Search

We would like to congratulate the authors on presenting us a nice methodology on how to perform statistical process monitoring (SPM) on point cloud, mesh and voxel data. The key idea is to consider...

In in most cases, the distribution of communications is unknown and one may summarize social network communications with categorical attributes in a contingency table. Due to the categorical nature...

Invited discussion paper for "A Review of Some Sampling and Aggregation Strategies for Basic Statistical Process Monitoring" by Zwetsloot and Woodall.

I would like to congratulate Drs. Zwetsloot and Woodall (2019) for writing this comprehensive and timely review. The issue of sampling and aggregation is becoming increasingly important in conjunct...

The article by Zwetsloot and Woodall (2019) opportunely addresses an updated and relevant topic that has been escaping the attention of the data-centric/evidenced based communities. As data collect...

Abstract Sensing networks provide nowadays massive amounts of data that in many applications provide information about curves, surfaces and vary over a continuum, usually time, and thus, can be suitably modelled as functional data. Their proper modelling by means of functional data analysis approaches naturally addresses new challenges also arising in the statistical process monitoring (SPM). Motivated by an industrial application, the objective of the present paper is to provide the reader with a very transparent set of steps for the SPM of functional data in real-world case studies: i) identifying a finite dimensional model for the functional data, based on functional principal component analysis; ii) estimating the unknown parameters; iii) designing control charts on the estimated parameters, in a nonparametric framework. The proposed SPM procedure is applied to a real-case study from the maritime field in monitoring CO2 emissions from real navigation data of a roll-on/roll-off passenger cruise ship, i.e., a ship designed to carry both passengers and wheeled vehicles that are driven on and off the ship on their own wheels. We show different scenarios highlighting clear and interpretable indications that can be extracted from the data set and support the detection of anomalous voyages.

I discuss the article “Real-time monitoring of events applied to syndromic surveillance” by Sparks and collaborators. This discussion focuses on how statistical network modeling and inferen...

Nowadays, due to the increasing role of social networks in our daily life, monitoring and forecasting social trends have attracted the attention of many researchers. To the best of the authors' knowledge, the literature includes few studies of monitoring social networks. Existing researches have focused on analyzing only the existence of communications between people and have neglected to monitor the number of such communications. In this paper, first counts of communications between people are modeled using Poisson regression profiles. Then, 3 Phase I monitoring methods, extended T2, F, and a standardized likelihood ratio test method is suggested to detect step changes, drift, and outliers in the parameters of Poisson regression profiles. The proposed methods are evaluated via simulation studies in terms of signal probability criterion. The results show that in most out-of-control situations the standardized likelihood ratio test method outperforms the T2 and F methods. Then, a numerical example and a case study based on Enron email data are presented to illustrate the application of the extended methods.

Metal additive manufacturing (AM) technologies enable the production of complex shapes, lightweight structures and novel functional features. Such increased complexity of the products imposes various challenges in terms of statistical process monitoring and quality assessment. However, one great potential of AM processes, compared to conventional ones, consists of the possibility of gathering a large amount of data layer by layer. This study investigates a data fusion methodology to combine in-situ data from multiple sensors embedded in Electron Beam Melting (EBM) systems to automatically detect faults and process errors. The aim consists of making sense of information already available from the system to enhance its embedded intelligence via novel data mining techniques. A real case study in EBM is presented and discussed.

Social network analysis is an important analytic tool to forecast social trends by modeling and monitoring the interactions between network members. This paper proposes an extension of a st...

This paper presents an overview of wavelet-based techniques for statistical process monitoring. The use of wavelet has already had an effective contribution to many applications. The increase of data availability has led to the use of wavelet analysis as a tool to reduce, denoise, and process the data before using statistical models for monitoring. The most recent review paper on wavelet-based methods for process monitoring had the goal to review the findings up to 2004. In this paper, we provide a recent reference for researchers and engineers with a different focus. We focus on: (i) wavelet statistical properties, (ii) control charts based on wavelet coefficients, and (iii) wavelet-based process monitoring methods within a machine learning framework. It is clear from the literature that wavelets are widely used with multivariate methods compared to univariate methods. We also found some potential research areas regarding the use of wavelet in image process monitoring and designing control charts based on wavelet statistics, and listed them in the paper.

