Process Compressor Re-design

FEA stress and vibration analysis was performed for three process compressor impellers installed in a highly erosive service. The objective of the analysis was to design a rerated impeller that would result in accurately predicted peak stresses with at least a factor of safety of 2.0 below both the tensile yield strength and the conservatively estimated fatigue endurance limit for the selected impeller material. In addition, optimal shaft attachment pressfits and bolt-down torques were determined. Finally, all natural frequencies that were close to potentially strong excitation frequencies were investigated.

This investigation included evaluation of the shape of the excitation force distribution in space, such as would be created for example by the presence of diffuser vanes. This shape was compared to the number of nodal diameters associated with a potentially resonant natural frequency. Only if shapes as well as frequencies matched on the basis of direct or sum-and-difference excitations was a natural frequency considered potentially resonant. If it was considered potentially resonant, then the structural performance was further evaluated, and would be judged acceptable only if either the natural frequency in question was lightly excited compared to the available damping, or if it possessed a margin of at least 10 percent relative to the excitation frequency.

An acceptable configuration to avoid structural fatigue was determined for each impeller, as well as appropriate shaft attachment fits and torques to resist loosening by running torque and combined thermal and centrifugal deflections. Stresses were well within acceptable limits. The highest stresses were in the impeller backface and bore, which is typical, and were substantially below the material yield and fatigue strengths everywhere, so that cracking was determined to be unlikely, even in the case of the severe erosion damage which was expected in this service. On the vibration Campbell Diagram (which plots natural frequencies and excitation frequencies versus speed), all anticipated excitation force frequencies were shown to miss significantly excitable natural frequencies by more than the 10% margin.

A new design of motor-driven shop compressor in France was needed to run reliably and quietly. MSI analyzed the casing and shaft/rotor assembly (both laterally and torsionally) for fatigue under expected worst-case variable loading. A fatigue problem in the casing discharge end corner fillet was identified under circumstances of rapid load cycling. The manufacturer at first did not agree that the analysis was realistic, but agreed to perform a shop test under cyclic conditions which were worst-case, but possible in many services. After accelerated cycling that represented several months of operation, a fatigue crack occurred, and a through-crack punctured the compressor casing, causing an explosion hazard. The design was revised per the original recommendations, and performed without structural hazards.