Multivariate Process Monitoring Research Articles

A hybrid root transformation and decision on belief approach to monitor multiattribute Poisson processes

Most of industrial applications of statistical process control involve more than one quality characteristics to be monitored. These characteristics are usually correlated, causing challenges for the monitoring methods. These challenges are resolved using multivariate quality control charts that have been widely developed in recent years. Nonetheless, multivariate process monitoring methods encounter a problem when the quality characteristics are of the attribute type and follow nonnormal distributions such as multivariate binomial or multivariate Poisson. Since the data analysis in the latter case is not as easy as the normal case, more complexities are involved to monitor multiattribute processes. In this paper, a hybrid procedure is developed to monitor multiattribute correlated processes, in which number of defects in each characteristic is important. Two phases are proposed to design the monitoring scheme. In the first phase, the inherent skewness of multiattribute Poisson data is almost removed using a root transformation technique. In the second phase, a method based on the decision on belief concept is employed. The transformed data obtained from the first phase are implemented on the decision on belief (DOB) method. Finally, some simulation experiments are performed to compare the performances of the proposed methodology with the ones obtained using the Hotelling T2 and the MEWMA charts in terms of in-control and out-of-control average run length criteria. The simulation results show that the proposed methodology outperforms the other two methods.

Diagnosing Problems of Distribution-Free Multivariate Control Chart

In order to solve the problem of only have a few historical data that can be used in multivariate process monitoring, a new distribution-free multivariate control chart has been proposed. And in the control chart structure the control limits are determined on-line with the future observations and the historical data. Therefore, the proposed control chart has very important application in practice. However, the research doesn’t study the problem of the fault diagnosis after the control chart alarms. So we use LASSO-based diagnostic framework to identify when a detected shift has occurred and to isolate the shifted components.

Reference set selection with generalized orthogonal Procrustes analysis for multivariate statistical process monitoring of multiple production processes

Multivariate process monitoring is important in industry to ensure that production processes perform as close as possible to optimal operation. However, the selection of a reference set of optimal or expected performance is required for efficient process monitoring in real time. In this paper we present the method of generalized orthogonal Procrustes analysis to select a reference set for the multivariate monitoring of multiple production processes simultaneously. We combine generalized orthogonal Procrustes analysis with principal component analysis (PCA) and biplots to illustrate the implementation of the method and the interpretation of the results which provide important information on the relationships between many process variables and differences between the production processes. The work is motivated by an industrial problem involving the multivariate monitoring of a coal gasification production facility considering many process variables monitored across multiple reactors.

Relationships between PCA and PLS-regression

This work aims at comparing several features of Principal Component Analysis (PCA) and Partial Least Squares Regression (PLSR), as techniques typically utilized for modeling, output prediction, and monitoring of multivariate processes. First, geometric properties of the decomposition induced by PLSR are described in relation to the PCA of the separated input and output data (X-PCA and Y-PCA, respectively). Then, analogies between the models derived with PLSR and YX-PCA (i.e., PCA of the joint input–output variables) are presented; and regarding to process monitoring applications, the specific PLSR and YX-PCA fault detection indices are compared. Numerical examples are used to illustrate the relationships between latent models, output prediction models, and fault detection indices. The three alternative approaches (PLSR, YX-PCA and Y-PCA plus X-PCA) are compared with regard to their use for statistical modeling. In particular, a case study is simulated and the results are used for enhancing the comprehension of the PLSR properties and for evaluating the discriminatory capacity of the fault detection indices based on the PLSR and YX-PCA modeling alternatives. Some recommendations are given in order to choose the more appropriate approach for a specific application: 1) PLSR and YX-PCA have similar capacity for fault detection, but PLSR is recommended for process monitoring because it presents a better diagnosing capability; 2) PLSR is more reliable for output prediction purposes (e.g., for soft sensor development); and 3) YX-PCA is recommended for the analysis of latent patterns imbedded in datasets.

Online monitoring of nonlinear multivariate industrial processes using filtering KICA–PCA

In this paper, a novel approach for processes monitoring, termed as filtering kernel independent component analysis–principal component analysis (FKICA–PCA), is developed. In FKICA–PCA, first, a method to calculate the variance of independent component is proposed, which is significant to make Gaussian features and non-Gaussian features comparable and to select dominant components legitimately; second, Genetic Algorithm is used to determine the kernel parameter through minimizing false alarm rate and maximizing detection rate; furthermore, exponentially weighted moving average (EWMA) scheme is used to filter the monitoring indices of KICA–PCA to improve monitoring performance. In addition, a novel contribution analysis scheme is developed for FKICA–PCA to diagnosis faults. The feasibility and effectiveness of the proposed method are validated on the Tennessee Eastman (TE) process.

