Optimal Neural Network Model for Characterization of Process‐Induced Damage in Drilling Carbon Fiber Reinforced Epoxy Composites

In this paper, a method for robust design of a neural network (NN) model for prediction of delamination (Da), damage width (Dw), and hole surface roughness (Ra) during drilling in carbon fiber reinforced epoxy (BMS 8‐256) is presented. This method is based on a parametric analysis of neural network models using a design of experiments approach. The effects of number of neurons (N), hidden layers (L), activation function (AF), and learning algorithm (LA) on the mean square error (MSE) of model prediction are quantified. Using the aforementioned method, a robust NN model was developed that predicted process‐induced damage with high accuracy.     

A Metamodeling Method Based on Support Vector Regression for Robust Optimization

Metamodeling techniques have been used in robust optimization to reduce the high computational cost of the uncertainty analysis and improve the performance of robust optimization problems with computationaUy expensive simulation models.Existing metamodels main focus on polynomial regression(PR),neural networks(NN)and Kriging models,these metamodeis are not well suited for large-scale robust optimization problems with small size training sets and high nonlinearity.To address the problem,a reduced approximation model technique based on support vector regression(SVR)is introduced in order to improve the accuracy of metamodels.A robust optimization method based on SVR is presented for problems that involve high dimension and nonlinear.First appropriate design parameter samples are selected by experimental design theories,then the response samples are obtained from the simulations such as finite element analysis,the SVR metamodel is constructed and treated as the mean and the variance of the objective performance functions.Combining other constraints,the robust optimization model is formed which can be solved by genetic algorithm(GA).The applicability of the method developed is demonstrated using a case of two-bar structure system study.The performances of SVR were compared with those of PR,Kriging and back-propagation neural networks(BPNN),the comparison results show that the prediction accuracy of the SVR metamodel was higher than those of other metamodels under uncertainty.The robust optimization solutions are near to the real result,and the proposed method is found to be accurate and efficient for robust optimization.This reaserch provides an efficient method for robust optimization problems with complex structure.     

钢丝绳隔振系统的神经网络杂交建模

提出一种基于神经网络的杂交建模方法,并对某电子设备的钢丝绳隔振系统滞后恢复力进行建模研究。利用周期载荷试验数据,通过参数识别确定系统非线性滞后恢复力的骨架模型。采用神经网络对系统恢复力中难以参数建模的特性进行学习训练,从而得到系统恢复力的神经网络杂交模型。利用得到的杂交模型对隔振系统在周期载荷和宽频随机载荷下的响应进行预测分析与比较,结果显示杂交模型具有较好的预测精度。     

荧光光谱法结合神经网络优化定量分析蒽芘混合物

采用荧光光谱法结合神经网络优化对蒽芘双组分混合物进行了定量分析,提出了利用留一法交叉验证（leave one out cross validation,LOOCV）的模型训练方法,以解决混合样品光谱定量分析中用有限样品建立非线性回归模型的问题。蒽和芘混合样品的激发波长为320nm,发射波长范围为320～450nm,以混合样品光谱数据的主分量作为输入、混合样品浓度作为输出进行LOOCV训练,对神经网络进行优化设计。在LOOCV实验结果中,预测10个测试质量较好的样品,平均相对误差（ARE）为3.39%,比预测所有12个样品的ARE低0.46%,样品最小相对误差可达到1.25%,10次重复实验相对标准偏差小于0.84%。该方法具有所需样品少、容错性好、分析精度高和稳定的特点。     

Robust integral of neural network and precision motion control of electrical–optical gyro-stabilized platform with unknown input dead-zones

