Search NASA⌕ Search

SEARCH · Search NASA

Results for “Feedforward 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 73 records · Page 4

The case for digital twins in metal additive manufacturing

The digital twin (DT) is a relatively new concept that is finding increased acceptance in industry. A DT is generally considered as comprising a physical entity, its virtual replica, and two-way digital data communications in-between. Its primary purpose is to leverage the process intelligence captured within digital models—or usually their faster-solving surrogates—towards generating increased value from the physical entities. The surrogate models are created using machine learning based on data obtained from the field, experiments and digital models, which may be physics-based or statistics-based. Anomaly detection and correction, and diagnostic closed-loop process control are examples of how a process DT can be deployed. In the manufacturing industry, its use can achieve improvements in product quality and process productivity. Metal additive manufacturing (AM) stands to gain tremendously from the use of DTs. This is because the AM process is inherently chaotic, resulting in poor repeatability. However, a DT acting in a supervisory role can inject certainty into the process by actively keeping it within bounds through real-time control commands. Closed-loop feedforward control is achieved by observing the process through sensors that monitor critical parameters and, if there are any deviations from their respective optimal ranges, suitable corrective actions are triggered. The type of corrective action (e.g. a change in laser power or a modification to the scanning speed) and its magnitude are determined by interrogating the surrogate models. Because of their artificial intelligence (AI)-endowed predictive capabilities, which allow them to foresee a future state of the physical twin (e.g. the AM process), DTs proactively take context-sensitive preventative steps, whereas traditional closed-loop feedback control is usually reactive. Apart from assisting a build process in real-time, a DT can help with planning the build of a part by pinpointing the optimum processing window relevant to the desired outcome. Again, the surrogate models are consulted to obtain the required information. In this article, we explain how the application of DTs to the metal AM process can significantly widen its application space by making the process more repeatable (through quality assurance) and cheaper (by getting builds right the first time).

36 MATERIALS SCIENCE↗

Multi-Stage System for Automatic Target Recognition

A multi-stage automated target recognition (ATR) system has been designed to perform computer vision tasks with adequate proficiency in mimicking human vision. The system is able to detect, identify, and track targets of interest. Potential regions of interest (ROIs) are first identified by the detection stage using an Optimum Trade-off Maximum Average Correlation Height (OT-MACH) filter combined with a wavelet transform. False positives are then eliminated by the verification stage using feature extraction methods in conjunction with neural networks. Feature extraction transforms the ROIs using filtering and binning algorithms to create feature vectors. A feedforward back-propagation neural network (NN) is then trained to classify each feature vector and to remove false positives. The system parameter optimizations process has been developed to adapt to various targets and datasets. The objective was to design an efficient computer vision system that can learn to detect multiple targets in large images with unknown backgrounds. Because the target size is small relative to the image size in this problem, there are many regions of the image that could potentially contain the target. A cursory analysis of every region can be computationally efficient, but may yield too many false positives. On the other hand, a detailed analysis of every region can yield better results, but may be computationally inefficient. The multi-stage ATR system was designed to achieve an optimal balance between accuracy and computational efficiency by incorporating both models. The detection stage first identifies potential ROIs where the target may be present by performing a fast Fourier domain OT-MACH filter-based correlation. Because threshold for this stage is chosen with the goal of detecting all true positives, a number of false positives are also detected as ROIs. The verification stage then transforms the regions of interest into feature space, and eliminates false positives using an artificial neural network classifier. The multi-stage system allows tuning the detection sensitivity and the identification specificity individually in each stage. It is easier to achieve optimized ATR operation based on its specific goal. The test results show that the system was successful in substantially reducing the false positive rate when tested on a sonar and video image datasets.

