基于事件触发的神经网络控制器稳定性分析

针对非线性系统稳定性问题,设计了基于事件触发的神经网络控制器来稳定非线性系统,控制器只有在触发规则满足的条件下,才更新控制参数,降低了网络传输率。算法开始先建立非线性系统模型,在采样过程中引入事件触发机制,并且设计了神经网络控制器,对于系统中包含的传输时滞,引入系统时滞模型,运用输入延迟法将同步控制器求解问题转化为时滞系统的稳定性问题。再通过构造分段Lyapunov-Krasovskii泛函并结合Jensen不等式,给出了非线性系统稳定条件。与传统数据采样系统相比,本文所提出的方法有效地增大了采样间隔,结尾通过例子仿真验证了所提出方法的有效性。     

一种基于神经网络的入侵检测系统研究

提出在入侵检测系统模型设计中引入神经网络技术,建立了一个基于神经网络的入侵检测系统模型。针对传统BP算法存在的一些固有缺点,提出增加动量项与自适应调节学习速率相结合的改进算法,一定程度上克服了BP神经网络存在的问题。实验结果表明,基于改进的BP神经网络的入侵检测方法具有良好的检测性能。     

基于PNDSC 的欠驱动AUV编队控制器设计

针对含有模型不确定与未知海洋环境扰动下的欠驱动自主水下航行器(AUV)的编队控制问题, 提出一种基于预估器的神经网络动态面(PNDSC) 控制算法. 将动态面法引入控制器的设计中, 采用神经网络逼近AUV模型中的不确定项与海洋环境的扰动, 并结合预估器设计了神经网络权值的离散迭代更新率. Lyapunov 稳定性分析表明, 闭环系统所有信号是一致最终有界的. 仿真结果验证了所提出方法的有效性.

     

一类复杂适应系统的建模研究

引入复杂适应系统理论,将股市、房地产系统视为一类复杂适应系统。采用多智能体方法进行建模,设计随机决策Agent、模仿Agent、BP神经网络Agent等模块,引入订单簿技术实现Agent之间的信息交互。通过仿真模拟该类复杂系统的动态演化过程,验证了模型的有效性。     

基于神经索引实例与知识推理的混合型智能CAPP策略

提出了一种新的基于耦合神经网络实例与知识的混合推理策略,采用面向对象的方法表达实例和知识,将神经索引模型引入实例推理中,在此基础上实现了ＣＡＰＰ的变异设计。同时,建立了基于零件及其工艺数据知识的分层知识表达与层次式推理机制,从而实现了ＣＡＰＰ的创成式设计。在给出这种知识化的ＣＡＰＰ系统的总体结构之后,详细讨论了基于神经网络的实例索引模型及实例推理和知识基推理的实现过程,由于吸收了派生法的类比设计思     

具有不确定性的非线性系统自适应神经网络L2增益控制

针对存在不确定性的非线性系统,提出了自适应神经网络L2增益控制器设计方法,将基于Hamilton—Jacobi—Issacs（HJI）不等式和自适应神经网络策略相结合,有效地克服了需要被控对象精确建模的局限性．神经网络对系统模型的偏差进行拟合;为了补偿拟合误差,引入补偿控制器和神经网络权值自适应调节律,通过在线自适应修正神经网络权值,来保证闭环系统满足相应的L2性能准则．仿真结果表明提出的控制器设计方法是有效的,克服了一般方法需要被控对象精确建模的局限性．     

聚合物分子量分布的多变量动态系统模型

为了解决聚合产品分子量分布控制的难题,将神经网络引入对其进行了无需任何系统内部先验知识的黑箱建模。所使用的神经网络是由B样条神经网络和非线性递归神经网络(DRNN)组合而成,并使用误差反传算法对网络进行训练和学习,从而建立了多变量动态系统的分子量分布模型。在模型建立中将控制变量与分布参数的函数关系利用非线性递归神经网络描述,分子量分布函数使用B样条神经网络表示,仿真研究结果证明该方法取得了预期的建模效果,具有一定的推广实用价值。     

基于多Agent系统设计原理的神经网络结构设计

将多Agent系统设计原理引入到神经网络的结构设计中,提出了基于多Agent系统设计原理的网络结构设计算法(MANN,multi-Agent neural network)。该算法基于多Agent系统设计的原理,将较大神经网络的结构分解成一组结构较小的子神经网络,每一个子神经网络都具有相对较少的隐节点数。将这种网络结构应用到热工过程建模中,仿真结果表明MANN网络具有较高的泛化能力和运算速度。     

人工神经网络专家系统在门窗选型中的应用

本文提出了一个将神经网络专家系统的设计思想引入门窗选型过程的应用模型,从而提高了门窗选型的合理性,也充分发挥了神经网络及专家系统的各自优点。文中主要介绍了基于神经网络专家系统的门窗选型模型的结构和推导过程,此模型在实际应用中确实取得了令人满意的效果。     

