Aerodynamics is the starting point of the compressors because it’s directly connected to the efficiency and the LNG production: a 1% increase in the compressor efficiency is equal to 1% increase in the LNG production for this kind of plant capacity. New aerodynamic solutions are always under development to increase the overall efficiency. State-of-the-art tools, tests and operating experience contribute to the design. For aerodynamics, the best tool is computational fluid dynamics (CFD). CFD is a relatively young science within the turbomachinery industrial world as it was only introduced at the beginning of the 1990s. Since then, much effort has been made to calibrate the tool and assess the level of accuracy of CFD, in order to better understand the amount of margin that needs to be applied within industrial applications. All OEMs need to dedicate a considerable amount of time to compare test results and CFD code results, benchmarking different types of soft- ware. It is commonly known that each type of software has multiple parameters that can be tuned in order to better fit the test results. After a good calibration, a new stage should be tested.
Scale model testing is very useful for performance estimation and validation of any computational tool. The geometry of the scale model test stage is reproduced and dynamic similarity is achieved as far as concerns flow coefficient, Mach and density ratio. The OEMs have developed good correlation to predict performance for the relative full-scale impeller. A model test is a limited-power-consumption test, and this limitation can raise concerns in terms of not being able to correctly identify the left limit (surge or stall of the stage). To avoid this limitation, the use of picocoulomb (PCB) probes is advisable. A PCB probe is able to make unsteady pressure measurements. If a mistake is made in the design of the compressor and the model test has resulted in a lack of performance — such as shortness of operating range (choke or stall) — or under performance, the stage will need to be redesigned in some part. All the measurements taken provide a good baseline from which to check when something goes wrong, allowing a quick redesign of the compressor that goes in parallel with the manufacturing of a new impeller. Within one month’s time, a new and optimized impeller can be tested. Scale model testing is a low-cost, low-time, effective way to anticipate performance.
OEMs have developed rules and analyses to predict the impact of geometrical scaling on performance with no surprises during the full-load test. For LNG, the most critical stages are within the pre-cooling compressor as we have the combination of low temperature with high molecular weight and large volume flows that make for a high Mach number.
Continuous cost reduction in the oil and gas market has obliged OEMs to optimize the rotating equipment, for example by reducing machinery size. This leads to the reduction of the impeller exit diameter and an increase of flow coefficient. Traditional, pure centrifugal stages start suffering from poor efficiency and range when the flow coefficient exceeds about 0.12, even though some have produced stages up to flow coefficient 0.18. To improve efficiency and range for flow coefficient range (0.12:0.24) and relative Mach number range (0.8:1.05), an optimized family of stages has been developed. Performance characterization up to a Mach number of 1.05 was considered following the demonstration by Grimaldi, et al. (2007) that reasonable performance could still be achieved even with such a high Mach level.
This development was done following a three-step approach:
1. Reviewed all internal test data and correlation of performance decay of pure centrifugal stages with “simple” geometrical parameters (specific impeller curvature, hub to shroud blade length ratio).
2. Identical activity was done comparing CFD RANS (Reynolds Averaged Navier Stokes) predictions with test data.
3. A base design respecting all criteria identified in step 1 was constructed and only local CFD optimization was performed.
A validation test plan was developed, and the comparison between global performance parameters expected from CFD and the early test results showed excellent agreement. Noticeable improvements of range, efficiency and head were visible. Figure 6 compares efficiency obtained by the five tested stages plus an additional test performed on a lower flow coefficient impeller. The new design proved to be aligned with existing stages at a flow coefficient lower than 0.1 and more efficient at high flow coefficient.
Seals: Compressor efficiency can also be increased by minimizing internal linkages. Standard stationary labyrinths, in aluminum or steel material, are not effective to achieve this target, since their designs have to take into account multiple aspects, such as operating conditions or rotordynamic behavior, that affect the seal gap sizing.
As an alternative solution, abradable seals can be used (Figure 7). They consist of an integral rotating labyrinth that works against a statoric insert arranged into the diaphragm. Abradable seals don’t suffer the issues mentioned above for stationary labyrinths. An abradable coating allows for tighter clearances compared with traditional aluminum seals, and prevents the possibility of rotor damage in case of abnormal vibrations. This solution is easier to maintain because the inserts can be renewed by recoating the abradable material.
Vaned: A further option to increase the efficiency is to use vaned parts. Vanes can be introduced upstream of the impeller blades — inlet guide vanes — or downstream in the diffuser. Here, there are multiple choices, including a wedge diffuser (typical for IG compressors), low solidity diffuser (LSD) or rib diffuser.
The IGV is a good solution because it is able to straighten the large distortion coming from the volute, increasing the efficiency of the downstream impeller.
A vaned diffuser can increase the efficiency at average operating conditions, but can lower it for alternate cases. For LNG service, the operating point moves along the curve during a 24-hour period, so a good average efficiency is preferred.
