September 2018, Vol. 245, No. 9


Machinery Torsional Issues in Pipeline Stations

By Amin Almasi, Engineer

Pipeline compressor trains and pipeline pump trains are the backbone of gas pipelines. The torsional issues and problems reported for pipeline machineries, such as gas compressor trains and pump trains, can sometimes cause operational problems, even damages or failure.  

Tools such as advanced modeling and modern simulation programs are employed to optimize the machinery trains in pipeline stations. These trains should meet the highest efficiency and the proper flexibility in the operation. 

The results of various studies should be used to assess and manage the risk of any mechanical, operational, control, and electrical issues such as torsional vibration issues, rotor-dynamics problems, performance shortfalls, anti-surge issues, different dynamic instabilities, alternative operating scenarios or resonance cases. 

For pipeline machinery trains, there are a wide variety of planned and unplanned incidents related to electrical, mechanical and aerodynamic excitations, which can cause torsional issues. Torsional responses will generally be multimodal with a slow decay rate because of the light damping in typical pipeline machineries. 

The fact that the damping of torsional modes is very low can lead to the generation of high torques in the machinery shafts as result of resonances or transient excitations. The effect on the amount of shaft fatigue expenditure may be much higher that expected, because of the nonlinear nature of the fatigue process. As a result, unexpected fatigue failures could occur. 

The torsional modal damping values may increase as result of the aerodynamic damping (inside machinery, such as a compressor and turbines), or by using some kind of flexible elements in the connections, particularly non-metallic flexible couplings. Some of the more significant damping mechanisms are:

  • Fluid forces on the impellers, blades, shafts, rotors, seals, and other components 
  • Shaft material hysteresis, particularly at high levels of oscillating strain 
  • Energy dissipation from coupling such as slippage (friction) during high torsional oscillation or damping in non-metallic elements 
  • Bearing oil losses or magnetic bearing controlled-damping effects
  • Various mechanisms of electrical damping

The magnitude of the modal damping varies with the mode number. The modal damping is depending particularly on the machinery train load, torsional mode number and train configuration. 

Most of the individual damping mechanisms are complex and could not be accurately predicted at the design stage. Usually, a comprehensive sensitivity study should be performed to identify the damping effects. 

Torsional Vibration 

To achieve a good accuracy in torsional analytical results, the required number and distribution of elements along the axial length of a rotating machinery train should carefully be selected. An optimum number of elements should be identified. The best guidelines are most often gained through experiences. 

Considerations including the vibration response frequency range of interest, number of locations that have distinctly different diameters, geometric discontinuities, and relative values of stiffness/inertia for discrete span of each rotor should be considered for proper torsional modeling and interpretations. 

Generally, the main body regions of an individual rotor or shaft have significant larger diameter sections than the rotor extension at each end. These shaft extensions often contain seals and bearings and may terminate with connection to other shafts or couplings. Therefore, proper modeling should be used to accurately consider all these.

Great care should be taken for torsional modeling of the electric motor rotor assembly. Particularly, in some rotors of electric motors, the cross-section does not remain planar during twisting; this is known as the “wrapping effect”. 

This effect results in a form factor much lower than the second polar moment of area of the cross-section. In addition, for rotors that contain materials embedded in slots or cavities, because of centrifugal stiffening effects, the stiffness of the rotor might be a function of the rotor speed. 

These effects are difficult to determine analytically. Rotating torsional shaker tests could provide useful information that qualifies this effect and help to properly assign parameters to better simulate or interpret torsional behaviors. 

One of the forcing function properties in the torsional vibration for a rotating machinery train is the torque and its effect of electric motor air gap. Torsional excitation torques should be calculated using electromagnetic mathematical models that simulate the rotating machineries, control systems, auxiliary system and all equipment in the loop. For electrical motor driven trains, because electrical system linked with torsional effects, a large portion of electrical system such as a transformer, electric transmission network and other important items should be included. 

