1,721,002 research outputs found
Constitutive modeling and characterization of C/C-SiC CMC material
No full textCeramic matrix composites (CMCs) are advanced materials that consist of a ceramic matrix reinforced with a high-strength, high-stiffness material, such as carbon fibers. They offer excellent thermal and chemical stability while exhibiting low weight and exceptional mechanical properties. A novel CMC material is the C/C-SiC produced with 2/2 twill weave fabric. It consists of a carbon fiber-reinforced carbon (C) and silicon carbide (SiC) matrix. In this study, a macroscopic non-linear constitutive model accounting for the damage-induced plasticity is proposed for the 2/2 twill weave C/C-SiC composite.In the context of this thesis, a computational model is developed, based on the framework of continuum damage mechanics and general plasticity theory. A potential function inspired by the Tsai-Wu criterion combined with a damage model is used to predict the strain and damage evolution. An exponential damage evolution law is introduced while the coupling of different damage modes is also considered. Moreover, an experimental investigation on the macroscopic mechanical behavior and damage mechanisms of C/C-SiC under in-plane onand off-axis loading conditions is performed. Specimens with 0𝑜, 30𝑜 and 45𝑜 on- and off-axis angles were manufactured and tested under monotonic and cyclic tensile and compression loads. Furthermore, the microstructure of the pristine material and the fracture surfaces of the tested specimens are studied through scanning electron microscopy (SEM). A Bayesian optimization algorithm is finally used to optimize simultaneously the different material parameters based on the experimental test data. The predicted stress-strain curves are in good agreement with the experimental curves, especially in the case of monotonic tensile loading. Both damage initiation and evolution are predicted accurately by the chosen laws and coupling functions. Moreover, the combination of the Tsai-Wu criterion with a damage evolution law is proven to predict the ultimate strength well. Fiber pull-out is observed in tension, while interlaminar and translaminar cracks in compression.This study thus provides an accurate constitutive model, a complete mechanical characterization of the in-plane behavior and a better understanding of the fracture mechanisms of C/C-SiC.Materials Science and Engineerin
Data-Driven Material Characterisation of Multicomponent Atherosclerotic Arteries
No full textCardiovascular diseases continue to be the primary cause of death worldwide, where the buildup of plaque within arterial walls, known as atherosclerosis, is a major contributor to various acute cardiovascular events. Determining the material properties and the resulting stress distributions is crucial in the risk assessment of atherosclerotic plaques, as stress is considered an indicator of plaque vulnerability. Material models can be found with stress-strain pairs, but experimentally determining stress tensors is challenging. To address this limitation, we use a recently developed technique called EUCLID (Efficient Unsupervised Constitutive Law Identification and Discovery) for material characterisation of a two-dimensional multicomponent atherosclerotic plaque, based solely on displacement and force data. A finite element model was developed to simulate the mechanical behaviour of the plaque using the neo-Hookean hyperelastic model, and noisy data was introduced into the model by applying Gaussian noise on the displacements. An L-BFGS gradient descent optimiser was used to minimise the objective function, which is the residual error between predicted internal forces and true external forces. Results showed that at the expected noise level in clinical imaging modalities, no physically relevant stress distributions were obtained, where the plaque’s heterogeneity was observed to affect the accuracy. Clinical imaging was further emulated by systematically removing data to determine the effect of missing data on the model. No significant deterioration of the accuracy of obtained parameters was seen until using 10% of the total data, indicating good robustness to missing data. While the study has limitations, the proposed approach could have implications for the future diagnosis and treatment of atherosclerosis. Future research could explore alternative optimisation algorithms or techniques to improve the model’s accuracy under these conditions.Materials Science and Engineerin
Design of Ferritic-Bainitic High Strength Steels: A Physical Metallurgy Guided Machine Learning