Chao, Tien-Hsin↗

Vector-level feedforward control of LPBF melt pool area using a physics-based thermal model

Laser powder bed fusion (LPBF) is an additive manufacturing technique that has gained popularity thanks to its ability to produce geometrically complex, fully dense metal parts. However, these parts are prone to internal defects and geometric inaccuracies, stemming in part from variations in the melt pool. Here, this paper proposes a novel vector-level feedforward control framework for regulating melt pool area in LPBF. By decoupling part-scale thermal behavior from small-scale melt pool physics, the controller provides a scale-agnostic prediction of melt pool area and efficient optimization over it. This is done by operating on two coupled lightweight models: a finite-difference thermal model that efficiently captures vector-level temperature fields and a reduced-order, analytical melt pool model. Each model is calibrated separately with minimal single-track and 2D experiments, and the framework is validated on a complex 3D geometry in both Inconel 718 and 316L stainless steel. Results showed that feedforward vector-level laser power scheduling reduced geometric inaccuracy in key dimensions by 62%, overall porosity by 16.5%, and photodiode root-mean-squared deviation by 38.5% on average. Overall, this modular, data-efficient approach demonstrates that proactively compensating for known thermal effects can significantly improve part quality while remaining computationally efficient and readily extensible to other materials and machines.

Additive manufacturing↗

An Efficient Bayesian Method for Advancing the Application of Deep Learning in Earth Science

Despite their wide and successful applications, deep learning (DL) models are prone to overfitting for small training datasets, produce a poor predictive performance for uncertain data, and provide point estimations without any indication of the accuracy and credibility. These limitations of the deterministic DL models hinder their effective application in Earth system science where the labelled data are sparse, noisy and incomplete with large uncertainty and where the predictive uncertainty quantification is needed for scientific understanding and policy decision making. Integration of Bayesian inference into DL models adds an estimate of uncertainty and regularization in the predictions. However, traditional Bayesian methods are computationally unaffordable and inflexible for high-dimensional problems, which limits their application to DL systems that typically have millions of model parameters. In this effort, we propose an efficient and general-purpose Bayesian inference method to advance DL model optimization and uncertainty quantification, so as to facilitate the adoption of DL in Earth sciences. In a demonstration, we integrate the proposed Bayesian method with a feedforward neural network (NN) to build a fast-to-evaluate surrogate of the complex Energy Exascale Earth System Land Model for efficient modeling. The formulated Bayesian NN, using a small number of training data, produces an accurate prediction with high credibility, whereas with the same small training size, the deterministic NN cannot yield a reasonable estimation and does not provide confidence information. The proposed Bayesian method is computationally efficient and flexible, capable of integration with diverse network variants such as convolutional NNs and recurrent NNs to advance the application of DL in Earth sciences.

Lu, Dan↗

Short-Term Forecasting Across a Network for the Autonomous Wind Farm: Preprint

In an autonomous wind farm, turbines will use information from nearby turbines to achieve wind farm-level objectives such as optimizing the overall performance of a wind farm, ensuring resiliency when other sensors fail, and adapting to changing local conditions. In this paper, the wind farm can be modeled as a network within which turbines (nodes) share information across designated communication channels, with a focus on turbines at the outside of the wind farm capturing local effects and sharing that information with downstream turbines. Understanding of varied inflow conditions can be especially important in complex terrain. This information can be used to monitor turbines, self-organize turbines into groups, and predict the power performance of a wind farm. In particular, this paper describes an autonomous wind farm that incorporates information from local sensors in real time to predict wind speed and wind direction at each turbine over a short-term horizon. Results indicate that the estimate of wind direction can be used to improve the knowledge of the wind speed and direction over the persistence method on a 10- 15-minute time horizon. These short-term forecasts can also be used to facilitate advanced control methods such as feedforward control within a wind farm.

17 WIND ENERGY↗

Short-Term Forecasting Across a Network for the Autonomous Wind Farm

In an autonomous wind farm, turbines will use information from nearby turbines to achieve wind farm-level objectives such as optimizing the overall performance of a wind farm, ensuring resiliency when other sensors fail, and adapting to changing local conditions. In this paper, the wind farm can be modeled as a network within which turbines (nodes) share information across designated communication channels, with a focus on turbines at the outside of the wind farm capturing local effects and sharing that information with downstream turbines. Understanding of varied inflow conditions can be especially important in complex terrain. This information can be used to monitor turbines, self-organize turbines into groups, and predict the power performance of a wind farm. In particular, this paper describes an autonomous wind farm that incorporates information from local sensors in real time to predict wind speed and wind direction at each turbine over a short-term horizon. Results indicate that the estimate of wind direction can be used to improve the knowledge of the wind speed and direction over the persistence method on a 10-15-minute time horizon. These short-term forecasts can also be used to facilitate advanced control methods such as feedforward control within a wind farm.