Support vector data description (SVDD), has attracted many researchers’ attention in statistical process monitoring. For batch process fault detection, based on the process data analysis of the three-way structural, a novel SVDD method integrating both generalized additive kernels and local models is proposed in this paper, which is Multi-local support vector data description with Generalized Additive Kernels (MLGAK-SVDD). It can obtain both the convenient on-line batch process fault detection model and the end-of-batch fault detection model at the same time. Finally, a case study based on a fed-batch penicillin fermentation process is conducted to verify the validity of the proposed MLGAK-SVDD method.

In statistical process monitoring, the presence of measurement errors is known to impact the performance of control charts. This paper makes an attempt to investigate the performance of the Hotelli...

In this paper, an integrated fault diagnosis method is proposed to deal with fault location and propagation path identification. A causality graph is first constructed for the system according to the a priori knowledge. Afterward, a correlation index (CI) based on the partial correlation coefficient is proposed to analyze the correlation of variables in causality graph quantitatively. To achieve accurate fault detection results, the proposed CI is monitored by probability principal component analysis. Moreover, the concept of weighted average value is introduced to identify fault propagation path based on reconstruction-based contribution and causality graph after detecting a fault. Finally, the new proposed scheme would be practiced with real industrial HSMP data, where the individual steps as well as the complete framework were extensively tested.

Autoencoders (AEs) are an effective means for nonlinear feature extraction and dimension reduction. Variant autoencoders are an improvement over traditional AEs in terms of robustness. This paper proposes a novel nonlinear and robust process-monitoring approach based on variant autoencoders (variant AEs), which include denoising autoencoders (DAE) and contractive autoencoders (CAE). The CAE and DAE are powerful for extracting robust and nonlinear feature representations or manifold structures underlying data from industrial processes. Next, an online monitoring model is built through constructing new test statistic H2 based on the robust feature representations. The control limits are determined by kernel density estimation. The proposed method was applied to the Tennessee Eastman process (TE process) to evaluate its monitoring performance, and it demonstrated outstanding process-monitoring performance through the experimental results, especially for the barely detectable faults, such as 3, 5, 9, 10, 11, 15, 19, 20 and 21. Variant AEs monitoring provides a simple and very effective process-monitoring method for industrial processes.

Methods based on principal component analysis (PCA) are widely used for statistical process monitoring of high-dimensional processes. Allowing the monitoring model to update as new observations are acquired extends this class of approaches to non-stationary processes. The updating procedure is governed by a weighting parameter that defines the rate at which older observations are discarded, and therefore, it greatly affects model quality and monitoring performance. Additionally, monitoring non-stationary processes can require adjustments to the parameters defining the control limits of adaptive PCA in order to achieve the intended false detection rate. These two aspects require careful consideration prior the implementation of adaptive PCA. Towards this end, approaches are given in this paper for both parameter selection challenges. Results are presented for a simulation and two real-life industrial process examples. Copyright © 2016 John Wiley & Sons, Ltd.

Multivariate data analysis is a powerful tool for process monitoring and data analysis. The theoretical methodology of real-time multivariate data analysis has been studied in the last decade. However, the effect of dataset size on modeling structure and fault detection ability has not been reported yet. In this paper, requirements for a minimum dataset for multivariate data analysis modeling are studied, and a practical approach is provided to evaluate the modeling structure. A method based on statistical index g2 and cross-validation is proposed to determine a minimum dataset size of a valid model for statistical process monitoring. The proposed method was built on the linear PLS model and elaborated by case studies using both batch and continuous processes. This paper provides theoretical development of multivariate data analysis and demonstrates its application in chemical processes.

