Hierarchical K-means Clustering Research Articles

Prediction of groundwater contamination in an open landfill area using a novel hybrid clustering-based ai model

Groundwater is a vital resource that provides drinking water to over half of the world’s population. However, groundwater contamination has become a serious issue due to human activities such as industrialization, agriculture, and improper waste disposal. The impacts of groundwater contamination can be severe, including health risks, environmental damage, and economic losses. A list of unknown groundwater contamination sources has been developed for the Wang-Tien landfill using a groundwater modeling system (GMS). Further, AI-based models have been developed which accurately predict the contamination from the sources at this site. A serious complication with most previous studies using artificial neural networks (ANN) for contamination source identification has been the large size of the neural networks. We have designed the ANN models which use three different ways of presenting inputs that are categorized by hierarchical K-means clustering. Such an implementation reduces the overall complexity of the model along with high accuracy. The predictive capability of developed models was assessed using performance indices and compared with the ANN models. The results show that the hybrid model of hierarchical K-means clustering and ANN model (HCA-ANN) is a highly accurate model for identifying pollution sources in contaminated water.

Hierarchical K-means clustering method for accelerated Lorentzian estimation (KALE) in chemical exchange saturation transfer-magnetic resonance imaging quantification.

Quantification of in vivo chemical exchange saturation transfer (CEST) magnetic resonance signals is challenging due to contamination from coexisting effects, including the direct water effect and asymmetric magnetization transfer. Fitting-based analysis allows the calculation of multiple types of signals from the line shape of Z-spectra. However, the conventional voxelwise method has several drawbacks, including its long computation time and its susceptibility to image noise and Z-spectra oscillations, and it is difficult to determine the initial fitting parameters. Herein, we propose a K-means clustering method for accelerated Lorentzian estimation (KALE) in CEST quantification. Briefly, voxels in CEST images are clustered into K groups according to their Z-spectra characteristics. A 'groupwise' fitting process is then performed with preset initial values, yielding a set of fitted spectra and fitted parameters for each group. With the updated initial values, each group is further clustered into subgroups, and groupwise fitting is performed again. This hierarchical K-means clustering and parameter updating process continues until the pixel number or intensity error meets the termination criteria. Voxelwise fitting could be further conducted to improve the quantification images (termed voxel-K) by utilizing the previous groupwise KALE results as the initial values (termed group-K). Incorporated with Lorentzian difference (LD) quantification, KALE was first optimized and evaluated on 5 healthy human brain datasets at 3 Tesla. Compared with traditional voxel-by-voxel LD quantification, the computation times of group-K and voxel-K were significantly reduced by ~85% and ~70%, respectively (P<0.001). Furthermore, the group-K images exhibited better denoising performance than traditional LD and voxel-K. KALE was further validated on six ischemic rat brains acquired at 7 Tesla, with both LD_group-K and LD_voxel-K displaying almost identical contrast maps with traditional voxelwise maps. When incorporated with the five-pool Lorentzian fitting (LF), KALE exhibited an improved contrast-to-noise ratio (CNR) for amplitude maps of each pool [P=0.003, 0.015, 0.047, and 0.047 for amide, nuclear Overhauser effect (NOE), magnetic transfer (MT) and guanidine amine, respectively] and improved fitting goodness (P=0.033). KALE quantification provides comparable or even superior contrast maps to traditional voxelwise fitting, with significantly reduced computation time. The 'smart' and hierarchical voxel-clustering and parameter updating process of KALE may facilitate more preclinical and clinical CEST applications.