Reconstruction‐based fault prognosis for flue gas turbines with independent component analysis

ABSTRACTOnline detection and prognosis are very important for the safe operation of flue gas turbines. Compared with univariate monitoring of the process, multivariate process monitoring is more effective and can capture abnormal situation in the early stage. This paper proposes a new multivariate fault prognosis framework for the flue gas turbine with a hidden fault process based on independent component analysis (ICA). ICA is a statistical method for identifying underlying independent factors or components in multivariate data. First of all, the non‐Gaussian measurements are modeled by the ICA model, and three indices are used for fault detection. Once several faulty samples are collected, fault directions can be extracted for the reconstruction. Then, a reconstruction‐based method is proposed to estimate the fault magnitude with the most sensitive index. At last, the fault magnitude is predicted by the support vector machine model. Test results for K‐103 catalytic flue gas turbine clearly showed that the proposed fault prognosis method is more efficient than traditional methods. The results also demonstrate support vector machine method has advantages over auto regression approach. © 2013 Curtin University of Technology and John Wiley & Sons, Ltd.

Multivariate process monitoring and fault identification using multiple decision tree classifiers

Machine learning based algorithms, such as a decision tree (DT) classifier, have been applied to automated process monitoring and fault identification in manufacturing processes, however the current DT-based process control models employ a single DT classifier for both mean shift detection and fault identification. As many manufacturing processes use automated data collection for multiple process parameters, a DT classifier would have to handle a large number of classes. Previous research shows that a large number of classes can degrade the accuracy of a DT multiclass classifier. In this study we propose a new process monitoring model using multiple DT classifiers with each handling a small number of classes. Moreover, we not only detect mean shifts but also identify process variability levels that may cause out-of-control signals. Experimental results show that our proposed model achieves satisfactory performance in process monitoring and fault identification with various parameter settings. It achieves better ARL performance compared with the baseline method based on a single DT classifier.

A Framework for Multivariate Process Monitoring and Diagnosis

Monitoring and diagnosis of mean shifts in manufacturing processes become more challenging when involving two or more correlated variables. Unfortunately, most of the existing statistical process control frameworks are only effective in shift detection but suffers high false alarm, that is, imbalanced performance monitoring. The problem becomes more complicated when dealing with small shift particularly in identifying the causable variables. In this research, a framework to address balanced monitoring and accurate diagnosis was investigated. Design considerations involved extensive simulation experiments to select input representation based on raw data and statistical features, recognizer design structure based on synergistic model, and monitoring-diagnosis approach based on two stages technique. The study focuses on correlated process mean shifts for cross correlation function, ρ = 0.1 ~ 0.9 and mean shift, μ = ± 0.75 ~ 3.00 standard deviations. The proposed design, that is, an Integrated Multivariate Exponentially Weighted Moving Average with Artificial Neural Network gave superior performance, namely, average run length, ARL1 = 3.18 ~ 16.75 (for out-of-control process), ARL0 = 452.13 (for in-control process) and recognition accuracy, RA = 89.5 ~ 98.5%. This research has provided a new perspective in realizing balanced monitoring and accurate diagnosis of multivariate correlated process mean shifts.

Modeling and performance monitoring of multivariate multimodal processes

A multimodal modeling and monitoring approach based on maximum likelihood principal component analysis and a component‐wise identification of operating modes are presented. Analyzing each principal component individually allows separating components describing the variation within the individual modes from those capturing variation which the modes commonly share. On the basis of the former set, a Gaussian mixture model produces a statistical fingerprint that describes the production modes. The advantage of the component‐wise analysis is a simple identification of the mixture model parameters, which does not rely on the computationally cumbersome expectation maximization. The proposed method diagnoses abnormal process conditions by defining statistics relating to the components describing (1) between‐cluster variation, (2) within cluster variation, and (3) model residuals. The article demonstrates the benefits of this approach over existing work by an application to a continuous stirred tank reactor (CSTR) simulator and the analysis of recorded data from a furnace and a chemical reaction process. © 2012 American Institute of Chemical Engineers AIChE J, 59: 1557–1569, 2013

