January 2020
January 2020

No products in the cart.

Jl. Cakung Cilincing KM.1 No 7 (021) 4604687


Axial compressors are used in large-scale processes where a significant amount of gas is required at a relatively low pressure. Typical applications are at the oil refinery’s fluid catalytic cracking unit (FCCU) as the main air blower, at the steel-mill as the blower for the blast furnace or as part of an industrial or aeroderivative gas turbine. Centrifugal compressors are more commonly used due to their versatility: they can compress a wide range of flows to a high discharge pressure. The stable operation of axial and centrifugal compressors is limited by surge and choke. Surge is a violent flow reversal that occurs when the process restricts the compressor flow below a certain minimum value.

Choke occurs when the process does not create enough restriction to the compressor flow and the compressor operates at its maximum flow for a given performance level. Control systems are available to protect the compressor from surge, but not always from choke. Investigation of several catastrophic failures using diagnostics tools point to choke (not surge) as the root cause of the failure due to fatigue of a rotating blade or fixed vane. Operation in deep choke, especially of axial compressors during unloaded operation, therefore, should be avoided. The effect of choke in centrifugal compressors is generally overlooked and is typically of no major concern. This is in great part due to the natural shape of the performance curve that represents a significant change in flow from the limit of surge to the limit of choke (Figure 1).


In axial compressors, however, this distance is significantly less (Figure 2). The more stages a compressor has, the higher the pressure ratio but the smaller the operational margin between surge and choke regions. Figure 2 shows the performance curves of a 12-stage axial compressor. The operational margin of a performance curve is significantly less than the one of a 6-stage centrifugal compressor shown in Figure 1.

OEMs are not always clear about the precise location of the choke line as most attention is given to the surge line. The same mistake is transferred to control systems, which focus exclusively on protection from surge. However, ignoring the effect of choke in axial compressors has cost refinery owners months of production loss due to the catastrophic failure of the axial main air blower (MAB), one the most critical pieces of machinery in the refinery. Compressor choke or stonewall is an unstable operating condition, which occurs when the compressor is operating at low discharge pressure and high f low rate. This leads to increased gas velocity in the compressor.

The increase in gas velocity occurs until it reaches sonic velocity or resonance at the blade throat (Mach 1). At this point, no more flow can pass through the compressor, causing high frequency and low amplitude vibration of the rotor blades or stator vanes. During choke, the flow channels between blade rows may experience blockage effects.

Choke can occur at any performance level, i.e., speed or position of the variable stator vanes (VSV) for a constant speed machine. Although the choke increases with the performance level, long-term operation even at low-performance level in a deep-choke condition can be damaging due to its low visibility and cumulative effect.

Choke is difficult to detect by conventional vibration monitoring systems. When a compressor transits back and forth from the stable area to the choke area, a minor change in noise frequency can sometimes be heard specially at higher performance levels. Currently, no dedicated instruments are used for choke detection. Therefore, an axial compressor can be operating in a deep choke condition for a long period of time during unload condition at low performance levels (low speed or closed VSV position) without being noticed by the machinery operator.

The opposite is true when the compressor is operating in surge condition. The negative impact on the process and the potentially damaging effects of compressor surge are clearly observable. The cumulative effect of choke means that the most vulnerable part (rotor blade or stator vane) has a limited number of cycles prior to failure by fatigue. Material analysis of failed rotor blades has concluded that the initiation points of fracture related to operation in deep choke are located at the blade’s suction side (convex or curved side).

Fracture begins to develop causing a continuous decline in the blade’s moment of resistance, and bending stresses increase permanently until they exceed material yield strength. Fracture eventually occurs during operation at high load. Therefore, choke can easily go unnoticed for years until it produces enough cumulative cycles leading to component failure by fatigue. In the case of an FCCU, operation in choke condition typically occurs during the startup of the MAB prior to or during the dry-out period. Since the process is not yet ready to accept the full amount of air, no attention is given to the fact that the compressor requires some back-pressure to operate safely. Consequently, no pressure resistance is created from the process side. Further, the anti-surge valve may remain open.

