Search

In most data analytic applications in manufacturing, understanding the data‐driven models plays a crucial role in complementing the engineering knowledge about the production process. Identifying relevant input variables, rather than only predicting the response through some "black‐box" model, is of great interest in many applications. There is, therefore, a growing focus on describing the contributions of the input variables to the model in the form of "variable importance", which is readily available in certain machine learning methods such as random forest (RF). Once a ranking based on the importance measure of the variables is established, the question of how many variables are truly relevant in predicting the output variable rises. In this study, we focus on the Boruta algorithm, which is a wrapper around the RF model. It is a variable selection tool that assesses the variable importance measure for the RF model. It has been previously shown in the literature that the correlation among the input variables, which is often a common occurrence in high dimensional data, distorts and overestimates the importance of variables. The Boruta algorithm is also affected by this resulting in a larger set of input variables deemed important. To overcome this issue, in this study, we propose an extension of the Boruta algorithm for the correlated data by exploiting the conditional importance measure. This extension greatly improves the Boruta algorithm in the case of high correlation among variables and provides a more precise ranking of the variables that significantly contribute to the response. We believe this approach can be used in many industrial applications by providing more transparency and understanding of the process. [ABSTRACT FROM AUTHOR]

In many industrial applications, obtaining labeled observations is not straightforward as it often requires the intervention of human experts or the use of expensive testing equipment. In these circumstances, active learning can be highly beneficial in suggesting the most informative data points to be used when fitting a model. Reducing the number of observations needed for model development alleviates both the computational burden required for training and the operational expenses related to labeling. Online active learning, in particular, is useful in high‐volume production processes where the decision about the acquisition of the label for a data point needs to be taken within an extremely short time frame. However, despite the recent efforts to develop online active learning strategies, the behavior of these methods in the presence of outliers has not been thoroughly examined. In this work, we investigate the performance of online active linear regression in contaminated data streams. Our study shows that the currently available query strategies are prone to sample outliers, whose inclusion in the training set eventually degrades the predictive performance of the models. To address this issue, we propose a solution that bounds the search area of a conditional D‐optimal algorithm and uses a robust estimator. Our approach strikes a balance between exploring unseen regions of the input space and protecting against outliers. Through numerical simulations, we show that the proposed method is effective in improving the performance of online active learning in the presence of outliers, thus expanding the potential applications of this powerful tool. [ABSTRACT FROM AUTHOR]

As a direct consequence of production systems' digitalization, high‐frequency and high‐dimensional data has become more easily available. In terms of data analysis, latent structures‐based methods are often employed when analyzing multivariate and complex data. However, these methods are designed for supervised learning problems when sufficient labeled data are available. Particularly for fast production rates, quality characteristics data tend to be scarcer than available process data generated through multiple sensors and automated data collection schemes. One way to overcome the problem of scarce outputs is to employ semi‐supervised learning methods, which use both labeled and unlabeled data. It has been shown that it is advantageous to use a semi‐supervised approach in case of labeled data and unlabeled data coming from the same distribution. In real applications, there is a chance that unlabeled data contain outliers or even a drift in the process, which will affect the performance of the semi‐supervised methods. The research question addressed in this work is how to detect outliers in the unlabeled data set using the scarce labeled data set. An iterative strategy is proposed using a combined Hotelling's T2 and Q statistics and applied using a semi‐supervised principal component regression (SS‐PCR) approach on both simulated and real data sets. [ABSTRACT FROM AUTHOR]

Abstract A novel data driven approach to batch process monitoring is presented, which combines the k-Nearest Neighbour rule with the dynamic time warping (DTW) distance. This online method (DTW-NN) calculates the DTW distance between an ongoing batch, and each batch in a reference database of batches produced under normal operating conditions (NOC). The sum of the k smallest DTW distances is monitored. If a fault occurs in the ongoing batch, then this distance increases and an alarm is generated. The monitoring statistic is easy to interpret, being a direct measure of similarity of the ongoing batch to its nearest NOC predecessors and the method makes no distributional assumptions regarding normal operating conditions. DTW-NN is applied to four extensive datasets from simulated batch production of penicillin, and tested on a wide variety of fault types, magnitudes and onset times. Performance of DTW-NN is contrasted with a benchmark multiway PCA approach, and DTW-NN is shown to perform particularly well when there is clustering of batches under NOC. [ABSTRACT FROM AUTHOR]

Abstract: The expanding availability of complex data structures requires development of new analysis methods for process understanding and monitoring. In manufacturing, this is primarily due to high‐frequency and high‐dimensional data available through automated data collection schemes and sensors. However, particularly for fast production rate situations, data on the quality characteristics of the process output tend to be scarcer than the available process data. There has been a considerable effort in incorporating latent structure–based methods in the context of complex data. The research question addressed in this paper is to make use of latent structure–based methods in the pursuit of better predictions using all available data including the process data for which there are no corresponding output measurements, ie, unlabeled data. Inspiration for the research question comes from an industrial setting where there is a need for prediction with extremely low tolerances. A semi‐supervised principal component regression method is compared against benchmark latent structure–based methods, principal components regression, and partial least squares, on simulated and experimental data. In the analysis, we show the circumstances in which it becomes more advantageous to use the semi‐supervised principal component regression over these competing methods. [ABSTRACT FROM AUTHOR]