The Consequentialist Scale: Translation and empirical investigation in a Greek sample

The Consequentialist Scale (Robinson, 2012) [89] assesses the endorsement of consequentialist and deontological moral beliefs. This study empirically investigated the application of the Greek translation of the Consequentialist Scale in a sample of native Greek speakers. Specifically, 415 native Greek speakers completed the questionnaire. To uncover the underlying structure of the 10 items in the Consequentialist Scale, an Exploratory Factor Analysis (EFA) was conducted. The results revealed a three-factor solution, where the deontology factor exhibited the same structure as the original work by Robinson (2012) [89], while the original consequentialism factor split into two separate factors. Significant Pearson's r correlations were observed between age and responses to the Consequentialist Scale. Separate EFAs were conducted for two age groups based on a medial split: younger (36 years old or less) and older (more than 36 years old). Interestingly, the younger group exhibited a two-factor solution with the same structure as the original work, while the older group showed a three-factor solution. A hierarchical k-means cluster analysis revealed that the cluster of participants who scored higher in deontology compared to consequentialism primarily consisted of older participants, whereas the two other clusters comprised of younger participants exhibited the reverse pattern. Neither gender nor previous experience with philosophy significantly affected scores on the Consequentialist Scale. Overall, our study provides evidence that the Consequentialist Scale is suitable for use in the Greek population.

Research on Location Selection of Mobile Communication Weak Coverage Network Based on Hierarchical K-means Clustering Algorithm

At present, the problem of weak coverage of mobile communication still exists. Firstly, this thesis determines the scale of the base station and the location of the network with the hierarchical clustering algorithm. Secondly, this thesis also determines the base station coverage with the sector area of the Archimedes screw, the location of the weak coverage point with the discriminative model of weak coverage point, and the angle of the sector with the coverage model of gradient descent method. Finally, this thesis provides better solutions to the weak coverage problem with the propagation model of cluster analysis, and optimizes the site selection with the correction model of the propagation pass loss. Within the certain cost, we aim to increase the volume of coverage business, decrease the time complexity of the model, and improve the efficiency of the site selection.

Comprehensive Dataset Building and Recognition of Isolated Handwritten Kannada Characters Using Machine Learning Models

In this work, an attempt is made to build a dataset for handwritten Kannada characters and also to recognize the isolated Kannada vowels, consonants, modifiers, and ottaksharas. The dataset is collected from 500 writers of varying ages, gender, qualification, and profession. This dataset will be used to recognize the handwritten kagunitas, ottaksharas, and other base characters, where the existing works have addressed very less on the recognition of kagunitas and ottaksharas. There are no datasets for the same. Hence, a dataset for handwritten 85 characters is built using an unsupervised machine learning technique, that is, K-means hierarchical clustering with run length count features. An accuracy of 80% was achieved with the unsupervised method. The dataset consists of 130,981 samples for 85 classes; these classes are further divided into upper, lower, and middle zones based on the position of the character in the dialect. After the dataset was built, support vector machine model with histogram of oriented gradients features was used for recognition and an accuracy of 99.0%, 88.6%, and 92.2% was obtained for the upper, middle, and lower zones, respectively, to increase the recognition rate, the convolutional neural network model is fine-tuned with raw input, and an accuracy of 100%, 96.15%, and 95.38% was obtained for the upper, middle, and lower zones, respectively. With the ResNet18 model, an accuracy of 99.88%, 98.92, and 97.55% was obtained for each of the zones, respectively. The dataset will be made available online for the researchers to carry out their research on handwritten characters, kagunitas, and word recognition with segmentation.

Wind Power Converter Fault Diagnosis Using Reduced Kernel PCA-Based BiLSTM

In this paper, we present a novel and effective fault detection and diagnosis (FDD) method for a wind energy converter (WEC) system with a nominal power of 15 KW, which is designed to significantly reduce the complexity and computation time and possibly increase the accuracy of fault diagnosis. This strategy involves three significant steps: first, a size reduction procedure is applied to the training dataset, which uses hierarchical K-means clustering and Euclidean distance schemes; second, both significantly reduced training datasets are utilized by the KPCA technique to extract and select the most sensitive and significant features; and finally, in order to distinguish between the diverse WEC system operating modes, the selected features are used to train a bidirectional long-short-term memory classifier (BiLSTM). In this study, various fault scenarios (short-circuit (SC) faults and open-circuit (OC) faults) were injected, and each scenario comprised different cases (simple, multiple, and mixed faults) on different sides and locations (generator-side converter and grid-side converter) to ensure a comprehensive and global evaluation. The obtained results show that the proposed strategy for FDD via both applied dataset size reduction methods not only improves the accuracy but also provides an efficient reduction in computation time and storage space.

