基于混沌时间序列的云工作流活动运行时间预测模型

针对线性时间序列方法无法有效预测云工作流活动的运行时间的问题,提出一种基于混沌时间序列的云工作流活动运行时间预测模型.该模型利用相空间重构理论和径向基函数神经网络实现对非线性时间序列的预测.相空间重构理论能够有效刻画云工作流活动的运行时间因受系统性能、网络状况等多种因素影响而呈现的非线性特征;径向基函数神经网络能够有效预测混沌时间序列.模拟实验分别考虑了计算密集型的科学工作流和实例密集型的商务工作流的情况.实验结果表明,无论长周期活动还是短周期活动,混沌时间序列模型明显优于其他有代表性的活动运行时间预测方法.     

混沌RBF神经网络对配变电系统气体传感的智能化预测

配电变站处在高压的运行环境下,保障环境中的气体质量非常重要。于是先构建阵列式传感器监测系统对配电变站气体进行监测,然后提出相空间重构的混沌RBF神经网络对其建立预测模型,提前做好安全监管措施。对传感器采集的的气体先用C-C法优化得到延迟时间τ和嵌入维数m,进而优化重构的相空间,然后再采用Lyapunov指数判断气体具有相空间混沌特性,对相空间重构进行混沌分析,有效地展示混沌吸引子动力学特性,通过吸引子相图反映出其混沌特性。通过重构后的吸引子训练径向基RBF神经网络进行训练和预测。通过对比RBF神经网路,发现混相空间重构的混沌RBF在这个预测中比RBF神经网路预测的精度更高。     

基于相空间重构的制造系统混沌研究

针对制造企业的生产效率时间序列体现的非线性的确定而又类似随机的特点,提出了基于相空间重构的制造系统的混沌研究方法。利用互信息法和CAO氏方法分别确定最佳延迟时间和最小嵌入维数,利用生产效率时间序列对制造系统相空间重构,通过Grassberger-Procac-cia算法和小数据量方法计算关联维数和最大Lyapunov指数这两个非线性特征量,结果表明制造企业在运行过程中出现混沌现象。     

反向传播神经网络在非线性预测中的应用

在相空间重构理论的基础上将神经网络理论引入混沌时间序列预测领域,给出了BP神经网络拟和重构后所得函数关系的具体实现过程,并应用训练后的网络进行预测,经计算机仿真表明该方法是行之有效的,并给出了具体应用实例。     

基于混沌理论和神经网络的瓦斯浓度预测研究

针对瓦斯浓度的非线性和混沌时间序列可预测的特点,根据Takens理论重构瓦斯浓度相空间,分别采用互信息最法计算时间延迟τ,G—P算法计算嵌入维数m,并采用BP神经网络对混沌时间序列进行预测,最后通过煤矿瓦斯浓度预测的实例说明预测结果,从而成功实现了对瓦斯浓度的预测。     

基于改进相空间重构的压缩机运行状态预测

精确的状态预测对于压缩机的平稳运行至关重要,同时,振动信号能够表征绝大多数压缩机的运行状态。为提高预测精度,文中提出改进的相空间重构法和神经网络相结合的预测模型。首先利用相空间重构技术,将一维时间序列振动信号扩展到高维空间。针对传统的G-P算法中无标度区识别过于依赖人工经验的不足,提出了DBSCAN聚类与粒子群优化算法相结合的无标度区自动识别方法,并以相关性指标最大和残差平方和最小为目标建立无标度区识别优化模型,使用粒子群算法获得最优解,实现无标度区的自动识别。使用BP神经网络对重构后的振动信号进行预测。预测结果表明相空间重构后的信号预测效果更好。     

基于小波包和混沌理论的状态预测方法研究

通过对某系统的压力时间序列进行多层次的小波包分解,对分解得到的各层次的时间序列进行分析,建立相应的混沌预测模型并进行预测,最后应用小波包重构算法,将各层次的预测结果进行重构,从而获得时间序列的预测结果。试验表明,该方法具有较高的预测精度,可有效地应用于系统的状态监测和故障预测。     

基于混沌理论的无人驾驶车辆行驶轨迹量化分析

针对无人驾驶车辆客观量化评价困难的问题,提出基于混沌理论的无人驾驶车辆行驶轨迹的量化分析方法。应用五次多项式方法结合优秀驾驶员驾驶车辆行驶的初始状态和目标状态设计无人驾驶车辆的理想轨迹,计算实际行驶轨迹与理想轨迹的偏差得到偏差时间数据序列;采用混沌理论的C-C方法对偏差时间数据序列重构相空间;计算偏差时间数据序列的Lyapunov指数,实现无人驾驶车辆行驶轨迹的量化表示。运用提出的方法对无人驾驶车辆避障换道的行驶轨迹偏差时间数据序列进行相空间重构,计算其Lyapunov指数为正值,表征人驾驶车辆系统的混沌性,还表征人驾驶车辆收敛到稳态响应的快慢程度。试验表明基于混沌理论的无人驾驶车辆行驶轨迹的量化分析是可行的,有效的。     