Parametric uncertainty associated with unmodeled disturbance always exist in physical electrical–optical gyro-stabilized platform systems, and poses great challenges to the controller design. Moreover, the existence of actuator deadzone nonlinearity makes the situation more complicated. By constructing a smooth dead-zone inverse, the control law consisting of the robust integral of a neural network (NN) output plus sign of the tracking error feedback is proposed, in which adaptive law is synthesized to handle parametric uncertainty and RISE robust term to attenuate unmodeled disturbance. In order to reduce the measure noise, a desired compensation method is utilized in controller design, in which the model compensation term depends on the reference signal only. By mainly activating an auxiliary robust control component for pulling back the transient escaped from the neural active region, a multi-switching robust neuro adaptive controller in the neural approximation domain, which can achieve globally uniformly ultimately bounded (GUUB) tracking stability of servo systems recently. An asymptotic tracking performance in the presence of unknown dead-zone, parametric uncertainties and various disturbances, which is vital for high accuracy tracking, is achieved by the proposed robust adaptive backstepping controller. Extensively comparative experimental results are obtained to verify the effectiveness of the proposed control strategy.     

改进的BP算法及其在镗孔加工中应用

对一般BP算法的激励函数、误差函数等进行了改进,将梯度下降法与遗传算法相结合来训练BP网络,并在此基础上,建立了镗孔加工的人工神经网络预报模型,实验结果表明,该模型能较好地对镗孔加工过程中的尺寸进行预测。     

Prediction modeling of Coriolis type mass flow: Sensor using neural network

The neural network (NN) technique has been utilized for prediction of performance of omegatube type Coriolis mass flow sensor. The results show that a well trained and well tested NN model has the capability to predict the performance of mass flow sensor for varying design parameters depending on the availability of the data and can be used as an alternative to the physical models in the sense that the results can be produced in a fast and cost effective way. The values of correlation coefficient (R) for the training, testing and whole datasets indicates that the NN results are in good agreement with the experimental results.     

基于神经网络的热轧卷取温度模型辨识

针对热轧带钢卷取温度模型具有高度非线性的特点，利用神经网络具有逼近任何非线性函数及预报的性质，采用附加动量BP算法，准确预报卷取温度，进而应用最小二乘辨识方法对卷取温度统计模型进行参数辨识，辨识结果与设定结果的比较表明此方法行之有效。这种神经网络预报与最小二乘线性辨识相结合的方法为热轧带钢卷取温度模型的辨识优化提供了新的途径。     

Dynamics Modeling and Robust Trajectory Tracking Control for a Class of Hybrid Humanoid Arm Based on Neural Network

In order to solve the problem of trajectory tracking for a class of novel serial-parallel hybrid humanoid arm(HHA), which has parameters uncertainty, frictions, disturbance, abrasion and pulse forces derived from motors, a multistep dynamics modeling strategy is proposed and a robust controller based on neural network(NN)-adaptive algorithm is designed. At the first step of dynamics modeling, the dynamics model of the reduced HHA is established by Lagrange method. At the second step of dynamics modeling, the parameter uncertain part resulting mainly from the idealization of the HHA is learned by adaptive algorithm. In the trajectory tracking controller, the radial basis function(RBF) NN, whose optimal weights are learned online by adaptive algorithm, is used to learn the upper limit function of the total uncertainties including frictions, disturbances, abrasion and pulse forces. To a great extent, the conservatism of this robust trajectory tracking controller is reduced, and by this controller the HHA can impersonate mostly human actions. The proof and simulation results testify the validity of the adaptive strategy for parameter learning and the neural network-adaptive strategy for the trajectory tracking control.     

基于神经网络的热轧带钢卷取温度预测

热轧带钢卷取温度是影响成品带钢性能指标的重要工艺参数之一,其层流控制系统具有高度的非线性。采用附加动量BP算法,建立了基于神经网络前馈与数学模型反馈的联合层流控制系统,仿真结果表明,采用神经网络预测的卷取温度与实测温度相近,结果可信,为层流数学模型参数的在线辨识打下了坚实的基础。     

神经网络并行进化设计及其在短期电力负荷预测中的应用

针对神经网络模型进化设计问题的特点 ,提出了改进的并行遗传算法 (IPGA) ,并分析其全局收敛性与应用策略。应用此方法于短期电力负荷预测系统中 ,取得了更高的预测精度。该方法设计简单通用、模型性能评价全面合理、全局优化能力强。实际应用表明该方法是有效的 ,具有应用推广价值。     

