Mechanical Engineering
Permanent URI for this communityhttp://hdl.handle.net/1903/2263
Browse
16 results
Search Results
Item Denoising the Design Space: Diffusion Models for Accelerated Airfoil Shape Optimization(2024) Diniz, Cashen; Fuge, Mark D; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Generative models offer the possibility to accelerate and potentially substitute parts of the often expensive traditional design optimization process. We present Aero-DDM, a novel application of a latent denoising diffusion model (DDM) capable of generating airfoil geometries conditioned on flow parameters and an area constraint. Additionally, we create a novel, diverse dataset of optimized airfoil designs that better reflects a realistic design space than has been done in previous work. Aero-DDM is applied to this dataset, and key metrics are assessed both statistically and with an open-source computational fluid dynamics (CFD) solver to determine the performance of the generated designs. We compare our approach to an optimal transport GAN, and demonstrate that our model can generate designs with superior performance statistically, in aerodynamic benchmarks, and in warm-start scenarios. We also extend our diffusion model approach, and demonstrate that the number of steps required for inference can be reduced by as much as ~86%, compared to an optimized version of the baseline inference process, without meaningful degradation in design quality, simply by using the initial design to start the denoising process.Item EFFECTS OF DIVERSE INITIALIZATION ON BAYESIAN OPTIMIZERS(2023) Kamrah, Eesh; Fuge, Mark D; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Design researchers have struggled to produce quantitative predictions for exactly why andwhen diversity might help or hinder design search efforts. This thesis addresses that problem by studying one ubiquitously used search strategy—Bayesian Optimization (BO)—on different ND test problems with modifiable convexity and difficulty. Specifically, we test how providing diverse versus non-diverse initial samples to BO affects its performance during search and introduce a fast ranked-DPP method for computing diverse sets, which we need to detect sets of highly diverse or non-diverse initial samples. We initially found, to our surprise, that diversity did not appear to affect BO, neither helping nor hurting the optimizer’s convergence. However, follow-on experiments illuminated a key trade-off. Non-diverse initial samples hastened posterior convergence for the underlying model hyper-parameters—a Model Building advantage. In contrast, diverse initial samples accelerated exploring the function itself—a Space Exploration advantage. Both advantages help BO, but in different ways, and the initial sample diversity directly modulates how BO trades those advantages. Indeed, we show that fixing the BO hyper-parameters removes the Model Building advantage, causing diverse initial samples to always outperform models trained with non-diverse samples. These findings shed light on why, at least for BO-type optimizers, the use of diversity has mixed effects and cautions against the ubiquitous use of space-filling initializations in BO. To the extent that humans use explore-exploit search strategies similar to BO, our results provide a testable conjecture for why and when diversity may affect human-subject or design team experiments. The thesis is organized as follows: Chapter 2 provides an overview of existing studies that explore the impact of different initial stimuli. In Chapter 3, we explain the methodology used in the subsequent experiments. Chapter 4 presents the results of our initial study on the diverse initialization of BO (Bayesian Optimization) applied to the wildcat wells function. In this chapter we also investigate the conditions under which less diverse initial examples perform better and expand on these findings in Chapter 5 by considering additional ND continuous functions. The final chapter discusses the limitations of our findings and proposes potential areas for future research.Item Ordering Non-Linear Subspaces for Airfoil Design and Optimization via Rotation Augmented Gradients(2023) Van Slooten, Alec; Fuge, Mark; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Airfoil optimization is critical to the design of turbine blades and aerial vehicle wings, among other aerodynamic applications. This design process is often constrained by the computational time required to perform CFD simulations on different design options, or the availability of adjoint solvers. A common method to mitigate some of this computational expense in nongradient optimization is to perform dimensionality reduction on the data and optimize the design within this smaller subspace. Although learning these low-dimensional airfoil manifolds often facilitates aerodynamic optimization, these subspaces are often still computationally expensive to explore. Moreover, the complex data organization of many current nonlinear models make it difficult to reduce dimensionality without model retraining. Inducing orderings of latent components restructures the data, reduces dimensionality reduction information loss, and shows promise in providing near-optimal representations in various dimensions while only requiring the model to be trained once. Exploring the response of airfoil manifolds to data and model selection and inducing latent component orderings have potential to expedite airfoil design and optimization processes. This thesis first investigates airfoil manifolds by testing the performance of linear and nonlinear dimensionality reduction models, examining if optimized geometries occupy lower dimensional manifolds than non-optimized geometries, and by testing if the learned representation can be improved by using target optimization conditions as data set features. We find that autoencoders, although often suffering from