基于混沌相空间重构的数控机床运动精度演化分析

数控机床运动精度状态的变化是一个复杂非线性动力系统的时空演化过程,难以通过数学建模方法分析其演化规律。提出基于混沌相空间重构理论的数控机床运动精度演化分析方法。根据混沌系统内在的规律性和有序性,分析基于相空间重构的数控机床精度演化原理,对数控机床运动精度的一维时间序列进行相空间重构,从而得到与原系统拓扑同构的高维相空间。提出用相点轨迹描述运动精度在相空间中的演化规律,构造一个非线性演化模型,用相点的多维分量构成输入向量,对其精度趋势进行演化。实验结果表明,基于混沌相空间重构的演化模型,可以很好地追踪数控机床运动精度的演变趋势和规律,且具有较高的演化精度。     

基于空时混沌分析的海面小弱目标检测

为了提升海杂波中小弱目标检测能力,通过研究海杂波在空域和时域上的动态特性,提出基于空时混沌分析的海面小弱目标检测方法。首先,在海杂波混沌动力系统相空间重构的基础上提取海浪序列图像的空域混沌参数和时域混沌参数,验证海杂波在空域与时域具备混沌性;然后,采用径向基函数神经网络学习海杂波的空域混沌重构函数、时域混沌重构函数和空时耦合系数,联合空域和时域混沌特性综合重构海杂波在时间与空间上的传播规律。多种复杂度海面目标检测试验结果表明,与空域混沌和时域混沌方法相比,预测误差可降低10%,检测概率可提高20%,基于空时混沌重构的海杂波抑制能力与小目标检测性都得以显著提升。     

A hybrid look-ahead SOM-FBPN and FIR system for wafer-lot-output time prediction and achievability evaluation

A hybrid system is constructed in this study for wafer-lot-output time prediction and achievability evaluation, which are critical tasks for a wafer fab (fabrication plant). In the first part of the hybrid system, a look-ahead self-organization map fuzzy-back-propagation network (SOM-FBPN) is constructed to predict the output time of a wafer lot. Compared with traditional approaches in this field, the look-ahead SOM-FBPN has three advanced features: incorporating the future release plan, classifying wafer lots, and incorporating expert opinions. According to experimental results, the prediction accuracy and efficiency of the look-ahead SOM-FBPN were significantly better than those of many existing approaches. In the second part of the hybrid system, a set of fuzzy inference rules (FIRs) are established to evaluate the achievability of an output time forecast, which is defined as the possibility that the fabrication on the wafer lot can be finished in time before the output time forecast. Achievability is as important as accuracy and efficiency but has been ignored in traditional studies. With the proposed methodology, both output time prediction and achievability evaluation can be concurrently accomplished.     

大数据驱动的晶圆工期预测关键参数识别方法

工期是晶圆制造中的重要性能指标,对其进行精准预测可促进系统运行优化,保证订单的准时交付率。针对晶圆工期影响参数多、数据体量大且作用机理复杂的特点,提出数据驱动的晶圆工期关键参数过滤方法,识别影响晶圆工期波动的关键参数。分析晶圆工期潜在影响参数,构建候选参数集;基于信息熵方法设计关键参数的入选测度,综合度量参数间的相关性、冗余性与互补性;提出过滤式的关键参数识别算法,滤取影响工期波动的关键参数子集。采用实例数据,从1 202个候选参数中过滤得到78个关键参数,并采用神经网络模型进行工期预测,结果表明,该方法在预测精度和稳定性上都优于采用全局参数的多元线性回归与神经网络方法。     

A daily production model for wafer fabrication

The complex process and high variation in wafer fabrication make its production management very difficult. Problems such as planned target achievement and line balancing are not unusual in the industry. Such problems reveal the importance of developing a daily production policy for wafer fabrication. Planned target achievement and line balancing are the major concerns of this investigation in developing a daily production model. This investigation divides the process of wafer fabrication into two sections, i.e., the front and the rear, according to the last sputtering operation step. In the rear section, the objective is attaining the planned output target. In the front section, the major focus is to satisfy the demand of the rear section so that the production line is balanced. Release and dispatch policies are incorporated in this study to achieve both objectives. A real-world numerical example is used as simulation data. Results show that the proposed daily production model gives a better performance in the achievement of monthly planned output but suffers a little in the performance of line balancing.     

Modeling of large-scale complex re-entrant manufacturing systems by extended object-oriented Petri nets

In this paper we propose an extended object-oriented Petri nets (EOPNs) for effective modeling of large-scale complex re-entrant manufacturing systems (ReMSs). To cope with their complexity because of the re-entrant process route and the mixed production mode, a special type of transition called a main-bus gate (MBG) is introduced, which can cause products to undergo every re-entrant manufacturing stage. A hierarchical approach is also applied to cope with the complexity. As a typical large-scale complex ReMS, the semiconductor wafer fabrication system (SWFS) is taken as an example for demonstrating the proposed EOPNs. A case study is provided to show the modelling procedures, and the resulting model validates that the EOPNs can cope well with the modelling complexity of large-scale complex ReMSs.     

