Multiway Principal Component Analysis Research Articles

Time-slice dynamic prediction and multiway serial PCA for batch industrial process monitoring

This paper introduces an innovative method for batch industrial process monitoring and fault detection. By combining time-slice dynamic prediction with an enhanced adaptive multiway principal component analysis (PCA) technique, it aims to balance the trade-off between high fault detection rates and low false alarm rates in monitoring batch industrial data. The method extracts dynamic features from 3D batch data time slices and utilizes a serial PCA framework with variable control limits for real-time monitoring. Comparative evaluations against traditional adaptive multiway PCA and dynamic time-slice canonical correlation analysis approaches demonstrate its superior performance. A case study using a novel cold rolling mill simulation model underscores its effectiveness for batch industrial process monitoring and fault detection.

A model-free-based control chart for batch process using U-statistics

Batch processes are known to generate time series of successive measurements of many process variables in each run and the main challenge is to capture and accommodate the variability in the batch domain (batch-to-batch variability) and in the time domain (data with serial correlation). Classical approaches are focused on the first goal by applying multivariate techniques in the columns of a data matrix, in which each row has the data for the entire batch run. Those approaches are grounded in the traditional Multiway Principal Component Analysis (MPCA). Recent approaches are focused on the balance between the two sources, taking into account the time series nature of the data mainly by using time series models and neural network methods to directly accommodate the variability in the time domain. This paper proposes a model-free approach based on U-statistics theory. Through this theory, we can build a group of control charts (named V charts) capable of monitoring batch processes considering both sources of variability. In general, the new batch under monitoring is evaluated according to a group of reference in-control samples, based on the distance measures from those batches. The proposed approach is really flexible since we avoid model identification, parameter estimates issues and it is able to accommodate the within and between batch variability. Additionally, our method is suitable to deal with a wide range of batch data features, since under weak conditions, the distribution of the V statistics remains well known. It makes the V-based charts able to be used in different kinds of processes. We are focused here on controlling changes caused by data drifts (including trends) and by data serial correlation (data dynamics). Through the simulated and real batch process, we show that the V charts work well, including a scenario with very few reference batches available.

Evolution stage identification of haze pollution episodes in beijing using constrained dynamic time warping and multiway principal component analysis

Evolution stage identification of haze pollution episodes in beijing using constrained dynamic time warping and multiway principal component analysis

Generalizability of Soft Sensors for Bioprocesses through Similarity Analysis and Phase-Dependent Recalibration.

A soft sensor concept is typically developed and calibrated for individual bioprocesses in a time-consuming manual procedure. Following that, the prediction performance of these soft sensors degrades over time, due to changes in raw materials, biological variability, and modified process strategies. Through automatic adaptation and recalibration, adaptive soft sensor concepts have the potential to generalize soft sensor principles and make them applicable across bioprocesses. In this study, a new generalized adaptation algorithm for soft sensors is developed to provide phase-dependent recalibration of soft sensors based on multiway principal component analysis, a similarity analysis, and robust, generalist phase detection in multiphase bioprocesses. This generalist soft sensor concept was evaluated in two multiphase bioprocesses with various target values, media, and microorganisms. Consequently, the soft sensor concept was tested for biomass prediction in a Pichia pastoris process, and biomass and protein prediction in a Bacillus subtilis process, where the process characteristics (cultivation media and cultivation strategy) were varied. High prediction performance was demonstrated for P. pastoris processes (relative error = 6.9%) as well as B. subtilis processes in two different media during batch and fed-batch phases (relative errors in optimized high-performance medium: biomass prediction = 12.2%, protein prediction = 7.2%; relative errors in standard medium: biomass prediction = 12.8%, protein prediction = 8.8%).

Online Monitoring of Penicillin Manufacture Based on Production Variables and Metabolic Fluxes

The existing data-based monitoring and diagnosis methods for microbial manufacturing processes only consider macroscopic production variables and ignore microscopic metabolic mechanisms and flux changes. This paper proposed a fault monitoring strategy based on metabolic flux variables (intracellular and extracellular exchange fluxes) by introducing microbial cellular metabolic network information. First, the metabolic network of the microbial manufacturing process was analyzed and the dynamic metabolic flux analysis (DMFA) was used to obtain the dynamic distribution of metabolic fluxes inside microbial cells based on measurable extracellular metabolites concentration. Then, production variables and flux variables were combined to constitute a sample dataset of the microbial cell growth process. On this basis, a multiway principal component analysis (MPCA) method was used to establish a fault diagnosis model to monitor the process. Finally, the method’s effectiveness was verified using the fault monitoring strategy proposed in this paper for online monitoring of the penicillin manufacturing process.

