• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.07a
• /
• pp.65-66
• /
• 2021

본 논문에서는 Smart Factory의 자동 공정에서 결함의 분류를 실시간으로 시도하여 자동 공정 제어를 위한 결함 분류 딥러닝 기법을 제안하고, Pooling 종류에 따른 분류 성능을 비교한다. Smart Factory 구축에 있어서 CNN을 이용한 공정 제어를 통해 제품 생산에 있어서 생산량의 증가와 불량률의 감소를 이루어내는 것이 가능하다. Smart Factory는 자동화 공정이므로 결함의 분류 속도가 중요하지만, 생산량의 증가와 불량률의 감소를 위해서는 정확하게 결함의 종류를 분류하여 Smart Factory의 공정을 제어하는 것이 더욱 중요하다. 본 논문에서는 Pooling을 Max Pooling과 Averrage Pooling을 복합적으로 설정하였을 때 높은 성능을 보였다.

• Journal of Korea Society of Industrial Information Systems
• /
• v.14 no.4
• /
• pp.116-127
• /
• 2009

This research aims to improve the efficiency of production by selecting production configuration with high yield rate and lower error rate based on production history of steel pipe. To achieve this, we identify the properties of various types of MTO(make-to-order) steel pipe products and determine properties affecting yield rate and error rate using decision tree technique. From experimental results, we find out that specification is critical to determine yield rate and error rate of ERW steel pipes with mostly small and medium caliber, and an external diameter range in case of roll benders or spiral steel pipes with mostly large caliber. This research classified and embodied the patterns of yield rate and error rate mathematically by product properties.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.8 no.4
• /
• pp.930-937
• /
• 2007

In this paper, we address a design problem and a case study of a K-stage inspection system, which is composed of K stages, each of which includes an inspection process and a rework process. Assuming the type I and II errors of inspectors and the re-inspection policy for items classified as good, we determine the smallest integer of K which can achieve a given target defective rate. If K does not exist, holding the current values of the type I, II errors, we search reversely a new vector, (the defective rate of an assembly line, the defective rate of a rework process), which can give the target defective rate. Our formulas and methodology based on our K-stage inspection system could be applied and extended to similar situations with slight modifications.

• Proceedings of the Korea Society for Industrial Systems Conference
• /
• 2009.05a
• /
• pp.144-149
• /
• 2009

본 논문은 스틸 파이프 생산시 저불량율, 고수율을 가진 제품을 분류하고 모형화하기 위하여 과거 스틸 파이프 생산이력을 비교, 분석하여 주요 특성들이 불량율, 수율에 어떠한 영향을 미치는가를 파악함으로써 향후 스틸 파이프 생산 공정에서 저불량율, 고수율의 제품을 생산하는데 주요한 지표로 활용코자 하는데 그 목적이 있다. 과거 스틸 파이프 산업에 대한 주요 특성별 수율 측정에 대한 연구가 미흡하였으나, 본 논문에서 이를 구체화 하여 주요 특성별 불량률과 수율이 어떠한 형태를 나타내는지를 분류하고, 그 영향정도를 구분하고자 한다.

• Journal of Advanced Navigation Technology
• /
• v.26 no.2
• /
• pp.113-118
• /
• 2022

Motivated by people highlighting important phrases while reading or taking notes we propose a neural network training method by highlighting defective pixel areas to classify effectively defect types of images with complex background textures. To verify our proposed method we apply it to the problem of classifying the defect types of tire band fabric images that are too difficult to classify. In addition we propose a backlight highlighting technique which is tailored to the tire band fabric images. Backlight highlighting images can be generated by using both the GradCAM and simple image processing. In our experiment we demonstrated that the proposed highlighting method outperforms the traditional method in the view points of both classification accuracy and training speed. It achieved up to 13.4% accuracy improvement compared to the conventional method. We also showed that the backlight highlighting technique tailored for highlighting tire band fabric images is superior to a contour highlighting technique in terms of accuracy.

