Troubleshooting and Rectification of a Giant C 3 Splitter Tower Problem Part 2: Rectification Daniel R. Summers, Sulzer Chemtech USA Inc., Tulsa, OK Brian Clancy-Jundt and Randy Miller, PetroLogistics Corp., Houston, Texas and Henry Z. Kister, Fluor, Aliso Viejo, CA Presented at the Distillation Topical Conference, AIChE Spring Meeting, New Orleans, Louisiana, March 30 - April 3, 2014 UNPUBLISHED Copyright 2014 Dan Summers, Brian Clancy-Jundt, Randy Miller, and Henry Z. Kister. The AIChE shall not be responsible for statements or opinions contained in its publications Page 1
Troubleshooting and Rectification of a Giant C 3 Splitter Tower Problem Part 2: Rectification By Daniel R. Summers, Sulzer Chemtech USA, Inc., Tulsa, OK, Brian Clancy-Jundt and Randy Miller, PetroLogistics Corp., Houston, Texas, and Henry Z. Kister, Fluor, Aliso Viejo, CA, Summary The PetroLogistics Propylene Dehydrogenation Unit (PDH) 28 feet ID 4-Pass trays C 3 Splitter started up in October 2010 and experienced low tray efficiencies and premature flooding in its first year of operation. Due to the low tray efficiency it could not produce on-spec polymer grade propylene. PetroLogistics, Fluor (who was not involved in the tower design), and Sulzer formed a taskforce to conduct a troubleshooting investigation to determine the root cause of the poor performance and to propose a fix. To investigate and solve the problem, a troubleshooting task force was formed comprising of representatives from PetroLogistics, Fluor, and Sulzer. The strategy was to conduct a field investigation aimed at identifying possible root causes to the operation difficulties, and proposing a fix. The PetroLogistics expertise in operating the C 3 Splitter was combined with other specialty resources. Fluor, who was not involved in the tower design, but has extensive experience in distillation design and troubleshooting, was requested to assist with investigating the root cause of the poor performance and with engineering the fix. Sulzer s expertise in tray design, evaluation, and modification was a crucial to the modifications and improvement. Tracerco was later brought in to provide diagnostic expertise in anticipation of extensive use of gamma scanning in identifying the root cause. Together, the task force identified the root cause and came up with an effective fix that solved the problem. Part 1 of our paper described the troubleshooting investigation. Part 2 here focuses on the rectification of the tower problem based on the troubleshooting diagnosis. PART 2 RECTIFICATION Background Based on the diagnosis established in Part 1, the key to solving the problem was to reduce the open area (from 15% to about 11% of the active area). In addition, it was determined that it would be beneficial to avoid maldistributed reflux liquid flow out of the false downcomers; to add anti-jump baffles to stop any cross-over of vapor, liquid or froth over the downcomers that could hurt tray capacity; and to add off-center downcomer clearance blocks to ensure that the liquid from the off-center downcomers was distributed more evenly. Additional modifications, including reducing outlet weir height, reducing downcomer clearances, and perforating the integral trusses were also proposed as potential improvements for the 4-Pass trays. Page 2
Right in the middle of the design modification discussions, the heat-pump main compressor experienced a high vibration incident on June 9 th 2011. On June 10 th it was discovered that the compressor had thrown a blade and that the unit would be out of service for several weeks. The team discussions turned immediately from what could we do to what will we do. Sulzer quickly determined that if it used all of its resources at its disposal worldwide that it still could not fabricate the needed 10,520 new tray panels in three weeks. On top of that there was not enough manpower and access points available to remove all the existing tray panels and install new ones in a 3 week window of opportunity. This left the team with only one available option to reduce open area and that was with blanking strips. Implementation The first thing that had to be done was to settle on a final open area and then determine how to achieve it. With the tower problem identified as interaction between vapor cross flow channeling (VCFC) with the inside-to-outside-pass maldistribution (see Part 1), it was necessary to mitigate the channeling and maldistribution. To mitigate channeling, the open area of the fixed valves needed to be reduced to 11% of the active area (1, 2) and this number was recommended by Fluor. Sulzer arrived at the same number independently. Recently, Sulzer published a paper regarding tray stability on single pass and multi-pass trays (3). The paper defines a stability factor criterion described by Eq. 1 below. With this equation, and assuming that 80% of design was the minimum loading the trays would need to see, it was determined that to maintain a 0.6 stability factor at turndown, the open area needed to be reduced to 10.9% of the active area Where, η = (ΔP DRY / H CL ) 0.5 (1) η = Stability Factor ΔP DRY = Dry Tray Pressure Drop, in Hot Liquid H CL = Hydrostatic Head of Liquid, in Hot Liquid The entire team unanimously adopted the 11% open area recommendation. This was nearly a 30% reduction in open area. The impact on the total tray pressure drop was expected to be quite significant at an 18% increase. Could the heat pump system and compressor handle it? The answer was yes because, the system was already accommodating the high pressure drop exhibited both by the channeling and flooding on the trays, which our modifications were expected to eliminate. The next question was how to reduce the open area. Sulzer entertained thoughts of replacing the center panels of each of the 4 passes, but they would be nearly devoid of any openings. Sulzer also experimented with crushing the MVG V-grids but they were fabricated from 10 gauge (0.1345 thick) carbon steel material and this was deemed impossible to perform in the field. Page 3
