RESEARCH ON THE IMPACT OF OPTIMIZATION CRITERIA ON THE EFFICIENCY OF EXTREME CONTROL SYSTEMS

Abstract

This paper addresses the pressing problem of enhancing the efficiency and reliability of automatic control systems (ACS) for industrial equipment operating under changing technical and operational conditions. It is substantiated that phenomena such as wear, overload, or changes in raw material properties cause alterations in the equipment’s acoustic characteristics (sound). To monitor these changes, the use of acoustic non-destructive testing (NDT) methods is proposed as a relatively inexpensive and flexible tool capable of detecting anomalous operating modes in real-time. The paper proposes an approach to ACS adaptation based on extremal control. In this structure, the acoustic NDT unit acts as a detector that identifies deviations from the nominal mode and initiates an optimization algorithm to retune the control system to the new conditions. Since the key factor determining the system’s behavior during such adaptation is the chosen optimization criterion, the aim of the work is to investigate the influence of this criterion on the key performance indicators of the transient process (response speed, accuracy, stability) of the extremal system. The research is focused on formulating recommendations for selecting a criterion based on specific production goals (e.g., minimizing energy consumption, maximizing productivity, minimum stabilization time). The paper conducts a comparative analysis of the extremal control system’s performance using various integral and specific optimization criteria. Classical integral criteria are examined, such as the Integral of Squared Error (ISE), which heavily penalizes large deviations, and the Integral of Time-weighted Absolute Error (ITAE), which is sensitive to long-lasting residual errors. One of the “strictest” criteria, the Integral of Time-weighted Squared Error (ITSE), which combines penalties for both the amplitude and duration of the error, is also considered. Additionally, the statistical criterion of Mean Squared Error (MSE) and two targeted approaches are analyzed: a criterion focused solely on error minimization (Error-only), and an energy-efficient criterion (Energy-aware), which balances accuracy against the “cost” of the control action (energy consumption, wear). The study was conducted by analyzing graphs of transient processes (changes in error, control input, and system output) after a programmed anomaly detection. The research results establish that the choice of optimization criterion fundamentally affects the system’s recovery dynamics. The ITSE and ITAE criteria demonstrate the best performance in terms of error suppression and the fastest convergence to a new steady state (passing the ±2% threshold). However, this is achieved at the cost of significant oscillations in the control input, which may be undesirable for actuators. The ErrorOnly and MSE criteria react more aggressively, accompanied by a higher initial overshoot of the output variable, but this is justified when accuracy is the sole priority and energy costs are not critical. In contrast, the EnergyAware criterion provides the smoothest control dynamics with minimal peaks, thereby reducing energy consumption and equipment wear; however, it is characterized by a much slower convergence to the new mode and a potential slight deviation from the reference steady state.