Actual Industrial Data Research Articles

Fault identification for quality monitoring of molten iron in blast furnace ironmaking based on KPLS with improved contribution rate

Blast furnace (BF) ironmaking is one of the most important production links in modern iron–steel making. The operation conditions of BF and the molten iron quality (MIQ) should be monitored and analyzed in real-time to realize high quality with low energy consumption. Aiming at the problems of strong nonlinearity and few fault samples in BF processes, a novel fault identification method for MIQ monitoring based on kernel partial least squares (KPLS) with improved contribution rate is proposed in this paper. First of all, a KPLS model is established with the actual historical data in order to detect the quality-related faults accurately. The T2 and SPE statistics are used to monitor the operation conditions of process from different aspects. Second, in view of the unclear physical meaning and complex computation of the existing fault identification methods based on KPLS with contribution rate, a scale factor vector is introduced into the new samples to calculate the T2 and SPE statistics, so as to construct the monitoring indicators functions. By performing Taylor approximation on these constructed functions near the scale factor with the value of 1, two new statistics are obtained from the absolute value of the first-order partial derivative representing the contribution rate of each variable. Finally, in order to further improve the effect of fault identification, the relative contribution rate of each variable is used to identify the final faulty variables. The tests of MIQ monitoring in BF ironmaking process using actual industrial data verify the validity and practicability of the proposed method.

Data-Driven Monitoring and Diagnosing of Abnormal Furnace Conditions in Blast Furnace Ironmaking: An Integrated PCA-ICA Method

Principal component analysis (PCA) and independent component analysis (ICA) have been widely used for process monitoring in process industry. Since the operation data of blast furnace (BF) ironmaking process contain both non-Gaussian distribution data and Gaussian distribution data, the above single PCA or ICA method hardly describes the data distribution information of the BF process completely, which makes the monitoring and diagnosis of abnormal working-conditions only with a single method prone to false positives and false negatives. In this article, a novel integrated PCA-ICA method is proposed for monitoring and diagnosing the abnormal furnace conditions in BF ironmaking by comprehensively considering and combining the characteristics of PCA and ICA. First, the process monitoring models of PCA and ICA are, respectively, established using the actual industrial BF data, while both them are using T 2 and squared prediction error statistics to monitor whether the process is abnormal. Based on this, in order to fully reveal the internal structure of actual BF ironmaking data, an integrated PCA-ICA strategy and algorithm is proposed for comprehensively monitoring and diagnosing the abnormal furnace conditions. The corresponding unified contribution charts indices and control limits for fault identification were also presented. Finally, data experiments using actual industrial BF data show that the proposed method can obtain good results in both monitoring and diagnosing the abnormal furnace conditions of BF ironmaking.

Robust stochastic configuration network multi-output modeling of molten iron quality in blast furnace ironmaking

Blast furnace ironmaking (BFI) is currently the most widely used method of pig iron smelting. In order to achieve efficient and reasonable control, how to quickly and accurately obtain the molten iron quality (MIQ) model is a key issue. Aiming at this problem, this paper applies robust stochastic configuration networks (RSCNs) based on kernel density estimation (KDE) into the BFI modeling to obtain the MIQ model with good modeling accuracy and strong robustness quickly and effectively. Firstly, the network model is incrementally constructed by adding neurons one by one using the conventional SCNs algorithm. Secondly, in order to solve the problem of insufficient robustness of conventional SCNs, kernel density estimation algorithm is introduced to obtain the corresponding probability density estimates of each training set, and it's used as the penalty weight introduced into constructing process of conventional SCNs. At the same time, the network output weight is obtained by an improved method to solve the problem that the output weight of the conventional RSCNs is abnormal in the multi-output modeling application. Finally, modeling experiments based on actual industrial data of BFI production verified that RSCNs can achieve good modeling accuracy and strong robust performance.

A Dynamic Grey-Box Model and its Application in the Sintering Process of Ternary Cathode Material

Soft-sensor technique is often used to estimate key variables in industrial manufacturing, of which the commonly used approaches as the first-principle modeling and data-driven modeling both have their limitations. To take full advantage of the modeling methods and overcome the problems of unmodeled dynamics in industrial manufacturing, a grey-box modeling method combining the first-principle analysis with dynamic data-driven model is developed in this paper. In the framework of the presented grey-box model, the unmodeled dynamics in the first-principle model are obtained by the dynamic probabilistic latent variable model. On this basis, availability of models can be improved. Finally, the actual industrial data in the roller kiln of ternary cathode material manufacturing is used for simulation to verify the validity of the model. The results have practical guiding significance.

