설비 제어 특성을 이용 학습형 설비 예지 보전 지원 Framework 개발
Development of a self-learning framework which support precognition maintenance using control peculiarity data of equipments
- 주제(키워드) Artificial neural network (ANN) , electrical energy usage monitoring , fault detection and isolation , industrial process modelling , manufacturing systems , programmable logic controller (PLC) , PLC Log data analysis
- 발행기관 아주대학교
- 지도교수 왕지남
- 발행년도 2018
- 학위수여년월 2018. 8
- 학위명 박사
- 학과 및 전공 일반대학원 산업공학과
- 실제URI http://www.dcollection.net/handler/ajou/000000028003
- 본문언어 영어
- 저작권 아주대학교 논문은 저작권에 의해 보호받습니다.
초록/요약
In modern factory maintenance department is one of most important departments. Maintenance work decide one factory’s production efficiency and the quality of the products, thereby influence the cost of products and the competitiveness of enterprises. There are many maintenance policy, but all researcher’s research direction focuses on prescient maintenance. But almost previous works using high performance measuring device to monitoring equipment health state, this kind of high performance device require dedicated communication line and it’s usually one-to-one monitoring. This determines previous framework is expensive and not easy to apply in large scale. In this paper describes one precognition maintenance framework which much different with previous studies. Proposed framework using controller(PLC) log data to determine the running state of the equipment, perception equipment health state through comprehensive analysis controller log data and energy consume data. The corresponding maintenance methods will recommend according to the relationship between the previous maintenance records and the health status of the equipment. Proposed framework supply one self-learning core to give one gradual improvement maintenance model for applied factory. Gradually improve predict accuracy rate to help maintenance manager make more applicable maintenance plan. When the framework runs for a long time, system will exactly predict maintenance time and required maintenance work. Efficiently decrease break-down time and maintenance cost.
more목차
1 INTRODUCTION 1
2 BACKGROUND AND RELATED WORK 6
3 INTRODUCTION OF MAINTENANCE 12
3.1 History of Maintenance 14
3.2 RCM(Reliability-Centered Maintenance) 18
4 INTRODUCTION OF NEURAL NETWORKS 24
4.1 Model of a Neuron 26
4.2 Network Architectures 29
4.2.1 Single -layer feedforward networks 29
4.2.2 Multilayer feedforward networks 31
4.2.3 Recurrent networks 32
4.2.4 Lattice structures 33
5 PROPOSED FRAMEWORK 34
5.1 Introduction Framework 36
5.1.1 Data gathering part 36
5.1.1.1 PLC log data gathering 36
5.1.1.2 Energy consume data gathering 38
5.1.2 Data analysis part 44
5.1.2.1 Energy consume data 44
5.1.2.2 Data preprocess 48
5.1.2.3 Data factor 50
5.1.2.4 Data grouping 53
5.1.3 Self-learning part 53
5.2 Computational Algorithm 55
5.2.1 limited minimum distance classification 55
5.2.2 Neural network 57
6 CASE STUDY 59
6.1 Hardware environment 59
6.2 Experimental process 63
6.3 Framework work flow 64
6.3.1 Data gathering and preprocess 64
6.3.2 Data analysis and grouping 69
6.4 Result 76
7 CONCLUSIONS 77
Reference 78
Appendix A: Abbreviation 82
Appendix B: Sample energy log data for motor 1 83
