Machine learning assisted design of new lattice core for sandwich structures with superior load carrying capacity

The spatial distribution of material phases within a periodic composite can be engineered to produce band gaps in its frequency spectrum. Applications for such composite materials include vibration and sound isolation. Previous research focused on utilizing topology optimization techniques to design two-dimensional periodic materials with a maximized band gap around a particular frequency or between two particular dispersion branches. While sizable band gaps can be realized, the possibility remains that the frequency bandwidth of the load that is to be isolated might significantly exceed the size of the band gap. In this paper, genetic algorithms are used to design squared bi-material unit cells with a maximized sum of relative band-gap widths over a prescribed frequency range of interest. The optimized unit cells therefore exhibit broadband frequency isolation characteristics. The effects of the ratios of contrasting material properties are also studied. The designed cells are subsequently used, with varying levels of material damping, to form a finite vibration isolation structure, which is subjected to broadband loading conditions. Excellent isolation properties of the synthesized material are demonstrated for this structure.

An investigation into parametric modeling of lattice structures is conducted in pursuit of an effective and efficient process for manufacturing lattice structures via masked stereolithography apparatus (MSLA). Principals of Design for Additive Manufacturing (DfAM) are applied, keeping in mind the manufacturing process to be used. Key design features are parameterized to decrease the effort required to design and implement changes to the structures. Unit cells are modeled as interconnected struts according to parameters such as unit cell side length and strut diameter or thickness. As well as parameterizing the unit cell geometry, the number of cells in each of the three cartesian directions are parameterized such that a three dimensional array of unit cells is constructed based on the single unit cell. Three unit cells are to be considered: the simple cubic (SC) unit cell, the body-centered cubic + simple cubic (BCC|SC) unit cell, and the face-centered cubic + simple cubic (FCC|SC) unit cell. A simple procedure for interfacing the lattice to bulk material is proposed. The proposed design process is validated through qualitative analysis of lattice samples designed with varying unit cell and overall lattice properties with results supporting future studies. Additionally, an example of replacing a solid model’s interior with the lattice structure is detailed to demonstrate the simplicity and power of the approach. It is found that even very thin features with maximum dimensions less than 10 pixels or layers depending on feature orientation (0.34mm and 0.5mm, respectively) are able to be manufactured using affordable consumer grade equipment, which allows for a multitude of potential applications from general purpose lightening to primitive object strength augmentation.

ABSTRACTThis paper extends the energy‐based homogenization method (EBHM) to evaluate the effective elastic properties of two‐dimensional arc unit cells, where the periodic boundary conditions are constructed based on the polar strain–displacement relationships, and the displacement field is solved by the polar finite element method. Combining the extended EBHM with the parametric level set method (PLSM) and the augmented Lagrangian multiplier method (ALM), it is able to design the topological configurations of arc unit cells for seeking superior effective elastic properties. In addition, the level set function is first filtered by compactly supported radial basis functions (CSRBFs) in order to eliminate small holes in the configuration. Numerical examples are performed to demonstrate the advantages of the proposed method. The periodic boundary condition performs well when the circumferential angle of the arc unit cell is in the range of [0.0002, 0.2]. Results indicate that the effective elastic properties of optimized arc unit cells override those of geometrically transformed optimal square unit cells. This superiority mainly benefits from the fact that unisymmetrical configurations are allowed in arc unit cells but not in square unit cells. Results also imply that the capability of nucleation of new holes is retained even though small holes are eliminated by filtering. It contributes to the trade‐off between performance and manufacturability.

Additive Manufacturing (AM) is rapidly evolving due to its unlimited design freedom to fabricate complex and intricate light-weight geometries with the use of lattice structure that have potential applications including construction, aerospace and biomedical applications, where mechanical properties are the prime focus. Buckling instability in lattice structures is one of the main failure mechanisms that can lead to major failure in structural applications that are subjected to compressive loads, but it has yet to be fully explored. This study aims to investigate the effect of surface-based lattice structure topologies and structured column height on the critical buckling load of lattice structured columns. Four different triply periodic minimal surface (TPMS) lattice topologies were selected and three design configurations (unit cells in x, y, z axis), i.e., 2 × 2 × 4, 2 × 2 × 8 and 2 × 2 × 16 column, for each structure were designed followed by printing using HP MultiJet fusion. Uni-axial compression testing was performed to study the variation in critical buckling load due to change in unit cell topology and column height. The results revealed that the structured column possessing Diamond structures shows the highest critical buckling load followed by Neovius and Gyroid structures, whereas the Schwarz-P unit cell showed least resistance to buckling among the unit cells analyzed in this study. In addition to that, the Diamond design showed a uniform decrease in critical buckling load with a column height maximum of 5193 N, which makes it better for applications in which the column’s height is relatively higher while the Schwarz-P design showed advantages for low height column maximum of 2271 N. Overall, the variations of unit cell morphologies greatly affect the critical buckling load and permits the researchers to select different lattice structures for various applications as per load/stiffness requirement with different height and dimensions. Experimental results were validated by finite element analysis (FEA), which showed same patterns of buckling while the numerical values of critical buckling load show the variation to be up to 10%.

