Chance-constrained Optimization: Approximations Algorithms and Applications.
Abstract not provided.
SEARCH · Search NASA
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.
Abstract not provided.
Abstract not provided.
Explore the source record for details and available documents.
With heterogeneous systems, the number of GPUs per chip increases to provide computational capabilities for solving science at a nanoscopic scale. However, low utilization for single GPUs defies the need to invest more money in expensive accelerators. Although related work develops optimizations to improve application performance, none studies how these optimizations impact hardware resource usage or average GPU utilization. Here, this paper takes a data-driven analysis approach in addressing this gap by (1) characterizing how hardware resource usage affects device utilization, execution time, or both, (2) presenting a multiobjective metric to identify important application-device interactions that can be optimized to improve device utilization and application performance jointly, (3) studying hardware resource usage behaviors of several optimizations for a benchmark application, and finally (4) identifying optimization opportunities for several scientific proxy applications based on their hardware resource usage behaviors. Furthermore, we demonstrate the applicability of our methodology by applying the identified optimizations to a proxy application, which improves the execution time, device utilization, and power consumption by up to 29.6%, 5.3% and 26.5% respectively.
Abstract The design of materials and identification of optimal processing parameters constitute a complex and challenging task, necessitating efficient utilization of available data. Bayesian Optimization (BO) has gained popularity in materials design due to its ability to work with minimal data. However, many BO-based frameworks predominantly rely on statistical information, in the form of input-output data, and assume black-box objective functions. In practice, designers often possess knowledge of the underlying physical laws governing a material system, rendering the objective function not entirely black-box, as some information is partially observable. In this study, we propose a physics-informed BO approach that integrates physics-infused kernels to effectively leverage both statistical and physical information in the decision-making process. We demonstrate that this method significantly improves decision-making efficiency and enables more data-efficient BO. The applicability of this approach is showcased through the design of NiTi shape memory alloys, where the optimal processing parameters are identified to maximize the transformation temperature.
As renewable power generation deployment increases, fossil fuel plants are increasingly required to operate more flexibly. Many coal-fired power plants were originally designed to operate at base load and do not operate optimally at partial load. Predictive first-principles plant-wide models can be employed to identify opportunities for flexibility improvements and diagnose low-load operating issues. This paper describes the application of the Institute for the Design of Advanced Energy Systems Integrated Platform (IDAES) to model and optimize flexible power plant operations. The key benefits of using IDAES are that it provides an open-source, fully equation-oriented modeling framework for efficient modular model construction, reuse, and customization, together with a mathematical optimization framework leveraging powerful, state-of-the-art solvers. The process systems engineering workflow from predictive process simulation to parameter estimation, model validation, and plant optimization is applicable to a variety of existing and next-generation energy systems as well as other chemical and environmental processes. Here, to demonstrate this capability, a physics-based, steady-state model was developed to improve full- and part-load performance of the Escalante Generating Station, a 245 MWe (net) subcritical pulverized coal-fired power plant owned and operated by Tri-State Generation and Transmission Association. Specifically, sixty-nine model parameters were simultaneously estimated from several months of operating data enabling prediction of flow rates, temperatures, pressures, and steam quality throughout the plant. The validated model was leveraged by Escalante to reduce the minimum operating load from 90 MW to 50 MW by diagnosing a low-load water-hammer issue, enabling coal usage and emissions reductions during periods of low power demand. Additionally, opportunities for heat rate reduction (i.e., efficiency improvement) through a steeper sliding-pressure approach to load-following and optimization of other boiler operating variables were also identified and quantified. For example, a potential efficiency improvement of 0.7 percentage points was observed at half-load operation.
Many new reactor designs, such as advanced reactors and micro reactors, have materials that lack nuclear data validation. This is also true for many other applications in criticality safety and global security. Both differential and integral experiments are needed to validate cross-section data. Without this, a user cannot have confidence in the predicted results of a radiation-transport code. This work describes an approach called ARCHIMEDES (Application Relevant Critical/Subcritical HEU/Pu-based Integral Measurements for Enhancing Data and Evaluating Sensitivities) to design new criticality experiments that have similar k eff cross-section sensitivities to an application of interest. This process involves simulations to generate cross-section sensitivities to a parameter of interest (such as k eff ), a gap analysis to determine which existing benchmarks are most similar to the application, and an experiment optimization. Recently, there has been a great deal of interest in the reactor physics community on advanced reactors, micro reactors, and accelerator driven systems (ADS). This work will apply the described method to specific examples in this area. The focus of this work will be on the sensitivity study and gap analysis, while future work will include experiment design.
