Simulation tools evaluate large-capacity furnace designs

March 23, 1998
Large-capacity cracking furnaces must be designed taking into account operational, process, and mechanical constraints. Kinetics Technology International BV (KTI) presents three case studies to evaluate potential constraints in furnace designs. Although a large-capacity furnace has a lower investment per ton of ethylene, flexibility and design considerations may limit the size of the furnace.
Wim Nouwen, Erik Dupon, Simon Barendregt, Frank Waterreus
Kinetics Technology International BV
Zoetermeer, The Netherlands
Large-capacity cracking furnaces must be designed taking into account operational, process, and mechanical constraints.

Kinetics Technology International BV (KTI) presents three case studies to evaluate potential constraints in furnace designs. Although a large-capacity furnace has a lower investment per ton of ethylene, flexibility and design considerations may limit the size of the furnace.

For large-capacity furnaces, the use of simulation and design tools, such as the Spyro (a registered trademark of KTI) program with a computational fluid dynamics (CFD) package, is necessary to avoid maldistribution problems.

The typical ethylene capacity per cracking furnace has increased from 20,000 metric tons/year (mty) in the early 1960s to capacities exceeding 120,000 mty today (Fig. 1, [45,800 bytes]) .

Cracking furnaces are the heart of the ethylene plant for various reasons:

  • Ethylene product yield is determined in the furnaces.
  • One-third of the cracker investment is consumed by the furnace section.
  • Most of the plant energy is consumed by the furnaces.
  • A furnace needs regular operation and maintenance attention, e.g. decoking operation.

Furnace modeling

KTI uses Spyro and CFD to study the complicated fluid flow and heat-transfer phenomena occurring in the firebox of a steam cracking furnace. Currently, about 80% of the ethylene-producing industry licenses Spyro for production optimization and design verification.

Spyro is a rigorous pyrolysis kinetic model contained in a computer simulation program representing the performance of a steam-cracking furnace, both from the radiant coil as well as the firebox side.

The kinetic model represents over 3,500 radical and molecular reactions. The model was developed and validated using an extensive data base of pilot plant and industrial data.

CFD comprises a series of methods to solve the general transport equations for heat, mass, and momentum. These transport equations are based on the fundamental equations of conservation subjected to their boundary conditions and are cast into a discrete form.

Take, for example, a rectangular box representing the geometry of the radiant section of a furnace. Given the profiles of mass, heat, and momentum at the inlets and outlets, the solution of the discrete transport equations over this box will show a three-dimensional (3D) field of pressure, velocity, and temperature. The fields are composed of values at a finite number of positions.

CFD's ability to predict heat transfer in 3D spaces extends to complicated geometries in which the prevailing boundary conditions are difficult to prescribe, which is frequently encountered in engineering.

These models are based on time-averaging of the fundamental transport equations. Thus, the solutions of these equations consist of time-averaged, rather than instantaneous, quantities. The time-averaging approach makes CFD feasible for engineering design purposes.

The transport of heat in furnaces mainly takes place by radiation heat exchange among the flue gas, process tubes, and refractory. To account for this phenomenon, the transport equation is extended by a term expressing the net radiation source.

To provide a realistic tube-skin temperature profile, a process model is necessary to supply the relation between the duty and the tube-skin temperature along the coil. Spyro models the steam-cracking process in a tubular reactor to predict this profile.

Furnace design

In the early 1960s, cracking furnace designs were based on horizontal radiant coils with long residence times and small capacities. The first vertical straight coils were installed in the late 1960s, and split coils followed in the 1970s.

During the 1980s, engineers focused on reducing radiant coil residence time to maximize ethylene and propylene selectivity. In the 1990s, with mature cracking-furnace technology, the designs focus on mechanical details and the minimization of both operating and investment costs. Some of these developments include:

  • The use of linear quench coolers, reducing the number of mechanical cleanings.
  • The reduction of the number of burners, using large-capacity, upshot-type burners.
  • The application of higher-grade radiant coil materials, such as 35Cr-45NiNb and 28Cr-48NiW, allowing higher temperatures and improved creep resistance.
  • Modifications to the radiant coil supporting system, such as the use of constant-load hangers to improve furnace robustness.
  • The implementation of advanced controls to optimize cracking severity, furnace loading, steam-dilution ratio control, pass balancing, firing control, etc.
With these improvements, a trend to increase the capacity of individual cracking furnaces has occurred. Larger furnaces result in lower investment costs per ton of ethylene, a reduction in operation and maintenance costs, and a saving of plot space (especially of importance for plant expansions).

