Identification of active magnetic bearing parameters in a rotor machine using Bayesian inference with generalized polynomial chaos expansion

Rotating machines are widely used in industry. They are composed of rotative components such as shaft and blades, which are connected to a static support structure by bearings. Rolling bearings and fluid lubricated bearings are commonly used for this function. However, in the last decades, active magnetic bearings (AMB) have gained importance in some applications. These bearings can support the shaft of such machines without contact and apply active control through electromagnetic forces. On the other hand, uncertainties are inherent to engineering systems and they should be quantified to obtain better models. Bayesian inference is an interesting option to identify or update the probability distributions of a random variable. Monte Carlo via Markov chains is usually implemented to solve the inference, but its processing time can be long. By using generalized polynomial chaos expansion, the solution process is accelerated. This work aims to identify the AMB parameters and unbalance force. After the identification, the stochastic response is evaluated and compared with experimental data from a test rig supported by AMB. The robustness of the identification is evaluated by inserting noise in the signal. A sensitivity analysis is performed through Sobol indices to evaluate if the AMB uncertainties should be considered in future analyses.