SEARCH · Engineering Papers

Results for “Differentiable predictive control”

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.

SourceResearch connectionFrom yearThrough year

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

At least 37 records · Page 2

All sources NASA NTRS DOE OSTI

NASA NTRSDate not supplied

COBRA-DDP: Trajectory Generation and Collision Avoidance Augmentations for eVTOL Vehicles

This paper presents a receding horizon model predictive control variation of the combined Bernstein polynomial optimal reciprocal collision avoidance (ORCA) differential dynamic programming (COBRA-DDP) algorithm for AAM vehicles. Collision avoidance in combination with effective trajectory replanning are expected to be core components of AAM vehicles operating within a crowded airspace. This environment necessitates the use of real-time trajectory planning algorithms that are capable of planning around large amounts of stationary and moving obstacles. Previous work on COBRA-DDP demonstrated the capability of the algorithm to produce dynamically feasible trajectories for AAM vehicles and general collision avoidance. This paper improves upon the previous work by increasing the number of stationary and moving obstacles, implementing a variation of COBRA-DDP that lends itself to real-time application. These advancements are demonstrated on a vertical takeoff and landing (VTOL) vehicle simulation with highly nonlinear vehicle dynamics.

COBRA-DDP↗

NASA NTRSDate not supplied

COBRA-DDP: Trajectory Generation and Collision Avoidance Augmentations for eVTOL Vehicles

COBRA-DDP↗

DOE OSTI2026

Differentiable multiphase flow model for physics-informed machine learning in reservoir pressure management

Accurate subsurface reservoir pressure control is extremely challenging due to geological heterogeneity and multiphase fluid-flow dynamics. Predicting behavior in this setting relies on high-fidelity physics-based simulations that are computationally expensive. Yet, the uncertain, heterogeneous properties that control these flows make it necessary to perform many of these expensive simulations, which is often prohibitive. To address these challenges, we introduce a physics-informed machine learning workflow that couples a fully differentiable multiphase flow simulator, which is implemented in the DPFEHM framework with a convolutional neural network (CNN). The CNN learns to predict fluid extraction rates from heterogeneous permeability fields to enforce pressure limits at critical reservoir locations. By incorporating transient multiphase flow physics into the training process, our method enables more practical and accurate predictions for realistic injection-extraction scenarios compared to previous works. To speed up training, we pretrain the model on single-phase, steady-state simulations and then finetune it on full multiphase scenarios, which dramatically reduces the computational cost. We demonstrate that high-accuracy training can be achieved with fewer than three thousand full-physics multiphase flow simulations – compared to previous estimates requiring up to ten million. This drastic reduction in the number of simulations is achieved by leveraging transfer learning from much less expensive single phase simulations.

25 ENERGY STORAGE↗

NASA NTRS2006

Model Predictive Optimal Control of a Time-Delay Distributed-Parameter Systems

This paper presents an optimal control method for a class of distributed-parameter systems governed by first order, quasilinear hyperbolic partial differential equations that arise in many physical systems. Such systems are characterized by time delays since information is transported from one state to another by wave propagation. A general closed-loop hyperbolic transport model is controlled by a boundary control embedded in a periodic boundary condition. The boundary control is subject to a nonlinear differential equation constraint that models actuator dynamics of the system. The hyperbolic equation is thus coupled with the ordinary differential equation via the boundary condition. Optimality of this coupled system is investigated using variational principles to seek an adjoint formulation of the optimal control problem. The results are then applied to implement a model predictive control design for a wind tunnel to eliminate a transport delay effect that causes a poor Mach number regulation.

Nguyen, Nhan↗

NASA NTRS2006

Adjoint Method and Predictive Control for 1-D Flow in NASA Ames 11-Foot Transonic Wind Tunnel

This paper describes a modeling method and a new optimal control approach to investigate a Mach number control problem for the NASA Ames 11-Foot Transonic Wind Tunnel. The flow in the wind tunnel is modeled by the 1-D unsteady Euler equations whose boundary conditions prescribe a controlling action by a compressor. The boundary control inputs to the compressor are in turn controlled by a drive motor system and an inlet guide vane system whose dynamics are modeled by ordinary differential equations. The resulting Euler equations are thus coupled to the ordinary differential equations via the boundary conditions. Optimality conditions are established by an adjoint method and are used to develop a model predictive linear-quadratic optimal control for regulating the Mach number due to a test model disturbance during a continuous pitch

