Perspectives from our Founder

Article 1: History of Process Flowsheet Simulation for Industrial Applications 



By Marshall Rafal

Founder of OLI Systems, Inc.

The 1940s

The seed that allowed the eventual simulation of complex chemical processes was planted during WWII at which time computing machines were sought to break complex codes used by enemies of the Allies.  In 1948 these efforts culminated in the world’s first digital computer, ENIAC, a vacuum tube-based machine that consumed a great deal of space in a laboratory at the University of Pennsylvania.

The 1950s and 1960s

During the following decade more and more powerful vacuum tube-based computers were built for commercial use largely for business purposes.  At one time, during my childhood, computers were not called such but, rather, UNIVACs, after the company that introduced commercial computers along with International Business Machines (IBM), which had been in the business of producing machines to facilitate managing business data.


All of this brings us to 1958, just a decade following the achievement ENIAC.  At this point scientists and engineers began to

see the possibility of utilizing digital computers for scientific applications involving simulation.  It is worth pointing out that the smart phone that we utilize today is many-fold more powerful in compute speed and storage than the IBM 7090, which defined the state of the art in digital computers in 1958.  The IBM 7090 filled a space the size of the floor of a warehouse, specially designed to accommodate this behemoth.  It took several computer operators to develop and execute applications.  IBM would provide, onsite, 24/7 specialists to deal with issues when the computer would “crash” and had to be brought back up … kind of a 60+ year old version of hitting the start button on a smart phone.


Although seen only by visionaries, software was destined to be the leading edge of this revolution.  For scientists, the power offered by these primitive machines could apply not just to storing and managing data but also to simulating real-life processes.  Within a very few years an intense space race was being fought not just on the launch pad but on the computer with evermore accurate simulations of space flight, to the point where by the end of the 60s we could thread a needle so fine that we could put men on the moon and bring them back.


In the late 50s, engineers concerned about the enormous cost of chemical processing in refineries and plants began to write programs, first in machine/assembler language but soon in an upstart, primitive version of FORTRAN, to predict the behavior of complex chemical systems, largely based on reactions and physical equilibrium, so that modules of actual plants could be predicted and plant performance improvements could be more safely implemented. 


In the latter part of the 50s, pioneers, one of whom I have been privileged to know, Dr. Mike Kessler, now in his mid-90s, envisioned a supervisory, infrastructure program that could put together full plant simulations.  The challenges were formidable.  In these earliest days, an engineer had at best 16K bytes of memory with which to work.  Techniques involving sharing core via overlays became ubiquitous, techniques that have all but faded from memory with today’s advent of virtually unlimited memory and compute speeds that are multimillions times faster than the “mainframe,” 7090-style machines of those days.

Yet, somehow practical problems got solved.  When I completed graduate studies in 1966 and went to work at Esso Math and Systems in NJ, Esso already had a fairly mature process flowsheet simulator by the name of COPE for simulation of refineries, focused on physical equilibrium (e.g., VLE) and chemical reactions (e.g., catalytic cracking).  Since Esso had a number of ammonia plants in its portfolio, I was assigned the job of enabling COPE to simulate ammonia plants.  This involved figuring out how to predict the VLE in ammonia-based systems and then implementing “unit operations” for things like reformers and shift converters.  This was really a dream assignment because it enabled me to develop and solve, via numerical methods like Newton-Raphson, sets of complex equations that would predict equilibrium and reactive chemistry.  I got to work with pioneers who had seen power in Newton-Raphson that would one day seed the ability to solve entire flowsheets based upon all of the equations depicting such flowsheets.

On into the 1960s as the US worked on simulating the equations of motion as they pertained to space travel, chemical engineers and computer specialists worked within very large companies in the chemical process industries in order to create simulators to predict the flowsheets of interest to these companies.  Esso had COPE, Mobil had QUIKBAL, Dupont had CPES, and many others (e.g., BASF, ICI, etc.) had their own as well.

Another product of the 60s was that these in-house simulators were enriched technologically in both chemical engineering technology (thermo and reactions) as well as algorithmically, the latter enabling the simulation of monolithic crude tower arrays.

The 1970s

The primary barrier to anyone launching a company to develop and offer flowsheet simulation was the cost of a computer, which entailed multiple millions of dollars.  As the decade of the 60s ended, however, operating systems were developed which would allow many users to use a mainframe computer simultaneously, which led to an industry called timesharing.  Timesharing enabled access to the most powerful mainframes of the day for affordable pricing and bridged the gap until the advent of minicomputers which would enable companies to acquire a powerful scientific machine for around $100K and sounded the death knell for timesharing … an industry that lasted for little more than a decade.

