Nonlinear Principal Component Analysis Research Articles

A multivariate-based variable selection framework for clustering traffic conflicts in a brazilian freeway

More than one million people die or suffer non-fatal injuries annually due to road accidents around the world. Understanding the causes that give rise to different types of conflict events, as well as their characteristics, can help researchers and traffic authorities to draw up strategies aimed at mitigating collision risks. This paper proposes a framework for grouping traffic conflicts relying on similar profiles and factors that contribute to conflict occurrence using self-organizing maps (SOM). In order to improve the quality of the formed groups, we developed a novel variable importance index relying on the outputs of the nonlinear principal component analysis (NLPCA) that intends to identify the most informative variables for grouping collision events. Such index guides a backward variable selection procedure in which less relevant variables are removed one-by-one; after each removal, the clustering quality is assessed via the Davies-Bouldin (DB) index. The proposed framework was applied to a real-time dataset collected from a Brazilian highway aimed at allocating traffic conflicts into groups presenting similar profiles. The selected variables suggest that lower average speeds, which are typically verified during congestion events, contribute to conflict occurrence. Higher variability on speed (denoted by high standard deviation, and speed’s coefficient of variation levels on that variable), which are also perceived in the assessed freeway near to congestion periods, also contribute to conflicts.

Resting state fMRI analysis using unsupervised learning algorithms

ABSTRACTRS-fMRI data analysis for functional connectivity explorations is a challenging topic in computational neuroimaging. Several approaches have been investigated to discover whole-brain data features. Among these, clustering techniques based on Competitive Learning (CL) and Spectral Methods (SM) have been shown effective in providing useful information in various contexts. We selected three clustering algorithms and two spectral methods, i.e the clustering algorithm are Self-organising Maps (SOM), Neural Gas (NG) and Growing Neural Gas (GNG), whereas the spectral methods are the classic Principal Component Analysis (PCA) and the Nonlinear Robust Fuzzy Principal Component Analysis (NRFPCA). We validated clustering with Davies–Bouldin Index (DBI) and we selected informative principal components using Random Matrix Theory (RMT). tools. We adopted these techniques to study the intrinsic functional properties of images coming from a shared repository of resting state fMRI experiments (1000 Functional Connectome Project).

Neural Networks Principal Component Analysis for estimating the generative multifactor model of returns under a statistical approach to the Arbitrage Pricing Theory. Evidence from the Mexican Stock Exchange

A nonlinear principal component analysis (NLPCA) represents an extension of the standard principal component analysis (PCA) that overcomes the limitation of the PCA’s assumption about the linearity of the model. The NLPCA belongs to the family of nonlinear versions of dimension reduction or the extraction techniques of underlying features, including nonlinear factor analysis and nonlinear independent component analysis, where the principal components are generalized from straight lines to curves. The NLPCA can be achieved via an artificial neural network specification where the PCA classic model is generalized to a nonlinear mode, namely, Neural Networks Principal Component Analysis (NNPCA). In order to extract a set of nonlinear underlying systematic risk factors, we estimate the generative multifactor model of returns in a statistical version of the Arbitrage Pricing Theory (APT), in the context of the Mexican Stock Exchange. We used an auto-associative multilayer perceptron neural network or autoencoder, where the ‘bottleneck’ layer represented the nonlinear principal components, or in our context, the scores of the underlying factors of systematic risk. This neural network represents a powerful technique capable of performing a nonlinear transformation of the observed variables into the nonlinear principal components, and to execute a nonlinear mapping that reproduces the original variables. We propose a network architecture capable of generating a loading matrix that enables us to make a first approach to the interpretation of the extracted latent risk factors. In addition, we used a two stage methodology for the econometric contrast of the APT involving first, a simultaneous estimation of the system of equations via Seemingly Unrelated Regression (SUR), and secondly, a cross-section estimation via Ordinary Least Squared corrected by heteroskedasticity and autocorrelation by means of the Newey-West heteroskedasticity and autocorrelation consistent covariances estimates (HEC). The evidence found shows that the reproductions of the observed returns using the estimated components via NNPCA are suitable in almost all cases; nevertheless, the results in an econometric contrast lead us to a partial acceptance of the APT in the samples and periods studied.

