상세정보
학술대회
Development of a CNN-based Expert System using Domain Knowledge
- 발행일
- 201906
- 출처
- International Technical Conference on Circuits/Systems, Computers and Communications (ITC-CSCC) 2019, pp.829-830
- 협약과제
- 19PT1100, 지능정보기술 기반의 제조혁신 및 최적운영 지원 시스템의 지식베이스 프레임워크 설계 개발, 김원종
- 초록
- This paper describes development of a CNN-based expert system which can be used in smart factory applications such as automatic facilities control. In real situation, human experts control facilities with different values for the same input conditions, since there are tolerances for the control rather than exact values. So, when we develop system, domain knowledge is important. We used these knowledges of experts in preprocessing. To consider experts knowledge, we used average and median values in min/max range for each input pattern. The core algorithm of the expert system uses CNN-model. Final results are also evaluated based on expert's knowledge. Experimental results show that the proposed expert system can recommend control values with accuracy of 81.8% for the values and 98% for the min/max ranges, respectively. Also our recommend system has less outlier values compared with expert's ones.
- KSP 제안 키워드
- Input pattern, Recommend System, Smart Factory, domain knowledge, expert system, real situation