The MAB, therefore, is accumulating a significant number of additional choke cycles that will eventually lead to failure. In many cases, catastrophic failure will not take place for ten or even twenty years of apparently stable operation after many apparently successful startups. The cumulative effect of choke-induced fatigue gradually leads to the failure of blower components. This is most likely to occur when the MAB is operating at a peak high load that is far away from the surge and choke regions. Without advanced diagnostic tools and proper analysis, damage is likely to be attributed to severe surging.

In industrial gas turbine applications, investigation has pointed to choke as the cause of premature fracture failure of blades in the first row of the axial compressor due to long and unstable operation of the combustion chamber during startup periods. Choke conditions Operators should be aware of the various operating conditions under which a compressor is likely to get into a choke condition such as during failure (open) of the anti-surge valve, or when the compressor is undersized for the desired operating conditions.

During start-up and under unloaded operation, too, choke can occur when the
anti-surge valve is opened too much or for too long, and no back-pressure is created by the process. If the anti-surge valve is oversized, high travel limitation (clamp) must be established in the controller to make sure the compressor operating point enters the stable operating envelope as soon as it starts and remains there during operation in unload condition. It is vital to
avoid compressor operation in a deep choke condition from the beginning of
startup (Figure 3).

Another choke condition takes place during loaded operation in combination
with excessive opening of the anti-surge valve due to manual operation or by valve low travel limitation (clamp) set by the operator. The operator sometimes applies low clamping during startup while the process is stabilizing to add extra cushion and prevent surge or fast opening of the antisurge valve by the control system.

In centrifugal compressors, low clamping wastes energy. But in axial compressors, it may also push the blower into the choke area (due to the narrow operating area between surge and choke). Therefore, clamping must be avoided or used with special care with axial compressors.

One further choke condition should be noted: In compressors operating in parallel, when one compressor trips, the natural response of a load sharing control system is to increase the performance of the running compressor to compensate for the loss. However, if process resistance remains unchanged, the running compressor will be pushed into the choke area.

To avoid such conditions, certain control measures should be implemented:
• Install an anti-choke valve immediately downstream of the anti-surge blowoff/recycle line and set up proper antichoke control, which must be independent from the anti-surge controller. Separate transmitters must be used as the distortion of a shared transmitter signal may have conflicting effects on the anti-surge and anti-choke controllers
• Make sure the process is always ready to provide enough back-pressure for
the compressor as soon as it is started. For example, place an initial set point of 10-15 psig at the FCCU regenerator. The slide valves should also be kept closed at the mechanical stop to help creating back-pressure. In the case of power recovery trains, the expander inlet and bypass valves can stay closed during startup of the train. If an anti-choke valve is present, start the compressor with that valve initially closed, then slowly ramp it open
• Perform a choke test in addition to a surge test to determine the choke line and establish a choke control line. Make sure the compressor can operate safely in the area close to the expected choke limit. A portable accelerometer can be used to determine if the axial blower is operating in a choke condition. In the author’s experience, radial vibration or axial displacement do not increase in a deep choke condition. An accelerometer placed in the casing
near the blower inlet has shown a clear acceleration peak when operating in a choke condition. At higher performance levels, choke can also be perceived by
a variation of the normal operating sound
• Graphically illustrate in the HMI the exact location of the choke line, surge line and control lines. As choke is not noticeable (unlike surge), it is
vital that it be graphically shown on a real-time compressor
• When operation near choke is detected, the control system should generate an alarm. In the absence of a choke valve, the operator may be able
to apply corrective measures, such as increasing the regenerator
• Properly size and test the anti-surge valve to make sure it does not allow for operation in choke when 100% open in an unloaded condition. The
selection criteria for the capacity of anti-surge valves may be different for axial compressors compared to centrifugal
• Have an advanced diagnostic tool that stores high resolution trends and events so catastrophic damage can be correctly attributed
• It is recommended that vendors develop and install an additional signal to the vibration system, such as an accelerometer from the compressor
case to generate an alarm and trip due to operation in choke condition for a certain period of time.

