Cleaning of Fouling in the Primary Reformer Convection Section and Combustion Chamber Firebrick Coating



PAWIRADIKRAMA, H. SUPRIYADI, Kaltim Parna Industri, Bontang, East Kalimantan, Indonesia

This article was first published at Nytrogen and Syngas Conference, 25-28 February 2007, Bahrain

Kaltim Parna Industry, a Joint Venture Company between Indonesian and Japanese companies, operates a single-train 1,500-t/d nameplate capacity ammonia plant in Bontang, East Kalimantan, Indonesia. The plant has been in commercial operation since February, 2002, serving the international market. Several months after the plant went into operation, a strange phenomenon was observed in the convection section coils, where the coils had been fouled by what appeared to be brick dust from degradation of fire bricks in the radiant section. These fouled coils greatly disturbed the plant operation as well as impairing energy recovery, and accordingly some adjustments were needed to maintain reasonably safe plant operation.

A temporary solution to the problem was to clean the convection section during the annual plant turn-around. While this noticeably improved the plant performance when it returned to operation, after a few months the operating conditions deteriorated as the fouling continued. It was obvious that a permanent countermeasure would have to be developed and the cause of the degradation of the firebrick in the radiant section that was the suspected source of the fouling material must be eliminated. A special coating to provide such protection against erosion has been studied and we have developed a specimen test accordingly. An experiment entailing spot coating was carried out during the February 2005 turnaround and the measurement results has been collected for observation. The following is a summary of the results observed to date:

  • The coating condition is relatively stable and we could not find any problem with the primary reformer operation accordingly;
  • Heat reflection/transfer could be expected better because of the increased inner wall temperature of the primary reformer;
  • Heat loss to the surroundings has been decreased, as indicated by the lower outside wall temperature.

Based on the above, the coating performance needs to be further evaluated, and more variations need to be exercised to ensure its performance and the expected plant performance improvement accordingly.

Currently 50% of the fire brick surface has been coated during the last turn around on November 2006, and so far no problem has been observed with plant operation. Unfortunately the impact on plant operation could not be examined properly because the plant is still being operated at reduced capacity for other reasons.

This paper shares our experience about the impact of the fouling on plant operation, the operational measures needed to mitigate the problem, the method of cleaning the convection section coil, and the result of the coating experiment.

The following terms are used in the paper:

Convection section: The upper section of the furnace and comprises a group of heat exchanger stacks used for recovering heat from the flue gas.

Radiant section: The main furnace box.

Fouling: The process by which dirt becomes attached to the surface of heat exchanger tubes.

Garnet: A natural mined sand, the main constituent of which is almandine ore [iron aluminum silicate, Fe3Al2(SiO4)3], used in sand blasting.

4 point nozzle: A specially stainless steel nozzle with four outlets, prepared in-house for use in sand blasting.

Furnace coating: Zirconium-based coating, capable of withstanding the high temperatures inside the furnace. There are two types: non-vitreous and vitreous. However, this paper refers to non-vitreous type only.


PT Kaltim Parna Industri operates a single-train, 1,500 t/d ammonia plant located in Bontang, East Kalimantan, Republic of Indonesia. The construction project commenced in the middle of 1999 and Mitsubishi Heavy Industry (MHI) Japan was the main contractor.

The ammonia process is based on the Haldor Topsøe low-energy process, and the CO2 removal section uses BASF’s aMDEA absorption medium. The ammonia synthesis uses the S-250 reactor series comprising a combination of the S-200 and S-50 reactors. The plant successfully passed the demonstration and performance test on December 14, 2001 (Table 1) and it has been in commercial operation since February 2002.

The raw material, natural gas is supplied from East Borneo gas fields that were explored by some foreign oil company under a production sharing contract scheme with Pertamina, a state-owned company for producing oil and gas. The natural gas averagely contents 87% CH4, 6% CO2 and 3 ppm sulfur compounds.

Table 2 shows annual production records until 2006.

