Can anybody update me, what is the acceptance criteria for torsional vibrations for typical centrifugal compressor train (motor driven with step up gearbox, approx. 10 MW rating) ?

You need an anaysis to determine the stresses (assuming you have reason to know the operating torsional deflection shape). Also, you probably need more than one measurement.

What is the concern? Shaft/coupling stresses - need operating deflections shape; a rigid body torsional deflection is a rotation that has no bending stress associated with it. Blade resonances - rigid deflections can affect these.

Bill, Thanks for reply. Concern is, we had/have torsional vibration issues. After plant is operational, it is very difficult to resolve torsional issue as torsional vibration and stress measurements needs outage for 4 - 5 days. Also there is little bit confusion on how much vibration / stress value is allowed. (1) If we have some predefined acceptance criteria, we can better tackle the issue on test bed itself.. (2) It is possible that we don't know presence of torsional vibration untill some parts fail.

Acceptance criteria should be based on a rotordynamics/torsional study with test measurements. Without knowing the relative torsional displacements or probable displacements from the analysis, one cannot assign stress values. Measurements will calibrate a transient analysis,also if one is concerned with starts (low cycle fatique, where the cycles are counted and measured in the analysis)

You need a rotordynamic analysis to help determine this.