39 research outputs found
Direct 3D printing of multi-level voxel models
No full textWe present a direct method for the additive manufacturing of multi-level voxelized models that achieves a better surface finish. We have developed a new multi-level marching squares algorithm to identify the boundary of the multi-level voxelized model. We have also developed methods to use the multi-level voxelization to perform the infill operation based on user-defined infill density. We directly generate the GCode that is input into the 3D printer for printing. Our method overcomes the issues associated with the slicing operation for standard CAD models. In addition, we can directly print thresholded voxel models that are output from CT or MRI scans to get a physical 3D representation of medical data. We show that our method performs well by directly printing test models of multi-level voxel representation of complex CAD geometries, and cardiac CT data.This article is published as Ghadai, Sambit, Anushrut Jignasu, and Adarsh Krishnamurthy. "Direct 3D printing of multi-level voxel models." Additive Manufacturing 40 (2021): 101929. doi: https://doi.org/10.1016/j.addma.2021.101929. © 2021 The Authors. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)
Learning and Visualizing Localized Geometric Features Using 3D-CNN: An Application to Manufacturability Analysis of Drilled Holes
No full text3D Convolutional Neural Networks (3D-CNN) have been used for object recognition based on the voxelized shape of an object. However, interpreting the decision making process of these 3D-CNNs is still an infeasible task. In this paper, we present a unique 3D-CNN based Gradient-weighted Class Activation Mapping method (3D-GradCAM) for visual explanations of the distinct local geometric features of interest within an object. To enable efficient learning of 3D geometries, we augment the voxel data with surface normals of the object boundary. We then train a 3D-CNN with this augmented data and identify the local features critical for decision-making using 3D GradCAM. An application of this feature identification framework is to recognize difficult-to-manufacture drilled hole features in a complex CAD geometry. The framework can be extended to identify difficult-to-manufacture features at multiple spatial scales leading to a real-time design for manufacturability decision support system.This is a proceeding preprint from Ghadai, Sambit, Aditya Balu, Adarsh Krishnamurthy, and Soumik Sarkar. "Learning and visualizing localized geometric features using 3d-cnn: An application to manufacturability analysis of drilled holes." arXiv preprint arXiv:1711.04851 (2017). doi: https://doi.org/10.48550/arXiv.1711.04851. Copyright 2017 The Authors
A framework for 3D x-ray CT iterative reconstruction using GPU-accelerated ray casting
No full textX-ray Computed Tomography (CT) is a powerful nondestructive evaluation (NDE) tool to characterize internal defects and flaws, regardless of surface conditions and sample materials. After data acquisition from a series of X-ray 2D projection imaging, reconstruction methods play a key role to convert raw data (2D radiography) to 3D models. For the past 50 years, standard reconstruction have been performed using analytical methods based on filtered back-projection (FBP) concepts. Numerous iterative methods that have been developed have shown some improvements on certain aspects of the reconstruction quality, but have not been widely adopted due to their high computational requirements. With modern high performance computing (HPC) and graphics processing unit (GPU) technologies, the computing power barrier for iterative methods have been reduced. Iterative methods have more potential to incorporate physical models and a priori knowledge to correct artifacts generated from analytical methods. In this work, we propose a generalized framework for iterative reconstruction with GPU acceleration, which can be adapted for different physical and statistical models in the inner iteration during reconstruction. The forward projection algorithm is an important part of the framework, and is analogous to the ray casting depth map algorithm that was implemented in an earlier work [I] and accelerated using the GPU. Within this framework, different sub-models could be developed in future to deal with different artifacts, such as beam hardening effect and limited angle data problem.This proceeding may be downloaded for personal use only. Any other use requires prior permission of the author and AIP Publishing. This proceeding appeared in Zhang, Zhan, Sambit Ghadai, Onur Rauf Bingol, Adarsh Krishnamurthy, and Leonard J. Bond. "A framework for 3D x-ray CT iterative reconstruction using GPU-accelerated ray casting." AIP Conference Proceedings 2102, no. 1 (2019): 070002, and may be found at DOI: 10.1063/1.5099749. Posted with permission.</p
NURBS-based microstructure design for organic photovoltaics