A comparative study on classification of edible vegetable oils by infrared, near infrared and fluorescence spectroscopy combined with chemometrics

Driven by economic benefits like any other foods, vegetable oil has long been plagued by mislabeling and adulteration. Many studies have addressed the field of classification and identification of vegetable oils by various analysis techniques, especially spectral analysis. A comparative study was performed using Fourier transform infrared spectroscopy (FTIR), visible near-infrared spectroscopy (Vis-NIR) and excitation-emission matrix fluorescence spectroscopy (EEMs) combined with chemometrics to distinguish different types of edible vegetable oils. FTIR, Vis-NIR and EEMs datasets of 147 samples of five vegetable oils from different brands were analyzed. Two types of pattern recognition methods, principal component analysis (PCA)/multi-way principal component analysis (M-PCA) and partial least squares discriminant analysis (PLS-DA)/multilinear partial least squares discriminant analysis (N-PLS-DA), were used to resolve these data and distinguish vegetable oil types, respectively. PCA/M-PCA analysis exhibited that three spectral data of five vegetable oils showed a clustering trend. The total correct recognition rate of the training set and prediction set of FTIR spectra of vegetable oil based on PLS-DA method are 100%. The total recognition rate of Vis-NIR based on PLS-DA are 100% and 97.96%. However, the total correct recognition rate of training set and prediction set of EEMs data based on N-PLS-DA method is 69.39% and 75.51%, respectively. The comparative study showed that FTIR and Vis-NIR combined with chemometrics were more suitable for vegetable oil species identification than EEMs technique. The reason may be concluded that almost all chemical components in vegetable oil can produce FTIR and NIR absorption, while only a small amount of fluorophores can produce fluorescence. That is, FTIR and NIR can provide more spectral information than EEMs. Analysis of EEMs data using self-weighted alternating trilinear decomposition (SWATLD) also showed that fluorophores were a few and irregularly distributed in vegetable oils.

Impact of cooking style and oil on semi-volatile and intermediate volatility organic compound emissions from Chinese domestic cooking

Abstract. To elucidate the molecular chemical compositions, volatility–polarity distributions, and influencing factors of Chinese cooking emissions, a comprehensive cooking emission experiment was conducted. Volatile organic compounds (VOCs), intermediate volatility, and semi-volatile organic compounds (I/SVOCs) from cooking fumes were analysed by a thermal desorption comprehensive two-dimensional gas chromatography coupled with quadrupole mass spectrometer (TD-GC × GC-qMS). Emissions from four typical Chinese dishes, i.e. fried chicken, Kung Pao chicken, pan-fried tofu, and stir-fried cabbage were investigated to illustrate the impact of cooking style and material. Fumes of chicken fried with corn, peanut, soybean, and sunflower oils were investigated to demonstrate the influence of cooking oil. A total of 201 chemicals were quantified. Kung Pao chicken emitted more pollutants than other dishes due to its rather intense cooking method. Aromatics and oxygenated compounds were extensively detected among meat-related cooking fumes, while a vegetable-related profile was observed in the emissions of stir-fried cabbage. Ozone formation potential (OFP) was dominated by chemicals in the VOC range. Of the secondary organic aerosol (SOA) estimation, 10.2 %–32.0 % could be explained by S/IVOCs. Pixel-based partial least squares discriminant analysis (PLS-DA) and multiway principal component analysis (MPCA) were utilized for sample classification and component identification. The results indicated that the oil factor explained more variance of chemical compositions than the cooking style factor. MPCA results emphasize the importance of the unsaturated fatty acid-alkadienal-volatile products mechanism (oil autoxidation) accelerated by the cooking and heating procedure.

