlo code MCNP (v4c). A convolution/superposition dose calculation program was tuned to match the Monte Carlo calculation. The modeled 2 MV x-rays and actual 6 MV flattened x-rays from existing Varian Linacs were used in integrated beam orientation and fluence optimization for a head and neck, a liver, a lung, and a partial breast treatment. A column generation algorithm was used for the intensity modulation and beam orientation optimization. Identical optimization parameters were applied in three different planning modes for each site: 2, 6 MV, and dual energy 2/6 MV. Results: Excellent agreement was observed between the convolution/superposition and the Monte Carlo calculated percent depth dose profiles. For the patient plans, overall, the 2/6 MV x-ray plans had the best dosimetry followed by 2 MV only and 6 MV only plans. Between the two single energy plans, the PTV coverage was equivalent but 2 MV x-rays improved organs-at-risk sparing. For the head and neck case, the 2MV plan reduced lips, mandible, tongue, oral cavity, brain, larynx, left and right parotid gland mean doses by 14%, 8%, 4%, 14%, 24%, 6%, 30% and 16%, respectively. For the liver case, the 2 MV plan reduced the liver and body mean doses by 17% and 18%, respectively. For the lung case, lung V20, V10, and V5 were reduced by 13%, 25%, and 30%, respectively. V10 of heart with 2 MV plan was reduced by 59%. For the partial breast treatment, the 2 MV plan reduced the mean dose to the ipsilateral and contralateral lungs by 27% and 47%, respectively. The mean body dose was reduced by 16%. Conclusions: The authors showed the feasibility of using flattening filter free 2 MV x-rays for extracranial treatments as evidenced by equivalent or superior dosimetry compared to 6 MV plans using the same inverse noncoplanar intensity modulated planning method.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

An algorithmic calibration approach to identify globally optimal parameters for constraining the DayCent model

he accurate calibration of complex biogeochemical models is essential for the robust estimation of soil greenhouse gases (GHG) as well as other environmental conditions and parameters that are used in research and policy decisions. DayCent is a popular biogeochemical model used both nationally and internationally for this purpose. Despite DayCent’s popularity, its complex parameter estimation is often based on experts’ knowledge which is somewhat subjective. In this study we used the inverse modelling parameter estimation software (PEST), to calibrate the DayCent model based on sensitivity and identifi- ability analysis. Using previously published N2 O and crop yield data as a basis of our calibration approach, we found that half of the 140 parameters used in this study were the primary drivers of calibration dif- ferences (i.e. the most sensitive) and the remaining parameters could not be identified given the data set and parameter ranges we used in this study. The post calibration results showed improvement over the pre-calibration parameter set based on, a decrease in residual differences 79% for N2O fluxes and 84% for crop yield, and an increase in coefficient of determination 63% for N2O fluxes and 72% for corn yield. The results of our study suggest that future studies need to better characterize germination tem- perature, number of degree-days and temperature dependency of plant growth; these processes were highly sensitive and could not be adequately constrained by the data used in our study. Furthermore, the sensitivity and identifiability analysis was helpful in providing deeper insight for important processes and associated parameters that can lead to further improvement in calibration of DayCent model.

Corn yield↗

THREE-DIMENSIONAL RECONSTRUCTION OF THE DENSITY FIELD: AN SVD APPROACH TO WEAK-LENSING TOMOGRAPHY

We present a new method for constructing three-dimensional mass maps from gravitational lensing shear data. We solve the lensing inversion problem using truncation of singular values (within the context of generalized least-squares estimation) without a priori assumptions about the statistical nature of the signal. This singular value framework allows a quantitative comparison between different filtering methods: we evaluate our method beside the previously explored Wiener-filter approaches. Our method yields near-optimal angular resolution of the lensing reconstruction and allows cluster sized halos to be de-blended robustly. It allows for mass reconstructions which are two to three orders of magnitude faster than the Wiener-filter approach; in particular, we estimate that an all-sky reconstruction with arcminute resolution could be performed on a timescale of hours. We find however that linear, non-parametric reconstructions have a fundamental limitation in the resolution achieved in the redshift direction.

79 ASTRONOMY AND ASTROPHYSICS↗

Maximum a posteriori CMB lensing reconstruction

Gravitational lensing of the CMB is a valuable cosmological signal that correlates to tracers of large-scale structure and acts as a important source of confusion for primordial B-mode polarization. State-of-the-art lensing reconstruction analyses use quadratic estimators, which are easily applicable to data. Yet, these estimators are known to be suboptimal, in particular for polarization, and large improvements are expected to be possible for high signal-to-noise polarization experiments. We develop a method and numerical code, LensIt, that is able to find efficiently the most probable lensing map, introducing no significant approximations to the lensed CMB likelihood, and applicable to beamed and masked data with inhomogeneous noise. It works by iteratively reconstructing the primordial unlensed CMB using a deflection estimate and its inverse, and removing residual lensing from these maps with quadratic estimator techniques. Roughly linear computational cost is maintained due to fast convergence of iterative searches, combined with the local nature of lensing. The technique achieves the maximal improvement in signal to noise expected from analytical considerations on the unmasked parts of the sky. Delensing with this optimal map leads to forecast tensor-to-scalar ratio parameter errors improved by a factor ≃2 compared to the quadratic estimator in a CMB stage IV configuration.

