While Part 1 of the Case Study, By using vibration response to a soft foot defect with associated DCS Process Data to illustrate why improved operation practice is essential for Mechanical Reliability - Case Study, Part 1, provided a broad overview on the integration of both online vibration and DCS process data to identify a soft foot defect, Part 2 expands upon the technical aspects of the case study.

Understanding Melt Index

To meet various customized specifications, especially viscosity, which impact extrusion gearbox loading and power consumption (kW), the polymer industry uses another term, melt index (MI). It is measured differently from the other more popular kinematic unit, centistokes (cST), as shown in Table-1.

Table 1 – Viscosity vs Melt Index

For this case study, five keys were selected: MI or g/10 minutes; capacity or pound per hour (PPH); electrical power (kW); start-up; and overshoot, both during start-up and online. Additionally, 15-minute cycle velocity overall (IPS) trending is used since only a motor 1x rpm peak exists. Although available, either spectra or waveform are for reference only. The other data source is from distributed control systems (DCS), which is updated every few seconds. Although both data are not synchronous, as long as an adequate time frame is picked up, the case study’s goal can be still focused on: How does vibration respond to each DCS’s respective variable?

Observing Vibration Peaks

Figure 1 (showing nine days duration in macro) includes all factors and relates MI, PPH and kW with the associated vibration response, which was picked up due to its weird pattern at first glance. Most importantly, each factor occurred only one at a time, which makes it easier to distinguish its individual contribution on vibration.

A Figure 1 break down looks like this:

1. On 11/12/2018 (green color highlighted):
   MI (or viscosity) online transit from 12.1 to 2.2 became thicker while PPH (feed rate) stayed stable (~ 28088);
2. On 11/13/2018 (orange color highlighted):
   PPH (feed rate) online transit without overshoot (+4004) went from 29985 to 33989 while MI (or viscosity) stayed stable (~ 2.2);
3. On 11/16/2018 (red color highlighted):
   PPH (feed rate) start-up transit without overshoot (+35985) went from zero to 35985 while MI (or viscosity) stayed stable (~ 2.2);
4. On 11/17/2018 (red color highlighted):
   PPH (feed rate) online transit (+831) went from 35985 to 36816 while MI (or viscosity) stayed stable (~ 2.2).

Figure 1: Overview macro from 11/10/2018 to 11/18/2018

Four vibration peaks from 11/10/2018 to 11/18/2018, as shown in Figure 1 with associated chronological DCS events, are listed in Table 2.

Table 2 – Vibration Peaks and Associated DCS Events

What was observed was not at all like a motor start-up transient, which is seconds from idle to a designated constant 1200 rpm and requires real-time sampling. That’s because it would never be caught in a 15-minute scanning cycle of overall velocity, let alone network latency for whatever reasons. Rather, the peak at 0.228 (inch/second) on 11/16/2018 start-up was definitely feedback to extruder loading status and not rotating speed.

Figure 2 is a 10-hour micro (magnified) view for Figure 1’s nine day macro view to provide a close-up for online and start-up transits, both without overshoot as follows:

The small, solid, black blocks shown in Figure 2 were used as a reference to get a better visual comparison for how long to complete the process procedure.

For example, both were without overshoot at the same MI (~ 2.2):

1. 11/16/2018 (upper Figure 2):
   Start-up PPH (feed rate) transit jumped +35985 (from zero to +35985) with much slower (~ 5x blocks) duration.
2. 11/13/2018 (lower Figure 2):
   Online PPH (feed rate) transit jumped +4004 (from 29985 to 33989) with much faster (~ 1.5x blocks) duration.

Figure 2: DCS micro after zooming in

1. Start-up transit (without PPH overshoot) while MI (viscosity) stayed the same;
2. Online transit (without PPH overshoot) while MI (viscosity) stayed the same;
3. MI (viscosity) transit while PPH (feed rate) stayed the same.

Electrical kW sharply increased as a response to MI (viscosity) transit, from 12.1 to 2.2 (or became thicker and more viscous), which can be clearly recognized on 11/12/2018 by the green arrow and was proportionally variable and dependent on PPH (feed rate) capacity on the other days.

Adding a New Factor

Since production was scheduled in advance and operations had no active initiation to support this study, a new factor, PPH overshoot, both online and start-up, was studied, as shown in Figure 3. A single sample is not convincing unless it is repeatable. To find only one or two similar process factors is easy, however, to include all the same events is difficult. After diligently searching frame by frame, Figure 3, showing another nine days in macro, was selected since it met the criteria of covering all four factors, plus an extra new factor, PPH overshoot. This not only confirms the repeatability, but also extends the exploration and its associated description, as chronologically listed in Table 3.

Figure 3: Overview macro 2/26/2019 to 3/6/2019

Table 3 – Vibration Response with PPH Factor

In Figure 3, six events are noticed and paired into three groups – start-up, online PPH, and MI – for two purposes and is summarized in Table 4:

• To examine if all phenomena in Figure 1 are repeatable (on left column);
• To add a new factor, overshoot, (on right column) and confirm all four plots (Figures 1-4) complement each other.

Table 4 – Comparison of All Six Events

Figure 4 is a 10-hour micro (magnified) view for Figure 3’s nine-day macro to confirm four events, both online and start-up, with and without overshoot.

Figure 4: DCS micro for 4x events after zooming in

Figure 4 shows:

1. 2/27/2019 (left/upper corner):
   Online PPH (feed rate) transit (+6000) without overshoot, from 30000 to 36000, in much longer duration (~6x blocks) and proved a positive influence (lower vibration response)
2. 2/28/2019 (right/upper corner):
   Online PPH (feed rate) transit (+ 3000) with overshoot, from 35000 to 38000 at first, then dropped to 36000
3. 2/26/2019 (left/lower corner):
   Start-up PPH without overshoot repeatedly indicated a minor peak at 0.1029
4. 3/6/2019 (right/lower corner):
   Start-up PPH with overshoot and the shortest duration (~ 1x block) indicated a peak at 0.3754 (> 3.5x times higher than 2/26/2019)

As stated in Part 1, the following conclusions were made:

1. Amplified or magnified video definitely saves time and labor, especially when the job scope is bigger and beyond normal route check. Most importantly, the dynamic image can be shared and lets the machine speak for itself without explanation. It is a great communication tool for reliability applications.
2. Magnified video can help look for defects beyond human vision limit, and its value is proved by numerous documented examples. When photographers proudly present the amazing dynamic images, most stories stop after all the observers give off their surprising “Wow” admiration, whereas it is just the beginning and this study intends to develop further exploration from another viewpoint.
3. Reliability common sense tells practitioners that a machine is supposed to be affected by the manner in which it is run by different operational practices or how the vibration is affected by different process factors. However, this intuition is never proven and supported by real data that integrates both online vibration and DCS data, which is the purpose of this case study.

Author’s Acknowledgments

Nick Ma, operation engineer in Formosa Plastics PP-1 Unit for extracting all the DCS plots. Without his contribution, this case study would not be informative.
Andy Lerche of Mechanical Solutions, Inc. for taking the video that discovered the hairline soft foot and solved my curiosity.
Greg Adams and Seth Rozner, colleagues in the Rotating Department, for reviewing the author’s work before submission.