Search

Highlights • A new robust nonlinear ensemble pattern trees model was developed. • A novel variable importance measure was proposed. • The new methods were applied to real operation data of a steelmaking plant. • The new methods achieved better performances than the conventional methods. Abstract In steel industry, it is crucial to predict hot metal temperature (HMT), which is strongly related to the product quality and the thermal state, to keep high productivity of the blast furnace. The present work proposes a novel ensemble pattern trees model to predict HMT. Ensemble pattern trees is a robust nonlinear modeling method, which aggregates a set of pattern trees models into a single predictive model via the bagging technique. Ensemble pattern trees overcomes the drawback of single pattern trees which may not be robust enough against the random variations such as process perturbations and noises in the blast furnace. In addition, a novel variable importance measure derived from the ensemble pattern trees is proposed to understand which process variables affect the final hot metal quality. The proposed method was validated through an industrial blast furnace ironmaking process, and the results have demonstrated its superiority to several conventional methods. [ABSTRACT FROM AUTHOR]

Independent component regression (ICR) is an efficient method for tackling non-Gaussian problems. In this work, the defects of the conventional ICR are analyzed, and a novel quality-relevant independent component regression (QR-ICR) method based on distance covariance and distance correlation is proposed. QR-ICR extracts independent components (ICs) using a quality-relevant independent component analysis (QR-ICA) algorithm, which simultaneously maximizes the non-Gaussianity of ICs and statistical dependency between ICs and quality variables. Meanwhile, two new types of statistical criteria, called cumulative percent relevance (CPR) and Max-Dependency (Max-Dep), are proposed to rank the order and determine the number of ICs according to their contributions to quality variables. The proposed QR-ICR (CPR) and QR-ICR (Max-Dep) methods were validated through a vinyl acetate monomer production process and a benchmark near-infrared spectral data. The results have demonstrated that the proposed QR-ICR (CPR) and QR-ICR (Max-Dep) provide simpler predictive models and give better prediction performances than PLS, ICR, ICR (CPR) , and ICR (Max-Dep) . [ABSTRACT FROM AUTHOR]

Soft-sensors are effective tools for predicting quality variables in many industries. In this work, a novel data-driven soft-sensor called stacked supervised Poisson autoencoders (SSPAE) is proposed to predict the number of defects in the steelmaking process. SSPAE is a deep learning-based soft-sensing model designed by integrating Poisson regression network layers into the deep autoencoders framework. In SSPAE, quality-related deep features can be progressively learned from data through the deep network architectures. During the feature extraction process, SSPAE takes the quality information into account, so that the extracted deep features are conducive to improving the accuracy of the prediction model. Additionally, due to the introduction of the Poisson regression network, SSPAE is more suitable for predicting the count-type quality variables. The proposed method is evaluated through a numerical example and real-world industrial data. The results demonstrated that SSPAE is superior to PLS, SVR, PR, SAE-FCL, and SAE-PR in prediction accuracy. • A new data-driven soft-sensor, SSPAE, was developed. • SSPAE is a deep learning model that extracts quality-related deep features of data. • SSPAE is more suitable for predicting the count-type quality variables. • SSPAE achieved better prediction performance than conventional methods. [ABSTRACT FROM AUTHOR]

Virtual sensing technology is crucial for monitoring product quality when real-time measurement is not available. To deal with both strong nonlinearity and time-varying dynamics of industrial processes, we propose a novel locally weighted kernel PLS (LW-KPLS) based on sparse nonlinear features in this research. Unlike the conventional locally weighted PLS (LW-PLS), the proposed method weights the training samples by using sparse kernel feature characterization factors (SKFCFs), which take account of the strength of nonlinear dependency between samples in the Hilbert feature space. By integrating the nonlinear features into the locally weighted regression framework, LW-KPLS not only can cope with the time-varying characteristics but also is more suitable for highly nonlinear processes. The proposed method was validated through a numerical example, a penicillin fermentation process, and a real industrial cleaning process for residual drug substances. The results have demonstrated that the proposed LW-KPLS outperforms the conventional PLS, KPLS, LW-PLS, and eLW-KPLS in the prediction performance. [ABSTRACT FROM AUTHOR]

Process uncertainties have been a big challenge to the stable operation and control of process industries. The current study is based on the use of an artificial intelligence model as a surrogate in the online optimization of process conditions of reactive units of a petroleum refinery under uncertainty. Initially, a steady-state Aspen model was used to perform exergy analysis for quantifying the exergy efficiency, irreversibility, and improvement potential of the plant. The process model was then transformed to a dynamic mode by inserting ±5 % uncertainty in process conditions, i.e., mass flow rate, pressure, and temperature, to generate a dataset of 216 samples. An artificial neural network (ANN) model was developed using the dataset to predict exergy efficiency. The ANN model was used as a surrogate in GA and PSO environments to achieve higher exergy efficiency under uncertainty. The optimized process condition derived through GA and PSO based approaches were fed to Aspen model for cross-validation. The Fourier Amplitude Sensitivity Test (FAST) was performed to identify the most influential process conditions in terms of their effect on process exergy efficiency. The overall plant had an exergy efficiency, irreversibility, and improvement potential of 50.57 %, 34,955.55 kW, and 17,276.98 kW, respectively. The correlation coefficient of ANN model was 0.97432. The use of ANN as a surrogate in the optimization frameworks outperformed the standalone (SA) Aspen model of the process in attaining the highest exergy efficiency. Overall the performance of both the GA and the PSO based approaches were comparable. Based on sensitivity analysis, inlet temperatures of reactors were the most sensitive variables affecting the process exergy efficiency. The current study helps in laying a foundation for the simulation of Refinery 4.0. [Display omitted] [ABSTRACT FROM AUTHOR]

Mathematical formulas play a prominent role in science, technology, engineering, and mathematics (STEM) documents; understanding STEM documents usually requires knowing the difference between equation groups containing multiple equations. When two equation groups can be transformed into the same form, we call the equation groups equivalent. Existing tools cannot judge the equivalence of two equation groups; thus, we develop an algorithm to judge such an equivalence using a computer algebra system. The proposed algorithm first eliminates variables appearing only in either equation group. It then checks the equivalence of the equations one by one: the equations with identical algebraic solutions for the same variable are judged equivalent. If each equation in one equation group is equivalent to an equation in the other, the equation groups are judged equivalent; otherwise, non-equivalent. We generated 50 pairs of equation groups for evaluation. The proposed method accurately judged the equivalence of all pairs. This method is expected to facilitate comprehension of a large amount of mathematical information in STEM documents. Furthermore, this is a necessary step for machines to understand equations, including process models. [ABSTRACT FROM AUTHOR]

