Quality prediction and analysis for large-scale processes based on multi-level principal component modeling strategy

This paper proposed a multi-level principal component regression (PCR) modeling strategy for quality prediction and analysis of large-scale processes. Based on decomposition of the large data matrix, the first level PCR model divides the process into different sub-blocks through uncorrelated principal component directions, with a related index defined for determination of variables in each sub-block. In the second level, a PCR model is developed for local quality prediction in each sub-block. Subsequently, the third level PCR model is constructed to combine the local prediction results in different sub-blocks. For process analysis, a sub-block contribution index is defined to identify the critical-to-quality sub-blocks, based on which an inside sub-block contribution index is further defined for determination of the key variables in each sub-block. As a result, correlations between process variables and quality variables can be successfully constructed. A case study on Tennessee Eastman (TE) benchmark process is provided for performance evaluation.     

台风最大风速预测的高斯过程回归模型

针对影响台风最大风速的输入变量较多以及输入变量与输出变量之间的非线性变化特点,首先计算各个输入变量与输出变量间的互信息,这些互信息间接地反映了各个输入变量与输出变量间的相关性;然后根据t检验法确定一个阈值,对于互信息小于阈值的输入变量作不相关变量处理,筛选出最佳的模型输入变量;最后采用高斯过程回归模型对筛选后的样本集进行拟合,在贝叶斯非参数建模的框架下,确定高斯过程回归模型的协方差函数.仿真结果表明,所得高斯过程模型能够满足绝对误差的预定要求,且具有较大的实用价值.     

An extended model for remaining time prediction in manufacturing systems using process mining

The ability to predict the remaining cycle-time in industrial environments is of major concern among production managers. An accurate prediction would enable managers to handle undesired situations with more control, thereby preventing future losses. However, making such predictions is no trivial task: there are many methods available to cope with this problem, including a recent research stream in process mining. Process mining provides tools for automated discovery of process models from event logs, and eventually, extend those models in driving predictions. In general, predictive models in process mining generally deals with business processes, and not directly with the industrial environment, which contains a full prism of particularities. In this paper we propose a hybrid predictive model based on transition-systems and statistical regression which is “product-oriented”, tailored to better predict online cycle-times on industrial environments. We propose a weight for each method, optimized by a linear programming model. We tested our new approach on an artificially created log that emulates an industrial environment, and on a real manufacture log. Results showed that our approach provides better accuracy measures for both test instances.     

基于半马尔柯夫过程的流量预测方法

提出了一种基于半马尔柯夫过程的流量预测方法。通过半马尔柯夫过程描述网络流量特性,将网络流量划分为四种状态：忙、空闲、上升和下降。通过各状态下的网络流量特性及各状态间的相互转换关系,推导了对忙状态下网络流速率上界的预测方法。对广域网和局域网的实际流量数据的分析和检验表明,95%的数据均服从半马尔柯夫过程相应状态下的随机分布;90%的流量预测以0.8或0.9的概率低于我们所预计的流量上界,且主干网流量预测的流量上界与实际流量之间的相对误差低于15%。     

Stream water temperature prediction based on Gaussian process regression

The prediction of stream water temperature presents an interesting topic since the water temperature has a significant ecological and economical role, such as in species distribution, fishery, industry and agriculture water exploitation. The prediction of stream water temperature is usually based on appropriate mathematical model and measurements of different atmospheric factors. In this paper, a probabilistic approach to daily mean water temperature prediction is proposed. The resulting model is a combination of two Gaussian process regression models where the first model describes the long-term component of water temperature and the other model describes the short-term variations in water temperature. The proposed approach is developed even further by modeling the short-term variations with multiple Gaussian process regression models instead with a single one. Apart from that, variable selection procedure based on mutual information is presented which is suitable for input variable selection when nonlinear models for stream water prediction are developed. The proposed approach is compared with traditional modeling approaches on the measurements obtained on the Drava river in Croatia. The presented methodology can be used as a basis of the predictive tools for water resource managers.     

基于预测模型的浮选过程pH值控制

矿浆pH值是泡沫浮选过程中的一个非常重要的被控量.目前,多数选厂的矿浆pH值控制基本是依靠现场工人定期对矿浆样本进行pH值测量,凭主观经验对pH调整剂进行调整.由于操作工人的主观性和随意性的影响以及矿浆样本pH值测量与药剂调整间存在的较长的时间滞后,矿浆pH值波动频繁,很难使矿物浮选保持在一个稳定最优生产状态下运行.为了使矿浆pH值保持在一个期望的生产状态,基于浮选泡沫表面视觉信息提出了一种新的矿浆pH值控制方法,分别采用基于泡沫视觉信息的自适应遗传混合神经网络AG-HNN和自适应遗传PID(AG-PID)控制方法建立了矿浆pH值预测模型和pH值控制模型,基于所建立预测和控制模型对浮选药剂用量进行调整,解决了浮选矿浆pH值波动问题.工业浮选现场的实验结果表明该方法可以使矿浆pH值保持在一个期望的范围内,有效提高浮选性能.     

