Adaptive-sparse polynomial chaos expansion for reliability analysis and design of complex engineering systems

This paper presents an adaptive-sparse polynomial chaos expansion (adaptive-sparse PCE) method for performing engineering reliability analysis and design. The proposed method combines three ideas: (i) an adaptive-sparse scheme to build sparse PCE with the minimum number of bivariate basis functions, (ii) a new projection method using dimension reduction techniques to effectively compute the expansion coefficients of system responses, and (iii) an integration of copula to handle nonlinear correlation of input random variables. The proposed method thus has three positive features for reliability analysis and design: (a) there is no need for response sensitivity analysis, (b) it is highly efficient and accurate for reliability analysis and its sensitivity analysis, and (c) it is capable of handling a nonlinear correlation. In addition to the features, an error decomposition scheme for the proposed method is presented to help analyze error sources in probability analysis. Several engineering problems are used to demonstrate the three positive features of the adaptive-sparse PCE method.     

Efficient surrogate method for predicting pavement response to various tire configurations

A computationally efficient surrogate model was developed based on artificial neural networks (ANN) to investigate the effect of the new generation of wide-base tires on pavement responses. Non-uniform tire contact stress measurements were obtained using a stress-in-motion instrument. The measured 3-D contact stresses were applied on two extreme 3-D flexible pavement finite element models representing low-volume (thin) and high-volume (thick) roads. Eleven critical pavement responses were modeled at two different material properties input levels—detailed and simplified—depending on data availability. The results rendered by the ANN surrogate models were highly accurate with average prediction error less than 5 % and R-square values higher than 0.95. In addition, two sensitivity analyses were performed to investigate the variables effect on pavement responses. It was found that the type of tire (wide-base vs. dual tire assembly) is more influential than the inflation pressure on pavement responses. However, the tire inflation pressure seemed to have a significant effect on near-surface responses. The developed models were incorporated into a tool to assist designers and engineers in investigating the effect of the pavement responses of wide-base versus dual tire assembly under typical loading conditions and pavement structures.

     

一种混合神经网络飞行控制方法研究

讨论了一种基于神经网络控制的飞行控制方法。针对复杂非线性系统难以建立精确模型的特点,利用神经网络的任意非线性逼近能力进行控制器设计,首先应用神经网络在线辨识对象逆模型,进行控制系统反馈线性化;接着利用circle theorem（圆定理）设计线性PID鲁棒控制器,控制系统输出跟随系统输入,然后应用神经网路自适应逆方法设计混合控制器,最后以F-8飞机纵向飞行控制模态为研究对象进行仿真。仿真结果表明,该控制方法具有较强的自适应和抗干扰能力。     

On-line system identification of complex systems using Chebyshev neural networks

This paper proposes a computationally efficient artificial neural network (ANN) model for system identification of unknown dynamic nonlinear discrete time systems. A single layer functional link ANN is used for the model where the need of hidden layer is eliminated by expanding the input pattern by Chebyshev polynomials. Thus, creation of nonlinear decision boundaries in the multidimensional input space and approximation of complex nonlinear systems becomes easier. These models are linear in their parameters and nonlinear in the inputs. The recursive least squares method with forgetting factor is used as on-line learning algorithm for parameter updation. The good behaviour of the identification method is tested on Box and Jenkins Gas furnace benchmark identification problem, single input single output (SISO) and multi input multi output (MIMO) discrete time plants. Stability of the identification scheme is also addressed.     

A two-phased SEM-neural network approach for consumer preference analysis

A fundamental task in the design of consumer products is consumer preference analysis. The primary focus of this task is establishing a mapping relationship between product parameters/attributes and consumer preferences. The key to connect the consumer space and the design space are user perceptions of the product. Among the many existing methods, the Structural Equation Model (SEM) is one of the most used methods because it explains the causal relationship between the input and the output variables explicitly. However, the relationship obtained from the conventional SEM is linear, which is usually not the case in practice. Fortunately, the Artificial Neural Network (ANN) provides a new perspective for building nonlinear models because of its nonlinear nature. Therefore, a two-phased SEM-NN approach for consumer preference analysis is introduced for identifying and mapping how product attributes affecting the fulfillment of user perceptions and ultimately their preferences. In this model, the consumer preference analysis is conducted in two phases: influence path construction, and path coefficient revision. The proposed method can reserve the original SEM topology that reflects the causal relationship between variables while using the training algorithm of ANN to obtain more accurate path coefficients. This model could help the designers to identify and map how product attributes affecting the consumer preferences, and to better understand the factors that affect user perceptions and the inner relationships between them. To demonstrate effectiveness of the model, a case study of smartphone is presented. It is shown that the SEM-NN model can make full use of the causal analysis of SEM and the nonlinear nature of ANN and ultimately provides more reliable results of consumer preference analysis.     