The Complexity of Interaction between Executive Board Gender Diversity and Financial Performance: A Panel Analysis Approach Based on Random Effects

This study examined the influence of the executive board of directors’ gender diversity on the financial performance of listed companies on the Bucharest Stock Exchange, for the period 2011 to 2019. The analysis of the composition and different characteristics of the board and the executive directors proved to be effective tools for corporate governance in countries with an emerging capital market. Therefore, a disclosure index on directors’ characteristics was used to moderate the interaction between gender diversity and financial performance, based on the theoretical framework provided by upper echelon theory. The study contributes to the enrichment of the literature both by using the composite indicator built by applying the multiway PCA method on panel data to express financial performance and by designing the ten EGLS panel models involving five financial indicators and two proxies for gender diversity. The results showed that there is a positive impact of the proportion of women on the executive board of directors on financial performance, measured through the composite index, ROA, ROE, and SOL. A statistically significant impact of gender diversity on financial performance was found only for SOL, in the case of the Blau index. Also, using the random-effects model to perform the panel data analysis, the results showed that a higher executive board size can be associated with better financial performance measured through the composite index, ROA, ROE, and EPS. Practical implications are significant for the board of executives’ composition, the complexity of the relationship with the board, and reshaping governance practices.

A novel unsupervised approach for batch process monitoring using deep learning

Process monitoring is an important tool used to ensure safe operation of a process plant and to maintain high quality of end products. The focus of this work is on unsupervised Statistical Process Control (SPC) of batch processes using Deep Learning (DL). A DL architecture referred as Multiway Partial Least Squares Autoencoder (MPLS-AE) is proposed and trained using a genetic optimization algorithm with a novel objective function that directly maximizes the average fault detection rate (FDR¯). The efficacy of the proposed method is demonstrated on an industrial scale Penicillin process. Comparisons of the proposed algorithm with linear Multiway Principal Component Analysis (MPCA) and Multiway Partial Least Squares (MPLS) based fault detection (FD) algorithm, trained with the same objective as used by the DL model, demonstrates the superiority of the deep learning based approach. The use of dynamic control limits significantly improves the detection rates for both the linear and DL models.

Investigation of partition coefficients and fingerprints of atmospheric gas- and particle-phase intermediate volatility and semi-volatile organic compounds using pixel-based approaches

Ambient gas- and particle-phase intermediate volatility and semi-volatile organic compounds (I/SVOCs) of Beijing were analyzed by a thermal desorption comprehensive two-dimensional gas chromatography quadrupole mass spectrometry (TD-GC × GC-qMS). A pixel-based scheme combing the integration-based approach was applied for partition coefficients estimation and fingerprints identification. Blob-by-blob recognition was firstly utilized to characterize I/SVOCs from the molecular level. 412 blobs in gas-phase and 460 blobs in particle-phase were resolved, covering a total response of 47.5% and 43.5%. A large pool of I/SVOCs was found with a large diversity of chemical classes in both gas- and particle-phase. Acids (8.5%), b-alkanes (5.8%), n-alkanes (C8-C25, 5.3%), and aromatics (4.4%) were dominant in gas-phase while esters (7.0%, including volatile chemical product compounds, VCPs), n-alkanes (C9-C34, 5.7%), acids (4.6%), and siloxanes (3.6%) were abundant in particle-phase. Air pollutants were then evaluated by a two-parameter linear free energy relationship (LFER) model, which could be further implemented in the two-dimensional volatility basis set (2D-VBS) model. Multiway principal component analysis (MPCA) and partial least squares-discriminant analysis (PLS-DA) implied that naphthalenes, phenol, propyl-benzene isomers, and oxygenated volatile organic compounds (OVOCs) were key components in the gas-phase under different pollution levels. This work gives more insight into property estimation and fingerprints identification for complex ambient samples.

