PTQ Q2 2026 Issue

At the core of the appli- cation is the health index, a composite metric combining deviation magnitude, data quality, and calibration his- tory. It provides a single metric that quantifies the ‘fitness’ of each digital twin. Open REST APIs make this fitness score accessible across the enter- prise, ensuring operational digital twins and planning models remain synchronised with the data historian. As shown in Figure 3 , this

Real-time optimisation Planning & scheduling Enhanced unit monitoring

What-ifs

Performance indicators & data quality parameters

Real-time health scores Recalibrated & returned Digital Twin

Process Twin

KBC Acuity Process Twin Pro

Actual

Figure 3 Real-time health scores update

closed-loop cycle links the new-generation digital twin to the site-based digital twin or the simulation model. The site continuously feeds new, real-time information into the application, and this new information generates updated health scores. Automating the lifecycle of the simulation models assures model integrity in near real time. It reduces manual work- load and improves responsiveness to process changes. Unified dashboards provide enterprise-wide visibility, alert - ing engineers to emerging deviations and offering actiona- ble recommendations. This approach supports safer, more stable operations and eliminates many of the bottlenecks associated with manual data reconciliation. Continuous model maintenance at scale Can automation really sustain digital twin accuracy at refin - ery scale? While the technology’s principles are universal, its impact becomes clearest in real-world deployment. A pilot project in early 2024 at a 400,000 bbl/d integrated refinery and petrochemical complex aimed to find out. The site operates multiple high-conversion units where accurate process simulation and LP alignment are critical to maintaining profitability and emissions compliance. Using the previously defined MPI, DQP, and KPI configuration, the project established a baseline to measure simulation accu- racy, data fidelity, and operational impact across multiple units. In practical terms, MPIs show whether the model accu - rately represents plant behaviour. DQPs evaluate whether the input data used for calibration and validation is reliable. KPIs connect model accuracy to operational outcomes. While MPIs and DQPs assess technical integrity, KPIs measure the impact on refinery performance, including energy consump - tion, throughput stability, hydrogen utilisation, and emissions intensity. In this way, KPIs indicate whether improved model fidelity supports better operational and economic decisions. For each digital twin, a health index quantified the devi - ation between model predictions and actual plant data. When this index exceeded thresholds, the system auto- matically triggered recalibration protocols and notified the engineering team. This action closed the feedback loop between process data and model assurance. The core of tuning leverages ML algorithms to exe - cute parameter optimisation and cross-prediction. The

application framework evaluated calibration performance across datasets and recommended parameter updates based on historical tuning behaviour. Sensitivity analysis identified variables with the greatest influence on accuracy. This step ensured recalibration focused on high-impact parameters, resulting in faster convergence and increased confidence in model adjustments. Operationally, continuous model maintenance improved planning accuracy and responsiveness. It allowed the LP model to remain synchronised with current constraints, such as energy integration efficiency and feed variability. From a workforce standpoint, automation reduced depend- ence on scarce domain expertise. It enabled fewer engi- neers to manage more complex systems while maintaining confidence in model integrity. Quantifiable performance and safety outcomes Refinery studies have begun to quantify the operational impact of digital twin deployment beyond conceptual benefits.3 Reported results include measurable reductions in unplanned downtime, energy consumption, and main- tenance labour requirements following digital twin imple- mentation, demonstrating that improved model fidelity and predictive capability translate directly into operational and financial performance improvements. These findings reinforce the importance of maintaining digital twin accu- racy over time, as the economic value of the technology depends on sustained alignment between the digital model and physical asset performance. As shown in Table 1 , early deployments show projected measurable benefits. Continuous monitoring of health indices prevents small misalignments from becoming major operational upsets. Automated recalibration minimises flaring, maintains stable Projected performance improvements from continuous automated model maintenance based on early refinery deployments

Metric

Projected result (per large refinery) More than 4,000 hours per year 10 times faster model tuning $0.5-2 million per asset per year ≈2,500 t per asset per year

Engineering hours saved Model-tuning speed Margin improvement

CO₂ reduction

Table 1

PTQ Q2 2026