Model based diagnostics and prognosis system for rotating machinery

Abstract

A PC-based automated mechanical vibration diagnostic and prognosis system for rotating machinery is under development by integrating AI-expert system-based interpretative capabilities with rotor dynamics based modeling and numerical optimization techniques. Presented here are the details involved while generating a rotor dynamic simulator: A turbine-generator is properly modeled using the finite element approach to compute steady-state response. The bearing stiffness and damping properties are computed by numerically solving the Reynold's equation. An optimizer software has been interfaced with the rotor dynamics code for the solution of nonlinear unconstrained function minimization. A finite element based closed form approach has been adopted to compute the gradients of the objective function. The model is perturbed by the optimizer to match the results with simulated field measurement. The model based optimization technique has been demonstrated on a 120 mass, 6 bearing rotor system. After optimization, the responses and mass unbalance converge to their goal (field) values. To date, feasibility studies show encouraging results to differentiate between mass unbalance and misalignment for realistic systems using this methodology.