The Effect of Learned Blindness or Mass Transfer Basics

Mass transfer in action — the unseen principle behind distillation, optimization, and big savings in chemical production

Introduction

The customer’s refrain was familiar: “How can you hope your work will give any result, considering that we have elaborated hundreds of optimization proposals for the unit?” We heard this many times as we stepped into the R&D project. Their skepticism set the tone: how could outsiders hope to deliver results in six months, when plant specialists had been working with this process for decades? Within half a year, however, we identified a change that required only a $100k investment but unlocked $2M in yearly savings.

Problem Context

Simplified process scheme of butadiene purification: extractive distillation, solvent regeneration and rectification

Butadiene purification is carried out by extractive distillation, which is then followed by a conventional distillation column. The C4 cut, including the recycle stream from the polymerization unit, enters the extractive distillation column. The overhead from this column consists mainly of butenes and butanes. The bottoms stream contains solvent rich in butadiene, 2-butene, and acetylenic hydrocarbons, which is then routed to solvent regeneration. After solvent removal, the resulting butadiene-rich stream is directed to a rectification column. Here, high-purity butadiene is obtained as overhead, vinylacetylene together with some butenes is withdrawn as a controlled side draw and sent to flare, and heavier acetylenic hydrocarbons are removed in the bottoms.

Our objective in this project was to identify optimization scenarios capable of reducing operating expenses or minimizing process losses in the existing unit. To start, we systematically gathered and reviewed all available information: operating regime lists, equipment passports, process flow schemes, technical regulations, and records of past optimization efforts. Four of our engineers were dispatched to the site to conduct an on‑site audit, speak directly with operators, and photograph thousands of pages from the paper archive.

Key Insight

We reproduced the unit in a simulation model and carefully verified it against historical operating data. At first glance, the unit appeared to be well heat‑integrated, and it seemed that all feasible optimizations had already been implemented, leaving little room for further improvement.

By the time two‑thirds of the project timeline had passed, we still had no promising optimization scenario in hand. We had considered complex options such as switching the extractant to N‑methylpyrrolidone or introducing hydrogenation of acetylenic hydrocarbons. However, these scenarios were already familiar to the customer, and their major drawbacks were high capital costs and long renovation periods. What we wanted was to deliver something unique and overlooked — but at that stage, we had yet to find it.

After repeatedly running through the unit's simulation model, our team turned our attention to the distillation column that separated butadiene from acetylenic hydrocarbons. We carefully monitored the yields and product compositions from this column, reviewing them again and again.

A small but noticeable discrepancy was identified between the side draw composition predicted by the model and the actual composition observed in the unit. We set out to determine the source of this difference.

Let's dive into the mass transfer basics here.

Distillation tray mass transfer illustrated in three views - tray scheme, Raoult's law interpretation and stage-to-stage enrichment diagram

Distillation would not work if the concentrations of components remained unchanged from tray to tray within the column. On each tray, the liquid contains a certain fraction of the more volatile component. From this liquid, vapor is released in which the content of the volatile component is somewhat higher. Rising upward, this vapor passes through the openings of the next tray and comes into contact with the liquid there, which has a slightly lower temperature. Part of the vapor condenses, enriching the liquid on that tray with the volatile component. From this enriched liquid new vapor is generated, whose composition in turn contains even more of the volatile component. Repeating this step by step from tray to tray, the concentration of the volatile component in the vapor phase gradually increases toward the top of the column. This principle is well described by Raoult’s law, which relates the partial pressure of each component in the vapor to its mole fraction in the liquid and the component’s vapor pressure, thereby governing how the more volatile component becomes enriched in the vapor phase.

The accompanying figure illustrates this principle in three complementary views: (1) a schematic of a distillation tray showing vapor–liquid contact, (2) a simple Raoult’s law interpretation linking a liquid composition xₙ to the corresponding vapor composition yₙ, and (3) a stage-to-stage diagram showing how successive contacts lead to enrichment from tray to tray. In this notation, x′ₙ and y′ₙ denote the equilibrium (ideal) liquid and vapor compositions, while xₙ and yₙ represent the actual compositions on a real tray.

In our project, the column under consideration separates a mixture of butadiene (BD), 2-butene, vinylacetylene (VA), and heavier acetylenic hydrocarbons. The bottom section becomes enriched with the heavier acetylenic compounds, VA concentrates in the middle zone, while butadiene moves toward the top. At the mid‑section, the design provides for a side draw of vinylacetylene (VA), with the draw point selected according to the column’s concentration profile. The VA‑rich stream is withdrawn under strict limits to ensure concentrations remain below the explosive margin; immediately after withdrawal it is diluted with fuel gas and directed to the flare. In practice, a portion of butadiene is also carried with the VA into the side draw and ultimately ends up being burnt in the flare.

Our model confirmed the appropriateness of the selected side draw point, but the predicted concentrations of key components (VA and BD) differed from actual measurements. It would have been easy to dismiss this as within the model’s tolerance, but in our case, we were determined to investigate even the smallest discrepancies. 

Seeking the origin of this discrepancy, we discovered that the practical implementation of the side draw differed from how it was represented in our model. In the unit, the side draw was taken through a nozzle located in the vapor space above the tray, while our simulation had assumed a liquid draw from the tray’s downcomer. This distinction explained the difference: the butadiene concentration in the vapor above the tray was slightly higher than in the liquid. Though the variation was barely one percent, given the unit’s large capacity and the high market value of butadiene, the annual economic benefit amounted to nearly USD 2 million.

Conclusion

It goes without saying that the proposed solution prevailed over the alternatives, offering the shortest payback period and the highest internal rate of return (IRR). While our report reviewed other, more conventional options, presenting such an easy‑to‑implement solution that had previously been overlooked came as a real surprise. We fully recognized and respected the many years of effort the customer’s team had invested in optimizing the plant. Yet sometimes it takes a fresh perspective - someone willing to start from scratch - to reveal insights that may be missed due to the well‑known effect of learned blindness.

Afterword

The customer arranged a field test, sampling both the vapor and liquid phases from the same tray. Beforehand his leading expert told us half-jokingly that he would “eat his hat” if our solution worked. After GC analysis confirmed the difference, we never heard from him again. The solution was successfully implemented after all.