Locally weighted partial least squares (LW-PLS) is one of Just-in-Time (JIT) modeling methods; PLS is used to build a local linear regression model every time when output variables need to be estimated. The prediction accuracy of local models strongly depends on the definition of similarity between a newly obtained sample and past samples stored in a database. To calculate the similarity, the Euclidean distance and the Mahalanobis distance have been widely used, but they do not take account of the relationship between input and output variables. This fact limits the achievable performance of LW-PLS and other locally weight regression methods. Thus, in the present work, covariance-based locally weighted PLS (CbLW-PLS) is proposed by integrating LW-PLS and a new similarity index based on the covariance between input and output variables. CbLW-PLS was applied to two industrial problems: soft-sensor design for estimating unreacted NaOH concentration in an alkali washing tower in a petrochemical process, and process analytical technology (PAT) for estimating concentration of a residual drug substance in a pharmaceutical process. The proposed similarity index was compared with six conventional indexes based on distances, correlations, or regression coefficients. The results have demonstrated that CbLW-PLS achieved the best prediction performance of all in both case studies. [ABSTRACT FROM AUTHOR]

In semiconductor manufacturing processes, virtual metrology (VM) has been investigated as a promising tool to predict important characteristics of products. Although partial least squares (PLS) is a well-known modeling technique that can cope with collinearity and therefore applied to construction of VM, its prediction performance deteriorates due to changes in process characteristics. In particular, maintenance of equipment strongly affects the process characteristics and the prediction performance. In this paper, VM was developed by using locally weighted PLS (LW-PLS), which is a type of just-in-time modeling technique, and it was used to predict the etching conversion differential of an actual dry etching process. The industrial application results have shown that the developed VM based on LW-PLS is superior to the conventional VM based on the sequential update model and artificial neural network model. In particular, it has been confirmed that the LW-PLS-based VM can keep its high prediction performance even after the maintenance, i.e., replacement of parts. [ABSTRACT FROM AUTHOR]

Input variable scaling is one of the most important steps in statistical modeling. However, it has not been actively investigated, and autoscaling is mostly used. This paper proposes two input variable scaling methods for improving the accuracy of soft sensors. One method statistically derives the input variable scaling factors; the other one uses spectroscopic data of a material whose content is estimated by the soft sensor. The proposed methods can determine the scales of the input variables based on their importance in output estimation. Thus, it can reduce the negative effects of input variables which are not related to an output variable. The effectiveness of the proposed methods was confirmed through a numerical example and industrial applications to a pharmaceutical and a distillation processes. In the industrial applications, the proposed methods improved the estimation accuracy by up to 63% compared to conventional methods such as autoscaling with input variable selection. [ABSTRACT FROM AUTHOR]

In process analytical technology (PAT) based on near infrared (NIR) spectroscopy, wavenumber selection is crucial to develop an accurate and robust calibration model. The present research proposes new efficient spectral dividing and wavenumber selection methods to significantly reduce the computational load required by conventional wavenumber selection methods such as interval partial least squares (iPLS). The proposed method, named spectral fluctuation dividing (SFD), divides a whole spectrum into multiple spectral intervals at local minimum points of the spectral fluctuation profile, which consists of the standard deviation of absorbance at each wavenumber in a calibration set. SFD is combined with PLS (SFD–PLS) to select the spectral intervals at which input variables have significant influence on a target response. The usefulness of SFD–PLS was demonstrated through its application to the problems of estimating water and drug content in granules. PLS models based on SFD–PLS achieved higher estimation accuracy than those based on conventional methods including iPLS, PLS-beta, and variable influence on projection (VIP). In addition, SFD–PLS was more than 10 times faster than the conventional variable selection methods including PLS-beta and VIP; in particular, SFD–PLS was more than 25 times faster than iPLS. Consequently, the proposed SFD–PLS is a promising wavenumber selection method. [ABSTRACT FROM AUTHOR]

Abstract: There has been a considerable research on the process analytical technology (PAT) and real-time monitoring based on NIR, but the model development is still an important issue and persons in charge have difficulty in building good models. In this study, to realize efficient NIR-based real-time monitoring of ingredient concentration and establish a model development method, we investigated the effect of a calibration set, spectral preprocessing, wavelengths, and other factors on the prediction error through pilot and commercial scale blending experiments. The results confirmed that the small prediction error was realized by a calibration set, including dynamic measurement spectra acquired with the target blender. In addition, the results demonstrated that locally weighted partial least squares (LW-PLS) achieved the smaller prediction error than conventional PLS. The present study has also clarified that spectral preprocessing methods and wavelengths selected to build a model affect the prediction error of ingredient concentration interactively. A wide wavelength range should be selected when the spectral preprocessing does not lessen the effect of baseline variation, while a narrow wavelength range should be selected when it strongly decreases the effect. [Copyright &y& Elsevier]

Highlights: [•] A new type of gray-box model is proposed to predict the tundish temperature, TD temp, in a steel making plant. [•] In addition, the proposed gray-box model is used in controlling of TD temp. [•] The results in prediction as well as control of TD temp show the advantage of the proposed gray-box models over the first-principle model, the statistical model and the conventional gray-box model. [•] The first-principle model covers the process between the secondary refining and the continuous casting process. [ABSTRACT FROM AUTHOR]

Many research works on soft-sensorshave been conducted. Although it is common practice to evaluate theestimation performance of soft-sensors by using industrial processdata, few papers have reported long-term application results of processcontrol using soft-sensors in real processes. In the present work,a practical configuration of an inferential control system was developedthat integrated a commercial model predictive control (MPC) softwareand a just-in-time (JIT) soft-sensor. The developed system has adoptedlocally weighted partial least squares (LW-PLS) to build soft-sensors.LW-PLS is a kind of JIT modeling method that can cope with changesin process characteristics as well as process nonlinearity. Thus,LW-PLS helps engineers to reduce their burden of model maintenance,which has been recognized as the most serious problem in practice.The usefulness of the developed LW-PLS-based soft-sensors and inferentialcontrol systems is demonstrated through their successful industrialapplications to a cracked gasoline (CGL) fractionator and a purificationsection for an acetyl plant. Inferential control systems have beenused for more than a year at Showa Denko K.K. (SDK) in Japan. Theoperation cost and environmental burden have been significantly reduced.In the CGL fractionator, for example, about 0.6% of operation costwas cut successfully. In addition, the present work aims to describechallenges, revealed by the long-term applications of JIT soft-sensors:the parameter tuning, the selection of input variables, the definitionof similarity in JIT modeling, the management of the database, andthe assessment and enhancement of soft-sensor reliability. [ABSTRACT FROM AUTHOR]

