A soft sensor for industrial melt index prediction based on evolutionary extreme learning machine

In propylene polymerization (PP) process, the melt index (MI) is one of the most important quality variables for determining different brands of products and different grades of product quality. Accurate prediction of MI is essential for efficient and professional monitoring and control of practical PP processes. This paper presents a novel soft sensor based on extreme learning machine (ELM) and modified gravitational search algorithm (MGSA) to estimate MI from real PP process variables, where the MGSA algorithm is developed to find the best parameters of input weights and hidden biases for ELM. As the comparative basis, the models of ELM, APSO-ELM and GSAELM are also developed respectively. Based on the data from a real PP production plant, a detailed comparison of the models is carried out. The research results show the accuracy and universality of the proposed model and it can be a powerful tool for online MI prediction.     

Melt index prediction by RBF neural network optimized with an adaptive new ant colony optimization algorithm

Melt index (MI) is a crucial indicator in determining the product specifications and grades of polypropylene (PP). The prediction of MI, which is important in quality control of the PP polymerization process, is studied in this work. Based on RBF (radial basis function) neural network, a soft‐sensor model (RBF model) of the PP process is developed to infer the MI of PP from a bunch of process variables. Considering that the PP process is too complicated for the RBF neural network with a general set of parameters, a new ant colony optimization (ACO) algorithm, N‐ACO, and its adaptive version, A‐N‐ACO, which aim at continuous optimizing problems are proposed to optimize the structure parameters of the RBF neural network, respectively, and the structure‐best models, N‐ACO‐RBF model and A‐N‐ACO‐RBF model for the MI prediction of propylene polymerization process, are presented then. Based on the data from a real PP production plant, a detailed comparison research among the models is carried out. The research results confirm the prediction accuracy of the models and also prove the effectiveness of proposed N‐ACO and A‐N‐ACO optimization approaches in solving continuous optimizing problem. © 2010 Wiley Periodicals, Inc. J Appl Polym Sci, 2010     

Melt index prediction based on fuzzy neural networks and PSO algorithm with online correction strategy

A black‐box modeling scheme to predict melt index (MI) in the industrial propylene polymerization process is presented. MI is one of the most important quality variables determining product specification, and is influenced by a large number of process variables. Considering it is costly and time consuming to measure MI in laboratory, a much cheaper and faster statistical modeling method is presented here to predicting MI online, which involves technologies of fuzzy neural network, particle swarm optimization (PSO) algorithm, and online correction strategy (OCS). The learning efficiency and prediction precision of the proposed model are checked based on real plant history data, and the comparison between different learning algorithms is carried out in detail to reveal the advantage of the proposed best‐neighbor PSO (BNPSO) algorithm with OCS. © 2011 American Institute of Chemical Engineers AIChE J, 2012     

基于典型相关分析和数据自回归处理的BP神经网络在聚丙烯熔融指数预报中的应用

由于聚丙烯生产是一个大量参数相互耦合的强非线性过程,使得传统的机理建模受到一定的限制。提出基于典型相关分析和数据自回归处理的BP神经网络软测量建模,通过可测变量来推知聚丙烯熔融指数。应用典型相关分析选择与输出熔融指数关系较大的独立输入变量,数据自回归处理校正一系列带有误差的量测数据,而BP神经网络用来刻画过程的非线性特征。最后,将提出的算法应用到聚丙烯大型生产工艺中进行熔融指数的预报建模并进行实例仿真,仿真结果表明该算法有较强的建模精度。     

基于多尺度分析与主元分析的聚丙烯熔融指数神经元软测量预报模型

Prediction of melt index （MI）, the most important parameter in determining the product＇s grade and quality control of polypropylene produced in practical industrial processes, is studied. A novel soft-sensor model with principal component analysis （PCA）, radial basis function （RBF） networks, and multi-scale analysis （MSA） is proposed to infer the MI of manufactured products from real process variables, where PCA is carried out to select the most relevant process features and to eliminate the correlations of the input variables, MSA is introduced to a~quire much more information and to reduce the uncertainty of the system, and RBF networks are used to characterize the nonlinearity of the process. The research results show that the proposed method provides promising prediction reliability and accuracy, and supposed to have extensive application prospects in propylene polymerization processes.     