Recursive Learning-Based Bilinear Subspace Identification for Online Modeling and Predictive Control of a Complicated Industrial Process

In this paper, a recursive learning based bilinear subspace identification (R-B-SI) algorithm is proposed for online modeling and data-driven predictive control of blast furnace (BF) ironmaking process with strong nonlinear time-varying dynamics. Different from the existing linear SI algorithms, the R-B-SI algorithm can make full use of the process data information by adding the Kronecker product term of input data and the Kronecker product term between input and output data into the data block Hankel matrices. Moreover, the parameters of the R-B-SI model can be updated online with the latest data using recursive learning. In order to adjust the sensitivity of the algorithm to the data at different sampling times, the forgetting factor is introduced into the involved recursive learning as well. Therefore, the nonlinear time-varying dynamics of process can be well captured and described by the R-B-SI model. As a result, the data-driven predictive controller designed by the R-B-SI prediction model with parameter adaption is always capable of achieving satisfactory control performances regardless of the variation in various process operating conditions. Data experiments using actual industrial data have verified the practicability and superiority of the proposed methods.

Prediction of Plate Crown during Aluminum Hot Flat Rolling by Finite Element Modeling

The roll deflection during hot rolling can result in uneven thickness distribution across the width of a plate (crown). A conventional rolling mill is equipped with bending systems that can control this convex shape of the plate. However, the determination of the proper bending load is very complicated as the plate crown is influenced by the rolling conditions. In this paper, a thermo-mechanical Finite Element Model on LS-DYNA™ software was utilized to predict crown evolution based on the rolling conditions in order to determine the setting values for achieving the target crown. The simulation results were compared and verified with actual industrial data for rolling force, plate temperature and plate crown. This approach is essential for pass schedule design and process parameter optimization in order to achieve the desired product quality.

An integrated prediction model of heavy metal ion concentration for iron electrocoagulation process

Abstract In the electrocoagulation process of the heavy metal wastewater treatment, the acquisition of the heavy metal ions concentration at outlet requires long-term analysis, resulting in delayed control of the process and many other continuing problems. This study focuses on establishing the prediction model of heavy metal ions concentration for electrocoagulation process. Based on the mechanistic analysis of the electrode reaction and the adsorption kinetics, a novel kinetics model of the electrocoagulation process is proposed. Then, an industrial condition classification and equalization method, as well as a metaheuristic method named state transition algorithm are introduced to optimize the unknown model parameters. After that, a data-based ANNs-ARIMA model is applied to compensate the errors to overcome the shortcomings of the kinetics model caused by the industrial process uncertainties. Finally, an integrated prediction model of heavy metal ion concentration is established by connecting the kinetics model and the compensation model in parallel. All the experimental results showed that the predicted values of the kinetics model roughly agree with the actual industrial data, but the prediction accuracy of integrated model reaches to 90% and the mean relative errors is within ± 8.2 % , which indicates that the integrated model is highly accurate and has better predictive capacity. Especially when the industrial conditions fluctuate violently, the advantages of the integrated model can be more demonstrated.

Soft-Sensor Model for Chemical Processes Based on D-Vine Copula with Rolling Pin Transformation

Soft-sensing methods have been widely used in recent years to predict key variables that are difficult to measure or involve costly and time-consuming in chemical processes. Owing to the increasing complexity of industrial processes, industrial data often exhibit strong nonlinearities and self-correlation, and the data distribution fails to satisfy the Gaussian assumption. To address these problems, a vine copula-based soft-sensor model combined with the rolling pin method is proposed. This approach uses a D-vine copula to model a joint probability distribution of auxiliary variables and a key variable. In accordance with the joint distribution, the predicted value of the key variable can be determined by weighting the training samples. During modeling, the Bayesian information criterion is adopted to select the best-fitted copula pairs. Given the nonmonotonicity between variables of actual industrial data, the rolling pin monotonic transformation is simultaneously introduced to improve the suitability of...