17 WIND ENERGY↗

High-speed adaptive contact-mode atomic force microscopy imaging with near-minimum-force

In this paper, an adaptive contact-mode imaging approach is proposed to replace the traditional contact-mode imaging by addressing the major concerns in both the speed and the force exerted to the sample. The speed of the traditional contact-mode imaging is largely limited by the need to maintain precision tracking of the sample topography over the entire imaged sample surface, while large image distortion and excessive probe-sample interaction force occur during high-speed imaging. In this work, first, the image distortion caused by the topography tracking error is accounted for in the topography quantification. Second, the quantified sample topography is utilized in a gradient-based optimization method to adjust the cantilever deflection set-point for each scanline closely around the minimal level needed for maintaining stable probe-sample contact, and a data-driven iterative feedforward control that utilizes a prediction of the next-line topography is integrated to the topography feeedback loop to enhance the sample topography tracking. The proposed approach is demonstrated and evaluated through imaging a calibration sample of square pitches at both high speeds (e.g., scan rate of 75 Hz and 130 Hz) and large sizes (e.g., scan size of 30 μm and 80 μm). The experimental results show that compared to the traditional constant-force contact-mode imaging, the imaging speed can be increased by over 30 folds (with the scanning speed at 13 mm/s), and the probe-sample interaction force can be reduced by more than 15% while maintaining the same image quality.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Two-photon exchange effect studied with neural networks

An approach to the extraction of the two-photon exchange (TPE) correction from elastic ep scattering data is presented. The cross-section, polarization transfer (PT), and charge asymmetry data are considered. It is assumed that the TPE correction to the PT data is negligible. The form factors and TPE correcting term are given by one multidimensional function approximated by the feedforward neural network (NN). To find a model-independent approximation, the Bayesian framework for the NNs is adapted. A large number of different parametrizations is considered. The most optimal model is indicated by the Bayesian algorithm. The obtained fit of the TPE correction behaves linearly in {epsilon} but it has a nontrivial Q{sup 2} dependence. A strong dependence of the TPE fit on the choice of parametrization is observed.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Optimizing temperature distributions for training neural quantum states using parallel tempering

Parametrized artificial neural networks (ANNs) can be very expressive ansatzes for variational algorithms, reaching state-of-the-art energies on many quantum many-body Hamiltonians. Nevertheless, the training of the ANN can be slow and stymied by the presence of local minima in the parameter landscape. One approach to mitigate this issue is to use parallel tempering methods, and in this work, we focus on the role played by the temperature distribution of the parallel tempering replicas. Using an adaptive method that adjusts the temperatures in order to equate the exchange probability between neighboring replicas, we show that this temperature optimization can significantly increase the success rate of the variational algorithm with negligible computational cost by eliminating bottlenecks in the replicas' random walk. Furthermore, we demonstrate this using two different neural networks, a restricted Boltzmann machine and a feedforward network, which we use to study a toy problem based on a permutation invariant Hamiltonian with a pernicious local minimum and the 𝐽 1 −𝐽 2 model on a rectangular lattice.

Neural network simulations↗

Microgravity vibration isolation: Optimal preview and feedback control

In order to achieve adequate low-frequency vibration isolation for certain space experiments an active control is needed, due to inherent passive-isolator limitations. Proposed here are five possible state-space models for a one-dimensional vibration isolation system with a quadratic performance index. The five models are subsets of a general set of nonhomogeneous state space equations which includes disturbance terms. An optimal control is determined, using a differential equations approach, for this class of problems. This control is expressed in terms of constant, Linear Quadratic Regulator (LQR) feedback gains and constant feedforward (preview) gains. The gains can be easily determined numerically. They result in a robust controller and offers substantial improvements over a control that uses standard LQR feedback alone.

