Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Cost optimization”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 793 records · Page 44
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Scaling Up: Demonstrating Risk Reduction and Cost Compression for Commercial Heat Pump Water Heaters - CRADA 625 (Abstract)

Commercial heat pump water heater (CHPWH) systems significantly decarbonize the commercial and multifamily sectors by eliminating the reliance on gas-fired water heating. CHPWH systems are also well suited to include load shift controls that enable load-up and shed commands for supporting grid reliability and time-of-use pricing structure. However, they have not had wide adoption due to factors including price, complexity, and perceived risk. Although CHPWHs have been available in the US for decades, they have not made significant market gains in part because the systems have required significant and costly engineering design expertise and proved lackluster performance. Successful widespread market adoption requires a different approach; a shift from the current custom specialized expertise project design and installation to a repeatable approach that requires little specialized knowledge or expertise and can deliver persistent performance. Using this type of holistic systems approach requires effectively integrating four CHPWH system key components: primary air-to-water heat pumps; primary thermal storage tanks, a temperature maintenance system, and a control system which has capabilities to manage the primary heat pump cycles, any back-up, supplemental, or temperature maintenance heating, alarms, and grid connectivity allowing for demand response (DR), and/or load shifting. The project team has developed and will implement a suite of tools to support faster, less expensive, and more reliable field installations of CHPWH technology and with the resulting data used to further improve the tool set. These tools include: (1) A tool for optimizing system size and costs. (2) A tool that predicts annual energy use and overall system efficiency. (3) The Advanced Water Heater Specification (AWHS 8.0) defining the components of a full CHPWH system addressing performance requirements by climate zone. (4) The Qualified Products List: (QPL) of approved products that meet the specifications requirements. (5) Training materials including online on-demand modules, instructor-led training, and virtual interactive video tours of CHPWH installations in multifamily buildings. Demonstration site identification in low-income buildings in underserved communities is currently underway. Preliminarily, the team anticipates having three demonstrations in the Pacific Northwest and three in the Northeast for a total of six sites. After the demonstration sites are finalized, and M&V instrumentation installations are complete, the team will gather performance data and confirm whether the CHPWH systems perform as predicted and use the data to improve the existing tools.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2022

COMPOFF: A Compiler Cost model using Machine Learning to predict the Cost of OpenMP Offloading

The HPC industry is inexorably moving towards an era of extremely heterogeneous architectures, with more devices configured on any given HPC platform and potentially more kinds of devices, some of them highly specialized. Writing a separate code suitable for each target system for a given HPC application is not practical. The better solution is to use directive-based parallel programming models such as OpenMP. OpenMP provides a number of options for offloading a piece of code to devices like GPUs. To select the best option from such options during compilation, most modern compilers use analytical models to estimate the cost of executing the original code and the different offloading code variants. Building such an analytical model for compilers is a difficult task that necessitates a lot of effort on the part of a compiler engineer. Recently, machine learning techniques have been successfully applied to build cost models for a variety of compiler optimization problems. In this paper, we present COMPOFF, a cost model which uses the multi-layer perceptrons to statically estimates the Cost of OpenMP OFFloading. We used six different transformations on a parallel code of Wilson Dslash Operator to support GPU offloading, and we predicted their cost of execution on different GPUs using COMPOFF during compile time. Our results show that this model can predict offloading costs with a root mean squared error in prediction of less than 0.5 seconds. Our preliminary findings indicate that this work will make it much easier and faster for scientists and compiler developers to port legacy HPC applications that use OpenMP to new heterogeneous computing environment.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

A Stochastic Gradient Descent Approach for Stochastic Optimal Control

In this work, we introduce a stochastic gradient descent approach to solve the stochastic optimal control problem through stochastic maximum principle. The motivation that drives our method is the gradient of the cost functional in the stochastic optimal control problem is under expectation, and numerical calculation of such an expectation requires fully computation of a system of forward backward stochastic differential equations, which is computationally expensive. By evaluating the expectation with single-sample representation as suggested by the stochastic gradient descent type optimisation, we could save computational efforts in solving FBSDEs and only focus on the optimisation task which aims to determine the optimal control process.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2007

