In recent years, digital twins have gained importance in both science and industry. Digitization and artificial intelligence offer potential for more efficient production and competitive advantages for the rubber industry. In rubber production , the main raw materials are either synthetic or natural rubbers, which may have different chemical compositions (e.g. protein content in natural rubber) and, even in the case of synthetic rubber, different distribution of molar mass in the polymers between different batches. These variations can result in decreased product quality, which could have been prevented by proper adaptation of the processing steps. In order to address this challenge, we performed resource quality control as well as online quality control after extrusion.
The overarching goal is the fusion of various optical sensor data and correlation with macroscopic properties (e.g., viscosity) as well as properties of the final products (e.g., elasticity and durability) by means of an artificial intelligence (AI). Subsequently, the resulting AI will be used to adapt the processing steps adequately in order to obtain a consistent product quality.
Here, we performed resource quality control on various rubbers by application of three optical methods: photometry, Raman- and Infrared-spectroscopy. From the acquired data, we determined the minimal necessary data processing steps to develop chemometric models based on a sensor data fusion on the feature level, which allows to classify raw rubber samples into their grades.
Due to their dark color and thus high absorption, the intermediates of rubber production usually are not accessible to the aforementioned spectroscopic methods. Thus, we applied laser induced breakdown spectroscopy for online analysis. With this element-sensitive technique, we observed the distribution quality of the crosslinking chemical zinc oxide in the extrudate on a moving conveyor belt.