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.
A method for aerodynamic shape optimization based on a genetic algorithm approach is demonstrated. The algorithm is coupled with a transonic full potential flow solver and is used to optimize the flow about transonic wings including multi-objective solutions that lead to the generation of pareto fronts. The results indicate that the genetic algorithm is easy to implement, flexible in application and extremely reliable.
A new method for aerodynamic shape optimization using a genetic algorithm with real number encoding is presented. The algorithm is used to optimize three different problems, a simple hill climbing problem, a quasi-one-dimensional nozzle problem using an Euler equation solver and a three-dimensional transonic wing problem using a nonlinear potential solver. Results indicate that the genetic algorithm is easy to implement and extremely reliable, being relatively insensitive to design space noise.
Materials discovery is increasingly being impelled by machine learning methods that rely on pre-existing datasets. Where datasets are lacking, unbiased data generation can be achieved with genetic algorithms. Here a machine learning model is trained on-the-fly as a computationally inexpensive energy predictor before analyzing how to augment convergence in genetic algorithm-based approaches by using the model as a surrogate. This leads to a machine learning accelerated genetic algorithm combining robust qualities of the genetic algorithm with rapid machine learning. The approach is used to search for stable, compositionally variant, geometrically similar nanoparticle alloys to illustrate its capability for accelerated materials discovery, e.g., nanoalloy catalysts. The machine learning accelerated approach, in this case, yields a 50-fold reduction in the number of required energy calculations compared to a traditional “brute force” genetic algorithm. This makes searching through the space of all homotops and compositions of a binary alloy particle in a given structure feasible, using density functional theory calculations.
The initial structures for the search for the global minimum of TiO₂ nanoclusters were generated by combining a tree growth (TG) algorithm with a hybrid genetic algorithm (HGA). In the TG algorithm, the clusters grow from a small seed to the size of interest stepwise. New atoms are added to the smaller cluster from the previous step, by analogy to new leaves grown by a tree. The addition of the new atoms is controlled by predefined geometry parameters to reduce the computational cost and to provide physically meaningful structures. In each step, the energies for the various generated structures are evaluated, and those with the lowest energies are carried into the next step. The structures that match the formulas of interest are collected as HGA candidates during the various steps. Low energy candidates are fed to the HGA component to search for the global minimum for each formula of interest. The lowest energy structures from the HGA are then optimized by using density functional theory to study the dissociation energies of the clusters and the evolution in the structure as the size of the cluster increases. The optimized geometries of the (TiO₂)n nanoclusters for n = 2-13, do not show the character of a TiO₂ bulk crystal with a hexacoordinate Ti. The average clustering energy (ΔEn) converges slowly to the bulk value for rutile. The TiO₂ dissociation energies for (TiO₂)n clusters approach the bulk value for rutile more quickly but show larger variations. The (TiO₂)12 cluster appears to be quite stable, and the (TiO₂)13 cluster is quite unstable on a relative scale.
This paper investigates the feasibility of applying genetic algorithms to solve optimization problems that are implemented entirely in reconfgurable hardware. The paper highlights the pe$ormance/design space trade-offs that must be understood to effectively implement a standard genetic algorithm within a modem Field Programmable Gate Array, FPGA, reconfgurable hardware environment and presents a case-study where this stochastic search technique is applied to standard test-case problems taken from the technical literature. In this research, the targeted FPGA-based platform and high-level design environment was the Starbridge Hypercomputing platform, which incorporates multiple Xilinx Virtex II FPGAs, and the Viva TM graphical hardware description language.
The problem of designing and flying a trajectory for successful recovery of a reusable launch vehicle is tackled using fuzzy logic control with genetic algorithm optimization. The plant is approximated by a simplified three degree of freedom non-linear model. A baseline trajectory design and guidance algorithm consisting of several Mamdani type fuzzy controllers is tuned using a simple genetic algorithm. Preliminary results show that the performance of the overall system is shown to improve with genetic algorithm tuning.