With the emergence of smart factories, large volumes of process data are collected and stored at high sampling rates for improved energy efficiency, process monitoring and sustainability. The data collected in the course of enterprise-wide operations consists of information from broadly deployed sensors and other control equipment. Interpreting such large volumes of data with limited workforce is becoming an increasingly common challenge. Principal component analysis (PCA) is a widely accepted procedure for summarizing data while minimizing information loss. It does so by finding new variables, the principal components (PCs) that are linear combinations of the original variables in the dataset. However, interpreting PCs obtained from many variables from a large dataset is often challenging, especially in the context of fault detection and diagnosis studies. Sparse principal component analysis (SPCA) is a relatively recent technique proposed for producing PCs with sparse loadings via variance-sparsity trade-off. Using SPCA, some of the loadings on PCs can be restricted to zero. In this paper, we introduce a method to select the number of non-zero loadings in each PC while using SPCA. The proposed approach considerably improves the interpretability of PCs while minimizing the loss of total variance explained. Furthermore, we compare the performance of PCA- and SPCA-based techniques for fault detection and fault diagnosis. The key features of the methodology are assessed through a synthetic example and a comparative study of the benchmark Tennessee Eastman process. [ABSTRACT FROM AUTHOR]

This study compares four models for industrial condition monitoring including a principal components analysis (PCA) approach and three deep learning models, one of which is a new, lightweight version of another. We also propose a simple attention mechanism for enchancing deep learning models with better predictions and feature importance. Two datasets are used, one simulated from the Tennessee Eastman Process, the other from two feedwater pumps at a Danish combined heat and power plant. Our final results show evidence in favour of the PCA-based approach as it has detection ability comparable to the deep learning approaches as well as faster training time, fewer hyperparameters, as well as robustness to changing operating conditions. We conclude the paper by putting into perspective the importance of building up complexity incrementally with a recommendation to start modelling with simpler and well-tested models before the adoption of more advanced, less transparent models. • A statistical model is compared with deep learning techniques for fault detection. • The principal components model's fault detection is as good as deep learning. • Deep learning method, ARGUELite, shows promise for single regime datasets. • A unique dataset from a power plant feedwater pump was used for model evaluation. [ABSTRACT FROM AUTHOR]

When principal component analysis (PCA) is used for statistical process monitoring it relies on the assumption that data are time independent. However, industrial data will often exhibit serial correlation. Dynamic PCA (DPCA) has been suggested as a remedy for high-dimensional and time-dependent data. In DPCA the input matrix is augmented by adding time-lagged values of the variables. In building a DPCA model the analyst needs to decide on (1) the number of lags to add, and (2) given a specific lag structure, how many principal components to retain. In this article we propose a new analyst driven method to determine the maximum number of lags in DPCA with a foundation in multivariate time series analysis. The method is based on the behavior of the eigenvalues of the lagged autocorrelation and partial autocorrelation matrices. Given a specific lag structure we also propose a method for determining the number of principal components to retain. The number of retained principal components is determined by visual inspection of the serial correlation in the squared prediction error statistic, Q (SPE), together with the cumulative explained variance of the model. The methods are illustrated using simulated vector autoregressive and moving average data, and tested on Tennessee Eastman process data. [ABSTRACT FROM AUTHOR]

Principal component analysis (PCA) is a widely accepted procedure for summarizing data through dimensional reduction. In PCA, the selection of the appropriate number of components and the interpretation of those components have been the key challenging features. Sparse principal component analysis (SPCA) is a relatively recent technique proposed for producing principal components with sparse loadings via the variance-sparsity trade-off. Although several techniques for deriving sparse loadings have been offered, no detailed guidelines for choosing the penalty parameters to obtain a desired level of sparsity are provided. In this paper, we propose the use of a genetic algorithm (GA) to select the number of non-zero loadings (NNZL) in each principal component while using SPCA. The proposed approach considerably improves the interpretability of principal components and addresses the difficulty in the selection of NNZL in SPCA. Furthermore, we compare the performance of PCA and SPCA in uncovering the underlying latent structure of the data. The key features of the methodology are assessed through a synthetic example, pitprops data and a comparative study of the benchmark Tennessee Eastman process. [ABSTRACT FROM AUTHOR]

Most of today's complex systems and processes involve several stages through which input or the raw material has to go before the final product is obtained. Also in many cases factors at different stages interact. Therefore, a holistic approach for experimentation that considers all stages at the same time will be more efficient. However, there have been only a few attempts in the literature to provide an adequate and easy-to-use approach for this problem. In this paper, we present a novel methodology for constructing two-level split-plot and multistage experiments. The methodology is based on the Kronecker product representation of orthogonal designs and can be used for any number of stages, for various numbers of subplots and for different number of subplots for each stage. The procedure is demonstrated on both regular and nonregular designs and provides the maximum number of factors that can be accommodated in each stage. Furthermore, split-plot designs for multistage experiments with good projective properties are also provided. [ABSTRACT FROM AUTHOR]

A basic assumption when using principal component analysis (PCA) for inferential purposes, such as in statistical process control (SPC), is that the data are independent in time. In many industrial processes, frequent sampling and process dynamics make this assumption unrealistic rendering sampled data autocorrelated (serially dependent). PCA can be used to reduce data dimensionality and to simplify multivariate SPC. Although there have been some attempts in the literature to deal with autocorrelated data in PCA, we argue that the impact of autocorrelation on PCA and PCA-based SPC is neither well understood nor properly documented. This article illustrates through simulations the impact of autocorrelation on the descriptive ability of PCA and on the monitoring performance using PCA-based SPC when autocorrelation is ignored. In the simulations, cross-correlated and autocorrelated data are generated using a stationary first-order vector autoregressive model. The results show that the descriptive ability of PCA may be seriously affected by autocorrelation causing a need to incorporate additional principal components to maintain the model's explanatory ability. When all variables have equal coefficients in a diagonal autoregressive coefficient matrix, the descriptive ability is intact, while a significant impact occurs when the variables have different degrees of autocorrelation. We also illustrate that autocorrelation may impact PCA-based SPC and cause lower false alarm rates and delayed shift detection, especially for negative autocorrelation. However, for larger shifts, the impact of autocorrelation seems rather small. © 2015 The Authors. Quality and Reliability Engineering International published by John Wiley & Sons Ltd. [ABSTRACT FROM AUTHOR]

