Equation Based Modeling - Time to renew the vows
Updated: Mar 28
“Modeling is the process of mapping reality into a representation that is thought to be useful for understanding that reality” wrote Kirk Abbott in his PhD thesis titled “Very Large Scale Modeling” (Abbott, 1996). This is apt in process engineering, design and operations.
1990’s was the decade when significant emphasis was laid on various types of process modeling techniques for chemical processes in academia and in industry. The core methodology of process modeling has been the sequential modular solving techniques. Most commercial process simulators (Aspen Plus, Aspen HYSYS, PRO-II, Unisim, Petrosim) use sequential modular solution techniques or a more enhanced version of it called simultaneous modular (Chen and Stadtherr, 1985). Also, in the 1990’s other commercial simulation technologies such as SPEEDUP1, RT-OPT2 (both from AspenTech), ROMEO (from Simulation Sciences, now AVEVA) and gPROMS (from Process Systems Enterprise, now Siemens) became available. These tools used an alternate solution methodology formally classified as equation-based modeling. While RT-OPT and ROMEO focused mostly on Online Closed Loop Real Time Optimization, SPEEDUP mostly focused on dynamic simulation (back in the 1990’s). Other technologies in the Refinery Planning space such as PIMS, GRTMPS, Spiral Plan as well as software such as GAMS, IMPL all have characteristics of equation-based modeling systems. In that context, they all separate the formulation of the problem from the solution of the problem. While there was significant academic and industrial discussion and research upon the benefits and strengths of each of the approaches, i.e. sequential modular vs equation-based back in the 1990’s especially in the first principles modeling space, that discussion has taken a back-seat over the last two decades. Most simulation practitioners have primarily adopted the sequential modular approach for offline modeling/design, while equation-based modeling has prevailed in online Real Time Optimization (RTO). The use of closed loop RTO itself has dwindled somewhat in refining due to the cost of implementation and maintenance, as well as significant improvements in Advanced Process Control (APC) which somewhat blurs the boundary between APC and RTO.
We believe it is time to renew and revive the interest in Equation Based Modeling in the first principle-based modeling space. More and more commercial tools such as Aspen HYSYS and Honeywell Unisim now have this capability. As mentioned above, Aspen Plus and Aspen Dynamics, ROMEO, GPROMS all have equation-based first principles modeling capabilities already. A the same there is an push towards digitalization as part of Industrial Revolution 4.0 in most large and medium sized companies. Digital Twins which play a key role in most digitalization strategy, and this is a great opportunity for process and control engineers to re-emphasize the value of first-principles model based digital twins for process optimization. We believe that Equation-Based first-principles models as digital twins are not only desirable but almost necessary in today’s complex chemical plant operations. With Digital Twins comes the need to keep the digital twins “current” and integrated with plant data and operations, and this step of keeping the models “current” has always been a challenge with traditional sequential modular modeling technologies. However, most equation-based process model technologies have inherent data reconciliation and parameter estimation capabilities, which are key capabilities for models to adapt to changing conditions. This is a key factor for their high applicability in online RTO. These capabilities enable the process and control engineer to continuously keep the plant calibrated against current operations, in a seamless and robust manner. Once you have a well calibrated digital twin, the possibilities of their application for process optimization are enormous. We can use it for performance monitoring, or for updating LP planning models. We can also optimization our processes while obeying plant equipment and quality constraints and set new APC limits. And we can also perform what-ifs with new feeds and new operating conditions.
So, let us renew our vows – our love for equation-based models. We believe that both software vendors and service providers such as ourselves should increase the awareness and hands-on training for equation-based modeling. While the software capabilities in this area are mature, there remain opportunities to improve failure diagnostics messaging and understanding. Sequential Modular techniques are always easier to diagnose in case of failure relative to equation-based techniques and therefore some investment in this area would be useful. But the biggest need of today in terms of generating greater value for process industry in simply awareness and utilization.
1 SPEEDUP to the best of our knowledge is subsumed into Aspen Custom Modeler by Aspen Technology, Inc.
1 RT-OPT to the best of our knowledge is subsumed into Aspen Plus by Aspen Technology Inc.
1. Abbott, K. A.; Very Large Scale Modeling. Ph. D. Thesis. Department of Chemical Engineering Carnegie Mellon University, Pittsburgh, PA, 1996.
2. Chen, H.S., Stadtherr, M.; A simultaneous‐modular approach to process flowsheeting and optimization. Part I: Theory and implementation, AIChE J. 1985, 31, 1843.
1. Aspen Plus, Aspen HYSYS, Aspen Custom Modeler, PIMS, refer to www.aspentech.com
2. GAMS refer to www.gams.com/products/gams/gams-language/
3. gPROMS refer to www.psenterprise.com/products/gproms
4. IMPL refer to www.linkedin.com/company/industrial-algorithms/
5. Petrosim refer to www.kbc.global/software/simulation-and-optimization/
6. PRO-II, ROMEO, Spiral Plan refer to www.aveva.com/en/products
7. Unisim refer to www.honeywellprocess.com/en-US/explore/products