ACHIEVEMENT OF DRASTIC DOWNSIZING AND COST REDUCTION FOR LNG OPEN RACK VAPORIZERS BY USING OPTIMUM DESIGN METHODS

Masaru Sekiguchi, Production Engineering Sect. Production Dept. Tokyo Gas Co., Ltd. Norihiro Hisada, Thermal Energy Systems Engineering Dept. Sumitomo Precision Products Co., Ltd. 1. INTRODUCTION LNG vaporizers are some of the most important pieces of equipment for ensuring the stable and cost-effective production of city gas to be found at LNG receiving terminals. Since most such terminals are sited at or near the seacoast, currently the majority of LNG vaporizers are ORVs designed to utilize the thermal energy of seawater. In 1969, we introduced an ORV at the Tokyo Gas Negishi Terminal, which was Japans first LNG receiving terminal. Since then, we have continued the technical development of ORV design. Comparing ORVs developed in recent years with those from 1969, improvements in heat transfer performance and other areas of ORV technology have resulted in a reduction of some 80% in the floor space required for ORV installation, and significant cost reductions have also been made. It is difficult, however, to achieve any further cost reductions using only empirical approaches based on the designers experience using conventional design methods. As Fig. 1 shows, the reasons for this include complex heat-transfer phenomena such as boiling of multicomponent LNG inside the heat exchanger tubes, and ice formation on the outside of the heat exchanger tubes resulting in resistance to heat transfer. For example, even though heat is efficiently transferred to the external surface of the heat exchanger tubes by a falling film of seawater, over-efficient heat transfer at the internal surface leads to the formation of a thick ice film on the external surface. This decreases the heat transfer area resulting in reduced thermal performance. The heat exchanger tubes should be designed so that heat transfer at the internal surface properly balances that of the external surface. However, the conventional design method necessitates considerable labor, expense, and time to be spent on the production and comparative testing of a variety of test models in order to select the optimum shape of the heat exchanger tubes.
NG

Seawater

Seawater

Boiling phenomenon Acceleration of heat transfer and reduction of film boiling due to swirl flow

Acceleration of heat transfer by falling film

Ice forming phenomenon (change of external shape due to icing) Heat transfer accelerator (spiral core)

LNG

Heat exchanger tube

Inner fin Radial outer fin (star-fin)

Fig. 1

ORV heat exchanger tube construction and heat transfer phenomena

Therefore, we have developed a technology that makes it possible to predict the performance of heat exchanger tubes of a wide variety of shapes using only mathematical calculations. The new

design method also enables the efficient search for and determination of the optimum shape of the heat exchanger tubes according to the desired design criteria, such as the minimization of construction cost.

2. OUTLINE OF NEW DESIGN METHOD

Fig. 2 shows the flow chart of the newly developed optimum design method. This section outlines the key points of the new design method such as the formula for the single-phase heat transfer coefficient on the LNG side, the empirical formula for the LNGs boiling heat transfer coefficient correlation, analysis of the shape of ice formed, and analysis of optimum shape of heat exchanger tubes.
Start

Change in external shape

Change in internal shape Formula for single-phase heat transfer coefficient on LNG side

Empirical formula for heat transfer coefficient on seawater side

Empirical formula for LNG's boiling heat transfer coefficient correlation

Analysis of shape of ice formed by BEM (boundary element method)

Calculation of numb er required Analysis of numb er required

Programmed calculation of performance and cost

Analysis of optimum shape by response surface methodology Determination of optimum shape

Fig. 2

ORV optimization flow

2.1. Formula for single-phase heat transfer coefficient on LNG side LNG flows upward from the bottom of the heat exchanger tubes, where it is vaporized and superheated. The internal structure of the heat exchanger tubes features fins that provide a large internal surface area to increase the heat transfer available (see Fig. 1). To further accelerate the heat transfer, a spiral core (twisted a crisscross section extruded bar) and other features are also incorporated, which generate a swirl flow inside the tube to accelerate the heat transfer and prevent dry out due to film boiling as well. In this case, the forced convection heat transfer coefficient in turbulent flow is determined as follows using the Colburn factor j: (2/3) Equation 1 = j Cp Gm / Pr where: = heat transfer coefficient [W/m2K] Gm = mass velocity [kg/m2s] Cp = heat capacity [J/kgK] Pr = Prandtl number

The pressure drop is given by the following equation using the friction factor f:

Equation 2 P = ( 4 f L / De ) x ( v / 2 ) where: P = pressure drop [Pa] length of channel [m] L= De = equivalent diameter [m] = density [kg/m3] flow velocity [m/s] v=
Because factors j and f differ depending on the internal shape of the heat exchanger tube and shape and twisting pitch of the spiral core, there has until now been no alternative but to determine these factors by taking actual measurements by means, for example, of actual tests using LNG or air as a substitute. This time, however, flow was calculated using numerical flow analysis software (Star-CD), and the heat transfer and pressure loss properties determined and compared with actual measurements. One example of the results of analysis is shown in Fig. 3.

