Case 97-03

Massachusetts Institute of Technology

Synthesis and Optimization of Chemical Processes

A Case Study of the R&D Value Mapping Project

Institute for Policy Research and Development

School of Public Policy

Georgia Institute of Technology

Atlanta, GA 30332-0345

Unedited Draft

This case was written by Juan D. Rogers. Comments or questions should be directed to Juan D. Rogers at (404) 894-6697 or, by email, juan.rogers@pubpolicy.gatech.edu. The research was sponsored by the Department of Energy, Office of Basic Energy Sciences, Contract ER45562. The views presented here are the case author’s and do not necessarily represent those of the Department of Energy or Georgia Institute of Technology.

Massachusetts Institute of Technology

Synthesis and Optimization of Chemical Processes

Project Summary

The Synthesis and Optimization of Chemical Processes case analyses the stream of research led first by Larry Evans, from around 1976 to 1991, and then by Paul Barton, from 1992 to the present. The main concerns of this research are the flows of energy in industrial chemical plants. It seeks to develop systematic methods to synthesize industrial chemical processes and provide optimal solutions for the required overall energy flows. The visible products of this research are specifications and software packages that implement the methods and algorithms to reach optimal solutions.

The most interesting features of this case are, on the one hand, the great impact it has had in the chemical industry by providing methods and software implementations that are widely used, allowing it to save hundreds of millions of dollars in production efficiencies. On the other hand, the career paths of graduate students who were part of the sponsored research have very peculiar patterns. This is due mainly to the significant role of a spin-off company, Aspen Technologies Inc. that hired and re-hired a number of them at various stages of their professional career. A virtual "traffic in students" helped both the company keep abreast of developments in basic research and the research team obtain direct experience of the problems that the industry faces. The role of the BES program manager was instrumental to the importance of the industrial impact of this research program.

Project Description

The research activities described in this case have some important differences with other basic research activities because, a) it is engineering research rather than a "classical" scientific field (such as physics or chemistry, for example), b) the content of the research seems to bear little resemblance with the range of phenomena that are involved in chemical engineering, c) the close relationship with industry is integral to the conduction of basic research, and d) the career paths of graduate students who participated in the research follow very peculiar patterns.

          The sponsorship of DOE went through several stages. The first, from 1976 to 1985, was not awarded through the Office of Basic Energy Sciences (BES). Rather, it was part of a synthetic fuels effort that grew out of the energy crisis of the 1970s. The second stage, from 1985 to the present, was channeled through BES and it coexisted with a very close relationship with industry.

1. Technical Background

The name given to this field of work by the practitioners is "Computer-Aided Chemical Engineering." It involves the use of powerful software tools for the development, design, operation and control of chemically-based manufacturing processes. Even though the substantive applications of this work fall inside the field of chemical engineering, the research activities related to this case are truly interdisciplinary. The main contributions are not in the chemical processes themselves. The research topics involve mathematical modeling techniques and specifications for chemical processes of interest to industry. Chemical engineering provides the background know-how about the chemical processes, including physical properties, thermodynamics, and the chemistry and physics of the processes. Numerical Mathematics provides the algorithms for solving the equations and other mathematical problems that result from the implementation of the processes at an industrial scale. Computer Software Engineering enables the implementation of computer software that is robust, reliable, and effective.

The development of techniques for computer-aided chemical engineering fall into three areas: simulation, optimization, and synthesis. The goal of simulation is to predict the performance of a proposed or existing plant using mathematical models to carry out the heat and material balances. Steady state-simulation and plant wide dynamic simulation are already available commercially. However, the simulation of batch and semi-continuous processes is still in its research stages.

Optimization is at a higher level than simulation. In any proposed flowsheet there are many degrees of freedom in the selection of design and operation parameters and many treadeoffs in the proper choice of values for these parameters. Optimization seeks to balance these competing forces to maximize an objective function such as profit.

Before optimization can be performed, it is necessary to know the structure of the flowsheet. The conception of the structure of the manufacturing process is a task in process synthesis, which is at the highest level.

