Alexander Mitsos

Jan G. Rittig, Karim Ben Hicham, Artur M. Schweidtmann et al.

Ionic liquids (ILs) are important solvents for sustainable processes and predicting activity coefficients (ACs) of solutes in ILs is needed. Recently, matrix completion methods (MCMs), transformers, and graph neural networks (GNNs) have shown high accuracy in predicting ACs of binary mixtures, superior to well-established models, e.g., COSMO-RS and UNIFAC. GNNs are particularly promising here as they learn a molecular graph-to-property relationship without pretraining, typically required for transformers, and are, unlike MCMs, applicable to molecules not included in training. For ILs, however, GNN applications are currently missing. Herein, we present a GNN to predict temperature-dependent infinite dilution ACs of solutes in ILs. We train the GNN on a database including more than 40,000 AC values and compare it to a state-of-the-art MCM. The GNN and MCM achieve similar high prediction performance, with the GNN additionally enabling high-quality predictions for ACs of solutions that contain ILs and solutes not considered during training.

Artur M. Schweidtmann, Jan G. Rittig, Jana M. Weber et al.

Graph neural networks (GNNs) are emerging in chemical engineering for the end-to-end learning of physicochemical properties based on molecular graphs. A key element of GNNs is the pooling function which combines atom feature vectors into molecular fingerprints. Most previous works use a standard pooling function to predict a variety of properties. However, unsuitable pooling functions can lead to unphysical GNNs that poorly generalize. We compare and select meaningful GNN pooling methods based on physical knowledge about the learned properties. The impact of physical pooling functions is demonstrated with molecular properties calculated from quantum mechanical computations. We also compare our results to the recent set2set pooling approach. We recommend using sum pooling for the prediction of properties that depend on molecular size and compare pooling functions for properties that are molecular size-independent. Overall, we show that the use of physical pooling functions significantly enhances generalization.

Eike Cramer, Dirk Witthaut, Alexander Mitsos et al.

Electricity is traded on various markets with different time horizons and regulations. Short-term intraday trading becomes increasingly important due to the higher penetration of renewables. In Germany, the intraday electricity price typically fluctuates around the day-ahead price of the European Power EXchange (EPEX) spot markets in a distinct hourly pattern. This work proposes a probabilistic modeling approach that models the intraday price difference to the day-ahead contracts. The model captures the emerging hourly pattern by considering the four 15 min intervals in each day-ahead price interval as a four-dimensional joint probability distribution. The resulting nontrivial, multivariate price difference distribution is learned using a normalizing flow, i.e., a deep generative model that combines conditional multivariate density estimation and probabilistic regression. Furthermore, this work discusses the influence of different external impact factors based on literature insights and impact analysis using explainable artificial intelligence (XAI). The normalizing flow is compared to an informed selection of historical data and probabilistic forecasts using a Gaussian copula and a Gaussian regression model. Among the different models, the normalizing flow identifies the trends with the highest accuracy and has the narrowest prediction intervals. Both the XAI analysis and the empirical experiments highlight that the immediate history of the price difference realization and the increments of the day-ahead price have the most substantial impact on the price difference.

Jan G. Rittig, Martin Ritzert, Artur M. Schweidtmann et al.

Fuels with high-knock resistance enable modern spark-ignition engines to achieve high efficiency and thus low CO2 emissions. Identification of molecules with desired autoignition properties indicated by a high research octane number and a high octane sensitivity is therefore of great practical relevance and can be supported by computer-aided molecular design (CAMD). Recent developments in the field of graph machine learning (graph-ML) provide novel, promising tools for CAMD. We propose a modular graph-ML CAMD framework that integrates generative graph-ML models with graph neural networks and optimization, enabling the design of molecules with desired ignition properties in a continuous molecular space. In particular, we explore the potential of Bayesian optimization and genetic algorithms in combination with generative graph-ML models. The graph-ML CAMD framework successfully identifies well-established high-octane components. It also suggests new candidates, one of which we experimentally investigate and use to illustrate the need for further auto-ignition training data.

Marah Almanasreh, Alexander Mitsos, Eike Cramer

Accurate prediction of polymerization dynamics is essential for process design, control, and optimization. Yet, purely mechanistic models require labor-intensive parameterization of partially characterized kinetics, while purely data-driven models demand large, diverse datasets that are costly to obtain, particularly in early-design stages. We propose a hybrid Neural Ordinary Differential Equation (NODE) framework for data-efficient modeling of free-radical polymerization. Using batch polymerization of methyl methacrylate (MMA) as a case study, the mechanistic mass balances are retained explicitly, and only the partially-characterized effective radical concentration governing monomer consumption is learned from data through a neural network surrogate, while established reactions such as initiator decomposition, propagation, and termination remain physically modeled. The hybrid NODE is evaluated against a discrete-time feedforward neural network and a purely data-driven NODE under sparse data conditions, with models trained on as few as ten measurements under both regular and irregular sampling. The hybrid NODE consistently achieves lower prediction errors and more physically consistent extrapolations than both purely data-driven baselines. In a generalization scenario with noisy data and unseen operating conditions, the hybrid NODE achieves an RMSE of 0.013, compared to 0.31 for the data-driven NODE and 0.68 for the discrete-time model, demonstrating that learning only a closure term rather than the full dynamics is sufficient for reliable prediction under limited data availability.