Fig. 3

Example of results of numerical flow analysis (velocity vector)

Whereas the channel along the inner fins of the heat exchanger tube is straight, the heat transfer accelerator forms a spiral channel, causing some problems due to the difficulty associated with the mesh binding and the occurrence of numerical errors caused by distorted mesh. Selecting, for example, the setting for the proper calculation option, however, resulted in the coincidence of tendencies between the actual measurements and analysis results, showing it to be possible to perform a simulation using numerical analysis. To determine the impact on factors j and f of each dimension, numerical calculations were made of several dozen shapes, including those on which there were no actual test results, and the degree of effect assessed by applying the response surface method using the design of experiments [4]. From the results were determined the most dominant design variables, and practical methods of estimating factors j and f developed.

2.2. Empirical formula for LNGs boiling heat transfer coefficient correlation A database of the heat transfer properties in the boiling zone was constructed using an experimental approach. We had until now been using correlation values acquired under certain specific conditions. This time, however, we verified the results by comparing actual measurements using LNG of typical compositions and pressures with literature values of the heat transfer coefficient correlation formulas for nucleate boiling heat transfer, forced convection heat transfer in boiling zone and heat transfer in mist flow. On this basis, we provided database control over the applicable ranges and coefficients used for each boiling pattern, thus creating empirical formulas applicable under a wide range of conditions. One example is shown in Fig. 4.

Shrock-Grossman [Ref. 1] 2.5(1/Xtt )

0.75

Actual measurement

l-x : Single-phase heat transfer

TP l-x
Guerrieri-Talty [Ref. 2] 3.4(1/Xtt )
0.45

coefficient determined using the liquid properties and defining fluid velocity as (1-x)v

TP

metalLNG
0 .9

1 x tt x

g l

0 .5

l g

0 .1

Fig. 4

Example of comparison with boiling heat transfer coefficient correlation formula

2.3. Analysis of shape of ice formed on seawater side As Fig. 1 shows, the fins are configured in a radial pattern on the external surface of the heat exchanger tubes (known as star-fins), thus enabling seawater to equally distribute in circumference and smoothly fall down along the surface. When the seawater contacts the bottom portion of the tube containing the low-temperature LNG, the temperature of the seawater falls and ice is formed on the external surface of the heat exchanger tubes near their bottom. If the ice becomes thick, it will gradually fill the gaps between the fins causing the ice surface to become tabular, thus resulting in reduced heat transfer area. The high heat transfer coefficient achieved due to falling film has traditionally been determined by an empirical formula, and this method is sufficient for practical purposes. Our understanding of changes in the shape of the surface due to ice formation has been based on actual measurements so far. We have recently established a method for calculating the shape of ice formed using the boundary element method [3] of heat transfer analysis. By repeatedly calculating the temperature distribution and revising the assumed shape of ice, this method determines the shape of the ice that is formed once the ice surface temperature reaches the freezing temperature of seawater (see Fig.5). We verified that the calculated values nearly coincided with the actual measurement values in a variety of conditions. One example is shown in Fig. 6.

Start Shape of ice film: hypothetical

Shape of ice film: revised

Calculation of external heat transfer coefficient based on shape of ice coating Outer surface heat flux distribution Internal surface temperature distribution

Alteration of shape so that external surface becomes the freezing temperature Boundary element method (BEM) Calculation of heat conduction

External surface temperature = freezing temperature?

Is the change in the shape of ice coating sufficiently small?

End

Fig. 5

Ice coating shape analysis flowchart

Solid line: calculated shape

Actual measurement (1) Analysis (1) Actual measurement (2) Analysis (2)

Dotted line: measured shape

Thin ice

Circumference

Medium ice

Height

Thick ice Heat exchanger tube center line

Fig. 6

Example of results of ice coating shape analysis

2.4. Analysis of optimum shape of heat exchanger tubes We integrated the aforementioned numerical calculation methods into the design method. Using this design method makes it possible to predict the performance of hypothesized heat exchanger tubes with a wide range of shapes only by calculations with no empirical tests, design an entire ORV on the basis of forecast performance, and evaluate costs more efficiently. In order to determine the answers to questions such as what tube shapes offer the best performance or reduce costs the most, we considered the question of optimization where the design variables were the dimensions representing the shape of the heat exchanger tube, and the objective function was the maximization of performance, minimization of costs, or the like. An optimum design method was constructed applying the response surface methodology [5] (OPTIMUS, VisualDoc) as a means of effectively determining the best shape of the heat exchanger tubes, and calculations performed in the sequence shown in Fig. 2. Using the response surface methodology, the responses of multiple peaks were replaced by an approximation surface, and methods of non-linear programming such as sequential quadratic programming (SQP) were applied to this surface to find the optimum solution. Furthermore, in order to analyze only heat exchanger tubes having a practical shape and dimensions for the optimum shape, we provided a variety of constraint conditions that considered material strength and fabrication methods on the wall thickness, fin dimensions, and others of the heat exchanger tubes.