Low complexity, modulation-transparent and joint polarization and phase tracking scheme based on the nonlinear principal component analysis.

A low complexity, modulation-transparent and joint polarization and phase tracking scheme based on the nonlinear principal component analysis (NPCA) is proposed and demonstrated via both simulation and experiment. Based on high-order statistics, NPCA can achieve joint polarization and phase tracking successfully without any prior information of modulation format. Meanwhile, owing to fact that the estimated matrix is constrained to be a unitary matrix, NPCA can avoid the singularity problem. Compared with the format dependent scheme such as CMA/MMA + VVPE, NPCA shows comparable BER performance under the back-to-back case and shows fast-tracking capability over wide polarization rotation rates ranges. Moreover, for 16QAM signals, the proposed NPCA-based scheme has reduced around 30% computation resources compared with the format dependent scheme, which confirms the advantage of low implementation complexity.

A Hybrid Model for Addressing the Relationship between Financial Performance and Sustainable Development

Measuring financial performance has become an essential topic due to the potential decimating impacts on the corporation itself as well as to whole societies during financial turmoil. In order to provide an overarching description of the multidimensional nature for measuring a corporation’s operations, it is preferable to employ data envelopment analysis (DEA). Different from prior research that merely focuses on a singular DEA performance rank, this study extends it to multiple DEA specifications (i.e., it combines inputs and outputs in several different ways) so as to make judgments more complete and robust. We also execute fuzzy visualization technique (i.e., nonlinear fuzzy robust principal component analysis, NFRPCA) to represent the main characteristics of data so that non-specialists can have better access to the results. The analyzed result is then fed into the restricted Boltzmann machine (RBM) to establish a model to forecast a firm’s operating performance. Even a fraction of accuracy improvement can result in considerable future savings to a firm and investors. When examined using real cases, the model is a promising alternative for operating performance forecasting and can assist both internal and external market participants.

Building anatomically realistic jaw kinematics model from data

This paper considers a different aspect of anatomical face modeling: kinematic modeling of the jaw, i.e., the Temporo-Mandibular Joint (TMJ). Previous work often relies on simple models of jaw kinematics, even though the actual physiological behavior of the TMJ is quite complex, allowing not only for mouth opening, but also for some amount of sideways (lateral) and front-to-back (protrusion) motions. Fortuitously, the TMJ is the only joint whose kinematics can be accurately measured with optical methods, because the bones of the lower and upper jaw are rigidly connected to the lower and upper teeth. We construct a person-specific jaw kinematic model by asking an actor to exercise the entire range of motion of the jaw while keeping the lips open so that the teeth are at least partially visible. This performance is recorded with three calibrated cameras. We obtain highly accurate 3D models of the teeth with a standard dental scanner and use these models to reconstruct the rigid body trajectories of the teeth from the videos (markerless tracking). The relative rigid transformations samples between the lower and upper teeth are mapped to the Lie algebra of rigid body motions in order to linearize the rotational motion. Our main contribution is to fit these samples with a three-dimensional nonlinear model parameterizing the entire range of motion of the TMJ. We show that standard Principal Component Analysis (PCA) fails to capture the nonlinear trajectories of the moving mandible. However, we found these nonlinearities can be captured with a special modification of autoencoder neural networks known as Nonlinear PCA. By mapping back to the Lie group of rigid transformations, we obtain parameterization of the jaw kinematics which provides an intuitive interface allowing the animators to explore realistic jaw motions in a user-friendly way.

Parallel neural networks for improved nonlinear principal component analysis

In this paper, a parallel neural network architecture is proposed to improve the performance of neural-network-based nonlinear principal component analysis. There exist two typical approaches for such analysis: simultaneous extraction of principal components using a single autoassociative neural network (also known as autoencoder), and sequential extraction using multiple neural networks in series. The proposed architecture can be obtained by systematically pruning the network connections of a fully connected autoassociative neural network, resulting in decoupled neural networks. As a result, more independent (i.e., less correlated) principal components can be obtained than the simultaneous extraction approach. The proposed architecture can be also viewed as a rearrangement of multiple neural networks for the sequential extraction in a parallel setting, and thus, the network training becomes more efficient. Simulation case studies are performed to illustrate the advantages of the proposed architecture, and it was shown that it is particularly beneficial for deep neural networks.