Danimir T. Doncevic, Alexander Mitsos, Yue Guo et al.

Meta-learning of numerical algorithms for a given task consists of the data-driven identification and adaptation of an algorithmic structure and the associated hyperparameters. To limit the complexity of the meta-learning problem, neural architectures with a certain inductive bias towards favorable algorithmic structures can, and should, be used. We generalize our previously introduced Runge-Kutta neural network to a recursively recurrent neural network (R2N2) superstructure for the design of customized iterative algorithms. In contrast to off-the-shelf deep learning approaches, it features a distinct division into modules for generation of information and for the subsequent assembly of this information towards a solution. Local information in the form of a subspace is generated by subordinate, inner, iterations of recurrent function evaluations starting at the current outer iterate. The update to the next outer iterate is computed as a linear combination of these evaluations, reducing the residual in this space, and constitutes the output of the network. We demonstrate that regular training of the weight parameters inside the proposed superstructure on input/output data of various computational problem classes yields iterations similar to Krylov solvers for linear equation systems, Newton-Krylov solvers for nonlinear equation systems, and Runge-Kutta integrators for ordinary differential equations. Due to its modularity, the superstructure can be readily extended with functionalities needed to represent more general classes of iterative algorithms traditionally based on Taylor series expansions.

Jan G. Rittig, Qinghe Gao, Manuel Dahmen et al.

Molecular property prediction is of crucial importance in many disciplines such as drug discovery, molecular biology, or material and process design. The frequently employed quantitative structure-property/activity relationships (QSPRs/QSARs) characterize molecules by descriptors which are then mapped to the properties of interest via a linear or nonlinear model. In contrast, graph neural networks, a novel machine learning method, directly work on the molecular graph, i.e., a graph representation where atoms correspond to nodes and bonds correspond to edges. GNNs allow to learn properties in an end-to-end fashion, thereby avoiding the need for informative descriptors as in QSPRs/QSARs. GNNs have been shown to achieve state-of-the-art prediction performance on various property predictions tasks and represent an active field of research. We describe the fundamentals of GNNs and demonstrate the application of GNNs via two examples for molecular property prediction.

Eike Cramer, Leonard Paeleke, Alexander Mitsos et al.

We present a specialized scenario generation method that utilizes forecast information to generate scenarios for day-ahead scheduling problems. In particular, we use normalizing flows to generate wind power scenarios by sampling from a conditional distribution that uses wind speed forecasts to tailor the scenarios to a specific day. We apply the generated scenarios in a stochastic day-ahead bidding problem of a wind electricity producer and analyze whether the scenarios yield profitable decisions. Compared to Gaussian copulas and Wasserstein-generative adversarial networks, the normalizing flow successfully narrows the range of scenarios around the daily trends while maintaining a diverse variety of possible realizations. In the stochastic day-ahead bidding problem, the conditional scenarios from all methods lead to significantly more stable profitable results compared to an unconditional selection of historical scenarios. The normalizing flow consistently obtains the highest profits, even for small sets scenarios.

Eike Cramer, Felix Rauh, Alexander Mitsos et al.

To model manifold data using normalizing flows, we employ isometric autoencoders to design embeddings with explicit inverses that do not distort the probability distribution. Using isometries separates manifold learning and density estimation and enables training of both parts to high accuracy. Thus, model selection and tuning are simplified compared to existing injective normalizing flows. Applied to data sets on (approximately) flat manifolds, the combined approach generates high-quality data.

Daniel Mayfrank, Alexander Mitsos, Manuel Dahmen

(Economic) nonlinear model predictive control ((e)NMPC) requires dynamic models that are sufficiently accurate and computationally tractable. Data-driven surrogate models for mechanistic models can reduce the computational burden of (e)NMPC; however, such models are typically trained by system identification for maximum prediction accuracy on simulation samples and perform suboptimally in (e)NMPC. We present a method for end-to-end reinforcement learning of Koopman surrogate models for optimal performance as part of (e)NMPC. We apply our method to two applications derived from an established nonlinear continuous stirred-tank reactor model. The controller performance is compared to that of (e)NMPCs utilizing models trained using system identification, and model-free neural network controllers trained using reinforcement learning. We show that the end-to-end trained models outperform those trained using system identification in (e)NMPC, and that, in contrast to the neural network controllers, the (e)NMPC controllers can react to changes in the control setting without retraining.

Jan G. Rittig, Alexander Mitsos

We propose excess Gibbs free energy graph neural networks (GE-GNNs) for predicting composition-dependent activity coefficients of binary mixtures. The GE-GNN architecture ensures thermodynamic consistency by predicting the molar excess Gibbs free energy and using thermodynamic relations to obtain activity coefficients. As these are differential, automatic differentiation is applied to learn the activity coefficients in an end-to-end manner. Since the architecture is based on fundamental thermodynamics, we do not require additional loss terms to learn thermodynamic consistency. As the output is a fundamental property, we neither impose thermodynamic modeling limitations and assumptions. We demonstrate high accuracy and thermodynamic consistency of the activity coefficient predictions.

Karim K. Ben Hicham, Jan G. Rittig, Martin Grohe et al.

