Ensemble-based Quality Classification and Deep Reinforcement Learning-based Production Scheduling : Ensemble-based Quality Classification and Deep Reinforcement Learning-based Production Scheduling
Smart Factory has received worldwide attention as a future factory in the era of Industry 4.0. Recently, many manufacturing industries have adopted information and communication technologies to improve their competitive advantage, through what is known as a smart factory. The smart factory employs an advanced manufacturing environment in production and operations by embracing information technologies such as the Internet of Things (IoT), the cloud, big data analytics and cyber-physical systems (CPS). In the smart factories of the future, the convergence of information technology and factory automation will affect most manufacturing activities such as demand forecasting, production planning and control, scheduling, inventory and logistics management.
In order to substantially achieve the goal of Smart Factory, it is necessary to analyse the data collected at the manufacturing site and utilize the results for production management and quality control. Data-driven quality control techniques are being actively developed for implementation in smart factories. Quality prediction during manufacturing processes is a good example of how big data analytics can influence advanced manufacturing environments. In this thesis, the problem of classifying manufacturing process conditions into normal and defective products according to defect types is dealt with. Such a quality analysis data set is generally unbalanced because the defective rate is quite low in practice. To solve this imbalanced classification problem, a cost-sensitive decision tree ensemble algorithm is adopted to boost the small number of defective cases and assign a higher cost to the misclassification of defective products than that of normal products. C4.5 decision trees are used as base classifiers, and three cost-sensitive ensembles, AdaC1, AdaC2 and AdaC3, are tried to address the imbalanced classification. A few types of defect conditions in a real-world die-casting data set were predicted through the proposed methods. In these experiments, the cost-sensitive ensembles were able to classify the imbalanced data and detect the defect conditions more precisely and more exactly than 19 algorithms in other classification categories such as classic classifiers and ensembles, cost-sensitive single classifiers and sampling-based ensembles. Especially, the AdaC2-based method mainly outperformed all other classification algorithms in terms of performance measures such as F-measure, G-means and AUC for the die-casting quality condition classification problem.
A fast and adaptive production scheduling is another key activity of the smart factory. Traditional approaches to solving scheduling problems assume full knowledge of the problem where finding optimal solutions require a huge computation time. In this thesis, we adopt a reinforcement learning algorithm based on a Deep Q-Network on scheduling (DQN4S). The Deep Q-Network algorithm is one of deep reinforcement learning algorithms, which is well-known to learn successful policies directly from high-dimensional sensory inputs. We interpret a parallel machine scheduling problem as sequential decision processes and let an agent of the reinforcement learning to learn a policy of scheduling that makes decisions of job dispatching and improves dispatching actions by trial and error. We have first tested the proposed DQN4S algorithm for single machine problem, and it was proved to find the optimal solution. We then have applied the DQN4S to the parallel machine problem. In our experiments we compared our algorithm to the approaches of Dispatching Rules (DRs), Heuristic, Mixed Integer Programming (MIP), and Column Generation (CG). The performance of DQN4S was always better than DRs. Compared with Heuristic, MIP and CG, the performance of DQN4S was better as the number of jobs increased, when the number of machines was fixed. In addition, the computation time of DQN4S was superior to CG or MIP. This phenomenon shows the efficiency and effectiveness of the DQN4S models for solving the online scheduling problems. Finally, we dealt with multi-objective problems, which minimize the sum of makespan and total weighted completion time in parallel machine problem while adjusting the weight of each objective. DQN4S derived adaptive scheduling results by finding a scheduling policy that maximizes the objective function according to the change of the objective function.
분석정보
서지정보 내보내기(Export)
닫기소장기관 정보
닫기권호소장정보
닫기오류접수
닫기오류 접수 확인
닫기음성서비스 신청
닫기음성서비스 신청 확인
닫기이용약관
닫기학술연구정보서비스 이용약관 (2017년 1월 1일 ~ 현재 적용)
학술연구정보서비스(이하 RISS)는 정보주체의 자유와 권리 보호를 위해 「개인정보 보호법」 및 관계 법령이 정한 바를 준수하여, 적법하게 개인정보를 처리하고 안전하게 관리하고 있습니다. 이에 「개인정보 보호법」 제30조에 따라 정보주체에게 개인정보 처리에 관한 절차 및 기준을 안내하고, 이와 관련한 고충을 신속하고 원활하게 처리할 수 있도록 하기 위하여 다음과 같이 개인정보 처리방침을 수립·공개합니다.
주요 개인정보 처리 표시(라벨링)
목 차
3년
또는 회원탈퇴시까지5년
(「전자상거래 등에서의 소비자보호에 관한3년
(「전자상거래 등에서의 소비자보호에 관한2년
이상(개인정보보호위원회 : 개인정보의 안전성 확보조치 기준)개인정보파일의 명칭 | 운영근거 / 처리목적 | 개인정보파일에 기록되는 개인정보의 항목 | 보유기간 | |
---|---|---|---|---|
학술연구정보서비스 이용자 가입정보 파일 | 한국교육학술정보원법 | 필수 | ID, 비밀번호, 성명, 생년월일, 신분(직업구분), 이메일, 소속분야, 웹진메일 수신동의 여부 | 3년 또는 탈퇴시 |
선택 | 소속기관명, 소속도서관명, 학과/부서명, 학번/직원번호, 휴대전화, 주소 |
구분 | 담당자 | 연락처 |
---|---|---|
KERIS 개인정보 보호책임자 | 정보보호본부 김태우 | - 이메일 : lsy@keris.or.kr - 전화번호 : 053-714-0439 - 팩스번호 : 053-714-0195 |
KERIS 개인정보 보호담당자 | 개인정보보호부 이상엽 | |
RISS 개인정보 보호책임자 | 대학학술본부 장금연 | - 이메일 : giltizen@keris.or.kr - 전화번호 : 053-714-0149 - 팩스번호 : 053-714-0194 |
RISS 개인정보 보호담당자 | 학술진흥부 길원진 |
자동로그아웃 안내
닫기인증오류 안내
닫기귀하께서는 휴면계정 전환 후 1년동안 회원정보 수집 및 이용에 대한
재동의를 하지 않으신 관계로 개인정보가 삭제되었습니다.
(참조 : RISS 이용약관 및 개인정보처리방침)
신규회원으로 가입하여 이용 부탁 드리며, 추가 문의는 고객센터로 연락 바랍니다.
- 기존 아이디 재사용 불가
휴면계정 안내
RISS는 [표준개인정보 보호지침]에 따라 2년을 주기로 개인정보 수집·이용에 관하여 (재)동의를 받고 있으며, (재)동의를 하지 않을 경우, 휴면계정으로 전환됩니다.
(※ 휴면계정은 원문이용 및 복사/대출 서비스를 이용할 수 없습니다.)
휴면계정으로 전환된 후 1년간 회원정보 수집·이용에 대한 재동의를 하지 않을 경우, RISS에서 자동탈퇴 및 개인정보가 삭제처리 됩니다.
고객센터 1599-3122
ARS번호+1번(회원가입 및 정보수정)