Full text
| |
| Author(s): |
Gabriel Yuji Garoli
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | Campinas, SP. |
| Institution: | Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Mecânica |
| Defense date: | 2020-06-19 |
| Examining board members: |
Hélio Fiori de Castro;
Katia Lucchesi Cavalca Dedini;
José Roberto de França Arruda;
Thiago Gamboa Ritto;
Adriano Todorovic Fabro
|
| Advisor: | Hélio Fiori de Castro |
| Abstract | |
Rotating machines are expensive and complex equipment used in a range of industries. Whose model is essential and, even with more than one century of study, it is at constant development. However, the dynamical behaviour of such machines can have a stochastic characteristic and should be added in the model as uncertain parameters. This process of identification and quantification of uncertainties in rotating machines is recent and is not so well established as the deterministic approach. Monte Carlo methods are usually used to calculate the stochastic response, because they are simple to implement and allow the use of algorithms that evaluate the deterministic solution. The convergence is guarantee by a large number of samples, which is a negative aspect for the case of rotating machines, which have complex models that need a large amount of processing time. This work proposes the use of the generalized Polynomial Chaos Expansion as an alternative to approximate the stochastic response. For problems with a small number of uncertain parameters, the Stochastic Collocation can be used. The method can be used to evaluate the expansion coefficients. A deterministic algorithm can be used, as in the Monte Carlo method, but a smaller number of samples is needed due to the polynomial approximation. After the evaluation of the expansion coefficients, statistical information can be calculated. A sensitivity analysis can be performed as well, which allows the analysis of the effect of the input uncertainties on the output uncertainties and to evaluate which ones have more influence. The generalized Polynomial Chaos Expansion can also be used in the Bayesian Inference. The expansion turns the equation of Bayes Theorem into a product between a polynomial series and known probability density functions. The inference is used to identify the fault parameters in rotating machines. This is a generic method, which needs only the rotor system of interest and fault models. In this work, the application of the generalized Polynomial Chaos Expansion to evaluate the stochastic response of rotating machines is studied. For the cases in which the bearing parameters uncertainties are assumed, the equivalent dynamic coefficients, the fluid induced instability and its sensitivity to the parameter uncertainties are analyzed. For the cases in which the fault parameters are assumed uncertain, the frequency and time response of rotor systems are considered, and its sensitivity to the fault parameters uncertainty is analyzed. The Bayesian Inference with the polynomial expansion is also used to identify the fault parameters and active magnetic bearing parameters. Results are compared to deterministic method results and the inference solved by the Monte Carlo via Markov Chain method results. With the identified parameters, responses are evaluated and compared with experimental data (AU) | |
| FAPESP's process: | 16/13223-6 - Uncertainty estimation and quantification applied to fault models of rotating machines |
| Grantee: | Gabriel Yuji Garoli |
| Support Opportunities: | Scholarships in Brazil - Doctorate (Direct) |