Kyoung Jun Lee Jun Woo Kwon Soohong Min Jungho Yoon, Embedding Convolution Neural Network-Based Defect Finder for Deployed Vision Inspector in Manufacturing Company Frontec, IAAI 2020 (The Innovative Applications of Artificial Intelligence Conference), 2020. (Pdf)
In collaboration with Frontec, which produces parts such as bolts and nuts for the automobile industry, Kyung Hee University and Benple Inc. develop and deploy AI system for automatic quality inspection of weld nuts. Various constraints to consider exist in adopting AI for the factory, such as response time and limited computing resources available. Our convolutional neural network (CNN) system using large-scale images must classify weld nuts within 0.2 seconds with accuracy over 95%. We designed Circular Hough Transform based preprocessing and an adjusted VGG (Visual Geometry Group) model. The system showed accuracy over 99% and response time of about 0.14 sec. We use TCP / IP protocol to communicate the embedded classification system with an existing vision inspector using LabVIEW. We suggest ways to develop and embed a deep learning framework in an existing manufacturing environment without a hardware change.
이경전, 권준우, 이미지 기반 품질 관리 기법의 스마트 팩토리 현장 적용 이슈와 전략 (Issues and Strategy for Deploying Image-Based Quality Management for Smart Factory), Entrue Journal of Information Technology Vol.17, No.1 / December 2019. (Pdf)
콘볼루션 신경망 기술의 발전이 영상 기반 품질 경영에서의 전처리 부담을 많이 줄여주는 의미가 있지만, 콘 볼루션 신경망의 발전이 전처리 노력을 완전히 제거해주지는 못한다. 그러나, 조금만 훈련받으면 컴퓨터 비전 전문가가 아니더라도 영상 기반의 품질 관리를 할 수 있으며, 이에 기반하여 가변적인 생산체계에 빠르게 적응 할 수 있다. 스마트 팩토리에서 자동화된 품질관리를 현실에서 실제 적용하는 것은, 이 방법론들을 이해하고, 이를 일부 구현하여 적용하거나, 통합적으로 구현하여 완전 자동화하는 형태로 진행된다. 이 논문은 스마트 팩토리 환경에서 자동화된 품질 검사를 위한 이미지 기반 품질 관리 기법들을 개관하고 현실에 이러한 기법을 실제 적용하는 데에서 나타나는 이슈와 전략에 대해 토론한다.
김진호, 권준우, 황보유정, 이경전, 제조분야 자동 품질 검사를 위한 딥러닝 기반 대형 이미지 분류시스템 개발 및 구축 사례, 2018 한국지능정보시스템학회 추계학술대회, 2018. (Pdf)
딥러닝 알고리즘의 발전에 따라, 다양한 산업분야에서의 딥러닝의 도입이 가속되고 있다. 제조분야의 경우 딥러닝을 적용하는데 있어 기존 적용 기술의 정확도를 상회해야 하며, 제안 된 시간내에 프로세스가 완료되어야 한다. 본 연구에서는 나사 제조 분야에서의 제약 조건 및 성능 조건을 만족하는 딥러닝 기반 대형 이미지 분류 시스템을 구축하였다. 결품 분류 공정 중 획득할 수 있는 대형 이미지 사진을 기반으로, 요구 조건인 분류 정확도 95%와 연산시간 0.2초를 만족할 수 있는 딥러닝 모형을 구현하였다. 이미지의 전처리를 위해 Hough Circle과 PCA를 사용하고, VGG 모형을 기반으로 CNN의 구조를 설계하였으며, 연산 속도 0.2초 내에 분류 정확도 99% 가 가능함을 확인하였다.
최형림, 김현수, 박영재, 박병주, 이경전, “인터넷상의 가상생산 기반 부품판매 에이전트 개발”, 경영정보학연구, 12(4):193-213, 2002. (PDF) in Korean
The sales activity of most of small manufacturing companies is based on orders of buyers. The process of promotion, negotiation, receipt and selection of orders of the manufacturers is closely coupled with the load status of the production lines. The decision on whether to accept an order or not, or the selection of optimal order set among excessive orders is entirely dependent on the schedule of production lines. However, in the real world, since the production scheduling activity is mainly performed by human experts, most of small manufacturers are suffer from being unable to meet due dates, lack of rapid decision on the acceptance of new order. Recently, Internet based Electronic Commerce is recognized as one of the alternatives for strengthening sales power of small and medium companies. However, small and medium manufacturers can’t adjust properly to the new environment because they are in short of money, personnel, and technology. To cope with this problem, this paper deals with development of part sales agent coupled with virtual manufacturing in Internet environment that consist of selection agent, advertisement agent, selection agent, negotiation agent, and virtual manufacturing system. This paper develops a time-bounded negotiation mechanism for small and medium manufacturers in agent-based automated negotiation between customers and negotiation agents. Furthermore, to select optimal order set maximized profit, we first formulate the order selection problem with mixed integer programming, but the computation time of IP is not acceptable for real world scale problem. To overcome this problem and dynamic nature of virtual manufacturing, we suggest a genetic algorithm approach, which shows a reasonable computation time for real world case and good incremental problem solving capability.
이경전, 장용식, 최형림, 김현수, 박영재, 박병주, “복잡한 의사결정과 협상 환경을 위한 에이전트 기반 시스템: 가상생산 응용”, Information Systems Review, 4(2):223-236, 2002. (PDF) in Korean
In an agent-based system, each agent has its own decision making capability and competes, cooperates, and communicates each other with an agent interface. Virtual manufacturing has characteristics as a typical application area of agent-based system: complex and time-bound decision making and negotiation. The time-boundness influences the choice of decision making models and design of protocols for internal and external negotiation. In this paper, we provide a case study which suggests a time-bound framework for external and internal negotiation between the agents for virtual manufacturing environment. We illustrate decision making model selection strategy and the system architecture of the agent-based system for the complex and time-bound environment.