After several months, a strange phenomena was observed in the convection section of the primary reformer. Fouling was occurring in the preheating coils, and it was giving rise to some operational difficulties and frequently threatening the stability of the plant. Some operating conditions were adjusted to prevent unnecessary plant trouble, and the coils were regularly cleaned at every plant turn-around in an effort to normalize plant operation conditions. But the cleaning task was troublesome and costly, so a permanent solution was needed, both to reduce costs and to maintain plant stability. Applying a stabilising coating to the firebrick surface was considered, but before embarking on such an expedient wholesale it was obviously prudent to carry out some experiments to establish the best method and to identify any other problems.


The primary reformer is of Haldor Topsøe design and is based on the side-fired furnace concept. It contains 180 catalyst tubes mounted in two similar rows along the centre line of furnace and 504 burners are mounted in seven rows in the furnace walls, where the flames are directed towards to catalyst tubes.

The furnace is divided into two sections: radiant and convection. The radiant section is the lower section of the furnace containing the burners and catalyst tubes, and the convection area is the upper section, where heat is recovered from the flue gas, which leaves the radiant section at 1,000-1,030ºC, for preheating various process streams or generating steam in a series of heat exchanger coils. The cooled flue gas leaves the reformer through the induced-draft fan at around 150ºC and 1% O2 excess.

To avoid excessive heat loss from the furnace, firebrick is installed on the inside of the entire wall of the furnace, in both the radiant and the convection sections. Normally, the outside wall temperature in the radiant section will be around 50ºC.



As mentioned, a variety of streams are heated sequentially by the hot flue gas in the convection section.. Table 3 on the preceding page shows the type and size of each exchanger in the series, starting from the bottom (hottest). The diagram of the reformer in Fig. 1 shows the arrangement of the components of the convection section graphically.


In the initial operation, the operating conditions of the convection section were normal and close to design, but unfortunately, after several months, the temperature of some streams were difficult to control and could cause worse, even the worst, operating conditions.

The critical streams that need to be carefully monitored and controlled were the natural gas stream outlet from the last preheater, E-0204A, and the boiler feed water outlet from preheater E-0205.

The natural gas temperature outlet from the final preheater E-0204A was excessive even though all the control parameters were within their optimum. The normal temperature is 400ºC, but the worst condition was 420ºC.

The boiler feed water temperature outlet from the preheater E-0205 has to be held below its boiling point, otherwise dangerous conditions can result. Several months after the start of operation, the actual temperature nearly reached the boiling point.

Other streams suffered similar excursions, but the main cause for concern in their case was simply the material resistance. Fortunately from the design specifications the materials still had some allowance against temperature overruns.


Abnormal heat distribution in the convection section was observed several months after initial plant operation and can be summarized as follows:

Cleaning of Fouling in the Reformer Convection Section and Combustion Chamber Firebrick Coating

The heat duty of the mixed feed preheater heat exchanger E-0201 decreased. The heat duty of the upper parts of the E-0202 process air, E-0204 natural gas and E-0205 boiler feed water preheaters increased.

Of these discrepancies, the BFW temperature is the most critical one, because the impact on the plant operation stability is high. If the temperature reaches boiling point, it will be very difficult to normalize and therefore it will affect the steam drum level stability and probably cause the plant to shut down.

Having observed the above condition, it was suspected that the heat transfer in the lower part of convection section had decreased and the heat absorption had moved to the upper coils. However, this suspected fouling need to be confirmed during plant shut-down/turn-around. Table 4 shows the difference between the operating conditions of the plant in its new and fouled states.


Struggling with the above situation, the following countermeasures were developed, including both adjusting the operating parameters and cleaning the fouled tube, to minimize the unfavourable impact on overall plant operations.

Adjustment of operating conditions

The purpose of the operational adjustments is to maintain the critical operation parameters within the ranges considered safe until the nearest turn-around for cleaning. For best results, a combination of the following methods was used.