Life Cycle Impact Assessment of Polylactic Acid (PLA) Produced from Sugarcane in Thailand

In this study we provide up-to-date cradle-to-gate information on the environmental footprint of polylactic acid (PLA) produced in Thailand at commercial scale, covering emerging topics such as water footprint and direct land use change. The enormous potential to further reduce the environmental impacts of PLA through improvements in feedstock production as well as in the PLA manufacturing process is also demonstrated. Life cycle assessment (LCA) is performed according to the ISO 14040/44 standard methodology. The 16 environmental impact categories from ILCD 2011 Midpoint + were considered for the hotspot analysis. As primary data actual industrial data were used for the sugar production, lactic acid production (Corbion) and PLA production (Total Corbion PLA), including various recently developed process insights. The agricultural feedstock production and the manufacturing process of PLA from sugar contributed most to the LCA impacts of PLA production. The sugarcane crop production particularly affected the environmental impact categories analyzed, including global warming potential (GWP), water, eutrophication, acidification, particulate matter and, inevitably, land use. However, when combined with the results of a sustainability risk assessment study, it becomes clear that land use and water-related impacts represent a low risk for the feedstock-sourcing area. The environmental impact categories of PLA manufacturing are mostly linked to energy and chemicals usage. Improvements in the environmental performance of PLA can be achieved through improvements in the sugarcane farming practices, higher efficiency bagasse boilers at the sugarmill, reduced usage of auxiliary chemicals and increased usage of renewable energy in the conversion process of sugar to PLA. From a cradle-to-gate perspective, considering the uptake of carbon dioxide in the PLA molecule, the GWP is 501 kg CO2 eq/ton PLA.

Digital Twin for Monitoring of Industrial Multi-Effect Evaporation

Digital twins are rigorous mathematical models that can be used to represent the operation of real systems. This connection allows for deeper understanding of the actual states of the analyzed system through estimation of variables that are difficult to measure otherwise. In this context, the present manuscript describes the successful implementation of a digital twin to represent a four-stage multi-effect evaporation train from an industrial sugar-cane processing unit. Particularly, the complex phenomenological effects, including the coupling between thermodynamic and fluid dynamic effects, and the low level of instrumentation in the plant constitute major challenges for adequate process operation. For this reason, dynamic mass and energy balances were developed, implemented and validated with actual industrial data, in order to provide process information for decision-making in real time. For example, the digital twin was able to indicate failure of process sensors and to provide estimates for the affected variables in real time, improving the robustness of the operation and constituting an important tool for process monitoring.

Data modeling for quality prediction using improved orthogonal incremental random vector functional-link networks

Innumerable complex industrial processes, such as blast furnace (BF) ironmaking process, greatly rely on some well-performing quality models to realize prediction and control of product quality. This paper proposes an improved orthogonal incremental random vector functional-link networks (I-OI-RVFLNs) algorithm with compact structure and high computational efficiency and applies it to quality modeling of BF ironmaking process. First, in order to improve the convergence speed of the basic I-RVFLNs, Schmidt orthogonalization method is introduced to orthogonalize the output vectors of the hidden layer nodes obtaining the least squares solution. Then, the proposed I-OI-RVFLNs algorithm further improves the network construction mode of the existing I-RVFLNs simplifying the network structure. To cut out the hidden nodes with smaller output weights, the number of hidden nodes of the proposed method is first fixed in advance and then the output weights of each hidden node are iteratively updated to approximate the network output. In this way, the problem of low computational efficiency caused by too many hidden nodes in basic I-RVFLNs can be avoided in the proposed algorithm. Lastly, data experiments using actual industrial data show that the proposed algorithm has much better performance in terms of convergence speed and computational efficiency than other compared algorithms.

A distributed PCA-TSS based soft sensor for raw meal fineness in VRM system

Vertical roller mill (VRM) is an increasingly popular comminution equipment in cement plants. Raw meal fineness at the outlet of VRM is one of the most important indicators to measure product quality. A soft sensing model developed for powder fineness in real-time could assist operators to monitor the comminution process online. However, due to frequent fluctuation of raw material properties, intrinsic nonlinearity of the process and changeable operation conditions, the data present a multimodal characteristic. Therefore, this paper proposes an indicator to measure the similarity between variables, that is, the shortest distance between nodes in the constructed weighted network. By combining the Girvan Newman (GN) algorithm, the nodes in the variable network are divided into multiple groups, and based on this, the distributed PCA (DPCA) similarity is adopted for time series segmentation (TSS). Compared with traditional similarities between samples (distance, density, etc.), the similarity between time series focuses more on the dynamic characteristics of variables. And the implementation of DPCA similarity is equivalent to increasing the sparse characteristics of principal components, which is beneficial for enhancing the generalization of the model. Support vector regression (SVR) models along with a support vector machine (SVM) based classifier are built to obtain final predictions of the powder fineness. Effectiveness of the proposed method is verified by actual industrial data. It has a root mean square error (RMSE) index of 0.3451 on the test set, which is much smaller than that of the multi-modal soft sensor based on either clustering approaches (0.6524) or PCA similarity based TSS method (0.4282).

Robust Online Sequential RVFLNs for Data Modeling of Dynamic Time-Varying Systems With Application of an Ironmaking Blast Furnace.