Nguyen, Nhan↗

NASA NTRS1981

Experimental validation of a numerical model predicting the charging characteristics of Teflon and Kapton under electron beam irradiation

The effect differential charging of spacecraft thermal control surfaces is assessed by studying the dynamics of the charging process. A program to experimentally validate a computer model of the charging process was established. Time resolved measurements of the surface potential were obtained for samples of Kapton and Teflon irradiated with a monoenergetic electron beam. Results indicate that the computer model and experimental measurements agree well and that for Teflon, secondary emission is the governing factor. Experimental data indicate that bulk conductivities play a significant role in the charging of Kapton.

Hazelton, R. C.↗

NASA NTRS1983

Large space structure model reduction and control system design based upon actuator and sensor influence functions

A model reduction procedure based on aggregation with respect to sensor and actuator influences rather than modes is presented for large systems of coupled second-order differential equations. Perturbation expressions which can predict the effects of spillover on both the aggregated and residual states are derived. These expressions lead to the development of control system design constraints which are sufficient to guarantee, to within the validity of the perturbations, that the residual states are not destabilized by control systems designed from the reduced model. A numerical example is provided to illustrate the application of the aggregation and control system design method.

Yam, Y.↗

NASA NTRS1974

Vehicle control for fuel optimization

The problem of predicting the minimum fuel trajectory for a six degree of freedom vehicle which has a motion characterized by the first order differential equations of translational and rotational dynamics is considered. The thrust direction and center of gravity of the vehicle are assumed to be fixed with respect to the vehicle. Thrust magnitude and the control moment are used as control variables and appear linearly in the equation of motion. Pontryagin's Maximum Principle is used to solve the variational problem. With this formulation, the extremal controls are bang-bang with the exception of the singular case.

Dannenberg, K. D.↗

NASA NTRS1982

Pilot model hypothesis testing

The aircraft control time history predicted by the optimal control pilot model and actual pilot tracking data obtained from NASA Langley's differential maneuvering simulator (DMS) are analyzed. The analysis is performed using a hypothesis testing scheme modified to allow for changes in the true hypothesis. A finite number of pilot models, each with different hypothesized internal model representations of the aircraft dynamics, are constructed. The hypothesis testing scheme determines the relative probability that each pilot model best matches the DMS data. By observing the changes in probabilities, it is possible to determine when the pilot changes control strategy and which hypothesized pilot model best represent's the pilot's control behavior.

Broussard, J. R.↗

NASA NTRS2001

The Role of Convection and Growth Competition in Phase Selection in Microgravity: Controlled Convection in the Containerless Processing of Steel Alloys

Containerless processing using electromagnetic levitation (EML) is a powerful technique in the investigation of reactive molten metal systems. On ground, the power required to overcome the weight of the sample is sufficient to cause significant heating and induce substantial melt convection. In microgravity, the heating and positioning fields may be decoupled and the field strength may be varied to achieve the desired level of convection within the limits set by the geometry of the levitation coil and the sample size. From high-speed digital images of the double recalescence behavior of Fe-Cr-Ni alloys in ground-based testing and in reduced-gravity aboard the NASA KC-135 parabolic aircraft, we have shown that phase selection can be predicted based on a growth competition model. An important parameter in this model is the delay time between primary nucleation and subsequent nucleation of the stable solid within the liquid/metastable solid array. This delay time is a strong function of composition and a weak function of the undercooling of the melt below the metastable liquidus. From the results obtained during the first Microgravity Sciences Laboratory (MSL-1) mission, we also know that convection may significantly influence the delay time, especially at low undercoolings. Currently, it is unclear what mechanism controls the formation of a heterogeneous site that allows nucleation of the austenitic phase on the pre-existing ferrite skeleton. By examining the behavior of the delay time under different convective conditions, we hypothesize that we can differentiate between several of these mechanisms to gain an understanding of how to control microstructural. evolution. We will anchor these predictions by examining samples quenched at different times following primary recalescence in microgravity. A second important parameter in the growth competition model is the identification of the growth rate of the stable phase into the semi-solid array that formed during primary recalescence. Current dendritic growth theory is inadequate in predicting solidification behavior under these conditions as metallographic analyses show that stable phase growth proceeds along the interface between the metastable solid and residual liquid. Since growth velocity is independent of the initial undercooling relative to the metastable liquidus, we hypothesize that purely thermal effects can be separated from other important growth model parameters by careful selection of the liquid composition in a ternary system.