Topology optimization of magnetic composite microstructures for electropermanent magnet

Recent advances in topology optimization methods have driven the development of bandgap crystals. These artificial materials with maximized operational bandwidth provide the basis for wave manipulation and investigating the topological phase of matter. However, it is still challenging to efficiently design acoustic bandgap crystals via existing topology optimization methods. Most previous studies considering only a volume fraction constraint on the constituent material may have impractical wide band gaps due to the pseudo-air resonant modes. To solve this issue, this paper establishes a new topology optimization method for creating acoustic bandgap crystals periodically composed of solid and air. We adopt a constraint on air permeability to ensure the connective air channels within the periodic microstructures, which is more applicable in engineering practice. The optimized unit cells from the proposed method are further analyzed to realize topologically protected states, providing opportunities for multi-dimensional wave manipulation in acoustic systems. Numerical examples demonstrate the effectiveness of the proposed method in designing acoustic crystals with broad bandgaps on any given band orders, and gapped/gapless edge states and corner states can be achieved in resulting topological insulators.

The rise of additive manufacturing has enabled new degrees of freedom in terms of design and functionality. In this context, this contribution addresses the design and characterization of structural unit cells that are intended as building blocks of highly porous lattice structures with tailored properties. While typical lattice structures are often composed of gyroid or diamond lattices, this study presents stackable unit cells of different sizes created by a generative design approach to meet boundary conditions such as printability and homogeneous stress distributions under a given mechanical load. Suitable laser powder bed fusion (LPBF) parameters were determined for AlSi10Mg to ensure high resolution and process reproducibility for all considered unit cells. Stacks of unit cells were integrated into tensile and pressure test specimens for which the mechanical performance of the cells was evaluated. Experimentally measured material properties, applied process parameters, and mechanical test results were employed for calibration and validation of finite element (FE) simulations of both the LPBF process as well as the subsequent mechanical characterization. The obtained data therefore provides the basis to combine the different unit cells into tailored lattice structures and to numerically investigate the local variation of properties in the resulting structures.

Despite recent advancements in using machine learning (ML) techniques to establish the microstructure-property linkage for composites’ representative volume elements (RVEs), challenges persist in effectively characterising the effect of microstructural randomness on material properties. This complexity arises from the difficulty of expressing randomness as definitive variables and its intertwined relations with other factors, such as material constituents. Such complexities result in limitations in generalising ML models across different material constituents. Conventional solutions to these challenges usually necessitate large datasets, which require considerable computational resources, for an accurate and generalisable ML models to be trained. This paper presents an innovative approach to tackling these challenges by integrating a high-accuracy convolutional neural network (CNN) with a novel microstructure-factored constitutive model (MCM). The MCM, rooted from classic empirical constitutive modelling, effectively segregates the microstructural and constituting material effects, extending the generalisability and thus significantly enhancing the efficacy of the CNN. This new approach enabled a CNN trained on the transverse stress-strain curves of one set of material constituents (CF/PEEK at 270 °C) to be generalised for the rapid prediction of various sets of material constituents at different temperatures, unseen by the CNN during training, with an average mean absolute percentage error around 3 %.

Anisotropic yield models for lattice unit cell structures exploiting orthotropic symmetry

With the popularization of machine learning (ML) techniques and the increased chipset’s performance, the application of ML to pedestrian localization systems has received significant attention in the last years. Several survey papers have attempted to provide a state-of-the-art overview, but they usually limit their scope to a particular type of positioning system or technology. In addition, they are written from the point of view of ML techniques and their practice, not from the point of view of the localization system and the specific problems that ML techniques can help to solve. This article is intended to offer a comprehensive state-of-the-art survey of the ML techniques that have been adopted over the last ten years to improve the performance of pedestrian localization systems, addressing the applicability of ML techniques in this domain, along with the main localization strategies. It concludes by indicating the underlying open issues and challenges associated with the existing systems, and possible future directions in which ML techniques could improve the performance of pedestrian localization systems. Among other open issues, most previous authors have focused their attention on position estimation accuracy, which wastes the potential of ML techniques to improve other performance parameters (e.g., response time, computational complexity, robustness, scalability or energy efficiency). This study shows that there is a strong trend towards the application of supervised learning. Consequently, there are many potential research opportunities in the use of other learning types, such as unsupervised and reinforcement learning, to improve the performance of pedestrian localization systems.

