International Scientific Journals
of Scientific Technical Union of Mechanical Engineering "Industry 4.0"

Data fusion is the process of integrating multiple heterogeneous data sources to produce more accurate, comprehensive, and useful information than any single source alone. For the medical data fusion, this information may come from imaging, genomic data, clinical records, and physiological signals. It has emerged as a cornerstone of modern precision medicine. Information theory provides a rigorous mathematical framework for quantifying uncertainty, measuring information gain, and can thus be used to optimize fusion strategies across diverse clinical contexts. This paper presents a review of information-theoretic approaches suitable for medical data fusion, covering: fundamental concepts (entropy, mutual information, Kullback-Leibler divergence, rate-distortion), theoretical frameworks (information bottleneck, transfer entropy, partial information decomposition), and their application across major clinical domains. We synthesize recent advances in fusion-based architectures, discuss the critical challenges of uncertainty quantification, and provide practical guidelines for implementing information-theoretic fusion in clinical settings. Through a systematic analysis, we identify key challenges, including parameter sensitivity, missing modalities, and clinical interpretability, to outline promising directions for future research. This review aims to provide clinicians, researchers, and developers with a comprehensive understanding of how information theory can transform medical data for improved diagnostic accuracy, prognostic precision, and personalized patient care.

This study evaluates the performance of newly designed deep-learning model—bidirectional long short-term memory network (Bi-LSTM), with baseline to a conventional multilayer perceptron (MLP)—for classification faults of rolling-element bearings from raw vibration signals. The models are benchmarked against a previously optimised one-dimensional convolutional neural network (1-D CNN), originally obtained via Bayesian hyperparameter search. A carefully selected dataset of 3600 one-second segments was captured under varying speed conditions and dynamically enhanced with Gaussian noise during processing. On the test set, the Bi-LSTM achieves 100 % accuracy, the 1-D CNN 97.9 %, and the MLP 53.3 %. Training dynamics, confusion patterns, and model complexity were thoroughly analysed, highlighting the trade-offs between accuracy, latency and deployment cost in edge-computing scenarios.

FEM Simulation is widely used in engineering practice. The use in small and medium-sized companies is partially limited due to the high workload and the time required for the simulation calculations. In recent years, the use of Artificial Intelligence (AI) has been increasingly adopted, emerging as an exciting and promising area of research. This article presents a methodology for the implementation of artificial intelligence in the simulation process of a part and an assembly. This methodology includes phases to integrate AI into the CAD model preparation process, as well as the definition of contact conditions, fixtured reactions, and external forces. Artificial intelligence can process a large volume of previous calculations, allowing it to analyse and automate these preparation steps AND thus increase the accuracy of simulations.

This article examines the use of computer vision and deep learning to automatically determine key quality indicators of apples, enhancing product quality. It describes a digital method for measuring apple size, ripeness, and variety classification using an automated optoelectronic system, achieving an accuracy of at least 86%. Advantages, limitations, and potential productivity benefits for Kazakhstan’s apple production are discussed. An algorithm developed with OpenCV in Python analyzes apple images to determine diameter, height, surface area, red color proportion, and external defects. Tested on “Sinap Almaty” apples, the method measures linear dimensions, crosssectional area, and redness percentage.

Current paper presents development of system intended to classify human facial and body emotions. It is based on two deep learning neural networks (DNN): – first one used for facial emotion recognition (FER) and second one for body gesture emotion recognition (BER). Combination of the results obtained by the two modalities (facial expression data and body gestures language data) provides more accurate results instead of these obtained using only one modality. After brief analysis of the available pre-trained DNN and datasets for facial and body emotions recognition, based on previous authors’ developments, the selection of two DNN models has been done. They are used in the development and verification of present system.

Over the last few years, Deep learning has begun to play an important role in analytics solutions of big data. Deep learning is one of the most active research fields in machine learning community. It has gained unprecedented achievements in fields such as computer vision, natural language processing and speech recognition. The ability of deep learning to extract high-level complex abstractions and data examples, especially unsupervised data from large volume data, makes it attractive a valuable tool for big data analytics. In this paper, we review the deep learning architectures which can be used for big data processing. Next, we focus on the analysis and discussions about the challenges and possible solutions of deep learning for big data analytics. Finally, have been outlined several open issues and research trends.

The paper deals with implementation of artificial intelligence method for diagnostics of technological machines. The deep learning as a method of AI seems to be a very good candidate for solving complex problem of technical diagnostics. The method is now implemented for diagnostics for concrete production enterprise.