Nonlinear dynamic response structural optimization using equivalent static loads

It is well known that nonlinear dynamic response optimization using a conventional optimization algorithm is fairly difficult and expensive for the gradient or non-gradient based optimization methods because many nonlinear dynamic analyses are required. Therefore, it is quite difficult to find practical large scale examples with many design variables and constraints for nonlinear dynamic response structural optimization. The equivalent static loads (ESLs) method is newly proposed and applied to nonlinear dynamic response optimization. The equivalent static loads are defined as the linear static load sets which generate the same response field in linear static analysis as that from nonlinear dynamic analysis. The ESLs are made from the results of nonlinear dynamic analysis and used as external forces in linear static response optimization. Then the same response from nonlinear dynamic analysis can be considered throughout linear static response optimization. The updated design from linear response optimization is used again in nonlinear dynamic analysis and the process proceeds in a cyclic manner until the convergence criteria are satisfied. Several examples are solved to validate the method. The results are compared to those of the conventional method with sensitivity analysis using the finite difference method.     

Robust adaptive DSC of directly input‐coupled nonlinear systems with input unmodeled dynamics and time‐varying output constraints

In this paper, an adaptive robust dynamic surface control is proposed for a class of uncertain nonlinear interconnected systems with time‐varying output constraints and dynamic input and output coupling. The directly coupled inputs and control inputs are both of nonlinear input unmodeled dynamics. To counteract the instable impact of the nonlinear input unmodeled dynamics, normalization signals are designed on the basis of the convergence rates of their Lyapunov functions. With new state variables and control variables being defined, the real control inputs are obtained through solving the equations of intermediate control laws. The time‐varying constraints on output signals are implemented by introducing asymmetric barrier Lyapunov functions. In addition, dynamic signals and decentralized K‐filters are used to deal with the state unmodeled dynamics and to estimate the unmeasurable states, respectively. By the theoretical analysis, the signals in the closed‐loop system are proved to be semi‐globally uniformly ultimately bounded, and the output constraints are guaranteed simultaneously. A numerical example is provided to show the effectiveness of the proposed approach. Copyright © 2017 John Wiley & Sons, Ltd.     

可逆冷带轧机速度张力系统的动态滑模解耦控制

为削弱可逆冷带轧机速度张力系统中各变量间的非线性耦合影响,本文提出了一种基于幂指数趋近律的微分几何动态滑模解耦控制方法.首先,应用微分几何理论,通过非线性状态反馈和坐标变换,实现了可逆冷带轧机速度张力非线性耦合系统的输入/输出动态解耦和线性化.其次,针对解耦后得到的各独立线性子系统,综合考虑可逆冷带轧机速度张力系统的负载扰动、参数摄动和未建模动态等不确定部分的影响,基于幂指数趋近律设计了动态滑模控制器.理论分析表明,所提出的控制方法能够保证闭环系统渐近稳定,并能有效削弱系统抖振.最后,对某1422mm可逆冷带轧机速度张力非线性耦合系统进行仿真,并同其他解耦控制方法相比较,结果验证了所提出方法的有效性.     

Global optimization with multivariate adaptive regression splines.

This paper presents a novel procedure for approximating the global optimum in structural design by combining multivariate adaptive regression splines (MARS) with a response surface methodology (RSM). MARS is a flexible regression technique that uses a modified recursive partitioning strategy to simplify high-dimensional problems into smaller yet highly accurate models. Combining MARS and RSM improves the conventional RSM by addressing highly nonlinear high-dimensional problems that can be simplified into lower dimensions, yet maintains a low computational cost and better interpretability when compared to neural networks and generalized additive models. MARS/RSM is also compared to simulated annealing and genetic algorithms in terms of computational efficiency and accuracy. The MARS/RSM procedure is applied to a set of low-dimensional test functions to demonstrate its convergence and limiting properties.     