In general, the calculation of electrical torque on electric motors might be assumed uncoupled from the torsional vibration point of view. However, for some excitations, such as sub-synchronous excitations, there is a cross-coupling between the machinery torsional response and current oscillations in the electrical network. 

Coupling Selection

The couplings should be selected based on torsional and rotor-dynamics considerations. Selected couplings should also accommodate operation misalignment, such as thermal misalignment, as the temperature changes. 

Some machineries will be heat-up; this makes the coupling selection and sizing challenging, particularly for large pipeline machineries. Often a coupling is required to transmit or isolate thrust loads. In general, there are two main classes of coupling used on pipeline rotating machineries:

Flexible couplings (“non-metallic-element flexible couplings”): These couplings should only be used in special circumstances since they offer relatively high maintenance and sometimes operational problems. For some machinery, it is necessary to specify a non-metallic flexible coupling with a relatively low stiffness and relatively high damping ranges.

High-torsional-stiffness coupling (“quasi-rigid couplings”): Most couplings of large rotating machines fall into high-torsional-stiffness coupling category. However, for some special application they may offer some issues. 

With the ever-increasing demand in the large pipeline machinery trains, more reliable and better couplings are required. Modern applications have relied on high-torsional-stiffness coupling (or metallic-flexible couplings), which are designed for infinite life and require relatively low maintenance. There have been many different designs for metallic-flexible element couplings. However, coupling for pipeline machineries are usually disc or diaphragm coupling.

Each of these can be further separated into different styles (sub-groups). Disc type couplings accommodate misalignment by using flexible discs which are connected by alternating bolts to opposing flanges. The thickness of the discs and distance between bolts determine the amount of flexibility. These couplings can be circular, scalloped or straight-sided, depending on the shape of the flexible discs.

Diaphragm couplings accommodate misalignment by allowing movement of outer diameter relative to inner diameter. The diaphragm is plate shaped with a contoured profile machined into one or both sides. The amount of flexibility is determined by the thickness of contoured profile and the difference between outer and inner diameters. 

Both disc-type and diaphragm-type couplings can be arranged in a reduced moment configuration. In this configuration, the flexible elements are attached outboard of the shaft end, allowing one size of coupling to accommodate a large range of shaft sizes. In this configuration, the flexible elements are moved over the shaft end, as close to the shaft bearings as possible, to minimize loading on the machinery bearings. 

If a coupling is shrunk onto a shaft, the fit should be adequate after accounting for all centrifugal and thermal effects to prevent slippage under any expected operation condition. For many large machinery trains, often suitable couplings cannot be found, and the rigid connection should be used. This is the flange-to-flange connection without a coupling; it should be avoided. 

Study Inaccuracies

As a rough indication, about 70% of all torsional calculations are relatively inaccurate. Too often, the real machinery train behavior is completely different from the multi-page torsional calculation report. Therefore, in large numbers of occasions, things go wrong despite inaccurate torsional calculations. 

The weakest link in the chain usually is an inaccuracy of specified coupling torsional flexibility values. Many cases of these inaccurate flexibilities are reported. For non-metallic flexible couplings, such as couplings with elastic elements, inaccuracy could be high. 

Metallic-flexible couplings may be better in this regard. A good recommendation always is the torsional verification test as part of the machinery site performance test for each unit, even one-by-one test for identical units. 

With increasing the shaft power (increasing the load), the natural torsional frequencies slightly shift. Often, natural torsional frequencies move to higher values because flexible couplings often become stiffer with increasing the torque. 

A major reliability problem arises once these shifted torsional frequencies coincide with an excitation frequency. As result, high vibration can be observed or even in extreme cases coupling may fail. This failure could come with initial cracks in metallic or non-metallic components. 

At the standstill coupling, these cracks are most often not visible. These cracks may be viewed with a stroboscope when the coupling is under load. For example, couplings with rubber material elements, particularly coupling deigns that use rubber elements at the shear loading, are vulnerable against the crack propagation, premature failure and fatigue. 

