Search

Multivariate statistical process monitoring methods are often used in chemical process fault diagnosis. In this article, (I) the cycle temporal algorithm (CTA) combined with the dynamic kernel principal component analysis (DKPCA) and the multiway dynamic kernel principal component analysis (MDKPCA) fault detection algorithms are proposed, which are used for continuous and batch process fault detections, respectively. In addition, (II) a fault variable identification model based on reconstructed-based contribution (RBC) model that paves the way for determining the cause of the fault are proposed. The proposed fault diagnosis model was applied to Tennessee Eastman (TE) process and penicillin fermentation process for fault diagnosis. And compare with other fault diagnosis methods. The results show that the proposed method has better detection effects than other methods. Finally, the reconstruction-based contribution (RBC) model method is used to accurately locate the root cause of the fault and determine the fault path.

The kernel principal component analysis (KPCA) is widely used as a fault monitoring tool for complex nonlinear chemical processes in recent years. The cumulative contribution rate that extracts the kernel principal is usually obtained relied on a fixed model, which cannot be employed for time-varying chemical processes. Hence, a novel adaptive kernel principal component analysis (AKPCA) integrating grey relational analysis (GRA) (AKPCA-GRA) is proposed to dynamically monitor the fault occurrence. A moving window integrating the threshold method is used to adaptively extract the kernel principal for chemical processes. Then the corresponding T2 and Q statistics calculated by the selected kernel principal based on the AKPCA decides whether the fault has occurred. Moreover, the GRA method is used to analyze and calculate the correlation coefficient of abnormal features obtained based on the APKCA method, which provides the operational guidance for the nonlinear chemical process to find out the variable causing the fault. Finally, the proposed method is verified using the Tennessee Eastman (TE) process. The case results demonstrate that the proposed method outperforms the KPCA, the KPCA based on the threshold and the moving window principal component analysis, the support vector machine (SVM) and the Logistic Regression (LR) in terms of the missed alarm rate (MAR) and the false alarm rate (FAR), which can effectively analyze the variables causing the fault.

In this paper, we propose an efficient algorithm for denoising the degraded face image sequence in the principal component analysis (PCA) domain for recognizing this face. We first apply a temporal filter that performs motion compensation combined with a weighted average filter, then an adaptative spatial filter realized by PCA transformation that decomposes the image temporally filtered into two sub-images using a threshold calculated according to the variance of noise and intensity retained by the first eigenvectors. The first sub-image containing a great intensity variance expresses the small features, and the other with high noise level and low intensity variance describes the large feature. Therefore, the spatial filter can be adapted according to the amount of information in each area, so that small features are reprojected in the Kernel PCA domain where image details are reconstructed efficiently, and large features are denoised by an anisotropic diffusion filter to recover the homogenous regions. Finally, the restored image is used in recognition process. The experimental results obtained from the Cohen-Kanade facial expression (CKFE) database tested against different noise levels and three types of blur (Gaussian, motion and pillbox) show better restoration and recognition performance of this algorithm compared to other methods.

In building automation system, timely detecting the operational faults in water chillers is crucial to system operation management. As a linear data transformation technique, the performance of principal component analysis (PCA) based chiller fault detection method is limited, especially for the incipient ones, due to the nonlinearities in chillers. In addition, the control limit for the monitoring statistic, such as squared prediction error (SPE), is determined under the Gaussian assumption of the score variables, which can hardly be satisfied in water chillers. Therefore, an enhanced fault detection method with the application of kernel density estimation (KDE) and kernel entropy component analysis (KECA) algorithm is reported in this paper. Cauchy-Schwarz (CS) divergence was evaluated as monitoring statistic to measure the cosine of the angle between two data sets after being projected onto a dominated subspace, and then adopted as an index for dissimilarity. KDE with its bandwidth being optimized was also applied to estimate the distribution of the CS divergence, so that the control limit for fault monitoring could be determined. The proposed KDE based KECA-CS method was validated using the experimental data from ASHRAE RP-1043, and further compared to the PCA -SPE, kernel principal component analysis (KPCA)-SPE, and KECA-SPE methods. Results showed that the best performance could be realized when using the KDE based KECA-CS method. The reported fault detection ratio was over 68% for the seven typical chiller faults even at their corresponding least severity level. The average fault detection accuracy was over 90%.

In this work, we propose a progressive principal component analysis (PPCA) method for compressing deep convolutional neural networks. The proposed method starts with a prespecified layer and progressively moves on to the final output layer. For each target layer, PPCA conducts kernel principal component analysis for the estimated kernel weights. This leads to a significant reduction in the number of kernels in the current layer. As a consequence, the channels used for the next layer are also reduced substantially. This is because the number of kernels used in the current layer determines the number of channels for the next layer. For convenience, we refer to this as a progressive effect. As a consequence, the entire model structure can be substantially compressed, and both the number of parameters and the inference costs can be substantially reduced. Meanwhile, the prediction accuracy remains very competitive with respect to that of the baseline model. The effectiveness of the proposed method is evaluated on a number of classical CNNs (AlexNet, VGGNet, ResNet and MobileNet) and benchmark datasets. The empirical findings are very encouraging. The code is available at https://github.com/zhoujing89/ppca .

This paper presents a data-driven methodology for fault detection and diagnosis (FDD) by integrating the principal component analysis (PCA) with the Bayesian network (BN). Though the integration of PCA-BN for FDD purposes has been studied in the past, the present work makes two contributions for process systems. First, the application of correlation dimension (CD) to select principal components (PCs) automatically. Second, the use of Kullback-Leibler divergence (KLD) and copula theory to develop a data-based BN learning technique. The proposed method uses a combination of vine copula and Bayes’ theorem (BT) to capture nonlinear dependence of high-dimensional process data which eliminates the need for discretization of continuous data. The data-driven integrated PCA-BN framework has been applied to two processing systems. Performance of the proposed methodology is compared with the independent component analysis (ICA), kernel principal component analysis (KPCA), kernel independent component analysis (KICA), and their integrated frameworks with the BN. The comparative study suggests that the proposed framework provides superior performance.

Quality-related fault detection is the key subject under process monitoring and fault diagnosis (PM-FD) framework, which is an effective mean to guarantee safety production and obtain reliable product quality, and thus, has recently become active areas of process control fields. In this paper, a novel decentralized detection method for quality-related faults is proposed for manufacturing industrial processes, which will help field engineers to purposefully and real-time observe the operation state of the production process. Specifically, the main innovations are: 1) the linear and nonlinear dynamic interdependencies between process and quality variables are revealed and quantified by the minimal redundancy maximal relevance (mRMR) algorithm, and the most representative variables are selected in each subprocess; 2) a new mixed kernel function based dynamic mixed kernel principal component analysis (DMKPCA) model is designed for enhancing the generalization ability and maintaining the data characteristics of original sample space; 3) unified monitoring statistics are constructed in the subprocess level, and Bayesian fusion is implemented for forming monitoring decisions from the plant-wide level. Finally, a case study on an actual hot rolling mill process (HSMP) is finally given to validate the effectiveness of the proposed scheme, and several competitive methods are applied to carry out the quality-related fault detection process.