By dealing with robust modeling and online learning together in a unified random vector functional-link networks (RVFLNs) framework, this paper presents a novel robust online sequential RVFLNs for data modeling of dynamic time-varying systems together with its application for a blast furnace (BF) ironmaking process. First, to overcome the difficulties caused by the nonlinear time-varying dynamics of process and to enable the RVFLNs to learn online and to avoid data saturation, an improved online sequential version of RVFLNs (OS-RVFLNs) is presented by sequential learning with forgetting factor. It has been shown that the improved OS-RVFLNs with forgetting factor is not only suitable for the large-scale and real-time data transfer situation but also can adjust the sensitivity of the algorithm to different samples. Second, in order to solve the issue of modeling robustness when the dataset is contaminated with various outliers, a Cauchy distribution function weighted M-estimator is introduced to strengthen the robustness of the improved OS-RVFLNs. The non-Gaussian Cauchy distribution function is used to estimate the weights of different data and thus the corresponding contribution on modeling can be properly distinguished. Experiments using actual industrial data of a large BF ironmaking process have demonstrated that the proposed algorithm produces a much stronger robustness and better estimation accuracy than other algorithms.

Real-Time Weighted Data Fusion Algorithm for Temperature Detection Based on Small-Range Sensor Network.

Biological oxidation pretreatment, which can improve the yield of gold, is the main gold extraction technology for disposing refractory gold ore with high arsenic and sulfur. The temperature of the oxidation tank influences the oxidation efficiency between the ore pulp and bacteria, including the yield of gold. Therefore, measurement has consistently been an important subject for researchers. As an effective data processing method, data fusion has been used extensively in many fields of industrial production. However, the interference of equipment or external factors such as the diurnal temperature difference or powerful wind may constantly increase measurement errors and damage certain sensors, which may transmit error data. These problems can be solved by following a pretreatment process. First, we establish a heat transfer mechanism model. Second, we design a small-range sensor network for the pretreatment process and present a layered fusion structure of sharing sensors using a multi-connected fusion structure. Third, we introduce the idea of iterative operation in data processing. In addition, we use prior data for predicting state values twice in order to improve the effectiveness of extended Kalman filtering in one time step. This study also proposes multi-fading factors on the basis of a weighted fading memory index to adjust the prediction error covariance. Finally, the state estimation accuracy of each sensor can be used as a weighting principle for the predictive confidence of each sensor by adding a weighting factor. In this study, the performance of the proposed method is verified by simulation and compared with the traditional single-sensor method. Actual industrial measurement data are processed by the proposed method for the equipment experiment. The performance index of the simulation and the experiment shows that the proposed method has a higher global accuracy than the traditional single-sensor method. Simulation results show that the accuracy of the proposed method has a 55% improvement upon that of the traditional single-sensor method, on average. In the equipment experiment, the accuracy of the industrial measurement improved by 37% when using the proposed method.

Investigation of the efficiency of sorption-enhanced methanol synthesis process in circulating fast fluidized bed reactors

The concept of circulating fast fluidization with in-situ water removal using zeolite-4A particles is introduced to the methanol synthesis process. The mathematical model of a circulating fast fluidized bed reactor (CFFBR) is validated by an actual industrial plant data. Also, the industrial plant feed data is used to test the performance of a bubbling fluidized bed reactor (FBR). The CFFBR with varying adsorbent compositions is investigated. The system exhibits optimum conditions at a low feed temperature of 201.75 °C due to water removal and thermodynamic equilibrium shift. The methanol production has substantially increased about 14-folds. The increase of the adsorbent mass concentration >50.0% has a little effect on the maxima due to the total conversion of carbon. It is interesting to note that feeds rich in CO2 and lean in CO are susceptible to a significant improvement of methanol production. The sensitivity analysis shows that the increase of H2 in the feed decreases CO formation and consequently increases the carbon conversion to methanol. Moreover, the pressure is more effective at low ranges due to a high carbon conversion. The high efficiency shown by the CFFBR in this study suggests a promising application of these reactor generations in the methanol industry.

