Curvilinear Component Research Articles

A visual tool for monitoring and detecting anomalies in robot performance

In robotic systems, both software and hardware components are equally important. However, scant attention has been devoted until now in order to detect anomalies/failures affecting the software component of robots while many proposals exist aimed at detecting physical anomalies. To bridge this gap, the present paper focuses on the study of anomalies affecting the software performance of a robot by using a novel visualization tool. Unsupervised visualization methods from the machine learning field are applied in order to upgrade the recently proposed Hybrid Unsupervised Exploratory Plots (HUEPs). Furthermore, Curvilinear Component Analysis and t-distributed stochastic neighbor embedding are added to the original HUEPs formulation and comprehensively compared. Furthermore, all the different combinations of HUEPs are validated in a real-life scenario. Thanks to this intelligent visualization of robot status, interesting conclusions can be obtained to improve anomaly detection in robot performance.

Modeling and multi-class classification of vibroarthographic signals via time domain curvilinear divergence random forest

They distinguish information linked with knee-joints’ articular cartilage surfaces feasibly extracted from vibroarthographic (VAG) signals. Even though diverse researches have presented numerous types of specifications for the investigation and classification of VAG signals, minimal methods have been examined to constitute the differentiations between different knee joint disorder levels. In this paper, a multi-class classification method called Intrinsic Three Band Curvilinear Principal Component and Divergence Forest (ITBCPC-DF) is proposed to precisely differentiate between different levels of disorder. First, an Intrinsic Three Band Pre-processing is applied to the input VAG signal images to prevent aliasing and eliminate muscle contraction interference, therefore contributing to de-noised signal images. Next, Principal Time Domain Curvilinear Energy is applied to the de-noised signal images that extract the requisite patterns in the VAG signals associated with the knee-joint position. Finally, Probable Divergent Random Forest is applied for multi-class classification. Compared with other VAG signal analysis methods, the experimental results show that the proposed ITBCPC-DF method shows comparatively better accuracy, sensitivity, specificity, and AUC curve.

Induction Machine Stator Fault Tracking Using the Growing Curvilinear Component Analysis

Detection of stator-based faults in Induction Machines (IMs) can be carried out in numerous ways. In particular, the shorted turns in stator windings of IM are among the most common faults in the industry. As a matter of fact, most IMs come with pre-installed current sensors for the purpose of control and protection. At this aim, using only the stator current for fault detection has become a recent trend nowadays as it is much cheaper than installing additional sensors. The three-phase stator current signatures have been used in this study to observe the effect of stator inter-turn fault with respect to the healthy condition of the IM. The pre-processing of the healthy and faulty current signatures has been done via the in-built DSP module of dSPACE after which, these current signatures are passed into the MATLAB® software for further analysis using AI techniques. The authors present a Growing Curvilinear Component Analysis (GCCA) neural network that is capable of detecting and follow the evolution of the stator fault using the stator current signature, making online fault detection possible. For this purpose, a topological manifold analysis is carried out to study the fault evolution, which is a fundamental step for calibrating the GCCA neural network. The effectiveness of the proposed method has been verified experimentally.

A Topological Neural-Based Scheme for Classification of Faults in Induction Machines

This article presents a data-driven approach for the classification of faults in induction machines. The designed scheme involves newly engineered features extracted from the line current signals, which provides an improved fault discrimination. For this purpose, a topological-based fast projection technique (curvilinear component analysis) is used as a tool to reduce the dimensionality of the data and interpret the feature behavior. Consequently, a shallow convolutional neural network has been designed to classify the three-phase stator current signals. Experimental tests at different operating conditions have assessed the procedure, confirming its effectiveness and suitability for online and real-time diagnostics.

Shallow Versus Deep Neural Networks in Gear Fault Diagnosis

Accurate gear defect detection in induction machine-based systems is a fundamental issue in several industrial applications. At this aim, shallow neural networks, i.e., architectures with only one hidden layer, have been used after a feature extraction step from vibration, torque, acoustic pressure and electrical signals. Their additional complexity is justified by their ability in extracting its own features and in the very high-test classification rates. These signals are here analyzed, both geometrically and topologically, in order to estimate the class manifolds and their reciprocal positioning. At this aim, the different states of the gears are studied by using linear (Pareto charts, biplots, principal angles) and nonlinear (curvilinear component analysis) techniques, while the class clusters are visualized by using the parallel coordinates. It is deduced that the class manifolds are compact and well separated. This result justifies the use of a shallow neural network, instead of a deep one, as already remarked in the literature, but with no theoretical justification. The experimental section confirms this assertion, and also compares the shallow neural network results with the other machine learning techniques used in the literature.