Time prediction based on process mining

Process mining allows for the automated discovery of process models from event logs. These models provide insights and enable various types of model-based analysis. This paper demonstrates that the discovered process models can be extended with information to predict the completion time of running instances. There are many scenarios where it is useful to have reliable time predictions. For example, when a customer phones her insurance company for information about her insurance claim, she can be given an estimate for the remaining processing time. In order to do this, we provide a configurable approach to construct a process model, augment this model with time information learned from earlier instances, and use this to predict e.g., the completion time. To provide meaningful time predictions we use a configurable set of abstractions that allow for a good balance between “overfitting” and “underfitting”. The approach has been implemented in ProM and through several experiments using real-life event logs we demonstrate its applicability.     

Data-based multiple-model prediction of the production rate for hematite ore beneficiation process

The prediction of the production rate of the hematite ore beneficiation process is important to plant-wide optimization. This paper presents a data-based multi-model approach to predict the production rate with multiple operating modes. The inputs of the predictive model are the performance indices of each unit process, and the output is the global production index (the production rate) of the hematite ore beneficiation process. The multiple models are developed by integrating the fuzzy clustering algorithm and machine learning algorithm. A global model, Takagi–Sugeno–Kang fuzzy model, and multiple neural network model were compared using the data obtained from a practical industrial process, and the effectiveness of the proposed algorithm was proven.     

基于模糊聚类神经网络的化工过程预报模型研究

本文提出了在化工过程预报中的模糊聚类神经网络模型,该模型具有提取典型数据、优化模糊规则及优化参数的优点,在化工过程预报实验中与传统方法相比预报结果的精度提高,计算时间缩短。     

基于概念漂移检测的数字孪生流程预测模型

预测性流程监控可以在业务流程运行过程中提供及时的信息,以便采取措施来应对潜在风险,如何提高流程预测的准确度一直受到高度关注。现有的研究方法大部分都在静态环境下引入,很少有结合数字孪生技术在动态环境中的流程预测。为此,提出了一个基于概念漂移检测的方法,并构建数字孪生流程预测模型（digital twin based on concept drift,DTBCD）预测下一个活动。首先利用事件流行为关系和权重散度将流程中的活动进行特征提取,得到数据流的特征集,其次进行漂移检测,动态选择特征集输入人工智能模型中训练并预测下一个活动,然后运用物联网和云计算等先进技术创建数字孪生虚拟环境,最后得到基于概念漂移的数字孪生模型。通过公开可用的数据集进行评估分析,实验结果表明,提出的方法能够有效提高预测的准确性。     

用基于中值滤波的高斯过程预测3G网络性能

针对3G网络中主动监控和对性能指标数据进行预测的需要,提出了基于中值滤波的高斯回归模型的网络性能指标预测方法,将高斯回归模型与中值滤波法相融合,对样本空间中的性能指标数据先进行中值滤波预处理,再对处理过的数据进行高斯回归预测,预测结果作为主动告警机制的预测曲线。仿真实验结果表明,相对于其他预测方法,基于中值滤波的高斯过程预测结果更加有效,生成的预测曲线更精确,为3G及以上网络进行主动监控确定更有效的阈值提供理论依据。     

基于加权移动平均的数据流预测模型*

提出一种新的基于滑动窗口的预测模型。该模型仅存储当前滑动窗口中的数据并对其进行分析,提高了计算效率;同时,为了削减在较小数据集上回归预测所产生的偏差,提出一种基于加权移动平均的数据流预测算法WMA_LRA。实验采用FDS 4.0模拟一个房屋的火灾发生情况,运用WMA_LRA算法对火灾现场的局部温度进行短期预测,结果表明该算法可以有效地提高计算效率和预测精度。     

炼焦生产过程综合生产指标的改进神经网络预测方法

针对炼焦生产过程综合生产指标 (焦炭质量、产量和焦炉能耗)检测的严重滞后问题,提出一种改进BP神经网络预测方法.首先基于相关过程参数的主元分析和灰色关联分析,确定出预测模型的输入输出变量;然后采用基于改进差分进化算法的BP神经网络建立预测模型,并与基本BP神经网络预测模型进行比较;最后,对改进BP神经网络预测模型进行了验证.实验结果表明,改进BP神经网络预测模型具有较快的收敛速度和较高的预测精度,模型的预测效果可以满足生产工艺要求.     

基于小波分析和Gaussian回归的急性低血压预测

急性低血压是危害病人健康的并发症之一,对急性低血压发生的提早预测,能够帮助医生对重症病人找到更好的医疗处理方案。提出了一个基于趋势分量的Gaussian函数拟合预测模型,即用小波多尺度分析提取出信号的趋势分量;再根据Gaussian回归模型对趋势分量进行函数拟合,得到的函数参数作为特征值,用支持向量机SVM对数据分类。Gaussian回归模型使用的是数据驱动,用系数来描述数据之间的关系。通过在较大病人数据集上实验得到了较好的效果。     

Principal component case-based reasoning ensemble for business failure prediction

Case-based reasoning (CBR) has several advantages for business failure prediction (BFP), including ease of understanding, explanation, and implementation and the ability to make suggestions on how to avoid failure. We constructed a new ensemble method of CBR that we termed principal component CBR ensemble (PC-CBR-E): it, was intended to improve the predictive ability of CBR in BFP by integrating the feature selection methods in the representation level, a hybrid of principal component analysis with its two classical CBR algorithms at the modeling level and weighted majority voting at the ensemble level. We statistically validated our method by comparing it with other methods, including the best base model, multivariate discriminant analysis, logistic regression, and the two classical CBR algorithms. The results from a one-tailed significance test indicated that PC-CBR-E produced superior predictive performance in Chinese short-term and medium-term BFP.     