Evolution Stage Identification of Haze Pollution Episodes Over the North China Plain Using Constrained Dynamic Time Warping and Multiway Principal Component Analysis

Evolution Stage Identification of Haze Pollution Episodes Over the North China Plain Using Constrained Dynamic Time Warping and Multiway Principal Component Analysis

Feature extraction and improved denoising method for nonlinear and nonstationary high-rate GNSS coseismic displacements applied to earthquake focal mechanism inversion of the El Mayor–Cucapah earthquake

In this study, for feature extraction of seismic- and nonlinear trend terms of the nonlinear and nonstationary high-rate Global Navigation Satellite System (GNSS) seismic displacements, we present an adaptive complete ensemble empirical mode decomposition (CEEMD) method based on correlation coefficient. We also introduce an improved CEEMD (ICEEMD) based on multiway principal component analysis (MPCA), hereafter, ICEEMD-MPCA, denoising method for boosting the high-rate GNSS seismic displacements with/without nonlinear trend. The results show that the proposed method is suitable for extracting features of nonlinear and nonstationary high-rate GNSS seismic displacements. The ICEEMD-MPCA denoising method is able to obviously eliminate white noise at high frequencies, remove systematic errors, and effectively preserve the earthquake wave signals. The denoising method reflects that high-rate GNSS has millimeter-level accuracy in the vertical component and submillimeter-level accuracy in the horizontal component. To verify the performance characteristics of the feature extraction and denoising method, the focal mechanism inversion for the 2010 El Mayor–Cucapah earthquake is estimated using the high-rate GNSS data (5 Hz) and collocated strong motion data (50 and 200 Hz). The focal mechanism of the earthquake estimated using the denoised high-rate GNSS with/without nonlinear trend are more consistent with the teleseismic estimates than those estimated using the raw high-rate GNSS and strong motion data. For the GNSS observations, the inversion results are more affected by errors in low-frequency, which accumulate over time and display their signatures as trend in coseismic displacements. It can be concluded that with the adaptive CEEMD method and the ICEEMD-MPCA denoising method, high-rate GNSS at near-field stations only or combined with GNSS/strong-motion records can be applied to yield better information regarding the earthquake focal mechanism inversion, the lower bounds of earthquake magnitudes, and earthquake rupture parameters for small to moderate earthquakes.

A Real-Time Monitoring Framework for Wafer Fabrication Processes With Run-to-Run Variations

Real-time process monitoring is a vital technique to maintain safety and quality in the wafer fabrication processes. Hence, statistical process monitoring methods such as statistical process control chart, principal component analysis have been widely applied in wafer fabrication. However, these methods often suffer from the performance degradation problem caused by run-to-run (R2R) variations such as uneven duration and R2R drifts. In this paper, we propose a real-time monitoring framework for continuous wafer fabrication processes with uneven duration and R2R drifts. In this framework, processes are divided into several phases and further aligned in a real-time manner to solve the uneven duration problem. Then, R2R drifts are compensated by a down-sampling seasonal autoregressive integrated moving average model and a moving window multi-batch forecast strategy. Finally, multiphase multiway principal component analysis models are built on the compensated data to monitoring the processes. The efficiency of this framework is demonstrated by a case study on a plasma etch process. The results indicate that the proposed framework can diminish the influence of R2R variations and improve the monitoring model performance.

Fault Detection and Diagnosis for Plasticizing Process of Single-Base Gun Propellant Using Mutual Information Weighted MPCA under Limited Batch Samples Modelling

Aiming at the lack of reliable gradual fault detection and abnormal condition alarm and evaluation ability in the plasticizing process of single-base gun propellant, a fault detection and diagnosis method based on normalized mutual information weighted multiway principal component analysis (NMI-WMPCA) under limited batch samples modelling was proposed. In this method, the differences of coupling correlation among multi-dimensional process variables and the coupling characteristics of linear and nonlinear relationships in the process are considered. NMI-WMPCA utilizes the generalization ability of a multi-model to establish an accurate fault detection model in limited batch samples, and adopts fault diagnosis methods based on a multi-model SPE statistic contribution plot to identify the fault source. The experimental results demonstrate that the proposed method is effective, which can realize the rapid detection and diagnosis of multiple faults in the plasticizing process.