Soft computing applications in the field of industrial and environmental enterprises

Soft computing applications in the field of industrial and environmental enterprises

Degradation assessment of ball bearings utilizing curvilinear component analysis

The performance degradation assessment of ball bearings is of great importance to increase the efficiency and the reliability of rotating mechanical systems. The large dimensionality of feature space introduces a lot of noise and buries the potential information about faults hidden in the feature data. This paper proposes a novel health assessment method facilitated with two compatible methods, namely curvilinear component analysis and self-organizing map network. The novelty lies in the implementation of a vector quantization approach for the sub-manifolds in the feature space and to extract the fault signatures through nonlinear mapping technique. Curvilinear component analysis is a nonlinear mapping tool that can effectively represent the average manifold of the highly folded information and further preserves the local topology of the data. To answer the complications and to accomplish reliability and accuracy in bearing performance degradation assessment, the work is carried out with following steps; first, ensemble empirical mode decomposition is used to decompose the vibration signals into useful intrinsic mode functions; second, two fault features i.e. singular values and energy entropies are extracted from the envelopes of the intrinsic mode function signals; third, the extracted feature vectors under healthy conditions, further reduced with curvilinear component analysis are used to train the self-organizing map model; finally, the reduced test feature vectors are supplied to the trained self-organizing map and the confidence value is obtained. The effectiveness of the proposed technique is validated on three run-to-failure test signals with the different type of defects. The results indicate that the proposed technique detects the weak degradation earlier than the widely used indicators such as root mean square, kurtosis, self-organizing map-based minimum quantization error, and minimum quantization error-based on the principal component analysis.

Gaining deep knowledge of Android malware families through dimensionality reduction techniques

This research proposes the analysis and subsequent characterisation of Android malware families by means of low dimensional visualisations using dimensional reduction techniques. The well-known Malgenome data set, coming from the Android Malware Genome Project, has been thoroughly analysed through the following six dimensionality reduction techniques: Principal Component Analysis, Maximum Likelihood Hebbian Learning, Cooperative Maximum Likelihood Hebbian Learning, Curvilinear Component Analysis, Isomap and Self Organizing Map. Results obtained enable a clear visual analysis of the structure of this high-dimensionality data set, letting us gain deep knowledge about the nature of such Android malware families. Interesting conclusions are obtained from the real-life data set under analysis.

The Growing Curvilinear Component Analysis (GCCA) neural network

Big high dimensional data is becoming a challenging field of research. There exist a lot of techniques which infer information. However, because of the curse of dimensionality, a necessary step is the dimensionality reduction (DR) of the information. DR can be performed by linear and nonlinear algorithms. In general, linear algorithms are faster because of less computational burden. A related problem is dealing with time-varying high dimensional data, where the time dependence is due to nonstationary data distribution. Data stream algorithms are not able to project in lower dimensional spaces. Indeed, only linear projections, like principal component analysis (PCA), are used in real time while nonlinear techniques need the whole database (offline). The Growing Curvilinear Component Analysis (GCCA) neural network addresses this problem; it has a self-organized incremental architecture adapting to the changing data distribution and performs simultaneously the data quantization and projection by using CCA, a nonlinear distance-preserving reduction technique. This is achieved by introducing the idea of “seed”, pair of neurons which colonize the input domain, and “bridge”, a novel kind of edge in the manifold graph, which signals the data non-stationarity. Some artificial examples and a real application are given, with a comparison with other existing techniques.

