Machine Learning Techniques for Automatic Sensor Fault Detection in Airborne SHM Networks

Good data is key to the success of a structural health monitoring (SHM) program, and modern data acquisition systems allow for reliable, high fidelity data capture. Unfortunately SHM programs are often hindered by undetected sensor and wiring problems resulting in invalid data. Many authors have identified sensors as the weakest link in an entire SHM system, where the transducer and the transducer/structure interface can ‘‘make or break’’ an SHM system. Choosing long-life sensors is one approach for addressing this problem, however these high cost and high specification sensors are rarely economically viable. For airborne monitoring programs, there is an expectation that sensors will be replaced over time, and that dedicated data analysts will be available to spot subtle signs within data which indicate the onset of sensor faults. This approach does not scale well for large fleets and does not allow for robust automation. In this paper, we present a new sensor diagnostic approach based on ‘‘Machine Learning’’ techniques. These automated techniques allow for reliable measurement using practical cost sensors, installed in extended duration monitoring programs. Machine Learning fault detection techniques not only detect the obvious catastrophic sensor errors, but also the more subtle sensor issues that can easily go undetected for long periods of time, e.g. strain gauge de-lamination, accelerometer de-calibration, loose/dry solder joints, adhesive degradation etc. A key enabler for airborne SHM systems is the automation of human data analysis, allowing systems to operate reliably without intervention for many years. This paper explores how Machine Learning techniques can be used to detect subtle signs of sensor/wiring faults within captured data, essentially automating the experience of human analysts who must ensure captured data is good. In today’s airborne data acquisition systems, built in diagnostics are typically limited in scope to a single subsystem and in functionality to a small number of failure scenarios. The Machine Learning techniques explored in this paper offer new expanded diagnostic capabilities which go beyond individual electronics units to provide system-wide diagnostics, encompassing all critical parts of an SHM system.