一种基于新型蚁群算法的聚丙烯熔融指数预报模型

聚丙烯熔融指数的实时预报非常重要却十分困难,提出了一种经过新型蚁群算法优化后的PCA-RBF神经网络方法进行熔融指数预报。PCA将原始数据从高维空间映射到低维空间,剔除冗余信息和提取过程特征;RBF神经网络则用来拟合输入与输出之间的非线性关系;最后用适用于连续空间寻优问题的新型蚁群算法对RBF神经网络权值进行优化。实际生产数据的研究结果,表明了所提出的熔融指数预报模型的准确性和可靠性。     

Development of a vision‐based online soft sensor for oil sands flotation using support vector regression and its application in the dynamic monitoring of bitumen extraction

Extraction from oil sands is a crucial step in the industrial recovery of bitumen. It is challenging to obtain online measurements of process outputs such as bitumen grade and recovery. Online measurements are a prerequisite for innovating better process control solutions for process efficiency and cost reduction. We have developed a soft sensor to provide online measurements of bitumen grade and recovery in a flotation‐based oil sand extraction process. Continuous froth images were captured using a VisioFroth camera system on a batch flotation unit. A support vector regression (SVR) model with a Gaussian kernel was constructed to develop a soft sensor for bitumen grade and recovery using froth image features as the inputs. The model was trained and validated for batch flotation of different grades of oil sands ore at industry‐relevant process conditions. A Dean‐Stark analyzer was used to obtain offline grade and recovery measurements that were used to calibrate the soft sensor. Mean squared errors (MSE) of 62 and 74 were achieved for grade (%) and recovery (%), respectively, and this was obtained using 5‐fold cross validation. The developed soft sensor model has been applied successfully in the real‐time dynamic monitoring of flotation grade and recovery for different grades of ore and operating conditions.

     

Melt index prediction by RBF neural network with an ICO‐VSA hybrid optimization algorithm

Melt index (MI) is considered as one of the most significant parameter to determine the quality and the grade of the practical polypropylene polymerization products. A novel ICO‐VSA‐RNN (RBF neural network with ICO‐VSA algorithm) MI prediction model is proposed based on radial basis function (RBF) neural network and improved chaos optimization (ICO), and variable‐scale analysis (VSA), where the ICO is first added and then combined with the VSA to overcome the defects of ICO and VSA, then the parameters of the RBF neural network are optimized with them. At last, the RBF neural network model for MI prediction model is developed. Further researches on the optimal RBF neural network model of MI prediction are carried out with the data from a real industrial plant, and the prediction results show that the performance of this prediction model is much better than the RBF neural network model without optimization. © 2012 Wiley Periodicals, Inc. J Appl Polym Sci, 2012     

基于GMC(1,N)的聚丙烯熔融指数预报

引言聚丙烯是以丙烯单体为主聚合而成的一种合成树脂,是五大通用塑料之一,是塑料工业中的重要原料。世界丙烯的50%、我国丙烯的65%都用来生产聚丙烯。熔融指数(简称MI)是在一定温度、一定压力、一定负荷下,熔体在10min内通过标     

Influence of low density polyethylene quality on extrusion coating processability

In order to obtain explicit information about the influence of different low density polyethylene (LDPE) quality parameters on extrusion coating processability, a test run was made with an autoclave reactor and the products were investigated. All the grades manufactured had melt indices (MI), densities, molecular weight distributions (MWD), and degrees of long chain branching(LCB) typical of commercial extrusion coating grades. The processability characteristics studied were maximum line speed and neck-in. The influence of MI, density, and extrusion melt temperature were systematically investigated. It was found that the maximum line speed rose with increasing MI, density, and extrusion melt temperature, and that an increasing extrusion melt temperature led to a growing difference between the maximum line speed at a constant coating thickness and the maximum line speed at a constant screw speed. Neck-in was found to increase with increasing MI, increasing density, and increasing coating thickness. These effects were more pronounced at higher extrusion melt temperatures. When using the extrusion temperature needed to achieve a certain line speed for each grade, the influence of MI on neck-in was practically non-existent.     