A comprehensive review on detection of plant disease using machine learning and deep learning approaches

Machine Learning (ML) techniques are pervasive tools in various emerging commercial applications, but have to be accommodated by powerful computer systems to process very large data. Although general-purpose CPUs and GPUs have provided straightforward solutions, their energy-efficiencies are limited due to their excessive supports for flexibility. Hardware accelerators may achieve better energy-efficiencies, but each accelerator often accommodates only a single ML technique (family). According to the famous No-Free-Lunch theorem in the ML domain, however, an ML technique performs well on a dataset may perform poorly on another dataset, which implies that such accelerator may sometimes lead to poor learning accuracy. Even if regardless of the learning accuracy, such accelerator can still become inapplicable simply because the concrete ML task is altered, or the user chooses another ML technique.In this study, we present an ML accelerator called PuDianNao, which accommodates seven representative ML techniques, including k-means, k-nearest neighbors, naive bayes, support vector machine, linear regression, classification tree, and deep neural network. Benefited from our thorough analysis on computational primitives and locality properties of different ML techniques, PuDianNao can perform up to 1056 GOP/s (e.g., additions and multiplications) in an area of 3.51 mm^2, and consumes 596 mW only. Compared with the NVIDIA K20M GPU (28nm process), PuDianNao (65nm process) is 1.20x faster, and can reduce the energy by 128.41x.

Owing to the present scenario of industries, a massive demand for sustainable green materials made of natural fibre is provoking. Besides, the cost involved in experimental trails could be reduced. Perhaps, experimental never reflects the ideal conditions of any materials system due to their natural heterogeneity. In the present study, an attempt is made to develop a representative volume element (RVE)-based micromechanical model to evaluate mechanical properties of pineapple leaf fibre (PALF) composites numerically before being fabricated really. A 3D model of RVE is prepared using finite element analysis software ANSYS®15 in the unit cell. To model the perfect fibre–matrix bonding, RVE modelled with both the square and hexagonal array of packaging. Results on longitudinal modulus, transverse modulus, in-plane Poisson’s ratio and shear modulus of PALF composites as a function of varying fibre loading (10–50 wt% in steps of 10) have been done. Present numerical prediction (RVE) for PALF composites is compared with different analytical models like parallel and series model, Hirsah’s model and Halpin–Tsai model and concluded with proper agreements.

Machine learning (ML) techniques have numerous applications in many fields, including healthcare, medicine, finance, marketing, and cyber security. For example, ML techniques are being applied to determine whether to give a loan to a customer or whether the computing system has been attacked. However, the ML techniques themselves may be subject to attacks and may discriminate when determining who should get the loan. Therefore, the ML techniques have to be secure, ensure privacy of the individuals, incorporate fairness and be accurate. Such collection of ML techniques has come to be known as trustworthy machine learning (trustworthy ML). This article describes an architecture to support scalable trustworthy ML and describes the features that have to be incorporated into the ML techniques to ensure that they are trustworthy.

Code smells are considered symptoms of poor implementation choices, which may hamper the software maintainability. Hence, code smells should be detected as early as possible to avoid software quality degradation. Unfortunately, detecting code smells is not a trivial task. Some preliminary studies investigated and concluded that machine learning (ML) techniques are a promising way to better support smell detection. However, these techniques are hard to be customized to promote an early and accurate detection of specific smell types. Yet, ML techniques usually require numerous code examples to be trained (composing a relevant dataset) in order to achieve satisfactory accuracy. Unfortunately, such a dependency on a large validated dataset is impractical and leads to late detection of code smells. Thus, a prevailing challenge is the early customized detection of code smells taking into account the typical limited training data. In this direction, this paper reports a study in which we collected code smells, from ten active projects, that were actually refactored by developers, differently from studies that rely on code smells inferred by researchers. These smells were used for evaluating the accuracy regarding early detection of code smells by using seven ML techniques. Once we take into account such smells that were considered as important by developers, the ML techniques are able to customize the detection in order to focus on smells observed as relevant in the investigated systems. The results showed that all the analyzed techniques are sensitive to the type of smell and obtained good results for the majority of them, especially JRip and Random Forest. We also observe that the ML techniques did not need a high number of examples to reach their best accuracy results. This finding implies that ML techniques can be successfully used for early detection of smells without depending on the curation of a large dataset.