Data-driven predictive control of molten iron quality in blast furnace ironmaking using multi-output LS-SVR based inverse system identification

Abstract Control of blast furnace (BF) ironmaking process has always been a hot and difficult issue in metallurgic engineering and automation. In this paper, a novel data-driven inverse system identification based predictive control method is proposed for multivariate molten iron quality (MIQ) indices in BF ironmaking process. First, since the widely used least-square support regression (LS-SVR) algorithm cannot cope with the multi-output problem directly, this paper uses multi-task transfer learning technology to construct a novel multi-output LS-SVR (M-LS-SVR) for multivariable nonlinear systems. Then, this M-LS-SVR is adopted to identify the inverse system model of the controlled BF ironmaking process with the help of the presented modeling performance comprehensive evaluation and NSGA II based multi-objective parameter optimization. In order to better perform the control of the MIQ indices, the identified inverse system model is used to compensate the controlled nonlinear BF system to a compounded pseudo linear system with linear transitive relation. Such an inverse system based data-driven predictive control can effectively improve the control performance of the conventional nonlinear predictive control. Data experiments using actual industrial data from a large BF show that the proposed methods are effective, advanced and practical, and provide a solution to the operational control and optimization of the BF ironmaking process.

Active learning based support vector data description method for robust novelty detection

Practical industrial data usually has non-Gaussian data distribution and nonlinear variable correlation. Because support vector data description (SVDD) has no Gaussian limitations and can be extended to the nonlinear case by applying the kernel trick, it is one of the most widely used novelty detection methods. However, there is a great deal of actual industrial data that are mixed with much noise and uncertainty. Furthermore, SVDD may perform worse when the amount of data is too large and the data quality is poor. Describing the whole distribution with a small number of labeled samples has great practical significance and research value. This paper proposed an active learning-based SVDD method for robust novelty detection. It can reduce the amount of labeled data using an active learning framework, generalize the distribution of data and reduce the impact of noise by using the local density to guide the selection process. Experiments on two-dimensional synthetic distributions, UCI datasets and the Tennessee Eastman Process (TEP) show the effectiveness of the proposed method.

Reagent Predictive Control Using Joint Froth Image Feature for Antimony Flotation Process

The prerequisite for achieving automatic control and optimal operation in flotation process is to extract distinctive froth image features. With the fluctuation of flotation conditions, the morphological features of surface bubbles will change. It is not reliable to impose bubble size distribution as the decisive character to represent the complex flotation conditions. Therefore, a novel joint froth image feature was proposed via analysing bubble size and shape simultaneously. To consist with the observation process of operators in plant and overcome the deficiency of bubble samples, splicing froth images from sequential frames were segmented to obtain the joint distribution and a nonparametric estimation using B-spline functions was introduced to depict the shape of distribution. Then a multi-output least square support vector regressor (MLS-SVR) was implemented to construct the relationship between the weights of the B-spline functions and the reagent dosage. Finally, a joint feature based reagent predictive controller was constructed to validate the proposed method. Actual industrial data experiment results have shown its effectiveness and feasibility.

Multiphase equilibrium modeling of oxygen bottom-blown copper smelting process

Abstract A computational thermodynamics model for the oxygen bottom-blown copper smelting process (Shuikoushan, SKS process) was established, based on the SKS smelting characteristics and theory of Gibbs free energy minimization. The calculated results of the model show that, under the given stable production condition, the contents of Cu, Fe and S in matte are 71.08%, 7.15% and 17.51%, and the contents of Fe, SiO2 and Cu in slag are 42.17%, 25.05% and 3.16%. The calculated fractional distributions of minor elements among gas, slag and matte phases are As 82.69%,11.22%, 6.09%, Sb 16.57%, 70.63%, 12.80%, Bi 68.93%, 11.30%, 19.77%, Pb 19.70%, 24.75%, 55.55% and Zn 17.94%, 64.28%, 17.79%, respectively. The calculated results of the multiphase equilibrium model agree well with the actual industrial production data, indicating that the credibility of the model is validated. Therefore, the model could be used to monitor and optimize the industrial operations of SKS process.

Interval Type-2 Fuzzy Model Based on Inverse Controller Design for the Outlet Temperature Control System of Ethylene Cracking Furnace

Multivariable coupling, nonlinear and large time delays exist in the coil outlet temperature (COT) control system of the ethylene cracking furnace, which make it hard to achieve accurate control over the COT of the furnace in actual production. To solve these problems, an inverse controller based on an interval type-2 fuzzy model control strategy is introduced. In this paper, the proposed control scheme is divided into two parts: one is the approach structure part of the interval type-2 fuzzy model (IT2-FM), which is utilized to approach the process output. The other is the interval type-2 fuzzy model inverse controller (IT2-FMIC) part, which is utilized to control the output process to achieve the target value. In addition, on the cyber-physical system platform, the actual industrial data are used to test and obtain the mathematical model of the COT control system of the ethylene cracking furnace. Finally, the proposed inverse controller based on the IT2-FM control scheme has been implemented on the COT control system of the ethylene cracking furnace, and the simulation results show that the proposed method is feasible.