3. EXAMPLE OF DESIGN USING NEW DESIGN METHOD

Fig. 7 illustrates a conceptual diagram of the optimization results obtained using the new design method under the design conditions for the new ORV installation at Tokyo Gas Ohgishima Terminal. The structure used was that of a time-proven and reliable ORV design, and the design variables were the dimensions such as tube diameter, fin height, number of fins and twisting pitch. In Fig. 7 the tube diameter representing the design variables is shown on the horizontal axis, and on the vertical axis are shown the ideal values for performance per panel and construction cost obtained optimizing the other design variables, such as fin height, for each diameter. (The improvement of performance per panel allows the number of panel and space required for installation to be reduced.) It was observed that cost performance improved as the diameter was reduced, but that beyond a certain size there was no particular advantage gained in terms of cost. Regarding performance, there was also found to exist a diameter that offers best performance and a tendency for the performance of tubes of a greater or lesser diameter to deteriorate. The focus was as a consequence narrowed down to a few heat exchanger tubes and spiral cores. Actual tests were then conducted using LNG, yielding results much the same as predicted by the calculations.

Maximization

Volume of LNG processedper panel Optimum point Heat exchanger tube diameter
Fig. 7 Example of results of optimization calculations

Cost

Minimization

On the basis of these results, we developed a new type of ORV with better performance (named the HiPerV) which, when compared to conventional ORVs, has drastically increased the LNG throughput per heat exchanger tube. Tokyo Gas employed this type for the new vaporizers at the Ohgishima Terminal. The heat exchanger tube of the HiPerV is of a larger diameter than conventional ones in order to increase the seawater flow and LNG flow per tube, and 12 star-fins are arranged on the exterior in

order to improve water retentivity. Additionally, the internal and external heat transfer areas are balanced and provide the proper shapes for accelerating internal heat transfer. This creates performance characteristics whereby, even though some ice is formed, the gaps between the fins are hardly ever filled up with ice, which would deteriorate the heat transfer performance. The panels of the HiPerV at the Ohgishima Terminal are 6m in height in order to maintain the balance with existing ORVs. However, actual tests using LNG were conducted on small-sized panels of 6m and 8m in height (see Photo 1) to enable designs incorporating 8m panels necessitated by specification requirements, and these confirmed performance to be satisfactory and fatigue strength to be at least as good as that of conventional models. The HiPerV at the Ohgishima Terminal (150 tons/hour x 2 units) was completed in November 2000. As Table 1 indicates, we successfully cut the number of heat exchanger tubes by about 50% (reducing the heat transfer area by about 25%) and lowered the floor space required for installation by about 30% compared with conventional models. As a result of these improvements and further steps such as by revising the specifications, construction costs were finally reduced by about 20%. Seven HiPerVs are presently being employed. In addition to the two at Ohgishima Terminal, these consist of four 135 ton/hours units in Japan and one 180 ton/hour in Korea.

Table. 1 Specifications of Ohgishima Terminal ORV

4. CONCLUSION
Considering the future energy market, demand for LNG will continue to increase, thus requiring more efficient, reliable, and economical LNG vaporizers. To meet market needs, we optimized and upgraded a time-proven and reliable ORV design in widespread use for LNG purpose. By applying the new design method to develop the HiPerV, we not only achieved a drastic reduction in costs but also verified that the HiPerV provided substantially improved operational performance. By applying this optimum design method, the HiPerV can meet a variety of users needs, such as cost reduction and reduced floor space required for installation. With the HiPerV, therefore, we firmly believe that Tokyo Gas and Sumitomo Precision Products can make a significant contribution to the construction of ORVs in the 21st century.

5. ACKNOWLEDGEMENTS
The authors wish to thank Professor Kataoka of the Department of Mechanophysics Engineering at Osaka University for his advice on the evaluation of the fluid analysis results and study of boiling heat transfer.

REFERENCES
1. Schrock, V. E. and Grossman, L. M. (1959). USAEC Rep. TID-14639 2. Guerrieri, S. A. and Talty, R. D. (1956). Chem. Eng. Prog. Symp. Ser., 52-18 3. Tanaka, M., Matsumoto, T. and Nakamura, M. (1991). Boundary Element Method 4. Kashiwamura, T., Shiratori, M. and Yu, Q. (1998). Nonlinear Optimization with Design of Experiments 5. JSME (1999). Introduction of Nonlinear Optimization with Response Surface Methodology, JSME course No.99-73