Sludge bulking is very common in wastewater treatment process (WWTP), which will degrade the operation performance or even destroy the process. In order to diagnose sludge bulking accurately, a data-knowledge-driven diagnosis (DKD) method is proposed to identify the occurrence and cause variable in this paper. This proposed DKD method contains the following advantages. First, a data-driven detection model, using a recursive kernel principal component analysis (RKPCA) algorithm, is designed to capture the intrinsic nonlinear and time-varying characteristic of sludge bulking. Then, the occurrence of sludge bulking can be detected with high accuracy. Second, a DKD model, based on the Bayesian network (BN), is developed to extract the causality among process variables to identify the root cause variables of sludge bulking. Then, the root cause variables of sludge bulking can be diagnosed to improve the operation performance of WWTP. Finally, the proposed DKD method was tested on the measured data from a real WWTP. Experimental results confirmed the effectiveness of the proposed DKD method.

Electroencephalogram (EEG) is a non-invasive measurement of electrical signal on the scalp originated due to neuronal activity. EEG signals associated with cortical activity are orders of magnitude lower in amplitude compared to other parasitic signals such as eye-blink artifacts, which contaminate the recorded EEG data making it imperative for the investigators to adopt an effective artifact suppression strategy. In all the previous studies, suppression of the pattern related to the artifacts was performed based on the assumption of a linear interaction between the source of artifact (eye-blink) and the brain signal (EEG data). This paper presents a novel methodology by considering the non-linear interaction between the artifact signal associated with the eye-blink and the contaminated EEG data using kernel functions. In the present work, adaptive data-driven approach called Ensemble Empirical Mode Decomposition (EEMD) is hybridized with kernel Principal Component Analysis (kPCA) to decode the non-linear interaction. The contaminated segment of EEG data is decomposed by the EEMD technique into a series of basic building blocks called intrinsic mode functions (IMFs). The features of the eye-blink signals are captured by some of these IMFs; subsequently, effective extraction of the ocular artifact from the IMFs is performed by kPCA using non-linear kernel functions (radial basis function, second and third-order polynomial function). In the present study, technique used for artifact suppression relies on the extraction of ocular artifact signal from IMFs based on optimizing the different parameters of the kPCA. Compared with other techniques used in previous studies, the proposed method (based on third-order polynomial kernel function) is capable of effectively extracting the pattern related to the ocular artifacts from the single channel contaminated EEG data with low distortion of the EEG signal.

In order to detect the incipient faults of nonlinear industrial processes effectively, this paper proposes an enhanced kernel principal component analysis (KPCA) method, called multi-block probability related KPCA method (DMPRKPCA). First of all, one probability related nonlinear statistical monitoring framework is constructed by combining KPCA with Kullback Leibler divergence (KLD), which measures the probability distribution changes caused by small shifts. Second, in view of the problem that the traditional KLD ignores the dynamic characteristic of process data, the dynamic KLD component is designed by applying the exponentially weighted moving average approach, which highlights the temporal data changes in the moving window. Third, considering that the holistic KLD component may submerge the local statistical changes, a multi-block modeling strategy is designed by dividing the whole KLD components into two sub-blocks corresponding to the mean and variance information, respectively. Case studies on one numerical system and the simulated chemical reactor demonstrate the superiority of the DMPRKPCA method over the conventional KPCA method.

A method for speeding-up data projections onto kernel-induced feature spaces (derived using e.g. kernel Principal Component Analysis - kPCA) is presented in the paper. The proposed idea is to simplify the derived features, implicitly defined by all training samples and dominant eigenvectors of problem-specific generalized eigenproblems, by appropriate approximations. Instead of employing the whole training set, we propose to use a small pool of its appropriately selected representatives and we formulate a rule for deriving the corresponding weight vectors that replace the considered dominant eigenvectors. The representatives are determined via clustering of training data, whereas weighting coefficients are chosen to minimize original feature approximation errors. The concept has been experimentally verified for kernel-PCA using both artificial and real datasets. It has been shown that the presented approach provides reduction in feature-extraction complexity, which implies a proportional increase in data projection speed, by one-to-two orders of magnitude, without sacrificing data analysis accuracy. Therefore, the proposed approach is well-suited for kernel-based, intelligent data analysis applications that are to be executed on resource-limited systems, such as embedded or IoT devices, or for systems where processing time is critical.

A multivariate statistical process monitoring scheme should be able to describe multimodal data. Multimodality typically arises in process data due to varying production regimes. Moreover, multimodality may influence how easy it is for process operators to interpret the monitoring results. To address these challenges, this paper proposes an on-line monitoring framework for anomaly detection where an anomaly may either indicate a fault occurring and developing in the process or the process moving to a new operating mode. The framework incorporates the Dirichlet process, which is an unsupervised clustering method, and kernel principal component analysis with a new kernel specialized for multimode data. A monitoring model is trained using the data obtained from several healthy operating modes. When on-line, if a new healthy operating mode is confirmed by an operator, the monitoring model is updated using data collected in the new mode. Implementation issues of this framework, including the parameter tuning for the kernel and the selection of anomaly indicators, are also discussed. A bivariate numerical simulation is used to demonstrate the performance of anomaly detection of the monitoring model. The ability of this framework in model updating and anomaly detection in new operating modes is shown on data from an industrial-scale process using the PRONTO benchmark dataset. The examples will also demonstrate the industrial applicability of the proposed framework.

Based on the dimension reduction of hyperspectral remote sensing image, a new neural network method is used to classify the hyperspectral remote sensing image of carbon dioxide in detail. Firstly, the Kernel Principal Component Analysis (KPCA) and Genetic Algorithms (GA) are used to reduce the dimension of hyperspectral remote sensing images; secondly, the traditional remote sensing image classification methods (ISODATA, SVM), traditional neural networks (BP), and new neural networks are used to classify the hyperspectral remote sensing images. Finally, noise assessment method based on local mean and local standard deviation (LMLSD) of spectral image is used to evaluate the classification accuracy. In addition, hyperspectral remote sensing images are dimensionality reduction. Secondly, the comparison between the traditional remote sensing image classification method and the new neural network method is analyzed. Finally, a new neural network method is applied to classify hyperspectral remote sensing images.