Measurements, Monitoring

Commonly used torsional monitoring systems collect torsional data from torsional monitoring sensors which could be magnetic pick-up probing the teeth gear signal, strain gage method (usually with telemetry), encoders on shaft, or advanced measurement devices using laser optic systems. Torsional vibration measurements should be frequency-demodulated. 

The frequency spectrum of demodulated output can show the torsional excitation frequencies. In simple terms, the haystacks in frequency spectrum reveal the position of torsional resonance frequency. When such a haystack coincides with one of the torsional excitation frequencies then the premature failure of the weakest link in machinery train is inevitable. 

Torsional vibration measurements can provide feed-backs into the torsional models and studies, and even quantify torsional studies. Necessary measurements can provide some indications to improve the train torsional model and to correct the theoretical torsional natural frequencies and torsional responses. Corrections should be implemented to refine the models and parameters. 

These corrections should be based on theoretical knowledge and measured trending. Particularly, measurements could provide feed-backs to verify coupling torsional stiffness accuracy, damping values and other key torsional data. There is a possibility to dig-out the casing vibration, particularly gear unit casing vibration (in trains containing gear unit), to obtain some torsional vibration data. It is not easy, but it is often possible.

Identical Trains 

Identical machinery trains are commonly used in many pipeline stations. There is a classic torsional problem of identical machinery trains that do not behave identically. For example, on large machinery, a flexible coupling fails, vibration measurements are high, or a control system is instable while the other identical units run trouble-free. 

Often, the root-cause is an actual torsional coupling flexibility that is off the specified value. Particularly on one unit, the first torsional resonance frequency may coincide with one of the train excitation frequencies, whilst other units run resonance free (and consequently trouble-free). 

Surprisingly, modal damping values (and even sometimes the stiffness) have also been observed to be slightly different on nominally identical machinery trains working under the same operating conditions. This occurs presumably because of tolerances of manufacturing or installation.

Sometimes, with identical trains running in parallel and the same power load sharing, the first torsional resonance frequency may be varied over 10-20% which implies a variation of flexible coupling stiffness of 15-30%. This is far more than 5% accuracy that is claimed by many coupling manufacturers.

Compressor Case Study 

This case study is for a gas turbine driven gas pipeline reciprocating compressor using a gear unit for the speed match. 

Considering that the driver system included a gas turbine and a gear unit, after successful commissioning of the gas turbine separately, and after connecting the gas turbine to the gear unit and re-commissioning, it was confirmed that the gas-turbine and gear-unit driver system has no problem in the no-load condition, and the noise, and the vibration were in normal levels for both the gas turbine and gear unit. 

After this stage, the coupling spacer was set between the gear unit and compressor. Before the gas turbine speed reached 2,000 rpm, the vibration was normal in the gear unit and gas turbine. 

However, when the rotating speed got gradually around 2,000 rpm and pass it, the noise and the vibration in the gear unit grew. Before reaching to the rated speed of about 4,500 rpm, the vibration in the gear unit got the maximum allowed limit, and it caused the compressor train to stop by the high vibration trip. 

The commissioning procedure was repeated based on manufacturer’s instructions. However, that was not successful. After the initial study, it was concluded by the commissioning team that there might be a problem with the coupling spacer between the compressor and gear unit. 

After some modifications in the spacer coupling, the compressor could reach 4,500 rpm of operation speed with no load. Within the first minutes, the vibration was normal, but after the compressor was loaded, the vibration got much and caused trip. Despite many changes in the coupling and spacer between the gear unit and compressor, the problem was not solved. 

The next step was the inspections and verifications to make sure the correct gear teeth contact pattern was established. Because of dynamic effects, the static gear teeth contact check is not useful. The inspection was done to verify dynamic gear teeth contact pattern; it was finally confirmed that dynamic gear teeth contact pattern was correct. 

The following observations should be noted for this reciprocating compressor train:

The load change can cause a different vibrational behavior. The load change can change the torsional stiffness of components which could result in changes of torsional natural frequencies that can trigger a torsional resonance and high vibration. 

