Resonance inspection uses the natural acoustic resonances of a part to identify anomalous parts. Modern instrumentation can measure the many resonant frequencies rapidly and accurately. Sophisticated sorting algorithms trained on sets of good and anomalous parts can rapidly and reliably inspect and sort parts. This paper aims at using finite element-based modal analysis to put resonance inspection on a more quantitative basis. A production level automotive steering knuckle is used as the example part for our study. First, the resonance frequency spectra for the knuckle are measured with two different experimental techniques. Next, scanning laser vibrometry is used to determine the mode shape corresponding to each resonance. The material properties including anisotropy are next measured to high accuracy using resonance spectroscopy on cuboids cut from the part. Then, the finite element model of the knuckle is generated by meshing the actual part geometry obtained with computed tomography. The resonance frequencies and mode shapes are next predicted with a natural frequency extraction analysis after an extensive mesh size sensitivity study. The good comparison between the predicted and the experimentally measured resonance spectra indicates that finite element-based modal analyses have the potential to be a powerful tool in shortening the training process and improving the accuracy of the resonance inspection process for a complex production level part. The finite element-based analysis can also provide a means to computationally test the sensitivity of the frequencies to various possible defects such as porosity or oxide inclusions, especially in high stress regions that the part will experience in service.