Self-Optimizing Smart Control Engineering Enabled by Digital Twins
There are two critical questions in control engineering: how optimal, and how robust the control system is? However, digital transformation, Industry 4.0, and the advent of breaking technologies like artificial intelligence, deep learning, big data analytics, edge computing, and etc., contribute to increased system health knowledge, sensing capabilities, and automation on performance assessment metrics. For this reason, two new questions emerge: how smart and how developmental a control system could be? Therefore, a new frontier in control engineering emerges, and this dissertation defines it as smart control engineering (SCE), supported by three groundbreaking technologies digital twin (DT), industrial artificial intelligence (IAI) and self optimizing control (SOC).Thus, smart control engineering transforms classic control systems into smart control systems. It means that systems are aware of their capabilities and limitations (cognizant), able to learn from past experience to improve its future performance (reflective), supported by a substantial body of knowledge (knowledge-rich), handling high-level instructions based on human vague commands (taskable), and always adhering to social and legal norms (ethical). This thesis tries to establish the foundations of the smart control engineering framework and its combination with digital twin, industrial artificial intelligence, and self optimizing control for the development of smart control systems. A set of smart and developmental controllers supported by digital twin are developed using real-time zeroth-order optimization algorithms to enable smartness on real systems. Likewise, a set of enabling capabilities resulting from breaking technologies like smart controller design, control performance assessment, or parallel intelligence and controls are integrated into the SCE framework, powered by real-time data analytics provided by IAI methods. The embedded implementation of smart controllers with enabling capabilities is performed and demonstrated for single-input and multi-input control systems using edge computing devices. Obtained results show that smart control engineering is a new and effective framework that can systematically improve the performance, reliability, and robustness under varying internal and external conditions.