There can be limitations to building a large furnace. Potential constraints are found in operational flexibility, mechanical limits, and process performance.

Operational flexibility

To ensure that the desired quantity and quality of products are produced from the available feedstock at any given time, the furnace section should satisfy several criteria:
  • Furnaces should be flexible enough to handle the variety of available feedstocks.
  • The number of furnaces in operation should be such that in the case that any furnace trips, there is not much upset in the downstream separation section. For example, having a minimum of four furnaces in operation generally prevents the cracked-gas compressor from going into antisurge operation.
  • When one furnace is in decoking operation or maintenance, the other furnaces should still produce the desired plant capacity.
  • The furnaces shall be easy to start up, operate, and shut down. Good advanced controls are helpful. Also, a limited number of burners reduces operator intervention. This reduction is achieved by using bottom and sidewall-mounted, upshot-type burners.

Mechanical limitations

To make larger-capacity furnaces, the limits of length, height, and width need to be stretched. The mechanical design should be within applicable codes, standards, and the requirements of local authorities. Also, the mechanical design should allow easy access for maintenance.

For the radiant coil, the length of the individual tubes is limited because the weight of the coil can cause a high creep rate. With a given coil design, the capacity per coil can be varied within given boundaries. A larger-capacity furnace should contain a higher number of coils.

KTI coils can be treated as sets or "modules" and multiplied as needed. Also, the number of quench coolers can be increased relative to the number of radiant coils.

The convection section tubes can, in principle, be made as long as needed for the furnace capacity. Although no clear technical guidelines are available, a 12 m restriction is sometimes imposed to restrict the length so that no intermediate welds are required.

KTI's convection section design minimizes the required surface area while achieving high thermal efficiency. This is done by increasing the flue-gas mass velocity and shortening the tubes; i.e., the convection section is shorter than the radiant section.

Process performance

If the design criteria are not correct or if they are incorrectly interpreted, the process performance in actual operation may greatly differ from the theoretical performance. Design criteria used for a small-capacity furnace may be correct, but they may not be applicable for large-capacity furnaces. Mistakes can cause maldistribution of the flue gas in the radiant box or in the convection section. With the help of sophisticated computer-simulation tools, an experienced engineer can avoid these problems. To obtain technical and economic data on the feasibility of large furnace designs, KTI studied critical areas of the furnace using Spyro and CFD in three cases:
  • Single-radiant box design vs. double-box design
  • Flue-gas flow entering a reduced-length convection section
  • Separate cracking of various feeds in one firebox.

Single vs.double-box design

Increasing the capacity and reducing the investment cost of cracking furnaces can be done by increasing the length of a cracking furnace with one firebox (single box) or by combining two fireboxes into a so-called "double-box" design. In a double-box design, two fireboxes share one common convection section. Fig. 2 [55,716 bytes] outlines these two designs.

The advantage of the double-box design is that the furnace becomes shorter, although it requires more width. As will be shown, however, the double-box design does not allow independent operation of the two boxes as is required for independent decoking operations.

In an extreme case, one firebox runs on full load (normal mode) and the other firebox is on decoking operation. In Figs. 3 and 4, KTI investigates the effect of this extreme operation on the flue-gas flow at the flue-gas crossover and on the high-temperature coil of the combined convection section. The firing in the decoking operation firebox is assumed to be 30% of the full load of the firebox.

In Fig. 3 [49,627 bytes], the velocity vectors indicate the direction of the flue gas and its velocities. At the place of high velocity, there is prevailing flow. Fig. 4 [41,754 bytes] shows the flue-gas temperatures in the crossover and in the first part of the convection section. Blue is the lowest temperature, then green, yellow, and dark red.

If the tubes on the left side of the convection section go to the decoking firebox, those tubes will get little flow. The other tubes will get high flow and high flue-gas temperatures. The imbalance on the right side impacts both fireboxes.

In this extreme case, the two fireboxes impact each other, and they cannot be considered as two independent furnaces.

