Intelligent condition monitoring and prognostics system based on data-fusion strategy

This paper proposes an intelligent condition monitoring and prognostics system in condition-based maintenance architecture based on data-fusion strategy. Firstly, vibration signals are collected and trend features are extracted. Then features are normalized and sent into neural network for feature-level fusion. Next, data de-noising is conducted containing smoothing and wavelet decomposition to reduce the fluctuation and pick out trend information. The processed information is used for autonomic health degradation monitoring and data-driven prognostics. When the degradation curve crosses through the specified threshold of alarm, prognostics module is triggered and time-series prediction is performed using multi-nonlinear regression models. Furthermore, the predicted point estimate and interval estimate are fused, respectively. Finally, remaining useful life of operating machine, with its uncertainty interval, are assessed. The proposed system is evaluated by an experiment of health degradation monitoring and prognostics for a methane compressor. The experiment results show that the enhanced maintenance performances can be obtained, which make it suitable for advanced industry maintenance.