Chemical industry is clearly moving from fossil fuels to renewable feedstocks, but it's slow. One of the barriers is the complexity of estimating the viability of producing new bioproducts on a comercial scale, specially when no previous processes, bibliography or patents are available.
To achieve a transition to the bioeconomy, new processes will need to be evaluated, new flowsheets will have to be developed, and here's where digital technologies can help.

Artificial Intelligence

The power of learning systems

The main reason we started this whole journey is because we wanted something more than process simulation. We wanted a tool capable to explore and learn for itself, and use this knowledge to give recommendations to the user, as a copilot, to help obtain the best facility designs to transform raw materials into products according to economic and environmental criteria.

AI-assisted tools have demonstrated a lot of potential in the field of science, with applications like new drug discovery or protein folding. So why not to apply AI in process systems engineering?

Today's challenges

Sustainability

Companies are focusing in new plant based products, obtained from renewable resources instead of petroleum or animals.

New answers will be needed to this type of questions: Which are the best routes to transform these raw materials into my products of interest? Which equipment do I need? Will it be profitable? Which is the ROI? Which steps of the process are the most expensive? In a fast and evolving market, all these questions need fast and flexible answers, and here is where our platform helps.

FAQs

Frequently asked questions. Contact us if you have more questions about our software.

The mass and energy balances, the equipment characteristics, prices and consumptions, the process flow diagram, the economic analysis of the whole facility, and its LCA (life cycle analysis).

Short answer: If it can be simulated, yes.
Long answer: AI only can learn from what can be simulated, either through mechanistic, hybrid, or data-driven models. The user has to introduce the kinetic model of the (bio)reactions, if any. Flowsheet copilot contains its own thermodynamic model and a range of unit operations (distillation, filtration, mixers, heat exchangers, crystallization, centrifugation, etc...), based on real equipment. New unit operations can be created by the user.

In each state of the flowsheet design, a deep neural network reads all the information of each stream of the flowsheet (compositions of each chemical, temperatures, pressures, viscosities, densities, etc...), processes it and decides which are the best candidates to be the next unit operation. This neural network is trained with reinforcement learning, with the objective to maximize a multi-objective function (economical and environmental) at the end of the design. After millions of simulations, Flowsheet Copilot has a deep intuition about how to design new processes.

Our software is not yet on the market. We are developing the web platform so companies can use it by their own.
Despite this, we are performing internal trials with companies. We will provide a free demo to companies that are interested in our product.

No. Flowsheet Copilot is a powerful tool for exploring and optimizing from scratch. Once the flowsheet is generated, it is a good starting point to study in more detail the process, perform simulations with other process simulators, etc...

What do they say about us?

Our software is not yet on the market. We are developing the web platform so companies can use it by their own.
Despite this, we are performing internal trials with companies. We will provide a free demo to companies that are interested in our product.