The origins of quality engineering are in manufacturing, where quality engineers apply basic statistical methodologies to improve the quality and productivity of products and processes. In the past decade, people have discovered that these methodologies are effective for improving almost any type of system or process, such as financial, health care, and supply chains. This paper begins with a review of key advances and trends within quality engineering over the past decade. The second part uses the first part as a foundation to outline new application areas for the field. It also discusses how quality engineering needs to evolve in order to make significant contributions to these new areas. © 2015 The Authors Quality and Reliability Engineering International Published by John Wiley & Sons Ltd. [ABSTRACT FROM AUTHOR]

One of the basic assumptions for traditional univariate and multivariate control charts is that the data are independent in time. For the latter, in many cases, the data are serially dependent (autocorrelated) and cross-correlated because of, for example, frequent sampling and process dynamics. It is well known that the autocorrelation affects the false alarm rate and the shift-detection ability of the traditional univariate control charts. However, how the false alarm rate and the shift-detection ability of the Hotelling T² control chart are affected by various autocorrelation and cross-correlation structures for different magnitudes of shifts in the process mean is not fully explored in the literature. In this article, the performance of the Hotelling T² control chart for different shift sizes and various autocorrelation and cross-correlation structures are compared based on the average run length using simulated data. Three different approaches in constructing the Hotelling T² chart are studied for two different estimates of the covariance matrix: (i) ignoring the autocorrelation and using the raw data with theoretical upper control limits; (ii) ignoring the autocorrelation and using the raw data with adjusted control limits calculated through Monte Carlo simulations; and (iii) constructing the control chart for the residuals from a multivariate time series model fitted to the raw data. To limit the complexity, we use a first-order vector autoregressive process and focus mainly on bivariate data. [ABSTRACT FROM AUTHOR]

The proliferation of automated data collection schemes and the advances in sensorics are increasing the amount of data we are able to monitor in real-time. However, given the high annotation costs and the time required by quality inspections, data is often available in an unlabeled form. This is fostering the use of active learning for the development of soft sensors and predictive models. In production, instead of performing random inspections to obtain product information, labels are collected by evaluating the information content of the unlabeled data. Several query strategy frameworks for regression have been proposed in the literature but most of the focus has been dedicated to the static pool-based scenario. In this work, we propose a new strategy for the stream-based scenario, where instances are sequentially offered to the learner, which must instantaneously decide whether to perform the quality check to obtain the label or discard the instance. The approach is inspired by the optimal experimental design theory and the iterative aspect of the decision-making process is tackled by setting a threshold on the informativeness of the unlabeled data points. The proposed approach is evaluated using numerical simulations and the Tennessee Eastman Process simulator. The results confirm that selecting the examples suggested by the proposed algorithm allows for a faster reduction in the prediction error. [ABSTRACT FROM AUTHOR]

• An unsupervised fault detection and diagnosis scheme based on orthogonal autoencoders is proposed for the monitoring of industrial and chemical processes. • The use of integrated gradients allows for exploration of the bottleneck of the autoencoder, providing candidate variables for the root cause analysis. • The proposed method performs well compared to traditional methods in offering compelling and interpretable results. • The analysis shows how the difference in the way faults are introduced affects the detection and diagnosis performances of principal component analysis-based control charts. In recent years, there have been studies focusing on the use of different types of autoencoders (AEs) for monitoring complex nonlinear data coming from industrial and chemical processes. However, in many cases the focus was placed on detection. As a result, practitioners are encountering problems in trying to interpret such complex models and obtaining candidate variables for root cause analysis once an alarm is raised. This paper proposes a novel statistical process control (SPC) framework based on orthogonal autoencoders (OAEs). OAEs regularize the loss function to ensure no correlation among the features of the latent variables. This is extremely beneficial in SPC tasks, as it allows for the invertibility of the covariance matrix when computing the Hotelling T 2 statistic, significantly improving detection and diagnosis performance when the process variables are highly correlated. To support the fault diagnosis and identification analysis, we propose an adaptation of the integrated gradients (IG) method. Numerical simulations and the benchmark Tennessee Eastman Process are used to evaluate the performance of the proposed approach by comparing it to traditional approaches as principal component analysis (PCA) and kernel PCA (KPCA). In the analysis, we explore how the information useful for fault detection and diagnosis is stored in the intermediate layers of the encoder network. We also investigate how the correlation structure of the data affects the detection and diagnosis of faulty variables. The results show how the combination of OAEs and IG represents a compelling and ready-to-use solution, offering improved detection and diagnosis performances over the traditional methods. [ABSTRACT FROM AUTHOR]

Abstract: In this article we investigate two-level split-plot designs where the sub-plots consist of only two mirror image trials. Assuming third and higher order interactions negligible, we show that these designs divide the estimated effects into two orthogonal sub-spaces, separating sub-plot main effects and sub-plot by whole-plot interactions from the rest. Further we show how to construct split-plot designs of projectivity P≥3. We also introduce a new class of split-plot designs with mirror image pairs constructed from non-geometric Plackett–Burman designs. The design properties of such designs are very appealing with effects of major interest free from full aliasing assuming that 3rd and higher order interactions are negligible. [Copyright &y& Elsevier]