Optimisation of the Future Routine Orbit for Mars Express

Mars Express (MEX), the first planetary mission of the European Space Agency (ESA), reached Mars on December 25th 2003. Since then it is performing routine operations. Its operational phase had to cover one Martian year, with the possibility of an extension for a second Martian year (i.e. until November 2007). The end of the mission extension is approaching but, given the good health of the payload instruments and the high science return of the mission, there is a strong will to achieve further extensions. Mars Express is also seen as an important asset, capable to provide relay functions for future Martian missions. The ESA Science Program Committee has recently approved a second extension of the MEX mission until May 2009 and even further extensions are possible. Mars Express has an eccentric quasi-polar orbit with a period of approximately 6.72 hours and a pericentre height of about 300 km. Science observations are mainly performed at pericentre (but not only). In addition the orbit has a resonance of 11 revolutions per 3 Martian days. This means that ground tracks corresponding to orbits separated by 11 revolutions are adjacent, such that a given area can be covered by the on-board camera without leaving gaps. The J2 effect of Mars causes a drift of both ascending node and argument of pericentre. The drift of argument of pericentre makes it possible to observe periodically all Mars latitudes from close distance. Illumination conditions at pericentre are influenced by both the drift of the argument of pericentre and the drift of ascending node, as well as by the rotation of Mars around the Sun. The original MEX routine orbit was optimized for the duration of the nominal mission and extension, such that it produced a balanced share of day-side observations (for the optical instruments) and night-side observations (for the radar). The orbit was thus not optimized for the time beyond the assumed extension. Indeed, the evolution of the ascending node and argument of pericentre would cause in the following years a drift of the pericentre towards night-side observation conditions, hence uninteresting for the optical instruments. In order to prevent this an optimisation process for the future routine orbit has taken place. The share between day-side and night-side observations can be controlled by adjusting the drift of argument of pericentre and ascending node. This can in particular be achieved by changing the semimajor axis, eccentricity and/or inclination. A change of inclination is inefficient compared to a change in semimajor axis and eccentricity, and has therefore been discarded. An in-plane maneuvre can be performed to change both semi-major axis and eccentricity, and thus the period of the orbit. Although an apocentre manoeuvre is cheaper in terms of deltaV, it would result in raising the pericentre height, which is unfavourable for close observations. Hence a pericentre manoeuvre is proposed, which will increase the apocentre height. A repeat cycle is still required to allow mapping areas with adjacent ground tracks, so the change of semimajor axis must result in a new resonance. Resonances 18:5, 25:7 and 7:2 have been considered as potential candidates. The resulting long term evolution of the observation conditions has been analysed. Finally it has been decided to perform a change of orbit to reach the 18:5. Another aspect of the optimisation process is the control of the ground track. The previous MEX reference trajectory included regular maneuvres at every apocentre in order to adjust the orbital period, such that the separation of the ground tracks would be optimal, regardless of the latitude of pericentre. The implementation of the actual delatVs on-board was done partly by optimizing the attitude of reaction-wheel desaturation activities. Despite of it, this strategy has a significant propellant cost, because it prevents to optimize reaction wheel de-saturation activities to minimize propellant consumption. Therore, with the aim at preserving propellant resources for a long time extension it has been agreed to stop the ground-track control. This requires now a more accurate science operation planning, with improved attitude pointing control. Finally, the approach to phase Mars Express to provide back-up relay functions for NASA Phoenix landing is explained. In the context of the routine trajectory optimisation a new requirement for close fly-bys at Phobos, with different observation geometries, has been specified. The approach to fulfill this requirement is explained.

Carranza, Manuel↗
DOE OSTI2020

Optimal electric-distribution-grid planning considering the demand-side flexibility of thermal building systems for a test case in Singapore

The planning of district-scale electric grids, i.e., distribution grids, has traditionally relied on finding the most cost-effective design such that they are able to supply the peak loads in a district. With the advent of electric demand side flexibility (DSF), there is the opportunity to reshape peak loads such that the investment cost of the electric grid decreases in exchange for a minor increase in the operation cost. This paper formulates an optimal planning approach for the electric grid at the district scale, which incorporates the DSF from thermal building systems, e.g., heating ventilation and air-conditioning (HVAC) systems. The problem is formulated as a mixed-integer linear program (MILP) and aims at minimizing the investment cost for the grid along with the operation cost of the flexible loads. This is subjected to the fixed electricity demand and thermal comfort constraints of building occupants. To this end, linear models for the thermal comfort in the buildings and the power flow in electric grid are considered. The approach is tested on a district planning test case based in Singapore, where the results show up to 30.9 % reductions in investment cost and up to 3.7 % reduction in total annualized cost. Urban planning authorities, developers and utility companies can all benefit from the presented approach to make optimized investment decisions. For building operators, the results point to the need of adopting their control systems for DSF.

