Contenu connexe Similaire à CoOpReport_Fall15_JDD Similaire à CoOpReport_Fall15_JDD (20) 2. ii
EXECUTIVE SUMMARY
The Wiped Film Evaporator (WFE), in the plant I was assigned to sets the
plant production rate rather than the reactor. Because production needs have not
necessitated the de-bottlenecking of the WFE, the reactor will continue to outrun
the WFE. There is a resulting opportunity for the plant to lengthen the reaction time
by charging less catalyst without impacting the overall production rate of the plant,
resulting in catalyst cost savings.
Over the last few years, the price of the main reaction’s catalyst has increased
43% relative to 2014. Although the price was fairly recently negotiated down 31%
relative to 2014, there is still an opportunity to optimize the catalyst loading and
further reduce costs.
It seems that several years ago, 150% (weight % relative to the standard
charge today) of catalyst was charged and then at some point, operations switched
to the current 100% charges. At another time, it is believed that effectively, 50%
charges were being run. It is said that while running this “50%” batches, a
contaminant byproduct formed in the recycle tank and substantially reduced the
catalyst’s activity and led to out-of-spec product in the following batches. Because a
50% charge seemed to be historically too low, it is clear that the ideal catalyst
loading is between 50%-100% of the current loading. It was also clear that the
project would involve decreasing the loading slowly and monitoring the
contaminant levels as well as the product conversion (i.e. staying within spec).
Experimental batches were charged with 10% reductions of catalyst and the
formation rates of the product were modeled. The resulting model was used to
7. 1
REPORT
BACKGROUND
The Wiped Film Evaporator (WFE), in the plant I was assigned to sets the
plant production rate rather than the reactor. Because production needs have not
necessitated the de-bottlenecking of the WFE, the reactor will continue to outrun
the WFE. This results in excess hold time, when the reactor is full of crude product
and ready to transfer to the crude tank, and excess idle time, when the reactor is
empty and waiting to be charged for the next batch. Until the WFE is de-
bottlenecked, this excess time is an opportunity for the plant to lengthen the
reaction time by charging less catalyst, resulting in catalyst cost savings.
Over the last few years, the price of the main reaction’s catalyst has increased
43% relative to 2014. Although the price was fairly recently negotiated down 31%
relative to 2014, there is still an opportunity to optimize the Catalyst loading and
further reduce costs.
It seems that several years ago, 150% (weight % relative to the standard
charge today) of catalyst was charged and then at some point, operations switched
to the current 100% charges. At another time, it is believed the batch size was
doubled in the reactor and the catalyst charged remained the same, effectively
halving the catalyst/reactant ratio. It is said that while running this “50%” batches, a
contaminant byproduct formed in the recycle tank and substantially reduced the
catalyst’s activity and led to out-of-spec product in the following batches. No
8. 2
documentation of this was available. However, this previous event served as a sort
of ‘ceiling’ to how far the catalyst could be reduced as the project progressed.
For several years, operators have manually poured (2) 50% drums of
catalyst per batch into for a total charge of 100%. Because a 50% charge seemed to
be historically too low, it was clear that smaller bags of catalyst would have to be
ordered from the manufacturer and that long-term, the ideal catalyst loading would
be between 50%-100%. It was also clear that the project would involve decreasing
the loading slowly and monitoring the contaminant levels as well as the product
conversion (i.e. staying within spec).
If successful, each 10% reduction of catalyst/batch would result in roughly
10% catalyst purchasing savings. The project would model and better quantify the
relationship between catalyst loading (lb) and the reaction rate (hr) as well as
create a better understanding of the impact less catalyst and the resulting longer
batches would have on contaminants that poison later reactions. The ultimate goal
was to provide Ops with a tool that recommended the ideal catalyst charge given the
planned production rate.
OBJECTIVES
The objective of this project was to reduce the catalyst utilization for each
batch to safely maximize cost reduction without impacting the overall production
rate or quality of the product. In order to do this, the relationship of the amount of
catalyst charged (lb/batch) and the required reaction time in the front-end reactor
(hr) would need to be quantified and modeled. Ultimately, the deliverable to Ops
would be a chart that recommends how much catalyst (lb/batch) to charge given
9. 3
the current production rate, whether that be based on a monthly quota
(Mkg/month) or on the current production rate of the plant (lb/hr).