Abstract: Real-time monitoring techniques based on an NIR spectrometer and a thermal effusivity sensor for lubrication properties of magnesium stearate (Mg-St) are proposed. The lubrication properties of Mg-St are defined by its concentration distribution and flatting state in a mixture. The concentration distribution of Mg-St significantly affects the absorbance in the NIR wavelength region between 1128nm and 1240nm. Thus, the absorbance area in this region after baseline correction was selected as a monitoring index (MI). In laboratory-scale experiments, the difference of Mg-St concentration distribution could be detected by the proposed MI with high sensitivity. In addition, experimental results using spherical mannitol granules confirmed that the changes of the flatting state of Mg-St could also be detected by the proposed MI. Similar experiments with spherical mannitol granules and the thermal effusivity sensor confirmed that effusivity could also be used to detect the changes of the flatting state of Mg-St. The applicability of these monitoring techniques was verified using a 2000L commercial-scale blender equipped with the NIR spectrometer and the thermal effusivity sensor. The results showed that both lubrication properties could be monitored by the proposed MI, and that the flatting state of Mg-St could be monitored more sensitively by using the effusivity. [Copyright &y& Elsevier]

Various column structures, configurations, and energyconservation methods have been proposed to reduce the cost and energyconsumption in distillation sequences, such as thermally couplingor heat integration. In this work, a synthesis method of optimal separationsequences including thermally coupled columns is proposed. In theproposed method, vapor recompression of the flow to the condenserand pressure change within thermally coupled columns are taken intoaccount to increase the possibility of heat integration. Furthermore,the energy requirement (ER) was adopted as the objectivefunction as well as the total annual cost (TAC).By solving the multiobjective optimization problem, the relationshipbetween the required annual cost and the amount of energy can be explicitlyevaluated, i.e., the necessary cost increase per unit of energy inputto the sequence can easily be assessed. A superstructure approachwas adopted to enumerate all the candidate distillation sequenceswhich consisted of conventional and thermally coupled columns. Rigoroussimulations of each column were effectively used to assign reliabledata to the optimization stage. Then, the process synthesis problemwas formulated as a mixed integer linear programming (MILP) problemwhose objective is to minimize TACand ER. For two case study problems, Pareto-optimal solutions have beenderived. The results of two case studies indicated that nonconventionalsequences became the optimal for both cases. Vapor recompression waseffective for minimizing the energy requirement. [ABSTRACT FROM AUTHOR]

Abstract: Although soft-sensors have been used for estimating product quality, they do not always function well due to not only changes in process characteristics but also the individual difference of production devices. Correlation-based Just-In-Time (CoJIT) modeling has been proposed to cope with such changes in process characteristics; however it cannot deal with the individual difference. In the present work, a new pattern recognition method, referred to as the nearest correlation (NC) method is proposed to cope with the individual difference. The proposed NC method is integrated with CoJIT modeling. The advantages of the proposed methods are demonstrated through a case study. [Copyright &y& Elsevier]

Abstract: Development of quality estimation models using near infrared spectroscopy (NIRS) and multivariate analysis has been accelerated as a process analytical technology (PAT) tool in the pharmaceutical industry. Although linear regression methods such as partial least squares (PLS) are widely used, they cannot always achieve high estimation accuracy because physical and chemical properties of a measuring object have a complex effect on NIR spectra. In this research, locally weighted PLS (LW-PLS) which utilizes a newly defined similarity between samples is proposed to estimate active pharmaceutical ingredient (API) content in granules for tableting. In addition, a statistical wavelength selection method which quantifies the effect of API content and other factors on NIR spectra is proposed. LW-PLS and the proposed wavelength selection method were applied to real process data provided by Daiichi Sankyo Co., Ltd., and the estimation accuracy was improved by 38.6% in root mean square error of prediction (RMSEP) compared to the conventional PLS using wavelengths selected on the basis of variable importance on the projection (VIP). The results clearly show that the proposed calibration modeling technique is useful for API content estimation and is superior to the conventional one. [Copyright &y& Elsevier]

Abstract: The individuality of production devices should be taken into account when statistical models are designed for parallelized devices. In the present work, a new clustering method, referred to as NC-spectral clustering, is proposed for discriminating the individuality of production devices. The key idea is to classify samples according to the differences of the correlation among measured variables, since the individuality of production devices is expressed by the correlation. In the proposed NC-spectral clustering, the nearest correlation (NC) method and spectral clustering are integrated. The NC method generates the weighted graph that expresses the correlation-based similarities between samples, and the constructed graph is partitioned by spectral clustering. A new statistical process monitoring method and a new soft-sensor design method are proposed on the basis of NC-spectral clustering. The usefulness of the proposed methods is demonstrated through a numerical example and a case study of parallelized batch processes. [Copyright &y& Elsevier]

Abstract: A new quality design system for steel products using a type of Just-In-Time modeling is proposed. The present research focuses on how to determine the optimal manufacturing conditions for securing the target quality by using locally weighted regression, a type of Just-In-Time modeling, while only the quality estimation or prediction had been tackled in the past research on Just-In-Time modeling. The proposed method can improve design accuracy and shorten design time in comparison with the conventional technique. The developed system has been applied practically at JFE Steel''s West Japan Works and has contributed to reduction of quality deviations and production cost. [Copyright &y& Elsevier]

Abstract: In this age of globalization, the realization of production innovation and highly stable operation is the chief objective of the process industry in Japan. Obviously, modern advanced control plays an important role to achieve this target, but the key to success is the maximum utilization of PID control and conventional advanced control. This paper surveys how the three central pillars of process control – PID control, conventional advanced control, and linear/nonlinear model predictive control – have been used and how they have contributed toward increasing productivity. In addition to introducing eminently practical methods, emerging methods, and their applications, the authors point out challenging problems. In Japan, industry and academia are working in close cooperation to share their important problems and develop new technologies for solving them. Several methods introduced in this paper are results of such industry–academia collaboration among engineers and researchers in various companies and universities. Furthermore, soft-sensor or virtual sensor design is treated with emphasis on its maintenance, because soft-sensors must cope with changes in process characteristics for their continuous utilization. Maintenance is a key issue not only for soft-sensors but also for controllers. Finally, we will expand our scope and briefly introduce recent activities in tracking simulation and alarm management. A part of the results of our recent questionnaire survey of process control are also introduced; the results are extremely helpful in clarifying the state of the art in chemical process control in Japan. [ABSTRACT FROM AUTHOR]