  • The heat absorption in the lower coils was increased to decrease the heat absorption in the upper coils. This was achieved by increasing the cooling steam to the inlet final process air preheater coil E-0202A so as to increase the heat absorption in that coil and decrease the flue gas temperature leaving it. The cooling steam increased from 2.21 t/hour to 5.6 t/h.
  • The rate of BFW circulation back to the deaerator was also increased so as to maintain the BFW temperature from the preheater. The recycle valve (normally closed) was opened to around 35% open.
  • In order to increase heat absorption in the lower coils and reduce it in the upper coils, the combustion air surplus was reduced to the minimum permissible level of oxygen excess for acceptably efficient firing.
  • The H2/N2 ratio in the synthesis loop was from the stoichiometric 3.0 to 3.5 so as to reduce the quantity of fuel off gas from the hydrogen recovery unit that contains inert gas such as N2, even though it was acknowledged that such an action would decrease the efficiency of the synthesis loop.
  • The HP steam temperature from steam superheater coil E-0203 was lowered from its normal level of 515ºC to 480ºC by injecting more desuperheater water to absorb more heat and decrease the flue gas temperature leaving its coil, thus relieving the heat load in the upper coils.
  • The output of HP from the main waste heat boiler E-0208 downstream of secondary reformer was increased by allowing it to run at a lower output temperature of 335ºC. Normally this boiler cools the hot process gas from 940ºC to around 360ºC before it enters the HT shift converter.

Periodic cleaning of the convection section coils

Periodic cleaning activities

The reformer convection section was first opened in the first annual turn-around in order to determine about the inside conditions. It was only the start of our efforts to correct the convection section conditions; we conducted special cleaning activities in this section during the second and third annual turnarounds as well.

First annual turn-around

The first plant annual turn-around commenced on September 10, 2002 and lasted for 20 days. After opening the manhole, fouling on the surface of the coils was plain to see. The severest fouling was on the lowest coil, the mixed feed (natural gas and steam) preheater. The thickness of the fouling material was around 1-2 mm and it was physically very easy to remove with the bare hand.

Table 5 compares the analysis of the fouling material with that of the refractory bricks in the furnace lining.

After determining the most probable cause of fouling and getting an understanding about the fouling conditions, an attempt was made to clean up the coil to restore it to normal condition, but a proper procedure had not been worked out. The plant performance was better for several weeks after cleaning, but after that the performance began to fall off. From this evidence we concluded that the fouling was constantly occurring in the coils and so regular cleaning would be needed unless the fouling could be eliminated by finding a way of preventing the firebrick from dusting.

Second annual turn-around

The second turn-around lasted 25 days from January 6, 2004. A special cleaning programme had been prepared and scheduled.

The sand-blasting material was garnet, a derivative of silica sand, which is three times harder and has a higher melting point. This quality is necessary in order to avoid air pollution, and in case any garnet is left in the furnace it will not melt inside and contribute to new fouling of the coils.

Cleaning of Fouling in the Reformer Convection Section and Combustion Chamber Firebrick Coating

The plant performance was definitely better after cleaning. One of the indications was that the HP steam temperature returned to its design value of 510ºC. It was also possible to keep the natural gas temperature at the inlet of the desulphurizer at 400ºC.

Third annual turn-around

A similar sand-blasting was carried out in the convection section during the third annual turnaround. The severest fouling was again observed on the three bottom coils. After cleaning, the performance of the convection section was further improved, although the result was not as dramatic as the previous occasion.

Cleaning method

As mentioned, the method used was sand-blasting using garnet sand. It was chosen because of the characteristics of the fouling and the environment. Generally, the fouling is loosely attached to the tubes (except for E-0201 up to E-0203, the bottom exchangers), but too sticky for air blowing. On the other hand, refractory material is sensitive to water, ruling out wet cleaning methods.

Garnet 20/40 mesh (425-850 microns) was chosen because of its following advantages compared to silica.

  • On account of its high hardness (7.5-8 Moh), it is not easily broken up so will not make a mess in the convection section. Although the theoretical breakdown rate is 10-20% per cycle (5-10 recycles), we strictly did three recycles only. The material purchase was based on this value.
  • Its higher melting point (min. 1,350ºC) is well above the inlet temperature if the convection section (about 950ºC), so there is no danger of any that may have been left behind in the system fusing and causing scaling.
  • Because of its higher specific gravity (SG) and low breakdown rate, less dust is produced, thus ensuring good operator visibility. The higher SG also means less power is required to achieve a given result than with silica.
  • With minimal ferrous particle, there is less risk of ferrous contamination to the surface.
  • The cleaning speed is faster, thanks to the low power requirement.

