Explore the vast range of titles available.

The inverse approach is computationally efficient in aerodynamic design as the desired target performance distribution is prespecified. However, it has some significant limitations that prevent it from achieving full efficiency. First, the iterative procedure should be repeated whenever the specified target distribution changes. Target distribution optimization can be performed to clarify the ambiguity in specifying this distribution, but several additional problems arise in this process such as loss of the representation capacity due to parameterization of the distribution, excessive constraints for a realistic distribution, inaccuracy of quantities of interest due to theoretical/empirical predictions, and the impossibility of explicitly imposing geometric constraints. To deal with these issues, a novel inverse design optimization framework with a two-step deep learning approach is proposed. A variational autoencoder and multi-layer perceptron are used to generate a realistic target distribution and predict the quantities of interest and shape parameters from the generated distribution, respectively. Then, target distribution optimization is performed as the inverse design optimization. The proposed framework applies active learning and transfer learning techniques to improve accuracy and efficiency. Finally, the framework is validated through aerodynamic shape optimizations of the wind turbine airfoil. Their results show that this framework is accurate, efficient, and flexible to be applied to other inverse design engineering applications.

The discovery of novel materials and functional molecules can help to solve some of society's most urgent challenges, ranging from efficient energy harvesting and storage to uncovering novel pharmaceutical drug candidates. Traditionally matter engineering-generally denoted as inverse design-was based massively on human intuition and high-throughput virtual screening. The last few years have seen the emergence of significant interest in computer-inspired designs based on evolutionary or deep learning methods. The major challenge here is that the standard strings molecular representation SMILES shows substantial weaknesses in that task because large fractions of strings do not correspond to valid molecules. Here, we solve this problem at a fundamental level and introduce SELFIES (SELF-referencIng Embedded Strings), a string-based representation of molecules which is 100% robust. Every SELFIES string corresponds to a valid molecule, and SELFIES can represent every molecule. SELFIES can be directly applied in arbitrary machine learning models without the adaptation of the models; each of the generated molecule candidates is valid. In our experiments, the model's internal memory stores two orders of magnitude more diverse molecules than a similar test with SMILES. Furthermore, as all molecules are valid, it allows for explanation and interpretation of the internal working of the generative models.

By codesigning a metaoptical front end in conjunction with an image-processing back end, we demonstrate noise sensitivity and compactness substantially superior to either an optics-only or a computation-only approach, illustrated by two examples: subwavelength imaging and reconstruction of the full polarization coherence matrices of multiple light sources. Our end-to-end inverse designs couple the solution of the full Maxwell equations—exploiting all aspects of wave physics arising in subwavelength scatterers—with inverse-scattering algorithms in a single large-scale optimization involving degrees of freedom. The resulting structures scatter light in a way that is radically different from either a conventional lens or a random microstructure, and suppress the noise sensitivity of the inverse-scattering computation by several orders of magnitude. Incorporating the full wave physics is especially crucial for detecting spectral and polarization information that is discarded by geometric optics and scalar diffraction theory.

Injection molding is a widely used manufacturing technology for the mass production of plastic parts. Despite the importance of process optimization for achieving high quality at a low cost, process conditions have often been heuristically sought by field engineers. Here, we propose two systematic data-driven optimization frameworks for the injection molding process based on a multi-objective Bayesian optimization (MBO) framework and a constrained generative inverse design network (CGIDN) framework. MBO, an extension of Bayesian optimization, uses Gaussian process regression adopting a multidimensional acquisition function based on the concepts of hypervolume and Pareto front. The CGIDN, which is an improved version of the original generative inverse design network (GIDN), uses backpropagation to calculate the analytical gradients of the objective function with respect to design variables. Both methods can be used for multi-objective optimization with trade-off relationships, for example, between the cycle time and deflection after extraction. We demonstrate the applicability of the optimization methods utilizing simulation data from Moldflow software for the manufacturing process of a door trim part. We showed that the optimal process parameters which simultaneously minimized deflection and cycle time were obtained with a relatively small dataset. We expect that in a realistic manufacturing facility, the optimal conditions found from simulations can guide the process design of the injection molding machine, or the proposed methods can be directly utilized because they do not require a very large dataset. We also note that the proposed optimization schemes are readily applicable to the optimization of other types of plastic manufacturing processes.

Finding materials with specific properties is a hot topic in materials science. Traditional materials design relies on empirical and trial-and-error methods, requiring extensive experiments and time, resulting in high costs. With the development of physics, statistics, computer science, and other fields, machine learning offers opportunities for systematically discovering new materials. Especially through machine learning-based inverse design, machine learning algorithms analyze the mapping relationships between materials and their properties to find materials with desired properties. This paper first outlines the basic concepts of materials inverse design and the challenges faced by machine learning-based approaches to materials inverse design. Then, three main inverse design methods—exploration-based, model-based, and optimization-based—are analyzed in the context of different application scenarios. Finally, the applications of inverse design methods in alloys, optical materials, and acoustic materials are elaborated on, and the prospects for materials inverse design are discussed. The authors hope to accelerate the discovery of new materials and provide new possibilities for advancing materials science and innovative design methods.

The computational cost of iterative design methods has been a challenge in aerodynamics. In this research, the data-driven acceleration of an iterative inverse design method was implemented to reduce its computational cost. Although iterative design methods are robust, a lot of unwanted data is generated during their intermediate stages. Inverse design methods rely on correcting an initial geometry based on a given target parameter distribution. The generated data during the early iterations of the inverse design was incorporated into two deep-learning models to accelerate target geometry attainment. The deep learning models were used to recognize the correlation between the pressure distribution and corresponding geometry as well as the meaningful changes of geometry and pressure distribution toward their targets. The deep learning models were validated in viscous and inviscid compressible flows for various benchmark aerodynamics problems. In conclusion, between 70 to 80% computational cost decrease was observed for online uses of the machine learning module with the inverse design algorithm. This approach suggests incorporating machine learning techniques into design algorithms by exploiting the intermediate data for further improvement of them. We draw a new interpretation of learning dynamic changes through consecutive iterations instead of typical time-dependent problems in the use of LSTM network.