Recently, human healthcare from body sensor data has been getting remarkable research attentions by a huge range of human-computer interaction and pattern analysis researchers due to its practical applications such as smart health care systems. For example, smart wearable-based behavior recognition system can be used to assist the rehabilitation of patients in a smart clinic to improve the rehabilitation process and to prolong their independent life. Although there are many ways of using distributed sensors to monitor vital signs and behavior of people, physical human action recognition via body sensors provides valuable data regarding an individual's functionality and lifestyle. In this work, we propose a body sensor-based system for behavior recognition using deep Recurrent Neural Network (RNN), a promising deep learning algorithm based on sequential information. We perform data fusion from multiple body sensors such as electrocardiography (ECG), accelerometer, magnetometer, etc. The extracted features are further enhanced via kernel principal component analysis (KPCA). The robust features are then used to train an activity RNN, which is later used for behavior recognition. The system has been compared against the conventional approaches on three publicly available standard datasets. The experimental results show that the proposed approach outperforms the available state-of-the-art methods.

Human actions detection is very much investigated in utilization of artificial intelligence and computer vision. Numerous effective action recognition strategies have demonstrated and the action information are successfully gained from motion videos and still pictures. In order to get the equivalent actions, the proper activity information gained from various kinds of media like videos or pictures might be connected. The majority of the existing video activity action identification strategies experience the ill effects of inadequate marked recordings. In that cases, over-fitting should be a potential issue and the execution of activity acknowledgment is controlled. In this paper, image processing techniques are used in order to recognize the different hand poster of the human body, also the over-fitting can be eased and the execution of activity acknowledgment is improved. Initially, the human action video including hand waving, walking, jogging, clapping, boxing is converted into image of 2D frames and then it is preprocessed followed by feature extraction using LST and classification by KNN classifier has been done individually. The kernel principal component analysis (KPCA) technique is used in the proposed system for finding the image features and joined features.The extracted features from the frames are compared with trained quantized dataset in order to identify the actions. The advantage of quantized dataset is that it occupies very less space. Thus, the result shows which action is present in the examined data. Trials on open benchmark data sets and genuine world data sets demonstrate that our technique outflanks a few other cutting edge activity acknowledgment strategies.

Principal component analysis (PCA) and its modified methods have been widely applied in industrial process monitoring. In practice, industrial processes are with disparate characteristics, the process monitoring system should consider as many process characteristics as possible, such as dynamic and nonlinear characteristics. In this paper, a multi-feature extraction technique based on PCA is proposed for nonlinear dynamic process monitoring. The proposed method integrates dynamic inner PCA (DiPCA), PCA and kernel PCA (KPCA) methods through a serial structure to extract the dynamic, linear and nonlinear features among the process data. Along with the proposed method, the original data space is decomposed into several orthogonal subspaces, in which abnormal variations of different features can be monitored. For real-time process monitoring, a combined Hotelling’s T2 statistic based on the extracted multi-feature and a squared prediction error (SPE or Q) statistic are established. Case studies on a numerical example and the Tennessee Eastman process are carried out to demonstrate the superior process monitoring performance of the proposed method compared with other relevant methods.

In this paper, we consider clustering based on the kernel principal component analysis (KPCA) for high-dimension, low-sample-size (HDLSS) data. We give theoretical reasons why the Gaussian kernel is effective for clustering high-dimensional data. In addition, we discuss a choice of the scale parameter yielding a high performance of the KPCA with the Gaussian kernel. Finally, we test the performance of the clustering by using microarray data sets.

Currently, it is a great challenge to effectively acquire more widespread equipment health information for guaranteeing safe production and timely fault maintenance in the process of industrial informationization. Aimed at the present issue that the conventional machine learning algorithm cannot juggle local and global fault feature information of rotating machinery, a novel fault diagnosis method based on Local-Global Deep Neural Network (LGDNN) algorithm is proposed in this paper. First of all, this fault diagnosis method can directly use the proposed LGDNN algorithm to extract local and global structural features from the original vibration spectral signals. Subsequently, the extracted high-level features are applied to classify different fault conditions by using soft-max classifier. Crucially, the core of this fault diagnosis method is that the anterior local layer of LGDNN utilizes a novel local feature extractor of the improved Convolutional Deep Belief Network (CDBN) based on the Fisher parameter optimization criterion (called Fisher-CDBN) to efficiently extract local discriminant information of data. Afterwards, its secondary global layer uses the global feature extractor of Kernel Principal Component Analysis (KPCA) to reduce the redundancy attribute of data. Ultimately, the vibration signals of the rolling bearing and the fan's gear are used to validate the effectiveness of the proposed method. And the experimental results demonstrate that the proposed algorithm and fault diagnosis method can juggle local and global fault feature information of rotating machinery. Besides, it also provides a novel research reference for deep learning fault diagnosis of rotating machinery.

Kernel principal component analysis (KPCA) forms the basis for a class of methods commonly used for denoising a set of multivariate observations. Most KPCA algorithms involve two steps: projection and preimage approximation. We argue that this two-step procedure can be inefficient and result in poor denoising. We propose an alternative projection-free KPCA denoising approach that does not involve the usual projection and subsequent preimage approximation steps. In order to denoise an observation, our approach performs a single line search along the gradient descent direction of the squared projection error. The rationale is that this moves an observation towards the underlying manifold that represents the noiseless data in the most direct manner possible. We demonstrate that the approach is simple, computationally efficient, robust, and sometimes provides substantially better denoising than the standard KPCA algorithm.

This paper introduces a new structure kernel principal component analysis (KPCA) that can successfully model symbolic interval-valued data for fault detection. In the proposed structure, interval KPCA (IKPCA) method is proposed to deal with interval-valued data. Two IKPCA models are proposed. The first model is based on the centers and ranges of intervals IKPCA CR and the second model is based the upper and lower bounds of intervals IKPCA UL . Residuals are generated and fault detection indices are computed. The aim of using IKPCA is to ensure robustness to false alarm without affecting the fault detection performance. The proposed fault detection approach is carried out using simulation example and Tennessee Eastman Process (TEP). The obtained results demonstrate the effectiveness of the proposed technique.

Gearbox is a critical transmission component in the drivetrain of wind turbine having a dominant failure rate and a highest downtime loss among all wind turbine subsystems. Ensemble empirical mode decomposition and principal component analysis have been extensively investigated for signal decomposition and fault feature extraction from the signals of a wind turbine gearbox. However, presence of background noise in wind turbine signals restricts the applicability of these algorithms in real scenarios. To solve this problem, a novel performance degradation assessment method based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and kernel principal component analysis (KPCA) was proposed to de-noise and fuse vibration signals. A comparison is conducted between CEEMDAN-KPCA and other five cross combinations. After feature extraction, extreme learning machine (ELM) optimized by fruit fly of algorithm (FOA) is employed to predict the remaining use life (RUL) of wind turbine gearbox. To avoid the tedious hand-crafting hidden nodes, a dynamic adjustment strategy is presented for the improvement of search efficiency. The effectiveness of the proposed method is validated using simulated and experimental vibration signals. The results illustrate that the CEEMDAN-KPCA gets better result in signal de-noising. Compared with different swarm intelligence algorithms, the RUL prediction rates of wind turbine gearbox are improved by using the FOA-ELM prediction model with multiple parameters.

