INVESTIGATION INTO THE FAILURE OF A GAS WASHING FAN ON AN AMMONIA PLANT

 

 

J.C.Guild: Southern African Institute of Welding, Johannesburg, South Africa

 

 

Keywords: Failure investigation, Corrosion, Fatigue, Corrosion fatigue, Brittle fracture, Cracking, Fan, Martensitic stainless steel, Impeller, EN56B

 

 

ABSTRACT

 

The investigation into the failure of a washing fan impeller, which lead to a fatality, is described. A number of mechanisms were involved in the failure, including corrosion fatigue, hydrogen imbrittlement and brittle failure. The brittle failure was as a result of the martensitic stainless steel’s heat treatment being incorrect.

 

 

1. INTRODUCTION

 

At 17.55 hrs on 14th May 1985 a Gas Washing Fan failed catastrophically. A piece of one of the back plates of the fan flew through the roof of the gas washing fan house, over a section of the plant and came down through the asbestos roof of the compressor house, resulting in a fatality. Plant engineering staff recorded the locations where other pieces of the fan were found after bursting out of the cast iron casing.

 

Gas samples were taken immediately both upstream and downstream of the fan to check for oxygen content. These revealed normal oxygen content and a process gas explosion has been discounted as a reason for the fan failure.

 

The gas washing fans are used to draw gas from the raw gas holder for delivery to the sulphur boxes and synthesis plant. The gas arrives at the fan at a nominal temperature of 45°C and has the following basic composition:-

 

H2:                     43%

CO:                   34%

CO2:                  19%

Ar:                     0.1%

CH4:                  0.6%

N2:                     1%

H2S:                  0.3%

 

Water is introduced into the gas stream at the Lymn washers upstream of the raw gas holder and is injected at the fans.

 

The Gas Washing Fan had been overhauled prior to this last service campaign. It was commissioned on 10th May but only made operational on 11th May. It is understood that at around 15.30 hrs on 14th May the fan had been checked and was considered to be running smoothly.

 

The fans have a single stage double impeller of rivetted construction. They operate at 2950 rpm delivering 12000 cubic metres of gas per hour. Figure 1 shows an assembled rotor.

 

Text Box:  
Figure 1: General view of assembled 42” rotor illustrating double impeller construction.
 

 


Original materials of construction are shown on the manufacturer’s drawing as follows:

 

Backplates, Shrouds, Blades: En56B with Brinell Hardness between 174 to 187

 

Rivets, Stay Rods, Dist. Pieces, Nuts, Taper Washers: FDP Quality Stainless Steel

 

It should be noted that FDP stainless steel is a Firth Vickers trade name for a grade of titanium stabilised austenitic stainless steel similar to AISI 321 type.

 

In early 1971 a replacement drawing was made in which the material of construction was changed to AISI 304L grade stainless steel for all component parts. This material has better corrosion resistance than En56B with lower strength properties. The fan which failed is one of four in operation. The other fans have been shown to be made in austenitic material similar to AISI 304L. As they have performed satisfactorily for many years it is evident that the tensile properties of AISI 304L are satisfactory for the duty.

 

 

2. INVESTIGATION

 

2.1 Visual Examination

 

Half of the double impeller had remained intact on the shaft (see fig. 2) whereas the other half had completely disintegrated.

 

Text Box:  
Figure 2: Shows rotor after failure of half of impeller.
 

 


2.1.1 Failed Impeller

 

The pieces of the failed impeller were reassembled into their various positions (figs.3 and 4). It will be noted that sections of the impeller were missing and these have not been traced.

 

On the basis of visual examination it appeared that the back plate and shroud were different materials. Testing with a magnet confirmed the back plate and blades were magnetic and the shroud non-magnetic. Fixing pieces (rivets etc.) were found to be non-magnetic. It was evident that the back plate had failed in a brittle manner whereas the shroud had been severely buckled and plastically deformed.

 


 

 


2.1.1.1 Back Plate

 

Examination of the back plate indicated that both surfaces were suffering from pitting corrosion suggesting that the plate had been in service for some considerable period of time prior to the 3 day period of operation terminating in failure.

 

Initial examination of fractures revealed that some were clean and obviously new. Others or portions of them were dirty and spotted with light rusting or severely mechanically damaged. The dirt deposits and light rusting were considered to be the result of contamination in the fire and subsequent water spraying after failure.

 

The major portion of all the fracture faces was typical of a brittle type of fracture. Small ductile shear lips were noted on some fracture faces. Brittle fractures often have cleavage marks in the form of chevrons or radial facets which indicate the direction of propagation and it is possible to trace back to the point of origin. It was noted that the initiation area on fracture A (fig.3) had a fairly heavily corroded appearance (fig. 5) indicating that it was relatively old. The corroded surface prevented identification of the mechanism of initiation. Initiation had occurred on the outside surface (ie. surface opposite adjacent back plate) near to the hub (fig.6) where it would not have been visible unless the rotor was completely dismantled.