A Variable-Selection-Based Multivariate EWMA Chart for Process Monitoring and Diagnosis

Fault detection and root cause identification are both important tasks in Multivariate Statistical Process Control (MSPC) for improving process and product quality. Most traditional control charts, including Hotelling's T2 chart and the Multivariate Exponential Weighted Moving Average (MEWMA) chart, separate the two tasks into independent and successive procedures by signaling the existence of process faults followed by auxiliary methods to locate root causes. This paper proposes an integrated procedure, a Variable-Selection-based MEWMA (VS-MEWMA) chart, for multivariate process monitoring and fault diagnosis by utilizing dimensionality reduction techniques. The VS-MEWMA chart first locates potentially out-of-control variables via variable selection and then deploys such information in the monitoring statistics with the reduction in dimensionality providing increased sensitivity to out-of-control conditions. When a signal is given, the algorithm also identifies the suspected variables for further root cause diagnosis. Both numerical simulations and real examples are presented to illustrate the performance of the proposed chart, as well as design guidelines.

Directional Control Schemes for Multivariate Categorical Processes

We consider statistical process control of multivariate categorical processes and propose a Phase II log-linear directional control chart that exploits directional shift information and integrates the monitoring of multivariate categorical processes into the unified framework of multivariate binomial and multivariate multinomial distributions. We also suggest a diagnostic scheme for identifying the shift direction. Both the control chart and the diagnostic approach are simple and easily computed. Numerical simulations and real applications are presented to demonstrate their effectiveness.

Quality Relevant Data-Driven Modeling and Monitoring of Multivariate Dynamic Processes: The Dynamic T-PLS Approach

In data-based monitoring field, the nonlinear iterative partial least squares procedure has been a useful tool for process data modeling, which is also the foundation of projection to latent structures (PLS) models. To describe the dynamic processes properly, a dynamic PLS algorithm is proposed in this paper for dynamic process modeling, which captures the dynamic correlation between the measurement block and quality data block. For the purpose of process monitoring, a dynamic total PLS (T-PLS) model is presented to decompose the measurement block into four subspaces. The new model is the dynamic extension of the T-PLS model, which is efficient for detecting quality-related abnormal situation. Several examples are given to show the effectiveness of dynamic T-PLS models and the corresponding fault detection methods.

Integration of support vector machines and control charts for multivariate process monitoring

Statistical process control tools have been used routinely to improve process capabilities through reliable on-line monitoring and diagnostic processes. In the present paper, we propose a novel multivariate control chart that integrates a support vector machine (SVM) algorithm, a bootstrap method, and a control chart technique to improve multivariate process monitoring. The proposed chart uses as the monitoring statistic the predicted probability of class (PoC) values from an SVM algorithm. The control limits of SVM-PoC charts are obtained by a bootstrap approach. A simulation study was conducted to evaluate the performance of the proposed SVM–PoC chart and to compare it with other data mining-based control charts and Hotelling's T 2 control charts under various scenarios. The results showed that the proposed SVM–PoC charts outperformed other multivariate control charts in nonnormal situations. Further, we developed an exponential weighed moving average version of the SVM–PoC charts for increasing sensitivity to small shifts.

Multivariate process monitoring and analysis based on multi-scale KPLS

In the paper, a new multi-scale KPLS (MSKPLS) algorithm combining kernel partial least square (KPLS) and wavelet analysis is proposed for investigating the multi-scale nature of nonlinear process. The MSKPLS first decomposes the process measurements into separated multi-scale components using on-line wavelet transform, and then the resultant multi-scale data blocks are modeled in the framework of multi-block KPLS algorithm which can describe the global relationships across the entire scales as well as the localized features within each scale. To demonstrate the feasibility of the MSKPLS method, its process monitoring abilities were tested for a real industrial data set, and compared with the monitoring abilities of the standard KPLS method. The results clearly showed that the MSKPLS was superior to the standard KPLS, especially in that it could provide additional scale-level information about the fault characteristics as well as more sensitive fault detection ability.