Modern industrial units collect large amounts of process data based on which advanced process monitoring algorithms continuously assess the status of operations. As an integral part of the development of such algorithms, a reference dataset representative of normal operating conditions is required to evaluate the stability of the process and, after confirming that it is stable, to calibrate a monitoring procedure, i.e., estimate the reference model and set the control limits for the monitoring statistics. The basic assumption is that all relevant &ldquo, common causes&rdquo, of variation appear well represented in this reference dataset (using the terminology adopted by the founding father of process monitoring, Walter A. Shewhart). Otherwise, false alarms will inevitably occur during the implementation of the monitoring scheme. However, we argue and demonstrate in this article, that this assumption is often not met in modern industrial systems. Therefore, we introduce a new approach based on the rigorous mechanistic modeling of the dominant modes of common cause variation and the use of stochastic computational simulations to enrich the historical dataset with augmented data representing a comprehensive coverage of the actual operational space. We show how to compute the monitoring statistics and set their control limits, as well as to conduct fault diagnosis when an abnormal event is declared. The proposed method, called AGV (Artificial Generation of common cause Variability) is applied to a Surface Mount Technology (SMT) production line of Bosch Car Multimedia, where more than 17 thousand product variables are simultaneously monitored.

The complexity of modern problems turns increasingly larger in industrial environments, so the classical process monitoring techniques have to adapt to deal with those problems. This is one of the reasons why new Machine and Statistical Learning methodologies have become very popular in the statistical community. Specifically, this article is focused on machine learning kernel methods techniques in the process monitoring field. After explaining the idea of kernel methods we thoroughly examine the process monitoring articles that make use of kernel models and the way in which these models are combined with other Machine Learning approaches. Finally, we summarize the whole picture of the literature and mention some remarkable points.

This work is focused on the data based modelling and monitoring of a family of modular systems that have multiple replicated structures with the same nominal variables and show temporal behaviour with certain periodicity. These characteristics are present in many systems in numerous fields such as the construction or energy sector or in industry. The challenge for these systems is to be able to exploit the redundancy in both time and the physical structure. In this paper the authors present a method for representing such granular systems using N-dimensional data arrays which are then transformed into the suitable 2-dimensional matrices required to perform statistical processing. Here, the focus is on pre-processing data using a non-unique folding-unfolding algorithm in a way that allows for different statistical models to be built in accordance with the monitoring requirements selected. Principal Component Analysis (PCA) is assumed as the underlying principle to carry out the monitoring. Thus, the method extends the Unfold Principal Component Analysis (Unfold-PCA or Multiway PCA), applied to 3D arrays, to deal with N-dimensional matrices. However, this method is general enough to be applied in other multivariate monitoring strategies. Two of examples in the area of energy efficiency illustrate the application of the method for modelling. Both examples illustrate how when a unique data-set folded and unfolded in different ways, it offers different modelling capabilities. Moreover, one of the examples is extended to exploit real data. In this case, real data collected over a two-year period from a multi-housing social-building located in down town Barcelona (Catalonia) has been used This work has been carried out by the research group eXIT (http://exit.udg.edu), funded through the following projects: MESC project(Ref. DPI2013-47450-C21-R) and its continuation CROWDSAVING (Ref.TIN2016-79726-C2-2-R), both funded by the Spanish Ministerio de Industria y Competitividad within the Research, Development and Innovation Program oriented towards the Societal Challenges, and also the project Hit2Gap of the Horizon 2020 research and innovation program under grant agreement N680708. The author Llorenç Burgas would also like to thank Girona University for their support through the competitive grant for doctoral formation IFUdG2016

One-class classification (OCC) has attracted a great deal of attentions from various disciplines. Few attempts are made to extend the scope of such application for process monitoring. In the present work, the Principal Component Analysis (PCA) and Variational Bayesian Principal Component Analysis (VBPCA) approach provides a powerful tool to project original data into lower data set as well as spreading different types of faults with different directions. This, along with multiple types of one-class classifiers (density-based, boundary-based, reconstruction-based and combination-based) that are able to isolate abnormal data from normal one, supported the design of process monitoring. These methodologies have been validated by process data collected from a Wastewater Treatment Plant (WWTP). The results showed that the proposed methodology is capable of detecting sensor faults and process faults with good accuracy under different scenarios.