Leyden Lopez is Director of Leyden Turbomachinery Corporation of Houston, TX, providing commissioning and consulting services on turbomachinery controls. For more information, visit




Most gas turbines have axial flow compressors with multiple rows of rotor blades and stator vanes that number into the thousands. Failure of any blade or vane can lead to severe collateral damage in both the compressor and turbine section of the engine.

Root cause analyses (RCA) often find foreign and domestic object damage and operational issues as among the usual suspects. However, shim protrusion and liberation and stator vane fretting wear are also becoming common reasons for failure. Other factors that can accelerate fretting wear include flow disruptions from casing geometry changes, air extraction points, rotor blade clocking, and higher frequency of start and stop cycles due to the demand for plant flexibility to accommodate solar and wind power.

Protruding and missing shims are regularly found during boroscope inspections and overhauls on GE gas turbine compressors. In some cases, a protruding shim can create a flow blockage of more than 15%. Blockage results in a force pulse which produces an alternating force or stimulus on downstream rotating blades. Since shims are generally located near horizontal joints, this produces one or two pulses per revolution. At these locations, force impulse on the rotating blades is the result of the area change in the stationary vane flow passage. This is exacerbated by trailing edge vortices that can upset the flow path following the protruding shim.

As the force profile is narrow, there are multiple harmonics which can contribute to circumferential force distribution on rotating blades. If any harmonic frequencies are resonant with blade natural frequency, blade vibratory response is amplified at this resonant condition.

At a resonant condition, blade vibration stress is directly proportional to the magnitude of the stimulus or alternating force on the blade; it is inversely proportional to the damping of the bladed system, and proportional to the resonant response factor and steady gas bending load on the blade.

The resonant response factor is the attenuation in response due to the phase relationship between the alternating force, or stimulus on the blade, and the particular mode shape. This method for predicting blade response can be used to evaluate the dynamic response of a compressor blade due to protruding shims. But it is important to understand that blade natural frequencies are dependent on several variables. They should not be predicted as discrete frequencies but as a range.

Temperature affects frequency due to the dependency of Young’s modulus on material temperature. Furthermore, manufacturing tolerances can affect frequency due to vane envelope tolerances and dovetail fit variations. This must be addressed by calculating all frequency ranges for the fixation point at the top and bottom of the contact area in the dovetail slot across the full range of operating temperatures.

This information can then be used to construct a Campbell diagram to illustrate the relationship of blade natural frequencies with the stimulus frequencies from the rotational speed of the compressor. As an example, consider the evaluation of the row 15 rotor blade of a 7FA compressor. One potential resonance was found with eight pulses per revolution (there is a stimulus force that occurs eight times for every revolution of the rotor) at the first tangential mode of vibration. Blade vibratory response was calculated as a function of different protruding shim conditions and assuming various values of damping (Q). Damping can be variable and sensitive to various factors such as dovetail fit. However, previous research and testing suggests that the damping for this blade design should be in the range of Q=200– 300 (Morgan P. Hanson. Effect of Blade- Root Fit and Lubrication of Vibration Characteristics of Ball-Root-Type Axial- Flow-Compressor Blades. Cleveland, Ohio; Lewis Flight Propulsion Laboratory, National Advisory Committee for Aeronautics, June 1950).

The blade vibratory stress at the eight pulses per revolution resonance would be in the 20,000 to 30,000 psi range if there was only one shim protruding 0.75 inches. This response is about 2.5× the predicted response at resonance without a protruding shim. For the case without a protruding shim, prediction was based on a random variation of stator forces (namely vane variable spacing and orientation) with a maximum variation of 1.5%. For two shims protruding at 180 degrees separation, the response is in the range of 40,000 to 60,000 psi. The blade failure mode due to stimulus from the protruding shims is high cycle fatigue.

A Goodman Diagram is typically used to evaluate the fatigue strength of a material under the combined influence of steady and alternating stresses. It shows the potential blade response due to protruding shims is higher than the deteriorated endurance strength for the material. High cycle fatigue is a possibility depending on the operational profile and the number and height of the protruding shims. Since the resonant frequency is 480 hz, fatigue cracking could occur in as little as ten million stress cycles which is less than six hours of operation on resonance.

