논문 상세보기
Operational State-Aware Early Process Deviation Detection in Slurry Pumps KCI 등재
운영 상태 기반 슬러리 펌프의 조기 공정 이탈 탐지
- 언어KOR
- URLhttps://db.koreascholar.com/Article/Detail/453352
pp.187-199
This study proposes an operational state-aware early process deviation detection method for predictive maintenance of slurry pump systems. In industrial applications, slurry pumps are typically operated under fixed-speed conditions, where process variables such as pressure, flow rate, and motor current are expected to remain stable during normal operation. However, in real-world environments, subtle process deviations may occur even during stable operation, excluding start-up, shutdown, and setpoint-change transients. Such deviations are often difficult to manage using conventional threshold-based alarms or vibration-centered predictive maintenance approaches. To address this issue, an operational state-aware framework was developed to classify pump operation into OFF, START_TRANSIENT, STEADY, END_TRANSIENT, and SETPOINT_TRANSIENT based on campaign segmentation and time-based preprocessing. Within the quantitatively defined STEADY state, process deviation candidates characterized by a simultaneous decrease in flow rate and increases in discharge pressure and motor current under a fixed rotational speed were defined as WEAK_DEVIATION. A total of 312,483 data points were collected from an industrial slurry pump system between August and December 2025, of which 113,653 operational data points remained after excluding the OFF state. As a result, eight independent WEAK_DEVIATION events were identified during STEADY operation. These events exhibited sustained multivariate process-signal changes rather than instantaneous fluctuations, with a consistent pattern of decreased flow rate and increased discharge pressure and motor current, indicating their potential relevance to early process deviation. A direct comparison identified 13 independent WEAK_DEVIATION events under the State-unaware condition, including three events during SETPOINT_TRANSIENT periods. The results demonstrate that combining multivariate process signals with operational state awareness helps distinguish normal transient responses from process-signal changes during STEADY operation, thereby improving the selectivity of early process deviation detection without relying on complex models or additional sensors. The proposed framework provides a practical and interpretable approach for industrial predictive maintenance applications.
2. 공정 및 데이터 구성
2.1 슬러리 펌프 공정 개요
2.2 데이터 수집 환경 및 센서 구성
2.3 데이터 전처리 및 운전 상태 식별
3. 운영 상태 기반 프레임워크
3.1 운영 상태 분류 개요
3.2 운영 상태 정의 및 분류
3.3 STEADY 구간의 통계적 변동성 평가
3.4 기동․정지 및 RpmSet 변경 전이 상태 분리
3.5 WEAK_DEVIATION 탐지 기준
4. 실험 결과 및 사례 분석
4.1 운영 상태 분류 결과
4.2 기동 전이 구간 사례 분석
4.3 WEAK_DEVIATION 탐지 결과
4.4 WEAK_DEVIATION 사례 분석
4.5 결과 요약 및 해석
5. 결론 및 향후 연구
References
