Iterative neural networks (INN) are rapidly gaining attention for solving inverse problems in imaging, image processing, and computer vision. INNs combine regression NNs and an iterative model-based image reconstruction (MBIR) algorithm, often leading to both good generalization capability and outperforming reconstruction quality over existing MBIR optimization models. This paper proposes the first fast and convergent INN architecture, Momentum-Net, by generalizing a block-wise MBIR algorithm that uses momentum and majorizers with regression NNs. For fast MBIR, Momentum-Net uses momentum terms in extrapolation modules, and noniterative MBIR modules at each iteration by using majorizers, where each iteration of Momentum-Net consists of three core modules: image refining, extrapolation, and MBIR. Momentum-Net guarantees convergence to a fixed-point for general differentiable (non)convex MBIR functions (or data-fit terms) and convex feasible sets, under two asymptomatic conditions. To consider data-fit variations across training and testing samples, we also propose a regularization parameter selection scheme based on the “spectral spread” of majorization matrices. Numerical experiments for light-field photography using a focal stack and sparse-view computational tomography demonstrate that, given identical regression NN architectures, Momentum-Net significantly improves MBIR speed and accuracy over several existing INNs; it significantly improves reconstruction quality compared to a state-of-the-art MBIR method in each application
This content will become publicly available on February 1, 2023
Inverse Design Framework With Invertible Neural Networks for Passive Vibration Suppression in Phononic Structures
Abstract Automated inverse design methods are critical to the development of metamaterial systems that exhibit special user-demanded properties. While machine learning approaches represent an emerging paradigm in the design of metamaterial structures, the ability to retrieve inverse designs on-demand remains lacking. Such an ability can be useful in accelerating optimization-based inverse design processes. This paper develops an inverse design framework that provides this capability through the novel usage of invertible neural networks (INNs). We exploit an INN architecture that can be trained to perform forward prediction over a set of high-fidelity samples and automatically learns the reverse mapping with guaranteed invertibility. We apply this INN for modeling the frequency response of periodic and aperiodic phononic structures, with the performance demonstrated on vibration suppression of drill pipes. Training and testing samples are generated by employing a transfer matrix method. The INN models provide competitive forward and inverse prediction performance compared to typical deep neural networks (DNNs). These INN models are used to retrieve approximate inverse designs for a queried non-resonant frequency range; the inverse designs are then used to initialize a constrained gradient-based optimization process to find a more accurate inverse design that also minimizes mass. The INN-initialized optimizations are found more »
- Publication Date:
- NSF-PAR ID:
- 10301028
- Journal Name:
- Journal of Mechanical Design
- Volume:
- 144
- Issue:
- 2
- ISSN:
- 1050-0472
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Topology optimization has been proved to be an efficient tool for structural design. In recent years, the focus of structural topology optimization has been shifting from single material continuum structures to multimaterial and multiscale structures. This paper aims at devising a numerical scheme for designing bionic structures by combining a two-stage parametric level set topology optimization with the conformal mapping method. At the first stage, the macro-structural topology and the effective material properties are optimized simultaneously. At the second stage, another structural topology optimization is carried out to identify the exact layout of the metamaterial at the mesoscale. The achieved structure and metamaterial designs are further synthesized to form a multiscale structure using conformal mapping, which mimics the bionic structures with “orderly chaos” features. In this research, a multi-control-point conformal mapping (MCM) based on Ricci flow is proposed. Compared with conventional conformal mapping with only four control points, the proposed MCM scheme can provide more flexibility and adaptivity in handling complex geometries. To make the effective mechanical properties of the metamaterials invariant after conformal mapping, a variable-thickness structure method is proposed. Three 2D numerical examples using MCM schemes are presented, and their results and performances are compared. The achievedmore »
-
Abstract. We detail a new prediction-oriented procedure aimed at volcanic hazardassessment based on geophysical mass flow models constrained withheterogeneous and poorly defined data. Our method relies on an itemizedapplication of the empirical falsification principle over an arbitrarily wideenvelope of possible input conditions. We thus provide a first step towards aobjective and partially automated experimental design construction. Inparticular, instead of fully calibrating model inputs on past observations,we create and explore more general requirements of consistency, and then weseparately use each piece of empirical data to remove those input values thatare not compatible with it. Hence, partial solutions are defined to the inverseproblem. This has several advantages compared to a traditionally posedinverse problem: (i) the potentially nonempty inverse images of partialsolutions of multiple possible forward models characterize the solutions tothe inverse problem; (ii) the partial solutions can provide hazard estimatesunder weaker constraints, potentially including extreme cases that areimportant for hazard analysis; (iii) if multiple models are applicable,specific performance scores against each piece of empirical information canbe calculated. We apply our procedure to the case study of the Atenquiquevolcaniclastic debris flow, which occurred on the flanks of Nevado de Colimavolcano (Mexico), 1955. We adopt and compare three depth-averaged modelscurrently implemented in the TITAN2D solver, available frommore »
-
Cellular materials have been widely used in load carrying lightweight structures. Although lightweight increases natural frequency, low stiffness of cellular structures reduces natural frequency. Designing structures with higher natural frequency can usually avoid resonance. In addition, because of the less amount of materials used in cellular structures, the energy absorption capability usually decreases such as under impact loading. Therefore, designing cellular structures with higher natural frequency and higher energy absorption capability is highly desired. In this study, machine learning and novel inverse design techniques enable to search a huge space of unexplored structural designs. In this study, machine learning regression and Generative Neural Networks (GANs) were used to form an inverse design framework. Optimal cellular unit cells that surpass the performance of biomimetic structures inspired from honeycomb, plant stems and trabecular bone in terms of natural frequency and impact resistance were discovered using machine learning. The discovered optimal cellular unit cells exhibited 30–100% higher natural frequency and 300% higher energy absorption than those of the biomimetic counterparts. The discovered optimal unit cells were validated through experimental and simulation comparisons. The machine learning framework in this study would help in designing load carrying engineering structures with increased natural frequency and enhancedmore »
-
Design optimization of metamaterials and other complex systems often relies on the use of computationally expensive models. This makes it challenging to use global multi-objective optimization approaches that require many function evaluations. Engineers often have heuristics or rules of thumb with potential to drastically reduce the number of function evaluations needed to achieve good convergence. Recent research has demonstrated that these design heuristics can be used explicitly in design optimization, indeed leading to accelerated convergence. However, these approaches have only been demonstrated on specific problems, the performance of different methods was diverse, and despite all heuristics being correct'', some heuristics were found to perform much better than others for various problems. In this paper, we describe a case study in design heuristics for a simple class of 2D constrained multiobjective optimization problems involving lattice-based metamaterial design. Design heuristics are strategically incorporated into the design search and the heuristics-enabled optimization framework is compared with the standard optimization framework not using the heuristics. Results indicate that leveraging design heuristics for design optimization can help in reaching the optimal designs faster. We also identify some guidelines to help designers choose design heuristics and methods to incorporate them for a given problem at hand.