We thought of plugging a number of MVG valves with a chewing gum type substance but no satisfactory substance could be identified. This left blanking as the only feasible option. Sulzer brainstormed ways to install the blanking strips. Since there are no through openings on a V-grid tray (such as using the holes on a sieve tray), therefore simply holding the blanking plates with bolting assemblies was out of the question. This conclusion was reached because drilling (or punching) through the 10,000 plus tray panels with multiple holes would be impossible to achieve in the time frame given. The workshop made several T bolts, see Figure 1, that could be placed within the underside of the V-Grids, but they needed to be aligned perfectly perpendicular to the V-grid or they would fall out, see view 3 of Figure 1. Figure 1 T Bolt Idea to hold Blanking to Trays The time it would take to weld all the necessary T bolts together was found to be several weeks and this idea was abandoned. One of the shop people made a special size washer and found that by tilting this washer (with bolt in) it could be placed easily within the underside of the V-grid and it would not fall out as long as the bolt remained vertical, see Figure 2. Page 4
Figure 2 Blanking Strip Hardware The next issue was the size of the blanking plates. It was agreed that blanking strips placed on the underside of the tray with the strips aligned perpendicular to the liquid flow direction per recommended industry practice (4),was the proper style blanking to employ. Since the V-grids were all placed on the trays with a triangular pitch, it was entertained to make the blanking strips notched such that whole v-grids would become covered. The workshop indicated that providing notched blanking strips would make the supply of sufficient blanking strips impossible in the given time frame. The blanking strips would have to have 4 straight sides so that the workshop could have them fabricated in time to start installation. This would result in numerous V-grids being only partially covered. A close review by Sulzer determined that if half or more of the V-grid bottom opening was covered, then this area would be controlling and could be correctly used in the calculation of the overall open area of the tray. Figure 3 shows a sketch of a 5-1/2 wide blanking strip application in comparison to the V-Grid tray pattern. The target was to have the equivalent of 8,125 active MVG valves remaining on each tray. The original MVG valve count was 11,216 MVG valves on each odd numbered tray and 11,248 MVG valves on each even numbered tray. Sulzer s idea was endorsed by the rest of the team. To avoid inactive regions on the tray, the team placed the constraint of making the blanking strips no larger than 5.5 inches wide, covering at most three rows of MVG valves. This number is only slightly larger than the 4-inch maximum blanking strip practice recommended in the literature for sieve trays (4). The final blanking patterns are shown in Figures 4 and 5 for the even and odd trays respectively. The final blanking strip design was to utilize several different sizes especially with the shape of the Modified Arc Downcomers (MOAD). As a safety measure, blanking was employed at the inlet from the side MOADs (see the angled blanking strips on the side panel diagram in Figure 4) to ensure there was no bypassing of vapor up those downcomers. Those particular downcomers had 3 clearances.. Page 5
Figure 3 Example of Blanking Layout on the MVG V-Grid Design Figure 4 Even Tray Blanking Layout (different colors are different part numbers) Page 6
Figure 5 Odd Tray Blanking Layout (different colors are different part numbers) The blanking strips were fabricated from 14 gauge Carbon Steel (some ended up being 10 gauge because Sulzer ran out of 14 gauge material) and altogether they weighted about 45,000 lbs. There were a total of 18,240 blanking strips provided along with 54,720 bolts and 109,440 washers and 109,440 nuts. There were 11 different size blanking strips as shown by the various colors in Figures 4 and 5. Fabrication was started in stages because initially it was thought that the compressor could be fixed in approximately 2 weeks. As Murphy s Law would have it, it took closer to 26 days for the compressor to be back in service and provided the team with sufficient time to fabricate all the needed blanking strips and hardware. Tower work (including closure of all external flanges, manways, etc.) took a total of 29 days. In addition to blanking strips being provided for the tray decks, it was determined that the offcenter downcomer clearance could have blocks added on the inboard side within the window of opportunity. Hundreds of Downcomer Blocking Plates were provided, see Figure 6. As stated above, these blocking plates were needed to make sure that the clearance length (and area) under the off-center downcomers were equal on both sides. Page 7
Figure 6 One of 3 sets of blocking plates per individual off-center downcomer The tower was ready for personnel entry on June 15 th and installation of these parts commenced on June 16 th. Initial inspection at all the vessel manholes (including the very bottom of the tower) showed all the trays to be intact and undamaged. The top tray was chosen to be the place to start the blanking process. Keep in mind that there are 12 manways per tray because of the 4 tray passes and the 3 lattice trusses. It was slow going at first, trying to establish which valve row (in a dark tower) was to receive the blanking strip, how best to place the bolt/washer arrangement in the tray, and finally how the blanking strip was to be oriented because there was plenty of play with all the parts. Eventually things started moving and there was roughly a crew of two at every tray manway and at 4 different elevations removing manways and placing blanking strips. After one day it became apparent that an inspection team was needed just to keep up with all the activity in the tower. At any given time, day or night, there were more than 80 workers inside the tower along with 20 support staff and ground crew outside. To enhance worker productivity, lunches were encouraged to be brought up onto the tower and portable toilets were installed on several platforms on the tower itself. Page 8