Highlights: • A two-step (strategic and detailed optimal planning) methodology is used for solving complex MSW management problems. • A software package is outlined, which can be used for generating detailed optimal plans. • Sensitivity analysis compares alternative scenarios that address objections and/or wishes of local communities. • A case study shows the application of the above procedure in practice and demonstrates the results and benefits obtained. - Abstract: The paper describes a software system capable of formulating alternative optimal Municipal Solid Wastes (MSWs) management plans, each of which meets a set of constraints that may reflect selected objections and/or wishes of local communities. The objective function to be minimized in each plan is the sum of the annualized capital investment and annual operating cost of all transportation, treatment and final disposal operations involved, taking into consideration the possible income from the sale of products and any other financial incentives or disincentives that may exist. For each plan formulated, the system generates several reports that define the plan, analyze its cost elements and yield an indicative profile of selected types of installations, as well as data files that facilitate the geographic representation of the optimal solution in maps through the use of GIS. A number of these reports compare the technical and economic data from all scenarios considered at the study area, municipality and installation level constituting in effect sensitivity analysis. The generation of alternative plans offers local authorities the opportunity of choice and the results of the sensitivity analysis allow them to choose wisely and with consensus. The paper presents also an application of this software system in the capital Region of Attica in Greece, for the purpose of developing an optimal waste transportation system in line with its approved waste management plan. The formulated plan was able to: (a) serve 113 Municipalities and Communities that generate nearly 2 million t/y of comingled MSW with distinctly different waste collection patterns, (b) take into consideration several existing waste transfer stations (WTS) and optimize their use within the overall plan, (c) select the most appropriate sites among the potentially suitable (new and in use) ones, (d) generate the optimal profile of each WTS proposed, and (e) perform sensitivity analysis so as to define the impact of selected sets of constraints (limitations in the availability of sites and in the capacity of their installations) on the design and cost of the ensuing optimal waste transfer system. The results show that optimal planning offers significant economic savings to municipalities, while reducing at the same time the present levels of traffic, fuel consumptions and air emissions in the congested Athens basin.
Reactor designs with mixed catalysts play an important role in transforming a multiple reactor system to single-shot reactors. In addition to savings in capital and ease of implementation, single-shot reactors are useful to break equilibrium limitations, thereby increasing the yield and selectivity of desired product as shown in previous studies. However, the nonlinear and highly exothermic nature of mixed-catalyst systems makes it difficult for commercial process simulation and optimization tools to optimize these systems. This study describes the development and application of optimization strategies for mixed-catalyst, single-shot reactors for syngas to olefin (STO) processes. Finding the optimal catalyst distribution is challenging and requires advanced solution strategies for singular optimal control problems, which are poorly conditioned and often lead to flat response surfaces. The graded bed and partial-moving finite-element approaches are used to find the optimal catalyst distribution that maximizes the olefins yield. A 1.3% increase in the yield is observed from one zone to three zones. The yield further improves from three zones to the exact infinite dimensional solution by 0.2%. This improvement can be realized in practice by changing only the catalyst distribution, without any extra investment. Lastly, the results suggest that a mixed-catalyst single shot reactor bed can be applied to other reaction mechanisms to increase reactor performance.
This paper introduces applications of the GenOpt optimizer coupled with a vapor compression system model for auto-calibration and control strategy determination towards the development of a variable-speed ground-source heat pump water heating unit. The GenOpt optimizer can be linked with any simulation program using input and output text files. It effectively facilitates optimization runs. Using our GenOpt wrapper program, we can flexibly define objectives for optimizations, targets, and constraints. Those functionalities enable running extensive optimization cases for model calibration, configuration design and control strategy determination. In addition, we describe a methodology to improve prediction accuracy using functional calibration curves. Using the calibrated model, we investigated control strategies of the ground-source heat pump water heater, considering multiple control objectives, covering the entire operation range.
