Papers by Keyword: Classification

Abstract: Crackles is one of the types of adventitious lung sound heard in patients with interstitial pulmonary fibrosis or cystic fibrosis. Pulmonary crackles of discontinuous short duration appear on inspiration, expiration, or both. To differentiate these pulmonary crackles, the medical staff usually uses a manual method, called auscultation. Various methods were developed to recognize pulmonary crackles and distinguish them from normal pulmonary sounds to be applied in digital signal processing technology. This paper demonstrates a feature extraction method to classify pulmonary crackle and normal lung sounds using Support Vector Machine (SVM) method using several kernels by performing spectrograms of the pulmonary sound to generate the frequency profile. Spectrograms with various resolutions and 3-fold cross-validation were used to divide the training data and the test data in the testing process. The resulting accuracy ranges from 81.4% - 100%. More accuracy values of 100% are generated by a feature extraction in several SVM kernels using 256 points FFT with three variations of windowing parameters compared to 512 points, where the best accuracy of 100% was produced by STFT-SVM method. This method has a potential to be used in the classification of other biomedical signals. The advantages of that are that the number of features produced is the same as the N-point FFT used for any signal length, the flexibility in the STFT parameters changes, such as the type of window and the window's length. In this study, only the Keiser window was tested with specific parameters. Exploration with different window types with various parameters is fascinating to do in further research.

Abstract: Polymer-based sealing materials currently play a large and important role in almost all areas of industry. The use of sealants can increase the energy saving of building structures, the safety of vehicles, etc. Sealants are equally used in such specialized applications as the aerospace field, and are also used to solve everyday problems, such as sealing plumbing. The main consumption of sealants is in construction and automotive. This review provides data on the global and domestic market for sealing materials; The basic terms and definitions used in the field of adhesive and sealing materials are analyzed. The main differences between the terms adhesive and sealant are discussed. The classification of sealants is given in accordance with modern international regulatory documents. It is disassembled that they are one-component reactive hot-melt sealants. Their main advantages and disadvantages are listed. It is noted that the development of the industry of sealing materials is likely to move towards the development of moisture-curing compositions based on silane-modified polymers. This assumption is supported by a fairly detailed classification of silane-modified polymers in the FEICA classification guide for sealants and adhesives.

Abstract: Tapioca starch adulterated with dolomite is sold in markets, but this adulteration cannot be identified by normal visual inspection. Near infrared (NIR) hyperspectral imaging has been successfully used as a non-destructive method of identifying various characteristics of food, therefore it was tested to identify dolomite adulteration. Adulterated tapioca starch samples were prepared by adding dolomite in the range of 0.5-100% (wt/wt). Samples (N=400) of pure tapioca starch (0) and adulterated tapioca starch (1) were divided into calibration set (N=300) and a prediction set (N=100). All samples were scanned using NIR hyperspectral imaging (935-1720 nm) and spectra were pre-processed using Savitzky-Golay first derivative differentiation pretreatment in order to obtain the optimal conditions for establishing a classification model. Partial least squares-discriminant analysis was carried out to evaluate the accuracy of classification tapioca starch adulterated with dolomite. The results showed the total accuracy of prediction for classification was 100%. Therefore, NIR hyperspectral imaging was demonstrated to have a potential for use in detecting adulteration of tapioca starch with dolomite.

Abstract: Multi-channel defect detection in epitaxial SiC layers is used to detect and classify various extended defects. The effects of the killer defect category are analyzed on hundreds of production diode and MOSFET wafers. Invariably, the number of these killer defects sets the yield entitlement for each wafer, and electrically, they fall into a few expected failure bins depending on their position on the die. A variety of non-killer defects like Bar Stacking Faults (BSFs), V-type defects, isolated micro-pipe related bumps, stacking faults, scratches and deep pits are identified in over thousand die each, and their effect on fully fabricated diode and MOSFET products are determined at wafer sort. A very high percentage of the die containing these non-killer defects pass all electrical tests including the rigorous Unclamped Inductive Switching (UIS) testing. Specifically, a population of electrically passing die, containing BSFs, are identified and packaged for High Temperature Reverse Bias (HTRB) tests. Every BSF containing die passes both forward and reverse 1000-hour HTRB drift tests.