A systematic and quantitative evaluation of plantar stimulation: The effect of type, pattern, force of stimulation in eliciting an accurate plantar response

ObjectivesSystematic and quantitative evaluation of the plantar reflex has been infrequently studied in the past and can help assess the vexing variables encountered in its elicitation. The objective of this study was to determine the effect of type, pattern and force of stimulation in eliciting an accurate plantar response in patients with pyramidal dysfunction and healthy individuals. Patients and methodsA special instrument was designed to give a predesigned force of stimulus. The plantar surface of foot was divided into nine parts and point and stroke stimulations were studied systematically in pyramidal weakness feet (cases) and healthy control feet (controls) with predefined forces. Results were tabulated and statistically analysed. ResultsStroke stimulation was superior to point stimulation in eliciting plantar response. There was no significant difference in the responses to the three predefined forces used for stroke stimulations. Sensitivity of Babinski sign was 72.9% and specificity was 100%. In pyramidal weakness feet, extensor responses were significantly elicited from lateral starting stroke patterns (67%) and on moving medially they were replaced by flexor responses (44%). Withdrawal responses increased with the stimulations reaching the distal foot and with the curvilinear component of stimulations. Sensitivity responses (related to the sensitivity of an individual) contaminate the plantar response and occasionally make its interpretation difficult. In subjects with bilateral sensitivity with unilateral disease, knowing the sensitivity pattern on the normal side improved the interpretation of plantar response on the abnormal side. ConclusionsBased on this study, the optimal method for eliciting Babinski sign accurately is to stroke the lateral aspect of the sole of the foot in a straight line up to mid foot. This should be performed in both feet three times, and if the weakness is unilateral, it should be performed in the normal leg first to aid interpretation of the affected leg.

Beta Hebbian Learning as a New Method for Exploratory Projection Pursuit.

In this research, a novel family of learning rules called Beta Hebbian Learning (BHL) is thoroughly investigated to extract information from high-dimensional datasets by projecting the data onto low-dimensional (typically two dimensional) subspaces, improving the existing exploratory methods by providing a clear representation of data's internal structure. BHL applies a family of learning rules derived from the Probability Density Function (PDF) of the residual based on the beta distribution. This family of rules may be called Hebbian in that all use a simple multiplication of the output of the neural network with some function of the residuals after feedback. The derived learning rules can be linked to an adaptive form of Exploratory Projection Pursuit and with artificial distributions, the networks perform as the theory suggests they should: the use of different learning rules derived from different PDFs allows the identification of "interesting" dimensions (as far from the Gaussian distribution as possible) in high-dimensional datasets. This novel algorithm, BHL, has been tested over seven artificial datasets to study the behavior of BHL parameters, and was later applied successfully over four real datasets, comparing its results, in terms of performance, with other well-known Exploratory and projection models such as Maximum Likelihood Hebbian Learning (MLHL), Locally-Linear Embedding (LLE), Curvilinear Component Analysis (CCA), Isomap and Neural Principal Component Analysis (Neural PCA).

Beacon deployment strategy for guaranteed localization in wireless sensor networks

In wireless sensor networks, beacons are always treated as infrastructures for localization. After beacons are deployed, non-beacon nodes can be located by simple schemes such as multilateration and multidimensional scaling (MDS). Deploying as many beacons as needed is an efficient way to improve localization accuracy where a global positioning system does not work well or a higher location accuracy is required. With more beacons to be deployed, the configuration of beacons' positions will have to be done manually. Therefore position auto-configuration using measured distances between these beacons can save a lot of efforts for the deployment. One challenge of this auto-configuration is that the positions should be uniquely determined based on the measured distances. In graph theory, it is a problem of unique realization in which the positions of vertices are determined by edges between them. Addressing this problem is one major aspect of this paper. To determine whether the topology of a network is a unique realization, this paper proposes a novel category of topology named Uniquely Determined Topology, with which edges in a d-dimensional space can be reduced from $$d+1$$d+1 to d in each extension, which is less strict and more suitable for beacon deployment. The other aspect of this paper is to improve localization accuracy of the deployed beacons. In MDS and curvilinear component analysis, a shortest-path algorithm is adopted to approximately reconstruct the distance matrix between each two nodes, and our proposed Uniquely Determined Topology has a feature that a distance calculation model can be adopted to replace the shortest-path algorithm, therefore that the local distance matrix can be reconstructed more accurately. Theoretical analysis shows that it has a low computational complexity to determine whether a deployment is a Uniquely Determined Topology. Simulations show the advantages of the improved localization scheme, in that they do not depend on the connectivity level of the networks, and they can provide accurate localization when the estimation accuracy of distances is high.