stability issues, have increased performance over linear methods such as PCA in low dimensional representations of airfoil databases. We also find that the use of optimized geometry and the addition of performance parameters have little effect on the intrinsic dimensionality of the data. This thesis then explores a recently proposed approach for inducing latent space orderings called Rotation Augmented Gradient (RAG) [1]. We extend their algorithm to nonlinear models to evaluate its efficacy at creating easily-navigable latent spaces with reduced training, increased stability, and improved design space preconditioning. Our extension of the RAG algorithm to nonlinear models has potential to expedite dimensional analyses in cases with near-zero gradients and long training times by eliminating the need to retrain the model for different dimensional subspacesItem Single- and Multi-Objective Feasibility Robust Optimization under Interval Uncertainty with Surrogate Modeling(2022) Kania, Randall Joseph; Azarm, Shapour; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This dissertation presents new methods for solving single- and multi-objective optimization problems when there are uncertain parameter values. The uncertainty in these problems is considered to come from sources with no known or assumed probability distribution, bounded only by an interval. The goal is to obtain a single solution (for single-objective optimization problems) or multiple solutions (for multi-objective optimization problems) that are optimal and “feasibly robust”. A feasibly robust solution is one that remains feasible for all values of uncertain parameters within the uncertainty interval. Obtaining such a solution can become computationally costly and require many function calls. To reduce the computational cost, the presented methods use surrogate modeling to approximate the functions in the optimization problem.This dissertation aims at addressing several key research questions. The first Research Question (RQ1) is: How can the computational cost for solving single-objective robust optimization problems be reduced with surrogate modelling when compared to previous work? RQ2 is: How can the computational cost of solving bi-objective robust optimization problems be improved by using surrogates in concert with a Bayesian optimization technique when compared to previous work? And RQ3 is: How can surrogate modeling be leveraged to make multi-objective robust optimization computationally less expensive when compared to previous work? In addressing RQ1, a new single-objective robust optimization method has been developed with improvements over an existing method from the literature. This method uses a deterministic, local solver, paired with a surrogate modelling technique for finding worst-case scenario of parameter configurations. Using this single-objective robust optimization method, improved large-scale performance and robust feasibility were demonstrated. The second method presented solves bi-objective robust optimization problems under interval uncertainty by introducing a relaxation technique to facilitate combining iterative robust optimization and Bayesian optimization techniques. This method showed improved feasibility robustness and performance at larger problem sizes over existing methods. The third method presented in this dissertation extends the current literature by considering multiple (beyond two) competing objectives for surrogate robust optimization. Increasing the number of objectives adds more dimensions and complexity to the search for solutions and can greatly increase the computational costs. In the third method, the robust optimization strategy from the bi-objective second method was combined with a new Monte Carlo approximated method. The key contributions in this dissertation are 1) a new single-objective robust optimization method combining a local optimization solver and surrogate modelling for robustness, 2) a bi-objective robust optimization method that employs iterative Bayesian optimization technique in tandem with iterative robust optimization techniques, and 3) a new acquisition function for robust optimization in problems of more than two objectives.Item Diversity and Novelty: Measurement, Learning and Optimization(2019) Ahmed, Faez; Fuge, Mark; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)The primary objective of this dissertation is to investigate research methods to answer the question: ``How (and why) does one measure, learn and optimize novelty and diversity of a set of items?" The computational models we develop to answer this question also provide foundational mathematical techniques to throw light on the following three questions: 1. How does one reliably measure the creativity of ideas? 2. How does one form teams to evaluate design ideas? 3. How does one filter good ideas out of hundreds of submissions? Solutions to these questions are key to enable the effective processing of a large collection of design ideas generated in a design contest. In the first part of the dissertation, we discuss key qualities needed in design metrics and propose new diversity and novelty metrics for judging design products. We show that the proposed metrics have higher accuracy and sensitivity compared to existing alternatives in literature. To measure the novelty of a design item, we propose learning from human subjective responses to derive low dimensional triplet embeddings. To measure diversity, we propose an entropy-based diversity metric, which is more accurate and sensitive than benchmarks. In the second part of the dissertation, we introduce the bipartite b-matching problem and argue the need for incorporating diversity in the objective function for matching problems. We propose new submodular and supermodular objective functions to measure diversity and develop multiple matching algorithms for diverse team formation in offline