Other major problems for compressors are loose vanes due to fretting wear in the casing T-slot groove, and severe wear on the vane attachment hooks. The average and maximum wear is always greater in the upper half (UH) casing and vanes. The worst location is at the horizontal joint on the right side looking downstream.

The reasons: (1) Aerodynamic forces push all the vanes over against the left side keeper bar in the UH, thus allowing the right-side vane to be the last one in the pack, making it rather loosely held. (2) This right-side vane in the upper half is notched to accommodate the keeper bar. Therefore, it does not have a full attachment hook in the base. This results in accelerated fretting wear. Typically, the average upper half vane looseness is nearly three times that of lower half vanes.

Variations in vane spacing and orientation must be kept small to have acceptably small harmonic excitations on downstream rotor blades. Platform lift creates a flow disturbance which will impact performance. An analysis was performed assuming a vane lift of 0.04″ in half the rows of vanes can result in a 1⁄3% loss in efficiency. Left unresolved, tip rock can lead to vane clashing with the compressor rotor blades, vane liberation, and compressor failure. Vane fretting wear and shim migration into the flow path are addressed by preventing vane movement. The lower half casing is where many vanes can be found solidly locked in place due to corrosion and deposits from the flow path.

One way to lock vanes is to pin them together. This procedure uses dowel spring pins inserted into holes drilled into the circumferential faces of the vane bases. Other vanes are attached, and more pins are added until a half ring vane segment is formed in both the upper and lower casing haves. This forces the vanes into an evenly spaced condition that reduces the average harmonic stimuli on rotor blades. In addition, this greatly reduces fretting wear, the potential for liberation of vanes and shims, and protruding shims that can lead to rotor blade high cycle fatigue stimulation and failure. This process has proven reliable and has successfully been implemented on more than 200 turbines over a period of more than 17 years.

Other solutions include: • A casing patch ring or weld repair and re-machining of the casing, as well as a complete replacement of the compressor vanes with a shim retention method. However, this approach does not address tip rock, i.e., wear of the hook fit and vane base due to fretting, which can occur over time • Replacing individual vanes with a multi-vane segment with a shim retention method, i.e., multiple vanes are machined in to a common or continuous solid base.

Across the large, worldwide fleet of gas turbines, there will be a statistically small number of compressor failures due to variability and wear impacting Goodman stress capability. When potential blade stimuli are identified, such as protruding shims and vane looseness, it is a best practice to eliminate them. The key to success of any solution is to have and maintain sufficient dampening over the wide variety of operating conditions.

Taking care of shim protrusion, shim liberation, vane fretting wear and tip rock will improve the general health and performance of the fleet. ■

Robert Traver is Senior Engineer at CTTS, a company established in 2019 to provide on-going fleet support for compressor vane pinning. Rodger Anderson is a Consultant who recently retired from DRS Technologies. For more information visit:




An obvious objective is to confirm the compressor will perform as predicted. But an equally important goal is to minimize the risk of unexpected aero-mechanical issues during field installation, especially in remote locations such as offshore platforms. It is far better to uncover potential problems before the equipment leaves the original equipment manufacturer’s (OEM) facility.

High-pressure and high-power density compressors typically experience rotordynamic (instability) and aerodynamic (stall) complications more often than compressors intended for petrochemical service. Thus, the demand for more rigorous in-house testing has grown. In response, OEMs and the turbomachinery community developed a variety of test programs and facilities to demonstrate that equipment meets end-user requirements.

These options include:

• Volume reduction, inert gas tests using small open- or closed-loops (typically called Type 2 testing according to PTC 10, “Performance Test Code on Compressors and Exhausters,” ASME International, 1997)

• Full-load, full-pressure, full-power testing with the contract drivers, gears, auxiliaries, complex piping systems, and so on (typically labeled Type 1 testing)

• Each type of test offers insight into a compression system’s aerodynamic and mechanical behavior. Type 2 low-pressure inert gas performance tests are set up based on compressor volume reduction (i.e., exit volumetric flow divided by inlet volumetric flow).

