CS Energy, a Queensland Government owned energy provider, whose asset portfolio comprises the Kogan Creek, Callide and Wivenhoe power stations, identified unstable steam temperatures entering the reheater collection headers in the Kogan Creek 750 MW supercritical boiler.
Knowing that fluctuating temperatures can cause thermal fatigue damage, CS Energy engaged Aurecon’s structural integrity team to perform a detailed investigation. The aim was to determine if CS Energy should be concerned for the structural integrity and future life of the 9Cr-1Mo-V (P91) reheater headers.
Unstable steam temperatures have been acknowledged by industry as a significant contributor to thermal fatigue damage and cracking, especially in the ligaments between tube holes on the internal surface of collection headers. Stresses are generated by through-wall temperature gradients, which form when the metal surface is heated or cooled relative to the average wall temperature. Experience has shown that the severity of the stress is caused by variables such as tube-hole spacing and ligament dimensions, temperature ramp rate, the material’s thermal expansion coefficients and header design thickness.
The cracking of ligaments has been a leading cause of header replacements in superheater headers, reheater headers, economiser headers in coal and gas fired boilers, as well as High Pressure, Intermediate Pressure and Low Pressure superheaters and evaporator headers in heat recovery steam generators.
Aurecon’s approach to investigating the risk of ligament cracking was to:
Aurecon engineers accessed CS Energy’s data collection systems (PI database) to obtain thermocouple data from key locations associated with the headers. The key locations were tube metal temperature, header steam temperature and header buried (mid-wall) and surface thermocouples. Data was obtained at one minute intervals for a number of two-day blocks of plant operations at different periods since commissioning.
A comprehensive finite element analysis (FEA) model was constructed which consisted of eight inlet tubes, a header stub, and a section of the manifold. The model incorporated temperature-dependant material properties (such as thermal conductivity, elastic modulus and expansion coefficient). The model was constructed in ANSYS v14 using 3D solid geometry with a numerically converged mesh comprising second order elements.
Buried and mid-wall metal thermocouples had been previously installed in the header during plant construction, and these were used to thermally calibrate the model and validate the correlation between steam temperature and the metal temperature predicted by the finite element heat transfer model. Calibration allowed identification of the heat transfer (surface film) coefficients to be defined.
Because the steam data showed highly unstable and irregular temperature spikes, Aurecon developed a method to simulate the transient behaviour across an array of temperature ramp rates, and severity of temperature spikes. The resulting stresses at key locations produced a complex 3D surface response when plotted against the two critical input parameters – temperature change and temperature change rate. The highest stresses were confirmed to occur in the tube ligaments.
To simplify the processing of the many tens-of-thousands of ‘steam temperature versus time’ data points obtained from CS Energy’s PI system, a non-linear power-law equation was developed to correlate steam temperature ramp rate and overall temperature change to ligament stress. This equation was then applied to the raw temperature data, which enabled ligament stress to be estimated as a function of time. This enabled the stress ranges to be determined for all thermal cycles based on the file of steam temperatures downloaded from the plant Distributed Control System computer. The stress relationship takes the form of:
Coefficients A, B, C and exponents n and m were identified based on interrogation of the finite element model’s results. The errors produced when comparing the results of the equation (red dots in Illustration 1) to the finite element model (solid lines in Figure 1) were negligible for all temperature steps and ramp rate combinations.
The resulting (large) table of stress was then processed by Rainflow counting algorithms, which determined the statistical distribution of the magnitude of all stress ranges and the occurrence frequency, both of which are required for a fatigue crack initiation analysis.
Fatigue analysis was performed using the 'Coffin-Manson' method, with a mean stress offset (Morrow adjustment). The form of the equation is shown below as Equation 2. The solution to the equation (Nf) is iteratively solved and is considered to represent the number of cycles to initiation. Coefficients to Equation 2 were calculated using the ‘Universal Slopes’ method, but making adjustments for high temperatures, where fatigue crack initiation was expected to occur approximately 20 000 times faster than at room temperature.
The creep model employed an integrated time-at-temperature approach based on the data collected from the CS Energy PI system. The creep damage model employed was the MPC/Omega model from API-579-1/ASME FFS-1, as the traditional Larson Miller Parameter (LMP) model for P91 was considered inaccurate at the stresses experienced in the ligament (<50 MPa).
Figure 2 - Creep-Fatigue ‘knee plot’
For each damage mechanism, a Damage Fraction was calculated for both Fatigue and Creep. The damage fractions could then be plotted on the Creep-Fatigue ‘knee plot’ shown in Figure 2), which enabled assessment of combined creep-fatigue damage.
A similar method to that presented here was applied to other locations within the header assembly, namely the header-to-manifold weld, and the stub tubes entering the headers. The analysis approach was further adapted from the method presented above, which accounts for high frequency temperature changes resulting in dynamic through-wall thermal gradients, as well as longduration temperature excursions that last for many days, and produce sustained stresses that result in an additional creep and fatigue fraction.
Analysis of the data suggested the time required for initiation of ligament cracks significantly exceeded the expected life of the plant and, as such, it is unlikely that thermal fatigue cracking would develop as a result of the constantly variable steam temperatures. However, it is still considered prudent to perform periodic inspections of the ligament to satisfy the requirements of AS/NZS 3788 “Pressure Equipment – In service inspection”.
Aurecon’s investigation enabled CS Energy to understand the risks associated with certain plant operations, and gave confidence that in-service damage is unlikely to occur.
This article first appeared in Engineers Australia magazine, May 2013.