Accurate molecular property prediction is central to drug discovery, catalysis, and process design, yet real-world applications are often limited by small datasets. Molecular foundation models provide a promising direction by learning transferable molecular representations; however, they typically involve task-specific fine-tuning, require machine learning expertise, and often fail to outperform classical baselines. Tabular foundation models (TFMs) offer a fundamentally different paradigm: they perform predictions through in-context learning, enabling inference without task-specific training. Here, we evaluate TFMs in the low- to medium-data regime across both standardized pharmaceutical benchmarks and chemical engineering datasets. We evaluate both frozen molecular foundation model representations, as well as classical descriptors and fingerprints. Across the benchmarks, the approach shows excellent predictive performance while reducing computational cost, compared to fine-tuning, with these advantages also transferring to practical engineering data settings. In particular, combining TFMs with CheMeleon embeddings yields up to 100\% win rates on 30 MoleculeACE tasks, while compact RDKit2d and Mordred descriptors provide strong descriptor-based alternatives. Molecular representation emerges as a key determinant in TFM performance, with molecular foundation model embeddings and 2D descriptor sets both providing substantial gains over classic molecular fingerprints on many tasks. These results suggest that in-context learning with TFMs provides a highly accurate and cost-efficient alternative for property prediction in practical applications.

Mehmet Velioglu, Song Zhai, Alexander Mitsos et al.

Separating liquid-liquid dispersions in gravity settlers is critical in chemical, pharmaceutical, and recycling processes. The dense-packed zone height is an important performance and safety indicator but it is often expensive and impractical to measure due to optical limitations. We propose to estimate phase heights using only inexpensive volume flow measurements. To this end, a physics-informed neural network (PINN) is first pretrained on synthetic data and physics equations derived from a low-fidelity (approximate) mechanistic model to reduce the need for extensive experimental data. While the mechanistic model is used to generate synthetic training data, only volume balance equations are used in the PINN, since the integration of submodels describing droplet coalescence and sedimentation into the PINN would be computationally prohibitive. The pretrained PINN is then fine-tuned with scarce experimental data to capture the actual dynamics of the separator. We then employ the differentiable PINN as a predictive model in an Extended Kalman Filter inspired state estimation framework, enabling the phase heights to be tracked and updated from flow-rate measurements. We first test the two-stage trained PINN by forward simulation from a known initial state against the mechanistic model and a non-pretrained PINN. We then evaluate phase height estimation performance with the filter, comparing the two-stage trained PINN with a two-stage trained purely data-driven neural network. All model types are trained and evaluated using ensembles to account for model parameter uncertainty. In all evaluations, the two-stage trained PINN yields the most accurate phase-height estimates.

Moritz Wedemeyer, Eike Cramer, Alexander Mitsos et al.

With the increasing flexibilization of processes, determining robust scheduling decisions has become an important goal. Traditionally, the flexibility index has been used to identify safe operating schedules by approximating the admissible uncertainty region using simple admissible uncertainty sets, such as hypercubes. Presently, available contextual information, such as forecasts, has not been considered to define the admissible uncertainty set when determining the flexibility index. We propose the conditional flexibility index (CFI), which extends the traditional flexibility index in two ways: by learning the parametrized admissible uncertainty set from historical data and by using contextual information to make the admissible uncertainty set conditional. This is achieved using a normalizing flow that learns a bijective mapping from a Gaussian base distribution to the data distribution. The admissible latent uncertainty set is constructed as a hypersphere in the latent space and mapped to the data space. By incorporating contextual information, the CFI provides a more informative estimate of flexibility by defining admissible uncertainty sets in regions that are more likely to be relevant under given conditions. Using an illustrative example, we show that no general statement can be made about data-driven admissible uncertainty sets outperforming simple sets, or conditional sets outperforming unconditional ones. However, both data-driven and conditional admissible uncertainty sets ensure that only regions of the uncertain parameter space containing realizations are considered. We apply the CFI to a security-constrained unit commitment example and demonstrate that the CFI can improve scheduling quality by incorporating temporal information.

Daniel Mayfrank, Kayra Dernek, Laura Lang et al.

With our recently proposed method based on reinforcement learning (Mayfrank et al. (2024), Comput. Chem. Eng. 190), Koopman surrogate models can be trained for optimal performance in specific (economic) nonlinear model predictive control ((e)NMPC) applications. So far, our method has exclusively been demonstrated on a small-scale case study. Herein, we show that our method scales well to a more challenging demand response case study built on a large-scale model of a single-product (nitrogen) air separation unit. Across all numerical experiments, we assume observability of only a few realistically measurable plant variables. Compared to a purely system identification-based Koopman eNMPC, which generates small economic savings but frequently violates constraints, our method delivers similar economic performance while avoiding constraint violations.

Karim K. Ben Hicham, Moreno Ascani, Jan G. Rittig et al.

Accurate prediction of phase equilibria remains a central challenge in chemical engineering. Physics-consistent machine learning methods that incorporate thermodynamic structure into neural networks have recently shown strong performance for activity-coefficient modeling. However, extending such approaches to equilibrium data arising from an extremum principle, such as liquid-liquid equilibria, remains difficult. Here we present DISCOMAX, a differentiable algorithm for phase-equilibrium calculation that guarantees thermodynamic consistency at both training and inference, only subject to a user-specified discretization. The method is rooted in statistical thermodynamics, and works via a discrete enumeration with subsequent masked softmax aggregation of feasible states, and together with a straight-through gradient estimator to enable physics-consistent end-to-end learning of neural $g^{E}$-models. We evaluate the approach on binary liquid-liquid equilibrium data and demonstrate that it outperforms existing surrogate-based methods, while offering a general framework for learning from different kinds of equilibrium data.

