ISU Chemical, a leading chemical producer in Korea, set out to overcome persistent challenges in optimizing reactor performance and managing catalyst lifecycles. Traditional models couldn’t fully capture the complexity of its operations, leading to unnecessary downtime and increasing costs. By implementing AVEVA™ Process Simulation and an innovative AI-driven hybrid modeling approach, the company unlocked greater precision and efficiency, transforming how it plans and performs.
Challenges
Optimize reactor performance to improve production outcomes
Establish an efficient catalyst replacement plan to minimize downtime and costs
Build feed and product composition structures from sample assay data
Address limitations of first principles modeling for complex reaction prediction and catalyst behavior
Results
99.7%
accuracy in predicting reactor yield for different recipes and operating environment
Enabled prediction of catalyst performance to facilitate an efficient plan for catalyst replacement
Engineers and operators can proactively simulate the plant through an external HMI built in Excel
ISU Chemical, a leading chemical producer in Korea, sought to enhance its reactor performance and catalyst lifecycle management. Leveraging the latest capabilities of AVEVA Process Simulation integrated with AI through a hybrid modeling approach, ISU Chemical achieved breakthrough accuracy in predicting reactor yields and catalyst decay. This high accuracy leads to data-driven replacement schedules that minimize downtime and extend catalyst life while also reducing simulation times from up to 20 seconds to near-instant responses. Now, engineers can make faster, more informed decisions.
Seeking to realize potential performance optimization
ISU Chemical’s operations rely on precise modeling and control of complex chemical reactions. However, several persistent challenges limited efficiency and optimization potential:
- Reactor performance optimization: Existing first-principles models could not fully capture the complexity of reaction dynamics under varying feed and operational conditions, resulting in conservative operation.
- Catalyst lifecycle management: Catalyst replacement decisions were often based on empirical data or past experience rather than predictive insights, leading to unnecessary downtime and higher costs.
- Data integration complexity: Building feed and product composition structures from sample assay data was a slow, manual process that limited agility.
- Simulation speed: The detailed thermodynamic and kinetic model used for post-combustion carbon capture required 10–20 seconds per run, constraining scenario testing and slowing decision-making.
If these issues persisted, ISU Chemical risked higher operating costs, reduced throughput, and missed opportunities for performance optimization, particularly in fast-changing market conditions.
Solution
Using AVEVA™ Process Simulation’s hybrid modeling capabilities, ISU Chemical integrated a machine learning reactor model directly into the process simulation environment via an ONNX adaptor.
Integrating AI into the solution
ISU Chemical selected AVEVA Process Simulation for its advanced hybrid modeling capabilities, which allow seamless integration of physics-based process models with AI and ML components. The platform’s Open Neural Network eXchange (ONNX) adaptor was a critical enabler, allowing the company to deploy ML models developed in external frameworks directly into AVEVA Process Simulation. This meant ISU Chemical could combine the strengths of traditional process engineering with the flexibility of AI without rebuilding or duplicating existing infrastructure. In addition, the solution’s open architecture and compatibility with widely used tools, such as Excel, made it easy for engineers and operators to access and leverage the hybrid models across workflows.
AVEVA Process Simulation presents the very rigorous and powerful hybrid model combined with AI that can predict reaction yield, catalyst decay, and operation performance.
-DH Kim, Process Engineer, ISU Chemical
The project focused on ISU Chemical’s post-combustion carbon capture (PCCC) process using Monoethanolamine (MEA).
Post-combustion carbon capture using Monoethanolamine (MEA)
This process removes CO₂ from flue gas generated by combustion, but its high energy consumption and complex reaction kinetics make optimization challenging. To overcome these challenges, ISU Chemical deployed a hybrid modeling strategy within AVEVA Process Simulation:
- Automated data generation: Using AVEVA’s scripting interface, engineers generated approximately 2,000 simulation cases through Latin hypercube sampling to capture diverse operating conditions for model training.
- Neural network development: A machine learning model was trained on the simulation data to accurately reproduce the carbon capture process behavior, including absorption, regeneration, and heat integration.
- Integration via ONNX: The trained ML model was exported in ONNX format and embedded directly within the AVEVA Process Simulation environment. This allowed it to operate alongside first-principles models without needing external interfaces.
- Operator access via Excel HMI: Engineers built a custom Excel-based interface that connects directly to the hybrid model, enabling instant simulation runs and easy experimentation with different process scenarios.
This implementation allowed ISU Chemical to test different feed compositions, reactor recipes, and catalyst conditions with speed and precision, transforming what once took minutes into real-time predictions.
Unprecedented accuracy and enhanced efficiency
ISU Chemical’s deployment of AVEVA Process Simulation with integrated, AI-driven hybrid modeling has delivered significant operational improvements for the organization. These include:
- Unprecedented accuracy: It achieved 99.7% reactor yield prediction accuracy, even across varying operational environments and feed compositions.
- Predictive catalyst management: The model forecasts catalyst decay with high confidence, supporting data-driven replacement schedules that minimize downtime and extend catalyst life.
- Real-time decision-making: Simulation times were reduced from up to 20 seconds to near-instant responses, enabling engineers to make faster, more informed decisions.
- Enhanced efficiency and sustainability: Optimized reactor operation reduced energy usage and improved overall plant efficiency, directly contributing to sustainability goals.
- Improved collaboration: The Excel-based HMI allows engineers, operators, and decision-makers to access hybrid models seamlessly, fostering a shared understanding of plant performance.
Unprecedented accuracy and enhanced efficiency
Through the integration of AI with AVEVA Process Simulation, ISU Chemical has transformed how it approaches process optimization. The combination of first-principles modeling, machine learning, and user-friendly deployment delivered measurable accuracy, speed, and sustainability improvements, positioning ISU Chemical as a pioneer in AI-driven industrial simulation.
Product highlights
AVEVA Process Simulation
Formerly Known As SimCentral Simulation Platform
Design sustainable processes, products, and plants at the speed the market demands. AVEVA Process Simulation moves beyond linear, wasteful workflows to enable a circular, sustainable world.
Related success stories
Stay in the know: Keep up to date on the latest happenings around the industry.
