The paper presents a comparative analysis of various machine learning algorithms for the classification of epoxy composites reinforced with basalt fiber and modified with inorganic fillers. The classification is based on key thermophysical characteristics, in particular, the mass fraction of the filler, temperature, and thermal conductivity coefficient. A dataset of 16,056 interpolated samples was used to train and evaluate more than a dozen models. Among the tested algorithms, the MLP neural network model showed the highest accuracy of 99.7% and balanced classification metrics F1-measure and G-Mean. Ensemble methods, including XGBoost, CatBoost, ExtraTrees, and HistGradientBoosting, also showed high classification accuracy. To interpret the results of the MLP model, SHAP analysis was applied, which confirmed the predominant influence of the mass fraction of the filler on decision-making for all classes. The results of the study confirm the high effectiveness of machine learning methods for recognizing filler type in composite materials, as well as the potential of interpretable AI in materials science tasks.

Drying of pellets is an important stage of production, as it ensures their quality, long-term storage and proper culinary properties. At the same time, the products can be produced in large quantities and stored without refrigeration. Nowadays, various types of dryers are used in the production of pellets, including convection, infrared, vacuum. As well as we have to achieving a certain level of moisture in the finished product, it is also important to ensure the quality of the product. After all, too fast or uneven drying can cause internal stresses in the dough, which can lead to the appears of cracks and increased brittleness. High temperatures or uneven drying can affect the protein structure and change the taste characteristics of the products. The theoretical definition of drying kinetics includes many interrelated processes. It has been studied that the drying process of pellets is similar to the drying process of pasta, and is well described by models created to describe the drying of pasta. Achieving quality indicators of the final product is possible only if drying is controlled and corrected throughout the entire technological process. Typically, control systems are built in such a way that drying modes are set based on predetermined conditions of the drying environment, namely the required air humidity and temperature in the drying process, and the air flow rate. These parameters are formed based on previous experience for each specific type of product. This approach can be called predictive temperature and humidity control. In this case, it is expected that the properties of the raw material remain unchanged from batch to batch, or that the discrepancy is within acceptable limits, which allows maintaining the same modes for different batches of products of the same type. However, even small changes in the structure or shape of the product can cause significant deviations in the drying rate. This article considers a control system for convection drying of pellets, in which the drying process is carried out along a predetermined trajectory of product humidity change. Due to the nonlinearity of the system and the long duration in time, ensuring the required trajectory requires not only a response to changes, but also a forecast of the process depending on the change in parameters. Such a forecast can be made on the basis of a mathematical model. Model forecasting allows to adjust the drying parameters depending on the actual properties of the product. In order to provide feedback, real-time product weighing was used. Using forecasting based on a mathematical model and introducing additional feedback loops into the classical drying control system allowed automatically adjust the operating modes of the drying oven and ensure a given drying speed by adjusting the characteristics of the drying environment.