In statistical process monitoring, contribution plots are commonly used by operators and experts to identify the root cause of abnormal events. Because contribution plots suffer from fault smearing - an effect that possibly masks the cause of an upset - this paper investigates whether automated fault identification can be improved by using process data instead of contributions. Hereto, both approaches (i.e., using either the sensor measurements or their contributions as inputs for a classification model) are tested on the benchmark penicillin fermentation process Pensim, implemented in RAYMOND. To optimize the performance of each approach, different manipulations of both the process data and the variable contributions are introduced based on the nature of the occurring faults. It is observed that these manipulations have a large influence on the classification performance. Furthermore, this paper demonstrates that fault smearing negatively affects the classification based on the variable contributions. It is concluded that automated fault identification is improved by using the process data rather than the variable contributions as model inputs for the case study investigated.

Considering that the global and local structures of process data would probably be changed in some abnormal states, a multi-manifold projection (MMP) algorithm for process monitoring and fault diagnosis is proposed under the graph embedded learning framework. To exploit the underlying geometrical structure that contains both global and local information of sampled data, the global graph and local graph are designed to characterize the global and local structures, respectively. A unified optimization framework, i.e. global graph maximum and local graph minimum, is then constructed to extract meaningful low-dimensional representations for high-dimensional process data. In the proposed MMP, the neighborhood embedding is used in both global and local graphs and the extracted features are faithful representations of the original data. The feasibility and validity of the MMP-based process monitoring scheme are investigated through two case studies: a simple simulation process and the Tennessee Eastman process. The experimental results demonstrate that the whole performance of MMP is better than those of some traditional preserving global or local or global and local feature methods.

This paper provides an overview and perspective of recent research and applications of statistical process monitoring.

Multivariate statistical process monitoring (MSPM) can conduct dimensionality reduction on process variables and can obtain low-dimensional representations that capture most of the information in t...

An overview on the effect of measurement errors on different areas of SPM.Providing a comprehensive classification of articles in this area.Presenting an analytical overview on the researches in this area.Introducing research gaps in this area to motivate future studies. In most industrial applications, the measures performed on inspected units are often strongly contaminated by either the inspector or the measuring device leading to measurement errors. It is recognized that the measurement errors affect the performance of control charts in various statistical process monitoring applications. In this paper, we present a conceptual classification scheme based on content analysis method to analyze and categorize the researches which have explored the effect of measurement errors on different aspects of statistical process monitoring (SPM). Moreover, based on 60 relevant papers in this field, the research gaps are mentioned and some directions to motivate the future studies are provided.

© 2016 American Institute of Chemical Engineers. High-dimensional and time-dependent data pose significant challenges to Statistical Process Monitoring. Most of the high-dimensional methodologies to cope with these challenges rely on some form of Principal Component Analysis (PCA) model, usually classified as nonadaptive and adaptive. Nonadaptive methods include the static PCA approach and Dynamic Principal Component Analysis (DPCA) for data with autocorrelation. Methods, such as DPCA with Decorrelated Residuals, extend DPCA to further reduce the effects of autocorrelation and cross-correlation on the monitoring statistics. Recursive Principal Component Analysis and Moving Window Principal Component Analysis, developed for nonstationary data, are adaptive. These fundamental methods will be systematically compared on high-dimensional, time-dependent processes (including the Tennessee Eastman benchmark process) to provide practitioners with guidelines for appropriate monitoring strategies and a sense of how they can be expected to perform. The selection of parameter values for the different methods is also discussed. Finally, the relevant challenges of modeling time-dependent data are discussed, and areas of possible further research are highlighted. ispartof: Aiche Journal vol:62 issue:5 pages:1478-1493 ispartof: location:Ankara status: published

Monitoring high-dimensional multistage processes becomes crucial to ensure the quality of the final product in modern industry environments. Few statistical process monitoring (SPC) approaches for monitoring and controlling quality in high-dimensional multistage processes are studied. We propose a deviance residual-based multivariate exponentially weighted moving average (MEWMA) control chart with a variable selection procedure. We demonstrate that it outperforms the existing multivariate SPC charts in terms of out-of-control average run length (ARL) for the detection of process mean shift.