In order to realize real-time and accurate monitoring of the tool wear in machining process, this paper presents a tool wear predictive model based on the integrated radial basis function based kernel principal component analysis (KPCA_IRBF) and relevance vector machine (RVM). The traditional methods such as partial least squares regression (PLS), artificial neural network (ANN) and support vector machine (SVM) can only provide predicted values which have no probabilistic significance. As a sparse probabilistic model, RVM can provide both the predicted value and the corresponding confidence interval (CI). However, the existence of process noises and redundancy will seriously affect the prediction accuracy and the stability of CI. As a new dimension-increment technique, KPCA_IRBF helps to weaken the negative effects of process noises and redundancy by increasing the dimensionality of monitoring features. The fused features obtained by KPCA_IRBF are more sensitive to the change of tool wear. Two different cutting experiments are carried out to verify the effectiveness of KPCA_IRBF in improving the prediction accuracy and ameliorating the CI of RVM. The experimental results show that KPCA_IRBF can reduce the root mean square error (RMSE) of RVM by more than 30% and compress the average width of CI by more than 90%. To further show the advantages of RVM, the traditional methods such as PLS, ANN and SVM are also utilized to realize tool wear prediction. This paper lays the foundation for the application of RVM to industrial field.

Raman spectroscopy has been widely used in discriminant analysis. In order to improve the accuracy of Raman spectroscopy discrimination, a model combining kernel principal component analysis (KPCA) and support vector machine (SVM) is proposed. Firstly, the Raman spectral discriminant data is collected, which is subjected to the fifth-order polynomial smoothing and vector normalization preprocessing to eliminate the influence of noise. Then, the collected unbalanced data is oversampled by the Synthetic Minority Over-sampling Technique algorithm, and the KPCA method is used to extract the features of the balanced data. The SVM discriminant model is established by selecting different kernel functions for the extracted features. In order to evaluate the performance of the KPCA-SVM discriminant model, it is compared with the PCA-SVM discriminant model under the same experimental conditions. The experimental results show that the KPCA-SVM discriminant model achieves a discriminative accuracy rate of 93.75%, which is better than that of the PCA-SVM discriminant model (87.5%). This study provides a new idea for improving the discrimination accuracy of Raman spectroscopy.

Accurate and efficient rotating machinery fault diagnosis is crucial for industries to guarantee the productivity and reduce the maintenance cost. This paper systematically proposes a new fault diagnosis approach including signal processing techniques and pattern recognition method. In order to reveal more useful details in a fault residing signal, a novel automatic signal segmentation method named Grassmann manifold – angular central Gaussian distribution is proposed to divide a raw signal into several segments, resulting in a significant improvement of diagnosis accuracy. An improved empirical mode decomposition, wavelet transform – ensemble empirical mode decomposition, is also designed which could adequately solve the problems of mode mixing and end effects. Moreover, a morphological method usually used in image processing is investigated and adopted to change the shape of the intrinsic mode functions to further reveal the faulty impulses. In order to reduce the high dimension of the extracted features and improve the computational efficiency and accuracy, a deep belief network is designed to conduct information fusion, and compared with widely adopted kernel principal component analysis. For classification, a pairwise coupling strategy is proposed and combined with sparse Bayesian extreme learning machine. The experiments conducted using the proposed approach demonstrate the effectiveness of the proposed system.

Motivated by the need of processing non-point-evaluation functional data, we introduce the notion of functional reproducing kernel Hilbert spaces (FRKHSs). This space admits a unique functional reproducing kernel which reproduces a family of continuous linear functionals on the space. The theory of FRKHSs and the associated functional reproducing kernels are established. A special class of FRKHSs, which we call the perfect FRKHSs, are studied, which reproduce the family of the standard point-evaluation functionals and at the same time another different family of continuous linear (non-point-evaluation) functionals. The perfect FRKHSs are characterized in terms of features, especially for those with respect to integral functionals. In particular, several specific examples of the perfect FRKHSs are presented. We apply the theory of FRKHSs to sampling and regularized learning, where non-point-evaluation functional data are used. Specifically, a general complete reconstruction formula from linear functional values is established in the framework of FRKHSs. The average sampling and the reconstruction of vector-valued functions are considered in specific FRKHSs. We also investigate in the FRKHS setting the regularized learning schemes, which learn a target element from non-point-evaluation functional data. The desired representer theorems of the learning problems are established to demonstrate the key roles played by the FRKHSs and the functional reproducing kernels in machine learning from non-point-evaluation functional data. We finally illustrate that the continuity of linear functionals, used to obtain the non-point-evaluation functional data, on an FRKHS is necessary for the stability of the numerical reconstruction algorithm using the data.

Terahertz (THz) spectroscopy and integrated learning method were utilized to identify ten types of soybean seeds in this work. Before soybean seeds were identified, seed sample of nine thicknesses were tested to verify THz spectra of soybeans prepared by grinding and polishing is feasible, and a statistical method of single factor variance was calculated to prove THz spectra have significant influence on soybean varieties. Then, the integrated learning method (DT_A) based on adaboost algorithm and decision tree (DT) was studied to identify soybean varieties by five pretreatment methods, thirty basic integrated classifiers and three compared methods. DT_A method combined with Savitzky Golay smoothing (SGS) and kernel principal component analysis (KPCA) obtained the best result of 99.24%. The best average accuracy of the proposed method was 89.29% for THz time-domain spectrum. The improved accuracy shows THz spectroscopy combined with integrated learning method may be an effective candidate technology for soybean detection.

We describe an approach to multivariate analysis, termed structured kernel principal component regression (sKPCR), to identify associations in voxel-level connectomes using resting-state functional magnetic resonance imaging (rsfMRI) data. This powerful and computationally efficient multivariate method can identify voxel-phenotype associations based on the whole-brain connectivity pattern of voxels, and it can detect linear and non-linear signals in both volume-based and surface-based rsfMRI data. For each voxel, sKPCR first extracts low-dimensional signals from the spatially smoothed connectivities by structured kernel principal component analysis, and then tests the voxel-phenotype associations by an adaptive regression model. The method's power is derived from appropriately modelling the spatial structure of the data when performing dimension reduction, and then adaptively choosing an optimal dimension for association testing using the adaptive regression strategy. Simulations based on real connectome data have shown that sKPCR can accurately control the false-positive rate and that it is more powerful than many state-of-the-art approaches, such as the connectivity-wise generalized linear model (GLM) approach, multivariate distance matrix regression (MDMR), adaptive sum of powered score (aSPU) test, and least-square kernel machine (LSKM). Moreover, since sKPCR can reduce the computational cost of non-parametric permutation tests, its computation speed is much faster. To demonstrate the utility of sKPCR for real data analysis, we have also compared sKPCR with the above methods based on the identification of voxel-wise differences between schizophrenic patients and healthy controls in four independent rsfMRI datasets. The results showed that sKPCR had better between-sites reproducibility and a larger proportion of overlap with existing schizophrenia meta-analysis findings. Code for our approach can be downloaded from https://github.com/weikanggong/sKPCR. [Abstract copyright: Copyright © 2018 Elsevier Inc. All rights reserved.]