A contrastive learning-based attention generative adversarial network for defect detection in colour-patterned fabric

The pattern style of colour-patterned fabrics is varied. Defective fabric samples are scarce in the production of small batches of colour-patterned fabrics. Therefore, the unsupervised defect-detection method for colour-patterned fabric has attracted wide attention. Several unsupervised defect-detection methods for colour-patterned fabrics based on convolutional neural networks have been proposed. However, convolutional neural network methods cannot learn long-range semantic information interaction well because of the intrinsic locality of convolution operations. Besides, as the number of layers in the convolutional neural network increases, the feature maps become more and more complex. Convolutional neural networks experience difficulties in coordinating numerous parameters and extracting key features from complex feature maps. Both these problems reduce the accuracy of the model for detecting defects in colour-patterned fabrics. In this paper, we propose a Contrastive Learning-based Attention Generative Adversarial Network (CLAGAN) for defect detection in colour-patterned fabrics. The CLAGAN possesses two important parts: contrastive learning and a channel attention module. Contrastive learning captures long-range dependencies by calculating the cosine similarity between different features. The channel attention module assigns different weights to each channel of the feature maps, and it enables the model to extract key features from those feature maps. The experimental results verified the effectiveness of the CLAGAN. It obtained values of 38.25% for intersection over union and of 51.67% for the F1-measure on the YDFID-2 public dataset.     

Adaptive multiscale convolutional neural network model for chemical process fault diagnosis

Intelligent fault recognition techniques are essential to ensure the long-term reliability of manufacturing.Due to the variations in material,equipment and environment,the process variables monitored by sensors contain diverse data characteristics at different time scales or in multiple operating modes.Despite much progress in statistical learning and deep learning for fault recognition,most models are constrained by abundant diagnostic expertise,inefficient multiscale feature extraction and unr...     

Feasibility Study of a Solar Reactor for Phenol Treatment by the Photo‐Fenton process in Aqueous Solution

Solar reactors can be attractive in photodegradation processes due to lower electrical energy demand. The performance of a solar reactor for two flow configurations, i.e., plug flow and mixed flow, is compared based on experimental results with a pilot‐scale solar reactor. Aqueous solutions of phenol were used as a model for industrial wastewater containing organic contaminants. Batch experiments were carried out under clear sky, resulting in removal rates in the range of 96–100 %. The dissolved organic carbon removal rate was simulated by an empirical model based on neural networks, which was adjusted to the experimental data, resulting in a correlation coefficient of 0.9856. This approach enabled to estimate effects of process variables which could not be evaluated from the experiments. Simulations with different reactor configurations indicated relevant aspects for the design of solar reactors.     

Experimental and neural model analysis of styrene removal from polluted air in a biofilter

BACKGROUND: Biofilters are efficient systems for treating malodorous emissions. The mechanism involved during pollutant transfer and subsequent biotransformation within a biofilm is a complex process. The use of artificial neural networks to model the performance of biofilters using easily measurable state variables appears to be an effective alternative to conventional phenomenological modelling. RESULTS: An artificial neural network model was used to predict the extent of styrene removal in a perlite‐biofilter inoculated with a mixed microbial culture. After a 43 day biofilter acclimation period, styrene removal experiments were carried out by subjecting the bioreactor to different flow rates (0.15–0.9 m³ h^?1) and concentrations (0.5–17.2 g m^?3), that correspond to inlet loading rates up to 1390 g m^?3 h^?1. During the different phases of continuous biofilter operation, greater than 92% styrene removal was achievable for loading rates up to 250 g m^?3 h^?1. A back propagation neural network algorithm was applied to model and predict the removal efficiency (%) of this process using inlet concentration (g m^?3) and unit flow (h^?1) as input variables. The data points were divided into training (115 × 3) and testing set (42 × 3). The most reliable condition for the network was selected by a trial and error approach and by estimating the determination coefficient (R²) value (0.98) achieved during prediction of the testing set. CONCLUSION: The results showed that a simple neural network based model with a topology of 2–4–1 was able to efficiently predict the styrene removal performance in the biofilter. Through sensitivity analysis, the most influential input parameter affecting styrene removal was ascertained to be the flow rate. Copyright © 2009 Society of Chemical Industry     

基于Ms-LWPLS的化工过程网络化性能分级评估方法

针对化工过程输入输出数据间非线性关系问题，提出一种基于多数据空间局部加权潜结构映射(multi-space locally weighted projection to latent structures，Ms-LWPLS)的网络化性能分级评估方法。该方法将历史数据分成不同性能等级的集合，利用Ms-LWPLS方法提取不同性能等级训练数据的过程变化，获得训练数据与性能等级标签之间的非线性映射结构，实现输入数据与性能等级之间的网络化“离线建模”。得到模型后，以数据滑动时间窗为评估单元，将滑动窗口数据输入到训练好的神经网络模型中，根据网络输出划分过程当前性能等级，并构造过渡性能系数，将稳态性能等级和过渡性能等级进行识别和区分。最后，将该方法应用到乙烯裂解过程在线性能评估中，说明此性能评估方法的有效性和准确性。     