Robust parameter design is one of the important tools used in Design for Six Sigma. In this article, we present an application of the generalized linear mixed model (GLMM) approach to robust design and analysis of signal–response systems. We propose a split-plot approach to the signal–response system characterized by two variance components—within-profile variance and between-profile variance. We demonstrate that explicit modeling of variance components using GLMMs leads to more precise point estimates of important model coefficients with shorter confidence intervals. Copyright © 2009 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

In this paper we present a modeling framework for analyzing computer models with two types of variations. The paper is based on a case study of an orthopedic surgical unit, which has both controllable and uncontrollable factors. Our results show that this structure of variation can be modeled effectively with linear mixed effects models and generalized additive models. Copyright © 2009 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

Inserting estimates for the missing observations from split-plot designs restores their balanced or orthogonal structure and alleviates the difficulties in the statistical analysis. In this article, we extend a method due to Draper and Stoneman to estimate the missing observations from unreplicated two-level factorial and fractional factorial split-plot (FSP and FFSP) designs. The missing observations, which can either be from the same whole plot, from different whole plots, or comprise entire whole plots, are estimated by equating to zero a number of specific contrast columns equal to the number of the missing observations. These estimates are inserted into the design table and the estimates for the remaining effects (or alias chains of effects as the case with FFSP designs) are plotted on two half-normal plots: one for the whole-plot effects and the other for the subplot effects. If the smaller effects do not point at the origin, then different contrast columns to some or all of the initial ones should be discarded and the plots re-examined for bias. Using examples, we show how the method provides estimates for the missing observations that are very close to their actual values. Copyright © 2007 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

There is an increasing use of on-line data acquisition systems in industry. This usually leads to autocorrelated data and implies that the assumption of independent observations has to be re-examined. Most decision procedures for capability analysis assume independent data. In this article we present a new way of performing capability analysis when data are autocorrelated. This method is based on what can be called the ‘iterative skipping’ strategy. In that, by skipping a pre-determined number of observations, e.g. considering every fifth observation, the data set is divided into subsamples for which the independence assumption may be valid. For each such subsample of the data we estimate a capability index. Then traditional tests, assuming independence, can be performed based on each estimated capability index from the subsamples. By combining the information from each test statistic based on the subsamples in a suitable way, a new and efficient decision procedure is obtained. We discuss different ways of combining the information from these individual tests. A main appeal of our proposed method is that no time-series model is needed. Copyright © 2007 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

Screening experiments are typically used when attempting to identify a few active factors in a larger pool of potentially significant factors. In general, two-level regular factorial designs are used, but Plackett–Burman (PB) designs provide a useful alternative. Although PB designs are run-efficient, they confound the main effects with fractions of strings of two-factor interactions, making the analysis difficult. However, recent discoveries regarding the projective properties of PB designs suggest that if only a few factors are active, the original design can be reduced to a full factorial, with additional trials frequently forming attractive patterns. In this paper, we show that there is a close relationship between the partial confounding in certain PB designs and their projective properties. With the aid of examples, we demonstrate how this relationship may help experimenters better appreciate the use of PB designs. Copyright © 2006 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

Blocking is commonly used in experimental design to eliminate unwanted variation by creating more homogeneous conditions for experimental treatments within each block. While it has been a standard practice in experimental design, blocking fractional factorials still presents many challenges due to differences between treatment and blocking variables. Lately, new design criteria such as the total number of clear effects and fractional resolution have been proposed to design blocked two-level fractional factorial experiments. This article presents a flexible matrix representation for two-level fractional factorials that will allow experimenters and software developers to block such experiments based on any design criterion that is suitable with the experimental conditions. Copyright © 2006 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

The investigation of aliases or biases is important for the interpretation of the results from factorial experiments. For two-level fractional factorials this can be facilitated through their group structure. For more general arrays the alias matrix can be used. This tool is traditionally based on the assumption that the error structure is that associated with ordinary least squares. For situations where that is not the case, we provide in this article a generalization of the alias matrix applicable under the generalized least squares assumptions. We also show that for the special case of split plot error structure, the generalized alias matrix simplifies to the ordinary alias matrix. [ABSTRACT FROM AUTHOR]

When some factors are hard to change and others are relatively easier, split-plot experiments are often an economic alternative to fully randomized designs. Split-plot experiments, with their structure of subplot arrays imbedded within whole-plot arrays, have a tendency to become large, particularly in screening situations when many factors are considered. To alleviate this problem, we explore, for the case of two-level designs, various ways to use orthogonal arrays of the Plackett-Burman type to reduce the number of individual tests. General construction principles are outlined, and the resulting alias structure is derived and discussed. [ABSTRACT FROM AUTHOR]

In this article we present a procedure to analyze split-plot experiments with mirror image pairs as sub-plots when third- and higher-order interactions can be assumed negligible. Although performing a design in a split-plot manner induces correlation among observations, we show that with such designs the essential search for potentially active factors can be done in two steps using ordinary least squares. The suggested procedure is tested out on a real example and on two simulated screening examples; one with a split-plot design based on a geometric design and one with a split-plot design based on a non-geometric Plackett and Burman design. The examples also illustrate the advantage of using non-geometric designs where the effects are partially aliased instead of being fully aliased as in highly fractionated fractional factorials. Copyright © 2005 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

Industrial experiments are frequently performed sequentially using two-level fractional factorial designs. In this context, a common strategy for the design of follow-up experiments is to switch the signs in one column. It is well known that this strategy, when applied to two-level fractional factorial resolution III designs, will clear the main effect, for which the switch was performed, from any confounding with any other two-factor interactions and will also clear all the two-factor interactions between that factor and the other main effects from any confounding with other two-factor interactions. In this article, we extend this result and show that this strategy applies to any orthogonal two-level resolution III design and therefore specifically to any two-level Plackett- Burman design . [ABSTRACT FROM AUTHOR]