Jan C. Schulze, Danimir T. Doncevic, Nils Erwes et al.

Achieving real-time capability is an essential prerequisite for the industrial implementation of nonlinear model predictive control (NMPC). Data-driven model reduction offers a way to obtain low-order control models from complex digital twins. In particular, data-driven approaches require little expert knowledge of the particular process and its model, and provide reduced models of a well-defined generic structure. Herein, we apply our recently proposed data-driven reduction strategy based on Koopman theory [Schulze et al. (2022), Comput. Chem. Eng.] to generate a low-order control model of an air separation unit (ASU). The reduced Koopman model combines autoencoders and linear latent dynamics and is constructed using machine learning. Further, we present an NMPC implementation that uses derivative computation tailored to the fixed block structure of reduced Koopman models. Our reduction approach with tailored NMPC implementation enables real-time NMPC of an ASU at an average CPU time decrease by 98 %.

Muhammad bin Javaid, Hasham Hussain, Ashima Khanna et al.

Molecular design encompasses tasks ranging from de-novo design to structural alteration of given molecules or fragments. For the latter, state-of-the-art methods predominantly function as "Instance Optimizers'', expending significant compute restarting the search for every input structure. While model-based approaches theoretically offer amortized efficiency by learning a policy transferable to unseen structures, existing methods struggle to generalize. We identify a key failure mode: the high variance arising from the heterogeneous difficulty of distinct starting structures. To address this, we introduce GRXForm, adapting a pre-trained Graph Transformer model that optimizes molecules via sequential atom-and-bond additions. We employ Group Relative Policy Optimization (GRPO) for goal-directed fine-tuning to mitigate variance by normalizing rewards relative to the starting structure. Empirically, GRXForm generalizes to out-of-distribution molecular scaffolds without inference-time oracle calls or refinement, achieving scores in multi-objective optimization competitive with leading instance optimizers.

Eike Cramer, Luis Kutschat, Oliver Stollenwerk et al.

Bayesian optimization (BO) has gained attention as an efficient algorithm for black-box optimization of expensive-to-evaluate systems, where the BO algorithm iteratively queries the system and suggests new trials based on a probabilistic model fitted to previous samples. Still, the standard BO loop may require a prohibitively large number of experiments to converge to the optimum, especially for high-dimensional and nonlinear systems. We present a hybrid model-based BO formulation that combines the iterative Bayesian learning of BO with partially known mechanistic physical models. Instead of learning a direct mapping from inputs to the objective, we write all known equations for a physics-based model and infer expressions for variables missing equations using a probabilistic model, in our case, a Gaussian process (GP). The final formulation then includes the GP as a constraint in the hybrid model, thereby allowing other physics-based nonlinear and implicit model constraints. This hybrid model formulation yields a constrained, nonlinear stochastic program, which we discretize using the sample-average approximation. In an in-silico optimization of a single-stage distillation, the hybrid BO model based on mass conservation laws yields significantly better designs than a standard BO loop. Furthermore, the hybrid model converges in as few as one iteration, depending on the initial samples, whereas, the standard BO does not converge within 25 for any of the seeds. Overall, the proposed hybrid BO scheme presents a promising optimization method for partially known systems, leveraging the strengths of both mechanistic modeling and data-driven optimization.

Artur M. Schweidtmann, Dominik Bongartz, Daniel Grothe et al.

Gaussian processes~(Kriging) are interpolating data-driven models that are frequently applied in various disciplines. Often, Gaussian processes are trained on datasets and are subsequently embedded as surrogate models in optimization problems. These optimization problems are nonconvex and global optimization is desired. However, previous literature observed computational burdens limiting deterministic global optimization to Gaussian processes trained on few data points. We propose a reduced-space formulation for deterministic global optimization with trained Gaussian processes embedded. For optimization, the branch-and-bound solver branches only on the degrees of freedom and McCormick relaxations are propagated through explicit Gaussian process models. The approach also leads to significantly smaller and computationally cheaper subproblems for lower and upper bounding. To further accelerate convergence, we derive envelopes of common covariance functions for GPs and tight relaxations of acquisition functions used in Bayesian optimization including expected improvement, probability of improvement, and lower confidence bound. In total, we reduce computational time by orders of magnitude compared to state-of-the-art methods, thus overcoming previous computational burdens. We demonstrate the performance and scaling of the proposed method and apply it to Bayesian optimization with global optimization of the acquisition function and chance-constrained programming. The Gaussian process models, acquisition functions, and training scripts are available open-source within the "MeLOn - Machine Learning Models for Optimization" toolbox~(https://git.rwth-aachen.de/avt.svt/public/MeLOn).

Jan C. Schulze, Alexander Mitsos

We use Koopman theory for data-driven model reduction of nonlinear dynamical systems with controls. We propose generic model structures combining delay-coordinate encoding of measurements and full-state decoding to integrate reduced Koopman modeling and state estimation. We present a deep-learning approach to train the proposed models. A case study demonstrates that our approach provides accurate control models and enables real-time capable nonlinear model predictive control of a high-purity cryogenic distillation column.