No full textBainite steels are in high demand in many application areas owing to their outstanding mechanical properties, mainly due to the presence of a combination of fine bainite plates and retained austenite. Understanding the complicated mechanism of bainite transformation is crucial to creating an optimum design. Researchers have addressed this challenge via computational modelling, where the transformation start temperatures of bainite (Bs) and martensite (Ms) are key indexes when designing highperformance steels and their heat treatments.This thesis aims to explore the potential that data collection has to create a model, using experimental results of the bainite transformation process. A list of metallurgical accepted claims is elaborated to assess the quality of the data. Principal component analysis and clustering techniques are used to identify patterns, most important features and main relationships in the dataset. The results confirm the exponential carbon dependence that bainite and martensite transformation temperatures have, therefore requiring nonlinear models to predict them.Following, regression models and machine learning algorithms based solely on the chemical composition are used to predict Bs and Ms. Train-test split series and cross-validation are used to evaluate the prediction and consistency of each model. The results show that the ensemble learning algorithms outperform the regression techniques. Random forest and gradient boosting decision tree provide excellent Ms prediction on the validation set with R2 values of 0.92 and 0.93. The smaller dataset size adds up to the complexity of bainite transformation, resulting in worse prediction models of Bs, where the random forest and gradient boosting decision tree R2 values are 0.68 and 0.67 respectively. Even though the models are showing signs of learning, the impact that outliers have demonstrates that the data is not good by itself to create a predictive model. The incorporation of microstructural and process parameters would provide significant advances to the models for designing bainitic steels.Materials Science and Engineerin
Combining spectral and lateral information during the evaluation of hyperspectral data
No full textHyperspectral techniques have found application for the investigation of objects and samples in many scientific fields. Data evaluation approaches commonly either process them as stacks of randomly ordered spectra or as flat images. This thesis aims at combining both, lateral and spectral, aspects of hyperspectral data during data evaluation. Different approaches, such as lateral distance and neighbor pixel augmentation, will be explored with modern factorization techniques, such as t-stochastic neighbor embedding and self-organizing maps, and supported by artificial neural networks. The routines developed are applied on selected X-ray fluorescence data (XRF) sets from the field of materials science and cultural heritage. The altered augmented data set concept consists in augmenting the spectrum of a central pixel with the mean spectrum of its eight neighboring pixels. This new method is applied to the data set of a bi-modal Ti-6Al-6V-2Sn alloy. Clustering the optimized augmented data set with t-SNE reveals the existence of four components: phase with strong titanium signal, phase with strong vanadium signal, the first phase on top of the second phase and the second phase on top of the first phase. The later was not identified by clustering with t-SNE the data set augmented with all the spectra of its eight neighboring pixels. Clustering the XRF data set of a 13th century B.C. Egyptian mural painting with Self Organising Map (SOM) indicates areas with different thicknesses of copper and iron containing pigments. Clustering with Fast interpolation-based t-SNE (FIt-SNE) also reveals information about the painting sequence and the composition of a mixture of pigments. The pigments are not separable with a simple observation of the XRF elemental distribution images. Moreover, FIt-SNE dramatically accelerates t-SNE and can provide well-separated clusters. These properties are especially useful for large data sets, containing mixture of pigments. Finally, I introduce an optimised Artificial Neural Network (ANN) for training the photograph of a painting with the elemental distribution images, obtained by its XRF data set. The elemental distribution images are upscaled with Laplacian Pyramid Super-Resolution Network. In this way, we can identify with more accuracy the exact location of the damages and retouches on a 17th century A.C. easel painting (portrait of Hortense Manchini), since the result image is of high resolution.Materials Science and Engineerin
On the applicability of selective laser melting on pistons for the oil & gas industry: Fatigue limit and fracture toughness of selective laser melted TI6AL4V
No full textPistons for reciprocating compressors for industrial applications are often made of specialised materials. These prove to have problems with manufacture due to the high quality and short production times needed in combination with a low production volume per design. Additive manufacturing, specifically selective laser melting, could solve the production problems, provided that the material retains the needed mechanical properties. Ti6Al4V is the most appropriate material for this application. The most important mechanical properties for the application are the fatigue limit and the stress intensity factor, which are not well established properties for printed materials. For this reason fatigue limit and stress intensity factor tests were performed for both stress relieved and hot-isostatic pressed test pieces on longitudinal and transverse directions. Strength, toughness and fatigue limit is higher in hot-isostatic pressed test pieces of Ti6Al4V, and these are proven to be appropriate for application in compressor pistons. However the fatigue limit of stress relieved Ti6Al4V is lower, anisotropic, and has more scatter, and as such is insufficient for the application, which is due to deleteriously oriented microstructure and the presence of porosities. This can be solved by changing the printing parameters – laser power, cooling rate or heat treatment – although the exact combination of parameters for optimised values is not known and will be part-specific.Materials Science and Engineerin