A consequence of the nature of this work is that it results in generic or infrastructural technologies that can be applied in other fields completely unrelated to the one that guides or motivates the original research. The sets of equations that model a chemical process may be found, with some variations, in other systems and the mathematical techniques and optimizing algorithms to solve them are portable. Similarly, computer software implementation requires solutions that are not inherently dependent on the particular chemical problem it is originally designed for and can also be transferred to other situations. The research program studied in this case has not fully exploited this, but the awareness of this possibility suggests that many links will probably develop in the near future. According to the PIs, the reason for this is difficulty of communication across disciplinary lines. Problems in the design of VLSI integrated circuits, for instance, are very similar to the ones encountered in designing large chemical process plants. However, the techniques and algorithms remain quite different and most innovations in modeling and simulation in this area have originated in Chemical Engineering rather than Electrical Engineering.

2. Technical Focus and Historical Development

The origins of this stream of research can be traced back to the second half of the decade of the 1970s when DOE sponsored a significant amount of work in alternative fuels due to the energy crisis. The first grant was awarded to Larry Evans by another office of DOE, not BES, that was interested in coal liquefaction and synthetic fuels. The funding support at this stage was 5 million dollars for 5 years, between 1976 and 1981, plus 1.5 million provided by industry. The goal of the project was to develop software for synthetic fuels processes and optimization of energy flows in chemical plants. Both subjects were directly related to priorities set by the energy crisis. They developed a simulator for large scale, steady state, commodity plants. The program would calculate the solution to the energy balance equations for the entire flow sheet of a plant.

By the end of the 5 year period, the emphasis on energy crisis priorities began to wane and DOE discontinued support for this line of work. Evans and his team had developed the simulator and wanted to continue developing their ideas and programs, including their relationship with the industries interested in the results of their work. Given that it was not exclusively research, there were perceived conflicts of interest in housing the continuation of this work entirely within the institutional framework of MIT. Evans, then, decided to start a business to pursue the commercialization of the simulator and its future versions. As a result, Aspen Technologies Inc. was founded in 1981 while Evans was still a professor at MIT. The company developed the simulator for commercialization and also developed plant control versions of the software that could not only simulate the energy balance conditions but also introduce integrated control loops to control the plant once it was implemented.

The fundamentals of the Aspen technology were used by other commercial enterprises that were founded to provide competing products. Therefore, the impact of the original research into energy balance sheet simulation was not simply a single spin-off company to commercialize the product but an entire industry segment that provided a family of products that varied in sophistication, price, completeness, etc. Aspen Technologies was challenged by the competition to conduct its own R&D to improve its product line as well as learn from the basic research which continued at MIT. This situation illustrates how sensitive the potential conflicts of interest Evans was faced with must have been.

After the start up period of the enterprise, Oscar Manley, a program manager for the division of Engineering at BES suggested that Evans develop a partnership effort with the Idaho National Engineering Laboratories. BES would fund it to continue the research component of plant energy integration and batch process modeling and optimization. The conversations for this initiative began in 1983 and BES awarded the first grant in 1985.

The work under this cooperative arrangement lasted 3 funding cycles and, in 1994, the INEL component was discontinued. The MIT team received BES funding to continue the research program independently.

During the period beginning in 1985, the oversight role of Manley was very important because he managed a coordinated program at the engineering division of BES. The projects he supported were not simply selected among a population of unsolicited proposals based on their individual merit. Manley took a proactive role in encouraging submissions that were complementary and contributed to overall goals in energy related engineering research. He was very supportive of research that ultimately was used in some way by industry and gave Evans and his team great flexibility in licensing results of BES supported research for commercial ventures. It must be kept in mind that, during the entire period of BES support for the basic research in expert systems and process modeling conducted by Evans, he was also the president of Aspen Technologies. This commercial involvement proved to be instrumental in the high industry impact of this research. However, Evans maintained that it was an intellectual agenda rather than commercial or industrial priorities that determined the direction of research under BES grants.

