This line of research aims to develop methodologies for automated identification of polymers in plastic waste from Fourier transform infrared (FTIR) spectra, employing advanced machine learning techniques. Different model architectures are investigated, including Transformers-based approaches (BERT), convolutional neural networks (CNNs), and recurrent networks (LSTM), in order to increase the robustness and accuracy of classification.

The database used in this line is continuously expanded in partnership with Prof. Douglas Alexandre Simon (IFRS – Farroupilha campus), and currently has hundreds of FTIR spectra of virgin and recycled polymers. This collection covers both homopolymers and copolymers, including materials such as polypropylene, polyethylene terephthalate, polyethylene (low and high density), acrylonitrile butadiene styrene, polystyrene, thermoplastic polyurethane and polyamides, representing a diversity of polymeric classes of great industrial and environmental relevance.

The research directly contributes to advances in advanced recycling, circular economy and innovation in material characterization techniques, reinforcing both the potential for industrial application and the positive environmental impact associated with the efficient identification of polymers in plastic waste.

Among the results already achieved in this line of research, the following publication stands out:

» NETO, J. G.; SIMON, D. A.; FIGUEIREDO, K. T.; BRANDÃO, A. L. T. Framework for data-driven polymer characterization from infrared spectra. Spectrochimica Acta Part A, v. 300, p. 122841, 2023.

In partnership with the Applied Macromolecular Engineering Group at the University of Alberta, coordinated by Prof. João Soares, this line of research explores the use of convolutional neural networks to correlate microstructure properties such as short-chain branch distributions (SCBD), molar mass (MWD) and chemical composition distributions (CCD) with polymerization conditions of ethylene copolymers with α-olefins and dienes.

The main objective is to reverse engineer polymerization processes, allowing both the optimization of production routes and the development of materials with controlled properties.

The advancement of this line directly contributes to the acceleration of polyolefin design, reducing the dependence on time-consuming experimental steps and favoring the creation of materials tailored to different applications. In addition, it reinforces the potential of artificial intelligence in the area of polymers, bringing laboratory modeling closer to the real demands of the chemical and materials industry.

This line of research aims to develop methodologies based on Physics-Informed Neural Networks (PINNs) for the modeling of copolymerization processes. The proposal consists of integrating microstructure properties and experimental data to previously established kinetic models, in order to reproduce key process variables, such as molar mass distribution (MWD), Chemical Composition Distribution (CCD), conversion profiles and numerical and weight mean molar masses (Mn and Mw) over the reaction time.

The methodological differential is to use the consolidated kinetic model of at least one of the homopolymers as an initial physical restriction in the training of the network. This strategy allows you to guide the learning of the PINN, reduce the need for large amounts of experimental data, and increase robustness in the face of uncertainties.

It is, therefore, a general approach that can be adapted to different copolymer systems, and is not restricted to a specific class of materials. The development of this line contributes to the advancement of hybrid modeling of polymerization processes, favoring phenomenological understanding and paving the way for the targeted development of new polymeric materials of industrial interest, such as elastomers and special rubbers.

Project funded by FAPERJ (Public Notice No. 21/2024 – Support Program for Young Researchers in Rio de Janeiro). The objective is to investigate the sustainable production of biopolymers synthesized from monomers from renewable sources, such as crude glycerol, a byproduct of biodiesel production. The integration of green chemistry and biotechnology principles will allow the conversion of this low-value waste into high value-added biopolymers, contributing to the reduction of dependence on fossil fuels and promoting greater sustainability in the Brazilian biofuel industry.

The project employs machine learning models to optimize the polymerization process, predicting the experimental conditions needed to produce biopolymers with specific properties and reducing the need for extensive testing. In addition, a Life Cycle Assessment (LCA) will be carried out to ensure alignment with sustainability objectives, while simulations in Aspen Plus® and Python will be used to assess the economic and environmental feasibility of the process.

This research seeks to promote the development of high-performance biopolymers, providing essential technical data for future large-scale production in biodiesel biorefineries.

This project involves the modeling and simulation of co-pyrolysis processes of waste tires and biomass, using the Aspen Plus software in a design that seeks energy self-sufficiency through the recycling of the gas produced. The steps include the development of a kinetic model capable of predicting the composition of the products, the simulation of the process, the automation of the economic evaluation via integration with Python, as well as the analysis of the effects of operational parameters on the minimum selling price of pyrolysis oil.

This line contributes to the development of sustainable technological routes for waste recovery, combining energy efficiency, economic viability and reuse of materials that are difficult to dispose of.

This research proposes an integrated and sustainable process for the production of 1,3-butadiene from ethanol, using bifunctional heterogeneous zirconium-based catalysts supported in mesoporous matrices. The central objective is to overcome the technical and environmental limitations of conventional naphtha-based petrochemical routes, characterized by high energy consumption, high CO₂ emissions and dependence on fossil resources.

The investigated route involves a single conversion step, more efficient and aligned with the principles of sustainability. The methodology integrates process simulation in Aspen HYSYS,® preliminary technical-economic evaluation and life cycle analysis (ISO 14040/14044 standards). The simulations point to ethanol conversions greater than 90%, selectivity to 1,3-butadiene of up to 85% and product purity greater than 97.9%, meeting the requirements of the ASTM D2593-19 standard.