Applying input variables selection technique on input weighted support vector machine modeling for BOF endpoint prediction

Basic oxygen furnace (BOF) steelmaking is a complex process and dynamic model is very important for endpoint control. It is usually difficult to build a precise BOF endpoint dynamic model because many input variables affect the endpoint carbon content and temperature. For this problem, two effective variables selection steps: mechanism analysis and mutual information calculation are proposed to choose appropriate input variables according to a variable selection algorithm. Then, the selected inputs are weighted on the basis of mutual information values. Finally, two input weighted support vector machine BOF endpoint dynamic models are constructed to predict endpoint carbon content and temperature. Results show that the variable selection for BOF endpoint prediction model is essential and effective. The complexity and precise of two endpoint prediction models are improved.     

Nonlinear channel equalization for QAM signal constellation usingartificial neural networks

Application of artificial neural networks (ANN's) to adaptive channel equalization in a digital communication system with 4-QAM signal constellation is reported in this paper. A novel computationally efficient single layer functional link ANN (FLANN) is proposed for this purpose. This network has a simple structure in which the nonlinearity is introduced by functional expansion of the input pattern by trigonometric polynomials. Because of input pattern enhancement, the FLANN is capable of forming arbitrarily nonlinear decision boundaries and can perform complex pattern classification tasks. Considering channel equalization as a nonlinear classification problem, the FLANN has been utilized for nonlinear channel equalization. The performance of the FLANN is compared with two other ANN structures [a multilayer perceptron (MLP) and a polynomial perceptron network (PPN)] along with a conventional linear LMS-based equalizer for different linear and nonlinear channel models. The effect of eigenvalue ratio (EVR) of input correlation matrix on the equalizer performance has been studied. The comparison of computational complexity involved for the three ANN structures is also provided.     

高炉炼铁过程多元铁水质量非线性子空间建模及应用

高炉炼铁是一个物理化学反应复杂、多相多场耦合的大滞后、非线性动态系统,其关键工艺指标——铁水质量参数的检测、建模和控制一直是冶金工程和自动控制领域的难题.本文提出一种面向控制的数据驱动高炉炼铁多元铁水质量非线性子空间建模方法.首先,为了提高建模效率和降低计算复杂度,采用数据驱动典型相关性分析与相关性分析相结合的方法提取与铁水质量相关性最强的关键可控变量作为建模的输入变量;同时,为了更好地反映高炉非线性动态特性,将相关输入输出变量的时序和时滞关系在建模过程进行考虑;最后,采用基于最小二乘支持向量机（Least square support vector machine,LS-SVM）的非线性Hammerstein系统子空间辨识方法建立数据驱动的多元铁水质量非线性状态空间模型.同时,将核函数表示的模型非线性特性用多项式函数拟合,在仅损失很小模型精度的前提下大大降低模型的计算复杂度.基于实际数据的工业试验验证了所提建模方法的准确性、有效性和先进性.     

Design sensitivity and reliability-based structural optimization by univariate decomposition

This paper presents a new univariate decomposition method for design sensitivity analysis and reliability-based design optimization of mechanical systems subject to uncertain performance functions in constraints. The method involves a novel univariate approximation of a general multivariate function in the rotated Gaussian space for reliability analysis, analytical sensitivity of failure probability with respect to design variables, and standard gradient-based optimization algorithms. In both reliability and sensitivity analyses, the proposed effort has been reduced to performing multiple one-dimensional integrations. The evaluation of these one-dimensional integrations requires calculating only conditional responses at selected deterministic input determined by sample points and Gauss–Hermite integration points. Numerical results indicate that the proposed method provides accurate and computationally efficient estimates of the sensitivity of failure probability, which leads to accurate design optimization of uncertain mechanical systems.     