Abstract: The individuality of production devices should be taken into account when soft-sensors are designed for parallelized devices. Since it is expressed as differences of the correlation among measured variables, it is useful to cluster samples on the basis of the correlation among variables for adopting a multi-model approach. In addition, changes in process characteristics can be coped with in the same way. In the present work, a new clustering method, referred to as NC-spectral clustering, is proposed by integrating the nearest correlation (NC) method and spectral clustering. Spectral clustering is a graph partitioning method that can be used for sample classification when an affinity matrix of a weighted graph is given. The NC method can detect samples that are similar to the query from the viewpoint of the correlation without a teacher signal. In the proposed method, the NC method is used for constructing the weighted graph that expresses the correlation-based similarities between samples and the constructed graph is partitioned by using spectral clustering. In addition, a new soft-sensor design method is proposed on the basis of the proposed NC-spectral clustering. The usefulness of the proposed methods is demonstrated through a numerical example and a case study of parallelized batch processes. The performance of the proposed correlation-based method is better than that of the conventional distance-based methods. [Copyright &y& Elsevier]

Abstract: Softsensors or virtual sensors are key technologies in industry because important variables such as product quality are not always measured on-line. In the present work, two-stage subspace identification (SSID) is proposed to develop highly accurate softsensors that can take into account the influence of unmeasured disturbances on estimated key variables explicitly. The proposed two-stage SSID method can estimate unmeasured disturbances without the assumptions that the conventional Kalman filtering technique must make. Therefore, it can outperform the Kalman filtering technique when innovations are not Gaussian white noises or the characteristics of disturbances do not stay constant with time. The superiority of the proposed method over the conventional methods is demonstrated through numerical examples and application to an industrial ethylene fractionator. [Copyright &y& Elsevier]

The issue of how to improve product quality and product yield in a brief period of time becomes more critical in many industries. Even though industrial processes are totally different in appearance, the problems to solve are highly similar: how to build a reliable model from a limited data, how to analyze the model and relate it to first principles, how to optimize operating condition, and how to realize an on-line monitoring and control system and maintain it. In this paper, statistical process monitoring and control methodologies are briefly surveyed, and our application results in steel facilities are presented. The achievements of the present work are as follows: (1) the development of a new method that can cope with qualitative quality information and relate operating conditions to product quality or product yield, (2) the simultaneous analysis of multiple processing units including a converter, a continuous caster, a blooming process, and rolling processes, and (3) the successful application results in the steel industry. [Copyright &y& Elsevier]

Abstract: One of the critical problems in the operation of microchemical processes is blockage in microchannels. Therefore, a process monitoring system which can detect and diagnose blockage is indispensable for effective and stable operation of microchemical processes. In the present research, two types of diagnosis systems, a data-based blockage diagnosis system (DB-BDS) and a model-based blockage diagnosis system (MB-BDS), are proposed. To realize efficient diagnosis for various degrees of blockage, the proposed DB-BDS uses the ratios of temperature differences between normal and abnormal operating conditions at one sensor to those at the other sensor. On the other hand, MB-BDS, which uses simple physical models of the process, diagnoses blockage by calculating the correlation coefficients between the prepared temperature distribution vectors and the actual temperature distribution vector when blockage is detected. The performance of the proposed systems is evaluated with their applications to a stacked microreactor. DB-BDS and MB-BDS are applied to various types of blockage, i.e., various locations and degrees. The results show that both DB-BDS and MB-BDS can diagnose the blockage location successfully even when blockage degree is less than 10%. [Copyright &y& Elsevier]

Univariate and multivariate statistical process control (USPC and MSPC) methods have been widely used in process industries for fault detection. However, their practicability and achievable performance are limited due to the assumptions that a continuous process is operated in a particular steady state and that variables are normally distributed. In the present work, external analysis is proposed to distinguish faults from normal changes in operating conditions. In addition, to further improve the monitoring performance, a new MSPC method based on independent component analysis (ICA) is used. The simulation results of a linear multivariable system and a CSTR process have clearly shown the superiority of ICA-based SPC over USPC and PCA-based SPC, and also the usefulness of the proposed external analysis. [Copyright &y& Elsevier]

This article is concerned with a classical problem of how to design an inferential control system for enhancing the performance of distillation composition control. In this article, a new inferential control, termed “predictive inferential control,” is proposed. In the predictive inferential control system, future compositions predicted from on-line measured process variables are controlled instead of the estimates of current compositions. The key concept of the predictive inferential control is to realize feedback control with a feedforward effect by the use of the inherent nature of a distillation column. The detailed dynamic simulation results show that the proposed predictive inferential control scheme integrated with cascade control works considerably better than other control schemes. Furthermore, the improvement of the control performance through iterative modeling or control-relevant identification is also demonstrated. [Copyright &y& Elsevier]

To improve the performance of multivariate statistical process control (MSPC), two advanced methods, moving principal component analysis (MPCA) and DISSIM, have been proposed. In MPCA and DISSIM, an abnormal operation can be detected by monitoring the directions of principal components and the dissimilarity index, respectively. Another important extension of MSPC was made with Multiscale PCA (MS-PCA). The present work investigates the characteristics of several statistical monitoring methods. The monitoring performance is compared with applications to simulated data obtained from a 2×2 process and the Tennessee Eastman process. The superiority of MPCA and DISSIM over the conventional methods comes from the fact that those methods focus on changes in the distribution of process data. Furthermore, the advantages of MPCA or DISSIM over the conventional MSPC and that of MS-PCA are combined, and new methods, termed MS-MPCA and MS-DISSIM, are proposed. [Copyright &y& Elsevier]

The paper addresses a closed-loop identification method based on generalized minimum variance (GMV) evaluation. Since the proposed method uses routine operation data, it requires no extra experiment with an external excitation signal. The model parameters of the plant and the disturbance are obtained simultaneously using a single set of input–output data generated by stochastic disturbance. A new variance criterion for closed-loop identification is derived through the conversion of the GMV evaluation function that has originally been developed for data-driven regulatory control. In the conversion, the feedback invariant polynomial, which is estimated from time series analysis of the closed-output signal, plays a key role. The features of the proposed approach lead to bridge closed-loop identification with control performance assessment as well as data-driven controller parameters tuning. The paper proves that the optimization of the proposed criterion results in the unique optimal solution, which corresponds to the true plant and disturbance model parameters. In numerical examples, the proposed method is applied to datasets obtained from a continuous stirred tank reactor (CSTR), which is operated around an unstable steady state. The result illustrates the effectiveness of the proposed closed-loop identification method. • Closed-loop identification method derived from data-driven regulatory control. • Simultaneously obtains the model parameters of the plant and the disturbance. • Use routine operation data and require no extra experiment for identification. • Bridge identification with performance assessment as well as data-driven control. • Analyze a new variance criterion and prove existence of the unique optimal solution. [ABSTRACT FROM AUTHOR]