Troitzsch, Sebastian↗
DOE OSTI2024

Department of Energy/IDEA: More best-practices research

This article continues the overview of the U.S. Department of Energy and IDEA collaboration by highlighting additional case studies that demonstrate innovative engineering approaches in district energy systems. It focuses on how campuses are improving efficiency, integrating advanced technologies, and reducing energy use and emissions through data-driven strategies. The University of Cincinnati is featured for its highly efficient chilled water system, using advanced algorithms and upgraded “smart chillers” to optimize performance, reduce energy consumption, and lower costs. Wake Forest University emphasizes a data-informed modernization process, using system optimization and targeted building improvements to significantly boost efficiency and reduce overall energy use. Arizona Western College showcases creative solutions to extreme cooling demands, including system redesign, smart controls, and thermal storage to improve reliability and cut peak energy costs. Overall, the article demonstrates that combining innovative engineering, real-time data analysis, and system optimization can significantly enhance performance, reduce emissions, and deliver cost savings in district energy systems.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
NASA NTRS1982

Conceptual design of the 6 MW Mod-5A wind turbine generator

The General Electric Company, Advanced Energy Programs Department, is designing under DOE/NASA sponsorship the MOD-5A wind turbine system which must generate electricity for 3.75 cent/KWH (1980) or less. During the Conceptual Design Phase, completed in March, 1981, the MOD-5A WTG system size and features were established as a result of tradeoff and optimization studies driven by minimizing the system cost of energy (COE). This led to a 400' rotor diameter size. The MOD-5A system which resulted is defined in this paper along with the operational and environmental factors that drive various portions of the design. Development of weight and cost estimating relationships (WCER's) and their use in optimizing the MOD-5A are discussed. The results of major tradeoff studies are also presented. Subsystem COE contributions for the 100th unit are shown along with the method of computation. Detailed descriptions of the major subsystems are given, in order that the results of the various trade and optimization studies can be more readily visualized.

Barton, R. S.↗
NASA NTRS2013

Software Tools to Support the Assessment of System Health

This presentation provides an overview of three software tools that were developed by the NASA Glenn Research Center to support the assessment of system health: the Propulsion Diagnostic Method Evaluation Strategy (ProDIMES), the Systematic Sensor Selection Strategy (S4), and the Extended Testability Analysis (ETA) tool. Originally developed to support specific NASA projects in aeronautics and space, these software tools are currently available to U.S. citizens through the NASA Glenn Software Catalog. The ProDiMES software tool was developed to support a uniform comparison of propulsion gas path diagnostic methods. Methods published in the open literature are typically applied to dissimilar platforms with different levels of complexity. They often address different diagnostic problems and use inconsistent metrics for evaluating performance. As a result, it is difficult to perform a one ]to ]one comparison of the various diagnostic methods. ProDIMES solves this problem by serving as a theme problem to aid in propulsion gas path diagnostic technology development and evaluation. The overall goal is to provide a tool that will serve as an industry standard, and will truly facilitate the development and evaluation of significant Engine Health Management (EHM) capabilities. ProDiMES has been developed under a collaborative project of The Technical Cooperation Program (TTCP) based on feedback provided by individuals within the aircraft engine health management community. The S4 software tool provides a framework that supports the optimal selection of sensors for health management assessments. S4 is structured to accommodate user ]defined applications, diagnostic systems, search techniques, and system requirements/constraints. One or more sensor suites that maximize this performance while meeting other user ]defined system requirements that are presumed to exist. S4 provides a systematic approach for evaluating combinations of sensors to determine the set or sets of sensors that optimally meet the performance goals and the constraints. It identifies optimal sensor suite solutions by utilizing a merit (i.e., cost) function with one of several available optimization approaches. As part of its analysis, S4 can expose fault conditions that are difficult to diagnose due to an incomplete diagnostic philosophy and/or a lack of sensors. S4 was originally developed and applied to liquid rocket engines. It was subsequently used to study the optimized selection of sensors for a simulation ]based aircraft engine diagnostic system. The ETA Tool is a software ]based analysis tool that augments the testability analysis and reporting capabilities of a commercial ]off ]the ]shelf (COTS) package. An initial diagnostic assessment is performed by the COTS software using a user ]developed, qualitative, directed ]graph model of the system being analyzed. The ETA Tool accesses system design information captured within the model and the associated testability analysis output to create a series of six reports for various system engineering needs. These reports are highlighted in the presentation. The ETA Tool was developed by NASA to support the verification of fault management requirements early in the Launch Vehicle process. Due to their early development during the design process, the TEAMS ]based diagnostic model and the ETA Tool were able to positively influence the system design by highlighting gaps in failure detection, fault isolation, and failure recovery.