A novel, real-valued Genetic Algorithm (GA) was designed and implemented to minimize the reflectivity and/or transmissivity of an arbitrary number of homogeneous, lossy dielectric or magnetic layers of arbitrary thickness positioned at either the center of an infinitely long rectangular waveguide, or adjacent to the perfectly conducting backplate of a semi-infinite, shorted-out rectangular waveguide. Evolutionary processes extract the optimal physioelectric constants falling within specified constraints which minimize reflection and/or transmission over the frequency band of interest. This GA extracted the unphysical dielectric and magnetic constants of three layers of fictitious material placed adjacent to the conducting backplate of a shorted-out waveguide such that the reflectivity of the configuration was 55 dB or less over the entire X-band. Examples of the optimization of realistic multi-layer absorbers are also presented. Although typical Genetic Algorithms require populations of many thousands in order to function properly and obtain correct results, verified correct results were obtained for all test cases using this GA with a population of only four.
Presented in this report is the description of a new method for neutron energy spectrum adjustment which uses a genetic algorithm to minimize the difference between calculated and measured reaction probabilities. The measured reaction probabilities are found using neutron activation analysis. The method adjusts a trial spectrum provided by the user which is typically calculated using a neutron transport code such as MCNP. Observed benefits of this method over currently existing methods include the reduction in unrealistic artefacts in the spectral shape as well as a reduced sensitivity to increases in the energy resolution of the derived spectrum. This report presents the adjustment results for various spectrum altering bucket environments in the central cavity of the Annular Core Research Reactor, as well as the adjustment results for the spectrum in the Sandia Pulse Reactor III. In each case, the results are compared to those generated using LSL-M2, which is a code commonly used for the purpose of spectrum adjustment. The genetic algorithm produces spectrum-averaged reaction probabilities with agreement to measured values, and comparable to those resulting from LSL-M2. The true benefit to this method, the reduction of shape artefacts in the spectrum, is difficult to quantify but can be clearly seen in the comparison of the final adjustments. Beyond these preliminary results, this report also gives a thorough description of the genetic algorithm and presents instructions for running the code using the graphical user interface. In its present state, the code does not provide uncertainties or correlations for the adjusted spectrum. This capability is currently being added, and will be presented in future work.
In this paper, a unique system level spacecraft design optimization will be presented. A Genetic Algorithm is used to design the global pattern of the reinforcing structure, while a gradient routine is used to adequately stiffen the sub-structure. The system level structural design includes determining the optimal physical location (and number) of reinforcing beams of a lunar pallet lander deck structure. Design of the substructure includes determining placement of secondary stiffeners and the number of rivets required for assembly.. In this optimization, several considerations are taken into account. The primary objective was to raise the primary natural frequencies of the structure such that the Pallet Lander primary structure does not significantly couple with the launch vehicle. A secondary objective is to determine how to properly stiffen the reinforcing beams so that the beam web resists the shear buckling load imparted by the spacecraft components mounted to the pallet lander deck during launch and landing. A third objective is that the calculated stress does not exceed the allowable strength of the material. These design requirements must be met while, minimizing the overall mass of the spacecraft. The final paper will discuss how the optimization was implemented as well as the results. While driven by optimization algorithms, the primary purpose of this effort was to demonstrate the capability of genetic algorithms to enable design automation in the preliminary design cycle. By developing a routine that can automatically generate designs through the use of Finite Element Analysis, considerable design efficiencies, both in time and overall product, can be obtained over more traditional brute force design methods.
A genetic-algorithm-based optimization approach is described and evaluated using a simple hill-climbing model problem. The model problem utilized herein allows for the broad specification of a large number of search spaces including spaces with an arbitrary number of genes or decision variables and an arbitrary number hills or modes. In the present study, only single objective problems are considered. Results indicate that the genetic algorithm optimization approach is flexible in application and extremely reliable, providing optimal results for all problems attempted. The most difficult problems - those with large hyper-volumes and multi-mode search spaces containing a large number of genes - require a large number of function evaluations for GA convergence, but they always converge.
