A5081019709- Analytical Chemistry and Chromatography
- Chemical and Physical Properties in Aqueous Solutions
- Crystallization and Solubility Studies
- Machine Learning in Materials Science
- Computational Drug Discovery Methods
- Phase Equilibria and Thermodynamics
- Metabolomics and Mass Spectrometry Studies
- Thermodynamic properties of mixtures
- Process Optimization and Integration
- Protein purification and stability
- Chromatography in Natural Products
- Catalysis and Oxidation Reactions
- Advanced Chemical Sensor Technologies
- NMR spectroscopy and applications
- Spectroscopy and Chemometric Analyses
- Advanced Thermodynamics and Statistical Mechanics
- Hydrocarbon exploration and reservoir analysis
- Surfactants and Colloidal Systems
- Statistical and Computational Modeling
- Calcium Carbonate Crystallization and Inhibition
- Advanced Data Processing Techniques
- Protein Interaction Studies and Fluorescence Analysis
- Software System Performance and Reliability
- Biofuel production and bioconversion
- Topological and Geometric Data Analysis
University of Kaiserslautern
2015-2025
Photonik-Zentrum Kaiserslautern
2025
Rheinland-Pfälzische Technische Universität Kaiserslautern-Landau
2023-2024
Daimler (Germany)
2023
E Ink (South Korea)
2023
Software (Spain)
2023
University of California, Irvine
2020-2022
Technical University of Munich
2021
Stadtwerke Straubing (Germany)
2021
University of California System
2020
Activity coefficients, which are a measure of the nonideality liquid mixtures, key property in chemical engineering with relevance to modeling and phase equilibria as well transport processes. Although experimental data on thousands binary mixtures available, prediction methods needed calculate activity coefficients many relevant that have not been explored date. In this report, we propose probabilistic matrix factorization model for predicting arbitrary mixtures. no physical descriptors...
Abstract Methods for predicting Henry's law constants H ij are important as experimental data scarce. We introduce a new machine learning approach such predictions: matrix completion methods (MCMs) and demonstrate its applicability using base that contains values 101 solutes i 247 solvents j at 298 K. Data on only available 2661 systems + . These stored in × matrix; the task of MCM is to predict missing entries. First, an entirely data‐driven presented. Its predictive performance, evaluated...
Although the pure component vapor pressure is one of most important properties for designing chemical processes, no broadly applicable, sufficiently accurate, and open-source prediction method has been available. To overcome this, we have developed GRAPPA - a hybrid graph neural network predicting pressures components. enables curve basically any organic molecule, requiring only molecular structure as input. The new model consists three parts: A attention message passing step, pooling...
The flexibility, affordability and ease of use benchtop 1H NMR spectroscopy makes it potentially very interesting for assessing the quality wine types monitoring fermentation process. However, low spectral resolution complexity mixtures hinder direct quantification important parameters and, thus, prevent its widespread as an analytical tool in wineries. We show here that these problems can be solved using model-based data processing. In a first step, accuracy new approach was evaluated by...
Large Language Models (LLMs) have made significant progress in reasoning, demonstrating their capability to generate human-like responses. This study analyzes the problem-solving capabilities of LLMs domain thermodynamics. A benchmark 22 thermodynamic problems evaluate is presented that contains both simple and advanced problems. Five different are assessed: GPT-3.5, GPT-4, GPT-4o from OpenAI, Llama 3.1 Meta, le Chat MistralAI. The answers these were evaluated by trained human experts,...
Accurate prediction of thermodynamic properties mixtures, such as activity coefficients, is essential for designing and optimizing chemical processes. While established physics-based methods face limitations in accuracy scope, emerging machine learning approaches, matrix completion (MCMs), offer promising alternatives. However, their performance can suffer data-sparse regions. To address this issue, we propose a novel hybrid MCM predicting coefficients at infinite dilution 298 K that not...
Activity coefficients describe the nonideality of liquid mixtures and are essential for calculating equilibria. The activity at infinite dilution in binary particularly important as finite concentrations can be predicted based on their knowledge not only but also multicomponent mixtures. available experimental data these is readily accessible databases organized a matrix with rows representing solutes columns solvents or vice versa. As lacking many mixtures, this sparsely populated. Filling...
We introduce HANNA, the first hybrid neural network model that strictly complies with all thermodynamic consistency criteria for predicting activity coefficients and outperforms current benchmark methods in terms of accuracy applicability.