Matson, D. M.↗

NASA NTRS1990

On the nature of the interaction between structures and actuators in vibration suppression

This lecture presents a summary of work on the analysis of the interaction between a structure, an actuator used to suppress the vibration of the structure, and the control law implemented by the actuator. Two control applications are considered. First, a proof-mass actuator with experimentally verified dynamics, capable of being used in a space structure configuration is examined. This is connected to a cantilevered beam modeled as a single degree of freedom system. Secondly, an electric motor, again with experimentally verified dynamics, is used to slew a beam modeled by a partial differential equation. Both the experimental and numerical configuration predict the presence of potential instabilities in system performance if proper consideration is not given to interactions between the control law, the structure and the actuator. In addition, an understanding of the interaction is shown to greatly effect performance. In particular it is shown that judicious choice of actuator parameters greatly improves closed loop performance.

Inman, Daniel J.↗

NASA NTRS1987

Flexible system model reduction and control system design based upon actuator and sensor influence functions

A model reduction technique based on aggregation with respect to sensor and actuator influence functions rather than modes is presented for large systems of coupled second-order differential equations. Perturbation expressions which can predict the effects of spillover on both the reduced-order plant model and the neglected plant model are derived. For the special case of collocated actuators and sensors, these expressions lead to the derivation of constraints on the controller gains that are, given the validity of the perturbation technique, sufficient to guarantee the stability of the closed-loop system. A case study demonstrates the derivation of stabilizing controllers based on the present technique. The use of control and observation synthesis in modifying the dimension of the reduced-order plant model is also discussed. A numerical example is provided for illustration.

Yam, Yeung↗

NASA NTRS1980

Development and validation of a hybrid-computer simulator for a transonic cryogenic wind tunnel

A study was undertaken to model the cryogenic wind tunnel process, to validate the model by the use of experimental data from the Langley 0.3 Meter Transonic Cryogenic Tunnel, and to construct an interactive simulator of the cryogenic tunnel using the validated model. Additionally, this model was used for designing closed loop feedback control laws for regulation of temperature and pressure in the 0.3 meter cryogenic tunnel. The global mathematical model of the cryogenic tunnel that were developed consists of coupled, nonlinear differential governing equations based on an energy state concept of the physical cryogenic phenomena. Process equations and comparisons between actual tunnel responses and computer simulation predictions were examined. Also included are the control laws and simulator responses obtained using the feedback schemes for closed loop control of temperature and pressure were also included.

Thibodeaux, J. J.↗

DOE OSTI2023

Long–short-term memory encoder–decoder with regularized hidden dynamics for fault detection in industrial processes

The ability of recurrent neural networks (RNN) to model nonlinear dynamics of high dimensional process data has enabled data-driven RNN-based fault detection algorithms. Previous studies have focused on detecting faults by identifying the discrepancies in data distribution between the faulty and normal data, as reflected in prediction errors generated by RNN models. However, in industrial processes, variations in data distribution can also result from changes in normal control setpoints and compensatory control adjustments in response to disturbances, making it hard to differentiate between normal and faulty conditions. This paper proposes a fault detection method utilizing a long short-term memory (LSTM) encoder–decoder structure with regularized hidden dynamics and reversible instance normalization (RevIN) to compactly represent high-dimensional measurements for effective monitoring. During training, the hidden states of the model are regularized to form a low-dimensional latent space representation of the original multivariate time series data. As a result, the prediction errors of the latent states can be used to monitor the abnormal dynamic variations, while the reconstruction errors of the measured variables are used to monitor the abnormal static variations. Furthermore, the proposed indices can reflect operating conditions, even when the distribution of test data changes, which helps distinguish faults from normal adjustments and disturbances that controllers can settle. Here, data from numerical simulation and the Tennessee Eastman process are used to illustrate the effectiveness of the proposed fault detection method.

