Beyond double-checking

End-user

The large petrochemical complex in the following case studies produces olefins (ethylene, propylene and butadiene), aromatics (phenol, acetone, benzene and cumene) and polyolefins (polyethylene, polypropylene). In the plant, 205 tpd of LDPE is currently produced for each train, using the high-pressure ethylene process (free-radical polymerisation).

Machine monitored

One of the primary machines in the LDPE production plant is the autoclave reactor, the focus of these case studies. It is a 6.5 m long, 530 mm dia. pressure vessel with a self-contained motor, which drives a long shaft with many mixer blades mounted on it (Figure 1). The mixer portion of the reactor stirs the ethylene and peroxide mixture at high pressure to initiate and control the polymerisation process. It is a critical machine for LDPE production just like the other machines used in the process, i.e. the primary and secondary compressors and extruders. If any of the bearings in the mixer fail, the entire process must be shut down.

Figure 1. Autoclave reactor core showing location of the ATEX-certified accelerometers and the bearings which are monitored by these accelerometers. The mixer shaft bearings are identical.

Temperature, pressure and the flow of peroxides are regulated in the reactor to control the LDPE properties. Because of the sensitivity of the polymerisation process, this must be carefully monitored and controlled. If there is any kind of disturbance in the reactor such as a plugged output line or even a hot spot in a bearing, this could trigger an uncontrolled reaction that could increase the pressure extremely fast. There are two rupture disks installed on the reactor to prevent over-pressuring.

Decomposition (decomp) of ethylene can occur under certain pressure temperature conditions and result in the formation of hydrogen and methane, which are highly explosive. If a decomp leads to a catastrophic failure, it takes one day to change the mixer. Production loss is approximately 205 tpd, which is over €225 500 (at February 2020 LDPE spot prices). If there is secondary damage, it will take a longer time and higher maintenance costs to return the reactor back to service. Some plants change the mixer bearings frequently to ensure that a bearing failure does not provoke a decomp, but this results in unnecessary downtime. The plant in this case study has adopted a condition monitoring strategy to carefully follow the condition of the bearings, thereby extending the time between bearing replacement.

Monitoring strategy

The monitoring system commissioned at the plant is used for protection and condition monitoring of the reactor in each of the LDPE lines at the plant. The same system is also monitoring the primary and secondary compressors and the extruders in the same plant. This system is also used at other olefin/polyolefin complexes.

As shown in Figures 1 and 2, the four rolling-element bearings in the reactor are monitored by two accelerometers. The primary fault detection measurement used is a bandpass acceleration vibration measurement. The acceleration detection setting for signal response is nearly the same as that for RMS and peak-peak measurement detection settings. A broadband measurement with a frequency range of 1 – 5 kHz is surprisingly accurate in capturing the rolling element bearing resonance frequencies, which give an early indication of a developing bearing fault.

Figure 2. Machine view screen showing the traffic light status of the two measurement points and a real-time acceleration bandpass vibration measurement display.

Alarm limits for this detection measurement are normally established through experience. The first alarm (alert) is set to occur when there is approximately one month until the second alarm (danger). When the danger alarm has been reached, it is necessary to change the entire mixer (reconditioned unit).

After a bearing fault has been detected, envelope analysis is carried out to identify the location of the bearing fault and determine its severity. The motor bearings are different, the mixer shaft bearings are identical.

A velocity spectrum can also be used to identify the bearing fault frequencies if the noise floor is not too high. Sometimes the rotating frequency and its harmonics increase due to process related conditions, such as when polymer sticks to the blade, resulting in unbalancing the entire reactor. The low-frequency vibration amplitude also increases if the polymerisation process is unstable, but usually this condition disappears in a relatively short time. If a bearing fault develops while there is a low-frequency vibration increase due to a process condition, this can be easily seen in the velocity bandpass trend and spectrum.