Appraisal of groundwater quality for drinking and irrigation suitability using multivariate statistical approach in a rapidly developing urban area, Tirunelveli, India.

The quantitative and qualitative stress on groundwater resources has been witnessed across the globe. The current study assesses the groundwater quality of Tirunelveli district which faces the hazard of groundwater contamination through seepage of toxins, considering the open dumping ofhuge volumes of solid waste. The findings from this study confirmed the presence of more than 20% samples in the "poor to very poor" quality with high concentrations of TDS, Cl-, and NO3-, unfit for drinking, and other domestic purposes. The spatial distribution of TDS and NO3- highlighted the potentialimpact of solid waste dumping in the nearby landfill sites. K-means hierarchical clustering and multivariate analysis suggested that salinization and nitrate pollution was highly influenced by anthropogenic sources in comparison to geogenic sources. Rock water interaction and evaporation processes emerged as the two major dominant natural mechanisms controlling the groundwater chemistry. Four hydro-chemical facies were identified in the order of Ca-HCO3 > Ca-Mg-Cl > Na-Cl > mixed Ca-Na-HCO3. Thus, this study creates an urgent need of mitigation measures towards curbing and management of solid waste disposal and hence, the potential hazard ofcontaminant seepage into the groundwater.

FAIR Machine Learning Model Pipeline Implementation of COVID-19 Data

Abstract Research and development are gradually becoming data-driven and the implementation of the FAIR Guidelines (that data should be Findable, Accessible, Interoperable, and Reusable) for scientific data administration and stewardship has the potential to remarkably enhance the framework for the reuse of research data. In this way, FAIR is aiding digital transformation. The ‘FAIRification’ of data increases the interoperability and (re)usability of data, so that new and robust analytical tools, such as machine learning (ML) models, can access the data to deduce meaningful insights, extract actionable information, and identify hidden patterns. This article aims to build a FAIR ML model pipeline using the generic FAIRification workflow to make the whole ML analytics process FAIR. Accordingly, FAIR input data was modelled using a FAIR ML model. The output data from the FAIR ML model was also made FAIR. For this, a hybrid hierarchical k-means (HHK) clustering ML algorithm was applied to group the data into homogeneous subgroups and ascertain the underlying structure of the data using a Nigerian-based FAIR dataset that contains data on economic factors, healthcare facilities, and coronavirus occurrences in all the 36 states of Nigeria. The model showed that research data and the ML pipeline can be FAIRified, shared, and reused by following the proposed FAIRification workflow and implementing technical architecture.

OS08.3.A Distinct systemic and tumor microenvironment immune landscapes discriminate across sellar tumor types and controls through a methylation-based deconvolution method

Abstract Background Systemic (Sys) and tumor microenvironment (TME) immune milieus play a pivotal role in tumor development, outcome and immunotherapy response predictions across a variety of central nervous system tumors. Genome-wide methylation profiling can reliably discriminate and estimate immune cell proportions present in the blood and within the tumor and has not been reported across sellar tumor types (STT). Material and Methods We estimated cell composition in liquid biopsy (LB, serum/plasma) and tissue specimens from 42 STT collections (i.e., pituitary neuroendocrine tumors [PitNETs; n=37] and craniopharyngiomas [CP; n=5]), and 26 nontumor controls (LB: 11; Tissue: 15) using MethylCIBERSORT, a methylation-based deconvolution algorithm and established immune cell signatures as reference. LB methylation was profiled with EPIC array. Correlations between estimated cell proportions across sample sources were explored (Spearman). Immune cell proportion hierarchical k-means clustering was performed across tissue and LB specimens. Similarly, mean comparisons between and across sample types and subgroups of interest were performed [Non-parametric Kruskal-Wallis, Wilcoxon rank-sum tests; p&amp;lt;0.05]. Results We identified three immune-clusters across tissue specimens which distinguished controls (k3-cluster) from sellar tumor specimens (k1- and k2- clusters), primarily attributable to differential B-cell and monocyte proportions. Interestingly, a subset of PitNET and CP, belonging to the k2-cluster, presented a distinct immune profile compared to their K1-sellar tumor counterparts. Analysis of plasma-derived immune clusters revealed that PitNETs were distributed across four distinct immune patterns and CP clustered together with controls and a PitNET subset. One of the PitNET clusters was enriched with patients that died during follow-up and presented an enrichment of CD4-(including the regulatory subtype), CD8 and CD56-T and depletion of natural killer cells. Differences across serum- and tissue-derived clusters were present but less prominent than their plasma counterparts. No correlation between immune cell proportions across other clinicopathological features within each tumor type (sex, age, histotypes, invasion etc) was observed. Conclusion Our results suggest that PitNETs are characterized by differential TME and systemic immune subtypes which also distinguish these tumors from CP and controls. Additionally, distinct systemic immune composition between tissue and LB sources, more readily observed in plasma, suggest that the systemic response to the presence of the tumor is distinct from the immune response noted in the TME. Tumor immune subtyping may allow the stratification of STT according to immunotherapy response vulnerabilities.

