Suppose a two-vane impeller is tested at several speeds. The torque trace has a large narrow peak in one speed band. After a coupling change, the peak moves even though the hydraulic parts have not changed. The impeller may still provide the periodic excitation, but the shaft train can amplify it at one of its natural frequencies. Record the frequency, speed and phase relationships before declaring that one impeller puts an unacceptable load on the motor.
A coupling is also a spring
A shaft train has rotating inertia on both sides of the coupling. The shafts and coupling twist slightly under torque; their stiffness and the rotating inertias determine torsional modes. Damping controls how strongly a mode responds. At a suitable frequency, a modest repeating torque can produce a large oscillation at the transducer. The measured amplitude can depend on which side of the coupling the transducer occupies. The average torque can remain similar while the dynamic trace looks very different.
In a simple two-inertia model, increasing torsional stiffness while keeping the inertias fixed tends to raise the relative torsional natural frequency. A real test rig has more parts, and local modes or damping can change too. Do not assume that a stiffer coupling always reduces every peak, or that a softer element protects all speeds. The useful starting information is the original and replacement coupling specification, shaft-train diagram, transducer location and actual speed range.
Identify an order, then watch for amplification
For illustration, 2,900 revolutions per minute is about 48.3 shaft revolutions per second. A two-vane impeller can generate a component near twice that rate, about 96.7 hertz. Suppose the bench has a torsional mode near 96 hertz. Passing through that region can magnify the two-per-revolution signal. If a controlled change to the shaft train shifts the narrow region where the amplitude is largest, the result supports a test-rig contribution. It does not make the impeller’s excitation disappear, and it is not a prediction for all two-vane pumps.
Use the tachometer as the speed basis rather than estimating speed from the supply frequency. Record synchronized torque, speed and hydraulic conditions. Track spectral lines against speed to distinguish a shaft order from unrelated fixed-frequency interference. Also plot order amplitude against speed; the order may still follow speed on both sides of the resonance while its amplitude rises only in a narrow range. If practical, compare the result with a second suitable shaft-train configuration or an independent measurement on the assembled pump. Keep fluid, duty, temperature, rotor and mounting comparable.
Protect both the test and the decision
A torsional response can damage a coupling or overload a transducer even if the average is well within range. Have the rig designer approve the coupling, guarding, maximum allowable dynamic torque, speed changes and test plan. Do not deliberately dwell at a severe resonance. Stop when the approved safety criteria are reached. Treat any change in the test rig as a new measurement condition and document it. The data should be comparable enough to explain why the peak moved.
If selecting an impeller, compare the repeatable hydraulic and average load behavior at the required duty as well as relevant dynamic loads on the intended pump assembly. If qualifying the bench, show that its own response does not distort the required measurement or declare the range in which it cannot be used. A frequency peak and its physical cause are separate findings.
Frequently asked questions
Is a peak at twice shaft speed always caused by a two-vane impeller?
No. The order is a clue. Other excitations can share the order, and the shaft train can amplify it.
Can I compare peaks from different coupling types directly?
Only after checking what changed in stiffness, inertia, damping and sensor position. Record those conditions with the result.
Is the average torque enough for an endurance decision?
It answers the mean load question. A large dynamic component can matter to fatigue, connections and instrumentation even when the mean is acceptable.