• Simulation of 28 process faults at different fault-magnitudes for six operating modes. • Simulation of the process' reaction to changes of the control setpoints. • Healthy simulation data for other configurations than the original six process modes. • Simulation of dynamic transitions between operating modes with different transition times. • Performance benchmark for statistical fault-detection schemes for the presented dataset. The Tennessee Eastman Process (TEP) is a frequently used benchmark in chemical engineering research. An extended simulator, published in 2015, enables a more in-depth investigation of TEP, featuring additional, scalable process disturbances as well as an extended list of variables. Even though the simulator has been used multiple times since its release, the lack of a standardized reference dataset impedes direct comparisons of methods. In this contribution we present an extensive reference dataset, incorporating repeat simulations of healthy and faulty process data, additional measurements and multiple magnitudes for all process disturbances. All six production modes of TEP as well as mode transitions and operating points in a region around the modes are simulated. We further perform fault-detection based on principal component analysis combined with T 2 and Q charts using average run length as a performance metric to provide an initial benchmark for statistical process monitoring schemes for the presented data. [ABSTRACT FROM AUTHOR]

Introduction: The current process for generating evidence in pharmacovigilance has several limitations, which often lead to delays in the evaluation of drug-associated risks. Objectives: In this study, we proposed and tested a near real-time epidemiological surveillance system using sequential, cumulative analyses focusing on the detection and preliminary risk quantification of potential safety signals following initiation of selective serotonin reuptake inhibitors (SSRIs) and serotonin-norepinephrine reuptake inhibitors (SNRIs). Methods: We emulated an active surveillance system in an historical setting by conducting repeated annual cohort studies using nationwide Danish healthcare data (1996–2016). Outcomes were selected from the European Medicines Agency's Designated Medical Event list, summaries of product characteristics, and the literature. We followed patients for a maximum of 6 months from treatment initiation to the event of interest or censoring. We performed Cox regression analyses adjusted for standard sets of covariates. Potential safety signals were visualized using heat maps and cumulative hazard ratio (HR) plots over time. Results: In the total study population, 969,667 new users were included and followed for 461,506 person-years. We detected potential safety signals with incidence rates as low as 0.9 per 10,000 person-years. Having eight different exposure drugs and 51 medical events, we identified 31 unique combinations of potential safety signals with a positive association to the event of interest in the exposed group. We proposed that these signals were designated for further evaluation once they appeared in a prospective setting. In total, 21 (67.7%) of these were not present in the current summaries of product characteristics. Conclusion: The study demonstrated the feasibility of performing epidemiological surveillance using sequential, cumulative analyses. Larger populations are needed to evaluate rare events and infrequently used antidepressants. [ABSTRACT FROM AUTHOR]

This paper presents a case study on condition monitoring of power generators at offshore wind turbines. Two fault detection models are proposed for detecting sudden changes in the sensed value of metallic debris at the generator. The first model uses an exponentially weighted moving average, while the second monitors first-order derivatives using a fixed threshold. This is expected to improve the maintenance activities by avoiding late or early part replacement. The economic impact of the proposed approach is also provided with a realistic depiction of the cost structure associated with the corresponding maintenance plan. While the specifics of the case study are supported by real-life data, considering the prevalence of the use of generators not only in offshore wind turbines but also in other production environments, we believe the case study covered in this paper can be used as a blueprint for similar studies in other applications. • Large scale condition monitoring demonstrates economic benefits. • Case study from real wind turbine population with economic impact assessment. • Sensor for monitoring fault mode results in lead time for maintenance planning. • Sensitivity analysis assesses results despite downtime and power price uncertainty. [ABSTRACT FROM AUTHOR]

Batch processes usually exhibit variation in the time at which individual batches are stopped (referred to as the harvest time). Harvest time is based on the occurrence of some criterion and there may be great uncertainty as to when this criterion will be satisfied. This uncertainty increases the difficulty of scheduling downstream operations and results in fewer completed batches per day. A real case study is presented of a bacteria fermentation process. We consider the problem of predicting the harvest time of a batch in advance to reduce variation and improving batch quality. Lasso regression is used to obtain an interpretable model for predicting the harvest time at an early stage in the batch. A novel method for updating the harvest time predictions as a batch progresses is presented, based on information obtained from online alignment using dynamic time warping. [ABSTRACT FROM AUTHOR]

The article calls for the submission of papers on data mining technologies and their applications in various domains sucha as engineering, health care, bioinformatics, social sciences and finance.

Industrial manufacturing processes often operate under closed-loop control, where automation aims to keep important process variables at their set-points. In process industries such as pulp, paper, chemical and steel plants, it is often hard to find production processes operating in open loop. Instead, closed-loop control systems will actively attempt to minimize the impact of process disturbances. However, we argue that an implicit assumption in most experimental investigations is that the studied system is open loop, allowing the experimental factors to freely affect the important system responses. This scenario is typically not found in process industries. The purpose of this article is therefore to explore issues of experimental design and analysis in processes operating under closed-loop control and to illustrate how Design of Experiments can help in improving and optimizing such processes. The Tennessee Eastman challenge process simulator is used as a test-bed to highlight two experimental scenarios. The first scenario explores the impact of experimental factors that may be considered as disturbances in the closed-loop system. The second scenario exemplifies a screening design using the set-points of controllers as experimental factors. We provide examples of how to analyze the two scenarios. © 2017 The Authors Quality and Reliability Engineering International Published by John Wiley & Sons Ltd [ABSTRACT FROM AUTHOR]