In the steel industry, defects may occur during the manufacturing process. Thus, it is important to predict the occurrence of defects online in steel products and identify the causal variable that may lead to defects. However, the unique characteristics of the observed defect count data, such as nonnegative integers and high overdispersion, have posed some difficulties to the traditional probability models. To deal with this issue, the present work employs random forests to model and analyze the observed defect count data. Random forests are a nonlinear ensemble learning technique, which constructs several regression trees during the training phase and then predicts the output by averaging the predictions of each tree. Unlike the traditional probability models which are based on the specific distribution assumption, random forests are a non-parametric or distribution-free model. Furthermore, random forests can ensure the nonnegativity of the prediction, and thus it is suitable for defect count data modeling. In addition, partial dependence analysis in conjunction with the variable importance measure was used to identify the causal variable. The application results on the real steelmaking process have demonstrated that random forests outperform the PLS, SVR, Poisson, and NB methods in prediction accuracy. And the most influential variables identified by random forests are in line with operator experience. • Prediction and causal analysis of defects in steel products were investigated. • Defect data are characterized by nonnegative integers and high overdispersion. • Random forest (RF) outperforms other linear/nonlinear methods in prediction accuracy. • The most influential variables identified by RF are in line with operator experience. [ABSTRACT FROM AUTHOR]

• In blast furnace ironmaking, it is crucial to predict hot metal temperature (HMT). • Deep and shallow predictive methods were applied to an industrial blast furnace. • Both the current time and multi-step-ahead HMT predictions were investigated. • The prediction performance and computational time of ten methods were evaluated. To realize stable operation of the ironmaking process, it is important to predict hot metal temperature (HMT) in a blast furnace. Recently, deep learning is emerging as a highly active area of research. Nonetheless, no thorough study has yet appeared comparing the performance of deep learning methods to the shallow learning methods in predicting HMT. This paper provides a comparative study on the deep and shallow predictive methods for the current time and multi-step-ahead HMT predictions. Three advanced deep predictive methods and seven effective shallow predictive methods are investigated from the application point of view. Both the deep and shallow predictive methods were applied to an industrial blast furnace, where the prediction performance and computational time of ten methods were evaluated. The results demonstrated that (1) shallow neural network is preferred for current time HMT prediction; (2) Gaussian process regression and support vector regression are preferred for multi-step-ahead HMT predictions. [ABSTRACT FROM AUTHOR]

The sustainability of petroleum-based fuel supply has gained broad attention from the global community due to the increase of usage in various sectors, depletion of petroleum resources, and uncertain around crude oil market prices. Additionally, environmental problems have also arisen from the increasing emissions of harmful pollutants and greenhouse gases. Therefore, the use of clean energy sources including biodiesel is crucial. Biodiesel is mainly produced from unlimited natural resources through a transesterification process. It presents various advantages over petro-diesel; for instance, it is non-toxic, biodegradable, and contains less air pollutant per net energy produced with low sulphur and aromatic content, apart from being safe. Considering the importance of this topic, this paper focuses on the use of palm oil, its by-products, and mill effluent for biodiesel production. Palm oil is known as an excellent raw material because biodiesel has similar properties to the regular petro-diesel. Due to the debate on the usage of palm oil as food versus fuel, extensive studies have been conducted to utilise its by-products and mill effluent as raw materials. This paper also discusses the properties of biodiesel, the difference between palm-biodiesel and other biodiesel sources, and the feasibility of using palm oil as a primary source for future alternative and sustainable energy sources. [ABSTRACT FROM AUTHOR]

• Sleep apnea syndrome (SAS) screening AI-based on R-R interval data was validated with a large clinical polysomnography dataset. • AUC of 0.92, a sensitivity of 0.80 and a specificity of 0.84 were achieved. • The SAS screening algorithm is easy to implement into a smartphone app. Easily detecting patients with undiagnosed sleep apnea syndrome (SAS) requires a home-use SAS screening system. In this study, we validate a previously developed SAS screening methodology using a large clinical polysomnography (PSG) dataset (N = 938). We combined R-R interval (RRI) and long short-term memory (LSTM), a type of recurrent neural networks, and created a model to discriminate respiratory conditions using the training dataset (N = 468). Its performance was validated using the validation dataset (N = 470). Our method screened patients with severe SAS (apnea hypopnea index; AHI ≥ 30) with an area under the curve (AUC) of 0.92, a sensitivity of 0.80, and a specificity of 0.84. In addition, the model screened patients with moderate/severe SAS (AHI ≥ 15) with an AUC of 0.89, a sensitivity of 0.75, and a specificity of 0.87. Our method achieved high screening performance when applied to a large clinical dataset. Our method can help realize an easy-to-use SAS screening system because RRI data can be easily measured with a wearable heart rate sensor. It has been validated on a large dataset including subjects with various backgrounds and is expected to perform well in real-world clinical practice. [ABSTRACT FROM AUTHOR]

This study develops a novel data-driven latent variable correlation analysis (LVCA) framework to achieve communication efficient distributed monitoring for industrial plant-wide processes. Process data of a local unit are first projected into a dominant latent variable subspace and a residual subspace to characterize the correlation within the local unit. Then, least absolute shrinkage and selection operator is used to determine communication variables from neighboring units that are beneficial for monitoring the local unit. Thereafter, canonical correlation analysis is performed between the dominant subspace and communication variables to characterize the correlation between units. Finally, a distributed monitor is established for each unit, which considers the correlation within the local unit and the correlation between different operation units. The proposed LVCA-based distributed monitoring scheme is applied on a numerical example, the Tennessee Eastman benchmark process, and a lab-scale distillation process. Comparison results with some state-of-the-art methods verify the effectiveness. Note to Practitioners—In the monitoring of a local operation unit, it is important to characterize the relationship among variables within the local unit and the relationship between the local unit and its neighboring units. However, not all variables from neighboring units are beneficial for the monitoring. Including nonbeneficial variables may cause considerable communication cost and model interpretation difficulty. Here a novel latent variable correlation analysis (LVCA)-based distributed local monitoring method, which considers simultaneous correlation within the local unit and between units, is proposed. The LVCA-based distributed monitoring preserves the fault detection ability and is more computationally efficient than the existing stochastic optimization-based methods, and therefore is more suitable for practical application. The superiority and characteristics are theoretically discussed and experimentally studied. MATLAB code is available upon request. [ABSTRACT FROM AUTHOR]