Case studies of bearing faults on the reactor mixer

There are several potential failure modes of the reactor mixer but those associated with the rolling-element bearings are top of the list. Generally speaking, the entire degradation and failure mechanism of a single rolling-element bearing fault during normal operation is well understood and somewhat predictable, and most condition monitoring systems can detect and diagnose this fault. It is difficult in the following scenarios:

• Multiple bearing faults are occurring at the same time.
• Process conditions and noise dominate over the bearing frequencies.
• Different size bearings are being monitored at the same time.
• Bearings are lubricated by the polymer media, thus making them susceptible to adhesive bearing faults.

These conditions were individually or collectively present in the three case studies described below, which demonstrate the importance of an effective condition monitoring of the LDPE reactor mixer.

Case 1: DE motor bearing fault – ball faults give little lead-time

The driving end (DE) motor bearing, the most robust of all the bearings in the reactor (comprising two sets of rolling element bearings), is the bearing that fails the most often. As seen in Figure 3, some of the bearings have a relatively short life cycle; there were 11 shutdowns to replace damaged/worn bearings during a five-year period.

Figure 3. Bearing vibration trend over a 5-year period primarily showing the DE motor bearing. (Acceleration bandpass measurement, 1 – 5 kHz).

Rolling element bearing faults in the inner or outer race, once detected, give several months lead-time to failure for planning maintenance. Ball bearing faults, however, have a much shorter lead time to failure. In this case study, there was only around 10 days from the time the vibration signal exceeded the alert alarm limit to when the reactor was shut down.

Case 2: multiple DE motor bearing faults – delayed replacement

In certain situations, production requirements may delay when a bearing can be serviced. In such a case, a bearing fault is carefully monitored to ensure that the risk of bearing failure is still minimal until, for example, the next scheduled shutdown. Figure 3 shows five examples of this happening, where the bearing was allowed to operate some time after it exceeded the danger alarm limits. A single example of this is shown in Figure 4, which corresponds to shutdown #10 in Figure 3. The risk of allowing bearings to operate too much time after extensive degradation is decomp, which can be caused by hot spots on the bearing. Experience plays a large role here.

Figure 4. An example of a DE motor bearing vibration trend exceeding the danger alarm limits, where repair was delayed due to production requirements. (Acceleration bandpass measurement).

Case 3: middle shaft bearing fault – unexpected start-up fault

Sometimes there are short shutdowns for process or other reasons not directly related to the reactor. Shutting down the reactor carries risks in itself. In this particular case, when the process was started up again, higher than normal vibrations were observed. Over the course of just a few days, the vibration amplitude doubled, as indicated in Figure 5. The mixer was changed the next day. Upon disassembly, it was seen that hardened polymer had found its way into the bearing and caused the bearing to fail prematurely.

Figure 5. Orange: lower accelerometer measurement trend for the mixer shaft bearings. White: upper accelerometer measurement trend for the motor bearings. The trend is normal up to the process shutdown but immediately after the reactor was put back into service, the vibration increased. Although the damage was limited to the middle mixer shaft bearing, this influenced the vibration on the upper accelerometer. (Acceleration bandpass measurement).

Conclusion

As the reactor mixer is a critical machine in the LDPE process, the bearing fault detection and diagnostic cases in this article prove that an effective condition monitoring solution is an absolute imperative for this machine. It not only avoids costly production losses due to downtime, but also avoids consequential damages that occur as a result of decomp, possibly provoked by defective bearings. The bearing fault detection and diagnostics can be challenging for the reactor. As can be seen in Case 1, the bearing fault can develop quickly with little lead-time, or even in a non-linear progression as seen in Case 3. If production operations delay servicing faulty bearings, these have to be carefully monitored to avoid catastrophic failure and a possible decomp, as seen in Case 2. Moreover, premature bearing failures can be instigated by factors other than wear. This could be due to process conditions such as deposits on the mixer blades or even polymer that enters the bearing housing and can develop very quickly, as seen in Case 3.

Written by Michael Hastings, Brüel & Kjær Vibro, Denmark.

Read the article online at: https://www.hydrocarbonengineering.com/special-reports/06092021/beyond-double-checking/