 


 

 


Several other areas of initiation were identified on other fracture faces. These were located on both inside and outside surfaces. Typically they were characterised by small areas of apparent fatigue with radial marks indicating the origin.

 

None of the other initiation sites had evidence of significant corrosion and could not be characterised as old even after careful cleaning.

 

Fracture I (fig.3) was of particular interest as it contained an area of very smooth appearance. However, on low power microscopical examination it became evident that the smooth metal was actually a separate tightly adherent non-magnetic metal which could be peeled off the back plate fracture. This non-magnetic material has subsequently been identified as Aluminium (see section 2.5) and it has been established that this particular piece of the back plate cut through some aluminium cladding on a plant steam line.

 

2.1.1.2. Shroud

 

The shroud had been severely deformed. Three substantial pieces were identified but at least one section was missing.

 

Examination indicated that the surface condition appeared to be almost new and was indicative of an austenitic stainless steel as suggested by the non-magnetic response. There was no evidence of significant corrosion and the mill finish was still evident as were scribe marks, rivet heat marks and ink marks made at the time of fabrication. In view of this and the absence of significant tar deposit it is considered that the shroud may well have had a service life of only the three days 11th - 14th May.

 

It was noted that the shroud had been made with a radial butt weld which is not shown on the replacement drawing. The weld would have been made for the convenience of forming a split plate.

 

All of the fracture faces that could be examined were typical ductile 45° shear type failures except in the radial butt weld. The first 30 mm of fracture in this weld adjacent to the hub was square edged. Poor penetration had been achieved even though welding had been carried out from both sides.

 

Low power microscopical examination confirmed poor weld penetration. The weld metal appeared porous. Most of the fracture face in the weld was mechanically damaged and hence it was not possible to comment further on the failure mode.

 

2.1.1.3. Fixings

 

As indicated previously these had been found to be non-magnetic indicating they had been made from ductile austenitic stainless steel. No evidence was found of any abnormality, and those parts that had failed and could be examined had broken in a ductile manner with associated deformation.

 

2.1.1.4. Blades

 

These were magnetic but were severely deformed indicating ductile failure and those fractures which could be examined showed typical 45° shear fractures. No significant abnormalities were detected.

 

2.1.2 Intact Impeller

 

This was essentially undamaged and was taken by plant engineering staff for cleaning by sand blasting. Magnetic testing of the various components gave similar responses to those registered on the disintegrated impeller suggesting similar materials of construction. Visual examination revealed two interconnecting cracks in the back plate (see fig. 7). In order to gain additional information on the mode of failure the back plate was sectioned such that the cracks could be broken open and the fracture faces examined. Fig.8 shows one of the fractures and appeared to indicate fatigue initiating in the plate material.

 



 

 


2.2 Chemical Analysis

 

Samples of components were submitted for spectrographic analysis. Carbon content was determined by the combustion method. Results obtained are given in Table 1. From these results it is evident that within normal analytical accuracy limits both back plates are En56B material and both shrouds are AISI 304L. The blade of the failed impeller is En56B and the rivet is an AISI 304 type material. It would also appear that both back plates are from the same cast of steel and likewise both shrouds are from the same cast.

 

 

C

Mn

Si

S

P

Cr

Ni

Mo

Ti

Nb

Shroud Failed Impeller

0.03

1.48

0.62

0.016

0.027

18.67

9.33

0.13

0.015

0.01

Back Plate Failed Impeller

0.19

0.39

0.15

0.024

0.022

13.12

0.39

0.11

-

-

Blade Failed Impeller

0.18

0.36

0.27

0.019

0.020

13.30

0.56

0.13

-

-

Rivet Failed Impeller

0.10

-

-

-

-

17.14

8.15

1.19

0.01

-

Shroud Intact Impeller

0.04

1.45

0.60

0.015

0.026

18.69

9.38

0.13

0.014

0.01

Back Plate Intact Impeller

0.19

0.39

0.15

0.027

0.023

13.21

0.40

0.11

-

-

 

 

 

 

 

 

 

 

 

 

 

EN56B Specification

0.12

0.18

1.0

max

1.0

max

0.045

max

0.045

max

12.0

14.0

1.0

max

-

-

-

AISI 304L Specification

0.03

max

2.0

max

1.0

max

0.030

max

0.045

max

18.0

20.0

8.0

10.5

-

-

-

AISI 321(FDP) Specification

0.08

max

2.0

max

1.0

max

0.030

max

0.045

max

17.0

19.0

9.0

12.0

-

5xC

min

-

Table 1: Results of chemical analysis.

 

 

2.3 Metallographic Examination

 

Sections were cut from the following areas for metallographic examination:

 

(i) the initiation areas of fractures A and C (fig. 3).

 

(ii) the initiation areas of cracking in the intact impeller.