Dynamic latent variable modeling for statistical process monitoring

AbstractDynamic principal component analysis (DPCA) has been widely used in the monitoring of dynamic multivariate processes. In traditional DPCA, the dynamic relationship between process variables are implicit and hard to interpret. To extract explicit latent factors that are dynamically correlated, a new dynamic latent variable model is proposed. The new structure can improve modeling of dynamic data and enhance the process monitoring performance. Fault detection indices are developed based on the proposed model. A case study is given to illustrate the effectiveness of the proposed new dynamic factor model.

Multivariate process monitoring of an experimental blast furnace

AbstractProcess monitoring by use of multivariate projection methods has received increasing attention as it can reduce the monitoring problem for richly instrumented industrial processes with many correlated variables. This article discusses the monitoring and control of a continuously operating experimental blast furnace (EBF). A case study outlines the need for monitoring and control of the EBF and the use of principal components (PCs) to monitor the thermal state of the process. The case study addresses design, testing and online application of PC models for process monitoring. The results show how the monitoring problem can be reduced to following just a few PCs instead of many original variables. The case study highlights the problem of multivariate monitoring of a process with frequently shifting operating modes and process drifts and stresses the choice of a good reference data set of ‘normal’ process behavior. Possible solutions for adaptations of the multivariate models to process changes are also discussed. Copyright © 2009 John Wiley & Sons, Ltd.

Applied advanced process analytics in biopharmaceutical manufacturing: Challenges and prospects in real-time monitoring and control

Biopharmaceutical manufacturing processes are inherently complex due to their nonlinear bioprocess dynamics, variability in batch operations and manufacturing schedule, raw materials involved, and automatic process control. A typical processed lot generates large amounts of data that need to be analyzed and interpreted for process troubleshooting and continuous improvement purposes in addition to product release. Multivariate batch process modeling, monitoring and control approaches in real-time are elaborated by providing industrial examples from the commercial manufacturing processes. Examples and opportunities in cell culture (e.g., bioreactor applications) and purification (e.g., large-scale chromatography) operations are summarized. Impact of process analytical technologies (PAT), soft-sensor development, first principles modeling applications and commercial-scale examples are presented.

Support Vector Regression control charts for multivariate nonlinear autocorrelated processes

Statistical process control charts are one of the most widely used techniques in industry and laboratories that allow monitoring of systems against faults. To control multivariate processes, most classical charts need to model process structure and assume that variables are linearly and independently distributed. This study proposes to use a nonparametric method named Support Vector Regression to construct several control charts that allow monitoring of multivariate nonlinear autocorrelated processes. Also although most statistical quality control techniques focused on detecting mean shifts, this research investigates detection of different parameter shifts. Based on simulation results, the study shows that, with a controlled robustness, the charts are able to detect the different applied disturbances. Moreover in comparison to Artificial Neural Networks control chart, the proposed charts are especially more effective in detecting faults affecting the process variance.

Integrating independent component analysis and support vector machine for multivariate process monitoring

This study aims to develop an intelligent algorithm by integrating the independent component analysis (ICA) and support vector machine (SVM) for monitoring multivariate processes. For developing a successful SVM-based fault detector, the first step is feature extraction. In real industrial processes, process variables are rarely Gaussian distributed. Thus, this study proposes the application of ICA to extract the hidden information of a non-Gaussian process before conducting SVM. The proposed fault detector will be implemented via two simulated processes and a case study of the Tennessee Eastman process. Results demonstrate that the proposed method possesses superior fault detection when compared to conventional monitoring methods, including PCA, ICA, modified ICA, ICA–PCA and PCA–SVM.

MULTIVARIATE ANALYSIS AND STATISTICAL MONITORING OF APPLE FRUIT DEVELOPMENT ON THE TREE

In management of food production and processing, multivariate data analysis and process monitoring enhance the efficiency to maintain nutritional product quality and diminish the loss due to the decaying along the supply chain of perishable agrofood products. The limitation of traditional univariate methods in process monitoring is overcome by multivariate techniques, since the latter uses the correlation among the various variables in developing the monitoring statistics and hence, enables the visualization and interpretation of the 'big picture' about the process status. Multivariate statistical process monitoring (SPM) techniques are based on statistical methods such as principal components analysis (PCA). In this paper, PCA based multivariate techniques are applied to SPM in the site-specific monitoring of apples using nondestructively recorded apple spectra recorded during fruit development on tree. The fruit quality changes in time are used as normal reference for the fruit development and new growing samples are monitored in order to detect any anomaly in their growth conditions.