Wang and Chen (Qual. Eng. 1998; 11:21–27) have defined process capability indices (PCIs) for multivariate normal processes data using principal component analysis (PCA). Veevers (Statistical Process Monitoring and Optimization. Marcel Dekker: New York, NY, 1999; 241–256) has suggested a multivariate capability index based on the first principal component (PC). In this paper we demonstrate the problem in the definition of PCIs given by Wang and Chen (Qual. Eng. 1998; 11:21–27) and the non-suitability of PCI given by Veevers (Statistical Process Monitoring and Optimization. Marcel Dekker: New York, NY, 1999; 241–256) through some examples. We also suggest an alternative method for assessing multivariate process capability based on the empirical probability distribution of PCs. This method has been performed on industrial and simulated data. Copyright © 2008 John Wiley & Sons, Ltd.

Several diagnosis methods have been proposed for statistical process monitoring. They have been developed from different backgrounds and considerations. In this paper, five existing diagnosis methods are analyzed and compared. It is shown that these methods can be unified into three more general methods, making the original diagnosis methods special cases of the general methods. An analysis of the diagnosability of the general methods shows that some diagnosis methods guarantee correct diagnosis for simple sensor faults with large magnitudes, while others do not. For the case of sensor faults with modest fault magnitudes, a Monte Carlo simulation is used to compare the performance of the diagnosis methods.

A new on-line predictive monitoring and prediction model for periodic biological processes is proposed using the multiway non-Gaussian modeling. The basic idea of this approach is to use multiway non-Gaussian modeling to extract some dominant key components from daily normal operation data in a periodic process, and subsequently combining these components with predictive statistical process monitoring techniques. The proposed predictive monitoring method has been applied to fault detection and diagnosis in the biological wastewater-treatment process, which is based on strong diurnal characteristics. The results show the power and advantages of the proposed predictive monitoring of a continuous process using the multiway predictive monitoring concept, which is thus able to give very useful conceptual results for a daily monitoring process and also enables a more rapid detection of the process fault than other traditional monitoring methods.

Recently, a novel multivariate statistical process monitoring method, known as dissimilarity algorithm(DISSIM), has been developed based on the idea that a change of operating condition can be detected by monitoring a distribution of process data set, where a dissimilarity index is introduced to quantitatively evaluate the difference between distributions of process data. However, as a fixed-model monitoring technique, it inevitably gives false alarms when applied to real processes involving slow changes. In this paper, an adaptive DISSIM(ADISSIM) algorithm is proposed for on-line updating to consecutively adapt to process slow-varying behaviors. The key to the proposed method is that whenever the old model is judged to be inefficient to capture the current normal operation status, a new monitoring model is developed by integrating the old model and the new updating data. The effectiveness of ADISSIM algorithm is successfully illustrated when applied to simulated data collected from a simple 2×2 numerical process. The results clearly show that the proposed adaptive method is effective to accommodate the normal gradual changes and distinguish them from real process faults, thus providing a new feasible statistical monitoring method for the prevalent slow-varying processes.

A new framework for extending Statistical Process Monitoring (SPM) to simultaneous False Alarm Rejection and Fault Identification (FARFI) is presented in this paper. This is motivated by the possibly large negative impact on product quality, process safety, and profitability resulting from incorrect control actions induced by false alarms—especially for batch processes. The presented FARFI approach adapts the classification model already used for fault identification to simultaneously perform false alarm rejection by adding normal operation as an extra data class. As no additional models are introduced, the complexity of the overall SPM system is not increased.

