• 반도체디스플레이기술학회지
• /
• 제20권2호
• /
• pp.109-114
• /
• 2021

When we inspect scaffold defect using sight, inspecting performance is decrease and inspecting time is increase. We need for automatically scaffold defect detection method to increase detection accuracy and reduce detection times. In this paper. We produced scaffold defect classification models using densenet, alexnet, vggnet algorithms based on CNN. We photographed scaffold using multi dimension camera. We learned scaffold defect classification model using photographed scaffold images. We evaluated the scaffold defect classification accuracy of each models. As result of evaluation, the defect classification performance using densenet algorithm was at 99.1%. The defect classification performance using VGGnet algorithm was at 98.3%. The defect classification performance using Alexnet algorithm was at 96.8%. We were able to quantitatively compare defect classification performance of three type algorithms based on CNN.

• Smart Structures and Systems
• /
• 제23권1호
• /
• pp.71-79
• /
• 2019

In general, it may be advantageous to measure the pressure pulsation near a valve to detect a valve defect in a linear compressor. However, the acceleration signals are more advantageous for rapid classification in a mass-production line. This paper deals with the performance improvement of fault classification using only the compressor-shell acceleration signal based on the relation between the refrigerant pressure pulsation and the shell acceleration of the compressor. A transfer function was estimated experimentally to take into account the signal noise ratio between the pressure pulsation of the refrigerant in the suction pipe and the shell acceleration. The shell acceleration signal of the compressor was modified using this transfer function to improve the defect classification performance. The defect classification of the modified signal was evaluated in the acceleration signal in the frequency domain using Fisher's discriminant ratio (FDR). The defect classification method was validated by experimental data. By using the method presented, the classification of valve defects can be performed rapidly and efficiently during mass production.

• 반도체디스플레이기술학회지
• /
• 제20권1호
• /
• pp.64-69
• /
• 2021

In this paper, a method for improving the defect classification performance in low contrast, ununiformity and featureless steel plate surfaces has been studied based on deep convolution neural network and transfer-learning neural network. The steel plate surface images have low contrast, ununiformity, and featureless, so that the contrast between defect and defect-free regions are not discriminated. These characteristics make it difficult to extract the feature of the surface defect image. A classifier based on a deep convolution neural network is constructed to extract features automatically for effective classification of images with these characteristics. As results of the experiment, AlexNet-based transfer-learning classifier showed excellent classification performance of 99.43% with less than 160 seconds of training time. The proposed classification system showed excellent classification performance for low contrast, ununiformity, and featureless surface images.

• Smart Structures and Systems
• /
• 제25권2호
• /
• pp.255-264
• /
• 2020

This paper deals with a defect classification technique that considers the structural characteristics of a refrigerant compressor. First, the pressure pulsation of the refrigerant flowing in the suction pipe of a normal compressor was measured at the same time as the acceleration of the shell surface, and then the transfer function between the two signals was estimated. Next, the frequency-weighted acceleration signals of the defect classification target compressors were generated using the estimated transfer function. The estimation of the variance of the transfer function is presented to formulate the frequency-weighted acceleration signals. The estimated frequency-weighted accelerations were applied to defect classification using frequency-domain features. Experiments were performed using commercial compressors to verify the technique. The results confirmed that it is possible to perform an effective defect classification of the refrigerant compressor by the shell surface acceleration of the compressor. The proposed method could make it possible to improve the total inspection performance for compressors in a mass-production line.

• 반도체디스플레이기술학회지
• /
• 제19권3호
• /
• pp.77-81
• /
• 2020

In this paper, we create scaffold defect classification models using machine learning based data. We extract the characteristic from collected scaffold external images using USB camera. SVM, KNN, MLP algorithm of machine learning was using extracted features. Classification models of three type learned using train dataset. We created scaffold defect classification models using test dataset. We quantified the performance of defect classification models. We have confirmed that the SVM accuracy is 95%. So the best performance model is using SVM.

• 한국건축시공학회지
• /
• 제8권1호
• /
• pp.63-70
• /
• 2008

Management process of apartment buildings construction has increased because the after service of construction company meet the needs of customers. Many defect data, which was inspected by construction company or customers before moving into a new apartment house, were classified by field engineers and then communicated to corresponding subcontractors. The classification process needs to be performed by an expert engineer because there is so much data, it is unfeasible to complete in a short period of time. For this classification process, an automatic classification system using case base reasoning (CBR) should be considered. This research proposed a defect management system with automatic classification system using CBR. This constructed defect management system consists of cyber after service system for tenants and the whole defect management process of construction, preservation and management of apartment buildings. This system could improve the efficiency of expert work in terms of time and accuracy, as well as helping laymen users to conduct defect classification work as experts do.

