Turbomachinery Structural Load Cases
Definition and combination of steady, transient, start-stop, overspeed, aerodynamic, thermal, vibratory and abnormal turbomachinery load cases for structural qualification and life.
Load cases are operating states, not isolated loads
A useful turbomachinery load case represents a physically credible machine state: speed, pressure, temperature field, aerodynamic forcing, torque and support condition at the same instant or operating phase. Maximum values taken independently from different times should not automatically be combined. For example, peak thermal gradient during start-up may occur well before maximum steady temperature, while maximum speed may coincide with a different pressure ratio. Structural qualification should therefore be based on an operating-event matrix rather than a list of unrelated maxima.
Steady operating conditions
Normal continuous operation typically combines design rotational speed, pressure distribution, torque or transmitted power, steady aerodynamic loading and a stabilised temperature field. This case defines baseline stress, radial growth and interface reactions and often supplies the mean stress for high-cycle-fatigue assessment. More than one steady condition may be required across idle, cruise, take-off, full power or different process duties. The most severe structural state is not always the maximum power condition because temperature, pressure and speed may vary differently across the operating map.
Start-up and shutdown
Start-stop cycles are central to low-cycle fatigue because centrifugal stress changes from zero to operating level while temperatures lag behind speed. During a fast start, the rotor may reach high speed before thick sections have heated, producing thermal gradients and stress combinations not present at steady state. Shutdown reverses the sequence and can create a different stress range. The analysis should therefore preserve the time-dependent relationship between rotational speed and temperature, particularly for discs, blades and casings with large thermal inertia.
Overspeed and proof conditions
Overspeed events or tests deliberately exceed maximum normal speed and can govern rotating-component strength because centrifugal loading scales approximately with the square of speed. The required overspeed level and acceptance criteria come from the applicable design basis or programme and should not be replaced by a generic percentage. A static overspeed model may be sufficient for strength if the event is slow relative to structural dynamics, but transient acceleration, contact or speed-dependent thermal effects may need separate treatment.
Aerodynamic and vibratory cases
Steady aerodynamic pressure creates blade bending and torsion, while unsteady forcing from rotor-stator interaction, wakes, vane passing, distortion or flow instabilities creates alternating stress. High-cycle-fatigue assessment uses the dynamic stress amplitude superimposed on the pre-stressed mean state. Excitation frequencies depend on shaft speed and engine order, so the relevant case is often a speed sweep rather than one operating point. Resonance crossings may be brief yet still significant if forcing amplitude is high or damping is low.
Abnormal and fault events
Credible abnormal events can include compressor surge, stall, rub, foreign-object damage, loss of cooling, bearing malfunction, shaft seizure, blade release or rapid load rejection. These cases are machine-specific and should come from system safety, certification or programme requirements. Acceptance criteria may differ from normal-operation criteria: limited yielding may be permitted in a rare event while containment, safe shutdown or retained function remains mandatory. The structural model should reflect the actual fault sequence and load duration rather than merely multiplying a normal case.
Load sequence and nonlinearity
Sequence matters when contact, plasticity, preload, creep or thermal expansion changes the structural state. A blade root contact solution should normally establish centrifugal seating before applying small vibratory perturbations. A casing with bolted joints may redistribute load as temperature changes. In such problems, algebraically adding independent linear results can be invalid. The model should reproduce the order in which loads develop and preserve any residual stress, contact state or geometric deformation needed for the subsequent event.
Building the qualification matrix
Create a table mapping each event to speed, pressure, temperature field, mechanical loads, duration, number of cycles and acceptance criterion. Identify which cases drive static strength, clearance, LCF, HCF, creep, containment and support loads. This matrix prevents duplicated or missing analyses and makes assumptions visible. It also helps downstream review because every reported result can be traced back to a defined physical operating state rather than a solver file name.
Do not create a 'worst case' by combining unrelated independent maxima unless the design basis explicitly requires that envelope. Preserve physically credible simultaneity.
Source and quality of load data
Each load case should identify the source and maturity of its inputs. Speed may come from control-system limits, pressure from CFD or test, temperature from CHT or thermocouples, and vibratory forcing from unsteady CFD or rig measurement. Preliminary estimates and certified limits should not be mixed without distinction. Record uncertainty and whether a load already contains a factor or envelope. This avoids double conservatism and makes it clear which analysis should be revisited when higher-quality operating data become available.
Acceptance criteria by event
Normal operation, endurance, overspeed and rare fault events do not necessarily share the same allowable basis. One case may be checked against elastic stress and fatigue life, another against plastic collapse, and another against containment or safe shutdown. Assign the failure mode and acceptance criterion when the case is created. This prevents a common late-stage error in which all cases are compared with one generic yield limit even though the programme requires different structural evidence for different event classes.