Hampton, R. D.↗

Feedforward/feedback control synthesis for performance and robustness

Both feedforward and feedback control approaches for uncertain dynamical systems are investigated. The control design objective is to achieve a fast settling time (high performance) and robustness (insensitivity) to plant modeling uncertainty. Preshapong of an ideal, time-optimal control input using a 'tapped-delay' filter is shown to provide a rapid maneuver with robust performance. A robust, non-minimum-phase feedback controller is synthesized with particular emphasis on its proper implementation for a non-zero set-point control problem. The proposed feedforward/feedback control approach is robust for a certain class of uncertain dynamical systems, since the control input command computed for a given desired output does not depend on the plant parameters.

Wie, Bong↗

Robot-Arm Dynamic Control by Computer

Feedforward and feedback schemes linearize responses to control inputs. Method for control of robot arm based on computed nonlinear feedback and state tranformations to linearize system and decouple robot end-effector motions along each of cartesian axes augmented with optimal scheme for correction of errors in workspace. Major new feature of control method is: optimal error-correction loop directly operates on task level and not on joint-servocontrol level.

Bejczy, Antal K.↗

Robust control of flexible space vehicles with minimum structural excitation: On-off pulse control of flexible space vehicles

Both feedback and feedforward control approaches for uncertain dynamical systems (in particular, with uncertainty in structural mode frequency) are investigated. The control objective is to achieve a fast settling time (high performance) and robustness (insensitivity) to plant uncertainty. Preshaping of an ideal, time optimal control input using a tapped-delay filter is shown to provide a fast settling time with robust performance. A robust, non-minimum-phase feedback controller is synthesized with particular emphasis on its proper implementation for a non-zero set-point control problem. It is shown that a properly designed, feedback controller performs well, as compared with a time optimal open loop controller with special preshaping for performance robustness. Also included are two separate papers by the same authors on this subject.

Wie, Bong↗

Dynamic modeling and control of a solid-sorbent CO{sub 2} capture process with two-stage bubbling fluidized bed adsorber reactor

Solid-sorbent-based CO{sub 2} capture processes have strong potential for reducing the overall energy penalty for post-combustion capture from the flue gas of a conventional pulverized coal power plant. However, the commercial success of this technology is contingent upon it operating over a wide range of capture rates, transient events, malfunctions, and disturbances, as well as under uncertainties. To study these operational aspects, a dynamic model of a solid-sorbent-based CO{sub 2} capture process has been developed. In this work, a one-dimensional (1D), non-isothermal, dynamic model of a two-stage bubbling fluidized bed (BFB) adsorber-reactor system with overflow-type weir configuration has been developed in Aspen Custom Modeler (ACM). The physical and chemical properties of the sorbent used in this study are based on a sorbent (32D) developed at National Energy Technology Laboratory (NETL). Each BFB is divided into bubble, emulsion, and cloud-wake regions with the assumptions that the bubble region is free of solids while both gas and solid phases coexist in the emulsion and cloud-wake regions. The BFB dynamic model includes 1D partial differential equations (PDEs) for mass and energy balances, along with comprehensive reaction kinetics. In addition to the two BFB models, the adsorber-reactor system includes 1D PDE-based dynamic models of the downcomer and outlet hopper, as well as models of distributors, control valves, and other pressure-drop devices. Consistent boundary and initial conditions are considered for simulating the dynamic model. Equipment items are sized and appropriate heat transfer options, wherever needed, are provided. Finally, a valid pressure-flow network is developed and a lower-level control system is designed. Using ACM, the transient responses of various process variables such as flue gas and sorbent temperatures, overall CO{sub 2} capture, level of solids in the downcomer and hopper have been studied by simulating typical disturbances such as change in the temperature, flowrate, and composition of the flue gas. To maintain the overall CO{sub 2} capture at a desired level in face of the typical disturbances, two control strategies were considered–a proportional-integral-derivative (PID)-based feedback control strategy and a feedforward-augmented feedback control strategy. Dynamic simulation results show that both the strategies result in unacceptable overshoot/undershoot and a long settling time. To improve the control system performance, a linear model predictive controller (LMPC) is designed. In summary, the overall results illustrate how optimizing the operation and control of carbon capture systems can have a significant impact on the extent and the rate at which commercial-scale capture processes will be scaled-up, deployed, and used in the years to come.

