Metallurgical Lab Case Study:

Cracking of Grade 23 Steel Furnace Wall Tubes

Grade 23 is a creep strength enhanced ferritic (CSEF) steel that was designed to offer similar creep strength to Grade 91 but with lower Cr content and, in the original concept, fabrication without pre- and post-weld heat treatment making the material attractive for the furnace wall tubes of ultra-supercritical coal plants where T12 has insufficient strength and T91 would be too complex to fabricate. Experience gained with T23 has shown that pre-heat is necessary and that post-weld heat treatment should also be performed when the material is employed in “high restraint” applications such as furnace wall tubes. Like other CSEF steels, T23 is very sensitive to heat treatment, and care must be taken to ensure that hard, brittle microstructures do not enter service – particularly in high restraint applications such as furnace wall tubes.

More generally, T23 has been employed in some superheater and reheater tubes of coal-fired boilers and HRSGs. In these loose tube applications (low restraint), the experience with the material has been much better (fewer failures), but occasional issues can occur. Most notably, several HRSGs have experienced T23 tube failures at the tube to header connections due to the combination of local bending loads and carbon migration associated with welding to a higher chromium content material.

This article explores a recent tube failure in an ultra-supercritical boiler with T23 furnace wall tubes. Leaks were found on two membrane-welded furnace wall tubes after approximately 37,000 hours of service. Sections of the two tubes containing the leaks were removed and sent to SI’s Materials Science Center for metallurgical analysis to determine the cause of the leaks.

Visual Examination

Overviews of the furnace sides of the two as-received tubes (Tube A and Tube B) are shown in Figure 1. The plant had circled the location of a leak on each tube. There were no other indications of mechanical or corrosion damage on either tube; the original paint was still present in many areas.

FIGURE 1. Furnace side of the as-received T23 furnace wall tubes. The leak areas are marked with red circles. TOP Tube A BOTTOM Tube B.

FIGURE 2. Tube A: Close views of leak area on the tube outer TOP and inner BOTTOM surfaces. An axially-oriented through-wall crack was present (arrowed).

FIGURE 3. Tube B: Close views of leak area on the tube outer TOP and inner BOTTOM surfaces. An axially-oriented through-wall crack was present (arrowed). In Tube B the crack was covered by an oxide/deposit crust on the inner tube surface.

The tubes were split longitudinally along the furnace membrane weld to reveal the tube inner surfaces. Figure 2 and Figure 3 show close views of the leak areas as seen on the outer and inner tube surfaces. An axially oriented crack was present at the leak area along with the furnace side tube crown in each tube. The cracks were approximately 0.5 inches long. The appearance of the cracks on the two tubes was similar: the crack lengths were more significant on the inner tube surface than on the outer surface, and the cracks were tight, relatively straight, and unbranched. There was no evidence of abnormal corrosion associated with the cracks on either the outer or inner tube surface, although an oxide nodule covered the crack on the inner surface of Tube B (Figure 3).


Metallographic Examination

A transverse metallographic section was prepared through each tube at the center of the visible crack. Figure 4 shows macro views of the furnace side of the tubes. In both tubes, the crack was found to run relatively straight through the tube wall with no branching, and there was a large zone of altered microstructure, centered on the tube crown outer surface, in which the crack was present.

FIGURE 4. Macroscopic overviews of the prepared transverse metallographic cross-sections through the cracks on Tube A TOP and Tube B BOTTOM. Both cracks were associated with a local zone of altered microstructure with relatively high hardness (325 to 340 HV); in Tube A local melting appeared to have occurred. Vilella’s etch.

FIGURE 5. Close view of crack in Tube A at approximately the mid-wall location, showing the transition from the being oxidized toward the ID TO BOTTOM to being unoxidized toward the OD TO TOP.

In Tube A, the outer tube wall at the crown had a columnar grain structure, and the profile of the outer tube surface was irregular, indicating that local melting had occurred. The heat-affected zone (HAZ) associated with the melting extended through the tube wall. The crack itself was more open on the inner half of the tube and tighter on the outer half of the tube; the transition points roughly corresponded to the transition from the HAZ to a re-melted structure. As shown in Figure 5, the inner length of crack was significantly oxidized, whereas there was very little visible oxidation on the outer length of the crack. The thickness of the oxide in the crack was approximately the same as found along the inner tube surface.

