Fault Diagnosis and Identification of Electro Mechanical Systems - Methods for Closed Loop Systems.

We are on the verge of the fourth industrial revolution also known as industry 4.0[1]. The goal of industry 4.0 is to increase the incorporation of sensor information into decision making for machinery. This in turn increases the popularity of feedback control, and by extension, closed loop schemes. With the increasing popularity of closed loop control it is important that the impact of the feedback loop is handled appropriately. Because of the feedback loop, signals that might normally be uncorrelated are suddenly not and assumptions often used for identification and fault diagnosis schemes are no longer realistic to achieve.
The thesis aims at introducing the reader to design methods with proper handling of noise for closed loop systems. In order to achieve this goal, it is investigated how to transform a closed loop identification problem into an open loop identification problem. Such a transformation is already well known, however the excitation signal design is not intuitive when applying such a transformation. The shape of the excitation signal is of paramount importance for the quality of the identified model. By making the design of the excitation signal more intuitive, it should be possible to increase the quality of identified models.
Another interesting closed loop application is fault diagnosis. More and more systems will be part of a closed loop scheme in the future in accordance with industry 4.0. Often, systems are designed without sensor redundancy, and with disturbance rejecting controllers. Methods which are not limited in isolability due to sensor redundancy, and which decouple the effect of the disturbance rejecting controller, are therefore of huge interest. Active fault diagnosis obtains the required information through a known excitation signal instead of the sensor redundancy. Design of detectors based on active fault diagnosis can therefore make fault diagnosis possible for systems where installation of extra sensors are too cost demanding.
The methods were developed with a piezoelectric rotor-bearing application in mind. The bearing is using air as the lubricant between the bearing and the shaft and is therefore referred to as a gas bearing. Gas bearings have relative low damping compared to high friction bearings such as ball bearings. Feedback control is therefore employed to increase the damping of the Gas Bearing. This makes Gas Bearings a prime example of technology following with the industry 4.0 standard.