Andrew J. Meade

Improved Adaptive–Reinforcement Learning Control for Morphing Unmanned Air Vehicles

This paper presents an improved adaptive-reinforcement learning control methodology for the problem of unmanned air vehicle morphing control. The reinforcement learning morphing control function that learns the optimal shape change policy is integrated with an adaptive dynamic inversion control trajectory tracking function. An episodic unsupervised learning simulation using the Q-learning method is developed to replace an earlier and less accurate actor-critic algorithm. Sequential function approximation, a Galerkin-based scattered data approximation scheme, replaces a K-nearest neighbors (KNN) method and is used to generalize the learning from previously experienced quantized states and actions to the continuous state-action space, all of which may not have been experienced before. The improved method showed smaller errors and improved learning of the optimal shape compared to the KNN.

Numerical investigation of nanoparticle-assisted laser-induced interstitial thermotherapy toward tumor and cancer treatments

In this work, we numerically investigated nanoparticle-assisted laser-induced interstitial thermotherapy for tumor/cancer treatments. The goal of the study was to investigate the therapeutic effects of treatment conditions including laser wavelength, power, exposure time, concentrations of tailored nanoparticles, and optical/thermal properties of the tissue that is under treatment. It was found that using absorbing preferential nanoparticles as the photothermal agent weakens fluence rate distributions in terms of lowering fluence rate peaks and reducing laser penetration depths. However, the local enhancement in laser photon absorption induced by nanoparticles is so significant that the reduced fluence rate will be balanced out, and the eventual medical hyperthermia is greatly prompted by using nanoparticles. Also, the results of numerical simulations indicated that with constant laser illumination, an increase in nanoparticle concentration beyond a certain range has only an insignificant impact on hyperthermia.

Approximation properties of local bases assembled from neural network transfer functions

The adaptive data-driven emulation and control of mechanical systems are popular applications of artificial neural networks in engineering. However, multilayer perceptron training is an ill-posed nonlinear optimization problem. This paper explores a method to constrain network parameters so that conventional computational techniques for function approximation can be used during training. This was accomplished by forming local basis functions which provide accurate approximation and stable evaluation of the network parameters. It is noted that this approach is quite general and does not violate the principles of network architecture. By employing the concept of shift-invariant subspaces, this approach yields a new and more robust error condition for feedforward artificial neural networks and allows one to both characterize and control the accuracy of the local bases formed. The two methods used are: (1) adding bases while altering their shape and keeping their spacing constant, and (2) adding bases while altering their shape and decreasing their spacing in a coupled fashion. Numerical examples demonstrate the usefulness of the proposed approximation of functions and their derivatives.

Enhancement of heat transfer in square helical ducts

Abstract A three-dimensional numerical investigation of steady laminar flow and heat transfer is undertaken to determine the developing as well as fully developed temperature fields. Physical interpretation is given for the enhancement of the heat transfer coefficient at the thermal entrance region and the overall increase in heat transfer in helical ducts compared to straight ducts. The variation of peripherally averaged Nusselt number is studied for the constant wall temperature boundary condition. Detailed analysis is given for the peripheral variation of Nusselt number and the temperature field at different cross-sectional planes. The effect of the Prandtl number on the temperature field is also studied. The problem is solved using a segregated solution approach which reduces the total computer memory requirements.

Low Dean Number Convective Heat Transfer in Helical Ducts of Rectangular Cross Section

Flow in a torroidal duct is characterized by increased convective heat transfer and friction compared to a straight duct of the same cross section. In this paper the importance of the nonplanarity (torsion) of a helical duct with rectangular cross section is investigated. A previously determined low Dean number velocity solution is used in the decoupled energy equation for the hydrodynamically fully developed, thermally developing case. Torsion, known to increase the friction factor, is found to cause a decrease in the fully developed Nusselt number compared to pure torroidal flow. Therefore, it is recommended that torsion be minimized to enhance heat transfer.

Series solution of low Dean and Germano number flows in helical rectangular ducts

Abstract The flow in a helical duct with rectangular cross-section is analyzed. A series solution based on curvature and torsion is introduced. The components of the series are determined analytically using appropriate eigenfunction expansions. The resulting solution is limited to flows in the low Dean number, low Germano number regime. An analytical friction factor relation is established and compared with previous numerically determined correlations.

Application of Scattered Data Approximation to a Rotorcraft Health Monitoring Problem

An adaptive and matrix–free scheme has been developed for interpolating and approximating sparse multi–dimensional scattered data and has been applied to a timeseries problem in rotorcraft. The Sequential Function Approximation (SFA) method is based on a sequential Galerkin approach to artificial neural networks and requires neither ad–hoc parameters for the user to tune, nor rescaling of the inputs. It is linear in storage with respect to the number of samples. The SFA method has been used to model and extrapolate the time series data from a critical temperature sensor in a 1/4 scale model of the V-22 Osprey. The SFA regression model, constructed with radial basis functions, has also been used satisfactorily to evaluate the sensitivity to 74 system health and safety–of–flight parameters during a series of wind–tunnel tests. An upper bound on the error convergence rate that is exponential and does not explicitly depend on the dimensionality of the approximation was derived and confirmed for the time series data.

Learning Air-Data Parameters for Flush Air Data Sensing Systems

An adaptive scattered data approximation scheme was developed to calibrate the Flush Air Data System (FADS) of a surface vessel. An array of pressure sensors were mounted flush with the deckhouse periphery and the airdata parameters were extracted from the pressure measurements. The developed Galerkin derived selfadaptive greedy function approximation scheme gave reliable and robust surrogates for predicting wind speed and direction. The resulting surrogates were also used to evaluate the sensitivity to each of the flush mounted pressure sensor. Fault tolerance of the proposed surrogates were also studied with respect to pressure sensor failure.

Numerical solution of a calculus of variations problem using the feedforward neural network architecture

Abstract It is demonstrated, through theory and numerical example, how it is possible to construct directly and noniteratively a feedforward neural network to solve a calculus of variations problem. The method, using the piecewise linear and cubic sigmoid transfer functions, is linear in storage and processing time. The L 2 norm of the network approximation error decreases quadratically with the piecewise linear transfer function and quartically with the piecewise cubic sigmoid as the number of hidden layer neurons increases. The construction requires imposing certain constraints on the values of the input, bias, and output weights, and the attribution of certain roles to each of these parameters. All results presented used the piecewise linear and cubic sigmoid transfer functions. However, the noniterative approach should also be applicable to the use of hyperbolic tangents and radial basis functions.

Adaptive Algorithm for Classifying Launch and Recovery of the HH-60H Seahawk

The pilot ratings from at-sea dynamic interface launch/recovery flight tests of the HH-60H Seahawk helicopter have been modeled satisfactorily by an adaptive, nonparametric classification scheme. Results were compared against those from three popular pattern classification tools and confirmed the authors’ choice of algorithm. The resulting pilot rating classification model has also been used to evaluate the sensitivity to the 13 dependent parameters that include ship, aircraft, and environmental conditions.