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.
Necessary conditions of optimality for control problems with state variable inequality constraints, using separating hyperplane theorem
Nonnegative Least Squares (NNLS) is a fundamental constrained optimization problem encountered in many applications such as image deblurring, signal processing, nonnegative matrix factorization, magnetic microscopy, and hyperspectral imaging. Active-set based methods are a common class of algorithms for solving NNLS which identify the optimal variable set of the NNLS solution. They do so by iteratively solving a series of unconstrained least squares problems, identifying which variables violate the nonnegativity constraints, and then swapping variables in/out of consideration until the optimal set of variables is found. Several variations improving upon this method exist in the literature. In this work, we propose an active-set swap heuristic which further improves upon existing active-set based methods for NNLS. Our optimizations are based upon adding multiple variables to the passive set within a threshold of the smallest gradient value and removing variables within a similar threshold of the closest boundary constraint. We leverage these optimizations to yield a Fast Active-Set Thresholding NNLS (FAST-NNLS) algorithm which significantly outperforms the existing state-of-the-art NNLS algorithms for a wide range of problems. Rigorous convergence guarantees are proven for the proposed method. We demonstrate the effectiveness of our proposed method on multiple synthetic datasets and two realworld text analysis applications. In doing so, we present the most comprehensive NNLS solver comparison in the literature to date.
The new concept of using a continuum model in the multilevel decomposition of structural optimization problems is presented. The practicality of this concept is demonstrated by application in a scheme for the optimization of beam-like space structures. This scheme is tested against traditional optimization procedures for savings in computational cost. Results from both optimization methods are presented for comparison.
It has been well documented that decisions made in the early stages of Conceptual and Pre-Conceptual design commit up to 80% of total Life-Cycle Cost (LCC) while engineers know the least about the product they are designing [1]. Once within Preliminary and Detailed design however, making changes to the design becomes far more difficult to enact in both cost and schedule. Primarily this has been due to a lack of detailed data usually uncovered later during the Preliminary and Detailed design phases. In our current budget-constrained environment, making decisions within Conceptual and Pre-Conceptual design which minimize LCC while meeting requirements is paramount to a program's success. Within the arena of launch vehicle design, optimizing the ascent trajectory is critical for minimizing the costs present within such concerns as propellant, aerodynamic, aeroheating, and acceleration loads while meeting requirements such as payload delivered to a desired orbit. In order to optimize the vehicle design its constraints and requirements must be known, however as the design cycle proceeds it is all but inevitable that the conditions will change. Upon that change, the previously optimized trajectory may no longer be optimal, or meet design requirements. The current paradigm for adjusting to these updates is generating point solutions for every change in the design's requirements [2]. This can be a tedious, time-consuming task as changes in virtually any piece of a launch vehicle's design can have a disproportionately large effect on the ascent trajectory, as the solution space of the trajectory optimization problem is both non-linear and multimodal [3]. In addition, an industry standard tool, Program to Optimize Simulated Trajectories (POST), requires an expert analyst to produce simulated trajectories that are feasible and optimal [4]. In a previous publication the authors presented a method for combatting these challenges [5]. In order to bring more detailed information into Conceptual and Pre-Conceptual design, knowledge of the effects originating from changes to the vehicle must be calculated. In order to do this, a model capable of quantitatively describing any vehicle within the entire design space under consideration must be constructed. This model must be based upon analysis of acceptable fidelity, which in this work comes from POST. Design space interrogation can be achieved with surrogate modeling, a parametric, polynomial equation representing a tool. A surrogate model must be informed by data from the tool with enough points to represent the solution space for the chosen number of variables with an acceptable level of error. Therefore, Design Of Experiments (DOE) is used to select points within the design space to maximize information gained on the design space while minimizing number of data points required. To represent a design space with a non-trivial number of variable parameters the number of points required still represent an amount of work which would take an inordinate amount of time via the current paradigm of manual analysis, and so an automated method was developed. The best practices of expert trajectory analysts working within NASA Marshall's Advanced Concepts Office (ACO) were implemented within a tool called multiPOST. These practices include how to use the output data from a previous run of POST to inform the next, determining whether a trajectory solution is feasible from a real-world perspective, and how to handle program execution errors. The tool was then augmented with multiprocessing capability to enable analysis on multiple trajectories simultaneously, allowing throughput to scale with available computational resources. In this update to the previous work the authors discuss issues with the method and solutions.
A simple approach for optimizing the jig-shape is proposed in this study. This simple approach is based on an unconstrained optimization problem and applied to a low-boom supersonic aircraft. In this study, the jig-shape optimization is performed using the two-step approach. First, starting design variables are computed using the least squares surface fitting technique. Next, the jig-shape is further tuned using a numerical optimization procedure based on in-house object-oriented optimization tool.
