Yau Shu Wong

Neural Network Approach for Nonlinear Aeroelastic Analysis

A new approach is proposed, based on the use of artificial neural networks, for predicting nonlinear aeroelastic oscillations. Our objective is to reconstruct the asymptotic state of the nonlinear behavior of an aeroelastic model when only a limited segment of the transient data is known. An original neural network architecture is proposed and is used to predict the nonlinear motions of an aeroelastic system modeling a self-excited two-degree-of-freedom airfoil oscillating in pitch and plunge. When a segment of the transient state of the given signal is used for training, the neural network is capable of correctly predicting the corresponding limit-cycle oscillations, damped oscillations, or unstable divergent oscillations. The network training set consists of numerically generated data or data obtained from a wind-tunnel experiment. A neural network used in conjunction with a wavelet decomposition is presented, which proves to be capable of extracting the values of the damping coefficients and frequencies from the predicted signal. Neural networks, thus, prove to be useful tools in nonlinear aeroelastic analysis.

Nested Monte Carlo EM algorithm for switching state-space models

Switching state-space models have been widely used in many applications arising from science, engineering, economic, and medical research. In this paper, we present a Monte Carlo Expectation Maximization (MCEM) algorithm for learning the parameters and classifying the states of a state-space model with a Markov switching. A stochastic implementation based on the Gibbs sampler is introduced in the expectation step of the MCEM algorithm. We study the asymptotic properties of the proposed algorithm, and we also describe how a nesting approach and the Rao-Blackwellized forms can be employed to accelerate the rate of convergence of the MCEM algorithm. Finally, the performance and the effectiveness of the proposed method are demonstrated by applications to simulated and physiological experimental data.

AN IMPROVED NEURAL NETWORK MODEL FOR NONLINEAR AEROELASTIC ANALYSIS

An improved neural network-based method for predicting nonlinear oscillations in the aeroelastic response is presented. An articial neural network is trained using the limited available information of a short transient data set, and the asymptotic state of the signal is reconstructed by a multi-step (or recursive) prediction process. An enhanced two-layer feedforward neural network with features that control the propagation of the prediction errors is designed. Methods for consistently choosing the number of network inputs and of neurons in the hidden layer for a given application are reported. The proposed predictor has been applied to wind-tunnel experimental data that model an oscillating airfoil with polynomial restoring forces, as well as to signals generated numerically by solving a dieren tial system that models a self-excited two-degree-of-freedom airfoil oscillating in pitch and plunge.

A parallel alternating direction implicit preconditioning method

Abstract The alternating direction implicit (ADI) iterative method is an efficient iterative method to solve systems of linear equations due to its extremely fast convergence. The ADI method has also been used successfully as a preconditioner in some other iterative methods, such as the preconditioned conjugate gradient. In this paper a parallel algorithm for the ADI preconditioning is proposed. In this algorithm, several steps of the ADI iteration are computed simultaneously. This means that several tridiagonal systems that are traditionally solved sequentially are now solved concurrently. The high performance of this algorithm is achieved by increasing the degree of parallelism and reducing memory contention. The algorithm can easily be implemented in a multiprocessor architecture. Experiments have been conducted on the Myrias SPS-2 computer with 64 processors and good performance of this algorithm is observed.

Uncertainty investigations in nonlinear aeroelastic systems

In this paper, the stochastic collocation method (SCM) is applied to investigate the nonlinear behavior of an aeroelastic system with uncertainties in the system parameter and the initial condition. Numerical case studies for problems with uncertainties are carried out. In particular, the performance of the SCM is compared with solutions based on other computational techniques such as Monte Carlo simulation, Wiener chaos expansion and wavelet chaos expansion. From the computational results, we conclude that the SCM is an effective tool to study a nonlinear aeroelastic system with random parameters.

Nonlinear statistical approach for aeroelastic response prediction

Aging aircraft and combat aircraft that carry heavy external stores potentially face problems arising from nonlinearities in structure. An expert data mining system is proposed that is capable of predicting the asymptotic behavior of an aeroelastic system with structural nonlinearities represented by polynomial restoring forces or freeplay models. The input is represented only by a limited set of transient data. The output provides a long-term nonlinear aeroelastic response, and the prediction is made when certain rule-based reasoning conditions are satisfied. An attractive feature of this new approach is that no information about the system parameters is needed. In the prediction module, we propose two methods, based on nonlinear time series models and the unscented Kalman filter. To our knowledge, these approaches have not been reported so far for predicting the long-term nonlinear aeroelastic responses. Compared with the classical extended Kalman filter, the unscented filter does not require differentiability and can be applied to nonlinear aeroelastic models with freeplay and hysteresis. The performances of the expert data mining system are demonstrated for simulated data and wind-tunnel experimental aeroelastic data resulting from a two-degree-of-freedom airfoil oscillating in pitch and plunge.