Adaptability Planning of Transmission Grid with Integration of High-proportion Renewable Energy

The large-scale integration of renewable energy to power grid is an important feature of future power system development, but renewable energy has strong fluctuation and high uncertainty, which will have a strong impact on power grid. In order to ensure the safe, efficient and reliable operation of power system, improve the acceptability of renewable energy in power grid planning, it is urgent to evaluate the adaptability of power grid structure to the strong fluctuation and uncertainty of renewable energy. Considering that the adaptability of power grid has a broad meaning and is difficult to quantify, this paper analyses the characteristics and actual operation state of the high-penetration renewable energy system and establishes an adaptability index series of power gird structure considering operation safety, efficiency and stability. Based on the adaptability indexes, a multi-objective transmission planning model is put forward. The improved chaotic crossover genetic algorithm and the nonlinear PCA method are used to solve the planning model. Finally, the simulation of Gaver-18 bus system demonstrates the feasibility and effectiveness of the adaptability indexes and planning model.

Decomposition of passive smoking inequality in Iranian children and adolescents: the CASPIAN-V Study

To investigate the socioeconomic inequality in passive smoking in Iranian children and adolescents. Through a multistage random cluster sampling method, a representative sample of 14,400 school students, aged 6-18 years, were enrolled from urban and rural areas of 30 provinces of Iran. Using a non-linear principal component analysis (NLPCA), the correlated variables were summarized as socioeconomic status (SES). Normalized concentration index (NCI) was used to measure inequality in passive smoking at national and regional SES levels. We decomposed total socioeconomic inequality in passive smoking into explanatory variables to identify the main contributors of inequality in passive smoking in the population studied. Data of 12,327 students and parents were complete for the current study. The response rate was 85.6% (50.9% boys, 71.2% urban residents). About 44% of Iranian students were exposed to passive smoking. The NCI for passive smoking at a national level was - 0.043 (95% confidence interval - 0.035, - 0.012), indicating that passive smoking was more concentrated among poorer children and adolescents. This inequality was statistically significant at a national level and in all regions except for regions with lower middle- and lowest SES levels. Considering the decomposition analysis, the household SES (63%), mothers' educational level (37%), fathers' educational level (29%), and school type (18%) made the largest positive contribution to inequality in passive smoking of children and adolescents. Passive smoking was distributed unequally among Iranian children and adolescents; it was more concentrated among socioeconomically disadvantaged families. Public health policies attentions should be given to reduce passive smoking among low SES children and adolescents living with illiterate or low-educated parents.

A Method for Identifying Most Significant Vehicle Parameters for Controller Performance of Autonomous Driving Functions

<div class="section abstract"><div class="htmlview paragraph">In this paper a method for the identification of most significant vehicle parameters influencing the behavior of a lateral control system of autonomous car is presented. Requirements for the design stage of the controller need to consider many uncertainties in the plant. While most vehicle properties can be compensated by an appropriate tuning of the control parameters, other vehicle properties can change significantly during usage. The control system is evaluated based on performance measures. Analyzed parameters comprise functional tire characteristics, mass of the vehicle and position of its center of gravity. Since the parameters are correlated, but Sobol’ sensitivity analysis assumes decorrelated inputs, random variation yields no reasonable results. Furthermore, the variation of each parameter or set of parameters is not applicable since the numbers of required simulations is increased significantly according to input dimension. Therefore, the proposed methodology determines directly the influence of the parameters on performance measures of the lateral control function. First, a hierarchical nonlinear principal component analysis using artificial neural networks is used to decouple inputs and reduce them to a lower dimensional subspace. In a second step Sobol’ sensitivity analysis is used to identify the significant parameters by sampling in the decoupled subspace and their evaluation by a full-blown vehicle dynamics simulation. The most significant parameters are merged in a static and dynamic variable resulting in four worst-case vehicle states. Applying the methodology of the Cooperative Design Approach, performance measures of the lateral control function (rising time, overshoot, number of oscillations) can be analyzed for each vehicle state yielding a robust control design.</div></div>

Farm-level risk factors for bovine mastitis in Dutch automatic milking dairy herds

