In carbon fiber (CF) production, the stabilization process step is the most energy- and time-consuming step in comparison with carbonization and graphitization. To develop optimization routes for energy and productivity, the stabilization needs to be monitored continuously via inline analysis methods. To prognose the evolution of high-performance CF, the density of stabilized fibers has been identified as a robust pre-indicator. As the offline analysis of density is not feasible for inline analysis, a density-soft sensor based on the stabilization indices of Fourier Transform Infrared spectrum (FTIR)-analysis and Electron Paramagnetic Resonance (EPR) Spectroscopy could potentially be used for inline monitoring. In this study, a Polyacrylonitrile-based precursor fiber (PF) stabilized in a continuous thermomechanical stabilization line with varying stretching profiles was incrementally analyzed using density, FTIR-based relative cyclization index (RCI), and EPR-based free radical concentration (FRC). Our findings show RCI and EPR dependencies for density, correlated for RCI with sensitivity by stretching to cubic model parameters, while FRC exhibits linear relationships. Therefore, this study identifies two possible soft sensors for inline density measurement, enabling autonomous energy optimization within industry 4.0-based process systems.

Within carbon fiber manufacturing, the stabilization process is the most time‐ and energy‐intensive process due to the complexity of chemical structure transformation. Therefore, optimization is strictly required to enable cost‐efficient stabilization processes. For the first time, a hybrid semi‐parametric model for continuous stabilization is developed to prognose stabilization progress and density of the stabilized fiber. The proposed model takes the process parameters like dwell time, stabilization temperature, and fiber stretching as well as precursor properties, such as fiber density, into account. Finally, the proposed hybrid semi‐parametric model offers a novel plant‐transferable optimization solution for model‐based energy optimization in the stabilization process.

Understanding the probabilistic distribution of the electrical properties of carbon fibers (CFs) at the filament level is crucial for the development of function‐integrative composite materials and the design of tailored CFs. While models describing the distribution of CF tensile strength are well established, the distribution of electrical resistivity remains largely unexplored. This work presents a comprehensive quantitative description of resistivity distribution at the single filament level and extends the established correlation with the tensile modulus from the scale of fiber bundles to individual filaments. The study is based on extensive experimental characterization of two commercially available CF, PAN‐based HTA40 and pitch‐based YSH‐70A‐10S. Maximum likelihood estimation (MLE) and the Akaike information criterion (AIC) are applied to evaluate the distribution of filament resistivities. The results indicate that the electrical resistivity distributions can be accurately described by either Weibull or Gamma distributions. Furthermore, the correlation between the electrical resistivity and tensile modulus of filaments is analyzed, and the influence of diameter variations on the observed distributions and correlations is discussed.

Within carbon fiber manufacturing, stabilization process is the most time- and energy-intensive process, due to the complexity of chemical structure transformation. Therefore, optimization is strictly required to enable cost-efficient stabilization processes. Within this study, for the first time a hybrid semi-parametric model for continuous stabilization is developed to prognose stabilization progress and density of the stabilized fiber. The proposed model takes the process parameters, like dwell time, stabilization temperature and fiber stretching, as well as precursor properties, such as fiber density, into account. Finally, the proposed hybrid semi-parametric model offers a novel plant transferable optimization solution for model-based energy optimization in the stabilization process.