Enhancing the clustering process in the category model load profiling

Load profiling based consumer categorisation is a procedure that involves stages such as the proper representation of the load data, the selection of one or more pattern recognition algorithms, the assessment of their performance, the formation of daily load curve clusters for each consumer and the formation of the final consumer clusters. This study proposes improvements for every stage, while a new stage is introduced to enhance the whole procedure. All the pattern representation techniques of the load profiling related literature are examined together with the most widely used clustering algorithm, namely the K-means. Two novel modified forms of the algorithm are introduced, specially designed by the authors for the problem under study. The aim is to optimally group together similar daily demand patterns to provide exploitable information to the utilities and the retailers for the efficient design and implementation of innovative energy services, targeted to the different consumer categories. Moreover, all the pattern representation techniques that have been proposed in the literature are examined. These techniques can be divided in three major categories, namely time-domain indices, load shape indices and frequency-domain indices. Furthermore, four proposed dimensionality reduction techniques are investigated, namely the principal components analysis (PCA), Sammon mapping (SM), curvilinear component analysis (CCA) and canonical variate analysis (CVA) (6, 7). To enhance the clustering performance, we examined reduction schemes from 0-98% of the initial patterns dimension. Therefore, this paper aims to help the user to select the most efficient load profiling approach in terms of minimising the clustering error, by presenting a detailed comparative analysis of existing algorithms and techniques.

Repairing uniform experimental designs: Detection and/or elimination of clusters, filling gaps

Construction of Space Filling Designs in high dimensional space remains difficult since powerful algorithms at low dimensions become difficult to use at higher dimensions that leads to non-uniform distribution in the factor space. We propose in this paper two approaches in order to repair designs: Curvilinear Component Analysis (CCA) and the Wootton, Sergent, Phan-Tan-Luu's algorithm called WSP in order to detect clusters and to fill gaps. Thus, CCA allows visualization of two or more very closely-spaced points in D dimensions by projecting them in a 2 dimensions space. Then identified clusters can be eliminated using the WSP algorithm. Moreover, the presence of gaps in input space could be very problematic since no information on the phenomenon is available and the WSP algorithm will be used in order to fill gaps by adding points in the “empty” zones. A new quality criterion has been proposed in order to follow the reparation steps. Examples in different dimensions are presented to illustrate these methods.

Bearing Fault Detection by a Novel Condition-Monitoring Scheme Based on Statistical-Time Features and Neural Networks

Bearing degradation is the most common source of faults in electrical machines. In this context, this work presents a novel monitoring scheme applied to diagnose bearing faults. Apart from detecting local defects, i.e., single-point ball and raceway faults, it takes also into account the detection of distributed defects, such as roughness. The development of diagnosis methodologies considering both kinds of bearing faults is, nowadays, subject of concern in fault diagnosis of electrical machines. First, the method analyzes the most significant statistical-time features calculated from vibration signal. Then, it uses a variant of the curvilinear component analysis, a nonlinear manifold learning technique, for compression and visualization of the feature behavior. It allows interpreting the underlying physical phenomenon. This technique has demonstrated to be a very powerful and promising tool in the diagnosis area. Finally, a hierarchical neural network structure is used to perform the classification stage. The effectiveness of this condition-monitoring scheme has been verified by experimental results obtained from different operating conditions.

Spatial Distance Preservation based Methods for Non-Linear Dimensionality Reduction

The preservation of the pairwise distances measured in a data set ensures that the low dimensional embedding inherits the main geometric properties of the data like the local neighborhood relationships. In this paper, distance preserving technique namely, Sammons nonlinear mapping (Sammon‟s NLM) and Curvilinear Component Analysis (CCA) have been discussed and compared for non-linear dimensionality reduction. Basic principle in both the technique is that local neighborhood relationship is maintained. The results have beencompared for both the techniques on artificially generated data set using MATLAB software. General Terms Neighborhood relationship, Variance, PCA, Manifold, Nonlinear, Convergence.