Kim, H. Lee, K., Criteria of Good Project Network Generator and its Fulfillment using a Dynamic CBR Approach, Lecture Notes in Computer Science 3155:630-644, September, 2004. – SCIE, ISSN:0302-9743. pdf
Most project-based industries such as construction, shipbuilding, and software development etc. should generate and manage project network for successful project planning. We suggest a set of criteria of good project network generator such as network generation efficiency, quality of network, and economics of system development. For the efficiency of the planning, the first criterion, we decided to take a CBR approach. However, using only previous cases is insufficient to generate a proper network for a new project. By embedding rules and constraints in the case-based system, we could improve the quality of the project network: the second criterion. The integration of CBR approach and the knowledge-based approach makes feasible the development of the project network generator and improves the quality of the network by mutual enhancement through crosschecking the knowledge and cases in the development and maintenance stages. For some complex project network planning, a single-case assumed project network generation methodology is refined into Dynamic Leveled Multiple Case approach. The methodology contributes again the efficiency and effectiveness of project network generation and reduces the efforts of the system development.
Lee, J. K., Lee, K. J., Hong, J. S., Kim, W. J., Kim, E. Y., Choi, S. Y., Kim, H. D., Yang, O. R., Choi, H. R., “DAS: Intelligent Scheduling Systems for Shipbuilding”, AI Magazine, vol. 16, no. 2, pp. 78-94, Winter, 1995.pdf
Daewoo Shipbuilding Company, one of the largest shipbuilders in the world, has experienced a great deal of trouble with the planning and scheduling of its production process. To solve the problems, from 1991 to 1993, Korea Advanced Institute of Science and Technology (KAIST) and Daewoo jointly conducted the Daewoo Shipbuilding Scheduling (das) Project. To integrate the scheduling expert systems for shipbuilding, we used a hierarchical scheduling architecture. To automate the dynamic spatial layout of objects in various areas of the shipyard, we developed spatial scheduling expert systems. For reliable estimation of person-hour requirements, we implemented the neural network–based person-hour estimator. In addition, we developed the paneledblock assembly shop scheduler and the longrange production planner. For this large-scale project, we devised a phased development strategy consisting of three phases: (1) vision revelation, (2) data-dependent realization, and (3) prospective enhancement. The DAS systems were successfully launched in January 1994 and are actively being used as indispensable systems in the shipyard, resulting in significant improvement in productivity and visible and positive effects in many areas.
Lee, K. J., J. K. Lee, S. Y. Choi, “A Spatial Scheduling System and its Application to Shipbuilding: DAS-CURVE”, Expert Systems with Applications, vol. 10, no.3/4 , pp. 311-324, 1996.pdf
Spatial scheduling considers not only traditional scheduling constraints like resource capacity and due dates, but also dynamic spatial layout of the objects. Automation of spatial scheduling is particularly important when the spatial resources are critical bottleneck resources, as is the case in the shipbuilding industry. To develop a spatial scheduling expert system for shipbuilding, a methodology for spatial layout of polygonal objects within rectangular plates is first developed. This study is then extended to the methodology for spatial scheduling, including the time dimension. The methodology is applied to the scheduling of Daewoo shipbuilding to build a system DAS-CURVE. DAS-CURVE is successfully operational and its experimental performance is remarkable.
Lee, J. K., Lee, K. J., Park, H. K., Hong, J. S., Lee, J. S., “Developing Scheduling Systems for Daewoo Shipbuilding: DAS Project”, European Journal of Operational Research, vol. 97, no.2, pp.380-395, 1997.pdf
Daewoo Shipbuilding Company, one of the largest shipbuilders in the world, had difficulties with planning and scheduling its production process. To solve the problems, Korea Advanced Institute of Science and Technology (KAIST) and Daewoo have been jointly performing the DAS (DAewoo Shipbuilding Scheduling) Project for three years from 1991 to 1993. To develop the integrated scheduling systems, several technological breakthroughs were necessary such as hierarchical architecture between systems, constraint directed graph search, spatial scheduling, dynamic assembly line scheduling, and neural network based man-hours estimation. Besides these technological research issues, we adopted the phased development strategy, which consists of three phases of vision revelation, data dependent realization, and prospective enhancement. The DAS systems were successfully launched in January 1994 and are being actively used as indispensable systems in the shipyard resulting in a significant improvement in productivity and reengineering of the scheduling process.
Scheduling, Manufacturing Industries, Artificial Intelligence, Shipbuilding.
Lee, K. J., “Sufficient Search Space in Spatial Expert Systems”, Expert Systems with Applications, vol. 19, no. 1, pp.1-8, July, 2000.pdf
This paper seeks the sufficient search space for the expert systems locating rectangular and arbitrary-shaped objects placed without rotation within a two-dimensional rectangular space. We found that for the layout of rectangular objects, the convex vertex set of feasible allocation space is a sufficient space to determine a feasible layout. We also found that for the layout of arbitrary-shaped objects, the boundary point set of the feasible allocation space is a sufficient space to determine a feasible layout. These two theorems are proved by developing two respective parallel translation algorithms. These theorems show that the search space can be significantly reduced in finding a feasible layout. Since these theorems were discovered while we were developing a spatial scheduling expert system, we have empirically tested the performance of the reduced search space with real world examples. According to the empirical test for the convex polygonal objects, the vertex set of feasible allocation space is satisfactory enough as a search space although the vertex set is not a sufficient space.