The general procedure was initially developed by our contractor, Mitsubishi Heavy Industries, which is then modified to match our requirements. It consists of the following stages:

Installation of refractory protection

In order to prevent the sand-blasting from eroding the refractory, all surfaces must be completely covered. End tube protection is made from 5-mm board designed specifically to follow the contour, while side protection (longitudinal portion) is covered with a canvas tarpaulin. It is a modification from the original method using corbel protection.


End tube protection is installed according to the following sequence:

Side protection installation is just a simple insertion of the canvas-tarpaulin. A wire is required in the beginning, to guide the tarpaulin down. In our experience, the salient portion of the refractory is strong enough to handle the friction and will not collapse.

Installation of waste trap on every three rows exchangers

Waste traps are essential for recovering used garnet for recycle. The more there are the better the reat of recovery will be. However, this has to be offset against the longer installation time. Here, every three rows is a compromise between these factors.

Sandblasting from upper and lower side of tube banks

Because of the way the tubes are arranged, sand-blasting is conducted from two sides, upper and lower part. The special 4-point nozzle is very helpful in speeding up cleaning. Figure 3 is a sketch of sandblast direction is as follow:


Air blowing for garnet recovery and final cleaning

Although garnet is relatively dust-free, the fouling material is reduced to dust after blasting. To prevent the dust from settling back onto the tubes, air is blown in amongst the tubes using the same nozzle. At the same time, the blower (operating in reverse) is used to catch flying dust. This will also push the remaining garnet downwards. Recovered garnet on the waste trap is then separated from the fouling material by vacuuming off the dust, allowing it to be re-used for blasting until it has done three recycles.

Table 6 lists the equipment needed in this operation and Table 7 gives the specification of the cleaning material.

The special nozzle was prepared in contractor’s shop. Closed-end stainless steel is drilled at 4 points, at angle 30 deg to shape the nozzle head. Due to shop’s limitations, no ceramic protection is applied around the holes that usually be provided for reinforcement purpose. Hand grip at 50 cm long, for ease of operator maneuver, is prepared separately from the nozzle. This is to safe material usage as only the nozzle head will damage during the blasting process. Figure 4 is a simple sketch. The whole sandblast works above is completed within 14 days time frame (24 hour shift-works).



Result of cleaning

After cleaning to all coils in the convection section, some improvements are observed and plant efficiency is significantly improved (Table 8).

The following photographs show the contrast between the coil surfaces in the convection section before and after and after cleaning.

Fin tube before cleaning                                                                      Fin tube after cleaning







Bare tube before cleaning                                                                    Bare tube after cleaning







Based on the above experience during the first, second and third turn-arounds, the assumption that the fouling is occurring continuously is being verified, even though the intensity is gradually decreased. The first alternative solution is periodic cleaning and the other alternative is eliminating the source of fouling material by coating the firebrick surface.



Regular cleaning at every turn-around is only a palliative measure, and it has significant cost implications in terms of the cost of performing the cleaning operation every year and the cost of efficiency loss in between cleanings.

Based on a rough calculation, if the source of fouling in the convection section coil could be eliminated, the total saving would be more than US $1 million per year.

To achieve the this objective, it was proposed to apply a special coating to the firebrick surface. The other benefits that could be expected from the coating application are a reduction in fuel consumption by reducing heat loss and protection of the firebrick from potential deterioration.

  • Reduced fuel consumption

The glazed surface of coating will reflects more heat than usual firebrick surface. Higher heat reflection reduces the heat loss to the outside through the reformer wall and it increases the efficiency of the heating unit.

  • Firebrick protection

There is scarcely any material in plant operation that takes a greater beating than the refractory in boilers or furnaces, and refractory replacement is a big and costly job. A special firebrick coating provides protection to the refractory surface from slag adherence, thermal shock and chemical attack.

The idea of applying a coating application was first proposed in mid-2003 and became the subject for serious discussion and intensive research involving all related departments in the company.

