Feature Extraction for Condition Monitoring of a Hydraulic Complex Self-sensing Testbench Valve with Data Mining Approach
سال انتشار: 1401
نوع سند: مقاله کنفرانسی
زبان: انگلیسی
مشاهده: 113
فایل این مقاله در 20 صفحه با فرمت PDF قابل دریافت می باشد
- صدور گواهی نمایه سازی
- من نویسنده این مقاله هستم
استخراج به نرم افزارهای پژوهشی:
شناسه ملی سند علمی:
IRMA15_008
تاریخ نمایه سازی: 1 اسفند 1403
چکیده مقاله:
Nowadays Condition monitoring create more value for Maintenance enabling repair and replacement monitoring’s more cost effective at the same time reduce the machineries downtime. Implementation of the algorithms of Machine Learning in order to detect the faults of process or a specific component is a state-of-the-art subject debated in literature. In this research, we have used plenty of analysis based on sophisticated statistical data modeling extract the characteristics feature from sensor readings in a Hydraulic Complex Self-sensing Test rig. Afterwards the feature reduction was performed by Random Projection Method to cover more failures for a hydraulic switching valve. The raw dataset was obtained from Repository of UCI Center for Machine Learning and Intelligent Systems implemented in Linear Discrimination Analysis. The result of different process readings including Pressure, Flow and Temperature are compared by plenty of Methods and algorithm where state of the Pumps, Accumulators, Cooler and Valve conditions were artificially varied. It was required to simulate a feature reduction on valve component’s fault on the time and frequency domain between ۱ to ۱۰۰ Hz.The results our research showed similar to other studies (K. S. Helwig Nikolai ۲۰۱۵) that for the above test bench a high correlation for Cooler and Switching Valve component but lower correlation for Pumps and Accumulators due to their robustness of Process Outliers and Offsets. The results of Random Loading also more challenging for the Fault detection of internal leakage of the Pump and Accumulator Gas pressure which occurred gradually in a longer period.Finally, the reduced features were imported to a supervised Machine Learning model and their feature-fault correlation was measured by the Multinomial Machine with more than ۸۶% precision with OneR, Symmetrical Uncertainty and IBK Wrapper.
نویسندگان
Alireza Omidvar
Fleet Engineering Department, Tehran Urban and Suburban Railway Co. (TUSRC). Tehran, Iran,
Arash HajiKandi
Fleet Engineering Department, Tehran Urban and Suburban Railway Co. (TUSRC). Tehran, Iran,
Seyed Mehdi Hejazi
Fleet Engineering Department, Tehran Urban and Suburban Railway Co. (TUSRC). Tehran, Iran,
Mohammad Reza Hatami
Overhaul and Periodical Maintenance Department Line ۲, Tehran Urban and Suburban RailwayCo. (TUSRC). Tehran, Iran