We present new matrix completion methods for the prediction of binary liquid phase diffusion coefficients at infinite dilution, which are trained to a newly consolidated database in this work and outperform established semiempirical correlations.
Predictive models of thermodynamic properties mixtures are paramount in chemical engineering and chemistry. Classical successful generalizing over (continuous) conditions like temperature concentration. On the other hand, matrix completion methods (MCMs) from machine learning successfully generalize (discrete) binary systems; these MCMs can make predictions without any data for a given system by implicitly commonalities across systems. In present work, we combine strengths both worlds hybrid...
Group contribution (GC) methods are widely used for predicting the thermodynamic properties of mixtures by dividing components into structural groups. These groups can be combined freely so that applicability a GC method is only limited availability its parameters interest. For describing mixtures, pairwise interaction between prime importance. Finding suitable numbers these often impeded lack experimental data. Here, we address this problem using matrix completion (MCMs) from machine...
We present a generic way to hybridize physical and data-driven methods for predicting physicochemical properties. The approach 'distills' the method's predictions into prior model combines it with sparse experimental data using Bayesian inference. apply new predict activity coefficients at infinite dilution obtain significant improvements compared baselines established ensemble from machine learning literature.
Abstract This paper provides the first comprehensive evaluation and analysis of modern (deep‐learning‐based) unsupervised anomaly detection methods for chemical process data. We focus on Tennessee Eastman dataset, a standard litmus test to benchmark nearly three decades. Our extensive study will facilitate choosing appropriate in industrial applications. From benchmark, we conclude that reconstruction‐based are choice, followed by generative forecasting‐based methods.
RESS enables the transformation from α to β phase PVDF. Piezoresponse force microscopy confirmed piezoelectricity of obtained particles.
Mixtures that contain a known target component but are otherwise poorly specified important in many fields. Previously, the activity of component, which is needed, e.g., to design separation processes, could not be predicted such mixtures. A method was developed solve this problem. It combines thermodynamic group contribution for coefficient with NMR spectroscopy, used estimating nature and amount different chemical groups mixture. The knowledge speciation mixture required. Test cases...
Abstract Knowledge of thermodynamic properties mixtures is essential in many fields science and engineering. However, the experimental data usually scarce, so prediction methods are needed. Matrix completion have proven to be very successful predicting binary mixtures. In this approach, organized a matrix whose rows columns correspond two components, entries indicate value studied property at fixed conditions. present work, we extend concept tensor (TCMs). This allows account for variation...
In this work, we present the first method for predicting unlike binary interaction parameters in molecular simulations unstudied mixtures based on a matrix completion from machine learning.
Abstract Digitalization is about data and how they are used. This has always been a key topic in applied thermodynamics. In the present work, influence of current wave digitalization on thermodynamics analyzed. Thermodynamic modeling simulation changing as large amounts different nature quality become easily available. The power complexity thermodynamic models techniques rapidly increasing, new routes viable to link them data. Machine learning opens perspectives, when it suitably combined...
Mixtures of which the composition is not fully known are important in many fields engineering and science, for example, biotechnology. Owing to lack information on composition, such mixtures cannot be described with common thermodynamic models. In present work, a method this obstacle can overcome an class problems. The enables estimation activity coefficients target components poorly specified based combination NMR spectroscopy group contribution method. It therefore called NEAT (NMR...
Embeddings of high-dimensional data are widely used to explore data, verify analysis results, and communicate information. Their explanation, in particular with respect the input attributes, is often difficult. With linear projects like PCA axes can still be annotated meaningfully. non-linear projections this no longer possible alternative strategies such as attribute-based color coding required. In paper, we review existing augmentation techniques discuss their limitations. We present...
Predicting the physico-chemical properties of pure substances and mixtures is a central task in thermodynamics. Established prediction methods range from fully physics-based ab-initio calculations, which are only feasible for very simple systems, over descriptor-based that use some information on molecules to be modeled together with fitted model parameters (e.g., quantitative-structure-property relationship or classical group contribution methods), representation-learning methods, may,...
Nuclear magnetic resonance (NMR) spectroscopy is a powerful tool for elucidating the structure of unknown components and composition liquid mixtures. However, these tasks are often tedious challenging, especially if complex samples considered. In this work, we introduce automated methods identification quantification structural groups in pure mixtures from NMR spectra using support vector classification. As input, 1H spectrum 13C sample (pure component or mixture) that to be analyzed needed....