Automatic milking systems (AMS) are installed on a growing number of dairy farms worldwide. Management to support good udder health might be different on farms with an AMS compared with farms milking with a conventional milking system, as risk factors for mastitis on farms using an AMS may differ. The aim of this study was to identify farm level factors associated with mastitis on Dutch dairy farms using an AMS. In 2008, risk factor data were collected using a questionnaire combined with on-farm recordings of cow, stall, and AMS hygiene on 135 farms. These risk factor data were linked to 4 udder-health-associated dependent variables: average herd somatic cell count (HeSCCav), variance of the average herd somatic cell count (SCC) on test days (HeSCCvar), the average proportion of new high SCC cases (NHiSCC), and the farmer-reported annual incidence rate of clinical mastitis (IRCM). We employed regression models using multiple imputation to deal with missing values. Due to the high dimensionality of the risk factor data, we also performed nonlinear principal component analysis (NLPCA) and regressed the dependent variables on the principal components (PC). Good hygiene of cows and of AMS were found to be related to a lower HeSCCav and less NHiSCC. Effective postmilking teat disinfection was associated with a lower NHiSCC. A higher bulk tank milk SCC threshold for farmers' action was related to more NHiSCC. Larger farm size was related to lower HeSCCvar but higher NHiSCC. Negative attitude of farmers to animal health, higher frequency of checking AMS, and more time spent on viewing computer data were all positively related to higher IRCM. An NLPCA with 3 PC explained 16.3% of the variance in the risk factor variables. Only the first 2 PC were associated with mastitis. The first PC reflected older and larger farms with poor cow hygiene and AMS hygiene, and was related to higher HeSCCav and NHiSCC, whereas the second PC reflected newly built smaller farms with poor cow hygiene and low milk production, and was associated with higher HeSCCvar and NHiSCC, but lower IRCM. Our study suggests that many of the risk factors on conventional milking system farms are applicable to AMS farms, specifically concerning hygiene of the cows and the milking machine, but on large AMS farms, udder health may need more attention than on smaller AMS farms. Multiple imputation is instrumental to deal with missing values and NLPCA is a useful technique to process high dimensional data in our study.

Multi-Scenario Species Distribution Modeling.

Correlative species distribution models (SDMs) are increasingly being used to predict suitable insect habitats. There is also much criticism of prediction discrepancies among different SDMs for the same species and the lack of effective communication about SDM prediction uncertainty. In this paper, we undertook a factorial study to investigate the effects of various modeling components (species-training-datasets, predictor variables, dimension-reduction methods, and model types) on the accuracy of SDM predictions, with the aim of identifying sources of discrepancy and uncertainty. We found that model type was the major factor causing variation in species-distribution predictions among the various modeling components tested. We also found that different combinations of modeling components could significantly increase or decrease the performance of a model. This result indicated the importance of keeping modeling components constant for comparing a given SDM result. With all modeling components, constant, machine-learning models seem to outperform other model types. We also found that, on average, the Hierarchical Non-Linear Principal Components Analysis dimension-reduction method improved model performance more than other methods tested. We also found that the widely used confusion-matrix-based model-performance indices such as the area under the receiving operating characteristic curve (AUC), sensitivity, and Kappa do not necessarily help select the best model from a set of models if variation in performance is not large. To conclude, model result discrepancies do not necessarily suggest lack of robustness in correlative modeling as they can also occur due to inappropriate selection of modeling components. In addition, more research on model performance evaluation is required for developing robust and sensitive model evaluation methods. Undertaking multi-scenario species-distribution modeling, where possible, is likely to mitigate errors arising from inappropriate modeling components selection, and provide end users with better information on the resulting model prediction uncertainty.