and online cases. Finally, in the third part, we demonstrate filtering and ranking of ideas using diversity metrics based on Determinantal Point Processes as well as submodular functions. In real-world crowd experiments, we demonstrate that such ranking enables increased efficiency in filtering high-quality ideas compared to traditionally used methods.Item COMBINED ROBUST OPTIMAL DESIGN, PATH AND MOTION PLANNING FOR UNMANNED AERIAL VEHICLE SYSTEMS SUBJECT TO UNCERTAINTY(2019) Rudnick-Cohen, Eliot; Azarm, Shapour; Herrmann, Jeffrey W; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Unmanned system performance depends heavily on both how the system is planned to be operated and the design of the unmanned system, both of which can be heavily impacted by uncertainty. This dissertation presents methods for simultaneously optimizing both of these aspects of an unmanned system when subject to uncertainty. This simultaneous optimization under uncertainty of unmanned system design and planning is demonstrated in the context of optimizing the design and flight path of an unmanned aerial vehicle (UAV) subject to an unknown set of wind conditions. This dissertation explores optimizing the path of the UAV down to the level of determining flight trajectories accounting for the UAVs dynamics (motion planning) while simultaneously optimizing design. Uncertainty is considered from the robust (no probability distribution known) standpoint, with the capability to account for a general set of uncertain parameters that affects the UAVs performance. New methods are investigated for solving motion planning problems for UAVs, which are applied to the problem of mitigating the risk posed by UAVs flying over inhabited areas. A new approach to solving robust optimization problems is developed, which uses a combination of random sampling and worst case analysis. The new robust optimization approach is shown to efficiently solve robust optimization problems, even when existing robust optimization methods would fail. A new approach for robust optimal motion planning that considers a “black-box” uncertainty model is developed based off the new robust optimization approach. The new robust motion planning approach is shown to perform better under uncertainty than methods which do not use a “black-box” uncertainty model. A new method is developed for solving design and path planning optimization problems for unmanned systems with discrete (graph-based) path representations, which is then extended to work on motion planning problems. This design and motion planning approach is used within the new robust optimization approach to solve a robust design and motion planning optimization problem for a UAV. Results are presented comparing these methods against a design study using a DOE, which show that the proposed methods can be less computationally expensive than existing methods for design and motion planning problems.Item USING ENGINEERING DESIGN PROJECT JOURNALS TO INFER INDIVIDUAL ACTIVITY AND ASSESS TEAM BEHAVIOR(2018) Born, Werner Christian; Schmidt, Linda C; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Design is critical to our society and yet we still do not understand a great deal about the underlying processes which occur as the engineering designer solves a problem. Analyzing such information could provide a wealth of insight on not only how design works but also how individuals work design. Concurrently maintained design journals kept by mechanical engineering students enrolled in capstone design courses offer a rich avenue to explore this phenomena. By applying a classification scheme to design journals, the entries of design activity of both individuals and teams were analyzed to determine relative time on each activity. These proportions were then examined across multiple project intervals defined by key deliverable dates in addition to across the project as a whole. Recurring sequences within these activities were also identified. Beyond the use of the classification scheme, other aspects of design journal analysis were explored. Specific concepts were tracked across a team, which were also used to calculate ratios of how often a member of a team references concepts by their teammates, themselves, and the final concept. With the addition of entry dates this information will also be used to map the exchange of ideas across a team and to chart the life of each concept. Journal content was codified and tested. Discoveries were made not only about the activity frequencies but how much those frequencies varied from person to person and team to team. The most common activity recorded in journals was Analysis, which averaged 23.13% of journal segments. Students were moderately accurate at identifying the frequency of Analysis and Reflection work. The most common sequences of unique activity were Idea Generation and Analysis along with its reverse. There were 3 distinct patterns of how individual behavior related to team members. 3 of 13 teams examined had uniform proportions of activity, while at the other extreme another 7 teams had little to no uniformity. The three activity classes found to be most affected by a project's time line were Problem Understanding, Idea Generation, and Decision Making, which each were found to have a statistically significant change in recorded activity over time.Item Electroosmotic Soft Actuators(2017) Sritharan, Deepa; Smela, Elisabeth; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This dissertation details the research involved in creating the first paper-based soft actuator driven by electroosmosis. To accomplish this, research breakthroughs were made in the fields of electrokinetic pumping and device manufacturing using soft materials. Electroosmosis is an electrically induced microfluidic flow phenomenon. When an electric field is applied to the fluid, across the microchannels, electroosmotic flow occurs in the