Methods and Algorithms for Decision-Making Based on Expert System Recommendations Using Artificial Intelligence Technologies

Abstract

This article examines methodological approaches to developing control actions in the mining industry through the synthesis of expert systems and artificial intelligence technologies. The study aims to bridge the gap between the accumulation of process data arrays and the speed of their interpretation for operational control. The author substantiates the need to transition from classical algorithms to hybrid intelligent models that combine expert systems with AI. The scientific novelty of this study lies in the development of a mathematical decision-making model based on a hybrid intelligent model that integrates data on the state of an object with a knowledge base and utility functions. The proposed system algorithm ensures consistent primary wavelet signal filtering, classification of process situations, and logical inference of recommendations. The simulation results demonstrated the advantage of a hybrid model for ensuring industrial safety and improving the operational efficiency of mining operations. Analysis of the obtained data confirms that the hybrid approach minimizes the influence of subjective factors and prevents critical equipment failures. Implementation of the algorithms will help automate the identification of pre-emergency conditions, ensuring high reliability of decision-making, even in the presence of noisy input signals. A comparison of the time characteristics indicates a five-hundredfold acceleration of analytical operations, which is critical for stabilizing production flow. A conclusion is drawn regarding the potential of the developed mathematical software as a reliable tool for building self-organizing expert systems using artificial intelligence technologies. The proposed solutions will help personnel reallocate time resources toward strategic planning, ensuring the stability of enterprise performance in the long term. A prerequisite for the successful applicability of the model to real-world enterprises is the availability of a digital precedent archive for the initial configuration of neural network scales. The technological limitations of the proposed approach are related to the computing power of local servers when processing extremely large data streams in hard real-time.