Repetitive failures on exhaust duct.
Initial measurements ruled out tentative hypotheses related to pure structural excitation and excitation by noise. Detailed FEM, CFD and coupled FSI analyses gave a clear picture of the affect of flow on the structure.
The investigation showed that large fluctuations in the flow excited the natural frequencies of the walls of the exhaust duct, i.e. structural resonance excited by large fluctuations in flow.
A redesign of the exhaust duct based on the evaluation from the fluid-structure interaction analysis. This reduced fluctuations in the flow, raised the lowest natural frequencies of the duct walls and increased the damping, hereby solving the resonance problem.
Repetitive fatigue failures on the Åsgard B Platform led Statoil to invest in a full mechanical failure investigation and a redesign of their gas turbine exhaust ducts.
Detailed measurements combined with computational modelling and expert analyses showed that the failures were caused by a structural resonance, excited by flow. A thorough evaluation of the new design proposals and a subsequent design improvement meant that vibration levels in the final exhaust duct were reduced by a factor 30.
For some time, the Åsgard B Platform in the Norwegian Sea had experienced serious and repetitive fatigue failures on the exhaust ducts. After one very serious incident, where a piece broke off the exhaust duct wall venting 500 degrees hot exhaust gas to the surroundings, Statoil decided to redesign the exhaust duct completely to prevent future failures.
As part of Statoil’s troubleshooting team, we carried out a mechanical failure analysis to identify and describe the physical mechanism at the root of the problem and to ensure that the new design wouldn’t experience similar failures. For this project, we took the failure analysis through five interconnected stages in just four short months: initial measurements, modelling and analyses, design evaluation, design improvement, and final measurements.
Initial measurement campaign
The nature of the failure indicated that it was caused by excessive vibrations in the duct walls. Our job was to find out what had caused the duct to vibrate. A number of tentative failure hypotheses were proposed, relating to structural integrity, direct excitation by flow and excitation by pressure waves (sound) generated in the gas turbine. These hypotheses were used to motivate a field measurement campaign on the exhaust duct.
The measurements ruled out the hypotheses related to pure structural excitation and excitation by noise as the frequencies of the duct vibrations didn’t match the frequencies we would expect to have seen if these hypotheses were valid.
This left us with the hypothesis that the exhaust duct vibrations were excited by fluctuations in the flow at low frequencies exciting a natural (resonant) frequency of the duct walls.
Fluid-Structure Interaction modelling and analyses
The measurements had exposed a very complex problem, and to solve it we needed a deeper understanding of the physical mechanisms behind it. First of all, we wanted to model the exhaust duct to know how the flow and structure were behaving in the current design.
First, we carried out a set of 3D computational fluid dynamics (CFD) analyses, both steady and unsteady, of the flow in the exhaust duct. These gave a clear indication of large unsteady fluctuations at relatively low frequencies. Second, we used the finite element method (FEM) to model and analyse the structure of the exhaust duct in relation to stress distribution, mode shapes and natural frequencies. This showed that part of the duct walls had low natural frequencies within the same range as the frequencies of the unsteady flow fluctuations.
And third, we combined these two models in a coupled, fluid-structure interaction (FSI) calculation. By using this technique we were able to transfer the unsteady fluid loading onto the vibrating duct structure, which gave us a clear picture of exactly how and where the flow affected the structure. We then carefully verified and calibrated the FSI model with the measurement results to make it as accurate as possible.
Our model confirmed our initial hypothesis that large fluctuations in the flow were exciting natural frequencies of the duct walls. The consequent high vibration levels led to corresponding high levels of stress, which in turn caused the fatigue failure.
Having created a model that accurately predicted how the flow and structure behaved in the current exhaust duct, we then used this to predict how they would behave for the new design proposals to ensure that the chosen design would in fact prevent future failures. Statoil was considering two different design proposals for the new exhaust duct. Based on the information from our model, we were now able to evaluate the design proposals and recommend the one best suited. In the end Statoil chose the Mjørud design as this reduced both the large and unsteady fluctuations in the flow and showed much fewer problematic low frequency structural modes. However, our analysis had shown a low frequency resonance problem that required further design modifications of the lower part of the exhaust duct.
We considered several design features to solve the low frequency resonance problem. Our main focus was to increase the lowest natural frequencies of the duct walls so they wouldn’t be excited by the low frequency pressures created by the unsteady fluctuations in the flow. This was done by introducing a top circular section of the lower transition cone. We also discovered a new way of improving the structural damping, namely by compressing the thermal insulation material that was used on the cone. This reduced the amplitude of the vibrations and thereby helped to reduce the stress on the duct walls.
After evaluating and improving the final design, the exhaust ducts were finally installed on the Åsgard B Platform. To make sure that the new design was in fact behaving as predicted, we carried out a final measurement campaign on the exhaust ducts, similar to the initial measurement campaign. The results showed that not only had the damping improved, but there had also been a very large reduction in the vibration levels and, most importantly, the vibration levels at the natural frequencies were dramatically reduced by a factor of more than 30.