Machine housing as a sensor in service to bearing force estimation
The focus of this PhD topic is to estimate multiple resultant bearing forces of a drivetrain by means of a measurement grid of strain and acceleration sensors on the drivetrain housing. In the state of the art, performing a direct measurement of the bearing forces is done by placing a sensor in the transmission path which is not only very expensive and intrusive, it also changes the dynamics and requires an adaptation of the machine to host the sensor. Virtual sensing offers a solution to estimate the force rather than measuring it. The solution involves deploying a model of the machine housing and an estimation technique to derive the bearing forces based on the model and the sensor grid. As typically an extensive amount of bearings are present in the system, the problem becomes under-determined and additional information is required to obtain a reliable bearing force estimate. Therefore, compressive sensing techniques will be deployed that represent the forces by a limited number of shape functions selected from a dictionary. The challenges in this PhD involve amongst others; selection of an appropriate sensor strategy, devising an accurate dictionary of shape functions and devising measurement principles to increase the accuracy of the estimation by small non-intrusive design adaptations.