The reality of the plant operation was that the reactor outruns the
downstream process. Another reality was that the catalyst that officiates and
accelerates the reaction was relatively expensive. Cost reduction was to be achieved
by decreasing the catalyst utilization of the plant to slow down the front-end
reaction, thereby decreasing excess holding/idle time in the plant. The savings from
the project come from those cost savings directly resulting from using less catalyst
per batch as well as lowered maintenance costs that could result from less catalyst-
related pluggage problems in the back-end process.
By slowing down the reactor, the front-end process rate would approach that
of the back-end. The objective was to still keep the front-end process a bit faster
than the back-end to avoid slowing down the overall production rate unnecessarily.
As a result, the overall production rate would be unaffected by the project.
In order to avoid impacting the quality of the product, it was clear that a
means of monitoring and preventing contamination or loss of quality was necessary
for the project to proceed. As a result, it became an objective to identify
contaminants and watch-outs that would have to be monitored during the project.
A clear objective was that the project had to be carried out safely. Safety,
after all, is a top priority for Albemarle as well as for other corporations and
industries. It was a high priority that the safety of everyone involved in the project,
specifically operators, would be taken into consideration throughout the process.
10. 4
ACTIVITIES
The first step of the project was to investigate and become familiarized with
the plant in question. There are several plants in Albemarle’s South Plant in
Magnolia, AR and this plant in particular was the sole focus of this project. More
specifically, the reactor and catalyst charged to that reactor were the key
components of the plant that was of interest to this project. The procedures and past
articles written by startup engineers of the plant, especially those occurring in the
reactor, were studied. On-site chemists were consulted about the nature of the
catalyst compound used and plant Ops were consulted about the reactor and
process.
Before any experimentation could be conducted, it was necessary to find out
how much excess hold/idle time in the reactor was available to cut into as the
reaction time was to be increased. Programmable Logic Controller (PLC) data was
analyzed to determine how long the batch was in “hold” mode, when the reaction is
complete and the crude product is ready to transfer to the crude tank, and how long
the batch was in “idle” mode, when the crude product has been transferred to the
crude tank and the reactor is waiting for the next batch to begin. It was determined
that the idle time would be left out of the calculations for the ‘available’ time to cut
into because idle time is not necessarily caused by a slower back-end process: lower
demand and lower quotas for product could result in the idle time. The hold time,
however, occurs because the back-end process was slower than the front-end.
With this in mind, a year’s worth of batches were statistically analyzed and
the average hold time per batch was approximately 4 hours. This led to the
11. 5
conclusion that as long as the reaction time was not lengthened more than 4 hours,
production rates should not be significantly affected.
Before any experimentation or catalyst reductions could be done, a
presentation was given to Operations managers and engineers detailing the
feasibility of the project, the current cost of the catalyst raw material, and potential
savings. There was much concern associated with the production risk of a “dead
batch” occurring due to lack of catalyst, as the price of disposing of a bad batch far
outweighed the potential savings of catalyst reduction. Another concern was that of
the operators’ safety risk of charging more catalyst when the reaction could
suddenly begin and heat/pressurize the reactor. Ultimately, the experiment would
have to be designed with careful, small, reductions in catalyst with limited operator
contact and close monitoring of contaminants and reaction activity.
The next step was to design the experiment that would quantify the
relationship of catalyst charge (lb/batch) with the reaction time necessary to
convert the product to specification (hr). Because the resources to conduct bench-
top experiments in a small-scale lab setup were not available, trial runs had to take
place on the full-scale level in the plant itself. All trials had to be conducted by the
outside operators in the plant on top of their current workload.
Upon consultation with both Ops and Technology Resources (TR) engineers,
the overall experiment was designed as follows. The experiment began by charging
100% (lb/lb) of the standard catalyst charge for (3) batches, sampling every two
hours once the reaction in each batch began until the dimer product was formed to
the sufficient percent formation according to specification.
12. 6
Operators both caught these samples and shot them into a gas
chromatograph, to determine the percent conversion of the dimer product e.g. when
the reaction is complete. This form of sampling was chosen because Ops were
already familiar with this exact method of sampling and analyzing; they already
used the same method to determine when the reaction is complete before killing the
reaction and transferring the crude product. Once the GC samples were run, the
resulting graphs were interpreted. The GC picked up the limiting reactant, the
monomer precursor to the product, the dimer product, and the trimer species
formed from the product. After analyzing the resulting data and consulting with Ops
and Technology Resources, the catalyst charge was reduced by 10% and tested the
same way as described for (1) trial batch followed by (4) consecutive batches (5
total). (2) 10% reductions were tested in this manner until the co-op experience,
and therefore project, ended. Further directions for the project were submitted to
Ops and TR, suggesting further 10% reductions.