A Damage Classification Approach for Structural Health Monitoring Using Machine Learning

Inspection strategies with guided wave-based approaches give to structural health monitoring (SHM) applications several advantages, among them, the possibility of the use of real data from the structure which enables continuous monitoring and online damage identification. These kinds of inspection strategies are based on the fact that these waves can propagate over relatively long distances and are able to interact sensitively with and uniquely with different types of defects. The principal goal for SHM is oriented to the development of efficient methodologies to process these data and provide results associated with the different levels of the damage identification process. As a contribution, this work presents a damage detection and classification methodology which includes the use of data collected from a structure under different structural states by means of a piezoelectric sensor network taking advantage of the use of guided waves, hierarchical nonlinear principal component analysis (h-NLPCA), and machine learning. The methodology is evaluated and tested in two structures: (i) a carbon fibre reinforced polymer (CFRP) sandwich structure with some damages on the multilayered composite sandwich structure and (ii) a CFRP composite plate. Damages in the structures were intentionally produced to simulate different damage mechanisms, that is, delamination and cracking of the skin.

Profiling of Oncogenic Signaling in Multiple Myeloma — Association with Biology, Disease Progression and Prognosis

Introduction: The pathogenesis of multiple myeloma (MM) involves complex genetic and epigenetic alterations, which ultimately affect oncogenic signaling networks. Analysis of pathway activation in primary MM cells within their native microenvironment will not only increase our understanding of MM cell biology but may open up new avenues towards tailored therapies that target individually activated signaling cascades.Methods: We established a (phospho-) immunohistochemistry (IHC)-assay to analyze key activation markers of oncogenic pathways and networks, namely RAS/RAF/MEK/ERK, PI3K/AKT, JAK/STAT3, canonical NFκB, and cMyc, in an exploratory cohort of 459 bone marrow (BM) biopsies from patients with smoldering MM (SMM), newly diagnosed symptomatic (NDMM) and refractory MM (RRMM). An independent validation cohort comprised BM samples from 462 NDMM patients who were uniformly treated within clinical trials. To integrate activation intensity and spread scores for each pathway and sample, non-linear principal component analysis (PCA) was applied.Results: Pathway profiling in progressing stages of plasma cell disorders showed the canonical NFκB pathway as nearly universally activated in SMM (97% of patients); signals for AKT, ERK, STAT3, and cMyc were only seen in subsets of SMM patients (≤33%), with STAT3 being the least activated pathway (6%). In NDMM and RRMM, NFκB activation significantly decreased to 90% and 82% of patients, respectively (p=0.02). In contrast, we observed a significant increase in activation of all other pathways and cMyc from SMM to NDMM and further to RRMM patients (p<0.001). All of these pathways were activated in at least 50% of RRMM. Analysis of 34 patients with available longitudinal samples confirmed a significant increase in strong activation of STAT3 and cMyc with progressing disease stages.Assessment of pathway clustering and interdependencies showed a strong negative correlation between NFκB and ERK in all disease stages. In contrast, a positive correlation was seen between cMyc and STAT3 as well as between AKT and ERK activation with increasing strength over progressing disease stages. Interestingly, when focusing on strong activation signals as determined by PCA, virtual absence of simultaneous activation of more than one pathway in SMM was observed, while pathway co-activation significantly increased in NDMM and RRMM (22% and 34%, p<0.001).Activation of STAT3 and expression of cMyc were significantly associated with higher ISS stage in NDMM of both the exploratory and the validation cohort. STAT3 pathway activation was also significantly associated with gain(1q21). In the validation cohort with available gene expression profiling (GEP) data, both STAT3 (p=0.02) and cMyc (p<0.001) activation showed a significant association with high risk disease according to GEP risk scores.Assessment of the impact of pathway activation on survival in the validation cohort showed expression of cMyc (p=0.01), activation of STAT3 (p=0.005) and cMyc-STAT3 co-activation (p=0.005) to negatively influence PFS. Expression of cMyc (p=0.02) and cMyc-STAT3 co-activation (p=0.02) were also significantly associated with inferior OS. STAT3 activation showed a negative trend towards inferior OS (p=0.08). Accounting for GEP-defined risk status, STAT3 pathway activation defined an ultra-high risk cohort. Assessment of differential gene expression suggested an autocrine positive feedback loop of IL-6 signaling and an enrichment of cell cycle associated genes in patients with STAT3 activation.Conclusions: Activation of either NFkB or ERK/AKT signaling appears to be a hallmark of MM. Later stages of MM are characterized by activation of additional signaling pathways that may affect therapy response. Importantly, cMyc expression and JAK/STAT3 pathway activation show adverse prognostic impact and significantly increase with disease progression. DisclosuresGoldschmidt:Sanofi: Consultancy, Research Funding; Adaptive Biotechnology: Consultancy; Amgen: Consultancy, Research Funding; Janssen: Consultancy, Honoraria, Research Funding; Celgene: Consultancy, Honoraria, Research Funding; Bristol Myers Squibb: Consultancy, Honoraria, Research Funding; ArtTempi: Honoraria; Novartis: Honoraria, Research Funding; Mundipharma: Research Funding; Chugai: Honoraria, Research Funding; Takeda: Consultancy, Research Funding. Raab:BMS: Consultancy, Honoraria, Research Funding; Amgen: Consultancy, Honoraria, Research Funding; Novartis: Consultancy, Honoraria, Research Funding; Celgene: Consultancy, Honoraria.

