Papers by Keyword: Clustering

Abstract: The paper presents the results of the study of registered acoustic emission (AE) parameters during static deformation and damaging of polymer composite materials (PCM). Mechanical tests were done by a static tension and a static three-point bend, accompanied by an acoustic emission method. The assessment of the loading rate effect on defects formation processes was done by additional static tension test at rate equal half of recommended by the standard and static three-point bend test at rate ten times lower than that calculated by the standard. Clustering by frequency components of the recorded AE signals with a self-organizing Kohonen map was performed. The characteristics of the types of PCM structure damage by the centroids of the obtained clusters are given. Based on the clusters accumulation during mechanical tests, the stages of damage formation for static tension and static three-point bend, the loading rate effect on the process of damage formation are described.

Abstract: Assembly lines consist of chained or unchained stations, yet usually only single stations are regarded individually for process and quality analytics. Since the quality of the final product depends on interactions of process parameters along the assembly flow, it is insufficient to analyze process parameters of each station separately. Therefore, data of every single assembly station along the assembly line has to be collected and stored. To explore such a big amount of multidimensional data and their correlations, different techniques are established. In this paper, assembly flows and their respective data are visualized using a parallel coordinates plot (PCP). Here, this technique visualizes process parameter combinations along the whole assembly chain. The contribution of this paper is to prove that the presented approach enables a fast detection of stations with malicious impacts on the product quality, when it comes to complex assembly lines. The goal is to help users to detect global problems in those lines, not only single station problems. Furthermore, the relevance of various processes to the quality (good or defective) of the final good shall be revealed.

Abstract: Conventional serial and workshop productions use specific parameter ranges to evaluate the quality of a process. Our research showed that parameters within tolerances do not ensure good quality of the final product due to malicious parameter combinations along the assembly line. Therefore, data sets from assembly processes like force-way or force-time curves and quality measurements are evaluated in this novel approach. Using Fourier Transform, k-means, decision trees and a dynamic envelope curve, classification and process monitoring are processed in time and frequency domain. This enables new possibilities to characterize quality and process data, for advanced error detection as well as a more simplified tracing of faults. Here, holistic optimization and monitoring follows two strategies. First, a simplified tracing approach of malicious impacts regards quality results from test benches. Therefore, assembly processes are monitored and characterized by quality data. Second, defective influences, like tool break or calibration errors, are linked to variations of the usual process behavior. Here, the error detection approach focuses on process data from single assembly stations. This approach uses three different methods. First, Fourier Transform extracts additional information from process, energy and quality data. Second, k-means algorithm is used to cluster quality data and extend the data base. Third, a decision tree classifies the quality of the final good and characterizes assembly processes. Last, results of k-means clustering and selected classification methods are compared. This combination allows to increase process quality, improve product quality and reduce failure costs.

Abstract: Extraction of geometric information and reconstruction of a parametric model from the data points captured by various sensors or generated by various image preprocessing algorithms is a vital research issue for many computer vision and robotics applications. The aim is to reconstruct 3D objects, consisting of planar patches, in a scene from its point cloud captured by a sensor set. A reconstructed scene has many applications such as stereo vision, robot navigation, medical imaging, etc. Unfortunately, the captured point cloud often gets corrupted due to sensor errors/malfunctioning and preprocessing algorithms. The corrupted data pose difficulty in accurate estimation of underlying geometric model parameters. In this paper, a new algorithm has been proposed to efficiently and accurately estimate the model parameters in heavily corrupted data points. The method is based on forming clusters of estimated planes with reference to a fixed plane. Clustering is accomplished on the basis of angles and distances of estimated planes from the reference plane. The proposed method is implemented over a wide range of data points. It is a robust technique and observed to outperform the widely used RANSAC algorithm in terms of accuracy and computational efficiency.

Abstract: In this paper, a new filter network is presented that is based on Radial Base Function Networks (RBFNs). The output layer of the network is modified, in order to make it more effective in certain fuzzy control systems. The training of the network is solved by a clustering step, for which two different clustering methods are proposed. The suggested structure can efficiently be used for data classification.

Abstract: In this paper, we represent approaches for detecting users’ POIs and identifying personal routes based on temporal smartphone sensor data, including GPS. POIs and routes of users are factors that affect prediction of a traveling route. However, recording user destinations and the routes of training data is almost impossible when building a route model. Thus, we propose algorithms that automatically extract the points of destinations and routes using GPS.

Abstract: In this paper, a novel approach was proposed for the topic detection which combined the multi-models. We paid attention to the content similarity, time similarity and location similarity respectively, at the same time, the Bayesian model also was researched and the atomic characteristics words were extracted. Combined the expert knowledge and multi-models, the experiment was completed and the experimental results show that the approach is effective.

Abstract: Focus on the character of energy harvesting sensor network in heterogeneous sensor network and some shortage in SEP algorithm, an improved algorithm for EH-SEP is been proposed. EH-SEP considers both residual energy and energy support of nodes in cluster-head election process .Improved algorithm achieves higher probability that the advanced nodes with high residual energy to be cluster-head, and lower probability that the traditional nodes with low residual energy to be cluster-head. During the state of data sensing, this paper adopted multiple hop data transmission to avoid long distance communication between the cluster head and base station, so it can improve the network energy utilization. The simulation result shows that: EH-SEP algorithm is not only suitable for energy harvesting of wireless sensor network, but also effectively prolong the work time in the network stable stage.

Abstract: The algorithm for clustering based on neural network modeling using T. Kohonen's self-organizing maps for the analysis of the housing stock is considered. This analysis of housing stock is required for the planning of complex reproduction of housing and major repairs regional programs development. The mechanism of self-organization is submitted. The representative sample clustering of the housing stock is produced. Its result is 16 groups of objects with a high level of internal similarity. The basic advantages of this approach for monitoring and analysis of the city housing stock are described.

Abstract: On the basis of six kinds of air pollutant data provided by the Xi’an Environment Protection Bureau, Two kinds of air assessment model were presented in this paper. Firstly, air quality index (AQI), which has been adopted as a part of national standard in China, was used to assess the air quality of Xi’an in 2013 winter. We also introduced a fuzzy self-organizing feature map (SOFM) model to classify air quality in an unsupervised and comprehensive way. As a further research, a hybrid model was put forward based on Evolutionary Strategy (ES) and SOFM. With SOFM neural networks embedded into ES, the sensitivity of SOFM neural networks to the initial weight matrix and sequence of exemplar input is overcome by the global optimization of ES. The results of our work demonstrate that the ES-SOFM Hybrid Model is quite appropriate techniques for air quality assessment. Unlike AQI method, SOFM’s result is decided by all pollutant instead of only the most serious one. No matter what kind of method, all assessment results show the very serious air pollution in Xi’an. The government and every citizen must take steps at once to prevent air quality from further depravation.