The threat of cyber attacks motivates the need to monitor Internet traffic data for potentially abnormal behavior. Due to the enormous volumes of such data, statistical process monitoring tools, such as those traditionally used on data in the product manufacturing arena, are inadequate. ''Exotic'' data may indicate a potential attack; detecting such data requires a characterization of ''typical'' data. We devise some new graphical displays, including a ''skyline plot,'' that permit ready visual identification of unusual Internet traffic patterns in ''streaming'' data, and use appropriate statistical measures to help identify potential cyberattacks. These methods are illustrated on a moderate-sized data set (135,605 records) collected at George Mason University.

The emphasis of most PCA process monitoring approaches is mainly on procedures to perform fault detection and diagnosis given a set of sensors. Little attention is paid to the actual sensor locations to efficiently perform these tasks. In this paper, graph-based techniques are used to optimize sensor locations to ensure the observability of faults, as well as the fault resolution to a maximum possible extent. Meanwhile, an improved PCA that uses two new statistics of PVR and CVR to replace the Q index in conventional PCA is introduced. The improved PCA can efficiently detect weak process changes, and give an insight to the root cause about the process malfunction. Simulation results of a CSTR process show that the improved PCA with optimized sensor locations is superior to conventional methods in fault resolution and sensibility.

Multivariate statistical process control (MSPC) has been widely used for monitoring chemical processes with highly correlated variables. In this work, a novel statistical process monitoring method is proposed based on the idea that a change of operating condition can be detected by monitoring a distribution of process data, which reflects the corresponding operating conditions. To quantitatively evaluate the difference between two data sets, a dissimilarity index is introduced. The monitoring performance of the proposed method, referred to as DISSIM, and that of the conventional MSPC method are compared with their applications to simulated data collected from a simple 2 × 2 process and the Tennessee Eastman process. The results clearly show that the monitoring performance of DISSIM, especially dynamic DISSIM, is considerably better than that of the conventional MSPC method when a time-window size is appropriately selected.

In this paper, we describe the theory underlying an empirical Bayesian approach to monitoring two or more process characteristics simultaneously. If the data is continuous and multivariate in nature, often the multivariate normal distribution can be used to model the process. Then, using Bayesian theory, we develop techniques to implement empirical Bayes process monitoring of the multivariable process. Lastly, an example is given to illustrate the use of our techniques. Copyright © 2001 John Wiley & Sons, Ltd.

Multivariate statistical process control (MSPC) has been successfully applied to chemical procesess. In order to improve the performance of fault detection, two kinds of advanced methods, known as 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 (PCs) and the degree of dissimilarity between data sets, respectively. Another important extension of MSPC was made by using multiscale PCA (MS-PCA). In the present work, the characteristics of several monitoring methods are investigated. The monitoring performances are compared with using simulated data obtained from the Tennessee Eastman process. The results show that the advanced methods can outperform the conventional method. Furthermore, the advantage of MPCA and DISSIM over conventional MSPC (cMSPC) and that of the multiscale method are combined, and the new methods known as MS-MPCA and MS-DISSIM are proposed.

For monitoring chemical processes, multivariate statistical process control (MSPC) has been widely used. In the present work, a new process monitoring method is proposed. The proposed method utilizes a change in distribution of process data, since the distribution reflects the corresponding operating condition. In order to quantitatively evaluate the difference between two data sets, the dissimilarity index is defined. The proposed method and the conventional SPC methods are applied to monitoring problems of the Tennessee Eastman process. The results have clearly shown that the monitoring performance of the proposed method is considerably better than that of the conventional methods.

In statistical process monitoring, violating the assumption of independent data leads to increased control chart false alarms and trends on both sides of the centerline. Autocorrelation requires modification to traditional control chart techniques. Th..

Most industrial processes are characterized by a system of several variables, all of which are subject to drifts, disturbances, and assignable causes of variation. In the chemical and process industries, there are often inertial forces arising from raw material streams, reactors and tanks that introduce serial correlation over time into these variables. This autocorrelation can have a profound impact on the effectiveness of the statistical monitoring methods used for such processes. This paper reviews some of the available methodology for multivariate process monitoring and shows the effectiveness of principal components in this context. An application of the principal components approach with correlated observation vectors is presented. The effectiveness of this procedure to indicate process upsets is discussed.