Christoforos Brozos, Jan G. Rittig, Sandip Bhattacharya et al.

Surfactants are of high importance in different industrial sectors such as cosmetics, detergents, oil recovery and drug delivery systems. Therefore, many quantitative structure-property relationship (QSPR) models have been developed for surfactants. Each predictive model typically focuses on one surfactant class, mostly nonionics. Graph Neural Networks (GNNs) have exhibited a great predictive performance for property prediction of ionic liquids, polymers and drugs in general. Specifically for surfactants, GNNs can successfully predict critical micelle concentration (CMC), a key surfactant property associated with micellization. A key factor in the predictive ability of QSPR and GNN models is the data available for training. Based on extensive literature search, we create the largest available CMC database with 429 molecules and the first large data collection for surface excess concentration ($Γ$$_{m}$), another surfactant property associated with foaming, with 164 molecules. Then, we develop GNN models to predict the CMC and $Γ$$_{m}$ and we explore different learning approaches, i.e., single- and multi-task learning, as well as different training strategies, namely ensemble and transfer learning. We find that a multi-task GNN with ensemble learning trained on all $Γ$$_{m}$ and CMC data performs best. Finally, we test the ability of our CMC model to generalize on industrial grade pure component surfactants. The GNN yields highly accurate predictions for CMC, showing great potential for future industrial applications.

Christoforos Brozos, Jan G. Rittig, Sandip Bhattacharya et al.

The critical micelle concentration (CMC) of surfactant molecules is an essential property for surfactant applications in industry. Recently, classical QSPR and Graph Neural Networks (GNNs), a deep learning technique, have been successfully applied to predict the CMC of surfactants at room temperature. However, these models have not yet considered the temperature dependency of the CMC, which is highly relevant for practical applications. We herein develop a GNN model for temperature-dependent CMC prediction of surfactants. We collect about 1400 data points from public sources for all surfactant classes, i.e., ionic, nonionic, and zwitterionic, at multiple temperatures. We test the predictive quality of the model for following scenarios: i) when CMC data for surfactants are present in the training of the model in at least one different temperature, and ii) CMC data for surfactants are not present in the training, i.e., generalizing to unseen surfactants. In both test scenarios, our model exhibits a high predictive performance of R$^2 \geq $ 0.94 on test data. We also find that the model performance varies by surfactant class. Finally, we evaluate the model for sugar-based surfactants with complex molecular structures, as these represent a more sustainable alternative to synthetic surfactants and are therefore of great interest for future applications in the personal and home care industries.

Tai Xuan Tan, Alexander Mitsos, Eike Cramer

Batch processes are inherently transient and typically nonlinear, motivating nonlinear model predictive control (NMPC). However, adopting NMPC is hindered by the cost and unavailability of dynamic models. Thus, we propose to use Gaussian Processes (GP) in a model-learning NMPC scheme (GP-MLMPC) for batch processes. We initialize the GP-MLMPC using data from a single initial trajectory, e.g., from a PI controller. We iteratively apply the NMPC embedded with GPs to run batches and update the GP with new observations from each iteration, thereby achieving batch-wise improvements. Using uncertainty quantification from the GPs, we formulate chance constraints to enforce safe operation to the required confidence levels. We demonstrate our approach in \textit{silico} on a semi-batch polymerization reactor for tracking and economic objectives over durations of two hours, and the reactor temperature is constrained in a range of $\pm2^\circ C$ around its setpoint. After only four batch iterations, tracking error from the GP-MLMPC scheme converged to a reduction of $83\%$, compared to the initial trajectory. Furthermore, under an economic objective, the GP-MLMPC resulted in a 17-fold increase in final product mass by iteration 8, compared to the initial trajectory. In both cases, the resulting GP-MLMPC performance is on par with the full-model NMPC, which shows that the optimal controller can be learned by the approach. By collecting samples around the optimal trajectory, the GP-MLMPC remains sample-efficient across iterations and achieves quick convergence. Thus, the proposed GP-MLMPC scheme presents a promising data-efficient approach for the control of nonlinear batch processes without mechanistic knowledge.

Jonathan Pirnay, Jan G. Rittig, Alexander B. Wolf et al.

Generative deep learning has become pivotal in molecular design for drug discovery, materials science, and chemical engineering. A widely used paradigm is to pretrain neural networks on string representations of molecules and fine-tune them using reinforcement learning on specific objectives. However, string-based models face challenges in ensuring chemical validity and enforcing structural constraints like the presence of specific substructures. We propose to instead combine graph-based molecular representations, which can naturally ensure chemical validity, with transformer architectures, which are highly expressive and capable of modeling long-range dependencies between atoms. Our approach iteratively modifies a molecular graph by adding atoms and bonds, which ensures chemical validity and facilitates the incorporation of structural constraints. We present GraphXForm, a decoder-only graph transformer architecture, which is pretrained on existing compounds and then fine-tuned using a new training algorithm that combines elements of the deep cross-entropy method and self-improvement learning. We evaluate GraphXForm on various drug design tasks, demonstrating superior objective scores compared to state-of-the-art molecular design approaches. Furthermore, we apply GraphXForm to two solvent design tasks for liquid-liquid extraction, again outperforming alternative methods while flexibly enforcing structural constraints or initiating design from existing molecular structures.

