Welding Procedures and Type IV Cracking Tendency - an Experimental Study J.A. Francis School of Materials University of Manchester V. Mazur Manufacturing and Materials Technology CSIRO Australia H.K.D.H. Bhadeshia Materials Science and Metallurgy University of Cambridge J.A. Francis School of Materials University of Manchester V. Mazur Manufacturing and Materials Technology CSIRO Australia H.K.D.H. Bhadeshia Materials Science and Metallurgy University of Cambridge
Background Part of a collaborative project involving CSIRO Australia University of Cambridge. Concerned with failure of welds in 9-12 wt % Cr ferritic power plant steels. If implemented, these steels enable higher steam temperatures and pressures, greater thermodynamic efficiency, when compared with 2.25Cr-1Mo grades. Aim: to develop technologies, procedures that ameliorate type IV cracking. Part of a collaborative project involving CSIRO Australia University of Cambridge. Concerned with failure of welds in 9-12 wt % Cr ferritic power plant steels. If implemented, these steels enable higher steam temperatures and pressures, greater thermodynamic efficiency, when compared with 2.25Cr-1Mo grades. Aim: to develop technologies, procedures that ameliorate type IV cracking.
Features of Type IV failures in 9-12 Cr Steels During service, localised void formation in FGHAZ/ICHAZ Type IV limited creep life significantly lower than for parent plate. Type IV failures predominate when applied stress is below a threshold. During service, localised void formation in FGHAZ/ICHAZ Type IV limited creep life significantly lower than for parent plate. Type IV failures predominate when applied stress is below a threshold. Parent material austenitised at o C, tempered at o C Intended service temperatures of ~ 600 o C ICHAZ has lowest hardness after PWHT Rupture location tends from ICHAZ to FGHAZ as stress decreases Micrographs: V. Karthik et al., Welding Journal, 81 (12), 265s, 2002.
Analysis of Published Cross-Weld Stress-Rupture Data Most type IV studies have focused on the metallurgy. Neural networks can capture mathematical relationships when physical models do not exist. Neural networks can also perceive the relative importance of each input variable. In a Bayesian framework, the dangers in extrapolating non- linear functions are reduced. Most type IV studies have focused on the metallurgy. Neural networks can capture mathematical relationships when physical models do not exist. Neural networks can also perceive the relative importance of each input variable. In a Bayesian framework, the dangers in extrapolating non- linear functions are reduced. Why Bayesian Neural Networks?Typical Network Structure
The Database VariableMinimumMaximumVariableMinimumMaximum C wt. % Normalising Temp. ( o C) N Normalising Time (h)0.52 B00.003Tempering Temp. ( o C) Cr8.4512Tempering Time (h)16 Mo Heat Input (kJ/mm) V Preheat Temperature ( o C) Nb Preparation Angle (deg.)045 W02.21PWHT Temperature ( o C) Mn PWHT Time (h)0.258 Si Internal Pressure Test? (0/1)01 Cu03Test Temperature ( o C) Ni Test Duration (h) Al Rupture Stress (MPa)40150 53 type IV failures included in database. Overambitious set of variables can limit data available for analysis.
Results Significance of Input Variables Trends
P91 pipe welds, OD 356 mm, wall thickness 51 mm Root pass + hot pass : GTAW Filling passes: FCAW Filler: 1.2 mm Supercore F91 PWHT: 760 o C for 2 hours Creep Testing
Parameters varied: heat input, preheat temperature, joint preparation log (10) of creep life (h) Stress (MPa) 1.6 kJ/mm, 10 o, 350 o C 1.6 kJ/mm, 0 o, 250 o C 1.6 kJ/mm, 90 o, 250 o C 1.6 kJ/mm, 30 o, 250 o C 0.8 kJ/mm, 30 o, 250 o C 2.4 kJ/mm, 30 o, 250 o C
1 mm Typical rupture surface corresponding to 0.8 kJ/mm, 250 o C, 30 o
Tests in Progress 5 further tests being initiated (“follow-up programme”) Aim to separate effects of preheat temperature and joint preparation Initial programme 250 o C 350 o C 250 o C Follow up programme 250 o C 350 o C
Conclusions There is scope to improve “resistance” to type IV cracking through optimisation of welding procedures. The weld heat input does not have a significant influence on propensity for type IV failure, which is in agreement with neural network predictions. The joint preparation angle has a significant influence on type IV creep life, but the mechanism remains unclear. Higher preheat temperatures have been predicted to improve resistance to type IV failures. Preliminary results suggest this is plausible but further tests are required to confirm effect.