The firebox in decoking operation will receive highly uneven inlet temperatures, which will result in very uneven decoking; incomplete decoking of some of the coils may occur. The firebox in normal operation will also be affected because normal temperature profiles cannot be maintained; this operation can result in overcracking and extreme coking.

Higher-grade materials will be required in some parts of the convection section, resulting in higher investment costs.

In normal operations, when both fireboxes are in operation, although similar imbalanced behavior is expected, the behavior will be less pronounced and will depend on the firing rate in both boxes. If the firing rate is similar in both boxes, the resulting flue-gas flows will be similar, and minor interaction will occur between the two boxes.

Shortened convection section

In KTI designs for large-capacity furnaces, the length of the single firebox has increased significantly, to over 22 m, whereas the length of the convection section remains short. To have efficient heat transfer in the convection section, the flue-gas velocities are kept in the range of 4-6 kg/sq m-s. This design keeps the convection section shorter than the radiant section.

The different lengths affect the flue-gas crossover between the radiant and convection sections. The flue gas from the radiant section needs to be contracted before entering the heat transfer zone. If the flue-gas contraction disturbs the flow pattern of the radiant section, cold zones will occur on the radiant coil.

Detailed CFD simulations were made for two scenarios:

The flow regimes of the radiant section of the two convection section designs are the same. The only difference in the flow begins in the flue-gas crossover, mainly as a result of the beams which are blocking part of the crossover.

The shortened convection section is a way to reduce the investment cost of a large-capacity cracking furnace without compromising process or operational integrities.

Separate cracking in one firebox

An ethylene plant is generally designed for a certain feedstock slate. A different feedstock, however, should not pose a problem as long as it has been accounted for in the design stage. There are several ways to cope with a variation in feed types and rates:
  • Crack the individual feedstocks in dedicated furnaces and store excess quantities. This solution does not work when there is no dedicated storage tank available, when there is not enough feed available, or when a recycle furnace is on decoke.
  • Mix the feedstocks and co-crack them. This solution gives no penalization in a case using various naphthas and raffinates, but other mixed and co-cracked feedstocks may penalize yield or run length. For example, mixing ethane and propane significantly penalizes the selectivity towards propylene.
  • Crack the individual feedstocks in separate coils of one furnace.
If different feeds have to be cracked simultaneously, cracking the individual feedstocks in dedicated furnaces is less flexible. Separate cracking of various feeds in one firebox increases the flexibility to change feed types and rates for large-capacity cracking furnaces. Certain passes and associated radiant coils are designed to process different feeds at different cracking conditions.

This solution is not new. To stabilize the cracking furnace operation, throughput of the various passes is adjusted and in some cases, the dilution-steam ratio for certain passes is adjusted as well. The coil-outlet temperature can be very different because one feed may need a higher cracking temperature than another.

Fig. 7 [45,233 bytes] shows the temperature profile when two different feeds were cracked at different cracking severities in one firebox. The smooth temperature gradient reveals that this cracking is possible. However, the firebox sections have to be balanced by varying the individual coil throughputs (hydrocarbon and steam). In this way, the firing in both coil areas are kept the same and consequently, no flue-gas flow maldistributions will occur.

The successful application of this system requires the coil sections to be equipped with individual controllers and the furnace to be equipped with the controls that allow smooth operation and transfer between the different operational modes.

The Authors

Wim Nouwen is technology manager, ethylene, with KTI BV. He has held various positions with KTI since 1981. Nouwen holds an engineering degree in chemical technology.
Erik Dupon is sales engineer, ethylene technology, for KTI BV. Before this function, he held various positions within KTI involving activities for proposals, basic design, plant simulations, and optimization in the fields of petrochemicals, refining, gas treating, hydrogen, and ethylene. He has been working with KTI for more than 10 years.
Frank Waterreus is supervisor, computational fluid dynamics, with KTI BV. He has been working with KTI since 1988, mainly involved with software engineering. Waterreus holds an MS degree in chemical engineering.
Simon Barendregt is vice-president, technology development, in KTI. He is responsible for petrochemical and refinery-related product and technology development. He is also responsible for KTI's Pyrotec division, which develops and commercializes steam cracking technology. He has been with KTI for 24 years in various technical and management positions. Barendregt holds a masters degree of chemical technology from Delft University of Technology, The Netherlands.

Copyright 1998 Oil & Gas Journal. All Rights Reserved.