On-demand inverse design of acoustic metamaterials (AMs), which aims to retrieve the optimal structure according to given requirements, is still a challenging task owing to the non-unique relationship between physical structures and spectral responses. Here, we propose a probabilistic generation network (PGN) model to unveil this implicit relationship and implement this concept with an acoustic magic-cube absorber. By employing the auto-encoder-like configuration composed of a gate recurrent unit (GRU) and a deep neural network, our PGN model encodes the required spectral response into a latent space. The memory or feedback loop contained in the proposed GRU allows it to effectively recognize sequence characteristics of a spectrum. The method of modeling the inverse problem and retrieving multiple meta structures in a probabilistic generative manner skillfully solves the one-to-many mapping issue that is intractable in deterministic models. Moreover, to meet different sound absorption requirements, we tailored several representative spectra with low-frequency sound absorption characteristics, generating high-precision (MAE<0.06) predicted spectra with multiple meta structures. To further verify the effective prediction of the proposed PGN strategy, the experiment was carried out in a tailored broadband example, whose results coincide with both theoretical and numerical ones. Compared with other 5 networks, the PGN model exhibits higher accuracy and efficiency. Our work offers flexible and diversified solutions for multivalued inverse problems, opening up avenues to realize the on-demand design of AMs.

Chemical looping with oxygen uncoupling (CLOU) materials is actively sought for combustion of carbonaceous materials to achieve complete conversion and capture of carbon dioxide. These materials may play a vital role in reducing atmospheric carbon via negative carbon output. However, there is no one‐size‐fits‐all approach as different operating conditions and feedstocks may require different CLOU materials. As a result, the exploration and discovery of high‐performance CLOU materials can be a slow process. To address this challenge, a high‐throughput inverse machine learning workflow that identifies optimum materials from perovskite oxides for a given set of targets is developed—temperature and Gibbs free energy of oxygen formation. The model is trained on high‐throughput density functional theory calculations of CLOU materials and inverts the materials design process using a genetic algorithm to produce realistic substituted SrFeO3‐δ compositions as output. Using the inverse model, it is able to identify several interesting new families of CLOU materials: Sr1‐xAxFe1‐yByO3‐δ (e.g., A = Ca or K; B = Mg, Bi, Mn, Ni, Co, Cu, or Zn). These materials have shown promising properties, and some of them even outperform the benchmark material in terms of oxygen release kinetics under relevant CLOU operating conditions. Taking the calculated Gibbs free energy of oxygen vacancy formation as inputs, an inverse high‐throughput machine learning workflow is developed to predict chemical formula of perovskite for chemical looping applications. Using this generic algorithm model, several new families of perovskites are identified as useful oxygen carrier materials. Particularly, a predicted new high‐performance oxygen carrier material, Sr0.89K0.11Fe0.80Zn0.20O3‐δ, is experimentally verified.

Turbomachinery optimization based on the inverse design method (IDM) has been investigated in several previous studies, however, due to head constraints, most of these studies have adopted the constant impeller outlet angular momentum (IOAM) distribution, namely the free vortex design, which in turn reduces the optimization effect. To overcome these drawbacks, forced and compound vortex designs are proposed here by parameterizing the IOAM using a parabola, and the effects of different vortex designs on the optimization results of a mixed-flow pump impeller are investigated. First, a baseline mixed-flow pump is simulated and experimentally verified. Second, based on the IDM, the impeller is parameterized for the three vortex designs. Finally, it is optimized by artificial neural network and genetic algorithm to maximize the weighted efficiency at 0.8Q des , 1.0Q des and 1.2Q des , and the results are analyzed. The results show that the weighted efficiency of the forced and compound vortex design is improved by 1.33% and 1.69% respectively compared to free vortex design. The internal flow analysis reveals that the improved efficiency of the compound vortex design can be attributed to the improved impeller-outlet flow regime. Finally, the energy characteristics of the preferred and baseline models are compared using the entropy production method.

Injection molding process parameters have a great impact on plastic production quality, manufacturing cost, and molding efficiency. This study proposes to apply the method of Latin hypercube sampling, and to combine the response surface model and “Constraint Generation Inverse Design Network (CGIDN)” to achieve multi-objective optimization of the injection process, shorten the time to find the optimal process parameters, and improve the production efficiency of plastic parts. Taking the LSR lens array of automotive LED lights as the research object, the residual stress and volume shrinkage were taken as the optimization objectives, and the filling time, melt temperature, maturation time, and maturation pressure were taken as the influencing factors to obtain the optimization target values, and the response surface models between the volume shrinkage rate and the influencing factors were established. Based on the “Constraint-Generated Inverse Design Network”, the optimization was independently sought within the set parameters to obtain the optimal combination of process parameters to meet the injection molding quality of plastic parts. The results showed that the optimal residual stress value and volume shrinkage rate were 11.96 MPa and 4.88%, respectively, in the data set of 20 Latin test samples obtained based on Latin hypercube sampling, and the optimal residual stress value and volume shrinkage rate were 8.47 MPa and 2.83%, respectively, after optimization by the CGIDN method. The optimal process parameters obtained by CGIDN optimization were a melt temperature of 30 °C, filling time of 2.5 s, maturation pressure of 40 MPa, and maturation time of 15 s. The optimization results were obvious and showed the feasibility of the data-driven injection molding process optimization method based on the combination of Latin hypercube sampling and CGIDN.