Improving selective channel occlusion of complex hydrocarbons and fatty acid methyl esters in urea crystals by using an expendable 3D-printed microfluidic device for sample preparation in untargeted petroleomics

In this study, we describe a proof-of-concept investigation of the potential and limitations of employing channel occlusion for sample preparation in untargeted analysis in petroleomics. A middle petroleum distillate composed of fatty acid methyl esters (FAME) and a complex mixture of linear, branched, and cyclic hydrocarbons were selected as the model samples for this investigation. A microfluidic device was engineered to overcome the limitations of channel occlusion, resulting in a quick and robust method for sample preparation. The 3D-printed device using fused deposition modelling (FDM) allowed the combination of a 13-h multi-step sample handling protocol into a 2-min single-step procedure, which is also automation-friendly. Such developments were also evaluated using the analytical eco-scale to guide the development of a green analytical method. The relative standard deviation decreased 2-fold with method miniaturization. The efficiency of n-alkane removal was extended from tridecane (n-C13) to heptadecane (n-C17), compared to original method (n-C16 to n-C17). The analytical performance of the method was investigated for untargeted analysis. The tool used to probe the intra- and inter-class variance was multi-way principal component analysis (MPCA). MPCA modelling revealed that both methods generated equivalent chemical information, highlighting the benefits of reliable and reproducible sample preparation methods, especially for untargeted analysis. Such awareness is critical to avoid the generation of misleading results in fields that heavily rely on untargeted analysis and fingerprinting, such as petroleomics.

Multivariate statistical process control methods for batch production: a review focused on applications

ABSTRACT In this paper, we highlight the basic techniques of multivariate statistical process control (MSPC) under the dimensionality criteria, such as Multiway Principal Component Analysis, Multiway Partial Squares, Structuration à Trois Indices de la Statistique, Tucker3, Parallel Factors, Multiway Independent Component Analysis, Multiset Canonical Correlation Analysis, Slow Features Analysis, and Parallel Coordinates. Furthermore, we summarize the procedures of each statistical technique and the usage of multivariate control charts. In addition, we review the most significant proposals and applications in practical cases. Finally, we compare and discuss the benefits and limitations within methods.

Stable development of travel business in the Odessa region: Estimate using multivariate statistical analysis

The article examines the level of sustainable development of the travel business in the Odessa region of Ukraine. This level is evaluated as a latent indicator, which is characterized by a certain set of signs-symptoms. The methodology used for modeling is Partial Least Squares-Path Modelling (or PLS-PM) and the method Time-Wise Multi-Way Principal Component Analysis. This approach allowed us to identify and evaluate latent and explicit factors affecting the sustainable development of travel business in the Odessa region.

Hydrocarbon transformations in sediments from the Cathedral Hill hydrothermal vent complex at Guaymas Basin, Gulf of California – A chemometric study of shallow seep architecture

Guaymas Basin isa submarine depression in the Gulf of California marking the northern end of the East Pacific Rise mid-oceanic spreading ridge.The basin receives high input of sedimentary organic matter (SOM) from elevated productivity in the overlying surface waters and runoff from the surrounding continent. This, coupled with high sedimentation rates, produces near-uniform compositions of SOM.Various hydrothermal vent complexes occur along this margin. One of these is Cathedral Hill, a hydrothermal mound with sulfide chimneys surrounded by mats dominated by sulfide-oxidizing Beggiatoa, covering sediments stained by locally produced oil. We collected four push cores along a transect at this site (4 m × 0.21 m) extending from near the vent center to the ambient sediment just outside the microbial mat cover. Porewater temperatures near the mound center were projected to reach 155 °C by 21 cm below the sea floor (cmbsf). Within these conditions, a kinetic model based on vitrinite reflectance equivalence (%Re) predicts petroleum formation as shallow as 15–18 cmbsf, with metagenesis commencing at <60 cmbsf. Bulk extract data (total lipid extracts as well as polar and apolar fractions), solvent-extracted sediment TOC (herein referred to as protokerogen TOC), and molecular thermal maturation parameters support these generation estimates. In recent years, the application of chemometric techniques tocomprehensive two-dimensional gas chromatographic (GC × GC) analyses has allowed comparison of thousands of unique hydrocarbons within oils. Here we reconstruct the shallow subsurface petroleum system by applying multiway principal component analysis (MPCA) and hierarchical cluster analysis (HCA)directly to GC × GC chromatograms. We then compare the resulting multivariate models to a systematic survey of subtracted GC × GC chromatograms, a transect heat map of sample hydrocarbon compound diversities, and profiles of various thermal maturation parameters to elucidate how these hydrocarbon matrices are attenuated by production, migration, and/or thermochemical oxidation. Sample matrices have up to 5800 unique compounds spanning a range ofnormal and branched alkanes, saturated and unsaturated biomarkers, substituted and unsubstituted polycyclic aromatic hydrocarbons (PAHs), perhydro-PAHs, and benzothiopheneswith up to six ring-cycles (i.e., benzoperylenes, dibenzopyrenes, dibenzochrysenes, and indenopyrenes). These matrices display systematic temperature-dependent trends. Generation likely begins in 6–10 and 15–18 cmbsf where sediments are exposed to ∼115 °C vent porewater temperatures, which is shallower than predictions based on our kinetic model. Independent of these sites of generation is ubiquitous staining of the sediments from advected oil that is heavily dominated by PAHs as well as two stratigraphic bandsof migrated oil that extend horizontally across the transect at 0–2 and ∼6–10 cmbsf, respectively.The MPCA models along with non-statistical validation techniques show evidence of decreasingdiversities and concentrations of alkylated aromatic hydrocarbons concomitant with elevated abundances of dealkylated PAHs and/or the migration of unsubstituted PAHs from deeper basin depths as sediments become exposed to more severe hydrothermal conditions. These results indicate that even at relatively small spatial scales, the petroliferous sediments at hydrothermal vent sites can be highly complex.