融合MLP与循环神经网络算法的语法自动纠错方法研究

为了进一步提升语法自动纠错技术的实用性，研究对以循环神经网络为核心的Sep2Sep模型进行优化改进，引入双向LSTM循环神经网络，将基于双向LSTM的Sep2Sep模型与MLP神经网络相结合构建语法自动纠错系统，并通过测试实验验证语法自动纠错系统的准确率。研究结果表明，研究所设计的语法自动纠错系统F_0.5值为56.37,P值和R值分别为66.78和35.09,检测准确率较高。纠错系统的运行响应时间保持在1.34 s,能在多个检测目标并发情况下进行快速系统响应。研究利用双向LSTM和MLP神经网络解决传统纠错模型的梯度爆炸问题，并采用分布式架构提升自动纠错系统的运行能力，对进一步加强自动语法纠错技术的实用性具有重要意义。     

An expert system based on artificial neural network for predicting the tensile behavior of tailor welded blanks

The forming behavior of tailor welded blanks (TWB) is influenced by thickness ratio, strength ratio, and weld conditions in a synergistic fashion. In most of the cases, these parameters deteriorate the forming behavior of TWB. It is necessary to predict suitable TWB conditions for achieving better-stamped product made of welded blanks. This is quite difficult and resource intensive, requiring lot of simulations or experiments to be performed under varied base material and weld conditions. Automotive sheet part designers will be greatly benefited if an ‘expert system’ is available that can deliver forming behavior of TWB for varied weld and blank conditions. This work primarily aims at developing an artificial neural network (ANN) model to predict the tensile behavior of welded blanks made of steel grade and aluminium alloy base materials. The important tensile characteristics of TWB are predicted within chosen range of varied blank and weld condition. Through out the work, PAM STAMP 2G^® finite element (FE) code is used to simulate the tensile behavior and to generate output data required for training the ANN. Predicted results from ANN model are compared and validated with FE simulation for two different intermediate TWB conditions. It is observed that the results obtained from ANN are encouraging with acceptable prediction errors. An expert system framework is proposed using the trained ANN for designing TWB conditions that will deliver better formed TWB products.     

Logic programming for process planning in the domain of sheet metal forming with progressive dies

In this work an intelligent system pertaining to sheet metal part and process design has been developed, storing knowledge and prescribing ways to use this knowledge according to the ‘programming in logic’ paradigm. The sheet metal parts covered by the software are those having U shape and being manufactured by bending (folding), cutting and piercing with particular emphasis on progressive dies. The use envisaged and corresponding parts of the system are: checking the part design for manufacturability, planning process phases, and checking the configuration of press tools involved. Particular attention is paid to the presentation of knowledge that has been gathered from handbooks and verified / enhanced in industry. This is first presented in natural language and then its formal representation in Prolog is described and explained by examples. Part design and press tool checking knowledge is relatively straightforward to represent and structure ‘linearly’. Process planning knowledge is based on patterns that are captured in lists and activated in a case-by-case fashion exploiting the power of Prolog. Validation of the system was conducted using examples from industry.     

Neural network based modeling and optimization of deep drawing – extrusion combined process

A combined deep drawing–extrusion process is modeled with artificial neural networks (ANN’s). The process is used for manufacturing synchronizer rings and it combines sheet and bulk metal forming processes. Input–output data relevant to the process was collected. The inputs represent geometrical parameters of the synchronizer ring and the outputs are the total equivalent plastic strain (TEPS), contact ratio and forming force. This data is used to train the ANN which approximates the input-output relation well and therefore can be relied on in predicting the process input parameters that will result in desired outputs provided by the designer. The complex method constrained optimization is applied to the ANN model to find the inputs or geometrical parameters that will produce the desired or optimum values of TEPS, contact ratio and forming force. This information will be very hard to obtain by just looking at the available historical input–output data. Therefore, the presented technique is very useful for selection of process design parameters to obtain desired product properties.     

CHORD RECOGNITION USING NEURAL NETWORKS BASED ON PARTICLE SWARM OPTIMIZATION

A sequence of musical chords can facilitate musicians in music arrangement and accompaniment. To implement an intelligent system for chord recognition, in this article we propose a novel approach using artificial neural networks (ANN) trained bythe particle swarm optimization (PSO) technique and back-propagation (BP) learning algorithm. All of the training and testing data are generated from musical instrument digital interface (MIDI) symbolic data. Furthermore, in order to improve the recognition efficiency, an additional feature of cadencesis included. In other words, cadence is not only the structural punctuation of a melodic phrase but is considered as the important feature for chord recognition. Experimental results of our proposed approach show that adding a cadence feature significantly improves recognition rate, and the ANN-PSO method outperforms ANN-BP in chord recognition. In addition, because preliminary experimental recognition rates are generally not stable enough, we chose the optimal ANNs to propose a two-phase ANN model to integrate the results among many models.     