Melcher, Kevin J.↗
DOE OSTI2020

Optimal Sizing of Resilience Solutions for the U.S. Army Reserve

Power or water outages in buildings threaten the ability of the military to support surrounding communities during natural disasters. Outages can last for days, weeks or months. Typical solutions include very expensive batteries and onsite generation that are sized based on historical power needs. The U.S. Army Reserve (USAR) has teamed up with the Pacific Northwest National Laboratory (PNNL) to develop a simulation framework that optimizes future power needs to reduce the cost of resilience solutions. The approach starts with the building. Power needs during an emergency event are simulated at the building end-use level, then loads are reduced through the selection of life cycle cost-effective building-level technology improvements. These new optimized loads are then fed into a microgrid sizing tool that dynamically constructs many different combinations of solar photovoltaic, battery storage, and generator resources to meet the load for hundreds of statistically generated outage scenarios. Six site assessments completed by USAR and PNNL in 2019 have resulted in a 1-14% reduction in overall investment when optimizing the buildings first before determining the generation requirements. This approach helps the Army secure critical missions and provide a 14 day-minimum supply of necessary energy and water in the most efficient manner.

buidlings, efficiency, resilience, FEDS, MCOR↗
DOE OSTI2025

Proactive Assignment Strategy With Human Choice Models for Boosting Pooled Rideshare Service

This study analyzes various human factors considerations in estimating discounts for pooled rideshare trips. The discounts are utilized in an optimization-based rideshare assignment strategy (proactive strategy) and compared against each other, as well as a heuristic strategy attempting to replicate current real-world pooling rates. Simulations within Austin, Texas and Greenville, South Carolina, reveal the proactive strategy’s ability to increase average vehicle occupancy by 0.23 persons/mile in Austin and 0.52 persons/mile in Greenville. A significant ability to decrease trip rejections and increase profitability is also observed. Finally, the strengths of particular combinations of factors are discussed relative to their effectiveness in each region.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2022

Calibration of cloud and aerosol related parameters for solar irradiance forecasts in WRF-solar

Model parameters are a major source of uncertainty in numerical weather prediction. Recently, the Weather Research and Forecasting model with Solar extensions (WRF-Solar) has been upgraded by enhancing the treatment of sub-grid scale cloud and aerosols with augmentations of a sub-grid scale cloud scheme (CLD3) and an upgraded aerosol-aware Thompson-Eidhammer scheme (TE14). However, the value of model parameters associated with these parameterizations are assigned based on limited measurements or theoretical calculations. Calibrating the most sensitive parameters has the potential to improve solar irradiance predictions. Here, we adopted a multiobjective surrogate-based optimization (SBO) framework to calibrate nine parameters used in CLD3 and TE14 that lead to the largest sensitivity in simulated irradiance. The normalized mean-absolute-error (NMAE) of global horizontal irradiance (GHI) and direct normal irradiance (DNI) are minimized by calibrating WRF-Solar over two regions including the Southern Great Plains (SGP) and Central California, in order to focus on parameter calibration under cloudy conditions with different aerosol loading. The results show that generalized linear model (GLM)-based surrogate models approximate physical models well, particularly when the third order and three-way interaction terms are considered. The SBO framework efficiently searches the parameter space for optimal solutions with less computational costs than directly calibrating the physical model. We first calibrate CLD3 parameters over the less-polluted SGP region. Optimized CLD3 parameters alone result in NMAE reduction by 14% for the site-mean and up to 33% for individual cases over the SGP region. With further calibration of TE14 parameters over the Central California during active fire periods, the optimized parameters lead to over 20% reductions of NMAE. Our investigation reveals, however, that optimizing TE14 has a limited impact on irradiance simulations under less-polluted conditions in the SGP.