A simple approach for optimizing the jig-shape is proposed in this study. This simple approach is based on an unconstrained optimization problem and applied to a low-boom supersonic aircraft. In this study, the jig-shape optimization is performed using the two-step approach. First, starting design variables are computed using the least-squares surface fitting technique. Next, the jig-shape is further tuned using a numerical optimization procedure based on an in-house object-oriented optimization tool. During the numerical optimization procedure, a design jig-shape is determined by the baseline jig-shape and basis functions. A total of 12 symmetric mode shapes of the cruise-weight configuration, rigid pitch shape, rigid left and right stabilator rotation shapes, and a residual shape are selected as sixteen basis functions. After three optimization runs, the trim shape error distribution is improved, and the maximum trim shape error of 0.9844 inches of the starting configuration becomes 0.00367 inch by the end of the third optimization run.
A simple approach for optimizing the jig-shape is proposed in this study. This simple approach is based on an unconstrained optimization problem and applied to a low-boom supersonic aircraft. In this study, the jig-shape optimization is performed using the two-step approach. First, starting design variables are computed using the least squares surface fitting technique. Next, the jig-shape is further tuned using a numerical optimization procedure based on in-house object-oriented optimization tool.
Need: A need exists for a generalized, robust, user-friendly and accessible end-to-end mission design optimization tool. Solution: Our solution to developing this capability was to interface two JSC tools—Copernicus and Genesis. Each of these tools has a specific area of the mission design process that it excels at. By utilizing them both, we can gain performance benefits not seen by either on their own. - Copernicus is a trajectory design and optimization software used for in-space trajectories around multiple bodies. - Genesis is a flight mechanics tool used to model ascent, entry, descent, and landing trajectories around a single planetary body. Year 1 was focused on combining these 2 software packages—allowing Copernicus to incorporate the ascent/descent capabilities of Genesis into the optimization problem—and developing this end-to-end mission design capability. Year 2 we focused on increasing the robustness of this capability by building the initial guess generator (IGG), which produces initial guesses based on simplifying assumptions and the physics of the problem. Year 3 of our project focused on utilizing the end-to-end mission design and optimization capabilities developed in the previous 2 years to analyze specific mission scenarios—scaling up from proof of concept to real analyses—capturing any resulting performance benefits, as well as addressing the Big Data challenges we’re faced with—namely, how we’re going to manage and interpret all the data that’s generated.
Abstract Particle accelerator operation requires simultaneous optimization of multiple objectives. Multi-objective optimization (MOO) is particularly challenging due to trade-offs between the objectives. Evolutionary algorithms, such as genetic algorithms (GAs), have been leveraged for many optimization problems, however, they do not apply to complex control problems by design. This paper demonstrates the power of differentiability for solving MOO problems in particle accelerators using a deep differentiable reinforcement learning (DDRL) algorithm. We compare the DDRL algorithm with model-free reinforcement learning (MFRL), GA, and Bayesian optimization (BO) for simultaneous optimization of heat load and trip rates in the continuous electron beam accelerator facility. The underlying problem enforces strict constraints on both individual states and actions as well as cumulative (global) constraints on energy requirements of the beam. Using historical accelerator data, we develop a physics-based surrogate model which is differentiable and allows for back-propagation of gradients. The results are evaluated in the form of a Pareto-front with two objectives. We show that the DDRL outperforms MFRL, BO, and GA on high dimensional problems.
This paper describes a distributed software system for solving large-scale structural optimization problems. Distributing the software between a mainframe computer and a minicomputer takes advantage of some of the best features available on each computer. The described software system consists of a finite element structural analysis computer program, a general purpose optimizer program, and several small user-supplied problem dependent programs. Comparison with a similar system executing entirely on the mainframe computer reveals that the distributed system costs less, uses computer resources more efficiently and improves production through faster turnaround and improved user control. The system interfaces with interactive graphics software for generating models and displaying the intermediate and final results
Improving distribution system resilience against frequent extreme weather events is important for reliable power system operations. Especially when dealing with events such as hurricanes that have short-term predictions, proactive pre-event preparedness plays a vital role in system resilience performance. In this paper, we propose a novel approach to construct pre-event resource allocation plans for system operators to cope with upcoming threats through a sequential attacker-defender game framework. The sequential attacker-defender game is designed to model the interaction between the extreme weather and the system operator. In each round of the game, the attacker and the defender sequentially update their current strategies by accounting for the opponent’s action set. The attacker model is formulated as a bi-level problem to identify the severe outage scenarios, and the defender model is formulated as a two-stage optimization problem to determine the allocation of restoration resources including mobile responsive resources and repair crews. Two scale-reduction strategies are proposed to ensure the scalability of the game scheme. Finally, case studies on the IEEE 33-bus and a 7149-node practical utility system validate the effectiveness of the proposed sequential game and the efficiency of the scale-reduction strategies.