A gear unit can potentially translate the torsional vibration to the lateral vibration. In addition, a gear unit can cause the torsional-lateral coupled vibration.

Observations on many reciprocating compressors in the large size ranges for pipeline services have shown that for large reciprocating compressor trains – above 2 megawatts (MW) – when using high-torsional-stiffness couplings, the first torsional natural frequency might be in the range of the operating speed and the torsional resonance could easily be occurred. 

There should be significant pulsating forces and torques for large reciprocating compressors driven by steam turbines or gas turbines, above 2 MW. In typical large reciprocating compressors, which are usually driven by large electric motors, electric motor rotor assembly provides 55-70% of moment of inertia of the whole compressor train and the flywheel typically supplies only around 25-35% of moment of inertia of the whole compressor train. 

For relatively small reciprocating compressor trains, for example, in the range of 100-700 kW, the rigid connections or the high-torsional stiffness couplings (or metallic rigid coupling) can be used and the first torsional natural frequency could be maintained well above the reciprocating compressor operating speed (usually in the 300-400 rpm range) and its main harmonics. For the large reciprocating compressor trains (above 2 MW), the situation is completely different. There could be two options as explained below. 

For the direct-drive electric motor trains, it may be possible to use a rigid flange-to-flange connection to keep the first torsional natural frequency sufficient above the reciprocating compressor operating speed (often around 330) and its main exciting harmonics.  

This is a desired arrangement for reciprocating compressor trains. The key for success in this arrangement is the sufficient torsional rigidities for all components. 

For large machineries, the torsional stiffness of all shafts, connections and components could be relatively low (because of large sizes). Any carelessness or inaccuracy in torsional modeling or studies can result in torsional problems because first torsional natural frequency is marginally above the reciprocating compressor operating speed (or harmonics), separation margin is usually low, and the situation is fragile. 

If the actual compressor train behaves softer than the theoretical model, the torsional resonance could easily happen. 

The second option is using the flexible coupling, often in the form of rubber element couplings, for large reciprocating compressors. This method is commonly employed by some vendors for large reciprocating compressors (above 3 MW). 

This design uses a relatively soft train concept to reduce the first torsional natural frequency significantly below the reciprocating compressor operating speed (about 330), for example to 30-40% of the operating speed, considering that for these ranges the separation margin is significantly large and inaccuracy of modeling has much less effect on the actual performance of compressor.

An accurate torsional study was done on this gas turbine driven train that used the gear unit and high torsional stiffness coupling (metallic couplings). It was shown that the first torsional critical speed (which was re-calculated based on accurate data) was just 10-15% below the operating speed. The first torsional critical speed was around 279 rpm a bit lower than the reciprocating operating speed 330 Hz. Because of this insufficient margin, when the compressor was loaded, a slight shift in the stiffness and consequently a slight rise in the first torsional critical speed can cause a dangerous torsional resonance. 

On the other hand, due to low moment of inertia of the compressor train there were significant pulsating forces and torques that could worsen the situation which resulted in high vibration. To solve the issue, a rubber-element flexible coupling (jaw coupling) was select for the gear-unit and compressor connection to decease the first torsional critical speed; it is to provide a better separate margin and also provide significant damping effects. 

This additional damping is important considering insufficient moment of inertia of the compressor train. New coupling resulted in the first torsional critical speed of about 122 rpm, 37% of the reciprocating compressor speed (330 rpm). 

As the result of a better separation margin and more damping effects of new coupling (rubber-element flexible coupling), the reciprocating compressor train was successfully commissioned with a low vibration and trouble-free, satisfactory operation.  P&GJ


Author: Amin Almasi is an engineer and senior rotating equipment consultant in Australia. and an IMechE and registered professional engineer in Australia and Queensland. He specializes in rotating equipment, condition monitoring and reliability.


Related Articles


{{ error }}
{{ comment.comment.Name }} • {{ comment.timeAgo }}
{{ comment.comment.Text }}