Figure 7 Initial Bolt Placement Figure 8 Initial Blanking Strip Placement Page 9
Also on June 16 th several photos of the progress in the tower were distributed to the team members. One picture showed the top of the reflux flash box (see Figure 9) which clearly showed the reflux piping entering from the side but there was no enclosure around the pipe. This left an 18.5 x 12 gap, mostly above the pipe, through which liquid would pour out when the frothy mixture was high enough or if there was turbulence in the region. The drawings did not clearly show whether such enclosure existed. The troubleshooting investigation previously identified the upward movement of the reflux liquid and the turbulence due to the feed pipe entry into the false downcomers as the trigger initiating the multi-pass maldistribution (see Part 1). The photo gave a clear validation to this theory and provided a likely path for the initiation of the suspected reflux maldistribution. A cover plate enclosure was designed and shipped for the flash boxes and a notched weir with additional height was designed by Fluor and added to the top of the Flash Boxes to prevent liquid overflow out of the boxes and to ensure uniform liquid distribution should some overflow from the flash boxes still occur. See further discussion on this installation of these parts below. Figure 9 Reflux Distributor opening By June 20 th good progress was being made in the tower and the blanking operation was going reasonably well, see example of finished product in Figure 10. About this time it became apparent that the Compressor was going to be out service a bit longer than the original 2 weeks. This provided an opportunity to add the anti-jump baffles recommended by the troubleshooting team (see Part 1) for the center and off-center downcomers. These baffles were also originally specified in the tower design from the original design engineering firm, but Sulzer had indicated Page 10
that the weir loading was so low (less than 8 gpm/inch) and that the 18.5 center and 20.5 offcenter downcomer widths did not need them. The troubleshooting team recommended adding them as secondary priority. Originally, they were not included in the fix due to time shortage, but the delay in the compressor return to service provided an opportunity and sufficient time to fabricate these devices and get them shipped and installed during this window of opportunity, see Figures 11 and 12. Installation of the Anti-jump baffles commenced on June 24 th. The Reflux Flash Box modifications were last to be supplied because they would be the last item to install before closing up the tower at the top. Figure 10 Overall Blanking Strip Progress Page 11
Figure 11 Anti-Jump Baffle Supports using Existing Hardware Figure 12 Anti-Jump Baffles Installed Page 12
Blanking Strips are supposed to cover these V-Grids completely Figure 13 Inspection Results Poorly installed Blanking Plates As seen in Figure 13, by the time June 26 th came there was good progress on the blanking strips and final inspections were being performed. There were a number of misaligned blanking strips that, because of the play in the bolting assemblies, specific rows of MVG V-Grids that were supposed to be covered completely were not. These were all identified by the inspection team and corrected. The Reflux piping enclosure and the notched weirs on the reflux flash box were one of the last items to be installed. Figures 14 and 15 show their successful placement in the tower. Page 13
Figure 14 Reflux Flash Box notched Weir Additions Figure 15 Reflux Flash Box Enclosure around the Feed Piping The tower was closed up just before July 3 rd and the unit was started up just after the 4 th of July Holiday. Page 14
Post Modifications Tower Operation Approximate tray efficiency after revamp (based on a proprietary equation of state) has been observed to be between 80 and 90 % with downcomer velocities as high as 0.3 ft/sec, weir loadings of roughly 10 gpm/inch and C-factors (based on bubbling area) above 0.28 ft/sec. Conclusion The modified tower achieved tray efficiencies comparable to those obtained in well-operated, smaller-diameter, low pressure C3 Splitters. This is incredible considering that the flow path lengths were somewhat compromised by the blanking strips, that the blanking strips were somewhat wider than ideal, and that the time was too short to implement some of the additional ideas on our wish list. Flooding and instability were fully eliminated. The tower successfully produces on-spec polymer grade propylene. Our experience in Parts 1 and Part 2 shows that correct diagnostics, good engineering, and a multi-discipline team working together can solve even the most challenging problems with total success. References 1. Kister, H. Z., P. M. Mathias, D. E. Steinmeyer, W. R. Penney, B. B. Crocker, and J. R. Fair, Equipment for Distillation, Gas Absorption, Phase Dispersion and Phase Separation, in D. W. Green and R. H. Perry Perry s Chemical Engineers Handbook 8 th Ed., Section 14, McGraw-Hill, New York, 2008. 2. Kister, H. Z., K. F. Larson and P. E. Madsen Vapor Cross Flow Channeling on Sieve Trays: Fact or Myth? Chem. Eng. Prog., p.86, November 1992. 3. Summers, D. R., Tray Stability at Low Vapor Load, Conference Proceedings of Distillation and Absorption 2010, p.611, Eindhoven, The Netherlands, September 12-15, 2010. 4. Kister, H. Z., Distillation Operation, McGraw-Hill, NY, 1990. Page 15