CRACK PARAMETER IDENTIFICATION USING INVERSE ANALYSIS OF ON-LINE VIBRATION MEASUREMENTS

摘要

Crack size and location of a slow rotating drum with a transverse crack are estimated using an inverse analysis technique. Crack parameter estimation is achieved by minimizing an error function quantifying the difference of the measured and modeled vibration signal. Vibration measurements are made using eddy-current proximity probes. The modeled vibration signal is the result of a 3D solid Finite Element analysis. Model inputs consist of the applied load, crack size and location. Measured and modeled vibration signals are Fourier transformed and comparison of the Fourier coefficients yields the basis of the error function. The error function is minimized using a quadratic optimization algorithm until the model inputs represent the true crack parameters. A variety of combinations of crack size and location have been investigated. The results show the influence of various error functions and identifies the most important vibration components. For a known crack location the inverse analysis yields an estimated crack size within 10% of the true crack size. For unknown crack location and size the results are a combination of estimated crack size and a probability measure as a function of location. The method has been implemented on a test rotor where the analysis is performed while the rotor is in operation.