With every reduction and step from trial batch to consecutive batches, data
describing the reaction rate of product and product quality was presented to
Operations. An estimation and confidence interval of how the reaction would
behave in the next step was always given, using reaction time estimates projected
from the fitted model of the reaction. Operations had to be convinced to advance the
project each step of the way due to the importance of the catalyst facilitating the
reaction during full-scale trial runs. It was kept in mind that the project was always
second in priority to that of the plant producing the monthly quota of product up to
specification. It was also imperative to realize that operation management had to be
14. 8
Catalyst
Loading Batch
Rxn
Time
Actual
Product %
Modeled
Product%
∑ of Squares
Difference
100% 2 00:32 11.6% 2.9% 0.76%
100% 2 03:13 28.5% 27.4% 0.01%
100% 2 04:57 66.0% 65.0% 0.01%
100% 2 07:19 91.9% 92.3% 0.00%
100% 2 09:12 97.1% 96.3% 0.01%
100% 2 10:42 97.9% 96.9% 0.01%
100% 2 13:42 97.8% 97.1% 0.00%
100% 3 00:31 11.1% 2.9% 0.68%
100% 3 03:06 28.2% 25.3% 0.09%
100% 3 06:16 87.0% 85.1% 0.04%
100% 3 08:26 96.0% 95.4% 0.00%
100% 3 11:16 95.8% 97.0% 0.02%
100% 3 15:01 96.9% 97.1% 0.00%
100% 4 00:29 8.7% 2.8% 0.34%
100% 4 03:34 29.7% 34.4% 0.22%
100% 4 07:19 89.3% 92.3% 0.09%
100% 4 09:04 96.0% 96.2% 0.00%
90% 0 00:29 10.2% 3.0% 0.52%
90% 0 02:54 13.6% 19.6% 0.36%
90% 0 04:54 54.3% 56.5% 0.05%
90% 0 06:49 84.1% 85.0% 0.01%
90% 0 08:59 94.5% 94.7% 0.00%
90% 0 14:19 97.3% 96.8% 0.00%
90% 0 17:54 97.7% 96.8% 0.01%
90% 1 00:30 9.6% 3.1% 0.42%
90% 1 02:45 25.0% 17.7% 0.52%
90% 1 04:45 57.9% 53.5% 0.19%
90% 1 06:50 86.1% 85.2% 0.01%
90% 1 08:45 94.8% 94.3% 0.00%
90% 1 09:45 96.5% 95.7% 0.01%
90% 1 11:45 96.9% 96.6% 0.00%
90% 2 00:43 11.9% 3.7% 0.67%
90% 2 08:28 93.3% 93.6% 0.00%
90% 2 10:28 96.4% 96.2% 0.00%
90% 2 12:33 96.5% 96.7% 0.00%
90% 3 00:27 12.6% 3.0% 0.92%
90% 3 02:47 17.7% 18.2% 0.00%
90% 3 05:07 66.9% 61.0% 0.35%
90% 3 06:52 87.2% 85.5% 0.03%
90% 3 10:12 96.8% 96.1% 0.01%
90% 4 00:31 13.4% 3.1% 1.06%
90% 4 02:31 16.2% 15.1% 0.01%
90% 4 04:46 45.0% 53.9% 0.79%
90% 4 08:16 90.6% 93.1% 0.06%
90% 4 10:16 96.1% 96.1% 0.00%
80% 0 00:32 11.4% 2.5% 0.80%
80% 0 03:12 20.2% 20.2% 0.00%
80% 0 05:12 54.0% 57.9% 0.16%
80% 0 09:17 94.8% 94.6% 0.00%
80% 0 10:22 96.2% 95.7% 0.00%
80% 1 00:41 13.2% 2.8% 1.10%
80% 1 03:01 18.4% 17.7% 0.01%
80% 1 05:01 54.7% 54.0% 0.00%
80% 1 07:01 84.6% 84.7% 0.00%
80% 1 08:46 92.7% 93.6% 0.01%
80% 1 10:31 96.9% 95.8% 0.01%
80% 2 00:27 10.9% 2.3% 0.75%
80% 2 02:47 10.7% 14.9% 0.18%
80% 2 05:22 62.3% 61.0% 0.02%
80% 2 07:42 85.6% 89.6% 0.16%
80% 2 09:27 94.7% 94.8% 0.00%
80% 2 11:07 97.1% 96.0% 0.01%
80% 3 00:25 5.8% 2.2% 0.13%
80% 3 02:55 11.1% 16.4% 0.28%
80% 3 04:55 48.9% 51.9% 0.09%
80% 3 07:15 81.3% 86.6% 0.28%
80% 3 09:20 90.7% 94.6% 0.16%
80% 3 11:05 94.8% 96.0% 0.01%
80% 3 12:00 96.2% 96.2% 0.00%
80% 4 02:53 19.8% 16.1% 0.14%
80% 4 04:53 60.5% 51.3% 0.85%
80% 4 06:53 87.2% 83.5% 0.13%
80% 4 08:58 95.5% 94.0% 0.02%