A new hybrid model for power forecasting of a wind farm using spatial–temporal correlations

This paper develops a new hybrid forecasting model for the hourly power output of multiple turbines (or plants), aimed at making use of both the spatial and temporal correlations of weather factors and power outputs from wind turbines (or plants). To account for the time-varying nature of local pattern and wind propagation, a new clustering algorithm named Kmeans–Hierarchical Clustering (KHC) is first used to adaptively cluster the wind turbines (or plants). Then, singular value decomposition (SVD) is employed to extract the leading components of the wind power of the wind turbines (or plants) in the same cluster. Finally, the support vector regression (SVR) models are built for the leading components and the predictions of the components are transformed into the predictions of the power output of each turbine (or plant) with the inverse SVD. The comparison results show that the proposed model is able to improve the forecast accuracy over the existing short-term forecasting models.

Biopsy Characteristics, Subtypes, and Prognostic Features in 107 Cases of Feline Presumed Immune-Mediated Polyneuropathy.

Inflammatory polyradiculoneuropathy (IMPN) is one of the causes of sudden onset of neuromuscular signs such as para-/tetraparesis in young cats. Even though most cases have a favorable outcome, persistent deficits, relapses, and progressive courses are occasionally seen. As clinical presentation does not always appear to predict outcome and risk of recurrence, this study was initiated to screen for prognostic biopsy findings in a large cohort of histologically confirmed IMPN cases with clinical follow-up. In total, nerve and muscle specimens of 107 cats with biopsy diagnosis of presumed autoreactive inflammatory polyneuropathy and 22 control cases were reviewed by two blinded raters for a set of 36 histological parameters. To identify patterns and subtypes of IMPN, hierarchical k-means clustering of 33 histologic variables was performed. Then, the impact of histological parameters on IMPN outcome was evaluated via an univariate analysis to identify variables for the final multivariate model. The data on immediate outcome and follow-up were collected from submitting neurologists using a purpose-designed questionnaire. Hierarchical k-means clustering sorted the tissues into 4 main categories: cluster 1 (44/129) represents a purely inflammatory IMPN picture, whereas cluster 2 (47/129) was accompanied by demyelinating features and cluster 3 (16/129) by Wallerian degeneration. Cluster 4 (22/129) reflects normal tissues from non-neuropathic control cats. Returned questionnaires provided detailed information on outcome in 63 animals. They were categorized into recovered and non-recovered. Thereby, fiber-invasive infiltrates by mononuclear cells and mild fiber loss in intramuscular nerve branches correlated with higher probabilities of recovery. Remyelination in semithin sections, on the other hand, is correlated with a less favorable outcome. Animals grouping in cluster 1 had a tendency to a higher probability of recovery compared to other clusters. In conclusion, diagnosis of feline IMPN from nerve and muscle biopsies allowed for the identification of histologic features that were positively or negatively correlated with outcome.