A SOM-FBPN-ensemble approach with error feedback to adjust classification for wafer-lot completion time prediction

Predicting the completion time of a lot is a critical task to a wafer fabrication plant (wafer fab). Many recent studies have shown that pre-classifying a wafer lot before predicting the completion time was beneficial to prediction accuracy. However, most classification approaches applied in this field could not absolutely classify wafer lots. Besides, whether the pre-classification approach combined with the subsequent prediction approach was suitable for the data was questionable. For tackling these two problems, a self-organization map-fuzzy-back-propagation network-ensemble (SOM-FBPN-ensemble) approach with error feedback to adjust classification is proposed in this study. The proposed methodology has two advanced features: predicting the completion time using a FBPN-ensemble instead of a single FBPN, and feeding back the prediction error to adjust the classification result by the SOM. According to experimental results, the prediction accuracy of the proposed approach was significantly better than those of many existing approaches. Besides, the effects of the two advanced features were also evident.     

Resource Portfolio Planning Methodology for Semiconductor Wafer Manufacturing

Resource portfolio planning is a frequent task in semiconductor wafer fabrication plants, as process, machine and product technologies evolve rapidly and the plants go through capacity expansion. As wafer fabrication plants are complex integrated factories with conspicuous queuing effects, portfolio planning must take into consideration machine use, factory throughput, and total flow-time simultaneously. This paper describes a resource portfolio planning methodology for wafer fabrication foundry plants. An improved static capacity model is first presented. A portfolio planning procedure based on static capacity estimation and queuing analysis is next described. This procedure enables the solution space of resource portfolios to be explored effectively and has demonstrated a capability superior to the current planning method in an industry case study. A software implementation of the procedure is also used to clarify planning dilemmas. It is shown that empirical formulae can be used to estimate the efficiency of batch machines. It is also used to show three types of portfolio adjustment action: flow-time reduction, cost reduction and effectiveness improvement.     

基于混合式特征选择模型的晶圆允收测试关键参数识别方法

晶圆允收测试是晶圆加工过程的关键环节,对其中的关键测试参数进行准确分析识别有助于准确预测晶圆良率、及时发现工艺缺陷。针对测试参数维度高、数据冗余性强、关键特征不显著的特点,以最小化晶圆允收测试参数量和晶圆良率预测误差为目标,提出了过滤式与封装式相结合的混合式特征选择方法。在过滤式预筛选中,通过互信息度量各参数与晶圆良率的相关性,以及各参数之间冗余性,并根据最大相关、最小冗余准则,缩小候选参数规模;在封装式精选中,以遗传算法实现候选参数的编码、寻优,根据神经网络的晶圆良率预测误差进行适应度函数评价,进一步精选关键特征。最后,采用标准数据集和实例数据对所提方法进行了有效性验证。     

Due-date assignment in wafer fabrication using artificial neural networks

Due-date assignment (DDA) is the first important task of shop floor control in wafer fabrication. Due-date related performance is impacted by the quality of the DDA rules. Assigning order due dates and timely delivering the goods to the customer will enhance customer service and competitive advantage. A new methodology for lead-time prediction, artificial neural network (ANN) prediction is considered in this work. An ANN-based DDA rule combined with simulation technology and statistical analysis is developed. Besides, regression-based DDA rules for wafer fabrication are modelled as benchmarking. Whether neural networks can outperform conventional and regression-based DDA rules taken from the literature is examined. From the simulation and statistical results, ANN-based DDA rules perform a better job in due-date prediction. ANN-based DDA rules have a lower tardiness rate than the other rules. ANN-based DDA rules have better sensitivity and variance than the other rules. Therefore, if the wafer fab information is not difficult to obtain, the ANN-based DDA rule can perform better due-date prediction. The SFM_sep and JIQ in regression-based and conventional rules are better than the others.     

混沌预测模型改进及在电力日负荷预测中的应用

针对电力日负荷预测中基于最大Lyapunov指数的传统混沌预测模型的缺陷,提出以下改进思想:采用微分熵法同时确定嵌入维数和延迟时间,改善相空间重构质量的同时,有效减少计算量;引入夹角参数,在与中心点距离最短的点中,筛选夹角最小的点作为最终邻近点;基于相似性原理,引入取舍规则,使得计算结果唯一确定。本文方法解决了传统预测模型计算量大、运算速度慢、预测精度不高以及正负号选取等问题。通过对典型混沌系统数值验证和某地区实际电力负荷系统的预测分析证明了本文方法的可靠性和高效性,为电力日负荷预测提供了一种新的有效途径。