A critical parameter in assessing multi-stage compressor performance is the combined volume reductions of the individual stages to determine the overall volume reduction of the compressor. Thus, the volume reduction of stage 1 determines the volumetric flow into stage 2, which sets the flow into stage 3, and so on.

Volume reduction tests use readily available gases, such as nitrogen, helium- nitrogen mixtures, carbon dioxide and R-134A refrigerants. The selection of test gas or gases, operating speed, and so forth are made to match as closely as possible the volume reduction the compressor will experience in the field.

Type 2 tests are also run at lower pressures than the compressor will experience in field operation. A Type 1 test, per the ASME code, is run with a gas mixture close to the gas compressed at or near operating conditions expected at the site. If it is possible to conduct the test using the actual field gas and field inlet conditions there may be no deviations between the field and test conditions; however, this is rare.

In most instances, the inlet temperature at the specified condition cannot be achieved due to an OEM’s test stand cooling capacity limits, and it is often impossible to exactly match the field gas. While a different gas can be used, it must have a k-value (a measure of thermal conductivity) close to the field gas to ensure test and field Mach numbers are equal and the thermodynamic conversion of the work input to the gas produces the same pressure ratio.

Generally, Type 1 test results accurately reflect field performance levels and end users should expect little or no difference between the performance measured during Type 1 testing and those in the real world. The Type 1 test exposes compressors and their auxiliaries to near field conditions. It offers a greater likelihood of uncovering abnormalities in performance (mechanical or aerodynamic) compared to Type 2 tests. Therefore, end users must weigh the higher cost of the Type 1 tests (as much as 10 times) against potential production loss that might arise.

Test selection

Numerous factors can impact the aerodynamic and mechanical behavior of turbomachinery and the selection process for Type 1 or Type 2 tests. These mostly relate directly or indirectly to the higher pressure and aerodynamic loading experienced on the Type 1 test. Primary among these are deflections or relative movement of parts. As a result, there can be subtle differences in the primary and secondary flow path geometries under Type 1 and Type 2 test conditions.

For example, one of the most common geometric changes that might occur during a Type 1 test (but not under Type 2 test conditions) is bundle deflection due to manufacturing tolerance stack-up. In a typical re-injection compressor, the highest pressure occurs in the last-stage diffusers (for both straight-through and back-to-back arrangements). Thus, forces on the sidewalls of the last-stage diffuser will cause the diffuser walls to move apart. As the last-stage diffuser is also typically the narrowest in a compressor, its behavior is more sensitive to axial deflections.

The amount of deflection depends on the machining tolerances of the bundle components and the fits, as well as material deflections that might occur from pressure forces or thermal effects. Typically, maximum deflections are in the tens of thousandths of inches (or tenths of millimeters) and of little consequence in most compressors.

However, if the design diffuser width is narrow (often the case with high-pressure reinjection compressors), tens of thousandths of inches can be a large percentage of the diffuser width. It is possible for the width to increase sufficiently for diffuser rotating stall to occur, resulting in high levels of subsynchronous radial vibrations. This probably would not be evident during a lower pressure Type 2 test. In one instance, had a Type 1 test not been run at the OEM facility, this exact problem would not have been discovered until the compressor came on-line in the field. The significant warranty implications to the OEM and lost revenue to the end user due to production disruptions would likely dwarf the cost of a Type 1 test. It is advisable, then, that end users, process engineers, and OEMs discuss compressor testing at the OEM’s facility to address aero-mechanical concerns prior to shipping equipment to the site. ■

James Sorokes, is Principal Engineer at Siemens Oil & Gas in Olean, NY. A paper on this topic, “A Comparison of Type 1 versus Type 2 Testing – Recent Experiences Testing A High Pressure, Re-Injection Centrifugal Compressor,” (Sorokes, Kocur, et al) was presented at the 2019 Texas A&M Turbomachinery & Pump Symposium. For more information, visit