A neural network solution for LOM process performance

A neural network (NN) modeling approach is presented for the prediction of laminated object manufacturing (LOM) process performance. A NN was developed using experimental data which were conducted on a LOM 1015 machine according to the principles of Taguchi design of experiments (DoE) method. The process parameters considered in the experiment to investigate LOM process performance were nominal layer thickness (NLT), heater temperature (HT), platform retract (PR), heater speed (HS), laser speed (LS), feeder speed (FS), and platform speed (PS). LOM process performance is divided in dimensional errors in X and Y directions (Ex and Ey), actual layer thickness (ALT), average surface roughness of vertical supporting frame (VSF-R_a), and tensile strength in X direction (TSx). It was found that NN approach can be applied in an easy way on designed experiments and predictions can be achieved, fast and quite accurate. The developed NN is constrained by the experimental region in which the designed experiment is conducted. Thus, it is very important to select parameters’ levels as well as the limits of the experimental region and the structure of the orthogonal experiment. The above analysis is useful for LOM users when prediction of process performance is needed. This methodology could be easily applied to different materials and initial conditions for optimization of other Rapid Prototyping (RP) processes.     

三维模型特征识别中的神经网络方法

将神经网络方法运用到三维模型的特征识别问题是一种新的尝试，对于用神经网络解决拓扑性的、不易被形式化的这类问题具有积极意义。本文综述了近10年来各种基于神经网络的三维模型特征识别技术，介绍并分析了三维模型拓扑数据的矢量化方法、不同神经网络模型的识别算法，以及基于层分解技术的特征自组织识别等，可帮助相关领域的研究人员较为完整地了解该领域的研究成果和发展方向。     

Wind pressure data reconstruction using neural network techniques: A comparison between BPNN and GRNN

Structural health monitoring (SHM) technique is increasingly used in civil engineering structures, with which the authentic environmental and structural response data can be obtained directly. To get accurate structural condition assessment and damage detection, it is important to make sure the monitoring system is robust and the sensors are functioning properly. When sensor fault occurs, data cannot be correctly acquired at the faulty sensor(s). In such situations, approaches are needed to help reconstruct the missing data. This paper presents an investigation on wind pressure monitoring of a super-tall structure of 600 m high during a strong typhoon, aiming to compare the performance of data reconstruction using two different neural network (NN) techniques: back-propagation neural network (BPNN) and generalized regression neural network (GRNN). The early stopping technique and the Bayesian regularization technique are introduced to enhance the generalization capability of the BPNN. The field monitoring data of wind pressure collected during the typhoon are used to formulate the models. In the verification, wind pressure time series at faulty sensor location are reconstructed by using the monitoring data acquired at the adjacent sensor locations. It is found that the NN models perform satisfactorily in reconstructing the missing data, among which the BPNN model adopting Bayesian regularization (BR-BPNN) performs best. The reconstructed wind pressure dataset has maximum root mean square error about 23.4 Pa and minimum correlation coefficient about 0.81 in reference to the field monitoring data. It is also shown that the reconstruction capability of the NN models decreases as the faulty sensor location moves from center to corner of the sensor array. While the BR-BPNN model performs best in reconstructing the missing data, it takes the longest computational time in model formulation.     