A Nonlinear Statistical Approach for Aeroelastic Response Prediction

Aging aircraft and combat aircraft that carry heavy external stores potentially face problems arising from nonlinearities in structure. An expert data mining system is proposed that is capable of predicting the asymptotic behavior of an aeroelastic system with structural nonlinearities represented by polynomial restoring forces or freeplay models. The input is represented only by a limited set of transient data. The output provides a long-term nonlinear aeroelastic response, and the prediction is made when certain rule-based reasoning conditions are satisfied. An attractive feature of this new approach is that no information about the system parameters is needed. In the prediction module, we propose two methods, based on nonlinear time series models and the unscented Kalman filter. To our knowledge, these approaches have not been reported so far for predicting the long-term nonlinear aeroelastic responses. Compared with the classical extended Kalman filter, the unscented filter does not require differentiability and can be applied to nonlinear aeroelastic models with freeplay and hysteresis. The performances of the expert data mining system are demonstrated for simulated data and wind-tunnel experimental aeroelastic data resulting from a two-degree-of-freedom airfoil oscillating in pitch and plunge.

Machine learning algorithms for mode-of-action classification in toxicity assessment

BackgroundReal Time Cell Analysis (RTCA) technology is used to monitor cellular changes continuously over the entire exposure period. Combining with different testing concentrations, the profiles have potential in probing the mode of action (MOA) of the testing substances.ResultsIn this paper, we present machine learning approaches for MOA assessment. Computational tools based on artificial neural network (ANN) and support vector machine (SVM) are developed to analyze the time-concentration response curves (TCRCs) of human cell lines responding to tested chemicals. The techniques are capable of learning data from given TCRCs with known MOA information and then making MOA classification for the unknown toxicity. A novel data processing step based on wavelet transform is introduced to extract important features from the original TCRC data. From the dose response curves, time interval leading to higher classification success rate can be selected as input to enhance the performance of the machine learning algorithm. This is particularly helpful when handling cases with limited and imbalanced data. The validation of the proposed method is demonstrated by the supervised learning algorithm applied to the exposure data of HepG2 cell line to 63 chemicals with 11 concentrations in each test case. Classification success rate in the range of 85 to 95 % are obtained using SVM for MOA classification with two clusters to cases up to four clusters.ConclusionsWavelet transform is capable of capturing important features of TCRCs for MOA classification. The proposed SVM scheme incorporated with wavelet transform has a great potential for large scale MOA classification and high-through output chemical screening.

Modeling and simulation for toxicity assessment

The effect of various toxicants on growth/death and morphology of human cells is investigated using the xCELLigence Real-Time Cell Analysis High Troughput in vitro assay. The cell index is measured as a proxy for the number of cells, and for each test substance in each cell line, time-dependent concentration response curves (TCRCs) are generated. In this paper we propose a mathematical model to study the effect of toxicants with various initial concentrations on the cell index. This model is based on the logistic equation and linear kinetics. We consider a three dimensional system of differential equations with variables corresponding to the cell index, the intracellular concentration of toxicant, and the extracellular concentration of toxicant. To efficiently estimate the models parameters, we design an Expectation Maximization algorithm. The model is validated by showing that it accurately represents the information provided by the TCRCs recorded after the experiments. Using stability analysis and numerical simulations, we determine the lowest concentration of toxin that can kill the cells. This information can be used to better design experimental studies for cytotoxicity profiling assessment.

Symplectic Numerical Schemes for Stochastic Systems Preserving Hamiltonian Functions

We present high-order symplectic schemes for stochastic Hamiltonian systems preserving Hamiltonian functions. The approach is based on the generating function method, and we show that for the stochastic Hamiltonian systems, the coefficients of the generating function are invariant under permutations. As a consequence, the high-order symplectic schemes have a simpler form than the explicit Taylor expansion schemes with the same order. Moreover, we demonstrate numerically that the symplectic schemes are effective for long time simulations.