Page 349 - Failure Analysis Case Studies II
P. 349

334

            2.2.  Visual inspection

              To make a detailed observation of the cracked specimen, specimen D was cut in a direction
            perpendicular to the primary crack as depicted in Fig. 2, generating two small sub-specimens, D,
            and Dz. With the specimen D, the primary crack was opened to fracture with intention of observing
            the crack surface. The crack tip region of the specimen D2 which was not opened was observed by
            a microscope. Figure 3(a) and (b) show the inner and outer surfaces of the other specimen DZ.
              All other failed tube samples as well as specimen D showed many locally thinned areas at the
            inside of the tubes. As observed from Fig. 3(b), which shows the inner surface of the specimen D2,
            thinning was fairly localized and occurred irregularly. It should be  noted that thinning did not
            occur  gradually with  distance from  the  burner.  From  this  observation,  we  can  rule  out  the
            possibility of erosion damage by solid particles included in the burner combustion gas as the main
            cause of tube thinning. Hence, it may be predicted that local oxidation or local corrosion is the
            main causes of thinning.
              A thick oxide scale indicated by an arrow in Fig. 3(b) was attached at the locally thinned area.
            Similar oxide scales were also observed in a number of local oxidation pits. Most of the oxide
            scales contain several cracks formed in random directions. The oxide scale in Fig. 3(b) is enlarged
            in Fig. 4. Cracking of the scale must be mainly due to the difference in thermal expansion coefficient
            between the oxide scale and the tube metal on which the scale is attached.
              Generally at the initial stage of oxidation, an oxide film forms on the metal to prevent further
            oxidation. Therefore, when  a  stabilized oxide film is formed on the  surface, the  resistance to
            oxidation under high temperature conditions is increased. As for the radiant tube of this failure
            analysis, the  high content  (25%)  of  Cr enables the  formation of  a  Cr203 film  that  increases
            resistance to high temperature oxidation. If this oxide film is removed, the base metal of the tube
            will undergo repeated oxidation which results in continuous thickness reduction [2]. When cracking
            occurs in the oxide scale, as illustrated in Fig. 4, the crack tip area loses the protective effect of the
            oxide film and the base metal beneath the crack tip will be repeatedly oxidised. As a result, a sharp
            oxide spike will  be  gradually formed in the base metal under  the oxidation layer where oxide
            cracking occurred. Observation of the opened fracture surface of specimen D1 showed an oxide
            layer extending to the crack tip.


            2.3. Metallographic observation

              To confirm the forementioned crack initiation mechanism, a small metallographic sample was
            taken from the location of a local oxidation pit of specimen D2 where the oxide scale is attached,
            as shown in Fig. 4. It was mounted and the scale was ground out until the tube metal right beneath
            the cracked oxide scale appeared. An observation was made to see if  any tube metal cracking
            occurred at the location beneath the oxide scale crack. The specimen preparation  procedure is
            shown in Fig. 5. From this observation, a small crack of 4-mm length was found in the base metal
            right beneath the oxide scale and this crack was oriented in the same direction as the crack in the
            oxide scale. Each crack tip area of the tube metal was observed by a scanning electron microscope
            and is shown in Fig. 6(a) and (b). The crack tip was not sharp but looked like a blunt notch. Also,
            Fig. 6(b) shows that the matching crack surfaces are separated from  one another. These two
            observations confirm that the crack was not initiated by mechanical loading such as fatigue load
   344   345   346   347   348   349   350   351   352   353   354