As one of the results of this line of research, the following publication stands out:

» FERREIRA, P. H. G.; DE CARVALHO, R. B.; BRANDÃO, A. L. T. Modelagem, simulação e avaliação econômica preliminar de uma planta de produção de 1,3-butadieno a partir do bioetanol. Brazilian Journal of Development, v. 7, p. 84526-84546, 2021.

This line of research seeks the development of alternative and sustainable routes for the production of light olefins, integrating kinetic modeling, process simulation, energy recovery strategies and technical-economic and environmental evaluation. The objective is to investigate processes that reconcile industrial competitiveness, energy efficiency and carbon emission mitigation, contributing to the energy transition and the decarbonization of the chemical sector.

Within the scope of this line, the study of CO₂-assisted oxidative dehydrogenation of propane (ODPC) stands out, which involved the simulation of a plant with energy recovery sections for cogeneration and carbon capture, in addition to the evaluation of different scenarios to encourage CO₂ capture.

As one of the results of this line of research, the following publication stands out:

» ESPINOSA, G. V.; BRANDÃO, A. L. T. Economic and sustainability evaluation of green CO₂-assisted propane dehydrogenation design. Digital Chemical Engineering, v. 14, p. 100203, 2025.

This line of research aims to combine artificial intelligence and machine learning techniques with methods of techno-economic evaluation of chemical and biochemical processes, in order to accelerate cost analysis, reduce uncertainties, and support decision-making. The proposed approach allows exploring the use of hybrid modeling, integrating simulation data, experimental information and automated scenario generation routines.

As an example of application, the methodology has already been explored in batch manufacturing studies of active pharmaceutical ingredients (API), in which machine learning models were integrated with process simulations to predict production costs, estimate the minimum selling price, and identify uncertainties and critical cost drivers.

This line of research contributes to the more agile and robust development of industrial processes, supporting innovation strategies and production efficiency.

This is a line of research already consolidated in GAAP, which over the last few years has resulted in relevant international publications on mathematical modeling of polymerizations using different mathematical approaches, such as the Monte Carlo stochastic method, the Instantaneous Distribution model and the Method of Moments. With the simulation of these models, it is possible to predict microstructural properties of polymers, such as the Chemical Composition Distribution (CCD) and the Chain Length Distribution (CLD); in the case of Monte Carlo, it is also possible to obtain the Intramolecular Sequence Distribution of Comonomers.

Among the results of this line of research, the following publications stand out:

» REGO, A. S. C.; BRANDÃO, A. L. T. Parameter Estimation and Kinetic Monte Carlo Simulation of Styrene and n-Butyl Acrylate Copolymerization through ATRP. Industrial & Engineering Chemistry Research, v. 60, p. 8396-8408, 2021.

» REGO, A. S. C.; AMARAL, A. M.; BRANDÃO, A. L. T. Monte Carlo simulation of terpolymerization: Optimizing the simulation and post-processing times. Canadian Journal of Chemical Engineering, v. 1, p. 1, 2023.

Currently, within this line of research, the copolymerization of acrylamide with acrylic acid is being modeled, continuing the development and application of these methodologies.

This line of research covers three main axes:

• Production and Simulation: modeling of green hydrogen production by water electrolysis, using renewable sources such as solar and wind. The process is optimized through simulations in software such as Aspen Plus, covering everything from energy capture to hydrogen purification and compression;
• Digital Twins and Operational Optimization: application of digital twins as virtual replicas of real plants, enabling real-time monitoring, performance forecasting, and operational optimization, especially in scenarios of high variability of renewable sources. When there is no access to real data, detailed modeling and simulation can act as an alternative to the digital twin, ensuring that performance and optimization analyses remain feasible;
• Technical-Economic Evaluation: calculation of the Levelized Cost of Hydrogen (LCoH), considering CAPEX, OPEX and REPEX, in addition to sensitivity analyses for critical variables such as the price of renewable energy in Brazil.

The development of this line of research seeks to provide technical and methodological tools for the implementation of more sustainable, efficient and economically viable green hydrogen projects, strengthening Brazil's insertion in the global energy transition.

This line of research investigates the potential for valuing glycerol, a low-value by-product of the Brazilian biodiesel industry, as a raw material for the production of higher value-added chemicals. The approach combines process simulation, green chemistry principles, biotechnology, and technical-economic and environmental analyses, seeking routes that reconcile sustainability, industrial competitiveness, and integration in biorefinery chains.

Among the projects conducted in this line, the study of the production of succinic acid from glycerol, using Yarrowia lipolytica as a natural producer, stands out. Different scenarios were evaluated, including isolated production from crude glycerol, production from purified glycerol, and integration with biodiesel plants. The results pointed to reductions of up to 47% in the climate impact when using raw glycerol compared to pure glycerol, in addition to economic viability in all scenarios, with a payback of less than seven years. The integration with biodiesel plants showed gains in competitiveness, but also an increase in CO₂ emissions, highlighting the need for energy optimization and solvent recovery.

As one of the results of this line of research, the following publication stands out:

» ORDÓÑEZ, D. A. R.; STRUNCK, F. J. B. T. L.; DUTRA, L. S.; BRANDÃO, A. L. T. Upcycling glycerol into succinic acid: sustainable integration with biodiesel mills. Bioresource Technology, v. 433, p. 132716, 2025.

The valorization of glycerol represents a key strategy for the circular economy and the development of sustainable biorefineries, and new chemical routes from this input continue to be investigated in this line of research.