14 SOLAR ENERGY↗
DOE OSTI2021

Two Dimensional Topology Optimization of Heat Exchangers with the Density and Level-Set Methods

We design heat exchangers using two topology optimization approaches: the density, i.e. volume fraction and level set methods. Our goal is to maximize the heat exchange between two fluids in separate channels while constraining the pressure drop across each channel. The heat exchanger is modeled with a coupled thermal-flow formulation. The flow is governed by an isothermal and incompressible Stokes-Brinkman equation and the heat transfer is governed by a convection-diffusion equation with high Peclet number. We solve one set of Stokes-Brinkman equations per fluid. Each Brinkman term in the flow equation serves to model the other phase as a solid, thereby preventing mixing. We first represent the solid and fluid phases using a volume fraction variable and apply a SIMP-like penalization in the Brinkman term to drive the optimization to a discrete design. The cost and constraint function derivatives are automatically calculated with the library pyadjoint and the optimization is performed by the Method of Moving Asymptotes. In a second optimization formulation, we use the level set approach to define the interface that separates the two fluids. Pyadjoint calculates the shape derivatives of the cost and constraint functions and the Hamilton-Jacobi advects the interface, allowing for topological changes. We present results in two dimensions and discuss the advantages and disadvantages of each approach.

42 ENGINEERING↗
NASA NTRS2004

Minimizing Launch Mass for ISRU Processes

The University of Dayton and the Jet Propulsion Laboratory are developing a methodology for estimating the Earth launch mass (ELM) of processes for In-Situ Resource Utilization (ISRU) with a focus on lunar resource recovery. ISRU may be enabling for both an extended presence on the Moon, and for large sample return missions and for a human presence on Mars. To accomplish these exploration goals, the resources recovered by ISRU must offset the ELM for the recovery process. An appropriate figure of merit is the cost of the exploration mission, which is closely related to ELM. For a given production rate and resource concentration, the lowest ELM - and the best ISRU process - is achieved by minimizing capital equipment for both the ISRU process and energy production. ISRU processes incur Carnot limitations and second law losses (irreversibilities) that ultimately determine production rate, material utilization and energy efficiencies. Heat transfer, chemical reaction, and mechanical operations affect the ELM in ways that are best understood by examining the process's detailed energetics. Schemes for chemical and thermal processing that do not incorporate an understanding of second law losses will be incompletely understood. Our team is developing a methodology that will aid design and selection of ISRU processes by identifying the impact of thermodynamic losses on ELM. The methodology includes mechanical, thermal and chemical operations, and, when completed, will provide a procedure and rationale for optimizing their design and minimizing their cost. The technique for optimizing ISRU with respect to ELM draws from work of England and Funk that relates the cost of endothermic processes to their second law efficiencies. Our team joins their approach for recovering resources by chemical processing with analysis of thermal and mechanical operations in space. Commercial firms provide cost inputs for ELM and planetary landing. Additional information is included in the original extended abstract.

England, C.↗
NASA NTRS1977

Procedure for minimizing the cost per watt of photovoltaic systems

A general analytic procedure is developed that provides a quantitative method for optimizing any element or process in the fabrication of a photovoltaic energy conversion system by minimizing its impact on the cost per watt of the complete system. By determining the effective value of any power loss associated with each element of the system, this procedure furnishes the design specifications that optimize the cost-performance tradeoffs for each element. A general equation is derived that optimizes the properties of any part of the system in terms of appropriate cost and performance functions, although the power-handling components are found to have a different character from the cell and array steps. Another principal result is that a fractional performance loss occurring at any cell- or array-fabrication step produces that same fractional increase in the cost per watt of the complete array. It also follows that no element or process step can be optimized correctly by considering only its own cost and performance