Multidimensional data statistical processing of magnetic flow leakage signals from a Colombian gas pipeline

The hydrocarbon industry in Colombia is one of the principal pillars for the Colombian economy, representing around 5% of its gross domestic product. Since petroleum reserves have decreased, gas becomes one main alternative for economical growth. However, current gas pipelines have been in service for over 30 years and some of them are buried and phenomena, such as metal losses, corrosion, mechanical stress, strikes by excavation machinery, and another type of damages, are presented. The maintenance program of these structures is typically corrective type and is very expensive. To overcome this situation, the native research institute “Research Institute of Corrosion—Corporación para la Investigación de la Corrosión” recently developed an in-line inspection tool to be operated in Colombian gas pipelines to get valuable information of their current state along thousands of kilometers. A huge quantity of data is recorded (including tool movement, magnet, magnetic flow leakage, and caliper signals), which demand a high-computational cost and an adequate tool analysis to establish the current pipeline structural health condition. In this sense, authors have shown in several works that principal component analysis is an effective tool to detect and locate abnormal operational structural conditions from multidimensional data. In a previous analysis, multidimensional data were used to locate possible damages along the pipeline. However, most of the activated points belonged to weld points. Then, in this article, it is proposed to use the root mean square value of magnetic flux leakage signals to separate these points and to obtain sets of signals by sections removing the welds, and then multiway principal component analysis is applied for each set of signals of each gas pipeline section. The maximum values of damage indices (Q and -statistics) of each section are conserved to activate the sections of the gas pipeline with more probability of damages and then, they must be evaluated by experts.

Sub-period division strategies combined with multiway principle component analysis for fault diagnosis on sequence batch reactor of wastewater treatment process in paper mill

Fault diagnosis of sequential batch reactor (SBR), a widely applied wastewater treatment technology in papermaking industry with huge discharge, has been a significant challenge due to the inherent multi-period characteristics of the process. In this paper, based on the conventional multi-way principal component analysis (MPCA) method (Scenario 0), two sub-period division strategies based on the processing phases of SBR process (Scenario 1) and the similarities of the loading matrices between the adjacent time slices (Scenario 2) are proposed for the detection of faults. Combined with Scenario 0, using the field data of blower current, level of SBR reactor, dissolved oxygen of wastewater and the blower valve opening in the SBR process of paper mill, two different fault diagnosis models with Scenarios 1 and 2 are developed and evaluated, respectively. The study results revealed that, both the calculated statistics of T2 and sum of prediction errors (SPE) of the fault diagnosis models with Scenarios 1 and 2 could detect the faults and identify the fault locations and sources. Compared to Scenario 0 which neglects the correlations between the different stages of SBR process, the fault diagnosis model by Scenario 2 demonstrated a superiority ability in fault identification in terms of the fault’s time onset and fault’s sources with adequate accuracy. The results enable the feasible and reliable implementation of the developed sub-MPCA diagnosis model with Scenario 2 in the actual SBR plants.