Abstract: The term “roofing slate” comprises all natural rocks that can be used for roofing. Some of the characteristics that define their performance as construction materials are water absorption, bending strength, mineral composition, and aesthetics. From a sustainability point of view, roofing slates also have the largest life expectancy and lowest carbon emissions of all covering materials. This work reviews the constructive characteristics of roofing slates, from technical data provided by producers from all over the world. These data, supported by the petrological characterization of several roofing slate samples, are used to propose a general classification for this special type of rock. Thus, roofing slates are sorted in twelve lithotypes, depending of their petrographical characteristics, defined by the color. The resulting classification is named International Roofing Slate Classification (IRSC), formally defined in this paper.

Abstract: Though India being home of one out of every six people in the globe, is facing an arduous task of providing healthcare service, especially to the large number of patients in remote areas due to lack of diagnosis support systems and doctors. It is reported that hospitals in rural areas have an insufficient radiologist due to which thousands of cases are usually handled by single doctor. In this context, we aim to develop an AI based computer-aided diagnosis tool, which can classify abnormalities by reading chest X-ray so that it could assist the doctors in arriving at quick diagnosis. We have employed a Convolutional Neural Network (CNN) designed by Google known as XceptionNet to detect those pathologies in ChestX-ray14 data. Further, same data is being used for executing other CNN- ResNet. Finally, both the results obtained are compared to assess the superior CNN model for X-ray level diagnosis.

Abstract: The creation of self-compacting concrete (SCC) is associated with the introduction of plasticizing additives based on polycarboxylate esters (PCE). The first patent for a group of substances proposed for use as superplasticizers for concrete, was declared in the early 80-ies of the last century in Japan. In the mid 90-ies superplasticizers based on esters of polycarboxylates began to be used in Europe. In Russia, the use of self-compacting concrete began much later, and in bridge and tunnel construction almost a few years ago. Currently, in the European Union, 70-80% of reinforced concrete structures are made or erected from self-compacting concrete mixtures, and concrete is commonly called self-compacting. In Russia, such concretes are sometimes called self-leveling. The volume of their use is still small.

Abstract: Inline metrology tools are widely used to detect defects in SiC epitaxial layers. The defect statistics are used in a variety of ways to determine quality, pass/fail and screen affected die. In this work, we document the automated detection and classification of various epitaxial defects based on type and origin. We further classify these categories into killer and non-killer defects and compare them to the electrical yield of Schottky Diodes. The origins of these defects are determined in broad categories, resulting in a clustering and yield-scaling model, which agrees very closely to experimental data. Further, we look at on-wafer screening techniques of potential weak die by both defect tagging and unclamped inductive switching (UIS) stress testing. Successful 1000-hr reliability tests show the robustness of our detection and screening methods.

Abstract: Interfaces between two fluid phases are a potential barrier for particles. Certain particles may not be able to pass such an interface, because they have to overcome a certain resistance. The latter depends on the strength of the interface, which is the surface tension. The second relevant property is the three phase wetting angle, which shows the fluid with the preferred wetting to the particle surface. It depends on the particle properties, like chemical composition, surface structure and surface modification. The third relevant parameter is the particle size. From these three main influence parameters it emerges that fluid-fluid interfaces can show a selectivity to special particle properties, which enables a separation of a particle mixture. Since there are possibilities to address the governing effects, the separation cut, size or composition cut respectively, can be engineered in a certain range. Separation at boundaries is feasible when the driving force is in the same order of magnitude as the retaining resistance force of the interface. The driving force is either the Brownian movement for very small particles or any field force like gravity or the centrifugal force. To describe the separation at interfaces it is necessary to understand the process of the phase transfer of particles through the interface, either the gas-liquid or the liquid-liquid interface between two immiscible liquids. In addition to the effects mentioned above, also dynamic phenomena such as surfactant depletion of the interface may have to be taken into account.

Abstract: In the photovoltaic industry a total of 100,000 tons of silicon is lost as waste per year. This waste is originating from several cropping and sawing steps of the high purity silicon blocks and ingots during the solar cell wafer production, resulting in a silicon containing suspension. Among different approaches to recycle the silicon from this waste is the utilization of hydrocyclones, which can be used to separate or classify particles by weight and size. In this work the use of a hydrocyclone was evaluated to upgrade the silicon fraction from a typical sawing waste. A potential field of use for the recycled silicon particles might be as anode material for next generation lithium ion batteries.