Radial basis function neural network for approximation andestimation of nonlinear stochastic dynamic systems

This paper presents a means to approximate the dynamic and static equations of stochastic nonlinear systems and to estimate state variables based on radial basis function neural network (RBFNN). After a nonparametric approximate model of the system is constructed from a priori experiments or simulations, a suboptimal filter is designed based on the upper bound error in approximating the original unknown plant with nonlinear state and output equations. The procedures for both training and state estimation are described along with discussions on approximation error. Nonlinear systems with linear output equations are considered as a special case of the general formulation. Finally, applications of the proposed RBFNN to the state estimation of highly nonlinear systems are presented to demonstrate the performance and effectiveness of the method.     

NARX model based nonlinear dynamic system identification using low complexity neural networks and robust H∞ filter

This paper proposes NARX (nonlinear autoregressive model with exogenous input) model structures with functional expansion of input patterns by using low complexity ANN (artificial neural network) for nonlinear system identification. Chebyshev polynomials, Legendre polynomials, trigonometric expansions using sine and cosine functions as well as wavelet basis functions are used for the functional expansion of input patterns. The past input and output samples are modeled as a nonlinear NARX process and robust H_∞ filter is proposed as the learning algorithm for the neural network to identify the unknown plants. H_∞ filtering approach is based on the state space modeling of model parameters and evaluation of Jacobian matrices. This approach is the robustification of Kalman filter which exhibits robust characteristics and fast convergence properties. Comparison results for different nonlinear dynamic plants with forgetting factor recursive least square (FFRLS) and extended Kalman filter (EKF) algorithms demonstrate the effectiveness of the proposed approach.     

Flood forecasting at the Dadu River in China based on ANN

Mappings of the stimuli effects and the input and output estimates of artificial neural networks (ANN) are obtained via combinations of nonlinear functions. This approach offers the advantages of self‐learning, self‐organization, self‐adaptation, and fault tolerance as well as the potential for use in flood forecasting applications. Furthermore, the ANN technology allows the use of multiple variables in both the input and output layers. This capability is very important for flood calculation because the stage, discharge, and other hydrological variables often are functions of many influential variables. Herein, we propose a flood forecasting system with related application, based on ANN. This method offers better performance and efficiency.     

Adaptive control for a class of second-order nonlinear systems with unknown input nonlinearities

An adaptive controller is developed for a class of second-order nonlinear dynamic systems with input nonlinearities using artificial neural networks (ANN). The unknown input nonlinearities are continuous and monotone and satisfy a sector constraint. In contrast to conventional Lyapunov-based design techniques, an alternative Lyapunov function, which depends on both system states and control input variable, is used for the development of a control law and a learning algorithm. The proposed adaptive controller guarantees the stability of the closed-loop system and convergence of the output tracking error to an adjustable neighbour of the origin.     

Nonlinear dynamic system identification using Chebyshev functionallink artificial neural networks

A computationally efficient artificial neural network (ANN) for the purpose of dynamic nonlinear system identification is proposed. The major drawback of feedforward neural networks, such as multilayer perceptrons (MLPs) trained with the backpropagation (BP) algorithm, is that they require a large amount of computation for learning. We propose a single-layer functional-link ANN (FLANN) in which the need for a hidden layer is eliminated by expanding the input pattern by Chebyshev polynomials. The novelty of this network is that it requires much less computation than that of a MLP. We have shown its effectiveness in the problem of nonlinear dynamic system identification. In the presence of additive Gaussian noise, the performance of the proposed network is found to be similar or superior to that of a MLP. A performance comparison in terms of computational complexity has also been carried out.     

Variable Selection for Artificial Neural Networks with Applications for Stock Price Prediction

We propose a new Artificial neural network (ANN) method where we select a set of variables as input variables to the ANN. The selection is made so that the input variables may be informative for a target variable as much as possible. The proposed method compared favorably with the existing ANN methods when their performances were evaluated based on 488 stocks in S&P500 in terms of prediction accuracy.     

Identification of nonlinear dynamic systems using functional linkartificial neural networks

We have presented an alternate ANN structure called functional link ANN (FLANN) for nonlinear dynamic system identification using the popular backpropagation algorithm. In contrast to a feedforward ANN structure, i.e., a multilayer perceptron (MLP), the FLANN is basically a single layer structure in which nonlinearity is introduced by enhancing the input pattern with nonlinear functional expansion. With proper choice of functional expansion in a FLANN, this network performs as good as and in some cases even better than the MLP structure for the problem of nonlinear system identification.