• Proceedings of the Korean Information Science Society Conference
• /
• 2008.06c
• /
• pp.482-486
• /
• 2008

본 논문에서는 하드 디스크 드라이브(Hard Disk Drive, HDD) 생산 공정 과정에서 발생할 수 있는 불량 HDD의 결함 분포에 대해서 패턴을 자동으로 분류해주는 기법을 제시한다. 이를 위해서 표준 패턴 클래스로 분류되어 있는 불량 HDD의 각 클래스의 확률 모델을 GMM(Gaussian Mixture Model)로 가정한다. 실험은 전문가에 의해 분류된 실제 HDD 결함 분포로부터 5가지의 특징 값들을 추출한 후, 결함 분포의 클래스를 표현할 수 있는 GMM의 파라미터(Parameter)를 학습한다. 각 모델의 파라미터를 추정하기 위해 EM(Expectation Maximization) 알고리즘을 사용한다. 학습된 GMM의 분류 테스트는 학습에 사용되지 않은 HDD 결함 분포에서 5가지의 특징 값을 입력 값으로 추정된 모델들의 파라미터 값에 의해 사후 확률을 구한다. 계산된 확률 값 중 가장 큰 값을 갖는 모델의 클래스를 표준 패턴 클래스로 분류한다. 그 결과 제시된 GMM을 이용한 HDD의 패턴 분류의 결과 96.1%의 정답률을 보여준다.

• Journal of the Korea Convergence Society
• /
• v.13 no.2
• /
• pp.249-255
• /
• 2022

This study proposed an analysis framework for real-time prediction of CNC processing defects using machine learning-based models that are recently attracting attention as processing defect prediction methods, and applied it to CNC machines. Analysis shows that the XGBoost, CatBoost, and LightGBM models have the same best accuracy, precision, recall, F1 score, and AUC, of which the LightGBM model took the shortest execution time. This short run time has practical advantages such as reducing actual system deployment costs, reducing the probability of CNC machine damage due to rapid prediction of defects, and increasing overall CNC machine utilization, confirming that the LightGBM model is the most effective machine learning model for CNC machines with only basic sensors installed. In addition, it was confirmed that classification performance was maximized when an ensemble model consisting of LightGBM, ExtraTrees, k-Nearest Neighbors, and logistic regression models was applied in situations where there are no restrictions on execution time and computing power.

• Communications for Statistical Applications and Methods
• /
• v.16 no.2
• /
• pp.239-347
• /
• 2009

This paper presents Kolmogorov-Smirnov, mean difference, AUROC and AR, four well known statistics that have been widely used for evaluating the discriminatory power of credit rating models. Criteria for these statistics are determined by the value of mean difference under the assumption of normality and equal standard deviation. Alternative criteria are proposed through the simulations according to various sample sizes, type II error rates, and the ratio of bads, also we suggest the meaning of statistic on the basis of discriminatory power. Finally we make a comparative study of the currently used guidelines and simulated results.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.6
• /
• pp.374-382
• /
• 2020

The purpose of this study is to improve the learning speed of an ammunition stockpile reliability classification artificial neural network model by proposing a normalization method that reduces the number of input variables based on the characteristic of Ammunition Stockpile Reliability Program (ASRP) data without loss of classification performance. Ammunition's performance requirements are specified in the Korea Defense Specification (KDS) and Ammunition Stockpile reliability Test Procedure (ASTP). Based on the characteristic of the ASRP data, input variables can be normalized to estimate the lot percent nonconforming or failure rate. To maintain the unitary hypercube condition of the input variables, min-max normalization method is also used. Area Under the ROC Curve (AUC) of general min-max normalization and proposed 2-step normalization is over 0.95 and speed-up for marching learning based on ASRP field data is improved 1.74 ~ 1.99 times depending on the numbers of training data and of hidden layer's node.

• Journal of Korea Multimedia Society
• /
• v.14 no.7
• /
• pp.895-904
• /
• 2011

In welding machine that executes existing spot welding, wrong operation of system has often occurs because of their mechanical motion that can be caused by a number of supply like the welding object. In exposed working environment for various situations such as worker or related equipment moving into any place that we are unable to exactly distinguish between good and not condition of nut. Also, in case of defective welding of nut, it needs various evaluation and analysis through image processing because the problem that worker should be inspected every single manually. Therefore in this paper, if the object was not stabilization state correctly, we have purpose to algorithm implementation that it is to reduce the analysis time and exact recognition as to improve system of image processing. As this like, as image analysis for assessment whether it is good or not condition of nut, in his paper, implemented algorithms were suggested and list by group and that it showed the effectiveness through more than one experiment. As the result, recognition rate of normality and error according to the estimation time have been shown as 40%~94.6% and 60%~5.4% from classification 1 of group 1 to classification 11 of group 5, and that estimation time of minimum, maximum, and average have been shown as 1.7sec.~0.08sec., 3.6sec.~1.2sec., and 2.5sec.~0.1sec.