The problems they addressed in the first period of BES funding were general interest expert systems problems that had no direct applications to industry problems of interest to Aspen Technologies. They were in the area of energy integration in order to balance the uses of energy in a chemical process. The main uses of energy in these cases are for heating and/or cooling material to the temperatures required for each stage of the process. Later, the emphasis shifted to modeling batch processes and the overall design problem of plants to implement the batch processes. The feasibility of the latter result was clear when the use of their models in a plant simulator achieved results very similar to those used in plants that been subjected to very long processes of optimization. A design package based on these methods promised to achieve in a single iteration what took years of field adjustments in the plant.

The great impact on industry that this research has had is due mainly to the career paths of students that participated in this research and were hired by Aspen Technologies and other industries and developed commercial software with the skills acquired doing research as members of Evans’ team. Graduates in this field are in great demand both for the companies that develop commercial software based on the results of basic research and in the chemical industry at large in order to apply the software intelligently to the particular cases of a plant or a firm. Evans has a very large consulting portfolio as well and is able to bring particular case studies or new problems from industry for students to work on in their theses.

The feedback from industry available through Evans’ connections is interesting enough. However, a striking pattern of personnel exchange provides still another much more substantial feedback channel. Graduating students have been hired by competitors of Aspen Technology in the same industry segment or by firms in the chemical segment and later hired back either by Aspen Technologies or returned to research work. The communication network of graduates plus the actual career path returning to the MIT-Aspen center has maintained this stream of research in very close connection with the relevant industrial sectors magnifying its impact potential enormously.

3. Outputs and Evidence of Possible Impacts

There are outputs of this research stream in several categories. There are numerous papers and reports over the span of the DOE grants. An exact count is not available from our documentation. But an indication of this productivity is the performance of the last three years, after Paul Barton took over the leadership of the research component at MIT. The number of published papers in refereed journals was 8 plus 4 conference papers in refereed proceedings and 7 more papers in preparation.

Besides the number of papers, it is important to mention the fields of publication as well in order to recognize the nature of its contribution to scientific knowledge. Given the high impact on industry of this stream of research, the former could easily be masked. Many papers are contributions, as expected, in the field of modeling of chemical processes, which are directly implemented in computer programs. However, many of the important papers are in applied mathematics with contributions to numerical analysis, algorithms for special cases of algebraic systems and optimization methods. At the same time, this work on generic methods and technologies enables development of other application areas as, for instance, in the case of a student working on applied methods for a solvent recycling system to improve environmental performance of chemical plants with support received from NSF and EPA.

Training of students is a significant output category for this case. The main research work in which students developed their skills was conducted under the direction of a single PI at a time, Evans until 1991 and Barton since 1992. Two students received their PhD degree during the last BES funding cycle and the total number since 1985 was 7. The number of PhD students at any one time ranges from 5 to 7. We have already mentioned that the demand for graduates in this field is greater than the supply this research program can provide. According to the PI, BES funding is crucial to human resource development because it is the only source of funding that provides financial stability for graduate students throughout most of the doctoral studies. Several private companies give yearly unrestricted grants ranging between $10,000 and $30,000 to the research program just for the privilege of receiving a call from the PI every time someone is about to graduate, giving them an early chance at recruiting the student. Given that graduates of the program are systematically employed by both industry segments, the commercial developers and users of chemical process simulation and design software, the impact of this work on industry is substantial. And, as we have mentioned, they also provide feedback channels for the research program.

Software packages produced in this work represent product outputs of the research. The commercial versions are not directly developed by BES funded work, but the academic versions were licensed on several occasions both to Aspen Technologies and other universities and represent the knowledge base for all the commercial ventures in this industry.

The spin-off company, Aspen Technologies, was not the result of the BES funded research. It had been created by the time this research program began receiving BES support. However, BES did encourage the same pattern of interaction with a commercial venture through the active involvement of its program manager, Oscar Manley. Therefore, the continued growth and viability of the company and others in the same industry segment can be attributed to the work in this research program.