There were no microstructural features or significant corrosion attack observed at the tube inner or outer surfaces which might have led to the cracking, other than the gross microstructural alteration created by the melting. Figure 6 compares the typical columnar grained, bainitic microstructure found in the melted zone with the equiaxed, tempered bainite structure found on the casing side of the tube, typical of non-affected areas in both tubes and normal for Grade 23 material.

Similar microstructural features and cracking morphology were observed in Tube B. One difference was that in Tube B there were no indications of local melting, but rather the region of altered microstructure consisted of untempered bainite, indicating heating above the lower critical temperature. There was also an area where the original factory paint was present over the altered microstructure, as shown in Figure 4.

Two additional transverse sections were prepared from each tube to determine if the altered microstructure was localized or present over a large area. None of the other sections showed evidence of alteration, so the altered microstructures appear to be localized to the areas of the cracks.

FIGURE 6. Comparison of the columnar-grain, re-solidified microstructure near the OD of Tube A TOP with the normal tempered bainite T23 microstructure observed on casing side of Tube A BOTTOM. Vilella’s etch.

Hardness, Dimensional Measurements, and Chemical Analysis

The Vickers hardness was measured in the microstructurally altered areas and at a mid-wall location on the casing side of the tubes. The average hardness for the altered region in Tube A was 325 HV10; the average hardness for the altered region in Tube B was 342 HV10. The average hardness on the casing side of the tubes was 172 and 174 HV10 for Tubes A and B, respectively, both normal for T23. The high hardness values indicate the presence of untempered bainite and martensite, consistent with the observed microstructural features.

Checks on the dimensions and chemical compositions of the two tubes were performed; these were found to agree with the specified values.


In both tubes, the cause of the cracking appears to be severe local overheating, which will result in high local residual stresses and a relatively hard, untempered microstructure with reduced ductility and greater susceptibility to cracking mechanisms such as stress corrosion cracking and reheat cracking. The most likely cause of very localized heating to well above the lower critical temperature (A1) or melting temperature is a welding arc strike, although it is possible that some other electrical arc strike could have occurred during the manufacture of the membrane welded panels, for example, shorting.

Factory paint was found covering the microstructurally altered area in Tube B so that it appears that the overheating occurred in the boiler shop during the manufacturing of the panels. For Tube A it is not certain that the overheating happened in the factory since the factory paint was not present near the cracks, having been removed either by service exposure or during the investigation of the leaks. Overheating in Tube A may have occurred in the factory, during erection, or at some other later time. The lack of significant plastic deformation in the overheated area indicates that overheating did not occur while the tube was pressurized in service.

Because the inner halves of the cracks were oxidized to a similar extent as the tube inner surfaces, initial cracks about halfway through the tube wall appear to have formed either before or shortly after entering service, with these cracks then subsequently propagating through the remainder of the tube wall. The propagation mechanism is not clear; the outer halves of the cracks were separated, and the microstructural features on opposite sides of the cracks did not match, indicating that the crack faces were eroded by fluid flow once the cracks were through-wall, and the inter- or transgranular nature of the crack path in this area could not be determined. However, there was no evidence of gross plastic deformation or creep cavitation found associated with the crack. In this case, it was recommended that the manufacturer of the waterwall panels be consulted to determine the likely source of the overheating and other possible locations where it may have occurred. These areas could be checked for part-through-wall cracks using ultrasonic inspection techniques.


The root cause of the cracks found in these tubes was severe local overheating which led to both a local microstructural region susceptible to reheat and stress corrosion cracking and to high residual stress levels needed to drive the cracking. This case is a good example illustrating the need for higher manufacturing, handling, and quality control standards when working with CSEF steels versus conventional low-alloy boiler steels (e.g., Grades 11, 12, and 22), since Grade 23 is intrinsically more susceptible to failure modes such as reheat cracking and stress corrosion cracking. A local overheating event such as found in this investigation most likely would not have resulted in through-wall cracking for a membrane panel manufactured from a conventional furnace wall tubing alloy such as T12 (1 ¼ Cr – ½ Mo).

Contact Form