42 ENGINEERING↗

NASA NTRS1972

Low speed aerodynamic characteristics of a large scale model with a thin, highly swept, 2.67 aspect ratio wing having a cranked leading edge

The low speed aerodynamic characteristics of a large-scale model intended to represent advanced fixed-wing fighters have been investigated in the Ames 40 by 80 foot wind tunnel. The model possessed positive static longitudinal stability to nearly 28 deg angle of attack, the maximum tested, both with and without leading-edge flaps deflected. Lateral control with differentially deflected ailerons and a right wing spoiler simultaneously deployed as combined controls was only slightly greater than that with the differentially deflected ailerons deployed as a separate control without the spoiler. Measured lift and drag were in close agreement with that predicted by theory to about 14 deg angle of attack, including the prediction of lift due to 30 deg of trailing edge flap deflection. Estimated takeoff performance of an aircraft 4/3 the scale of the test model showed takeoff distances of less than 2000 ft.

Giulianetti, D. J.↗

NASA NTRS2020

Squash-Box Feasibility Driven Differential Dynamic Programming

Recently, Differential Dynamic Programming (DDP) and other similar algorithms have become the solvers of choice when performing non-linear Model Predictive Control (nMPC) with modern robotic devices. The reason is that they have a lower computational cost per iteration when compared with off-the-shelf Non-Linear Programming (NLP) solvers, which enables its online operation. However, they cannot handle constraints, and are known to have poor convergence capabilities. In this paper, we propose a method to solve the optimal control problem with control bounds through a squashing function (i.e., a sigmoid, which is bounded by construction). It has been shown that a naive use of squashing functions damage the convergence rate. To tackle this, we first propose to add a quadratic barrier that avoids the difficulty of the plateau produced by the sigmoid. Second, we add an outer loop that adapts both the sigmoid and the barrier; it makes the optimal control problem with the squashing function converge to the original control-bounded problem. To validate our method, we present simulation results for different types of platforms including a multi-rotor, a biped, a quadruped and a humanoid robot.

Navarro, Angel Santamaria↗

DOE OSTI2024

Power System Frequency Dynamics Modeling, State Estimation, and Control using Neural Ordinary Differential Equations (NODEs) and Soft Actor-Critic (SAC) Machine Learning Approaches

With the global energy transition of the electric power system, grid control, supervision, and protection is becoming more challenging. With the increasing integration of renewable energy sources (RES), the system dynamics are changing, causing traditional power system dynamic modeling with swing equation-based modeling approaches to fail. Additionally, the converter-dominated power grid is decreasing the system inertia, making the power system more fragile to the frequency swings. This paper first investigates and compares the application of a model-based Kalman filter state estimation approach with (i) a model-free machine learning approach --- neural ordinary differential equations (NODEs) --- and (ii) a data-driven system identification (SysId) approach to model and infer critical state values of the power system frequency dynamics. Then a model predictive control (MPC) framework is compared to a model-free Soft Actor-Critic (SAC) reinforcement learning (RL) control algorithm in providing efficient fast frequency response (FFR) to the power system frequency dynamics. The approaches are compared in terms of their performance goals as well as their per-timestep computational efficiency. Furthermore, the comparative study for state estimation shows that for the model-free requirement, both NODEs and SysId can provide accurate state estimates; however, with increasing model complexity, NODEs can be a better choice for model identification. Similarly, the results from the FFR comparative study show that the SAC RL-based FFR, once trained, outperforms MPC with better control signals and faster computation time, making the SAC RL-based FFR better option for providing FFR to the power system.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS1991

Life extending control: An interdisciplinary engineering thrust

The concept of Life Extending Control (LEC) is introduced. Possible extensions to the cyclic damage prediction approach are presented based on the identification of a model from elementary forms. Several candidate elementary forms are presented. These extensions will result in a continuous or differential form of the damage prediction model. Two possible approaches to the LEC based on the existing cyclic damage prediction method, the measured variables LEC and the estimated variables LEC, are defined. Here, damage estimates or measurements would be used directly in the LEC. A simple hydraulic actuator driven position control system example is used to illustrate the main ideas behind LEC. Results from a simple hydraulic actuator example demonstrate that overall system performance (dynamic plus life) can be maximized by accounting for component damage in the control design.

Lorenzo, Carl F.↗