Porting an industrial sheet metal forming code to a distributed memory parallel computer

The parallel version of the sheet metal forming semi-implicit finite element code ITAS3D has been developed using the domain decomposition method and direct solution methods at both subdomain and interface levels. IBM Message Passing Library is used for data communication between tasks of the parallel code. Solutions of some sheet metal forming problems on IBM SP2 computer show that the adopted DDM algorithm with the direct solver provides acceptable parallel efficiency using a moderate number of processors. The speedup 6.7 is achieved for the problem with 20000 degrees-of-freedom on the 8-processor configuration.     

An integrated knowledge based system for sheet metal cutting–punching combination processing

Cutting and punching are two important processes in the sheet metal industry; the former is flexible in processing whilst the latter is effective in production. To combine the advantages of both, the so-called combination machines that combine the cutting and punching processes are used in sheet metal processing. To support such machines, this paper presents an integrated knowledge based CAD/CAM system for sheet metal cutting–punching combination processing. The whole system consists of five functional modules, i.e., modeling, nesting, process planning, NC-programming and simulation and reporting module. These modules share a unified objected-oriented data structure, so they are integrated seamlessly in terms of data processing. The knowledge base is the core of the integrated system, based on which, the system runs in an efficient and intelligent manner. This system has been successfully applied in industry, and an application example is demonstrated.     

Automated driving for individualized sheet metal part production—A neural network approach

The manufacturing of individualized sheet metal components is one of the most important issues in industrial sheet metal working. Incremental forming methods, in particular driving, offer the opportunity for achieving this objective. However, these manual processes are very difficult to automate, as a result of their complexity and user interactivity. To resolve this problem, a knowledge-based approach is presented, which utilizes a special type of driving process. Initially, a neural network architecture is established which delivers manufacturing strategies allowing part production for simple component shapes. After providing a method for training data generation, training sessions are carried out. Strategies, computed by trained networks, are adopted for processing sheet blanks which are used for evaluating the framework. Finally, the developed procedure is generalized, and a concept is designed which allows a transfer, in order to facilitate the production of arbitrary individualized sheet metal parts.     

Tool path optimization for single point incremental sheet forming using response surface method

Incremental sheet forming (ISF) process is based on localized plastic deformation in a thin sheet metal blank. It consists to deform progressively and locally the sheet metal using spherical forming tool controlled by a CNC machine-tool. Although it is a slow process compared to conventional forming technique such as stamping. The cost reduction linked to the fact that punches and dies are avoided which makes it a very attractive process for small batch production and rapid prototyping. However, ISF process depends strongly on the forming tool path which influences greatly the part geometry and sheet thickness distribution. A homogeneous thickness distribution requires a rigorous optimization of the parameter settings, and an optimal parameterization of the forming strategy. This paper shows an optimization procedure tested for a given forming strategy, in order to reduce the manufacturing time and homogenize thickness distribution of an asymmetric part. The optimal forming strategy was determined by finite element analyses (FEA) in combination with response surface method (RMS) and sequential quadratic programming (SQP) algorithm.     

Three‐dimensional localization of thin‐walled sheet metal parts for robotic assembly

This article presents a technical demonstration of a system for determining the three‐dimensional spatial location of complexly shaped, thin‐walled sheet metal parts grasped by robots during assembly. For successful part assembly, the precise location of grasped parts (essential for successful mating of parts) must be achieved. A localization system is implemented to determine the accurate position and orientation of a sheet metal part that has been picked up by a robot from an arbitrary location. The proposed localization system employs a novel sensing method, utilizing laser‐based proximity and edge detectors, to extract the part feature data in real time. These geometrical feature data are incorporated into an existing localization algorithm, which is based on the singular value decomposition formulation of the part localization problem. The sensing method is particularly effective in measuring 3‐D feature geometry (i.e., thin edges) of sheet metal parts. An experimental single‐robot test bed has been developed to demonstrate the feasibility of the part localization concept for a single sheet metal part. The experimental results obtained from the test bed demonstrate that the system can be effectively used for the localization of thin‐walled sheet metal parts. © 2002 Wiley Periodicals, Inc.     

Next generation stamping dies — controllability and flexibility

In sheet metal forming, external energy is transferred to sheet metal through a set of tooling to plastically deform a workpiece. The design of the tooling and its associated forming process parameters play important roles in this manufacturing process since they directly affect the quality and cost of the final product. With increasing demands from customers, government regulations, and global competition, the controllability and flexibility of stamping dies have been challenged. In this paper, we will summarize the research activities conducted at the Advanced Materials Processing Laboratory at Northwestern University in the area of sheet metal forming. An overview of our approach towards the system will be given followed by a summary of individual projects in the areas of failure prediction, design and control of a variable binder force, and the segmented die design with local adaptive controllers.