Aims: The present case study is on maximizing the aqua feed properties using response surface methodology and genetic algorithm. Study Design: Effect of extrusion process variables like screw speed, L/D ratio, barrel temperature, and feed moisture content were analyzed to maximize the aqua feed properties like water stability, true density, and expansion ratio. Place and Duration of Study: This study was carried out in the Department of Agricultural and Food Engineering, Indian Institute of Technology, Kharagpur, India. Methodology: A variable length single screw extruder was used in the study. The process variables selected were screw speed (rpm), length-to-diameter (L/D) ratio, barrel temperature (degrees C), and feed moisture content (%). The pelletized aqua feed was analyzed for physical properties like water stability (WS), true density (TD), and expansion ratio (ER). Extrusion experimental data was collected by based on central composite design. The experimental data was further analyzed using response surface methodology (RSM) and genetic algorithm (GA) for maximizing feed properties. Results: Regression equations developed for the experimental data has adequately described the effect of process variables on the physical properties with coefficient of determination values (R2) of > 0.95. RSM analysis indicated WS, ER, and TD were maximized at L/D ratio of 12-13, screw speed of 60-80 rpm, feed moisture content of 30-40%, and barrel temperature of = 80 degrees C for ER and TD and > 90 degrees C for WS. Based on GA analysis, a maxium WS of 98.10% was predicted at a screw speed of 96.71 rpm, L/D radio of 13.67, barrel temperature of 96.26 degrees C, and feed moisture content of 33.55%. Maximum ER and TD of 0.99 and 1346.9 kg/m3 was also predicted at screw speed of 60.37 and 90.24 rpm, L/D ratio of 12.18 and 13.52, barrel temperature of 68.50 and 64.88 degrees C, and medium feed moisture content of 33.61 and 38.36%. Conclusion: The present data analysis indicated that WS is mainly governed by barrel temperature and feed moisture content, which might have resulted in formation of starch-protein complexes due to denaturation of protein and gelatinization of starch. Screw speed coupled with temperature and feed moisture content controlled the ER and TD values. Higher screw speeds might have reduced the viscosity of the feed dough resulting in higher TD and lower ER values. Based on RSM and GA analysis screw speed, barrel temperature and feed moisture content were the interacting process variables influencing maximum WS followed by ER and TD.
A Machine Learning-Genetic Algorithm (ML-GA) approach was developed to virtually discover optimum designs using training data generated from multi-dimensional simulations. Machine learning (ML) presents a pathway to transform complex physical processes that occur in a combustion engine into compact informational processes. In the present work, a total of over 2000 sector-mesh computational fluid dynamics (CFD) simulations of a heavy-duty engine were performed. These were run concurrently on a supercomputer to reduce overall turnaround time. The engine being optimized was run on a low-octane (RON70) gasoline fuel under partially premixed compression ignition (PPCI) mode. A total of nine input parameters were varied, and the CFD simulation cases were generated by randomly sampling points from this nine-dimensional input space. These input parameters included fuel injection strategy, injector design, and various in-cylinder flow and thermodynamic conditions at intake valve closure (IVC). The outputs (targets) of interest from these simulations included five metrics related to engine performance and emissions. Over 2000 samples generated from CFD were then used to train an ML model that could predict these five targets based on the nine input features. A robust super learner approach was employed to build the ML model, where results from a collection of different ML algorithms were pooled together. Thereafter, a stochastic global optimization genetic algorithm (GA) was used, with the ML model as the objective function, to optimize the input parameters based on a merit function so as to minimize fuel consumption while satisfying CO and NOx emissions constraints. The optimized configuration from ML-GA was found to be very close to that obtained from a sequentially performed CFD-GA approach, where a CFD simulation served as the objective function. In addition, the overall turnaround time was (at least) 75% lower with the ML-GA approach, as the training data was generated from concurrent CFD simulations and employing the ML model as the objective function significantly accelerated the GA optimization. Finally, this study demonstrates the potential of ML-GA and high-performance computing (HPC) to reduce the number of CFD simulations to be performed for optimization problems without loss in accuracy, thereby providing significant cost savings compared to traditional approaches.