基于多数据空间全潜结构映射的化工过程性能评估方法

针对化工过程运行状态在线评估的问题,提出多数据空间全潜结构映射（multi-space total projection to latent structures,MsT-PLS）性能评估方法。该方法采用“离线建模,在线评估”的评估策略。首先对历史多数据输入空间进行全面分解,结合多数据空间基向量提取方法,剔除多数据输入空间中与质量变量无关信息的干扰。在与质量变量相关的多数据输入空间上,建立不同运行性能等级的离线数据网络分类模型,实现“离线建模”。“在线评估”阶段,以数据滑动时间窗为评估单元,将过程性能分为稳定和过渡性能等级,把在线数据与历史性能等级进行相似度匹配。利用过程变量相对贡献度,对性能变化起决定性影响的过程变量进行识别和贡献度分析,为系统性能劣化原因的识别提供了参考。最后,应用到乙烯裂解过程在线性能评估中,说明了本评估方法可以对系统进行准确的在线性能评估。     

Optimization of Whole Milk Powder Processing Variables with Neural Networks and Genetic Algorithms

Spray-dried whole milk powder, one potential ingredient of milk chocolate, was exposed to high shear and elevated temperatures to increase the free fat content and to crystallize the lactose using a twin-screw continuous mixer/processor. Optimal process conditions were determined using neural networks and genetic algorithm optimization. Response surfaces methodology was used to design the experiments to collect data for the neural network modelling. A general regression neural network model was developed to predict the responses of lactose crystallinity and free fat content from the processor screw speed, process temperature, milk powder feed rate and lecithin addition rate. A genetic algorithm was used to search for a combination of the process variables for maximum free fat content and maximum crystallinity. The combinations of the process variables during genetic algorithm optimization were evaluated using the neural network model. The common optimum process conditions to maximize the free fat content and lactose crystallinity were determined to be 20 kg h⁻¹ feed rate, 284 rpm screw speed, 71.1°C process temperature and 0.01 kg h⁻¹ lecithin addition rate.     

基于互信息的分散式动态PCA故障检测方法

对现代大型复杂动态过程来讲,不同测量变量会存在不同的序列相关性,而且变量间的相互影响会体现在不同的采样时刻上。为此,结合利用分散式建模的优势,提出一种基于互信息的分散式动态过程故障检测方法。该方法在对每个测量变量都引入多个延时测量值后,利用互信息为每个变量区分出与其相关的测量值,并建立起相应的变量子块。这种变量分块方式使每个变量子块都能充分地获取与之相对应的自相关性与交叉相关性信息,较好地处理了数据的动态性问题。然后,利用主元分析（PCA）算法对每一变量子块进行统计建模从而建立起适于大规模动态过程的多模块化的故障检测模型。最后,通过实例验证该方法用于动态过程监测的可行性和有效性。     

Quality control in the polypropylene manufacturing process: An efficient,data‐driven approach

In the propylene polymerization process, the melt index (MI), as a critical quality variable in determining the product specification, cannot be measured in real time. What we already know is that MI is influenced by a large number of process variables, such as the process temperature, pressure, and level of liquid, and a large amount of their data are routinely recorded by the distributed control system. An alternative data‐driven model was explored to online predict the MI, where the least squares support vector machine was responsible for establishing the complicated nonlinear relationship between the difficult‐to‐measure quality variable MI and those easy‐to‐measure process variables, whereas the independent component analysis and particle swarm optimization technique were structurally integrated into the model to tune the best values of the model parameters. Furthermore, an online correction strategy was specially devised to update the modeling data and adjust the model configuration parameters via adaptive behavior. The effectiveness of the designed data‐driven approach was illustrated by the inference of the MI in a real polypropylene manufacturing plant, and we achieved a root mean square error of 0.0320 and a standard deviation of 0.0288 on the testing dataset. This proved the good prediction accuracy and validity of the proposed data‐driven approach. © 2014 Wiley Periodicals, Inc. J. Appl. Polym. Sci. 2015 , 132, 41312.