Going Beyond Counting First Authors in Author Co-citation Analysis
Get PDFThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Data-driven inverse design of growth-based Voronoi meta-materials
No full textMetamaterials derive their properties from microstructure rather than from bulk material properties. This opens property spaces that are difficult, or impossible, to access with traditional methods. However, exploring this vast design space remains challenging because classical techniques can be computationally inefficient. Recent years has seen many successful applications of data-driven methods to this problem. Data-driven models can be used to bypass expensive Finite Element (FE) simulations and experiments by exploiting large datasets. This requires the parameterization of microstructures such that they can be mapped to properties. The forward problem, in which properties of are determined from design parameters, is typically well-posed, which allows straightforward application of machine learning methods. The inverse problem, in which design parameters are identified to match specific properties, is ill-posed because multiple sets of design parameters can produce similar properties. The Voronoi growth method induced by star-shape metrics provides a way to explore a large geometrical design space using simple parameterization. It is used to generate 2D unit cells of void and material pixels with non-trivial topologies. The growth process enforces connected geometry while also allowing for smooth transition between different designs. Using homogenization techniques, we generate a large dataset of design parameters and stiffness properties. Machine learning techniques are first used to model the forward problem. By combining the trained forward model with the inverse model, the inverse problem is rendered well-posed. Both the forward and (deterministic) inverse model show excellent agreement between target and predicted values. The method is generalized beyond the design space by targeting properties of non-growth based structures. We investigate methods to create a stochastic inverse model, that can produce multiple designs to match target properties. Verification is done by comparing tensile tests of 3D printed samples to FE simulations.Mechanical Engineerin
A Compact Ultra-Linear Compliant Torsion Reinforced Sarrus Mechanism
No full textIn this study, a promising design for a compact, ultra-linear Compliant Torsion Reinforced Sarrus mechanism (CORS), capable of achieving ultra-linear motion is presented. The CORS prototype, made entirely of aluminum, is produced monolithically using electric discharge machining (EDM). The design incorporates four torsion-reinforced folded leaf springs, effectively reducing parasitic motion and enhancing support stiffness. To meet the specified requirements, a design optimization process is undertaken, carefully considering constraints to attain an optimal CORS configuration. Integration of the CORS with a voice coil actuator for driving force and confocal chromatic sensors for detecting parasitic motion is carried out. Experimental results demonstrate and validate the performance of the CORS.Mechanical Engineering | Mechatronic System Design (MSD
On-Chip Photonic Recurrent Neural Networks for Time Series: A Dynamical Exploration and Application Search
No full textArtificial intelligence has a strong need for faster and more energy-efficient solutions, especially for computation performed at the sensor edge. On-chip photonic neural networks (PNNs) offer a promising solution for high speeds and energy efficiency. A less explored side of PNNs is their application to time-series data, which is often the case for real-world sensor applications. While PNNs promise high speeds and energy-efficient solutions, no good use cases have been proposed. This report will first review the state of the art of PNNs. It will be seen that to solve time-series tasks, photonic continuous-time recurrent neural networks (CTRNNs) are required. The dynamics of CTRNNs are thoroughly explored to leverage obtained insights to recommend novel practical applications. This is done through simple examples, and applied to a real machine learning task. It was seen that on a real classification task the network learned two distinct fixed points corresponding to the classes. A link between the time constant of the continuous-time neurons and the temporal dynamics of the task is also found. Two general directions for novel applications are then proposed. Firstly, photonic PNNs can be slowed down to match the task. Opto-electronic PNNs allow for more control of the time constant, and on-chip photonic filter neurons are suggested. Secondly, the high speeds of photonic neural networks can also be directly leveraged. Extremely fast convergence of photonic CTRNNs can be utilized for Modern Hopfield networks, pathfinding algorithms, and time-dependent optimization problems such as for example Model Predictive Control.Mechanical Engineerin
- …