Incorporating visualisation quality measures to curvilinear component analysis

Curvilinear Component Analysis (CCA) is a useful data visualisation method. CCA has the technical property that its optimisation surface, as defined by its stress function, changes during the optimisation according to a decreasing parameter. CCA uses a variant of the stochastic gradient descent method to create a mapping of data. In the optimisation method of CCA, the stress function is only a general guide towards an acceptable mapping. In other multidimensional scaling methods such as Sammon’s mapping, the best mapping among multiple runs from different initialisations can be chosen by selecting the mapping with the lowest stress, whereas in CCA the embedding is simply the result of one run, surely we can have multiple starts. As a consequence of the absence of an objective function to be used as a selection criterion, embedding made by CCA can be poorly optimised. In this paper we present a new way of improving the optimisation of CCA by integrating non-stress data visualisation quality measures into the existing algorithm. We first use data visualisation quality measures to select the best mapping from multiple runs of a standard stochastic gradient descent implementation; then we tune various parameters involved to achieve further enhancement. A brief comparison with other dimensionality reduction methods is included.

Localization in Wireless Sensor Networks and Anchor Placement

Applications of wireless sensor network (WSN) often expect knowledge of the precise location of the nodes. Many different localization protocols have been proposed that allow nodes to derive their location rather than equipping them with dedicated localization hardware such as GPS receivers, which increases node costs. We provide a brief survey of the major approaches to software-based node localization in WSN. One class of localization protocols with good localization performance patches together relative-coordinate, local maps into a global-coordinate map. These protocols require some nodes that know their absolute coordinates, called anchor nodes. While many factors influence the node position errors, in this class of protocols, using Procrustes Analysis, the placement of the anchor nodes can significantly impact the error. Through simulation, using the Curvilinear Component Analysis (CCA-MAP) protocol as a representative protocol in this category, we show the impact of anchor node placement and propose a set of guidelines to ensure the best possible outcome, while using the smallest number of anchor nodes possible.

Questionnaire- versus voice-based screening for laryngeal disorders

The usefulness of questionnaire and voice data to screen for laryngeal disorders is explored. Answers to 14 questions form a questionnaire data vector. Twenty-three variables computed by the commercial “Dr.Speech” software from a digital voice recording of a sustained phonation of the vowel sound/a/constitute a voice data vector. Categorization of the data into a healthy class and two classes of disorders, namely diffuse and nodular mass lesions of vocal folds is the task pursued in this work. Visualization of data and automated decisions is also an important aspect of this work. To make the categorization, a support vector machine (SVM) is designed based on genetic search. Linear as well as nonlinear canonical correlation analysis (CCA) is employed, to study relations between the questionnaire and voice data sets. The curvilinear component analysis, performing nonlinear mapping into a two-dimensional space, is used for visualizing data and decisions. Data from 240 patients were used in the experimental studies. It was found that the questionnaire data provide more information for the categorization than the voice data. There are 3–4 common directions along which the statistically significant variations of the questionnaire and voice data occur. However, the linear relations between the variations occurring in the two data sets are not strong. On the other hand, very strong linear relations were observed between the nonlinear variates obtained from the questionnaire data and linear ones computed from the voice data. Questionnaire data carry great potential for preventive health care in laryngology.

Unsupervised neural models for country and political risk analysis

This interdisciplinary research project focuses on relevant applications of Knowledge Discovery and Artificial Neural Networks in order to identify and analyze levels of country, business and political risk. Its main goal is to help business decision-makers understand the dynamics within the emerging market countries in which they operate. Most of the neural models applied in this study are defined within the framework of unsupervised learning. They are based on Exploratory Projection Pursuit, Topology Preserving Maps and Curvilinear Component Analysis. Two interesting real data sets are analyzed to empirically probe the robustness of these models. The first case study describes information from a significant sample of Spanish multinational enterprises (MNEs). It analyses data pertaining to such aspects as decisions over the location of subsidiary enterprises in various regions across the world, the importance accorded to such decisions and the driving forces behind them. Through a projection-based analysis, this study reveals a range of different reasons underlying the internationalization strategies of Spanish MNEs and the different goals they pursue. It may be concluded that projection connectionist techniques are of immense assistance in the process of identifying the internationalization strategies of Spanish MNEs, their underlying motives and the goals they pursue. The second case study covers several risk categories that include task policy, security, and political stability among others, and it tracks the scores of different countries all over the world. Interesting conclusions are drawn from the application of several business intelligence solutions based on neural projection models, which support data analysis in the context of country and political risk analysis.