At the CAPE-OPEN 2020 Annual Meeting, CO-LaN Management Board asked Dr Michel PONS, Chief Technology Officer of CO-LaN, to deliver his views (access PDF here, access video here) on what could be version 2.0 of the CAPE-OPEN standard after the presentation Dr Mark STIJNMAN gave on this topic at the CAPE-OPEN 2019 Annual Meeting.
Dr Michel PONS has been the Chief Technology Officer of CO-LaN since January 2005 after having been involved in the CAPE-OPEN, Global CAPE-OPEN and GCO Support projects funded by the EU, as representative first of ELF and then of TOTAL in these projects. His involvement in CAPE-OPEN goes even back to 1995 when several operating companies, including ELF, investigated jointly the possibility (OO-CAPE and OS-CAPE common projects) to create an interoperability standard for process simulation software.
Michel PONS addressed different points raised by Mark:
- Changes in the computing landscape
- Changes in the process simulation landscape
- Pain points in the current standard:
• Unit operations
Dr Michel PONS points first that, so far, only the scope of version 1.0 of the CAPE-OPEN standard was worked upon and defined, as part of the Global CAPE-OPEN project, when this project was put together. So defining the scope of a version of the CAPE-OPEN standard is a completely new exercize for CO-LaN. Also the speed is slow at which CO-LaN is capable of releasing new interface specifications, very much like most Standard Development Organizations (SDOs) while CO-LaN operates with less resources than most SDOs. This should be kept in mind when setting goals for CAPE-OPEN 2.0 in order to avoid aiming too high. Dr Michel PONS sets October 2025 as a possible horizon for CAPE-OPEN 2.0, in sync with the new duration given to CO-LaN by the AGM on October 15, 2020.
Dr Mark STIJNMAN stated last year that even if the present in process simulation software is single-threaded, future is multi-threaded. Dr PONS joins Mark on this statement and points that version 1.2 of the CAPE-OPEN standard starts to consider the effects of different threads on the operation of CAPE-OPEN (modifications brought to ICapeUtilities::Edit). What is more important, as reported previously at CAPE-OPEN Annual Meetings: multi-threading is addressed by COBIA. The exact scope of Phase 3 of the COBIA development project is not frozen yet but the Methods & Tools Special Interest Group has already developed a threading model for COBIA and the rules attached to it.
Mark also stated that the future of process simulation is cloudy. Dr PONS remarks that several software vendors have indeed started to offer solutions on the cloud and named some of them. However user subroutines, and even less CAPE-OPEN plug-ins, are not yet part of the offers made. Following discussions with different software vendors, part of the reluctance to include, in the offers made, plug-ins such as CAPE-OPEN ones, is related to license issues.
Dr PONS asks for business cases to be developed on “CAPE-OPEN in the cloud” to permit its consideration in Phase 3 of the COBIA project. Dr PONS mentioned a business case demonstrated back at the CAPE-OPEN 2006 Annual Meeting by PSE Ltd: the Virtual Plant Demonstration Model (VPDM) project. This business case stayed as a demonstration project: what could make it more attractive today? It is the responsibility of software vendors and end-user organizations to develop plausible business cases that could drive the development of a solution within CO-LaN. While adding the cloud dimension to CAPE-OPEN 2.0 could make CAPE-OPEN more attractive, any business case should address the CPU performance issue which has been sometimes a sore point for CAPE-OPEN (no thanks to COM and to non-optimized implementations).
Next Dr PONS agrees with Mark’s view on “More dynamic and equation-oriented modelling in future than currently”. Dr PONS states first that equation-oriented modelling has been at the core of CAPE-OPEN since the start of the CAPE-OPEN project, at the same level of importance as Unit Operations and Thermodynamics. The design of CAPE-OPEN numerical interfaces was conducted by Imperial College, as partner in the CAPE-OPEN project and the Global CAPE-OPEN project with ELF as the operating company leading the corresponding workpackages. The conceptual design achieved then relies on the Equation Set Object (ESO) to encapsulate a system of equations to be solved and on the concept of various types of numerical solvers. This design / specification was implemented (with sometimes some adaptations) in one commercial tool (gPROMS) and in several academic applications (EMSO, CheOps, DIANA).