79 ASTRONOMY AND ASTROPHYSICS↗

MO-G-9A-01: Imaging Refresher for Standard of Care Radiation Therapy

Imaging techniques and technology which were previously the domain of diagnostic medicine are becoming increasingly integrated and utilized in radiation therapy (RT) clinical practice. As such, there are a number of specific imaging topics that are highly applicable to modern radiation therapy physics. As imaging becomes more widely integrated into standard clinical radiation oncology practice, the impetus is on RT physicists to be informed and up-to-date on those imaging modalities relevant to the design and delivery of therapeutic radiation treatments. For example, knowing that, for a given situation, a fluid attenuated inversion recovery (FLAIR) image set is most likely what the physician would like to import and contour is helpful, but may not be sufficient to providing the best quality of care. Understanding the physics of how that pulse sequence works and why it is used could help assess its utility and determine if it is the optimal sequence for aiding in that specific clinical situation. It is thus important that clinical medical physicists be able to understand and explain the physics behind the imaging techniques used in all aspects of clinical radiation oncology practice. This session will provide the basic physics for a variety of imaging modalities for applications that are highly relevant to radiation oncology practice: computed tomography (CT) (including kV, MV, cone beam CT [CBCT], and 4DCT), positron emission tomography (PET)/CT, magnetic resonance imaging (MRI), and imaging specific to brachytherapy (including ultrasound and some brachytherapy specific topics in MR). For each unique modality, the image formation process will be reviewed, trade-offs between image quality and other factors (e.g. imaging time or radiation dose) will be clarified, and typically used cases for each modality will be introduced. The current and near-future uses of these modalities and techniques in radiation oncology clinical practice will also be discussed. Learning Objectives: To review the basic physical science principles of CT, PET, MR, and ultrasound imaging. To understand how the images are created, and present their specific role in patient management and treatment planning for therapeutic radiation (both external beam and brachytherapy). To discuss when and how each specific imaging modality is currently used in clinical practice, as well as how they may come to be used in the near future.

60 APPLIED LIFE SCIENCES↗

SU-E-T-507: Internal Dosimetry in Nuclear Medicine Using GATE and XCAT Phantom: A Simulation Study

Purpose Monte Carlo simulations are routinely used for internal dosimetry studies. These studies are conducted with humanoid phantoms such as the XCAT phantom. In this abstract we present the absorbed doses for various pairs of source and target organs using three common radiotracers in nuclear medicine. Methods The GATE software package is used for the Monte Carlo simulations. A typical female XCAT phantom is used as the input. Three radiotracers 153Sm, 131I and 99mTc are studied. The Specific Absorbed Fraction (SAF) for gamma rays (99mTc, 153Sm and 131I) and Specific Fraction (SF) for beta particles (153Sm and 131I) are calculated for all 100 pairs of source target organs including brain, liver, lung, pancreas, kidney, adrenal, spleen, rib bone, bladder and ovaries. Results The source organs themselves gain the highest absorbed dose as compared to other organs. The dose is found to be inversely proportional to distance from the source organ. In SAF results of 153Sm, when the source organ is lung, the rib bone, gain 0.0730 (Kg-1) that is more than lung itself. Conclusion The absorbed dose for various organs was studied in terms of SAF and SF. Such studies hold importance for future therapeutic procedures and optimization of induced radiotracer.

60 APPLIED LIFE SCIENCES↗

Progress and R/D challenges for FCC-ee SRF

The FCC-ee machines present a huge challenge for the RF systems, which need to be adapted to very diverse beam conditions going from moderate energy and high current for the Z machine to high energy and low beam current for the ttbar. This inverse scaling results naturally from a fixed budget for the synchrotron radiation, which the SRF cavities need to compensate. A global solution was elaborated for the FCC Conceptual Design Report (Abada in Eur Phys J Spec Top 228):261–623, 2019), and is referred here as the baseline. Recently, further studies have led to a new optimized baseline, still based on traditional elliptical cavities. In parallel, a novel concept, named the Slotted Waveguide ELLiptical (SWELL), was proposed with the potential of greatly simplified logistics and reduced costs. Under several aspects, all these changes call for enhanced performance of the RF systems. A vigorous R&D program has therefore continued since the publication of the CDR, with the aim of pushing the performance and demonstrating the feasibility of a more advanced baseline and, more recently, of the SWELL option. The progress and challenges of this ambitious program were presented in the dedicated SRF sessions at FCC week 2022 and are summarized in this paper.

47 OTHER INSTRUMENTATION↗

Regularizing Invertible Neural Networks for Airfoil Design Through Dimension Reduction

This work examines how dimension reduction can improve the performance of invertible neural networks (INN) for airfoil design. Design workflows are typically expensive, relying on many evaluations of high fidelity computational fluid dynamics (CFD) models. Furthermore, the inverse design problem is typically ill-posed. That is, multiple valid solutions exist that satisfy the design criteria. Regularization can reduce this inverse design space and simplify the problem. We study the use of subspace-based input dimension reduction to act as a regularizer for the INN model and improve the recovery of new airfoil shapes with desired performance characteristics. We find that the dimension reduction identifies two dominant modes, relating to airfoil thickness and camber, that optimally determine the airfoil's aerodynamics. We demonstrate the capability of the proposed INN model to generate 100 airfoils that satisfy the specific aerodynamic and structural characteristics.

airfoil design↗