Gas path fault diagnostics using a hybrid intelligent method for industrial gas turbine engines

There are many challenges against an accurate gas turbine fault diagnostics, such as the nonlinearity of the engine health, the measurement uncertainty, and the occurrence of simultaneous faults. The conventional methods have limitations in effectively handling these challenges. In this paper, a hybrid intelligent technique is devised by integrating an autoassociative neural network (AANN), nested machine learning (ML) classifiers, and a multilayer perceptron (MLP). The AANN module is used as a data preprocessor to reduce measurement noise and extract the important features for visualisation and fault diagnostics. The features are extracted from the bottleneck layer output values based on the concept of the nonlinear principal component analysis (NLPCA). The nested classifier modules are then used in such a manner that fault and no-fault conditions, component and sensor faults, and different component faults are distinguished hierarchically. As part of the classification, evaluation of the fault classification performance of five widely used ML techniques aiming to identify alternative approaches is undertaken. In the end, the MLP approximator is utilised to estimate the magnitude of the isolated component faults in terms of flow capacity and isentropic efficiency indices. The developed system was implemented to diagnose up to three simultaneous faults in a two-shaft industrial gas turbine engine. Its robustness towards the measurement uncertainty was also evaluated based on Gaussian noise corrupted data. The test results show the derivable benefits of integrating two or more methods in engine diagnostics on the basis of offsetting the weakness of the one with the strength of another.

Analiza nieliniowych składowych głównych dla danych czasowo‑przestrzennych geograficznie ważonych

Schölkopf, Smola i Müller (1998) zaproponowali analizę nieliniowych składowych głównych (NPCA) dla ustalonych danych wektorowych. Niniejszy artykuł zawiera rozszerzenie tej metody na dane czasowo‑przestrzenne oraz czasowo‑przestrzenne geograficznie ważone. Każdy obiekt jest scharakteryzowany za pomocą macierzy Xi, rozmiaru T × p, zawierającej wartości p cech zaobserwowanych w T momentach czasowych, i = 1, …, n. Macierze te są przekształcane nieliniowo do przestrzeni Hilberta i budowana jest scentrowana macierz jądrowa. Ostatecznie macierz ta jest podstawą konstrukcji nieliniowych składowych głównych. W przypadku danych geograficznie ważonych macierz Xizostaje zastąpiona macierzą wiXi, gdzie wijest dodatnią wagą geograficzną związaną z i‑tym miejscem obserwacji, i = 1, …, n. Teoria zilustrowana jest przykładem dotyczącym stanu szkolnictwa wyższego w 16 polskich województwach, notowanego w latach 2002–2016.

Deep principal component analysis: An enhanced approach for structural damage identification

The structural health monitoring relies on the continuous observation of a dynamic system over time to identify its actual condition, detect abnormal behaviors, and predict future states. The regular changes in environmental factors have been reported as one of the main challenges for the application of structural health monitoring systems. These influences in the structural responses are in general nonlinear, affecting the damage-sensitive features in the most varied forms. The usual process to remove these normal changes is referred to as data normalization. In that regard, principal component analysis is probably the most studied algorithm in structural health monitoring, having numerous versions to learn strong nonlinear normal changes. However, in most cases, not all variability is properly accounted for via the existing nonlinear principal component analysis approaches, resulting in poor damage detection and quantification performances. In this article, a new paradigm based on deep principal component analysis, rooted in the deep learning field, is presented to overcome these limitations. This approach extracts the most salient underlying feature distributions by stacking multiple feedforward neural networks trained to learn an identity mapping of the input variables, where the network inputs are reproduced into the outputs. Similar to the traditional nonlinear principal component analysis–based approach, our approach identifies a nonlinear output-only model of an undamaged structure by comprising modal features into an internal bottleneck layer, which implicitly represents the independent environmental factors. The proposed technique is validated through the application on a progressively damaged prestressed concrete bridge and a three-span suspension bridge. The experimental results demonstrate that capturing the most slight nonlinear variations in the data can lead to improved data normalization and, consequently, better damage detection and quantification performances.