Christoforos Brozos, Jan G. Rittig, Elie Akanny et al.

Surfactants are key ingredients in foaming and cleansing products across various industries such as personal and home care, industrial cleaning, and more, with the critical micelle concentration (CMC) being of major interest. Predictive models for CMC of pure surfactants have been developed based on recent ML methods, however, in practice surfactant mixtures are typically used due to to performance, environmental, and cost reasons. This requires accounting for synergistic/antagonistic interactions between surfactants; however, predictive ML models for a wide spectrum of mixtures are missing so far. Herein, we develop a graph neural network (GNN) framework for surfactant mixtures to predict the temperature-dependent CMC. We collect data for 108 surfactant binary mixtures, to which we add data for pure species from our previous work [Brozos et al. (2024), J. Chem. Theory Comput.]. We then develop and train GNNs and evaluate their accuracy across different prediction test scenarios for binary mixtures relevant to practical applications. The final GNN models demonstrate very high predictive performance when interpolating between different mixture compositions and for new binary mixtures with known species. Extrapolation to binary surfactant mixtures where either one or both surfactant species are not seen before, yields accurate results for the majority of surfactant systems. We further find superior accuracy of the GNN over a semi-empirical model based on activity coefficients, which has been widely used to date. We then explore if GNN models trained solely on binary mixture and pure species data can also accurately predict the CMCs of ternary mixtures. Finally, we experimentally measure the CMC of 4 commercial surfactants that contain up to four species and industrial relevant mixtures and find a very good agreement between measured and predicted CMC values.

Anastasia Georgiou, Daniel Jungen, Luise Kaven et al.

Bayesian Optimization (BO) with Gaussian Processes relies on optimizing an acquisition function to determine sampling. We investigate the advantages and disadvantages of using a deterministic global solver (MAiNGO) compared to conventional local and stochastic global solvers (L-BFGS-B and multi-start, respectively) for the optimization of the acquisition function. For CPU efficiency, we set a time limit for MAiNGO, taking the best point as optimal. We perform repeated numerical experiments, initially using the Muller-Brown potential as a benchmark function, utilizing the lower confidence bound acquisition function; we further validate our findings with three alternative benchmark functions. Statistical analysis reveals that when the acquisition function is more exploitative (as opposed to exploratory), BO with MAiNGO converges in fewer iterations than with the local solvers. However, when the dataset lacks diversity, or when the acquisition function is overly exploitative, BO with MAiNGO, compared to the local solvers, is more likely to converge to a local rather than a global ly near-optimal solution of the black-box function. L-BFGS-B and multi-start mitigate this risk in BO by introducing stochasticity in the selection of the next sampling point, which enhances the exploration of uncharted regions in the search space and reduces dependence on acquisition function hyperparameters. Ultimately, suboptimal optimization of poorly chosen acquisition functions may be preferable to their optimal solution. When the acquisition function is more exploratory, BO with MAiNGO, multi-start, and L-BFGS-B achieve comparable probabilities of convergence to a globally near-optimal solution (although BO with MAiNGO may require more iterations to converge under these conditions).

Daniel Mayfrank, Na Young Ahn, Alexander Mitsos et al.

Mechanistic dynamic process models may be too computationally expensive to be usable as part of a real-time capable predictive controller. We present a method for end-to-end learning of Koopman surrogate models for optimal performance in a specific control task. In contrast to previous contributions that employ standard reinforcement learning (RL) algorithms, we use a training algorithm that exploits the differentiability of environments based on mechanistic simulation models to aid the policy optimization. We evaluate the performance of our method by comparing it to that of other training algorithms on an existing economic nonlinear model predictive control (eNMPC) case study of a continuous stirred-tank reactor (CSTR) model. Compared to the benchmark methods, our method produces similar economic performance while eliminating constraint violations. Thus, for this case study, our method outperforms the others and offers a promising path toward more performant controllers that employ dynamic surrogate models.

Jan G. Rittig, Manuel Dahmen, Martin Grohe et al.

We present a perspective on molecular machine learning (ML) in the field of chemical process engineering. Recently, molecular ML has demonstrated great potential in (i) providing highly accurate predictions for properties of pure components and their mixtures, and (ii) exploring the chemical space for new molecular structures. We review current state-of-the-art molecular ML models and discuss research directions that promise further advancements. This includes ML methods, such as graph neural networks and transformers, which can be further advanced through the incorporation of physicochemical knowledge in a hybrid or physics-informed fashion. Then, we consider leveraging molecular ML at the chemical process scale, which is highly desirable yet rather unexplored. We discuss how molecular ML can be integrated into process design and optimization formulations, promising to accelerate the identification of novel molecules and processes. To this end, it will be essential to create molecule and process design benchmarks and practically validate proposed candidates, possibly in collaboration with the chemical industry.

Eleni D. Koronaki, Luise F. Kaven, Johannes M. M. Faust et al.

