An Independent Component Analysis Based on Sliding Window Statistics and its Application to Power Plant Equipment State Monitoring

2013 ◽  
Vol 860-863 ◽  
pp. 1801-1806
Author(s):  
Yue Zhao ◽  
Feng Qi Si ◽  
Zhi Gao Xu
Keyword(s):  
Power Plant ◽  
Independent Component Analysis ◽  
Condition Monitoring ◽  
Sliding Window ◽  
Component Analysis ◽  
Independent Component ◽  
Engineering Practice ◽  
Monitoring Methods ◽  
Sliding Windows ◽  
Operation Conditions

Traditional condition monitoring methods are not suitable for the nonlinear operation parameters and time-variable operation conditions. We propose an independent component analysis method based on sliding window statistics (SSWICA). This method uses statistics in sliding windows of parameters as input samples, then uses a N-step forward sliding window ICA method to modeling. Then we monitor the operating state of the equipments by observing whether the SPE index of real-time parameters exceeds the control limits. SSWICA is applied to condition monitoring of condenser in 600MW unit, comparing with traditional ICA monitoring methods based on sliding window. The results show SSWICA can accurately reflect current operating state and related changes of condensers state parameters, recognize steady, unsteady and fault conditions effectively. It is valuable for engineering practice and suitable for the application to equipments condition monitoring in power plant.

scholarly journals Acoustical Source Tracing Using Independent Component Analysis and Correlation Analysis

2015 ◽  
Vol 2015 ◽  
pp. 1-8 ◽  
Author(s):  
Wei Cheng ◽  
Zhousuo Zhang ◽  
Jie Zhang ◽  
Jiantao Lu
Keyword(s):  
Fault Diagnosis ◽  
Correlation Analysis ◽  
Independent Component Analysis ◽  
Condition Monitoring ◽  
Component Analysis ◽  
Independent Component ◽  
Source Information ◽  
Independent Components ◽  
Source Tracing ◽  
Acoustical Signals

Acoustical signals from mechanical systems reveal the operating conditions of mechanical components and thus benefit for machinery condition monitoring and fault diagnosis. However, the acoustical signals directly measured by the sensors in essential are the mixed signals of all the sources, and normally it is very difficult to be used for source identification or operating feature extraction. Therefore, this paper studies the acoustical source tracing problem using independent component analysis (ICA) and identifies the sources using correlation analysis: the measured acoustical signals are separated into independent components by independent component analysis method, and thus all the independent information of all the sources is obtained; these independent components are identified based on the prior information of the sources and correlation analysis. Therefore, all the source information contained in the measured acoustical signals can be independently separated and traced, which can provide more purer source information for condition monitoring and fault diagnosis.

Sign in / Sign up

Export Citation Format

Share Document