Sensor recovery for robust multivariate condition monitoring

The ability to predict and prevent equipment failures is essential to various industrial processes and military operations. In recent years, Condition Monitoring (CM) has been recognized as an effective paradigm in this regard. CM can be performed via several sensor channels with broad coverage to enhance monitoring capabilities. However, loss of sensor readings due to malfunction of connectors and/or sensor abnormalities is the hurdle to reliable fault diagnosis and prognosis in multichannel CM systems. The problem becomes more challenging when the sensor channels are not synchronized because of different and/or time-varying sampling/transmission rates. This paper provides a new sensor recovery technique to improve the robustness of multichannel CM systems. Specifically, the associated sensor signals are modeled through Functional Principal Component Analysis (FPCA). Based on the FPCA results obtained from historical data, the relationships among the signals can be constructed. In on-line implementation, such relationships along with parameters updated by real-time CM data can be used to recover lost sensor signals. To this end, a two-stage approach is developed to estimate the Functional Principal Component (FPC) scores and construct a functional regression model. The flexibility of FPC based models furnishes them with substantial potential for sensor recovery in multichannel CM environments. A turbofan aircraft engine simulation study is used to demonstrate the sensor recovery technique.