There are two key reasons for aligning batch process data. The first is to obtain same-length batches so that standard methods of analysis may be applied, whilst the second reason is to synchronise events that take place during each batch so that the same event is associated with the same observation number for every batch. Dynamic time warping has been shown to be an effective method for meeting these objectives. This is based on a dynamic programming algorithm that aligns a batch to a reference batch, by stretching and compressing its local time dimension. The resulting ”warping function” may be interpreted as a progress signature of the batch which may be appended to the aligned data for further analysis. For the warping function to be a realistic reflection of the progress of a batch, it is necessary to impose some constraints on the dynamic time warping algorithm, to avoid an alignment which is too aggressive and which contains pathological warping. Previous work has focused on addressing this issue using global constraints. In this work, we investigate the use of local constraints in dynamic time warping and define criteria for evaluating the degree of time distortion and variable synchronisation obtained. A local constraint scheme is extended to include constraints not previously considered, and a novel method for selecting the optimal local constraint with respect to the two criteria is proposed. For illustration, the method is applied to real data from an industrial bacteria fermentation process. [ABSTRACT FROM AUTHOR]

An editorial is presented which discusses the aspects of Design for Six Sigma (DFSS). The authors consider the association between DFSS and Six Sigma in terms of principle and practice. They explain that the differences between Six Sigma and DFSS are not restricted to their reactive and proactive natures. They comment on the quantitative techniques used in a DFSS project.

This section calls for papers for a 2008 issue of the "Quality and Reliability Engineering International (QREI)" on design methodology and theory, program management, proess capability evaluation and other related topics.

As the goals of ensuring process safety and energy efficiency become ever more challenging, engineers increasingly rely on data collected from such processes for informed decision making. During recent decades, extracting and interpreting valuable process information from large historical data sets have been an active area of research. Among the methods used, principal component analysis (PCA) is a well-established technique that allows for dimensionality reduction for large data sets by finding new uncorrelated variables, namely principal components (PCs). However, it is difficult to interpret the derived PCs, as each PC is a linear combination of all of the original variables and the loadings are typically nonzero. Sparse principal component analysis (SPCA) is a relatively recent technique proposed for producing PCs with sparse loadings via the variance-sparsity trade-off. We propose a forward SPCA approach that helps uncover the underlying process knowledge regarding variable relations. This approach systematically determines the optimal sparse loadings for each sparse PC while improving interpretability and minimizing information loss. The salient features of the proposed approach are demonstrated through the Tennessee Eastman process simulation. The results indicate how knowledge and process insight can be discovered through a systematic analysis of sparse loadings. [ABSTRACT FROM AUTHOR]

The present study aims at using statistically designed computational fluid dynamics (CFD) simulations as numerical experiments for the identification of one-dimensional (1-D) advection-dispersion models – computationally light tools, used e.g., as sub-models in systems analysis. The objective is to develop a new 1-D framework, referred to as interpreted CFD (iCFD) models, in which statistical meta-models are used to calculate the pseudo-dispersion coefficient ( D ) as a function of design and flow boundary conditions. The method – presented in a straightforward and transparent way – is illustrated using the example of a circular secondary settling tank (SST). First, the significant design and flow factors are screened out by applying the statistical method of two-level fractional factorial design of experiments. Second, based on the number of significant factors identified through the factor screening study and system understanding, 50 different sets of design and flow conditions are selected using Latin Hypercube Sampling (LHS). The boundary condition sets are imposed on a 2-D axi-symmetrical CFD simulation model of the SST. In the framework, to degenerate the 2-D model structure, CFD model outputs are approximated by the 1-D model through the calibration of three different model structures for D . Correlation equations for the D parameter then are identified as a function of the selected design and flow boundary conditions (meta-models), and their accuracy is evaluated against D values estimated in each numerical experiment. The evaluation and validation of the iCFD model structure is carried out using scenario simulation results obtained with parameters sampled from the corners of the LHS experimental region. For the studied SST, additional iCFD model development was carried out in terms of (i) assessing different density current sub-models; (ii) implementation of a combined flocculation, hindered, transient and compression settling velocity function; and (iii) assessment of modelling the onset of transient and compression settling. Furthermore, the optimal level of model discretization both in 2-D and 1-D was undertaken. Results suggest that the iCFD model developed for the SST through the proposed methodology is able to predict solid distribution with high accuracy – taking a reasonable computational effort – when compared to multi-dimensional numerical experiments, under a wide range of flow and design conditions. iCFD tools could play a crucial role in reliably predicting systems' performance under normal and shock events. [ABSTRACT FROM AUTHOR]

The reliability of electronic devices depends on the environmental loads at which they are exposed. Climatic conditions vary greatly from one geographical location to another (from hot and humid to cold and dry areas), and the temperature and humidity vary from season to season and from day to day. High levels of temperature and relative humidity mean high water content in the air, but saturated conditions (i.e. 100 % RH) can also be reached at low temperatures. This paper analyses the relationship between temperature, dew point temperature, their difference (here called ΔT), and occurrence and time period of dew point closeness to temperature on transient condensation effects on electronics. This paper has two parts: (i) Data analysis of typical climate profiles within the different zones of the Köppen -Geiger classification to pick up conditions where ΔT is very low (for example ≤0.4 °C). Various summary statistics of these events are calculated in order to assess the temperature at which these events happen, their durations and their frequency and (ii) Empirical investigation of the effect of ΔT ≤ 0.4 °C on the reliability of electronics by mimicking an electronic device, for which the time period of the ΔT is varied in one set of experiments, and the ambient temperature is varied in the other. The effect of the packaging of the electronics is also studied in this section. The statistical study of the climate profiles shows that the transient events (ΔT ≤ 0.4 °C) occur in almost every location, at different temperature levels, with a duration of at least one observation (where observations were hourly in the database). The experimental results show that presence of the enclosure, cleanliness and bigger pitch size reduce the levels of leakage current, while similar high levels of leakage current are observed for the different durations of the transient events, indicating that these climatic transient conditions can have a big impact on the electronics reliability. • Statistical climate data analysis showed that almost all of the locations experience the transient risk of condensation. • Short durations of the transient events resulted in similar high levels of LC compared to longer durations. • The enclosure protection has buffered the humidity change experienced by the SIR PCB placed inside. • The hygroscopic nature of the weak organic acids has the highest impact on LC and ECM formation. [ABSTRACT FROM AUTHOR]

