“Aspen Hybrid Models are a major advance in the field of chemical engineering…  a game changer in process engineering and plant improvement.”   

Dr. Karuna Potdar, Vice President, Reliance Industries

 

On October 6, Aspen Tech released Aspen Hybrid Models for Aspen Plus^®, Aspen HYSYS^® and Aspen PIMS-AO™.  What’s the big deal?

When earlier this year, AspenTech revealed the details of how we could merge process simulation first-principles models capabilities with AI–based machine learning into one hybrid toolset, many of our most important global customers jumped at the opportunity to get an early look and to test the possibilities. They brought us some of their most troublesome and difficult asset optimization challenges to see how this new approach could help. The results have been exciting. The combination of first principles, domain expertise and AI has provided insight into improving the operability, uptime and performance of process equipment and complex processes. Hybrid models deliver these insights using an intuitive, easy-to-grasp workflow; as well as models that can be run at larger scale and much faster. The feedback from industry to us has been uniformly positive.

 

In May, the Wall Street Journal published an article titled, AI isn’t Magical and Won’t Help You Reopen Your Business. Experts observed that in the time of rapid change (such as in 2020), “… to carry out one of its primary applications, predictive analytics, today’s AI requires vast quantities of relevant data. When things change this quickly, there’s no time to gather enough…”  This explains perfectly why AspenTech’s hybrid approach of Industrial AI, which places analytics and machine learning into the context of chemical-physical first principles and domain knowledge, IS so well-suited and important in addressing today’s volatile and changeable world, which translates into rapid asset operational changes.

The Aspen Hybrid Models approach begins with the correlations, patterns and insights that lab, pilot plant, and asset operational data reveal. It applies data science, cleaning the data transparently; machine learning, providing predictive insight into complicated processes; then applies the guardrails of first principles knowhow, that have been built into the robust Aspen HYSYS and Aspen Plus simulators involving hundreds of years of chemical engineering and operations domain expertise. The result: A major step forward in chemical and process engineering.

This week, as AspenTech releases this capability deployable in our widely-used modeling environments, we have validated this new technology with more than 65 different use cases that span applications in upstream oil and gas, midstream, refining, petrochemicals and specialty chemicals. 

Hybrid Models are immediately applicable in both operating and design environments: they are designed so AI-based models can be built by process engineers and planners, not data scientists. They provide intuitive workflows that incorporate deceptively sophisticated data science and machine learning algorithms within a normal engineer’s work processes and paradigms. Best of all, this is all done in a manner such that the resulting models (that accurately describe complicated process equipment, reactions and process flows) can be inserted into existing Aspen HYSYS and Aspen Plus models (even those built in versions 10 and 11 of these simulation systems.)

The earliest application of hybrid models have already revealed very high-value use cases. This, though, is just the beginning. We are eager to work with companies in multiple industry sectors to apply this capability to:

• model assets which require models that closely match actual plant conditions to achieve higher levels of optimization and asset integrity
• model large scale complex problems, such as integrated oils-to-chemicals sites, oil and gas upstream carbon reduction
• deploy fast models online, and much more.

Here are a few sample use cases:

 

Create More Accurate Planning Models

Refining reactor models can now be rapidly turned into high-accuracy, reduced order, non-linear sub models and run in your existing planning models to achieve higher performance. Based on a wider data range, they are valid across more operating conditions. The SARAS refining organization told us, “Aspen Hybrid Models provide very efficient nonlinear planning model generation, taking information from Aspen HYSYS rigorous refining reactor models and offering a great deal of promise as a new approach for updating planning models.” The value can be $0.03 to $0.15 per barrel in today’s market.

 

Predict Chemical Qualities with Virtual Sensors

Hybrid Models will be able to infer characteristics such as color, pH, polymer melt index and hardness, elasticity, flashpoints, oil viscosity and assay properties. This provides value in optimizing operations to minimize lost or inefficient production and maximizing asset integrity. 

 

Improve Operability and Uptime in Polymer Lines

A customer came to us with an advanced polymer production line that was experiencing periodic and unexplained challenges in managing startups and shutdowns, resulting in longer than expected transition times between polymer runs, often causing wasted product and lost production.  Hybrid models look extremely promising in improving such polymer operations, with values easily in the $ 0.5 M to $1M per year benefit.

 

Predict Performance of Difficult-to-Model Units 

Hybrid models provide much more accurate performance predictions for unit operations involving membranes, fluidized beds and crystallization.

 

Develop Oils-to-Chemicals Integrated Production 

As companies navigate the energy transition, many refiners are looking at process technologies that can change their mix of products to add petrochemical units integrated with their refining sites. Reduced-order models of key process units can be combined into larger scale, fast running models to evaluate alternatives and optimize technology selections in design and operating conditions to achieve higher economics and flexibility, while minimizing carbon emissions and energy use.

 

Integrate Gas Field Modeling 

Gas production fields offer operators considerable optionality in terms of gathering systems, separation and contaminant removal processes, compression and transport. Production operation decisions have immediate economic impact. Hybrid models have been tested in gas plant simulations to create fast-running reduced order models that can provide both operator advice to save energy use and accurate production allocation financial information. Savings in production allocation effort alone is worth $0.5M per year per site.

 

Speed Time to Market for New Technologies 

Companies will be able to quickly and accurately model new technologies using lab and pilot plant data, to accelerate proof of scale up economics and commercial viability -- in particular for sustainable processes.

Hopefully this tour of some of the use cases that our customers and industry experts have surfaced will immediately spark ideas as to how you can apply this exciting technology to address your most important challenges. We expect this to be just the beginning of many valuable use cases for Aspen Hybrid Models.

 

 

To learn more about Aspen Hybrid Models, visit our solutions page.