But the CAPE-OPEN numerical interfaces have been kept out of the scope of COBIA Phase 2, therefore are not present in version 1.2 of CAPE-OPEN. Will they be in or out the scope of COBIA Phase 3 and consequently of CAPE-OPEN 2.0? The current reasoning within the Methods & Tools Special Interest Group is that there is almost no implementation of these interface specifications in commercial tools, and there is no Numerical Special Interest Group at CO-LaN to take care of the maintenance of these interfaces. Therefore there is no business case for maintaining the CAPE-OPEN numerical interfaces within CAPE-OPEN 2.0.
Dr PONS goes on by listing the past actions at CO-LaN with respect to numerical interfaces as well as some attempts to regain momentum in that domain. It more or less boils down to the lack of a Numerical Special Interest Group while the CO-LaN membership contains the necessary expertise to populate one. It is Dr PONS’ s analysis that without a Numerical SIG, there will be no CAPE-OPEN numerical interfaces in CAPE-OPEN 2.0, missing therefore what pertains to equation-oriented process modelling and simulation.
Next Dr PONS covers the situation with respect to dynamic simulation where CO-LaN has prototyped an interface specification for dynamic unit operations. However there are several issues with this specification, both about its design and its implementation. This specification falls within the purview of the UNIT SIG. Another approach is with the numerical interface specification for Differential Algebraic Equations but this takes us back to the absence of a Numerical Special Interest Group.
Dr STIJNMAN had also listed a number of sore points in the current standard in thermodynamics and in unit operations. Sore points that could be addressed in CAPE-OPEN 2.0. One dealt with the low level of functionality offered by the Property Package Manager. As advocated by the Thermo Special Interest Group and reported earlier within the CAPE-OPEN 2020 Annual Meeting, a new interface specification, a Manager interface specification, has already been sketched out and will be developed by the Thermo SIG in order to address, if not all, most of the issues raised by Mark in his talk last year. Evidently this new specification will find its way in version 2.0 of the CAPE-OPEN standard.
Mark also pointed out that “Property Packages are inherently unsafe for multi-threading“. Dr PONS remarked first that Property Packages are progressively made threadsafe and the presentation by KBC at the CAPE-OPEN 2020 Annual Meeting points that out. Mark advocated major changes in the design of the Thermodynamics and Physical Properties interface specification to eliminate the heavy reliance on Material Objects as data containers. Dr PONS pointed out that the current interface specification is widely implemented, that the new solution needs to be fully specified and therefore would require a major time investment within CO-LaN. Changing completely the approach in thermodynamics could mean the success or failure of CAPE-OPEN 2.0 in terms of scope and timing. CO-LaN needs to hear from software vendors about such a possible major change in design. Dr PONS stated that each design has its flaws and its advantages and this has to be weighted against the risks involved in a change that needs to be accepted and implemented widely.
Dr PONS lists next several points that could be covered by CAPE-OPEN 2.0 as they have been addressed to the Thermo SIG over the years before moving to the “need some way to support “classic” unit operations in an equation-oriented environment” mentioned by Mark. There one has to go back to the discussion on the equation-oriented approach made above and especially the proposal made 10 years ago by PSE Ltd in order to tackle this need. By taking the direction outlined by PSE Ltd, the support of classic unit operations in an equation-oriented environment could be present in CAPE-OPEN 2.0 provided copyright issues are resolved.
Dr PONS briefly touched upon the recent interactions with CFIHOS and DEXPI following the presentations made at the CAPE-OPEN 2019 Annual Meeting. While this represents a challenge because it is a domain largely outside the process modelling and simulation domain CAPE-OPEN is about, a solution of low complexity could be achievable within the timeframe of CAPE-OPEN 2.0 development.
Dr PONS ends up by summarizing what CAPE-OPEN 2.0 could be made of:
- Close integration between chemical reactions and thermodynamics
- Manager interface offering enhanced functionality compared to Property Package Manager
- Unit Operations
- Black-box modeling support
- Equation-oriented and dynamic simulation
- Strong basis exists that could be ported to CAPE-OPEN 2.0
- Needs a Numerical SIG to make it happen
- Opening to digital twin aspects
- Software engineering aspects
- Multi-threading / cloud: technical scope of COBIA Phase 3