基于行为轮廓矩阵增强的业务流程结果预测方法

预测性过程监控依赖于预测效果,针对如何增强预测性过程监控预测效果的问题,提出了一种基于行为轮廓矩阵增强的业务流程结果预测方法。首先,通过分析活动间的行为关系提取行为轮廓矩阵,并将其与事件序列一同输入到模型中。随后,结合卷积神经网络（CNN）和长短期记忆网络（LSTM）分别学习矩阵图像特征和序列特征。最后,引入注意力机制以整合图像特征和序列特征进行预测。通过真实事件日志进行验证,在预测事件日志结果方面,提出的增强方法对比基准的LSTM预测方法提高了预测效果,验证了方法的可行性。该方法结合行为轮廓矩阵增强了预测模型对事件日志中行为之间关系的理解,进而提升了预测效果。     

高斯过程回归模型多扩展目标多伯努利滤波器

针对复杂不确定性环境下不规则形状的多扩展目标跟踪问题, 本文提出了一种基于高斯过程回归(GPR) 模型的多扩展目标多伯努利(GPR–ETCBMeMBer)滤波算法. 首先, 在利用有限集统计理论(FISST)将多扩展目标的 状态集与量测集分别建模为多伯努利随机有限集(MBer RFS) 和泊松随机有限集(Poisson RFS) 的基础上, 通过 GPR方法建立多扩展目标随机超曲面的跟踪滤波模型. 然后, 基于容积卡尔曼滤波器(CKF)详细推导并提出GPR多 扩展目标多伯努利滤波算法的高斯混合(GM)实现. 最后, 通过构造对星凸形多扩展目标和多群目标跟踪的仿真实 验验证了本文所提算法的有效性.     

Dynamic process fault monitoring based on neural network and PCA

A newly developed method, NNPCA, integrates two data driven techniques, neural network (NN) and principal component analysis (PCA), for process monitoring. NN is used to summarize the operating process information into a nonlinear dynamic mathematical model. Chemical dynamic processes are so complex that they are presently ahead of theoretical methods from a fundamental physical standpoint. NN functions as the nonlinear dynamic operator to remove processes' nonlinear and dynamic characteristics. PCA is employed to generate simple monitoring charts based on the multivariable residuals derived from the difference between the process measurements and the neural network prediction. It can evaluate the current performance of the process. Examples from the recent monitoring practice in the industry and the large-scale system in the Tennessee Eastman process problem are presented to help the reader delve into the matter.     

过程神经网络水华预测方法研究

水华是水体富营养化的表现,会导致水体透明度下降,溶解氧降低,产生藻毒素,给人类居住环境和人体健康造成很大损害,水华已成为我国水资源保护急需解决的一个重大问题。在深入研究水华形成机理的基础上,通过化工正交实验分析和粗糙集理论,确定温度、溶解氧、叶绿素、氮磷比、总氮和光照作为水华预测的指标,叶绿素作为表征水华产生的指标,提出1种过程神经网络的水华预测模型。该模型将输入函数在给定精度下展开为1组正交基的有限项级数形式,将网络权函数表示为同1组基函数的展开形式,利用基函数的正交性来简化过程神经元对时间聚合运算的复杂性,同时通过变速率学习算法和加入动量项以提高网络的收敛速度,减少训练时所产生的振荡误差等问题。通过实验室数据的仿真,得到预测精确度为83.4%,证明本方法的有效性,为水华的预测提供1种有效途径。     

Multi-model ensemble prediction model for carbon efficiency with application to iron ore sintering process

Iron ore sintering is one of the most energy-consuming process in steel industry. Accurate prediction of carbon efficiency for this process is beneficial to energy savings and consumption reduction. Considering the sintering process exhibits strong nonlinearities, multiple parameters, multiple operating conditions, etc., a multi-model ensemble prediction model based on the actual run data is developed to achieve the high-precision prediction of carbon efficiency. It takes the comprehensive coke ratio (CCR) as a metric (index) of carbon efficiency in the sintering process. First, an affinity propagation clustering algorithm is used to realize the automatic identification of multiple operating conditions. Then, different models are established under different operating conditions by using the proposed least squares support vector machine (LS-SVM) with hybrid kernel modeling method. Finally, a partial least-squares regression method is employed as an ensemble strategy to combine the different models to form the multi-model ensemble prediction model for the CCR. The simulation results involving the actual run data demonstrate that the proposed model can predict the CCR accurately when compared with other prediction methods. The results of actual runs show that the coefficient of determination for the proposed model is 0.877. The proposed model satisfies the requirements of actual sintering process and enables the real-time prediction.