Doing a complete coating job from the start was considered too risky because there was no previous experience and references. So, to avoid the risk of creating new problems, and to find out more about the characteristics of coatings, it was decided to do a sample application before going to full


Coating sample application

The first specimen was applied during the annual turn-around in February 2005. On this occasion, eight samples were placed in the surroundings of 32 burners in different locations. In selecting the sample locations, consideration was given to obtaining as much data as possible representing the overall reformer operation. Ease in sample installation and easy observation were also considerations.

  • Samples were installed in the neighbourhood of burners that would be exposed to the highest temperature, so as to assess the stability of the coating under high-temperature attack.
  • Samples were located near the peep hole (lowest row burner) for easier installation and observation.

Based on the manufacturing data, the coating material compositions are

Zirconium 62.97%, Aluminium 1.63%, Calcium 1.37%, Magnesium 0.34%, Silicon 32.06%, Titanium 0.25%. Boron 1.1%, Fluorine 0.25%.

This coating material has a good adhesion to the firebrick surface. The coating was brushed on manually in a layer around 1 mm thick.

After coating, the sample was aerated to encourage natural drying so as to avoid thermal shock and probable damage when the burner was ignited. Following is the sample coating picture after several days in aeration before burner ignition. The photograph below shows the coated area before the furnace was restarted.

Observation and data collection

After start-up on February 18, 2005, visual observation and performance measurement of the sample was conducted.

  • Visual inspection revealed that the sample bonded tightly to the firebrick surface. From the 32 samples, minor deterioration was only observed in one, and intensive observation and comparison with the other samples revealed that the sample that had deteriorated was thicker than the others. Therefore we concluded that the thickness of the coating has an impact on performance.
  • Visually the coating surface is smoother than uncoated surfaces with a shiny surface appearance from which it may reasonably be deduced that, as predicted, the coating would improve heat reflection.

To evaluate the performance improvement, temperature measurement was conducted in several points of the coated (C) and non-coated areas (N). The non-coated (N) measurement points were selected in the nearest location of sample coating. The temperature measurements were made on the inside and outside of the primary reformer wall. Typical results are given in Table 9.

From the above observations and temperature measurements, it can be expected that application of the coating at the firebrick surface can increase the plant performance in the following ways.

  • Coating the fire brick surface will eliminate the source of fouling in the convection section coils, so plant performance can be maintained in optimum condition.
  • Increasing the inner wall temperature from 1,046.71ºC to 1,092.19ºC (a 45.48ºC increase) can be expected to increase heat reflection from the coated surface to the catalyst tube in proportion, so to achieve the same temperature in the catalyst tube, fuel consumption can be reduced.
  • Decreasing the outer wall temperature from 69.91ºC to 66.10ºC (a difference of 3.8ºC) indicating that the coating is also helping to prevent excessive heat loss to the surroundings.

The above evidence implies that application of an appropriate coating to the firebrick surface in the primary reformer will provide some advantages and can contribute to the improvement of the overall plant performance. Table 10 records the progress towards complete coating of the furnace lining during three turn-arounds.

The following photographs compare the appearance under normal service of coated and uncoated furnace linings.

Appearance of coated and uncoated fire brick surface under operation









  • Fouling in the convection section of the primary reformer has given rise to difficulty in maintaining plant stability in operation. Some operating conditions have been adjusted to overcome the problem. Based on laboratory analysis, the fouling material appears to come from the firebrick insulation.
  • Periodic cleaning in every turn-around is carried out to restore the plant operation conditions, but even so, after several months the condition is back to bad again.
  • Periodic cleaning of fouled tubes entails extra cost; therefore a permanent solution has to be developed by eliminating the source of the fouling material. Coating of the firebrick surface is an alternative solution to eliminate the source of the fouling.
  • From the coating sample experiment, the coating condition is stable and tightly bonded to the fire brick surface. The coating is also expected to provide the additional benefit of increasing fuel thermal efficiency.
  • The most important parameter of the coating is thickness, as thicker coatings will become unstable and probably crack during operation.


Ammonia Know How