Context Software defect prediction strives to detect defect-prone software modules by mining the historical data. Effective prediction enables reasonable testing resource allocation, which eventually leads to a more reliable software. Objective The complex structures and the imbalanced class distribution in software defect data make it challenging to obtain suitable data features and learn an effective defect prediction model. In this paper, we propose a method to address these two challenges. Method We propose a defect prediction framework called KPWE that combines two techniques, i.e., Kernel Principal Component Analysis (KPCA) and Weighted Extreme Learning Machine (WELM). Our framework consists of two major stages. In the first stage, KPWE aims to extract representative data features. It leverages the KPCA technique to project the original data into a latent feature space by nonlinear mapping. In the second stage, KPWE aims to alleviate the class imbalance. It exploits the WELM technique to learn an effective defect prediction model with a weighting-based scheme. Results We have conducted extensive experiments on 34 projects from the PROMISE dataset and 10 projects from the NASA dataset. The experimental results show that KPWE achieves promising performance compared with 41 baseline methods, including seven basic classifiers with KPCA, five variants of KPWE, eight representative feature selection methods with WELM, 21 imbalanced learning methods. Conclusion In this paper, we propose KPWE, a new software defect prediction framework that considers the feature extraction and class imbalance issues. The empirical study on 44 software projects indicate that KPWE is superior to the baseline methods in most cases.

In view of the fact that the traditional laser ultrasonic imaging test takes a long time and cannot achieve large area scanning of rail. This paper explores the possibility of combing the laser-ultrasonic technology and a hybrid intelligent method to fast achieve classification and evaluation of artificial rolling contact fatigue (RCF) defect in different depths. The laser ultrasonic scanning detection system is used to collect data samples from different locations of the defects quickly, and the signals are detected by an interferometer. Once the characteristic information of different rail defects is acquired and trained by Support Vector Machine (SVM), the high efficient and high-precision rail detection can be realized through the input of the feature in the detection process. The hybrid method is composed by Wavelet Packet Transform (WPT), Kernel Principal Component Analysis (KPCA) and SVM. The WPT is used to decompose the signal of surface defect in different frequency bands. The KPCA is used to eliminate the redundancy of the original feature set, thereby reducing the correlation among all the defect features. Wavelet packet time-frequency coefficient (X), energy (E) and local entropy (F) are generated and a new feature (Ynew) is created by fusing X, E and F, as a result of WPT and KPCA. Finally, a support vector machine (SVM) method is used to classify RCF defect in different depths. It implements a fast classification of small data. Compared with single features, fusion feature has the highest accuracy rate up to 98.73%.

In order to deal with the sensitivity of traditional kernel principal component analysis (KPCA) to the outliers and high computational complexity of the other existing robust KPCAs, a novel density-sensitive Robust Fuzzy Kernel Principal Component Analysis (DRF-KPCA) is proposed in this paper. First, the initial membership degree of the sample is determined by introducing the relative density. Second, the membership degree is formulated based on reconstruction error and updated by the optimal gradient descent approach, which can effectively solve the problem of the principal component skewing caused by the sensitivity of KPCA to the outliers. Ultimately, the computational complexity and running time of DRF-KPCA are significantly reduced by simplifying the calculation formula of the reconstruction error. In the experiments, as compared with KPCA and other modified approaches on both the datasets with outliers and the datasets without outliers, DRF-KPCA is evaluated to effectively eliminate the impacts of the outliers on the performance with low computational complexity. In addition, the influence of parameters on the performance of DRF-KPCA is also analyzed in detail and the suggestions of determination on the optimal coefficients are given. Eventually, the comparison results with the-state-of-art techniques on UCI benchmark datasets and other high-dimensional classification datasets demonstrate that the performance of DRF-KPCA is significantly improved.

Slope displacement prediction plays a critical role in the early warning system for landslide and greatly helps prevent property damage and loss of human lives. Given the non-stationary and complex characteristic of the slope deformation, this paper proposes a slope displacement prediction model and an early warming framework based on a set of sequential intelligent computing algorithms that can take advantages of Rough Set theory (RS), Kernel principal component analysis (KPCA), quantum particle swarm optimization (QPSO), least square support vector machine (LSSVM), and Markov chain (MC). Firstly, based on the analysis results of field monitoring data, the environmental parameters affecting the landslide displacement are used as the initial input variables and discretized to remove the effects of dimension and magnitude. After this process, RS is utilized to identify the important influence factors in order to eliminate the multi-collinearity and redundancy of the attributes that were selected initially. As the extracted parameters are still high-dimensional, KPCA is employed to fuse them into a comprehensive indicator to further reduce input dimension and computational cost. The nonlinear relation model between the indicator and displacement is established by using LSSVM, with the parameters optimized through QPSO that has much faster and better global search ability. Once the QPSO-LSSVM model is established, MC is integrated to refine the prediction results. Five months continuous field measurements from the real time monitoring system of Tuyang landslide is applied to evaluate the effectiveness of the proposed model. The results demonstrate that the proposed approach achieves higher prediction accuracy, faster convergence, and better generalization compared with existing prevalent models.

Feature extraction from vibration signals is considerably significant for condition monitoring and fault diagnosis. The Symbolic Aggregate approXimation (SAX) technique essentially transforming a real-valued time series into a symbol sequence, has been proven as a potential tool of feature extraction for machinery intelligent diagnosis. However, challenge still exists that the SAX cannot fulfill feature extraction tasks well since it is carried out only on the basis of mean value in time domain. To overcome this limitation, an improved SAX (ISAX) is proposed in this paper. This new method substitutes the feature of mean value in time domain with multiple features extracted from time, frequency and time-frequency domains in order to obtain comprehensive fault information. With the ISAX transformation, a vibration signal can be transformed into various symbol sequences according to the multi-domain features. Next the Shannon entropy technique is conducted on a symbol sequence to capture sequential patterns in local signals and then the Shannon entropy value is used as the eigenvalue of the symbol sequence. Various eigenvalues are obtained to describe a vibration signal from different perspectives, which leads to a better feature extraction. These eigenvalues are then fed into the Kernel Principal Component Analysis (KPCA) to reduce dimensions and extract principal features for classification tasks. Compared with SAX, the most significant advantage of ISAX is extracting comprehensive signal characteristics from multi-domain. Moreover, the ISAX captures fault information better considering the local fault patterns. The effectiveness and superiority of ISAX were validated by experimental studies using the fault signals of rolling bearings and reciprocating compressor valves with remarkably high classification rates.