In general, the modeling errors of dynamic system model are a set of random variables. The traditional performance index of modeling such as means square error (MSE) and root means square error (RMSE) can not fully express the connotation of modeling errors with stochastic characteristics both in the dimension of time domain and space domain. Therefore, the probability density function (PDF) is introduced to completely describe the modeling errors in both time scales and space scales. Based on it, a novel wavelet neural network (WNN) modeling method is proposed by minimizing the two-dimensional (2D) PDF shaping of modeling errors. First, the modeling error PDF by the tradional WNN is estimated using data-driven kernel density estimation (KDE) technique. Then, the quadratic sum of 2D deviation between the modeling error PDF and the target PDF is utilized as performance index to optimize the WNN model parameters by gradient descent method. Since the WNN has strong nonlinear approximation and adaptive capability, and all the parameters are well optimized by the proposed method, the developed WNN model can make the modeling error PDF track the target PDF, eventually. Simulation example and application in a blast furnace ironmaking process show that the proposed method has a higher modeling precision and better generalization ability compared with the conventional WNN modeling based on MSE criteria. Furthermore, the proposed method has more desirable estimation for modeling error PDF that approximates to a Gaussian distribution whose shape is high and narrow.
Metagenomics, the study of all microbial species cohabitants in an environment, usually produces large amount of sequence data varying from several GBs to a few TBs. Analyzing metagenomics data includes both data-intensive and compute-intensive steps, making the entire process hard to scale. Here we aim to optimize a metagenomics application that partitions the shortgun metagenomics sequences based on their species of origin. Our solution combines MapReduce-based BioPig analytic toolkit with MPI to provide scalability in respective to both data and compute. We also made some improvements to the existing BioPig toolkit by using simplified data types and compressed k-mer storage. These optimizations leads up to 193× speedup for the computing-intensive step and 9.6× speedup over the entire pipeline. Our optimized application is also capable of processing datasets that are 16 times larger on the same hardware platform. These conclusions indicate integrating heterogeneous technologies such as Hadoop and MPI is quite efficient to solve large genomics problems that are both data-intensive and compute-intensive.
This work evaluates different optimization algorithms for computational fluid dynamics (CFD) simulations of engine combustion. Due to the computational expense of CFD simulations, emulators built with machine learning algorithms were used as surrogates for the optimizers. Two types of emulators were used: a Gaussian process (GP) and a weighted variety of machine learning methods called SuperLearner (SL). The emulators were trained using a dataset of 2048 CFD simulations that were run concurrently on a supercomputer. The design of experiments (DOE) for the CFD runs was obtained by perturbing nine input parameters using a Monte-Carlo method. The CFD simulations were of a heavy duty engine running with a low octane gasoline-like fuel at a partially premixed compression ignition mode. Ten optimization algorithms were tested, including types typically used in research applications. Each optimizer was allowed 800 function evaluations and was randomly tested 100 times. The optimizers were evaluated for the median, minimum, and maximum merits obtained in the 100 attempts. Some optimizers required more sequential evaluations, thereby resulting in longer wall clock times to reach an optimum. The best performing optimization methods were particle swarm optimization (PSO), differential evolution (DE), GENOUD (an evolutionary algorithm), and micro-genetic algorithm (GA). These methods found a high median optimum as well as a reasonable minimum optimum of the 100 trials. Moreover, all of these methods were able to operate with less than 100 successive iterations, which reduced the wall clock time required in practice. Two methods were found to be effective but required a much larger number of successive iterations: the DIRECT and MALSCHAINS algorithms. A random search method that completed in a single iteration performed poorly in finding optimum designs but was included to illustrate the limitation of highly concurrent search methods. The last three methods, Nelder–Mead, bound optimization by quadratic approximation (BOBYQA), and constrained optimization by linear approximation (COBYLA), did not perform as well.
Mixed Variable Optimization (MVO) is an optimization method that uses discrete (categorical) and continuous variables in its optimization process. This method of optimization is applicable to highly nonlinear problems where the information used in the optimization is of poor quality of has minor inaccuracies in the data sets.