No full textThe microstructure-spatial distribution of electron donor and acceptor domains-plays an important role in determining the photo current in thin film organic solar cells (OSCs). Optimizing the microstructure can lead to higher photo current generation, and is an active area of experimental research. There has been recent progress in framing OSC microstructure design as a computational design problem. However, most current approaches to microstructure optimization are based on volumetric distribution of material, which makes the design space very large. In contrast, we frame the microstructure design optimization problem in terms of designing the interface between the donor and acceptor regions, and thus pose it as a surface representation and optimization problem. This results in substantially reduced number of design variables, thus enabling use of standard optimization tools. In this work, we address the efficient design of OSC microstructure by using surface and curve modeling techniques to model the donor-acceptor interface, and use meta-heuristic, gradient-free optimization techniques to optimize the microstructure for maximum short circuit current generation. Our modeling framework consists of three major components: 1) geometric modeling of OSC microstructure that uses Non-Uniform Rational B-spline (NURBS) curves and surfaces to construct the free-form donor-acceptor interface, 2) photo-current generation modeling that uses a parallel, finite-element based exciton-drift-diffusion (XDD) model, and 3) optimization that utilizes genetic algorithms (GA) to optimize the OSCs microstructure via exploration of the NURBS representation. We apply these methods for the optimization of both 2D and 3D microstructures. Results show substantial improvement in current density compared to the bulk-heterojunction microstructures. These results provide promising microstructures for experimental groups to fabricate. The proposed surface representation approach seems to be a promising approach for interface design in engineered systems.This is a manuscript of an article published as Noruzi, Ramin, Sambit Ghadai, Onur Rauf Bingol, Adarsh Krishnamurthy, and Baskar Ganapathysubramanian. NURBS-based microstructure design for organic photovoltaics. Computer-Aided Design 118 (2019): 102771. DOI: 10.1016/j.cad.2019.102771. Posted with permission.</p
Learning localized features in 3D CAD models for manufacturability analysis of drilled holes
No full textWe present a novel feature identification framework to recognize difficult-to-manufacturedrilled holes in a complex CAD geometry using deep learning. Deep learning algorithms have been successfully used in object recognition, video analytics, image segmentation, etc. Specifically, 3D Convolutional Neural Networks (3D-CNNs) have been used for object recognition from 3D voxel data based on the external shape of an object. On the other hand, manufacturability of a component depends on local features more than the external shape. Learning these local features from a boundary representation (B-Rep) CAD model is challenging due to lack of volumetric information. In this paper, we learn local features from a voxelized representation of a CAD model and classify its manufacturability. Further, to enable effective learning of localized features, we augment the voxel data with surface normals of the object boundary. We train a 3D-CNN with this augmented data to identify local features and classify the manufacturability. However, this classification does not provide information about the source of non-manufacturability in a complex component. Therefore, we have developed a 3D-CNN based gradient-weighted class activation mapping (3D-GradCAM) method that can provide visual explanations of the local geometric features of interest within an object. Using 3D-GradCAM, our framework can identify difficult-to-manufacture features, which allows a designer to modify the component based on its manufacturability and thus improve the design process. We extend this framework to identify difficult-to-manufacture features in a realistic CAD model with multiple drilled holes, which can ultimately enable development of a real-time manufacturability decision support system.This is a manuscript of the article Ghadai, Sambit, Aditya Balu, Soumik Sarkar, and Adarsh Krishnamurthy. "Learning localized features in 3D CAD models for manufacturability analysis of drilled holes." Computer Aided Geometric Design (2018). DOI: 10.1016/j.cagd.2018.03.024. Posted with permission.</p
A cybermanufacturing framework incorporating deep learning and multi-resolution voxel representations
No full textCybermanufacturing (CM) is a modern concept involving predictive analytic operations and information technology to aid the manufacturing industry in better decision making for design and manufacturing processes. This thesis presents a data-driven intelligent Cybermanufacturing framework for the effortless design and manufacturing of a product.
While traditional manufacturing systems are iterative and especially require skilled operators in the process, CM systems alleviate this issue by making intelligent predictions without specialists' involvement. CM systems operate with a network and data-rich environment involving interaction within and between virtual and physical spaces resulting in an effective decision support system. The broad objective of this research is to define and establish a framework of a cyber-physical system consisting of such virtual and physical systems to confront various departments of a manufacturing process.
The first stage in most of the iterative manufacturing processes is a product design that is compliant with certain design specifications and requirements. However, this is not a one-stop solution; to realize the final design, a product goes through multiple iterations between design and manufacturing stages to be compliant with the existing manufacturing paradigm. To tackle this issue, we have developed data-driven decision support for an intelligent design for manufacturing (DFM) framework using a volumetric representation (voxels) of 3D CAD models and deep neural networks to make high-quality predictions of the manufacturability of a part or product without requiring domain expertise of the user. We have developed a manufacturing process planning framework that detects such features irrespective of its size by hierarchically representing 3D CAD models as volumes on multiple scale levels (multi-level voxels) and facilitating scale-variant feature learning through the implementation of a multi-level Deep Neural Network to make decisions from hierarchical data.