Polymer particle size constitutes a crucial characteristic of product quality in polymerization. Raman spectroscopy is an established and reliable process analytical technology for in-line concentration monitoring. Recent approaches and some theoretical considerations show a correlation between Raman signals and particle sizes but do not determine polymer size from Raman spectroscopic measurements accurately and reliably. With this in mind, we propose three alternative machine learning workflows to perform this task, all involving diffusion maps, a nonlinear manifold learning technique for dimensionality reduction: (i) directly from diffusion maps, (ii) alternating diffusion maps, and (iii) conformal autoencoder neural networks. We apply the workflows to a data set of Raman spectra with associated size measured via dynamic light scattering of 47 microgel (cross-linked polymer) samples in a diameter range of 208nm to 483 nm. The conformal autoencoders substantially outperform state-of-the-art methods and results for the first time in a promising prediction of polymer size from Raman spectra.

Jan Pavšek, Alexander Mitsos, Elvis J. Sim et al.

Machine learning (ML) approaches have shown promising results for predicting molecular properties relevant for chemical process design. However, they are often limited by scarce experimental property data and lack thermodynamic consistency. As such, thermodynamics-informed ML, i.e., incorporating thermodynamic relations into the loss function as regularization term for training, has been proposed. We herein transfer the concept of thermodynamics-informed graph neural networks (GNNs) from the Gibbs-Duhem to the Clapeyron equation, predicting several pure component properties in a multi-task manner, namely: vapor pressure, liquid molar volume, vapor molar volume and enthalpy of vaporization. We find improved prediction accuracy of the Clapeyron-GNN compared to the single-task learning setting, and improved approximation of the Clapeyron equation compared to the purely data-driven multi-task learning setting. In fact, we observe the largest improvement in prediction accuracy for the properties with the lowest availability of data, making our model promising for practical application in data scarce scenarios of chemical engineering practice.

Jan Pavšek, Alexander Mitsos, Manuel Dahmen et al.

We propose a machine learning (ML) architecture to better capture the dependency of thermodynamic properties on the independent states. When predicting state-dependent thermodynamic properties, ML models need to account for both molecular structure and the thermodynamic state, described by independent variables, typically temperature, pressure, and composition. Modern molecular ML models typically include state information by adding it to molecular fingerprint vectors or by embedding explicit (semi-empirical) thermodynamic relations. Here, we propose to rather split the information processing on the molecular structure and the dependency on states into two separate network channels: a graph neural network and a multilayer perceptron, whose output is combined by a dot product. We refer to our approach as DeepEOSNet, as this idea is based on the DeepONet architecture [Lu et al. (2021), Nat. Mach. Intell.]: instead of operators, we learn state dependencies, with the possibility to predict equation of states (EOS). We investigate the predictive performance of DeepEOSNet by means of three case studies, which include the prediction of vapor pressure as a function of temperature, and mixture molar volume as a function of composition, temperature, and pressure. Our results show superior performance of DeepEOSNet for predicting vapor pressure and comparable performance for predicting mixture molar volume compared to state-of-research graph-based thermodynamic prediction models from our earlier works. In fact, we see large potential of DeepEOSNet in cases where data is sparse in the state domain and the output function is structurally similar across different molecules. The concept of DeepEOSNet can easily be transferred to other ML architectures in molecular context, and thus provides a viable option for property prediction.

Jan C. Schulze, Alexander Mitsos

Computationally cheap yet accurate enough dynamical models are vital for real-time capable nonlinear optimization and model-based control. When given a computationally expensive high-order prediction model, a reduction to a lower-order simplified model can enable such real-time applications. Herein, we review state-of-the-art nonlinear model order reduction methods and provide a theoretical comparison of method properties. Additionally, we discuss both general-purpose methods and tailored approaches for (chemical) process systems and we identify similarities and differences between these methods. As manifold-Galerkin approaches currently do not account for inputs in the construction of the reduced state subspace, we extend these methods to dynamical systems with inputs. In a comparative case study, we apply eight established model order reduction methods to an air separation process model: POD-Galerkin, nonlinear-POD-Galerkin, manifold-Galerkin, dynamic mode decomposition, Koopman theory, manifold learning with latent predictor, compartment modeling, and model aggregation. Herein, we do not investigate hyperreduction (reduction of FLOPS). Based on our findings, we discuss strengths and weaknesses of the model order reduction methods.

Daniel Mayfrank, Mehmet Velioglu, Alexander Mitsos et al.

Reinforcement learning (RL) can be used to tune data-driven (economic) nonlinear model predictive controllers ((e)NMPCs) for optimal performance in a specific control task by optimizing the dynamic model or parameters in the policy's objective function or constraints, such as state bounds. However, the sample efficiency of RL is crucial, and to improve it, we combine a model-based RL algorithm with our published method that turns Koopman (e)NMPCs into automatically differentiable policies. We apply our approach to an eNMPC case study of a continuous stirred-tank reactor (CSTR) model from the literature. The approach outperforms benchmark methods, i.e., data-driven eNMPCs using models based on system identification without further RL tuning of the resulting policy, and neural network controllers trained with model-based RL, by achieving superior control performance and higher sample efficiency. Furthermore, utilizing partial prior knowledge about the system dynamics via physics-informed learning further increases sample efficiency.

Mehmet Velioglu, Song Zhai, Sophia Rupprecht et al.