Purpose: The purpose was to demonstrate a developed acceleration technique of dose optimization and to investigate its applicability to the optimization process in a treatment planning system (TPS) for proton therapy. Methods: In the developed technique, the dose matrix is divided into two parts, main and halo, based on beam sizes. The boundary of the two parts is varied depending on the beam energy and water equivalent depth by utilizing the beam size as a singular threshold parameter. The optimization is executed with two levels of iterations. In the inner loop, doses from the main part are updated, whereas doses from the halo part remain constant. In the outer loop, the doses from the halo part are recalculated. We implemented this technique to the optimization process in the TPS and investigated the dependence on the target volume of the speedup effect and applicability to the worst-case optimization (WCO) in benchmarks. Results: We created irradiation plans for various cubic targets and measured the optimization time varying the target volume. The speedup effect was improved as the target volume increased, and the calculation speed increased by a factor of six for a 1000 cm3 target. An IMPT plan for the RTOG benchmark phantom was created in consideration of ±3.5% range uncertainties using the WCO. Beams were irradiated at 0, 45, and 315 degrees. The target’s prescribed dose and OAR’s Dmax were set to 3 Gy and 1.5 Gy, respectively. Using the developed technique, the calculation speed increased by a factor of 1.5. Meanwhile, no significant difference in the calculated DVHs was found before and after incorporating the technique into the WCO. Conclusion: The developed technique could be adapted to the TPS’s optimization. The technique was effective particularly for large target cases.
Three-dimensional characterization of materials at the nano- and meso-scale has become possible with transmission and scanning transmission electron microscopes (S/TEM). Its importance has extended to a wide class of nanomaterials such as hydrogen fuel cells, solar cells, industrial catalysts, new battery materials and semiconductor devices, as well as spanning high-tech industry, universities, and national labs. While capable instrumentation is abundant, this rapidly expanding demand for high-resolution tomography is bottlenecked by software that is instead tailored for lower-dose, biological applications and not optimized for higher-resolution materials applications. Existing tomography fails to utilize the chemical information provided by S/TEM spectrometers. To address this problem, this project delivered a scalable, fully functional, freely-distributable, open Source materials tomography package with a modern user interface that enables automated acquisition, alignment, and real-time reconstruction of raw tomography data, and provides advanced segmentation, three-dimensional chemical visualization and analysis optimized for materials applications. It has established an extendable framework capable of automation for high-throughput for the tomography of materials from data acquisition to visualization. Phase I and II delivered a full-featured, cross-platform, clean and integrated application for S/TEM materials tomography. It can read projection data from the microscope, with graphical tools for alignment of data, tomographic reconstruction, segmentation, and visualization of the reconstructed 3D volume. The entire pipeline can be saved to an XML state file, enabling fully reproducible data processing and analysis with a full Python environment for the development of custom algorithms, processing, and analysis requirements. Phase IIB extended the application to provide real-time tomographic reconstruction, "live updates" as data is processed, analysis of big data, and multi-channel tomography. Here, quantitative assessment of nanomaterials can occur as data is being recorded on a 3D visualization platform that accommodates additional information from multiple channels. We provide unmatched high-throughput tomography, where the complete tomographic pipeline from measurement to 3D visualization can occur rapidly. The project was extended to offer capabilities for micro-CT, X-ray, neutron, focused ion beam, atomic electron tomography and atom probe tomography. With over 600 transmission electron microscopes worldwide and approximately 50 coming online each year, the demand and impact of an open-source tomography tool is large. Significant opportunities exist in high-tech industry, universities, and national labs to enable or enhance three-dimensional imaging at the nanoscale and bring automated high-throughput approaches that will accelerate progress in materials characterization and metrology. The project supports a service based business model by enabling lab-specific acquisition and processing customization and integration-support and development that will be provided into Phase III and beyond.
The aim of the overall project was research to formulate and solve multi-scale optimization problems for energy storage applications. More specifically, we formulated multi-scale models for optimizing the internal structure of nanoporous materials, with the goal of developing models for related energy applications associated with sustainable sources of energy. The applications include supercapacitors, hydrogen storage, catalytic beds, and filters. The optimization algorithms we have developed can be applied to general hierarchical design problems arising from complex systems.
This project investigated computational methods for optimizing the design of nanoporous materials. These materials hold promise in energy storage applications such as electrical supercapacitors and hydrogen storage systems. A significant issue for nanoporous storage devices is the design of the channel network within these materials to allow rapid diffusion while retaining storage capacity.