This study proposes an efficient wavenumber selection method to develop calibration models based on near-infrared (NIR) spectroscopy. First, spectral fluctuation dividing (SFD) divides a whole NIR spectrum into multiple spectral intervals based on a spectral fluctuation profile, which consists of the standard deviation of spectral intensities at each wavenumber. Then, nearest correlation spectral clustering (NCSC) clusters those spectral intervals into spectral interval groups based on the correlation of the spectral intensities among the spectral intervals. Finally, the proposed method builds a partial least squares (PLS) model using the spectral intensities in each spectral interval group, and selects several spectral interval groups based on the estimation accuracy of each PLS model. This method was named SFD–NCSC–PLS. In developing calibration models to estimate water and drug contents in granules, SFD–NCSC–PLS achieved higher estimation accuracy than the commonly-used interval PLS, searching combination moving window PLS, and the methods using either SFD or NCSC. The results show that SFD–NCSC–PLS can properly select wavenumbers that reflect the target response. In addition, SFD–NCSC–PLS took only less than half the computation time compared with the wavenumber selection methods using either SFD or NCSC. Thus, the proposed SFD–NCSC–PLS is a promising wavenumber selection method. [ABSTRACT FROM AUTHOR]

Background Currently, in the era of post-genomics, immunology is facing a challenging problem to translate mutant phenotypes into gene functions based on high-throughput data, while taking into account the classifications and functions of immune cells, which requires new methods. Results Here we propose a novel application of a multidimensional analysis, Canonical Correspondence Analysis (CCA), to reveal the molecular characteristics of undefined cells in terms of cellular differentiation programmes by analysing two transcriptomic datasets. Using two independent datasets, whether RNA-seq or microarray data, CCA successfully visualised the cross-level relationships between genes, cells, and differentiation programmes, and thereby identified the immunological features of mutant cells (Gata3-KO T cells and Stat3- KO T cells) in a data-oriented manner. With a new concept, differentiation variable, CCA provides an automatic classification of cell samples, which had a high sensitivity and a comparable performance to other classification methods. In addition, we elaborate how CCA results can be interpreted, and reveal the features of CCA in comparison with other visualisation techniques. Conclusions CCA is a visualisation tool with a classification ability to reveal the cross-level relationships of genes, cells and differentiation programmes. This can be used for characterising the functional defect of cells of interest (e.g. mutant cells) in the context of cellular differentiation. The proposed approach fits with common hypothesis-oriented studies in immunology, and can be used for a wide range of molecular and genomic studies on cellular differentiation mechanisms. [ABSTRACT FROM AUTHOR]

Abstract: Recently, just-in-time (JIT) modeling, such as locally weighted partial least squares (LW-PLS), has attracted much attention because it can cope with changes in process characteristics as well as nonlinearity. Since JIT modeling derives a local model from past samples similar to a query sample, it is crucial to appropriately define the similarity between samples. In this work, a new similarity measure based on the weighted Euclidean distance is proposed in order to cope with nonlinearity and to enhance estimation accuracy of LW-PLS. The proposed method can adaptively determine the similarity according to the strength of the nonlinearity between each input variable and an output variable around a query sample. The usefulness of the proposed method is demonstrated through numerical examples and a case study of a real cracked gasoline fractionator of an ethylene production process. [Copyright &y& Elsevier]

Recent developments in energy conservationhave sought internalheat integration through side heat exchangers as an attractive energy-efficientalternative to reduce the energy consumption in conventional distillation.However, it is difficult to estimate the effect on the reduction ofreboiler duty through internal heat integration without executingrigorous simulations. Thus, the stages used for heat exchange havebeen selected by trial and error or some heuristics. For a fixed designwhere heat integration between a rectifying section and a strippingsection takes place, current research has derived the optimal structureby trial-and-error or systematic approaches. In this work, the combinationof superstructure representation, rigorous simulations, and mathematicalprogramming is effectively used to develop a systematic synthesisprocedure, which can derive the optimal distillation process structureincluding the internal and external heat integration. In the proposedprocedure, piece-wise linear functions between the amount of internalheat exchange and the reduction of reboiler and condenser heat dutiesare derived in advance by executing rigorous simulations. Then, forevery pair of rectifying and stripping sections of two columns, theoptimal structure with internal heat integration is derived by solvingan MILP problem. The cost of side heat exchangers, and the condenserand reboiler duties and areas resulted from internal heat integrationare used as parameters in the formulation of a synthesis problem whichcomprises all the possible distillation sequences. By adopting sucha procedure, the overall synthesis problem can also be formulatedas an MILP problem, and the optimal process structure can be derivedby solving it. The result of a case study to separate a ternary mixtureshows the effectiveness of the proposed synthesis method. [ABSTRACT FROM AUTHOR]

The understanding of pathological processes is based on the comparison between physiological and pathological conditions, and transcriptomic analysis has been extensively applied to various diseases for this purpose. However, the way in which the transcriptomic data of pathological cells relate to the transcriptomes of normal cellular counterparts has not been fully explored, and may provide new and unbiased insights into the mechanisms of these diseases. To achieve this, it is necessary to develop a method to simultaneously analyse components across different levels, namely genes, normal cells, and diseases. Here we propose a multidimensional method that visualises the cross-level relationships between these components at three different levels based on transcriptomic data of physiological and pathological processes, by adapting Canonical Correspondence Analysis, which was developed in ecology and sociology, to microarray data (CCA on Microarray data, CCAM). Using CCAM, we have analysed transcriptomes of haematological disorders and those of normal haematopoietic cell differentiation. First, by analysing leukaemia data, CCAM successfully visualised known relationships between leukaemia subtypes and cellular differentiation, and their characteristic genes, which confirmed the relevance of CCAM. Next, by analysing transcriptomes of myelodysplastic syndromes (MDS), we have shown that CCAM was effective in both generating and testing hypotheses. CCAM showed that among MDS patients, high-risk patients had transcriptomes that were more similar to those of both haematopoietic stem cells (HSC) and megakaryocyte-erythroid progenitors (MEP) than low-risk patients, and provided a prognostic model. Collectively, CCAM reveals hidden relationships between pathological and physiological processes and gene expression, providing meaningful clinical insights into haematological diseases, and these could not be revealed by other univariate and multivariate methods. Furthermore, CCAM was effective in identifying candidate genes that are correlated with cellular phenotypes of interest. We expect that CCAM will benefit a wide range of medical fields. [ABSTRACT FROM AUTHOR]