Modern products frequently feature monitors designed to detect actual or impending malfunctions. False alarms (Type I errors) or excessive delays in detecting real malfunctions (Type II errors) can seriously reduce monitor utility. Sound engineering practice includes physical evaluation of error rates. Type II error rates are relatively easy to evaluate empirically. However, adequate evaluation of a low Type I error rate is difficult without using accelerated testing concepts, inducing false alarms using artificially low thresholds and then selecting production thresholds by appropriate extrapolation, as outlined here. This acceleration methodology allows for informed determination of detection thresholds and confidence in monitor performance with substantial reductions over current alternatives in time and cost required for monitor development. Copyright © 2006 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

Data‐driven process monitoring and control techniques and their application to industrial chemical processes are gaining popularity due to the current focus on Industry 4.0, digitalization and the Internet of Things. However, for the development of such techniques, there are significant barriers that must be overcome in obtaining sufficiently large and reliable datasets. As a result, the use of real plant and process data in developing and testing data‐driven process monitoring and control tools can be difficult without investing significant efforts in acquiring, treating, and interpreting the data. Therefore, researchers need a tool that effortlessly generates large amounts of realistic and reliable process data without the requirement for additional data treatment or interpretation. In this work, we propose a data generation platform based on the Tennessee Eastman Process simulation benchmark. A graphical user interface (GUI) developed in MATLAB Simulink is presented that enables users to generate massive amounts of data for testing applicability of big data concepts in the realm of process control for continuous time‐dependent processes. An R‐Shiny app that interacts with the data generation tool is also presented for illustration purposes. The app can visualize the results generated by the Tennessee Eastman Process and can carry out a standard fault detection and diagnosis studies based on PCA. The data generator GUI is available free of charge for research purposes at https://github.com/dtuprodana/TEP. [ABSTRACT FROM AUTHOR]

The concurrent use of statistical process control and engineering process control involves monitoring manipulated and controlled variables. One multivariate control chart may handle the statistical monitoring of all variables, but observing the manipulated and controlled variables in separate control charts may improve understanding of how disturbances and the controller performance affect the process. In this article, we illustrate how step and ramp disturbances manifest themselves in a single‐input–single‐output system by studying their resulting signatures in the controlled and manipulated variables. The system is controlled by variations of the widely used proportional‐integral‐derivative(PID) control scheme. Implications for applying control charts for these scenarios are discussed. [ABSTRACT FROM AUTHOR]

This paper introduces a novel methodology for analyzing three-way array data with a multi-group structure. Three-way arrays are commonly observed in various domains, including image analysis, chemometrics, and real-world applications. In this paper, we use a practical case study of process modeling in additive manufacturing, where batches are structured according to multiple groups. Vast volumes of data for multiple variables and process stages are recorded by sensors installed on the production line for each batch. For these three-way arrays, the link between the final product and the observations creates a grouping structure in the observations. This grouping may hamper gaining insight into the process if only some of the groups dominate the controlled variability of the products. In this study, we develop an extension of the PARAFAC model that takes into account the grouping structure of three-way data sets. With this extension, it is possible to estimate a model that is representative of all the groups simultaneously by finding their common structure. The proposed model has been applied to three simulation data sets and a real manufacturing case study. The capability to find the common structure of the groups is compared to PARAFAC and the insights into the importance of variables delivered by the models are discussed. • A new methodology for the analysis of multi-group three-way datasets. • The model is an extension of the PARAFAC model that considers the multi-group structure of the data. • The model delivers loadings coefficients that account for the common variability of all the groups present in the data. • Simulations and a real-case study were used to showcase the aspects of the proposed model. [ABSTRACT FROM AUTHOR]

We analyse the movement of pig herds through video recordings at a slaughterhouse by using statistical analysis of optical flow (OF) patterns. Unlike the previous attempts to analyse pig movement, no markers, trackers nor identification of individual pigs are needed. Our method handles the analysis of unconstrained areas where pigs are constantly entering and leaving. The goal is to improve animal welfare by real-time prediction of abnormal behaviour through proper interventions. The aim of this study is to identify any stationary pig, which can be an indicator of an injury or an obstacle. In this study, we use the OF vectors to describe points of movement on all pigs and thereby analyse the herd movement. Subsequently, the OF vectors are used to identify abnormal movements of individual pigs. The OF vectors, obtained from the pigs, point in multiple directions rather than in one movement direction. To accommodate the multiple directions of the OF vectors, we propose to quantify OF using a summation of the vectors into bins according to their angles, which we call modified angular histograms. Sequential feature selection is used to select angle ranges, which identify pigs that are moving abnormally in the herd. The vector lengths from the selected angle ranges are compared to the corresponding median, 25th and 75th percentiles from a training set, which contains only normally moving pigs. We show that the method is capable of locating stationary pigs in the recordings regardless of the number of pigs in the frame. [ABSTRACT FROM AUTHOR]