direction of the applied electric field. In this work, liquid was electroosmotically displaced within a flexible microfluidic device to actuate an elastomeric membrane. The goal of this work was to create a fully sealed fluidic actuator. It was therefore necessary to encapsulate the pumping fluid within the device, and to maximize pressure it was necessary to eliminate compliance caused by trapped gases. Electrolytic gas formation is well known to disrupt pumping in DC electroosmotic systems that use water as the pumping liquid. In this work, electrolysis was eliminated by replacing water with propylene carbonate (PC): PC was determined to be electrochemically stable up to at least 10 kV, in the absence of moisture or salt contaminants. Bubble-free electroosmotic pumping with PC was achieved within sealed miniature actuators, which could be continuously operated for at least one hour. Benchtop fabrication techniques were developed to build encapsulated fluidic actuators composed entirely of soft, flexible materials. Stretchable electrochemically stable electrodes were made using a conductive paint made by mixing carbon nanoparticles into a silicone base. High-density microchannel networks were incorporated by using paper and other flexible porous materials, instead of conventional planar replica-molded microchannels. The device was filled with pumping fluid without the use of external tubing, and then encapsulated by casting a film of elastomer over the filled reservoir to form the actuating membrane. The resulting actuators were flexible and stretchable, demonstrating significant membrane deformations (hundreds of micrometers) within seconds of applying the electric field and ability to lift large loads (tens of grams). These polymeric electroosmotic actuators are unique among electroactive polymer actuators because they are able to simultaneously generate high force as well as large stroke. It is envisioned that this research will pave the way for the creation of artificial muscles and smart shape-changing materials that can be actuated by electroosmosis.Item Investigation into the Influence of Build Parameters on Failure of 3D Printed Parts(2016) Fornasini, Giacomo; Schmidt, Linda C; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Additive manufacturing, including fused deposition modeling (FDM), is transforming the built world and engineering education. Deep understanding of parts created through FDM technology has lagged behind its adoption in home, work, and academic environments. Properties of parts created from bulk materials through traditional manufacturing are understood well enough to accurately predict their behavior through analytical models. Unfortunately, Additive Manufacturing (AM) process parameters create anisotropy on a scale that fundamentally affects the part properties. Understanding AM process parameters (implemented by program algorithms called slicers) is necessary to predict part behavior. Investigating algorithms controlling print parameters (slicers) revealed stark differences between the generation of part layers. In this work, tensile testing experiments, including a full factorial design, determined that three key factors, width, thickness, infill density, and their interactions, significantly affect the tensile properties of 3D printed test samples.Item Mechanistic-based Design-Integrated Reliability Validation Framework for Mechanical Systems(2015) Yousif, Omer Hassan Awad; Modarres, Mohammad; Reliability Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)New product design challenges, related to customer needs, product usage and environments, face companies when they expand their product offerings to new markets; Some of the main challenges are: the lack of quantifiable information, product experience and field data. Designing reliable products under such challenges requires flexible reliability assessment processes that can capture the variables and parameters affecting the product overall reliability and allow different design scenarios to be assessed. These challenges also suggest a mechanistic (Physics of Failure-PoF) reliability approach would be a suitable framework to be used for reliability assessment. Mechanistic Reliability recognizes the primary factors affecting design reliability. This research views the designed entity as a “system of components required to deliver specific operations”; it addresses the above mentioned challenges by; Firstly: developing a design synthesis that allows a descriptive operations/ system components relationships to be realized; Secondly: developing component’s mathematical damage models that evaluate components Time to Failure (TTF) distributions given: 1) the descriptive design model, 2) customer usage knowledge and 3) design material properties; Lastly: developing a procedure that integrates components’ damage models to assess the mechanical system’s reliability over time. Analytical and numerical simulation models were developed to capture the relationships between operations and components, the mathematical damage models and the assessment of system’s reliability. The process was able to affect the design form during the conceptual design phase by providing stress goals to meet component’s reliability target. The process was able to numerically assess the reliability of a system based on component’s mechanistic TTF distributions, besides affecting the design of the component during the design embodiment phase. The process was used to assess the reliability of an internal combustion engine manifold during design phase; results were compared to reliability field data and found to produce conservative reliability results. The research focused on mechanical systems, affected by independent mechanical failure mechanisms that are influenced by the design process. Assembly and manufacturing stresses and defects’ influences are not a focus of this research.