Abstract: Soft-sensors have been widely used for estimating product quality or other key variables, and partial least squares (PLS) regression is accepted as a useful technique for soft-sensor design. To achieve high estimation performance, it is important to select appropriate input or explanatory variables. The present work proposes a new systematic methodology to select input variables for PLS using nearest correlation spectral clustering (NCSC), which is a clustering method based on the correlation among variables. The proposed method, referred to as NCSC-based variable selection (NCSC-VS), clusters the variables into some variable classes by using NCSC, and selects a few variable classes according to their contribution to estimates. That is, the input variables are not selected individually but some variables that have similar correlation are selected together. The usefulness of the proposed NCSC-VS is demonstrated through an application to soft-sensor design for an industrial chemical process. [Copyright &y& Elsevier]

Machine Learning (ML) is one of the major driving forces behind the fourth industrial revolution. This study reviews the ML applications in the life cycle stages of biofuels, i.e., soil, feedstock, production, consumption, and emissions. ML applications in the soil stage were mostly used for satellite images of land to estimate the yield of biofuels or a suitability analysis of agricultural land. The existing literature have reported on the assessment of rheological properties of the feedstocks and their effect on the quality of biofuels. The ML applications in the production stage include estimation and optimization of quality, quantity, and process conditions. The fuel consumption and emissions stage include analysis of engine performance and estimation of emissions temperature and composition. This study identifies the following trends: the most dominant ML method, the stage of life cycle getting the most usage of ML, the type of data used for the development of the ML-based models, and the frequently used input and output variables for each stage. The findings of this article would be beneficial for academia and industry-related professionals involved in model development in different stages of biofuel's life cycle. [ABSTRACT FROM AUTHOR]

• A gray-box model of the industrial Czochralski (CZ) process was developed. • The model was built for the CZ process producing 300 mm diameter single-crystal Si. • The model predicts a crystal radius, a crystal growth rate, and heater temperature. • Computing a temperature gradient by a statistical model improves prediction accuracy. • The results demonstrated the high prediction accuracy of the gray-box model. More than 95% of 300 mm diameter single-crystal silicon ingots, the raw material for semiconductors, are produced by the Czochralski process. The demand for improving yield, throughput, and control performance has been increasing. The present study developed a gray-box model that can predict controlled variables from manipulated variables with higher accuracy than the conventional first-principle model (Zheng et al., 2018), aiming at realizing model predictive control of the Czochralski process. The proposed gray-box model used a statistical model to predict the temperature gradient of the crystal at the solid–liquid interface G cry , which was constant in the first-principle model. The crystal length and the melt temperature are used as the input variables to predict G cry . The prediction accuracy of the proposed gray-box model was compared with that of the first-principle model using real process data obtained during the production of four silicon ingots. The results demonstrated that the proposed model reduced the root mean square errors of the crystal radius, the crystal growth rate, and the heater temperature by 94.1%, 62.7%, and 70.6% on average, respectively. [ABSTRACT FROM AUTHOR]

This study is based on gray-box modeling for the prediction of optimum mass flow rates of inlet streams of a Plate and Fin Heat Exchanger under uncertainty. A first principle model of the Plate and Fin Heat Exchanger was developed in Aspen Exchanger Design and Rating. Genetic algorithm was integrated with the first principle model to achieve the highest possible exit temperature of the cold process stream under uncertainty. A dataset of uncertain process conditions and their corresponding optimum inlet flow rates derived through the first principle-genetic algorithm integration was used to develop an artificial neural networks model. The artificial neural networks model was then integrated with the first principle model by replacing the genetic algorithm to form a novel gray-box framework. The proposed gray box model, i.e., artificial neural networks and first principle integration, achieved a higher effectiveness and higher outlet temperature than those derived through the straight run first principle model. The performance of the gray box framework was also comparable to the first principle integrated genetic algorithm approach and significantly minimized the computation time needed for estimating the optimum process conditions. First principle integrated genetic algorithm approach enhanced the effectiveness of the straight run model by 3.05%. Performance of the proposed gray-box model was comparable to the integrated framework of the genetic algorithm with the first principle model but was significantly faster. The developed artificial neural networks model was employed as a surrogate in Sobol and FAST sensitivity analysis framework to identify the impact of input variables on output variables. The proposed gray box based method enhanced the capability of the plate and fin heat exchanger to recover energy from the process stream and its robustness to cope with uncertainty. The proposed approach is suitable for real-time application and would contribute to laying a foundation for petroleum refinery 4.0. • A Gray-box (GB) modeling framework was developed for the optimization of process conditions of a plate and fin heat exchanger of a tile industry under uncertainty. • The GB framework was based on the integration of first principle (FP) model and Artificial Neural Networks (ANN) model. • The proposed GB approach attained the optimum conditions in a significantly less period of time than the conventional FP-GA framework. [ABSTRACT FROM AUTHOR]

Soft sensors have been used for estimation of research octane number (RON) of gasoline and to realize stability in process operation. In this study, a novel framework is proposed by incorporating uncertainty analysis in the conventional soft-sensor design for RON prediction. Ensemble learning method, i.e., boosting, and artificial neural networks (ANN) were used to realize highly efficient soft sensor. The individual impact of uncertainty of each process variable on RON was analyzed using Sobol technique and Fourier Amplitude Sensitivity Test (FAST). The accumulative effect of uncertainties on RON was investigated through nonintrusive polynomial chaos expansion (PCE). For realizing reliability in online application of the predictive distribution of RON, control limits were derived using a Just-in-Time method, i.e., k-d tree. The proposed method was validated through process data where the results were found suitable for industrial applications. [ABSTRACT FROM AUTHOR]

Soft sensors are mathematical models employed to estimate hard-to-measure variables from available easy-to-measure variables. These sensors are typically developed using either model-driven algorithms or data-driven counterparts. Data-driven algorithms are preferable since developing soft sensors via model-driven algorithms is in most cases unprofitable, technically difficult, or impossible. Soft sensors are desirable to be adaptive so that they can cope with the state and characteristics changes in industrial processes. Among the existing adaptive methods, just-in-time (JIT) models are gaining popularity due to their advantages of dealing with nonlinear, and time-varying processes with abrupt changes processes. In this paper, the existing JIT-based algorithms adopted to formulate adaptive soft sensors are critically reported and discussed. Furthermore, the limitations of these JIT-based algorithms are also highlighted. These JIT-based models are generally categorized into Gaussian and non-Gaussian distributed algorithms. A short discussion on algorithms considering nonlinear and missing data is also presented. Lastly, recommendations and future directions on JIT-based algorithms are outlined. It is envisaged that this manuscript will serve as a more comprehensive and up-to-date review and provide general guidelines for developing JIT models as soft sensors. • Reviews of JIT-based models are presented. • These models are critically discussed. • Future research directions are highlighted. [ABSTRACT FROM AUTHOR]