TABLE 1: Product Formation Raw Data
A logistic growth model, based on the equation in Figure 1 and
corresponding values in Table 2 was developed to describe the behavior of the
product formation. The corresponding modeled values associated with the real
values of product conversion are shown and a regression was completed to optimize
the fit of the model by manipulating the variables. A solver was used to find the
15. 9
necessary reaction time, t (days), for enough dimer product to form per
specification.
Product Formation % =
𝑎
1 + 𝑏𝑒−𝑐(𝑡−𝑑)
, 𝑡 = time(days)
FIGURE 1: Logistic Growth Equation Modeling the Formation of the Dimer Product
Catalyst Loading a b c d Necessary Reaction Time (hh:mm)
100% 97% 11% 22.7 06:30 08:12
90% 97% 11% 20.5 07:03 09:09
80% 96% 11% 20.9 07:17 09:36
TABLE 2: Product Formation Equation Coefficients and Necessary Reaction Times
The contaminant levels in the reactant 2 recycle stream were monitored via
daily sampling, per operation procedure. The levels remained lower than the
specification, 0.30%. The levels are graphed in Figure 2 during the time of
experimentation.
FIGURE 2: Contaminant Levels in Reactant 2 Recycle Tank During Trial Batches
0.15
0.20
0.25
0.30
0 20 40 60 80 100
Contaminant Concentration in Reactant 2
Recycle Tank
Time (Day)
Contaminant in Reactant 2 Recycle Stream
100% Catalyst Charge
90% Catalyst Charge
80% Catalyst Charge
18. 12
Recommended Cat Charge Based on WFE Rates and Split (with Estimated Necessary Conversion Time)
INPUTS
# lb/hr % lb/lb
Wkly Avg (PLC)
CALCULATIONS
% lb/lb # lb # Hr/Rxn = Constant (lb.h)/charge(lb)
# Constant
# lb Reac 1 # 100% # Hr
# lb Cat # 90% # Hr
# lb Reac 1 Charge 1 # 80% # Hr
# lb Reac 2 Charge 2 # 70% # Hr
# lb Solvent 60% #
# lb Product 2 50% #
OUTPUTS
# lb/Batch # hr
Input
Output
00/00/00 0:00 AM Reaction Start Time
WFE Overhead Flow WFE Split
Wkly Avg from PLC
(lb/hr)
hr
Recommended Cat Charge Estimated Reaction Time
Reac 2 Flashed off Crude Product 1 per Batch Max Reaction Time
Hr Batch Time
SF
Hr (Non-Reaction)
Buffer Hour
00/##/00 #:## AM
Estimated Reaction
Completion Time
FIGURE 5: Recommended Cat Charge Based on WFE Rates and Split (with Estimated
Necessary Conversion Time)
CONCLUSIONS
Overall, the catalyst charge was reduced to 80% and was found to not impact
production, product quality, or operator safety. This resulted in saving 18.5% of the
material cost of catalyst annually going forward, with promise of further savings in
the future.
A chart and an interactive spreadsheet recommending the optimum catalyst
charge based on production needs were delivered to operations. Operations
continues to run with 80% catalyst charges and further catalyst reduction of less
than 50% reduction of catalyst was recommended upon departure (e.g. 70% and
60%). Each 10% reduction in catalyst results in roughly 10% savings based off of
the cost of ordering the catalyst itself.
A crucial lesson was learned when working with Operations, both
management and operators. In order to overcome the resistance to change and