 

2.3.1 Initiation Area Fracture A

 

Corrosion was evident on the fracture face in the area of initiation. Initial cracking appeared to have been transgranular with some subsequent branching. Propagation at crack tips had an intergranular nature. The outer surface adjacent to the fracture exhibited multiple crack initiation from the base of pits (fig.9). Although the outer surface had suffered shallow intergranular corrosion the inroads of cracking from the outer surface were relatively broad and transgranular. Pitting and cracking was measured to a depth of 0,25mm on the outside surface and pitting to a depth of 0,05mm on the inside surface. Microscopical evidence is a positive indication of a corrosion fatigue mechanism having been the primary mode of crack initiation.

 

Text Box:  
Figure 9: Illustrating pitting and associated cracking on outside surface adjacent to initiation of fracture A.
 

 


2.3.2 Initiation Area Fracture C

 

The initiation area on fracture C was located on the inside surface of the back plate. The microsection showed no significant branching of the crack face at the initiation area. Initial cracking was transgranular which would be consistent with fatigue but subsequent propagation was of a mixed inter/transgranular nature. Adjacent to the point of initiation on the inside surface pitting was measured to a depth of 0,10 mm. Maximum depth of pitting on the outside surface was 0,04 mm.

 

2.3.3 Initiation Area of Crack in Intact Impeller

 

A section through the crack face in this initiation area indicated cracking to be intergranular with no significant branching. Initially some short discontinuous cracks were detected adjacent to the main fracture but after re-polishing these disappeared and it is possible they were facets of the main fracture. Pitting was measured to a depth of 0,09 mm on the outside surface and 0,07 mm on the inside surface. The intergranular nature of the cracking would not be typical of fatigue.

 

2.4 Hardness Tests

 

Vickers hardness tests were carried out with the following results:

 

Failed Back Plate  HV3O - 478/508 Approx. Equivalent Brinell = 460

Intact Back Plate  HV3O - 478/484  Approx. Equivalent Brinell = 450

 

The back plates are harder than the level specified on the original equipment manufacturer’s drawing.

 

2.5 Scanning Electron Microscope Semi-Quantitative Analytical Examination

 

A sample of the non-magnetic material found on fracture I was submitted for analysis and was shown to be an aluminium alloy.

 

Since it was evident from the visual and metallographic examination that corrosion had played a role in the failure, a sample of the initiation area of fracture “A” was examined. Analysis was carried out in an old area of fracture and a new area for comparison purposes. The results of these analyses are shown in Table 2. As the gas contains H2S, sulphur concentrations of the order detected were not unexpected. A high chloride level (8,3%) was detected on the old site. The measured figure cannot be taken as definitive quantitatively but indicates a high chloride level existed at the point of fracture initiation.

 

Sample

Si

(Wt%)

S

(Wt%)

Cl

(Wt%)

Ca

(Wt%)

Cr

(Wt%)

Fe

(Wt%)

Old area fracture A

0.60

2.57

8.32

0.43

12.88

75.21

New area fracture A

0.73

1.02

0.26

0.10

14.40

83.50

Table 2: Analytical results of SEM examination of old and new areas of fracture A.

 

 

3. DISCUSSION

 

Visual and metallographic evidence indicates that cracking has initiated primarily as a result of corrosion fatigue with subsequent propagation by brittle fracture. The old fracture site found is considered to have been in existence for some time before the fan was put back into service. This is indicated by the degree of corrosion found on the fracture face. Fracture faces at other initiation points showed no significant corrosion damage and may have initiated in the last period of operation. High strength quenched and tempered steels can be susceptible to hydrogen embrittlement cracking which is usually intergranular in nature. It is possible that this mechanism was involved at some initiation sites.

 

Text Box:  
Figure 10: Relationship between tensile strength and impact resistance (Izod) evident from above.
Initiation of fracture from the possible source causing failure was remote from rivet/bolt holes or other free surfaces and this is attributed to pitting attack. Subsequent propagation towards a bolt or rivet hole probably caused further initiations and propagation towards the impeller tip resulting in brittle fracture. The chloride levels found would not be unexpected taken into account the concentrating effect which can occur in water used in the gas washing process.

 


The back plates were En56B martensitic stainless steel which was the original equiment manufacturer’s specification. They had been heat treated to a higher hardness level than indicated on the original drawing. The specified hardness indicates a tensile strength level of around 36 tsi whereas the measured hardness indicates a tensile strength of higher than 100 tsi. The increased strength has an adverse effect on ductility as is shown in figure 10.

 

 

4. CONCLUSION

 

Failure occurred as a result of corrosion fatigue propagating such that a brittle fracture ensued. The corrosion fatigue is believed to have been initiated as a result of concentration of chlorides, over many years of operation, in water introduced into the gas stream for washing purposes.

 

With regard to the circumstances giving rise to this unfortunate accident the plant have decided to implement any and all additional quality assurance tests during and after repairs to the rotating parts of the fan. These will include stringent material tests in line with international standards for the construction of modern technology fans.