In chemical engineering, process data are expensive to acquire, and complex phenomena are difficult to fully model. We explore the use of physics-informed neural networks (PINNs) for modeling dynamic processes with incomplete mechanistic semi-explicit differential-algebraic equation systems and scarce process data. In particular, we focus on estimating states for which neither direct observational data nor constitutive equations are available. We propose an easy-to-apply heuristic to assess whether estimation of such states may be possible. As numerical examples, we consider a continuously stirred tank reactor and a liquid-liquid separator. We find that PINNs can infer immeasurable states with reasonable accuracy, even if respective constitutive equations are unknown. We thus show that PINNs are capable of modeling processes when relatively few experimental data and only partially known mechanistic descriptions are available, and conclude that they constitute a promising avenue that warrants further investigation.

Jan G. Rittig, Kobi C. Felton, Alexei A. Lapkin et al.

We propose Gibbs-Duhem-informed neural networks for the prediction of binary activity coefficients at varying compositions. That is, we include the Gibbs-Duhem equation explicitly in the loss function for training neural networks, which is straightforward in standard machine learning (ML) frameworks enabling automatic differentiation. In contrast to recent hybrid ML approaches, our approach does not rely on embedding a specific thermodynamic model inside the neural network and corresponding prediction limitations. Rather, Gibbs-Duhem consistency serves as regularization, with the flexibility of ML models being preserved. Our results show increased thermodynamic consistency and generalization capabilities for activity coefficient predictions by Gibbs-Duhem-informed graph neural networks and matrix completion methods. We also find that the model architecture, particularly the activation function, can have a strong influence on the prediction quality. The approach can be easily extended to account for other thermodynamic consistency conditions.

Eike Cramer, Leonardo Rydin Gorjão, Alexander Mitsos et al.

The design and operation of modern energy systems are heavily influenced by time-dependent and uncertain parameters, e.g., renewable electricity generation, load-demand, and electricity prices. These are typically represented by a set of discrete realizations known as scenarios. A popular scenario generation approach uses deep generative models (DGM) that allow scenario generation without prior assumptions about the data distribution. However, the validation of generated scenarios is difficult, and a comprehensive discussion about appropriate validation methods is currently lacking. To start this discussion, we provide a critical assessment of the currently used validation methods in the energy scenario generation literature. In particular, we assess validation methods based on probability density, auto-correlation, and power spectral density. Furthermore, we propose using the multifractal detrended fluctuation analysis (MFDFA) as an additional validation method for non-trivial features like peaks, bursts, and plateaus. As representative examples, we train generative adversarial networks (GANs), Wasserstein GANs (WGANs), and variational autoencoders (VAEs) on two renewable power generation time series (photovoltaic and wind from Germany in 2013 to 2015) and an intra-day electricity price time series form the European Energy Exchange in 2017 to 2019. We apply the four validation methods to both the historical and the generated data and discuss the interpretation of validation results as well as common mistakes, pitfalls, and limitations of the validation methods. Our assessment shows that no single method sufficiently characterizes a scenario but ideally validation should include multiple methods and be interpreted carefully in the context of scenarios over short time periods.

Eike Cramer, Alexander Mitsos, Raul Tempone et al.

Neural networks-based learning of the distribution of non-dispatchable renewable electricity generation from sources such as photovoltaics (PV) and wind as well as load demands has recently gained attention. Normalizing flow density models are particularly well suited for this task due to the training through direct log-likelihood maximization. However, research from the field of image generation has shown that standard normalizing flows can only learn smeared-out versions of manifold distributions. Previous works on normalizing flow-based scenario generation do not address this issue, and the smeared-out distributions result in the sampling of noisy time series. In this paper, we exploit the isometry of the principal component analysis (PCA), which sets up the normalizing flow in a lower-dimensional space while maintaining the direct and computationally efficient likelihood maximization. We train the resulting principal component flow (PCF) on data of PV and wind power generation as well as load demand in Germany in the years 2013 to 2015. The results of this investigation show that the PCF preserves critical features of the original distributions, such as the probability density and frequency behavior of the time series. The application of the PCF is, however, not limited to renewable power generation but rather extends to any data set, time series, or otherwise, which can be efficiently reduced using PCA.

Simon Olofsson, Eduardo S. Schultz, Adel Mhamdi et al.

Diverse domains of science and engineering use parameterised mechanistic models. Engineers and scientists can often hypothesise several rival models to explain a specific process or phenomenon. Consider a model discrimination setting where we wish to find the best mechanistic, dynamic model candidate and the best model parameter estimates. Typically, several rival mechanistic models can explain the available data, so design of dynamic experiments for model discrimination helps optimally collect additional data by finding experimental settings that maximise model prediction divergence. We argue there are two main approaches in the literature for solving the optimal design problem: (i) the analytical approach, using linear and Gaussian approximations to find closed-form expressions for the design objective, and (ii) the data-driven approach, which often relies on computationally intensive Monte Carlo techniques. Olofsson et al. (ICML 35, 2018) introduced Gaussian process (GP) surrogate models to hybridise the analytical and data-driven approaches, which allowed for computationally efficient design of experiments for discriminating between black-box models. In this study, we demonstrate that we can extend existing methods for optimal design of dynamic experiments to incorporate a wider range of problem uncertainty. We also extend the Olofsson et al. (2018) method of using GP surrogate models for discriminating between dynamic black-box models. We evaluate our approach on a well-known case study from literature, and explore the consequences of using GP surrogates to approximate gradient-based methods.