Exploration of Topic Classification in the Tourism Field with Text Mining Technology—A Case Study of the Academic Journal Papers

This study collects abstracts of SSCI tourism journal papers between 2010 and 2019 from the WoS (Web of Science) database and uses a novel method of topic classification to explore the vocabulary characteristics of the classified articles. The corpora of abstracts are given quantitative Term Frequency–Inverse Document Frequency (TF–IDF) weights. A hierarchical K-means cluster analysis is then performed to automatically classify the articles; co-word analysis techniques are used to show the characteristics of feature words for distinct clusters, titles, and the consistency of the classified articles. Based on the results for 5783 abstracts, cluster analysis classifies the number of K-means clusters into six categories: travel, culture, sustainability, model, behavior, and hotel. A cross-check method is applied to assess the consistency of the topic classifications, list titles and keywords of the documents with the three smallest distances in each category and apply a strategic diagram to present the features of the distinct categories.

Spatio-temporal characterization of microbial heat production on undisturbed soil samples combining infrared thermography and zymography

Passive infrared thermography (IRT) has already been applied in several approaches for high-resolution and non-contact imaging of microbial hot spots and hot moments on soil sample surfaces. The technique has only been used on homogenized disturbed samples to characterize the in-situ heterogeneity of heat development. In this study, undisturbed top- and subsoils samples from two forest sites were used in a substrate-induced approach to capture surface heat production using passive IRT with homogeneously applied glucose and water. The soil sample surface temperature was measured at 10-minute intervals and a spatial resolution of 0.17 mm per pixel. The soil samples were incubated for five days during passive IRT measurements under controlled ambient conditions with a relative air humidity >95% and constant ambient air temperature of 20 °C. Soil sample surface characterization was done by using active IRT for soil moisture approximation and surface structure, digital photography to estimate soil organic carbon (SOC) contents from soil color parameters, and zymography to get an indicator of initial microbial activity. In a first step, surface temperature dynamics were characterized using mathematical and geostatistical methods concerning microbial hot spots and hot moments. The characterization of hot moments was performed using a Gaussian curve fit. The spatial information of the hot spots was described using geostatistical semivariance. In a second step, a stepwise forward regression using sample surface properties was performed to find explanatory variables for surface heat production. Finally, a hierarchical k-means clustering was applied to the sequential thermal images and transferred to the other spatial datasets for deeper insights into small-scale variations in heat production and to also consider the temporal perspective. With an ANOVA combined with Tukey’s HSD post hoc test, significant differences between the cluster groups were calculated. This study showed that the temperature difference between averaged glucose- and water-treated sample surface temperature increased up to 0.2 K with a maximum hot moment duration of nearly 46 h. The variance in the samples could not be mapped, since the maximum spatial variability of the sample surface temperature could not be represented in the sample sections. The regression analysis revealed significant influences by soil moisture content, approximated SOC content, as well as initial β-glucosidase activity in the subsoil, which could be confirmed by the k-means cluster analysis. There was an indication that soil brightness (SOC contents) was influencing heat production due to the varying soil-borne substrate availability. The initial β-glucosidase activity showed that the initial biotic state of the soil contributed to the temperature increase concerning the provision of energy for microbial activity. In conclusion, passive IRT is a helpful mapping tool for the spatio-temporal characterization of hot spots and hot moments. In this context, additional mapping techniques such as zymography help to gain insights into the process level of heat generation by microbial activity.

Identification of Strata from Irregularities in Well Logs Using Chaos Quantifiers