A new nonlinear process monitoring algorithm called ensemble local kernel principal component analysis (ELKPCA) is proposed. Conventionally, the performance of kernel-based model depends on the width parameter selected empirically in Gaussian kernel function, which means a single parameter corresponds to a single model. Once a poor width parameter is decided, a single kernel-based model may be only effective for part faults. As a typical kernel-based method, the local kernel principal component analysis (LKPCA), considering both global and local structure information of the original data, faces the problem of width parameter selection as well. Since a single model is one-sided, an available way is to combine different single models and take advantage of them. The ensemble kernel principal component analysis (EKPCA) uses single KPCA models as its sub-models and ensemble learning approach is used to combine them. Due to inherit drawbacks from KPCA models, EKPCA only preserves global structure information of data, but ignores important local structure information. In this paper, to solve the above issues, both LKPCA and EKPCA is unified in the proposed framework. First, single LKPCA models are chosen instead and combined by using ensemble learning strategy. Then two monitoring statistics are turned into fault probabilities through Bayesian inference approach and weighted combination strategy, which makes the monitoring behavior easier and more clear. The result shows that ELKPCA model can not only take advantage of sub-LKPCA models effectively, allowing it selecting width parameter more easily and stably, but also retain both global and local structure information from input data by introducing the local structure analysis in the EKPCA model. Case studies on synthetic example and Tennessee Eastman process demonstrate the proposed method outperforms LKPCA and EKPCA and enhances monitoring performance significantly.

Variables in quality-related process monitoring can be divided into quality-relevant and quality-irrelevant groups depending on the correlation with the quality indicator. These variables can also be separated into multiple sets in which variables are closely relevant to one another because of the interdependence of the process. Block monitoring with reasonable variable partition and reliable model can distinguish quality-related and quality-unrelated faults and improve monitoring performance. A block monitoring method based on self-organizing map (SOM) and kernel approaches is proposed. After collecting and normalizing the sample data including process variables and quality ones, the data matrix is transposed. The inverted samples are used as the input of SOM, and variables with the same behavioral characteristic and a close correlation are topologically mapped in a similar area. Accordingly, samples can be visually blocked into quality-relevant and independent subspaces. Given the nonlinearity of industrial process, kernel partial least squares (KPLS) and kernel principal component analysis (KPCA) are employed to monitor the two types of blocks. The information provided by fault detection can reveal the effects on quality indicators and the location of faults. Finally, the effectiveness of SOM-KPLS/KPCA is evaluated using a numerical example and the Tennessee–Eastman process.

Wavelet-based feature extraction techniques are very promising for classifying epileptic electroencephalograph (EEG) but the determination of the optimal wavelet basis is intractable. To overcome this strait, maximal overlap discrete wavelet package transform (MODWPT) was introduced to characterize epileptic EEG for the first time and the multi-basis MODWPT-based feature extraction was further proposed in this study. Instead of merely using single wavelet basis, multiple bases were synchronously adopted in one-time implementation of multi-basis MODWPT. Six dimensionality reduction algorithms, namely principal component analysis (PCA), independent component analysis (ICA), kernel principal component analysis (KPCA), isometric feature mapping (ISOMAP), locally linear embedding (LLE) and Laplacian eigenmaps (LE), were then employed to map the extracted features into other domain in feature selection procedure. Finally, the mapped features were fed into a least squares support vector machine (LS-SVM) for classification and multiple kernel learning support vector machine (MKLSVM) was used as comparison. Experimental results show the combination of multi-basis MODWPT, ICA and LS-SVM with linear kernel provides the highest accuracy of 99.67% in classifying inter-ictal and ictal EEGs while MKLSVM in conjunction with multi-basis MODWPT-based PCA also leads to the maximal accuracy of 99.83%. Our presented scheme yields superior performance than the majority of newly-developed methods and is proven useful.

The nutritional value of goat milk is higher than that of cow milk, which is scarce and expensive. Therefore, many bad credit enterprises mix cow’s milk into goat milk and sell it as pure goat milk, which infringes on the rights and interests of consumers. The adulteration of dairy products has a long history. Many techniques have been applied for the detection of dairy products’ adulteration, but no obvious results have been achieved. The electronic tongue system, based on a virtual instrument, can provide fast detection of goat milk’s quality. The qualitative identification of different purities of goat milk by kernel principal component analysis (KPCA) has achieved 100% accuracy. By comparing the four algorithms of the optimal extreme learning machine (ELM), it was determined that the quantitative model for the prediction of adulterated goat milk, established by the improved artificial fish swarm optimized ELM method, had high prediction accuracy. The prediction set coefficient R2 was 0.998, the average absolute error (MAE) was 0.083, and the root mean square error (MSE) was 0.011. According to its unique advantages, the electronic tongue provides a new idea and method for the detection of food adulteration. As a modern intelligent sensory instrument, the electronic tongue has great potential for brand identification as well as the identification of goat milk and goat milk powder adulteration.

Rotation forest (RoF) is an ensemble classifier combining linear analysis theories and decision tree algorithms. In recent existing works, RoF was widely applied to various fields with outstanding performance compared to traditional machine learning techniques, given that a reasonable number of base classifiers is provided. However, the conventional RoF algorithm suffers from classifying linearly inseparable datasets. In this study, a hybrid algorithm integrating kernel principal component analysis (KPCA) and the conventional RoF algorithm is proposed to overcome the classification difficulty for linearly inseparable datasets. The radial basis function (RBF) is selected as the kernel for the KPCA method to establish the nonlinear mapping for linearly inseparable data. Moreover, we evaluate various kernel parameters for better performance. Experimental results show that our algorithm improves the performance of RoF with linearly inseparable datasets, and therefore provides higher classification accuracy rates compared with other ensemble machine learning methods.

Hybrid optimization algorithms which combine Evolutionary Algorithms (EAs) and Gradient-Based (GB) methods o refine the most promising solutions, are valuable tools for use in engineering optimization. Several hybrid methods can be found in the literature; differences among them are associated with the criteria used to select individuals for refinement through the GB method and the feedback the EA gets from the latter. GB methods require the gradient of the objective functions with respect to the design variables. By employing the adjoint method in problems governed by partial differential equations, the cost of computing the gradient becomes independent of the number of design variables. For multi-objective optimization problems this paper is exclusively dealing with, the availability of the gradients of all objective functions is not enough. Hybrid algorithms require the computation of descent directions in the objective space capable of improving the current front of non-dominated solutions. Using the sum of weighted objectives as the new objective function is ineffective. In this paper, a method which refrains from using arbitrarily defined weights is proposed. The method is driven by data obtained from the Principal Component Analysis (PCA) of the objective function values of the elite individuals at each generation of the EA. The PCA, with computational cost that of the solution of an eigenproblem, identifies the direction in the objective space along which the current front of non-dominated solutions should be improved. This along with the gradients computed by the adjoint method are used by the GB method to refine selected individuals. The efficiency of the proposed hybrid algorithm is further improved by employing online trained surrogate models or metamodels and Kernel PCA within the EA-based search. The proposed method is demonstrated in aerodynamic shape optimization problems, using in-house Computational Fluid Dynamics software and its adjoint.

