KBC, a Yokogawa company, has released Petro-SIM 7.7, a process simulation, optimization, and digital twin platform aimed at engineers and safety specialists in refining, petrochemical, chemical, and general process industries. The version number matters less than the architectural shift it represents: AI/ML-enabled hybrid modeling now runs inside the same environment as first-principles simulation, rather than in a separate data-science toolchain.

That consolidation addresses a practical gap. Process engineers who want hybrid models — machine-learned correlations anchored to engineering physics — have typically needed data science skills or a specialist collaborator to build and deploy them. Petro-SIM 7.7 embeds an ML Utility Hub that generates synthetic data and lets users develop, train, and deploy hybrid models within the simulation environment itself. The company positions this as the mechanism for making hybrid modeling accessible to engineering workflows rather than analytics teams.

"Industrial AI delivers the greatest value when it's grounded in engineering physics," said Philippa Hayward, product manager for Petro-SIM. "Process engineers need an accessible way to develop and apply hybrid models within a trusted simulation environment, helping them solve increasingly complex problems without specialist data science expertise."

The physics-versus-data distinction is not cosmetic in this domain. A pure first-principles flowsheet model of a refinery unit can take weeks to converge and may still miss unit-specific behavior that historical plant data captures well. A pure ML model, conversely, interpolates reliably within its training envelope but extrapolates poorly into operating regimes the plant has never visited. Hybrid modeling — physics providing structure and extrapolation, ML supplying residual accuracy — has been the industry's working answer to that trade-off. What KBC claims for 7.7 is not a new modeling paradigm but a lower barrier to entry for it, inside an environment whose thermodynamic and unit-operations basis engineers already trust.

Beyond the ML Utility Hub, the release widens the optimizer library and adds new optimization algorithms for complex refinery process operations. An expanded suite of reactor capabilities targets higher-fidelity modeling of refinery units and integrated value chains — the kind of end-to-end models that refineries use to track margins from crude slate to product slate.

The energy-transition portfolio gets its own set of additions. Version 7.7 supports bio-oil processing, electrolysis, and new decarbonization correlations, with functionality for pyrolysis and biomass gasification modeling. For refiners evaluating co-processing of biogenic feeds or building hydrogen capacity via electrolysis, the practical question is whether a single platform can model both the conventional asset and the emerging units in one consistent thermodynamic basis — which is the integration KBC says it provides.

Petro-SIM 7.7 also serves as the engineering foundation for KBC's process digital twins, including Acuity Process Twin Pro, a digital twin application. The two together, according to KBC, enable more accurate monitoring and operational decision-making across refinery and petrochemical value chains.

All performance characterizations in the announcement — "more accurate monitoring," "greater fidelity" — are vendor claims without published benchmark conditions. KBC has not yet released comparative validation runs showing hybrid-model accuracy against measured plant data, or convergence-time figures for the new reactor models. Buyers evaluating the release against competing platforms such as AspenTech's or AVEVA's hybrid offerings will want exactly that evidence.

The trust question sits at the center of the CTO's argument. "Industrial AI is only valuable when engineers can trust it," said Simon Rogers, chief technology officer at KBC. "That trust is built on transparency, engineering rigor, and results that can be validated. AI should strengthen engineering judgment, not replace it. Our vision is to embed intelligent technologies across our software portfolio, helping engineers solve increasingly complex challenges with confidence."

That framing sets the adoption bar. The release's success will rest on whether process engineers without data science backgrounds can actually train hybrid models that validate against plant data — and whether safety specialists, who carry the regulatory burden for any model used in decision-making, accept ML-augmented results as auditable. Validatability, not algorithm sophistication, is what determines whether hybrid simulation moves from pilot studies into the operating envelope of refineries subject to safety-instrumented-system governance.