Robust tracking and distributed synchronization control of a multi-motor servomechanism with H-infinity performance

The multi-motor servomechanism (MMS) is a multi-variable, high coupling and nonlinear system, which makes the controller design challenging. In this paper, an adaptive robust H-infinity control scheme is proposed to achieve both the load tracking and multi-motor synchronization of MMS. This control scheme consists of two parts: a robust tracking controller and a distributed synchronization controller. The robust tracking controller is constructed by incorporating a neural network (NN) K-filter observer into the dynamic surface control, while the distributed synchronization controller is designed by combining the mean deviation coupling control strategy with the distributed technique. The proposed control scheme has several merits: 1) by using the mean deviation coupling synchronization control strategy, the tracking controller and the synchronization controller can be designed individually without any coupling problem; 2) the immeasurable states and unknown nonlinearities are handled by a NN K-filter observer, where the number of NN weights is largely reduced by using the minimal learning parameter technique; 3) the H-infinity performances of tracking error and synchronization error are guaranteed by introducing a robust term into the tracking controller and the synchronization controller, respectively. The stabilities of the tracking and synchronization control systems are analyzed by the Lyapunov theory. Simulation and experimental results based on a four-motor servomechanism are conducted to demonstrate the effectiveness of the proposed method.     

神经网络遗传算法在粉体流量测量的应用

利用浓度和速度电容式传感器分别测量粉体的浓度和速度,结合阀门开度、喷吹罐压力、温度等参数构造神经网络粉体流量测量模型,考虑BP算法训练神经网络测量模型时收敛速度慢、动态特性不够理想等不足,用遗传算法来优化神经网络测量模型的参数,以提高测量系统的精度。在现场与电子秤比对,最大满量程误差小于4.2%,具有工程应用价值。     

基于神经网络的气缸压力识别研究

对传统的内燃机气缸压力识别方法存在的问题 ,提出了应用人工神经网络方法进行气缸压力识别的新方法。以 BP网络构造气缸压力识别模型。通过对网络的训练 ,用实测的缸盖振动信号识别气缸压力。结果表明 ,利用神经网络进行内燃机气缸压力识别 ,识别结果的重复性好 ,精度较高     

Distributed robust adaptive control of high order nonlinear multi agent systems

In this paper, a robust adaptive neural network based controller is presented for multi agent high order nonlinear systems with unknown nonlinear functions, unknown control gains and unknown actuator failures. At first, Neural Network (NN) is used to approximate the nonlinear uncertainty terms derived from the controller design procedure for the followers. Then, a novel distributed robust adaptive controller is developed by combining the backstepping method and the Dynamic Surface Control (DSC) approach. The proposed controllers are distributed in the sense that the designed controller for each follower agent only requires relative state information between itself and its neighbors. By using the Young's inequality, only few parameters need to be tuned regardless of NN nodes number. Accordingly, the problems of dimensionality curse and explosion of complexity are counteracted, simultaneously. New adaptive laws are designed by choosing the appropriate Lyapunov-Krasovskii functionals. The proposed approach proves the boundedness of all the closed-loop signals in addition to the convergence of the distributed tracking errors to a small neighborhood of the origin. Simulation results indicate that the proposed controller is effective and robust.     

Application of neural network interval regression method for minimum zone straightness and flatness

The goal of this paper is to develop an accurate, efficient, and robust algorithm for the minimum zone (MZ) straightness and flatness. In this paper, we use an interval bias adaptive linear neural network (NN) structure together with least mean squares (LMS) learning algorithm, and an appropriate cost function to carry out the interval regression analysis. From the results, we can see that both the straightness and flatness results from the interval regression method by NN can converge closer to the definition of the MZ straightness and flatness, respectively, than that of the least-squares (LSQ) method. The interval regression method by NN developed in this paper is applicable in the linear regression analysis that has a complicated constraint, and where the LSQ method cannot be used.     

基于并行核径向基神经网络的软测量与控制模型

转炉提钒过程中存在大量多元非线性因素,难以从统计学和机理上建立各操作参数与生产目标的优化控制模型,为优化转炉的操作参数,建立了基于径向基神经网络的半钢钒含量软测量和控制模型.径向基神经网络常用于非线性回归预测和控制,但是高维的核函数矩阵运算需要花费巨大计算资源.为了缩短计算时间,本文设计了并行算法用于计算径向基网络核函数矩阵,并将它用于转炉提钒软测量和控制模型,在以MPI构建的工作站机群上执行该算法,利用实际数据验证了该算法的加速性和准确性.