Along with virtual decision support systems for design and manufacturing, CM systems also involve actual manufacturing in the physical space using machines and robotic environments. We have developed an automated manufacturing module that includes an algorithm for direct 3D printing from voxels and optimization based robust reinforcement algorithm.</p
Multi-resolution 3D CNN for learning multi-scale spatial features in CAD models
No full textLearning multi-scale spatial features from 3D spatial geometric representations of objects such as point clouds, 3D CAD models, surfaces, and RGB-D data can potentially improve object recognition accuracy. Current deep learning approaches learn such features using structured data representations such as volume occupancy grids (voxels) and octrees or unstructured representations such as graphs and point clouds. Structured representations are generally restricted by their inherent limitations on the resolution, such as the voxel grid dimensions or the maximum octree depth. At the same time, it is challenging to learn directly from unstructured representations of 3D data due to non-uniformity among the samples. A hierarchical approach that maintains the structure at a larger scale while still accounting for the details at a smaller scale in specific spatial locations can provide an optimal solution for learning from 3D data. In this paper, we propose a multi-level learning approach to capture large-scale features at a coarse level (for example, using a coarse voxelization) while simultaneously capturing flexible sparse information
of the small-scale features at a fine level (for example, a local fine-level voxel grid) at different spatial locations. To demonstrate the utility of the proposed multi-resolution learning, we use a multi-level voxel representation of CAD models to perform object recognition. The multi-level voxel representation consists of a coarse voxel grid containing volumetric information of the 3D objects and multiple fine-level voxel grids corresponding to each voxel in the coarse grid containing a portion of the object boundary. In addition, we develop an interpretability-based feedback approach to transfer saliency information from one level of features to another in our hierarchical end-to-end learning framework. Finally, we demonstrate the performance of our multi-resolution learning algorithm for object recognition. We outperform several previously published benchmarks for object recognition while using significantly less memory during training.This article is published as Ghadai, Sambit, Xian Yeow Lee, Aditya Balu, Soumik Sarkar, and Adarsh Krishnamurthy. "Multi-resolution 3D CNN for learning multi-scale spatial features in CAD models." Computer Aided Geometric Design 91 (2021): 102038. doi: https://doi.org/10.1016/j.cagd.2021.102038
Design and Development of a Continuous Passive Motion Device for Physiotherapeutic Treatment of Human Knee
No full textSuccess of post-operative and post-traumatic therapy and rehabilitation of conditions related to major joints typically requires continuous passive movement of the affected joint. Physiotherapeutic devices are commonly used to promote rehabilitation of damaged or injured synovial joints. The research work aims to develop an assisted motion device for the physiotherapeutic treatment of the human knee joint. A cam-follower mechanism is proposed to reproduce actual gait cycle with varying flexion and extension of knee joint. The proposed model defines the passive motion device in terms of an improvised range of motion similar to the variation of the knee angle during normal walking gait cycle. Experiments are conducted on the proposed device to verify its angle variation and ease of use for the patients. The intuitive device finds its application in knee joint rehabilitation after knee replacement surgeries, fractures, injuries and other knee joint diseases to facilitate joint flexibility and promote well-being. The demonstrated works lays the groundwork for the prospective passive knee models and prosthesis desig
Multi-Level 3D CNN for Learning Multi-Scale Spatial Features
No full text3D object recognition accuracy can be improved by learning the multi-scale spatial features from 3D spatial geometric representations of objects such as point clouds, 3D models, surfaces, and RGB-D data. Current deep learning approaches learn such features either using structured data representations (voxel grids and octrees) or from unstructured representations (graphs and point clouds). Learning features from such structured representations is limited by the restriction on resolution and tree depth while unstructured representations creates a challenge due to non-uniformity among data samples. In this paper, we propose an end-to-end multi-level learning approach on a multi-level voxel grid to overcome these drawbacks. To demonstrate the utility of the proposed multi-level learning, we use a multi-level voxel representation of 3D objects to perform object recognition. The multi-level voxel representation consists of a coarse voxel grid that contains volumetric information of the 3D object. In addition, each voxel in the coarse grid that contains a portion of the object boundary is subdivided into multiple fine-level voxel grids. The performance of our multi-level learning algorithm for object recognition is comparable to dense voxel representations while using significantly lower memory.This is a preprint of a proceeding from S. Ghadai, X. Y. Lee, A. Balu, S. Sarkar and A. Krishnamurthy, "Multi-Level 3D CNN for Learning Multi-Scale Spatial Features," 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Long Beach, CA, USA, 2019, pp. 1152-1156, doi: 10.1109/CVPRW.2019.00150. Published version ©2019 IEEE. Preprint copyright 2019 The Authors
Recommended from our members
Growth Miracles and Growth Debacles: Exploring Root Causes
No full textIn this fascinating book, Sambit Bhattacharyya presents a detailed account of the socio-economic processes that create broad variations in living standards across the globe. The author examines the world's economic history over the last five centuries, replete with growth miracles and growth debacles: growth in Britain was steady, yet China lost her early advantage; North America settler colonies performed significantly better than those of Asia and Africa; Australia and Argentina were notably similar at the start of the twentieth century but delivered strikingly different growth outcomes. The book argues that these differences in growth rate are best explained by an interplay of factors, namely economic, political and geographical. In conclusion it presents long-run comparative growth narratives for Africa, China, India, the Americas, Russia and Western Europe. Presenting a unique and original analytical framework to explain economic growth and decline, and bridging empirical growth literature and economic history, this book will prove a stimulating read for both academic and professional economists, and scholars of economic history and economic growth. Other social scientists including sociologists, political scientists and economic historians will also find the book to be of great value