Humane handling and slaughter of livestock are of major concern in modern societies. Monitoring animal wellbeing in slaughterhouses is critical in preventing unnecessary stress and physical damage to livestock, which can also affect the meat quality. The goal of this study is to monitor pig herds at the slaughterhouse and identify undesirable events such as pigs tripping or stepping on each other. In this paper, we monitor pig behavior in color videos recorded during unloading from transportation trucks. We monitor the movement of a pig herd where the pigs enter and leave a surveyed area. The method is based on optical flow, which is not well explored for monitoring all types of animals, but is the method of choice for human crowd monitoring. We recommend using modified angular histograms to summarize the optical flow vectors. We show that the classification rate based on support vector machines is 93% of all frames. The sensitivity of the model is 93.5% with 90% specificity and 6.5% false alarm rate. The radial lens distortion and camera position required for convenient surveillance make the recordings highly distorted. Therefore, we also propose a new approach to correct lens and foreshortening distortions by using moving reference points. The method can be applied real-time during the actual unloading operations of pigs. In addition, we present a method for identification of the causes leading to undesirable events, which currently only runs off-line. The comparative analysis of three drivers, which performed the unloading of the pigs from the trucks in the available datasets, indicates that the drivers perform significantly differently. Driver 1 has 2.95 times higher odds to have pigs tripping and stepping on each other than the two others, and Driver 2 has 1.11 times higher odds than Driver 3. [ABSTRACT FROM AUTHOR]

The in vivo Comet assay is a sensitive method for evaluating DNA damage. A recurrent concern is how to analyze the data appropriately and efficiently. A popular approach is to summarize the raw data into a summary statistic prior to the statistical analysis. However, consensus on which summary statistic to use has yet to be reached. Another important consideration concerns the assessment of proper sample sizes in the design of Comet assay studies. This study aims to identify a statistic suitably summarizing the % tail DNA of mice testicular samples in Comet assay studies. A second aim is to provide curves for this statistic outlining the number of animals and gels to use. The current study was based on 11 compounds administered via oral gavage in three doses to male mice: CAS no. 110-26-9, CAS no. 512-56-1, CAS no. 111873-33-7, CAS no. 79-94-7, CAS no. 115-96-8, CAS no. 598-55-0, CAS no. 636-97-5, CAS no. 85-28-9, CAS no. 13674-87-8, CAS no. 43100-38-5 and CAS no. 60965-26-6. Testicular cells were examined using the alkaline version of the Comet assay and the DNA damage was quantified as % tail DNA using a fully automatic scoring system. From the raw data 23 summary statistics were examined. A linear mixed-effects model was fitted to the summarized data and the estimated variance components were used to generate power curves as a function of sample size. The statistic that most appropriately summarized the within-sample distributions was the median of the log-transformed data, as it most consistently conformed to the assumptions of the statistical model. Power curves for 1.5-, 2-, and 2.5-fold changes of the highest dose group compared to the control group when 50 and 100 cells were scored per gel are provided to aid in the design of future Comet assay studies on testicular cells. [ABSTRACT FROM AUTHOR]

• Developing a novel taxonomy for railway track maintenance planning and scheduling (RTMP&S) decision-making models. • Discussing the differences in planning and scheduling problems in railway maintenance. • Considering the structural characteristics of the railway track that can affect decision-making models. • Reviewing the attributes of maintenance management decisions in RTMP&S decision-making. • Summarising the optimisation frameworks for modelling the RTMP&S problems and the proposed solution approaches in the literature. • Discussing research trends can help researchers and practitioners to have a clear understanding of the state of the art of RTMP&S problems and future research directions. Railway track maintenance and renewal are vital for railway safety, train punctuality, and travel comfort. Therefore, having cost-effective maintenance is critical in managing railway infrastructure assets. There has been a considerable amount of research performed on mathematical and decision support models for improving the application of railway track maintenance planning and scheduling. This article reviews the literature in decision support models for railway track maintenance planning and scheduling and transforms the results into a problem taxonomy. Furthermore, the article discusses current approaches in optimising maintenance planning and scheduling, research trends, and possible gaps in the related decision-making models. [Display omitted] [ABSTRACT FROM AUTHOR]

• Modified cost-sensitive classification strategy for an industrial problem. • Decision rule for going through or skipping elements of last quality control stage. • Trade-off problem between cost and quality. • Modified strategy can also be applied on problems without an associated cost. In the current era of Industry 4.0, sensor data used in connection with machine learning algorithms can help manufacturing industries to reduce costs and to predict failures in advance. This paper addresses a binary classification problem found in manufacturing engineering, which focuses on how to ensure product quality delivery and at the same time to reduce production costs. The aim behind this problem is to predict the number of faulty products, which in this case is extremely low. As a result of this characteristic, the problem is reduced to an imbalanced binary classification problem. The authors contribute to imbalanced classification research in three important ways. First, the industrial application coming from the electronic manufacturing industry is presented in detail, along with its data and modelling challenges. Second, a modified cost-sensitive classification strategy based on a combination of Voronoi diagrams and genetic algorithm is applied to tackle this problem and is compared to several base classifiers. The results obtained are promising for this specific application. Third, in order to evaluate the flexibility of the strategy, and to demonstrate its wide range of applicability, 25 real-world data sets are selected from the KEEL repository with different imbalance ratios and number of features. The strategy, in this case implemented without a predefined cost, is compared with the same base classifiers as those used for the industrial problem. [ABSTRACT FROM AUTHOR]