Comparison between SARSA and Q-Learning algorithms based Fault-Tolerant Control

Processes and systems are continually subjected to fault or malfunctions because of age or sudden events, which might degrade the operation performance and even result in operation failure that is a vital issue in the safety essential system. Therefore, it is actuated to develop a Fault-Tolerant Control strategy so that the system will operate with tolerated performance degradation. A fascinating property in Fault-Tolerant Controllers is adaptability to system changes as they evolve throughout systems operations. In this paper, we proposed the Reinforcement-based Fault-Tolerant Control (RL- based FTC) methodology while not want of the system model and also the data of fault and we compared between the utilization of Q-Learning and SARSA algorithms in Fault-Tolerant Control strategy. The effectiveness of the two algorithms is demonstrated by the Centrifugal Compressor system by the Python software. Our experiment demonstrates that the Q- learning algorithm primarily based on Fault-Tolerant Control performs better than SARSA algorithmic program based on Fault-Tolerant Control underneath sudden fault.

Keywords:

Fault-Tolerant Control (FTC) , Reinforcement Learning (RL) , Q-Learning algorithm , SARSA algorithm , Neural Network (NN)