This final technical report summarizes the outcomes of a DOE-funded project on federated scientific machine learning (FL) under nonsmooth, nonconvex, and hierarchical optimization settings. The project develops new mathematical models, algorithms, and theoretical guarantees for decentralized stochastic, bilevel, and minimax optimization problems arising in DOE mission-relevant applications. A unified framework of randomized and zeroth-order federated optimization methods is introduced, providing provable convergence, communication efficiency, and sample-complexity guarantees. The report documents algorithmic design, theoretical analysis, and empirical validation of the proposed federated learning methods. The project also contributes to workforce development through graduate training and dissemination of results via publications and seminars.
The work that has been done in the last decade or so in the application of optimization techniques to vehicle design is discussed. Much of the work reviewed deals with the design of body or suspension (chassis) components for reduced weight. Also reviewed are studies dealing with system optimization problems for improved functional performance, such as ride or handling. In reviewing the work on the use of optimization techniques, one notes the transition from the rare mention of the methods in the 70's to an increased effort in the early 80's. Efficient and convenient optimization and analysis tools still need to be developed so that they can be regularly applied in the early design stage of the vehicle development cycle to be most effective. Based on the reported applications, an attempt is made to assess the potential for automotive application of optimization techniques. The major issue involved remains the creation of quantifiable means of analysis to be used in vehicle design. The conventional process of vehicle design still contains much experience-based input because it has not yet proven possible to quantify all important constraints. This restraint on the part of the analysis will continue to be a major limiting factor in application of optimization to vehicle design.
The approach to carrying out multi-discipline aerospace design studies in the future, especially in massively parallel computing environments, comprises of choosing (1) suitable solvers to compute solutions to equations characterizing a discipline, and (2) efficient optimization methods. In addition, for aerodynamic optimization problems, (3) smart methodologies must be selected to modify the surface shape. In this research effort, a 'direct' optimization method is implemented on the Cray C-90 to improve aerodynamic design. It is coupled with an existing implicit Navier-Stokes solver, OVERFLOW, to compute flow solutions. The optimization method is chosen such that it can accomodate multi-discipline optimization in future computations. In the work , however, only single discipline aerodynamic optimization will be included.
The Newton Modified Barrier Method (NMBM) is applied to structural optimization problems with large a number of design variables and constraints. This nonlinear mathematical programming algorithm was based on the Modified Barrier Function (MBF) theory and the Newton method for unconstrained optimization. The distinctive feature of the NMBM method is the rate of convergence that is due to the fact that the design remains in the Newton area after each Lagrange multiplier update. This convergence characteristic is illustrated by application to structural problems with a varying number of design variables and constraints. The results are compared with those obtained by optimality criteria (OC) methods and by the ASTROS program.
Problems related to interdisciplinary interactions in the design of a complex engineering systems are examined with reference to aerospace applications. The interdisciplinary optimization problems examined include those dealing with controls and structures, materials and structures, control and stability, structure and aerodynamics, and structure and thermodynamics. The discussion is illustrated by the following specific applications: integrated aerodynamic/structural optimization of glider wing; optimization of an antenna parabolic dish structure for minimum weight and prescribed emitted signal gain; and a multilevel optimization study of a transport aircraft.
The complete flightpath optimization problem for minimum fuel consumption from takeoff to landing including the initial and final turns from and to the runway heading is solved. However, only the initial and final segments which contain the turns are treated, since the straight-line climbout, cruise, and descent problems have already been solved. The paths are derived by generating fields of extremals, using the necessary conditions of optimal control together with singular arcs and state constraints. Results show that the speed profiles for straight flight and turning flight are essentially identical except for the final horizontal accelerating or decelerating turns. The optimal turns require no abrupt maneuvers, and an approximation of the optimal turns could be easily integrated with present straight-line climb-cruise-descent fuel-optimization algorithms. Climbout at the optimal IAS rather than the 250-knot terminal-area speed limit would save 36 lb of fuel for the 727-100 aircraft.
Recent developments in multilevel optimization are briefly reviewed. The general nature of the multilevel design task, the use of approximations to develop and solve the analysis design task, the structure of the formal multidiscipline optimization problem, a simple cantilevered beam which demonstrates the concepts of multilevel design and the basic mathematical details of the optimization task and the system level are among the topics discussed.