Redfield, D.↗
DOE OSTI2021

U.S. Domestic Microreactor Security-by-Design

U.S. nuclear power facilities face increasing challenges in meeting dynamic security requirements caused by evolving and expanding threats while keeping cost reasonable to make nuclear energy competitive. The past approach has often included implementing security features after a facility has been designed and without attention to optimization, which can lead to cost overruns. Incorporating security in the design process can provide robust, cost effective, and sufficient physical protection systems. The purpose of this work is both to develop a framework for the integration of security into the design phase of a microreactor and increase the use of modeling and simulation tools to optimize the design of physical protection systems. Specifically, this effort focuses on integrating security into the design phase of a model microreactor that meets current Nuclear Regulatory Commission (NRC) physical protection requirements and providing advanced solutions to improve physical protection and decrease costs. A suite of tools, including SCRIBE3D©, PATHTRACE© and Blender© were used to model a hypothetical, generic domestic microreactor facility. Physical protection elements such as sensors, cameras, barriers, and guard forces were added to the model based on best practices for physical protection systems. Multiple outsider sabotage scenarios were examined with four-to-eight adversaries to determine security metrics. The results of this work will influence physical protection system designs and facility designs for U.S. domestic microreactors. This work will also demonstrate how a series of experimental and modeling capabilities across the Department of Energy (DOE) Complex can impact the design of and complete Safeguards and Security by Design (SSBD) for microreactors. The conclusions and recommendations in this document may be applicable to all microreactor designs.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
DOE OSTI2020

Lignin-derived electrochemical energy materials and systems

Electrochemical energy storage systems such as supercapacitors, rechargeable batteries and fuel cells have been proven the most effective technologies for energy conversion, storage, and management at different scales. Although a large number of electrochemical energy technologies have been developed in the past and they will continue to be optimized in terms of cost, lifetime, and performance, there is a substantial growing demand for advanced electrochemical energy systems. To deploy these advanced systems, the electrode and electrolyte materials with higher performance, longer life, and lower cost, must be developed. Lignin is the second most abundant natural polymer after cellulose, a byproduct from emerging cellulosic biorefineries, and a waste product from pulp and paper industries. Numerous researches have successfully demonstrated that lignin from different sources can be used as precursors or feedstocks for preparing high-performance electrochemical energy materials and components such as electrodes, electrolytes, membrane separators, and additives. Moreover, techno-economic analyses indicate that it is possible to prepare cost-effective carbons from lignin at engineering scales, compared to current carbon products. These facts suggest that scalable conversion of lignin into high-value energy materials will offer a promising pathway to not only promote the utilization and valorization of lignin but also boost the development of the advanced electrochemical energy systems. This review presents state of the arts of renewable energy materials derived from various lignin and their applications in electrochemical energy systems with emphasis on supercapacitors, rechargeable batteries, and fuel cells. Meanwhile, this article also aims to carve out the critical barriers for lignin-derived high-performance materials for energy applications, intending to identify viable approaches for synthesis of sustainable new energy materials.

09 BIOMASS FUELS↗
DOE OSTI2020

Parallel performance of algebraic multigrid domain decomposition

Algebraic multigrid (AMG) is a widely used scalable solver and preconditioner for large-scale linear systems resulting from the discretization of a wide class of elliptic PDEs. While AMG has optimal computational complexity, the cost of communication has become a significant bottleneck that limits its scalability as processor counts continue to grow on modern machines. This article examines the design, implementation, and parallel performance of a novel algorithm, algebraic multigrid domain decomposition (AMG-DD), designed specifically to limit communication. The goal of AMG-DD is to provide a low-communication alternative to standard AMG V-cycles by trading some additional computational overhead for a significant reduction in communication cost. Numerical results show that AMG-DD achieves superior accuracy per communication cost compared with AMG, and speedup over AMG is demonstrated on a large GPU cluster.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Differential Property Prediction: A Machine Learning Approach to Experimental Design in Advanced Manufacturing

Advanced manufacturing techniques have enabled the production of materials with state-of-the-art properties. In many cases however, the development of physics-based models of these techniques lags behind their development in the lab. This means that material and process development proceeds largely via trial and error. This is sub-optimal since experiments are cost-, time-, and labor-intensive. In this work we propose a machine learning framework, differential property classification (DPC), which enables an experimenter to leverage machine learning's unparalleled pattern matching capability to pursue data-driven experimental design. DPC takes two possible experiment parameter sets and outputs a prediction of which will produce a material with a more desirable property specified by the operator. We demonstrate the success of DPC on AA7075 tube manufacturing process and mechanical property data using shear assisted processing and extrusion (ShAPE), an emerging solid phase processing technology. We show that by focusing on the experimenter's need to choose between multiple candidate experimental parameters, we can reframe the challenging regression task of predicting material properties from processing parameters, into a classification task on which machine learning models can achieve good performance.

advanced manufacturing, machine learning, ShAPE↗