A genetic algorithm is developed and applied to optimize hyperparameters of convolutional recursively determined dual neural network-Gaussian process (NNGP) kernels. As a specific application of the combined GPNN-GA algorithm, it is applied to image classification in publicly available data of Hyper Suprime-Cam Subaru Strategic Program. Matthews correlation coefficient is calculated based on results of binary star-galaxy classification and used as a fitting function of the GA module of the algorithm. The simulation results confirm significant improvement of the classification accuracy with optimized hyperparameters.
Plutonium nuclear material safeguards is essential to ensure that states do not divert plutonium to clandestine nuclear weapons programs. Traditional safeguards methods rely on correlated neutron measurements. Detection systems based on organic scintillators, which are sensitive to fast neutrons and photons, are gaining support for use in safeguards applications due to increased capabilities and shorter measurement times compared to the current state-of-the-art. A drawback of organic scintillators in plutonium measurements is data throughput limitations; 241 Am builds up in plutonium samples and emits 60-keV gamma rays with a high specific activity. In this work, a genetic algorithm is used to design graded shielding made of tin and copper to attenuate the 60-keV gamma rays. The shield is designed to limit the attenuation of the higher-energy, fission photons, because fission photons can be used as a signature. The genetic algorithm optimization takes days to run, while a complete search space exploration would take years. The genetic algorithm was run for different levels of 60-keV gamma-ray attenuation, and the thickness of the required shielding layers was determined at each level. Tin–copper shielding was placed in front of an array of organic scintillation detectors for a measurement of approximately 4.5 kg of alpha-phase, weapons-grade plutonium, an AmLi source, and a 252 Cf source. We show that the shield successfully attenuated low-energy, non-fission gamma rays, which significantly improved the organic scintillator performance.
We analyse scattering of light from multi-layer plasmonic nanowires and employ a genetic algorithm for optimizing the scattering cross section. We apply the mode-expansion method using experimental data for material parameters to demonstrate that our genetic algorithm allows designing realistic core-shell nanostructures with the superscattering effect achieved at any desired wavelength. This approach can be employed for optimizing both superscattering and cloaking at different wavelengths in the visible spectral range.
SPLICER computer program used to solve search and optimization problems. Genetic algorithms adaptive search procedures (i.e., problem-solving methods) based loosely on processes of natural selection and Darwinian "survival of fittest." Algorithms apply genetically inspired operators to populations of potential solutions in iterative fashion, creating new populations while searching for optimal or nearly optimal solution to problem at hand. Written in Think C.
Optimal design method of high-power microwave source using particle simulation and parallel genetic algorithms is presented in this paper. The output power, simulated by the fully electromagnetic particle simulation code UNIPIC, of the high-power microwave device is given as the fitness function, and the float-encoding genetic algorithms are used to optimize the high-power microwave devices. Using this method, we encode the heights of non-uniform slow wave structure in the relativistic backward wave oscillators (RBWO), and optimize the parameters on massively parallel processors. Simulation results demonstrate that we can obtain the optimal parameters of non-uniform slow wave structure in the RBWO, and the output microwave power enhances 52.6% after the device is optimized.
Traditionally, structural loads are measured using strain gages. A loads calibration test must be done before loads can be accurately measured. In one measurement method, a series of point loads is applied to the structure, and loads equations are derived via the least squares curve fitting algorithm using the strain gage responses to the applied point loads. However, many research structures are highly instrumented with strain gages, and the number and selection of gages used in a loads equation can be problematic. This paper presents an improved technique using a genetic algorithm to choose the strain gages used in the loads equations. Also presented are a comparison of the genetic algorithm performance with the current T-value technique and a variant known as the Best Step-down technique. Examples are shown using aerospace vehicle wings of high and low aspect ratio. In addition, a significant limitation in the current methods is revealed. The genetic algorithm arrived at a comparable or superior set of gages with significantly less human effort, and could be applied in instances when the current methods could not.