Soft sensor for real-time cement fineness estimation

This paper describes the design and implementation of soft sensors to estimate cement fineness. Soft sensors are mathematical models that use available data to provide real-time information on process variables when the information, for whatever reason, is not available by direct measurement. In this application, soft sensors are used to provide information on process variable normally provided by off-line laboratory tests performed at large time intervals. Cement fineness is one of the crucial parameters that define the quality of produced cement. Providing real-time information on cement fineness using soft sensors can overcome limitations and problems that originate from a lack of information between two laboratory tests. The model inputs were selected from candidate process variables using an information theoretic approach. Models based on multi-layer perceptrons were developed, and their ability to estimate cement fineness of laboratory samples was analyzed. Models that had the best performance, and capacity to adopt changes in the cement grinding circuit were selected to implement soft sensors. Soft sensors were tested using data from a continuous cement production to demonstrate their use in real-time fineness estimation. Their performance was highly satisfactory, and the sensors proved to be capable of providing valuable information on cement grinding circuit performance. After successful off-line tests, soft sensors were implemented and installed in the control room of a cement factory. Results on the site confirm results obtained by tests conducted during soft sensor development.     

一种基于傅里叶基神经网络的频谱分析方法

该文提出了一种用递推最小二乘法训练傅里叶基神经网络权值的频谱分析方法。其主要思想是采用递推最小二乘法训练傅里叶基神经网络权值,根据权值获得信号的幅度谱和相位谱。该方法不涉及复数的乘法运算和加法运算,便于软件和硬件实现,特别适合于DSP软硬件实现。仿真结果表明,该方法不仅计算精度高,计算速度快,而且具有噪声滤波功能,是一种有效的频谱分析方法。     

一种基于神经网络的模拟电路故障诊断方法

提出了一种基于模拟电路故障诊断的神经网络方法。这种方法利用小波分解、数据标准化、主成分分析对输入数据进行预处理,采用k个神经元输出的前馈神经网络结构进行有效训练。该方法检测和识别故障准确率高,系统的鲁棒性和稳定性强。     

无线传感器网络中基于神经网络的数据融合模型

数据融合技术通过减少传感器节点间的数据通信量,可以有效地节省传感器节点能耗,延长无线传感器网络的寿命.提出了独特的基于神经网络的数据融合模型(NNBA),该模型巧妙地将无线传感器网络的分簇层次结构与神经网络的层次结构相结合,将每个簇设计为一个三层感知器神经网络模型,通过神经网络方法从采集到的大量原始数据中提取特征数据,然后将特征数据发送给汇聚节点.以森林火灾实时监测网为应用实例,设计神经元模型及功能函数,并给出NNBA模型的仿真测试结果.     

基于门控图卷积神经网络的有机化学反应预测

随着现代医药技术和计算机技术的发展,采用人工智能技术来加速药物的研发进度成为了研究热点,而对有机化学反应产物的高效预测是药物逆合成路线设计中的关键问题。针对样本数据集中化学反应类型分布不均匀的问题,提出了一种主动采样训练下的门控图卷积神经网络（ASGGCN）模型。首先,输入化学反应物的简化分子线性输入规范（SMILES）编码,通过门控图卷积神经网络（GGCN）以及注意力机制预测反应中心所在位置;然后,根据化学约束条件和候选反应中心枚举出可能的化学键组合来生成候选产物,再通过门控图卷积差分网络对候选产物进行筛选;最终,得到反应产物。门控图卷积神经网络拥有三个权重参数矩阵并通过门控对信息加以融合,与传统的图卷积神经网络相比,它能获取更加丰富的原子隐藏特征信息。通过主动采样的方式进行训练,使得该模型能够兼顾较差样本和普通样本的分析能力。实验结果表明,所提模型对化学反应产物的Top-1预测准确率可达87.2%,对比Weisfeiler-Lehman差分网络（WLDN）模型提高了1.6个百分点,可见模型能够更准确地预测有机化学反应产物。     

磁盘读取系统的神经网络控制

该文根据磁盘读取系统的结构,建立了它的物理模型,并指出其存在非线性因素的原因。由于经典控制方法对于非线性系统效果不是很好,因此使用MATLAB神经网络工具箱中的NARMA—L2控制模块进行控制,并利用Simulink可视化建模工具平台设计了整个控制系统。对神经网络控制器进行了训练,并利用经过训练的神经网络控制磁盘读取系统。仿真的结果表明,神经网络NARMA—L2控制模块能够满足含有非线性因素的磁盘读取系统的控制要求。     

Soft analyzers for a sulfur recovery unit

This work deals with the design and implementation of soft sensors for a Sulfur Recovery Unit (SRU) in a refinery. Soft sensors are mathematical models able to emulate the behavior of existing sensors on the basis of available measurements. In this application, they are used when sensors are taken off for maintenance. The measurements considered in this work are very important for the environmental impact of the refinery, as they regard pollutant acid gas emissions. Four strategies have been implemented and compared: Multi-Layer Perceptrons (MLP) and Radial Basis Function neural networks, Neuro-Fuzzy networks and nonlinear Least-Squares (LSQ) fitting. The best performance is given by MLP neural networks and nonlinear LSQ, all of them implementing Nonlinear Moving Average models. The best soft sensors have been installed on the on-line distributed control systems of the refinery and on-line performance is highly satisfactory.     

Noncircular cutting on a lathe using tool position and differential motor current

In this paper, noncircular cutting on a lathe (NCL) using the signals of tool position and differential motor current is developed. First, a neural-network is used to learn a weighted combination of tool position and differential motor current of a voice coil servomotor system (VCSS). The differential motor current is achieved from the difference between motor current under real cutting and that under air cutting. Then a nonlinear inverse control based on the learned model is designed to obtain an acceptable cutting error. Although the nonlinear inverse control is accomplished from the consideration of tool position under real cutting, its performance cannot be ensured as the NCL is subjected to the uncertainties (e.g. noise, different cutting conditions, aging of system components). Under these circumstances, a fuzzy sliding-mode control is then synthesized with the previous nonlinear inverse control to reduce the cutting error. The experimental results of NCL using the proposed control including profile error of finished workpiece and surface roughness of finished workpiece, are better than those without the consideration of motor current.     

Super-resolution using neural networks based on the optimal recovery theory

An optimal recovery based neural-network Super Resolution algorithm is developed. The proposed method is computationally less expensive and outputs images with high subjective quality, compared with previous neural-network or optimal recovery algorithms. It is evaluated on classical SR test images with both generic and specialized training sets, and compared with other state-of-the-art methods. Results show that our algorithm is among the state-of-the-art, both in quality and efficiency. Y. Long: This work was supported in part by the National 211/985 Project of Shanghai Jiaotong University under Subgrant PRP[S03009009].     

BJEST: A reverse algorithm for the real-time predictive maintenance system

An algorithm, to estimate the machine system parameters from the motion current signature, based upon non-linear time series techniques for use in the real-time predictive maintenance system is presented in this paper. Earlier work has introduced the use of a neural-network approach to learn non-linear mapping functions for condition monitoring systems. However, the performance of the neural-network largely depends upon the quality of the training data, and that of the quality and type of the pre-processing of the input data. A reverse algorithm called BJEST (Bansal–Jones Estimation), for estimating the machine input parameters using the motion current signature, has been designed and proven to be successful in estimating the macro-dynamics of the motion current signature. This motivated the enhancement of the predictive analysis to incorporate non-linear characteristic of the motion current signature. The results show considerable improvement in the estimation of the parameters using the enhanced BJEST algorithm due to estimation consistency, hence, improving the real-time predictive maintenance system.