20 FOSSIL-FUELED POWER PLANTS↗

Real time thermal imaging of high temperature semiconductor melts

A real time thermal imaging system with temperature resolution better than + or - 1 C and spatial resolution of better than 0.5 mm was developed and applied to the analysis of melt surface thermal field distributions in both Czochralski and liquid encapsulated Czochralski (LEC) growth configurations. The melt is viewed in near normal incidence by a high resolution charge coupled device camera to which is attached a very narrow bandpass filter. The resulting image is digitized and processed using a pipelined pixel processor operating at an effective 40 million operations per second thus permitting real time high frequency spatial and temporal filtering of the high temperature scene. A multi-pixel averaging algorithm was developed which permits localized, low noise sensing of temperature variations at any location in the hot zone as a function of time. This signial is used to implement initial elements of a feedforward growth control scheme which is aimed at reducing disturbances to the melt caused by the batch nature of the growth process. The effect of magnetic melt stabilization on radial melt temperature distributions was measured using this technique. Problems associated with residual internal reflections and non-optimized path geometry are discussed.

Wargo, Michael J.↗

Lidar-Enhanced Wind Turbine Control: Past, Present, and Future

The main challenges in harvesting energy from the wind arise from the unknown incoming turbulent wind field. Balancing the competing interests of reduction in structural loads and increasing energy production is the goal of a wind turbine controller to reduce the cost of producing wind energy. Conventional wind turbines use feedback methods to optimize these goals, reacting to wind disturbances after they have already impacted the wind turbine. Lidar sensors offer a means to provide additional inputs to a wind turbine controller, enabling new techniques to improve control methods, allowing a controller to actuate a wind turbine in anticipation of an incoming wind disturbance. This paper will look at the development of lidar-enhanced controls and how they have been used for various turbine load reductions with pitch actuation, as well as increased energy production with improved yaw control. Ongoing work will also be discussed to show that combining pitch and torque control using feedforward nonlinear model predictive control can lead to both reduced loads and increased energy production. Future work is also proposed on extending individual wind turbine controls to the wind plant level and determining how lidars can be used for control methods to further lower the cost of wind energy by minimizing wake impacts in a wind farm.

17 WIND ENERGY↗

Reinforcement Learning for In-Spill Optimization of the Mu2e Resonant Extraction: Compensating Non-Stationarity

We present design considerations and challenges for the fast machine learning component of a third-order resonant beam extraction regulation system being commissioned to deliver steady beam rates to the mu2e experiment at Fermilab. Dedicated quadrupoles drive the tune toward the 29/3 resonance each spill, extracting beam at kV multiwire septa. The overall Spill Regulation System consists of (1) a “slow” process using ~100-spill averages to adjust the base quad ramp infrequently, (2) a feedforward harmonic content compensator, and (3) the “fast” ML agent reacting during each ongoing spill with on-the-fly additive corrections to the sum of (1) and (2). We have demonstrated improved beam-rate steadying for a fast ML agent compared to a PID controller using a quasi-physical spill simulation, and demonstrated distillation of that simulation into a predictive surrogate model. Current work includes a data-and-training pipeline to generate data-aware surrogates with real-world dynamics, even as the dynamics shift unpredictably. The surrogates are to act as RL environments against which to train our fast ML control agents before deploying them on FPGA in the live system. Further current efforts focus on modeling and controlling beam loss around the storage ring, understanding additional available hardware inputs to the model, and the interplay of these with beam-steadying performance.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