Summary Estimating a transition between rock types from well logs using traditional methods can be challenging and time-consuming. Developing new approaches to improving the quality of the estimation as well as saving time becomes necessary. This paper presents a new methodology that uses elements of chaos theory to evaluate the variability of well logs to identify rock layers. Four different parameters that quantify chaos were used in the present study: fractal correlation dimension, sample entropy, Hurst exponent, and Lyapunov exponent. Each of them describes a different property of a well log. The method presented in this paper uses all of them together for an extensive characterization of well log irregularities. The study was carried out on a set of 68 well logs from six wells in the Pluto gas field (Australia). The logs were divided into segments of 25 m. A computer program was written to calculate the chaos parameter values of each interval. The parameters were then analyzed statistically. Hierarchical methods and k-means clustering were used to create dendrograms and clusters. The statistical analysis of the results has shown that the well log variability can be used to successfully differentiate rock formations by showing which intervals on a log are similar. In addition, the intervals that correspond to Mungaroo sandstones, which are the reservoir rock of the Pluto gas field, were particularly distinguished from other parts of the log. Therefore, the presented methodology could prove useful to estimate zones of interest in terms of hydrocarbon potential. The presented algorithm accounts for the variability of the well log readings, not the log values themselves. It does not point exactly to a depth where rock layers interface, but it rather allows similar (in terms of irregularities), consecutive intervals to be grouped together. Based on that, one can draw a conclusion that a lithology differs between groups of intervals.

The genetic landscape of South African males: A Y-STR perspective

The profiles of 2188 SA males obtained with 10 Y-STR highly informative markers were analyzed for their information content for forensic and population studies. The samples comprised a total of 16 populations, represented by Bantu-speaking groups, KhoeSan descendants, out-of-Africa descendants and admixed groups. The country hosts approximately 58 million inhabitants, 80% native and the remaining with ancestry external to Africa and admixed. The forensic parameters indicated high levels of diversity in all populations and lower in the Nguni, who showed elevated number of repeated haplotypes, thus displaying the lowest DC values. Population comparative analysis with MDS showed concordant results with the historical record. Non-hierarchical and hierarchical AMOVA over ethnolinguistic groups and administrative divisions showed significant variation in all cases, with higher differences due to ethnicity than to geopolitical subdivision. The haplotypes were further analyzed by hierarchical kmeans clustering. The identified clusters differed in their relative contribution to the gene pool of the 16 analyzed populations. Geostatistical analysis of the clusters evidenced areas of higher density for some clusters in correspondence with language, while other clusters were more homogeneously distributed. In addition, a few rare microvariants were identified with very restricted geographic occurrence. The results emphasize the forensic value of a highly informative set of markers in a country with high genetic diversity and complex population history.

Emotion regulation in times of COVID-19: A person-centered approach based on self-determination theory.

Although the COVID-19 crisis is a worldwide threat to individuals’ physical health and psychological well-being, not all people are equally susceptible to increased ill-being. One potentially important factor in individuals’ vulnerability (versus resilience) to ill-being in the face of stress is emotion regulation. On the basis of Self-Determination Theory, this study examined the role of three emotion regulation styles in individuals’ mental health during the COVID-19 crisis, that is, integration, suppression, and dysregulation. Participants were 6584 adults (77% female, Mage = 45.16 years) who filled out well-validated measures of emotion regulation, depression, anxiety, life satisfaction, and sleep quality. To examine naturally occurring combinations of emotion regulation strategies, hierarchical k-means clustering was performed, yielding 3 profiles: (a) low scores on all strategies (indicating rather low overall levels of worry; 27%), (b) high scores on integration only (41%), and (c) high scores on suppression and dysregulation (32%). Participants in the profiles scoring high on suppression and dysregulation displayed a less favorable pattern of outcomes (high ill-being, low life satisfaction, and poorer sleep quality) compared to the other two groups. Between-cluster differences remained significant even when taking into account the corona-related worries experienced by people. Overall, the findings underscore the important role of emotion regulation in individuals’ mental health during mentally challenging periods such as the COVID-19 crisis. Practical implications and directions for future research are discussed.