Determining the geographical origin of wines is a major challenge in wine authentication, but little information is available regarding non-parametric statistical approaches for wines. In this study, we collected 33 domestic Chardonnay wines vinified on a small scale from grapes cultivated in Japan, and 42 Chardonnay wines imported from 8 countries, for oxygen stable isotope and multi-element analyses. Non-metric multidimensional scaling (NMDS), kernel principal component analysis (KPCA) and principal component analysis (PCA) were applied to the oxygen stable isotopic compositions (δ18O) and the concentrations of 18 elements in the wines to compare the extractions by parametric and non-parametric methods. The non-parametric methods, NMDS and KPCA, separated domestic from imported Chardonnay wines better than the parametric method, PCA. Of 19 variables, 18 were important for geographical discrimination, with the δ18O value being the most significant in all statistic methods. Non-parametric multivariate analyses will help discriminate domestic from imported Chardonnay wines.

The increased complexity of digitalized process systems requires advanced tools to detect and diagnose faults early to maintain safe operations. This study proposed a hybrid model that consists of Kernel Principal Component Analysis (kPCA) and DNNs that can be applied to detect and diagnose faults in various processes. The complex data is processed by kPCA to reduce its dimensionality; then, simplified data is used for two separate DNNs for training (detection and diagnosis). The relative performance of the hybrid model is compared with conventional methods. Tennessee Eastman Process was used to confirm the efficacy of the model. The results show the reduction of input dimensionality increases classification accuracy. In addition, splitting detection and diagnosis into two DNNs results in reduced training times and increased classification accuracy. The proposed hybrid model serves as an important tool to detect the fault and take early corrective actions, thus enhancing process safety.

The chemometric analysis of chromatographic data is commonly used for discriminating bee propolis based on their geographical origin. Traditional machine learning methods for this purpose include principal component analysis and hierarchical clustering. When viewed as a time-series, the key discriminatory features in chromatographic data are the peaks, which should have similar location patterns for the propolis of the same origin. However, the peaks between same-origin samples are not always exactly aligned. Without proper alignment, samples from the same origin may be perceived by the clustering method to be very different. In this paper, we propose a novel dynamic time warping kernel principal component analysis (DTW-KPCA) method for the chemometric discrimination of propolis. The proposed method uses a Gaussian dynamic time warping kernel to measure the similarity between chromatographic time-series which incorporates time-series alignment. Results show a better clustering of propolis samples compared to the currently used methods, as measured by the silhouette coefficient. Hence, the proposed method enables a more reliable clustering as to the origin of propolis samples.

In the era of the 4th industrial revolution, a key challenge for the industries is the efficient reduction of the production cost caused by malfunctioning equipment. This paper proposes a Fault Detection and Diagnosis (FDD) framework for Non-Linear Processes utilizing Dynamic Neural Networks and feature reduction methods. We investigate both types of dynamic neural models, ie. Recurrent Neural Networks -in particular Long Short-Term Memory (LSTM) models, and Time Delay Neural Networks (TDNN). Intending to mitigate the overfitting problem, we also investigated the use of feature reduction techniques such as Non-Negative Matrix Factorization (NMF), Principal Component Analysis (PCA), and kernel PCA (kPCA), as preprocessing steps in our Machine Learning pipeline. The Tennessee Eastman Process (TEP) is used to evaluate our proposed framework on 18 different faults. Our simulations demonstrate that our method outperforms state of the art methods in the majority of those faults.

Automatic sign language translation process relies mainly on dictionaries of signs to interpret the right meaning of gestures. Due to the lack of large multi sign language dictionaries covering all the aspect of sign languages, the collaborative approach to create signs becomes essential. In fact, the collaborative sign creation process based on Kinect motion capture tool requires the collaboration of non expert users to make sign language dictionaries. However, due to the availability constraint of sign language experts to validate the created signs and the huge amount of signs to be validated manually, the automatic sign language validation process becomes the most suitable solution. In this paper, we present a new automatic and unsupervised sign validation process based on machine learning techniques applied on sign replicas. Given a set of replicas (records) of the same sign created by different non expert sign language user, our main goal is to select the adequate sign records to be used to generate the closest sign signature compared to the one created by sign language expert. For this aim, we present an automatic sign selection and validation solution based on unsupervised clustering of sign motion parameters related to the different sign replicas. We conducted an experimental study to validate 300 ASL signs based on four unsupervised clustering methods, namely, Kernel PCA Kmeans, GMM, Spectral clustering and kernel Kmeans. We concluded that the use our sign validation process using Spectral clustering method allows us to select the right sign replicas to be used to generate the user sign signature. The use of our unsupervised sign validation process onto 3000 ASL sign replicas (300 sign * 10 replicas) lead us to enhance the R2 score average from 0.4830 without sign validation to 0.9123 with sign validation compared to expert sign signature.

Both linear and nonlinear relationships may exist among process variables, and monitoring a process with such complex relationships among variables is imperative. However, individual principal component analysis (PCA) or kernel PCA (KPCA) may not be able to characterize these complex relationships well. This paper proposes a parallel PCA–KPCA (P-PCA–KPCA) modeling and monitoring scheme that incorporates randomized algorithm (RA) and genetic algorithm (GA) for efficient fault detection for a process with linearly correlated and nonlinearly related variables First, to determine the included variables in the parallel PCA (P-PCA) and the parallel KPCA (P-KPCA) models, GA-based optimization is performed, in which RA is used to generate faulty validation data. Second, monitoring statistics are established for the P-PCA and the P-KPCA models, in which the process status is determined. The proposed monitoring scheme discriminates the linear and nonlinear relationships among variables in a process and deals with nonlinear processes efficiently. We provide case studies on a numerical example and the continuous stirred tank reactor process. These case studies demonstrate that the proposed P-PCA–KPCA monitoring scheme is better than conventional PCA- or KPCA-based methods at performing nonlinear process monitoring.