Online fault detection and isolation of an AIR quality monitoring network based on machine learning and metaheuristic methods

The dynamic process monitoring is discussed in this paper. Kernel principal component analysis (KPCA) is a nonlinear monitoring method that cannot be applied for dynamic systems. Reduced online KPCA (OR-KPCA)is used for fault detection of dynamic processes, which is developed to built a dictionary according to the process status and then, it update the KPCA model and uses it for process monitoring. Also the Tabu search metaheuristic algorithm is used in order to determine the optimal parameter of the kernel function. In this paper, new approaches for online fault isolation, which is a challenging problem in nonlinear PCA, are formulated. An extension of partial PCA and the elimination sensor identification (ESI) to the case of nonlinear systems are presented in a feature space. The partial OR-KPCA and the elimination sensor identification (ESI-KPCA) are generated based on the OR-KPCA method and they consist of developing a set of sub-models. The sub-models are selected according to a pre-designed fault-to-residual structure matrix and by eliminating sequentially one variable from the set of the variables. The proposed fault isolation methods are applied for monitoring an air quality monitoring network. The simulation results show that the proposed fault isolation methods are effective for KPCA.

Ensemble Learning Based Multiple Kernel Principal Component Analysis for Dimensionality Reduction and Classification of Hyperspectral Imagery

Classification is one of the most challenging tasks of remotely sensed data processing, particularly for hyperspectral imaging (HSI). Dimension reduction is widely applied as a preprocessing step for classification; however the reduction of dimension using conventional methods may not always guarantee high classification rate. Principal component analysis (PCA) and its nonlinear version kernel PCA (KPCA) are known as traditional dimension reduction algorithms. In a previous work, a variant of KPCA, denoted as Adaptive KPCA (A-KPCA), is suggested to get robust unsupervised feature representation for HSI. The specified technique employs several KPCAs simultaneously to obtain better feature points from each applied KPCA which includes different candidate kernels. Nevertheless, A-KPCA neglects the influence of subkernels employing an unweighted combination. Furthermore, if there is at least one weak kernel in the set of kernels, the classification performance may be reduced significantly. To address these problems, in this paper we propose an Ensemble Learning (EL) based multiple kernel PCA (M-KPCA) strategy. M-KPCA constructs a weighted combination of kernels with high discriminative ability from a predetermined set of base kernels and then extracts features in an unsupervised fashion. The experiments on two different AVIRIS hyperspectral data sets show that the proposed algorithm can achieve a satisfactory feature extraction performance on real data.

Linear and nonlinear features and machine learning for wind turbine blade ice detection and diagnosis

Abstract The mass of ice on wind turbines blades is one of the main problems that energy companies have in cold climates. This paper presents a novel approach to detect and classify ice thickness based on pattern recognition through guided ultrasonic waves and Machine Learning. To successfully achieve a supervised classification, it is necessary to employ a method that allows the correct extraction and selection of features of the ultrasonic signal. The main novelty in this work is that the approach considers four feature extraction methods to validate the results, grouped by linear (AutoRegressive (AR) and Principal Component Analysis) and nonlinear (nonlinear-AR eXogenous and Hierarchical Non-Linear Principal Component Analysis), and feature selection is done by Neighbourhood Component Analysis. A supervised classification was performed through Machine Learning with twenty classifiers such as Decision tree, Discriminant Analysis, Support Vector Machines, K-Nearest Neighbours, and Ensemble Classifiers. Finally, an evaluation of the classifiers was done in single frequency and multi-frequency modes, obtaining accurate results.