Fault Detection of a Rotating Shaft by Using the Electromechanical Impedance Method and a Temperature Compensation Approach

Condition-based maintenance systems that use sensor networks for damage detection in rotating machinery have been evolving constantly. Such strategies aim at detecting the presence and severity of damage on a statistical basis. The aim of this contribution relies on the correct detection of incipient faults in rotating shafts by using a real-time impedance-based structural health monitoring method, with a low-cost portable device. This technique monitors changes in the electric impedance of piezoelectric transducers, acting simultaneously as actuators and sensors, which are bonded to the host structure. With the use of damage metrics, these changes can be quantified so that the presence and severity of damage are detected. This is possible since the electrical impedance of the sensor is directly related to the mechanical impedance of the structure. However, the frequency response functions resulting from this method are susceptible to environmental and operational conditions that should be accounted for to avoid false diagnostics. Consequently, a temperature compensation technique is proposed based on a hybrid optimization method associated with different damage metrics. Additionally, a statistical model is used for threshold determination based on the statistical process control method. Experimental results show that an incipient fatigue crack associated with bearing wear was successfully detected with a probability of detection above 95% confidence for the majority of sensors used.