This paper presents a global manifold margin learning approach for data feature extraction or dimensionality reduction, which is named locally linear representation manifold margin (LLRMM). Provided that points locating on one manifold are of the same class and those residing on the corresponding manifolds are varied labeled, LLRMM is desired to identify different manifolds, respectively. In the proposed LLRMM, it firstly constructs both a between-manifold graph and a within-manifold graph. In the between-manifold graph, for any point, its k nearest neighbors and itself must belong to different manifolds. However, any node and its neighborhood points should be on the same manifold in the within-manifold graph. Then we use the minimum locally linear representation trick to reconstruct any node with their corresponding k nearest neighbors in both graphs, from which a between-manifold graph scatter and a within-manifold graph scatter can be reasoned, followed by a novel global model of manifold margin. At last, a projection will be explored to map the original data into a low dimensional subspace with the maximum manifold margin. Experiments on some widely used face data sets including AR, CMU PIE, Yale, YaleB and LFW have been carried out, where the performance of the proposed LLRMM outperforms those of some other methods such as kernel principal component analysis (KPCA), non-parametric discriminant analysis (NDA), reconstructive discriminant analysis (RDA), discriminant multiple manifold learning (DMML) and large margin nearest neighbor (LMNN).

The objective of this work is to investigate the pattern exhibited by detrended intra-day BSE Sensex data for the years 2006 to 2012. The detrended data are analysed using Principal Component Analysis (PCA) and its non-linear version, Kernel Principal Component Analysis (KPCA). The detrended data is found to display a high degree of correlation which indicates that the evolution of the detrended prices is restricted to a very low dimensional subspace of the original vector space in which the analysis is done. Different types of synthetic data are generated, which when subject to the same set of analyses, are found to give results along expected lines, thereby verifying the efficacy of the techniques employed. Hurst coefficients of the detrended data sets are calculated for different years using modified R/S analysis. The Hurst coefficient is also computed for the entire data set by gradually changing the scale of analysis and also by using the sliding window technique. In all cases the data set are found to be persistent in nature thereby reinforcing the conclusions obtained by the PCA/KPCA formalism.

In this study, a QCM sensor array with four different surface modified QCM sensors (i.e., multi-walled carbon nanotubes, graphene, copper oxide and polyaniline) was fabricated and was applied to evaluate the shelf life of eggs by sensing the volatiles. The morphologies of the sensitive materials on electrodes were analyzed by field-emission scanning electron microscope (FE-SEM), and sensor responses were monitored by a self-made frequency measurement system. Particularly, these four sensors exhibited relatively good sensitivity, reversibility, repeatability and long-term stability. Then, the sensor array was applied to detect volatiles of eggs with different shelf life. The result of linear discriminant analysis (LDA) outperformed principal component analysis (PCA), and exhibited excellent classification accuracy. Partial least square regression (PLSR) was employed to predict the shelf life of eggs and the fitting coefficient of determination (R2) of PLSR increased from 0.8474 to 0.9547 after kernel principal component analysis (KPCA) introduced, which showed satisfactory prediction performance. It could be concluded that the QCM sensor array is effective for the detection of eggs with different shelf life, offering an alternative strategy to estimate the freshness of eggs.

Background There has been growing interest towards naturalistic neuroimaging experiments, which deepen our understanding of how human brain processes and integrates incoming streams of multifaceted sensory information, as commonly occurs in real world. Music is a good example of such complex continuous phenomenon. In a few recent fMRI studies examining neural correlates of music in continuous listening settings, multiple perceptual attributes of music stimulus were represented by a set of high-level features, produced as the linear combination of the acoustic descriptors computationally extracted from the stimulus audio. New method fMRI data from naturalistic music listening experiment were employed here. Kernel principal component analysis (KPCA) was applied to acoustic descriptors extracted from the stimulus audio to generate a set of nonlinear stimulus features. Subsequently, perceptual and neural correlates of the generated high-level features were examined. Results The generated features captured musical percepts that were hidden from the linear PCA features, namely Rhythmic Complexity and Event Synchronicity. Neural correlates of the new features revealed activations associated to processing of complex rhythms, including auditory, motor, and frontal areas. Comparison with existing method Results were compared with the findings in the previously published study, which analyzed the same fMRI data but applied linear PCA for generating stimulus features. To enable comparison of the results, methodology for finding stimulus-driven functional maps was adopted from the previous study. Conclusions Exploiting nonlinear relationships among acoustic descriptors can lead to the novel high-level stimulus features, which can in turn reveal new brain structures involved in music processing.

Principal component analysis (PCA) is largely adopted for chemical process monitoring and numerous PCA-based systems have been developed to solve various fault detection and diagnosis problems. Since PCA-based methods assume that the monitored process is linear, nonlinear PCA models, such as autoencoder models and kernel principal component analysis (KPCA), has been proposed and applied to nonlinear process monitoring. However, KPCA-based methods need to perform eigen-decomposition (ED) on the kernel Gram matrix whose dimensions depend on the number of training data. Moreover, prefixed kernel parameters cannot be most effective for different faults which may need different parameters to maximize their respective detection performances. Autoencoder models lack the consideration of orthogonal constraints which is crucial for PCA-based algorithms. To address these problems, this paper proposes a novel nonlinear method, called neural component analysis (NCA), which intends to train a feedforward neural work with orthogonal constraints such as those used in PCA. NCA can adaptively learn its parameters through backpropagation and the dimensionality of the nonlinear features has no relationship with the number of training samples. Extensive experimental results on the Tennessee Eastman (TE) benchmark process show the superiority of NCA in terms of missed detection rate (MDR) and false alarm rate (FAR). The source code of NCA can be found in https://github.com/haitaozhao/Neural-Component-Analysis.git., 10 pages,11 figures

To realize and accelerate the pace of intelligent manufacturing, this paper presents a novel tool wear assessment technique based on the integrated radial basis function based kernel principal component analysis (KPCA_IRBF) and Gaussian process regression (GPR) for real-timely and accurately monitoring the in-process tool wear parameters (flank wear width). The KPCA_IRBF is a kind of new nonlinear dimension-increment technique and firstly proposed for feature fusion. The tool wear predictive value and the corresponding confidence interval are both provided by utilizing the GPR model. Besides, GPR performs better than artificial neural networks (ANN) and support vector machines (SVM) in prediction accuracy since the Gaussian noises can be modeled quantitatively in the GPR model. However, the existence of noises will affect the stability of the confidence interval seriously. In this work, the proposed KPCA_IRBF technique helps to remove the noises and weaken its negative effects so as to make the confidence interval compressed greatly and more smoothed, which is conducive for monitoring the tool wear accurately. Moreover, the selection of kernel parameter in KPCA_IRBF can be easily carried out in a much larger selectable region in comparison with the conventional KPCA_RBF technique, which helps to improve the efficiency of model construction. Ten sets of cutting tests are conducted to validate the effectiveness of the presented tool wear assessment technique. The experimental results show that the in-process flank wear width of tool inserts can be monitored accurately by utilizing the presented tool wear assessment technique which is robust under a variety of cutting conditions. This study lays the foundation for tool wear monitoring in real industrial settings.