[Display omitted] • A low-dimensional design space was constructed based on high-dimensional data. • Dimensionality reduction of the design space was achieved by reproducing the observed internal representations in the data for design space construction. • The expected correlation of process parameters, material attributes of intermediates, and the quality attributes at the target manufacturing condition were visualized. Implementation of the design space (DS) is a scientific concept for ensuring quality to be submitted as a part of the regulatory filing of a drug product for approval to market. An empirical approach is constructing the DS based on the regression model whose inputs are process parameters and material attributes over the different unit operations, i.e., a high-dimensional statistical model. While the high-dimensional model assures quality and process flexibility through a comprehensive process understanding, it has difficulty visualizing the feasible range of input parameters, i.e., DS. Therefore, this study proposes a greedy approach to constructing the extensive and flexible low-dimensional DS based on the high-dimensional statistical model and the observed internal representations that satisfies both comprehensive process understanding and the DS visualization capability. Introducing the observed correlation structure enabled the dimensionality reduction of the DS. The non-critical controllable parameters were fixed to the target values in visualizing the low-dimensional DS as a function of critical parameters. The expected variation of non-critical non-controllable parameters was considered the source of variation in prediction. The case study demonstrated the proposed approach's usefulness for developing the pharmaceutical manufacturing process. [ABSTRACT FROM AUTHOR]

Steam methane reforming (SMR) is a dominant technology for hydrogen production. For the highly energy-efficient operation, robust energy analysis is crucial. In particular, exergy analysis has received the attention of researchers due to its advantage over the conventional energy analysis. In this work, an exergy analysis based on the computational fluid dynamics (CFD)-based method was applied to a monolith microreactor of SMR. Initially, a CFD model of SMR was developed using literature data. Then, the design and operating conditions of the microreactor were optimized based on the developed CFD model to achieve higher conversion efficiency and shorter length. Exergy analysis of the optimized microreactor was performed using the custom field function (CFF) integrated with the CFD environment. The optimized catalytic monolith microreactor of SMR achieved higher conversion efficiency at a smaller consumption of energy, catalyst, and material of construction than the reactor reported in the literature. The exergy analysis algorithm helped in evaluating length-wise profiles of all three types of exergy, namely, physical exergy, chemical exergy, and mixing exergy, in the microreactor. [ABSTRACT FROM AUTHOR]

The crude distillation unit (CDU) is one of the most energy-intensive processes of a petroleum refinery. The composition of crude is subject to change on regular basis. The uncertainty in crude oil composition causes wastage of a substantial amount of energy in the CDU operation. In this study, a novel approach based on a multi-output artificial neural networks (ANN) model was devised to cope with variations (uncertainty) in crude composition. The proposed method is an extended version of another method of cut-point optimization based on hybridization of Taguchi and genetic algorithm. A data comprised of several hundred variations of crude compositions and their optimized cut point temperatures, derived from the hybrid approach, was used to train the ANN model. The proposed method was validated on a simulated CDU flowsheet for a Pakistani crude, i.e., Zamzama. The proposed method is faster and computationally less expensive than the hybrid method. In addition, it can efficiently predict optimum cut point temperatures for any variant of the crude composition. [ABSTRACT FROM AUTHOR]

Objective: Driver drowsiness detection is a key technology that can prevent fatal car accidents caused by drowsy driving. The present work proposes a driver drowsiness detection algorithm based on heart rate variability (HRV) analysis and validates the proposed method by comparing with electroencephalography (EEG)-based sleep scoring. Methods: Changes in sleep condition affect the autonomic nervous system and then HRV, which is defined as an RR interval (RRI) fluctuation on an electrocardiogram trace. Eight HRV features are monitored for detecting changes in HRV by using multivariate statistical process control, which is a well known anomaly detection method. Result: The performance of the proposed algorithm was evaluated through an experiment using a driving simulator. In this experiment, RRI data were measured from 34 participants during driving, and their sleep onsets were determined based on the EEG data by a sleep specialist. The validation result of the experimental data with the EEG data showed that drowsiness was detected in 12 out of 13 pre-N1 episodes prior to the sleep onsets, and the false positive rate was 1.7 times per hour. Conclusion: The present work also demonstrates the usefulness of the framework of HRV-based anomaly detection that was originally proposed for epileptic seizure prediction. Significance: The proposed method can contribute to preventing accidents caused by drowsy driving. [ABSTRACT FROM AUTHOR]

Abstract To reduce the heat variance and realize the stable operation of a blast furnace, an operation guidance system was developed. This system involves nonlinear model predictive control based on a new transient model and provides appropriate guidance to operators. Accurate future prediction is a key to success because the huge heat capacity of the blast furnace leads to slow, complex dynamics with long dead time. The online validation results in the actual furnace showed that the standard deviation of the hot metal temperature was reduced by 3.7 °C when the operators followed the guidance provided by the developed system. Highlights • A new transient model for the prediction of hot metal temperature was developed. • Hot metal temperature of 8 hours ahead was predicted successfully with the model. • Model predictive control based on the developed transient model was established. • Variance of the hot metal temperature was reduced by the predictive control. [ABSTRACT FROM AUTHOR]

The Czochralski (CZ) process is the dominant method for manufacturing large cylindrical single-crystal ingots for the electronics industry. Although many models and control methods for the CZ process have been proposed, they were only tested with small equipment and only a few industrial application were reported. In this research, we constructed a first-principle model for controlling industrial CZ processes that produce 300 mm single-crystal silicon ingots. The developed model, which consists of energy, mass balance, hydrodynamic, and geometrical equations, calculates the crystal radius and the crystal growth rate as output variables by using the heater input, the crystal pulling rate, and the crucible rise rate as input variables. To improve accuracy, we modeled the CZ process by considering factors such as changes in the positions of the crucible and the melt level. The model was validated with the operation data from an industrial 300 mm CZ process. We compared the calculated and actual values of the crystal radius and the crystal growth rate, and the results demonstrated that the developed model simulated the industrial process with high accuracy. [ABSTRACT FROM AUTHOR]