Segmentation of Natural Images with K-Means and Hierarchical Algorithm based on Mixture of Pearson Distributions

In this paper, an attempt has been made to analyze the performance of the image segmented algorithms with the addition of the Pearsonian Type III mixture model. By using the Type III Pearsonian system of distributions the image segmentation process was carried out in the current article which is a novel technique. With the help of K-component combination of Pearsonian Type III distribution, it is considered that the whole input images are characterized. The performance parameters PRI (Probabilistic Rand Index), GCE (Global Consistency Error) and VOI (Volume of Interest) for the currently considered model are estimated with the help of EM (Expectation Maximization) algorithm. For analyzing the proposed model’s performance, four random images are selected as input for the current model from Berkeley image database. The performance metric parameters PRI, GCE and VOI values given the results as the currently proposed method is providing more précise results for the input images where the regions of the input images selected are with tiles having long upper model and the left skewed images. By the help of image quality measures, the proposed method is performing well for the purpose of retrieving the images with respect to the picture segmenting process which is based on GMM (Gaussian Mixture Model). The current model performance was compared with the other existing models like the k-means hierarchical clustering model and the 3-paprameter regression models.

Reduced Gaussian process regression based random forest approach for fault diagnosis of wind energy conversion systems

This paper proposes a novel Reduced Gaussian Process Regression (RGPR)-based Random Forest (RF) technique (RGPR-RF) for fault detection and diagnosis (FDD) of wind energy conversion (WEC) systems. First, two RGPR models are proposed to deal with WEC features extraction and selection. The proposed RGPR models extract the most relevant information from the WEC system data while reducing the computation burden compared to the classical GPR model. The complexity reduction is ensured by the selection of the most effective samples through the dimensionality reduction (DR) metrics including Hierarchical K-means (HKmeans) clustering and Euclidean distance (ED). Next, in order to classify the WEC faults and improve the diagnosis abilities, RF classifier is developed. The proposed RGPR H K m e a n s -RF and RGPR E D -RF techniques boost the classification speed and accuracy using a reduced number of features where only the most relevant and sensitive characteristics are kept in case of redundancy. The open-circuit, wear-out, and short-circuit are the three transistor faults considered in order to illustrate the effectiveness and robustness of the developed techniques. The obtained results show that the proposed RGPR-RF technique is characterized by a low computation time and high diagnosis accuracy (an average accuracy of 99.9%) compared to the conventional RF classifiers.

Hierarchical K-means clustering for registration of multi-view point sets

As a long-standing research issue in computer vision and robotics, multi-view registration has attracted much attention in recent years. Most existing works are mainly focus on the estimating the point to point match correspondence, which usually suffers from the poor initial pose and data noise as well as leads to the inaccurate matches. To overcome the aforementioned limitation, we propose a novel Hierarchical K-means Clustering Registration (HKCR), which casts the multi-view registration as a hierarchical clustering task. Specifically, the proposed method employs a small number of clusters firstly, then increases the number of clusters during the registration process. Benefiting from the recursive partitioning process, more robust and more accurate results can be achieved with the increasing finer granularity. To show the effectiveness and robustness of HKCR, extensive experiments are conducted on several benchmark datasets and compared to several state-of-the-art methods.

On the application of machine learning for defect detection in L-PBF additive manufacturing

This paper investigates the performance of several Machine Learning (ML) techniques for online defect detection in the Laser Powder Bed Fusion (L- PBF) process. The research aims to improve the consistency in product quality and process reliability. The applications of acoustic emission (AE) sensor to receive elastic waves during the printing process is a cost-effective way of materializing such a demand. In this study, the process parameters were intentionally adjusted to create three different levels of defects in H13 tool steel samples. The first class was printed with minimum defects, the second class had only intentional cracks, and the last class included both intentional cracks and porosities. The AE signals were acquired during the samples' manufacturing, and three different machine learning (ML) techniques were applied to analyze and interpret the dataset. First, a hierarchical K-means clustering is employed for labeling the data, followed by a supervised deep learning neural network (DL) to match acoustic signal with defect type. Second, a principal component analysis (PCA) was used to reduce the dimensionality of the dataset. A Gaussian Mixture Model (GMM) was then employed to enable fast defect detection suitable for online monitoring. Third, a variational auto-encoder (VAE) approach is used to obtain a general feature of the signal that could be an input for the classifier. A supervised DL trained on the H13 tool steel dataset successfully detected quality trends in AE signals collected from 316L samples. The VAE approach presents a novel method to detect defects in L-PBF processes that eliminate the need for model training in different materials.