In the very late 1960s three entities of which I am aware, decided to develop their own process flowsheet simulators and thereby enable mid-sized companies in the CPI to access the power of flowsheet simulation.  I personally met (in the early 1980s) and developed a business relationship with the founder of Simulation Sciences (Simsci), Dr. Y.L. Wang, truly a pioneer and visionary in the field.  Simsci’s simulator named Process was introduced in 1969.  At about the same time, two other entities, a company called Chemshare offered a simulator called Design and another, an outgrowth of the University of Houston offered a simulator called CHESS.

These upstart commercial simulators developed their niche and sought ever-more sophisticated algorithms and broader based thermo.  OLI Systems began, in 1973, to address the gaping hole relative to electrolyte systems.  Within the commercial simulators a very few electrolyte systems were handled and these strictly via interpolation (e.g., sour water) without addressing the underlying thermodynamics.

The 1980s

I never got to ask Dr. Wang if he ever envisioned that the monolithic companies who invested in their own simulators would transition over to independents like Simsci, but that is what happened over the following two decades.  By the turn of the century I doubt that any of the majors in the CPI still ran their own simulators and if they did, I seriously doubt that it was their primary such tool.  The issue was cost of development, especially in the area of UI, and maintenance.  One company like Simsci, serving hundreds of clients, could invest far more in the resources (technology, software, and support) than could a single company.  The turning point was two seminal developments.  The first, offered by Aspen Technology, Aspen Plus, was developed with DOE funding in order to address chemical process simulation as opposed to oil and gas simulations.  The advent of Aspen Plus brought the largest chemical companies to consider a serious alternative to their in-house simulators.  Aspen Technology also developed an interpolative mixed solvent electrolyte model usable by technical experts within the majors who could fit systems of interest with data regression.  In the meantime, OLI continued to seek its “holy grail” of a truly predictive simulation model.

The most important change in the industry in the 80s was centered in UI development supported by the advent of the PC. A small embryonic group at University of Calgary, formed a company by the name of Hyprotech and created the first process flowsheet simulator for the PC, an early 16-bit machine.  In addition, Hyprotech turned to the emerging C++/Visual C++ to create a leap forward in UI, albeit with a simulator far less appointed technologically but focused solely on oil and gas and the idea that operators could use a simulator not just R&D folks.  It was the price point of the PC that enabled taking process flowsheet simulation out of the R&D center and into the plant.  By the end of the decade, powerful UIs were being undertaken by Simsci  and Aspen Technology because the PCs had grown into a 32-bit architecture.


The 80s and the 90s were a time of substantial leaps forward not just relative to UIs but relative to chemical engineering technology and algorithms.  Flowsheet optimization, costing, heat exchanger rating, equations-based simulators, and so much more emerged.  The 90s was a decade of profound change for OLI Systems as it forged the basis for a broadly, predictive model for simulating electrolytic chemical processes, an area beyond the reach of conventional process simulation and, heretofore, OLI itself.  Thermodynamics teaches us that all of the principal properties are a function of a standard state term (e.g., equilibrium “constants”) and an excess term (e.g., activity coefficients). 


In 1991, based upon the groundbreaking work of Hal Helgeson and co-workers in the 80s, OLI implemented a fully predictive standard state property facility.  It took the combination of OLI hiring Dr. Andre Anderko and his eventual co-workers along with the enabling funding of the DOE to launch OLI to produce first a comprehensive mixed solvent electrolyte model, and then over the next 20 years invest more than 100 man years to populate a supporting database of enormous scope.

In 1990, with funding from an industrial consortium, OLI accomplished ESP, the Electrolyte Simulation Program (originally Environmental Simulation Program), a powerful chemical and electrolyte simulator with a prior generation UI.

2000s and 2010s

Despite the advent of OLI flowsheet simulation, our core competence, predictive simulation of the very most complex chemical systems has resulted in the OLI Engine, which resides in all of the largest process flowsheet simulators of today including KBC’s Petro-SIM, Aspen Technology’s Aspen Plus and HYSYS, Honeywell’s Unisim, and Aveva’s Sim Central (tracing its origins back to Simsci) providing all of them the unique state of the art capability within the area of process flowsheet simulation.  In a manner analogous to “Intel Inside,” the catchphrase for the chips that drive the modern PCs, “OLI Inside” drives electrolyte simulation for process flowsheet simulation within the CPI.  The one difference in the analogy is that OLI has its own process flowsheet simulation product as well.

By 2015 OLI had undertaken the development of a modern UI for its ESP simulator resulting in Flowsheet Analyzer: ESP, a niche product that could be used heavily and even exclusively by companies focused wholly on electrolyte-based (in particular water-based) systems.  The majors like KBC, Honeywell, Aspen Technology, and Aveva were offering ever more sophisticated facilities technologically and algorithmically and for customers requiring those features not necessarily offered by OLI, there was “OLI inside.”

More information

Contact OLI at https:/ for more information or to schedule a meeting with an OLI expert to discuss how you can use OLI technologies to address your process modeling challenges.