Unfortunately, IGVs are a source of aerodynamic wakes that hit the impeller blades. In case of coincidence of frequency between IGV wakes and impeller natural frequencies, the impeller can break.
Side streams: Refrigeration compressors are typically equipped with injections between one stage and another to keep all the service within one casing, which complicates the stage-by-stage performance prediction . The same considerations for CFD design and model test validation can be applied to the statoric parts. OEMs have developed aerodynamic design practices to predict the impact of the shape of the injection stream onto the main stream. All major OEMs have validated side stream prediction rules with model tests performed both on test rigs or wind tunnels that have then been confirmed by full-scale tests.
However, real operating conditions may widely vary from design conditions for multiple reasons such as allowances on heat transfer equipment, calculated pipe diameters or equipment (pumps, vessels, etc.); favorable compressor performance relative to quoted polytropic efficiency; API margin; vendor margin; allowances on available turbine power; or any licensor allowances.
All the earlier considerations plus all the process and ambient variables make for operating conditions much different from the original design and reduce the real impact of the side-stream performance on the real compressor operation.
Casing: LNG compressors can be horizontally or vertically split. The main advantages of the horizontally split compressor is maintenance capability (compressor arrange between driver and another equipment) and larger volume flows. The inlet volume flow can reach 176,000 cfm (300,000 m3/hr) and require a 70 in (1.8 m) flange.
Compressor layout also impacts the downstream gas path, which should be studied in order to accommodate a smaller compressor inner diameter to minimize machinery dimensions and weight, without impacting overall bearing span. By reducing the size of the compressor, it is possible to increase the maximum allowable working pressure (MAWP), which can improve the plant layout and overall plant operation.
A new design adopts a bolted end-cover solution, similar to the barrel- type concept. This allows a step increase in design pressure. A recent project in Australia, characterized by an internal compressor casing diameter of 118 in. (3 m) and a design pressure greater than 580 psig (40 barg) has been successfully tested. An interesting collateral advantage provided by the new casing design is the reduction of rotor-bearing span, with consequent benefits on the lateral behavior and stability of the compressor. Reduced bearing span and higher design pressure can lead to future developments to boost profitability by reducing the number of casings and, therefore, capital expenditures.
The compressor casing must be compliant to API and ASME requirements, avoiding multiple design pressures and additional O-rings on the horizontal flange to meet the design pressure. O-rings have a limited life (five to 10 years) and need to be replaced from time to time, reducing machinery availability. Moreover, it cannot represent a definitive solution, because gas can bypass the O-ring and leak outside.
Compressor walls and flanges must be thick enough to grant a safe operation over the years, compensating for transient or out-of-design operation at the job site. The best way to measure casings is to use finite element (FE) tools that should be validated by each OEM. The worst condition to be analyzed is the hydro test pressure, which is 1.5 x MAWP. In order to further optimize the casing, a thermal transient analysis can be performed to determine the effect of operating temperature on the casing assembly.
Impeller: The Impeller can take advantage of new tools such as FE to improve its robustness, keep the peripheral speed limit at a reasonable value (984 fps [300 m/s]), reduce the weight and increase the torque transmission. Every new stage design starts with the aerodynamic path, but once it’s established, the optimum mechanical dressing can be achieved by the use of optimization tools and design of experiment analysis by modifying cover shapes.
Lateral: The most important advance in rotordynamics is the ability to measure the logarithmic decrement of the compressors by the use of a magnetic exciter installed at one end of the compressor. In main LNG compressors trains, typically both ends of the compressor are connected to other rotating equipment (electric motor or centrifugal compressor), so the logarithmic decrement measurement with the exciter is not possible during a full load test. A stability test must be accomplished during a performance test or a mechanical running test at reduced power on the effectiveness of the test.
Torsional: The most used technology for the helper/starter electric motor is the load-commutated inverter (LCI). This technology introduces a torque ripple in the shaft line that can lead to the failure of the coupling. In addition, the greater the size of the electric motor, the stronger the torque ripple. The torque ripple issue can be managed during a string test by instrumenting the coupling and measuring the torsional behavior. The electric motor parameters can then be tuned to minimize the impact of the torque ripple in the train.
Operating experience for rebundle
Some LNG plants have operated for 10 or more years, and some operational analysis can be done to evaluate and rebundle to maximize production. For example, equipment designed at the time may suffer low efficiency due to a technology gap. New high-Mach, high-efficiency stages have been designed to improve performance without impacting stage axial dimensions. In this way, new stages can be installed in the same casing, increasing the overall efficiency of the machine.
The development of new high-Mach stages now can offer more than four points of efficiency with respect to the old. A possible rebundle for a propane compressor can increase the efficiency from an average of 72 to 80% at roughly 10% lower power.
In another plant, a previous debottleneck of the plant brought the compressor to work beyond the choke limit in a region that potentially could be detrimental for the compressor. New stages with larger operating ranges can be installed to cover high flow points, extending the actual operating range.