Seong-Pyo Cheon

Bayesian networks based rare event prediction with sensor data

A Bayesian network is a powerful graphical model. It is advantageous for real-world data analysis and finding relations among variables. Knowledge presentation and rule generation, based on a Bayesian approach, have been studied and reported in many research papers across various fields. Since a Bayesian network has both causal and probabilistic semantics, it is regarded as an ideal representation to combine background knowledge and real data. Rare event predictions have been performed using several methods, but remain a challenge. We design and implement a Bayesian network model to forecast daily ozone states. We evaluate the proposed Bayesian network model, comparing it to traditional decision tree models, to examine its utility.

Measure of certainty with fuzzy entropy function

To measure the certainty, we use the meaning of entropy. For the selection of reliable data, fuzzy entropy through distance measure is proposed. The appropriateness of the proposed entropy is verified by the definition of entropy measure. To measure the fuzziness of 3-phase stator currents, membership functions are obtained by the Bootstrap method. Finally, the proposed entropy is applied to the membership function of 3-phase currents, and the fuzzy entropy values of phase current each are illustrated.

Reliable data selection with fuzzy entropy

In this paper, the selection of a data set from a universal set is carried out using a fuzzy entropy function. According to the definition of fuzzy entropy, the fuzzy entropy function is proposed and that function is proved through definitions. The proposed fuzzy entropy function calculates the certainty or uncertainty value of a data set; hence we can choose the data set that satisfies certain bounds or references. Therefore a reliable data set can be obtained using the proposed fuzzy entropy function. With a simple example we verify that the proposed fuzzy entropy function selects the reliable data set.

On-line fabric-defects detection based on wavelet analysis

This paper introduces a vision-based on-line fabric inspection methodology for woven textile fabrics. The current procedure for the determination of fabric defects in the textile industry is performed by humans in the off-line stage. The proposed inspection system consists of hardware and software components. The hardware components consist of CCD array camera, a frame grabber, and appropriate illumination. The software routines capitalize on vertical and horizontal scanning algorithms to reduce the 2-D image into a stream of 1-D data. Next, wavelet transform is used to extract features that are characteristic of a particular defect. The signal-to-noise ratio (SNR) calculation based on the results of the wavelet transform is performed to measure any defects. Defect detection is carried out by employing SNR and scanning methods. Learning routines are called upon to optimize the wavelet coefficients. Test results from different types of defect and different styles of fabric demonstrate the effectiveness of the proposed inspection system.

Learning Bayesian networks based diagnosis system for wastewater treatment process with sensor data

Contemporary technical capabilities allow an operator to easily monitor and control several remote wastewater treatment processes simultaneously but an on-line automatic diagnostic system has not yet been installed. In this paper, an on-line diagnostic system is proposed, designed and implemented for the lab-scale five-stage step-feed Enhanced Biological Phosphorus Removal plant based upon a learning Bayesian network. In order to practically diagnose wastewater treatment processes, a lab-scale pilot plant was built and the proposed on-line diagnostic method was applied to evaluate the performance of the algorithm. In experimental results, real abnormal conditions occurred 21 times in a three month period. The suggested on-line diagnosis system made correct predictions 14 times and incorrect predictions 7 times. Moreover, a comparison of the prediction results of the Bayesian model and learning Bayesian model clearly show that learning algorithm became more effective as time passed.

Performance Criterion-based Polynomial Calibration Model for Laser Scan Camera

The goal of image calibration is to find a relation between image and world coordinates. Conventional image calibration uses physical camera model that is able to reflect camera`s optical properties between image and world coordinates. In this paper, we try to calibrate images distortion using performance criterion-based polynomial model which assumes that the relation between image and world coordinates can be identified by polynomial equation and its order and parameters are able to be estimated with image and object coordinate values and performance criterion. In order to overcome existing limitations of the conventional image calibration model, namely, over-fitting feature, the performance criterion-based polynomial model is proposed. The efficiency of proposed method can be verified with 2D images that were taken by laser scan camera.

On-line diagnosis system with Bayesian networks for WWTP

Nowadays, due to development of automatic control devices and various sensors, one operator can freely handle several remote plants and processes. Automatic diagnosis and warning systems have been adopted in various fields, in order to prepare an operatorpsilas absence for patrolling plants. In this paper, a Bayesian networks based on-line diagnosis system is proposed for a wastewater treatment process. Especially, the suggested system is included learning structure, which can continuously update conditional probabilities in the networks. To evaluate performance of proposed model, we made a lab-scale five-stage step-feed enhanced biological phosphorous removal process plant and applied on-line diagnosis system to this plant in the summer.

Directed knowledge discovery methodology for the prediction of ozone concentration

Data mining is the exploration and analysis, by automatic or semiautomatic means, of large quantities of data in order to discover meaningful patterns and rules. Data mining consists of several tasks and each task uses a variety of methodologies. Some of these tasks are suited for a top-down method called hypothesis testing and others are suited for a bottom-up method called knowledge discovery. In this paper, we report our research procedures and results that concern and relate ozone concentration data in various factors and attributes. We use the general steps of directed knowledge discovery methodologies and intelligent modeling techniques. Next, we construct ozone concentration prediction system in order to reduce various adverse effects on human beings and life on the earth.

Modified Transformation and Evaluation for High Concentration Ozone Predictions

To reduce damage from high concentration ozone in the air, we have researched how to predict high concentration ozone before it occurs. High concentration ozone is a rare event and its reaction mechanism has nonlinearities and complexities. In this paper, we have tried to apply and consider as many methods as we could. We clustered the data using the fuzzy c-mean method and took a rejection sampling to fill in the missing and abnormal data. Next, correlations of the input component and output ozone concentration were calculated to transform more correlated components by modified log transformation. Then, we made the prediction models using Dynamic Polynomial Neural Networks. To select the optimal model, we adopted a minimum bias criterion. Finally, to evaluate suggested models, we compared the two models. One model was trained and tested by the transformed data and the other was not. We concluded that the modified transformation effected good to ideal performance In some evaluations. In particular, the data were related to seasonal characteristics or its variation trends.

Defect Analysis of the SBR Wastewater Treatment Plant for Unmanned Automation Based on Time-series Data Mining

This paper describes how to diagnose SBR plant equipment using time-series data mining. It shows the equipment diagnostics based upon vibration signals that are acquired front each device lot process control. Data transform techniques including two data preprocessing skills and data mining methods were employed in the data analysis. The proposed method is not only suitable for SBR equipment, but is also suitable for other Industrial devices. The experimental results performed on a lab-scale SBR plant show a good equip-ment-management performance.