4. Impact Maps

The impact maps for this case have multiple paths. There is a direct component common to most government funded basic research that has industry application of its results and which would ordinarily be interpreted in terms of the linear model of R&D. For this path, BES support enables research at MIT. The research program produces results that are published, contributing to increase in the stock of certified knowledge. These results are developed further into a prototype, which is demonstrated, tested and finally commercialized.

There are several paths that actually include feedback loops when we consider the human resource development component of this case. Students are hired by both the industry segment that develops software packages based on this research stream and by the industry sector that applies the packages to optimize, control and/or design large scale chemical commodity plants. Personnel has been hired back from one segment of industry to the other and from the software segment back to research constituting true feedback of experience and information that has an impact on problem formulation in research and product development in the software industry. Another feedback loop is constituted by the consulting position that Evans holds with most firms in the sector which allows him to bring back industry case studies for his students to develop in their thesis research. There is also a semi-permanent consulting exchange between Aspen Technology and the MIT research team, which brings industry information to bear on research as well. Actually, this link between the software industry segment and the research team, even though it has a privileged relationship with Aspen Technology, is present with its competitors as well because of the connection with students hired and the information about competing products that is acquired by Aspen Technology.

The economic impact on the chemical industry at large is very significant. Evans reports the following figures as an illustration of its magnitude:

   Union Carbide: eliminated the need for $40 Million expansion and reduced wastewater load in half.

   Mitsubishi Chemical: $10 Million per year savings in Ethylene; $6 Million per year in VCM; $2 Million per year in Acrylonitrile.

   Dow Chemical: $15 Million per year savings in process improvement.

   BASF: Avoided need to add fourth power station at Ludwigshafen.

   Dupont: Reduced need for capital investment of $30-50 Million.

   Sunoco: $8 Million reduction in operating costs.

Finally, there is an educational impact through licensing of the software to other universities for teaching in chemical engineering. The packages that are used educationally are the ones developed as the output of research and not the ones that are developed for commercialization at Aspen Technologies.

The attached schematic illustrates these impact paths in a pictorial diagram. It also includes the special facilitator role played by the BES program manager, Oscar Manley.

1. Conclusions

This case is a very rich example of multiple high payoff impact paths for BES funded research. The research team is relatively small, a single PI and half a dozen graduate students. The BES funding level has also been small, about $100,000 per year. However, the impacts are very diverse and sizable, even in sheer economic terms for the industrial end users of the research results.

Two important peculiarities of this case are worth restating because of their role in facilitating the high impact ratio. First, the deliberate "traffic" in graduate students that leads to their "spiraling" career paths. The return of students to the privileged industrial partner, Aspen Technologies, and the research team at MIT constitutes a very effective source of information to maintain the relevance of the research and innovation agenda. It must be noted that many of the problems that industry perceived had not been articulated in ways that allowed them to readily find technical solutions. Therefore, the interesting aspect of the mobility of personnel in this case is that it was by people who were able to translate these problems in terms of the problem solving capacity of the MIT research team.

Second, the active role of the BES program manager is remarkable. His contribution was both at the level of backing the risks that Evans was taking in pursuing a dual scientific-commercial agenda and in providing a rationale for this research program by fitting it within an integrated program of support for research at his BES division. Evans mentioned the two instances of accountability he was responsible before: his home academic institution, MIT, and the main funding agency, BES. MIT did not intervene in order to increase the chances of success of this research program with all its components. It operated as an ordinary academic institution that allowed a successful researcher to pursue his goals. In the cases of potential conflict of interest, it adopted a passive stance and did not obstruct Evans’ work. On the side of BES, Manley reassured Evans that the path he chose was acceptable and even a desirable use of agency funds. He also facilitated the open licensing agreements for the results of DOE sponsored research.

These institutional and strategic characteristics added to the high quality of research proved to be a winning combination and yielded a high impact research program.