• 한국멀티미디어학회논문지
• /
• 제13권8호
• /
• pp.1115-1127
• /
• 2010

TFT-LCD 생산 과정에서 발생하는 결함을 정확하게 분류하여 결함 유형에 따라 폐기, 사용가능 등의 의사결정을 적절하게 내리는 것은 수율 증가 및 생산성 향상에 필수적인 요소이다. 본 논문에서는 TFT-LCD 생산 라인에서 획득한 결함 영상에 대하여 명암도 분포(intensity distribution) 및 결함 영상의 형태 특징(shape feature)을 분석하여 효과적으로 필름 결함 유형을 분류하는 기법을 제시한다. 본 연구에서는 먼저 필름 결함 영상을 결함 영역과 결함이 아닌 배경 영역으로 이진화하고, 결함 영역에서 결함의 선형성(linearity), 명암도 분포를 고려한 형태 특징 등의 여러 가지 특징을 분석하여 기준 영상(referential image) 데이터베이스를 구축하였으며, 분류하고자 하는 결함 영상과 데이터베이스에 저장된 기준 영상과의 매칭 비용 함수(matching cost function)를 정의하여 적절히 매칭시킴으로써 결함의 유형을 결정하였다. 제시한 기법의 성능을 검증하기 위하여 실제 TFT-LCD 생산 라인에서 획득한 결함 영상들을 대상으로 분류 실험을 수행하였으며, 실험 결과 생산 라인에서 이용할 수 있을 정도의 상당한 수준의 분류 정확도를 달성하였음을 보여주었다.

• 반도체디스플레이기술학회지
• /
• 제19권3호
• /
• pp.1-6
• /
• 2020

In this paper, a method for improving the defect classification performance in steel plate surface has been studied, based on DWT(discrete wavelet transform) and SVM(support vector machine). Surface images of the steel plate have low contrast, uneven, and featureless, so that the contrast between defect and defect-free regions is not discriminated. These characteristics make it difficult to extract the feature of the surface defect image. In order to improve the characteristics of these images, a synthetic images based on discrete wavelet transform are modeled. Using the synthetic images, edge-based features are extracted and also geometrical features are computed. SVM was configured in order to classify defect images using extracted features. As results of the experiment, the support vector machine based classifier showed good classification performance of 94.3%. The proposed classifier is expected to contribute to the key element of inspection process in smart factory.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 1999년도 추계종합학술대회 논문집
• /
• pp.637-640
• /
• 1999

In this Paper we show the LCD simulator for defect inspection using image processing algorithm and neural network. The defect inspection algorithm of the LCD consists of preprocessing, feature extraction and defect classification. Preprocess removes noise from LCD image, using morphology operator and neural network is used for the defect classification. Sample images with scratch, pinhole, and spot from real LCD color filter image are used. The proposed algorithms show that defect detected and classified in the ratio of 92.3% and 94.6 respectively.

• 한국산학기술학회논문지
• /
• 제19권3호
• /
• pp.61-68
• /
• 2018

공동주택 하자는 유지보수에 막대한 비용이 들어가게 되며, 발주자, 시공자 그리고 입주자 등에게 심각한 피해를 입힌다. 이에 따라 하자분쟁을 최소화하고 철저한 품질관리를 통한 체계적이고 효율적인 하자관리를 위한 토대를 마련할 필요가 있다. 본 연구에서는 하자분쟁사례를 활용하여 공동주택의 공종/부위/현상에 따른 하자분류체계를 도출하고, 이를 기반으로 하자유형별 하자위험을 평가할 수 있는 방안을 제시하였다. 이를 위하여 본 논문에서는 경과년수 10년 이상 공동주택 하자분쟁사례 34건, 약 6000여개의 하자항목 자료를 토대로 분석을 실시하였다. 분석 결과를 정리하면, 하자분류체계는 하자 공종, 하자부위 및 하자현상으로 크게 분류한 후 세부적으로 총 157개 항목으로 세분화하였다. 하자분류체계를 토대로 하자 빈도, 하자비용 및 하자위험을 분석한 결과, RC공사 및 마감공사에 하자위험이 상당히 집중되어 있는 것으로 확인되었다. 이에 따라 이러한 하자위험에 대한 하자예방 활동이 우선적으로 고려되어야 할 것으로 판단된다. 본 연구를 토대로 하자위험을 관리할 수 있는 방안에 대한 추가적인 연구가 필요할 것으로 판단된다.