Fault Analysis of Electromechanical Systems using Information Entropy Concepts
MetadataShow full item record
Fault analysis of mechanical and electromechanical systems has been a subject of considerable interest in the systems and control research community. Entropy, under its various formulations is an important variable, which is unrivaled when it comes to measuring order (or organization) and/or disorder (or disorganization). Researchers have successfully used entropy based concepts to solve various challenging problems in engineering, mathematics, meteorology, biotechnology, medicine, statistics etc. This research tries to analyze faults in electromechanical systems using information entropy concepts. The objectives of this research are to develop a method to evaluate signal entropy of a dynamical system using only input/output measurements, and to use this entropy measure to analyze faults within a dynamical system. Given discrete-time signals corresponding to the three-phase voltages and currents of an electromechanical system being monitored, the problem is to analyze whether or not this system is healthy. The concepts of Shannon entropy and relative entropy come from the field of Information Theory. They measure the degree of uncertainty that exists in a system. The main idea behind this approach is that the system's dynamics may have regularities hidden in measurements that are not obvious to see. The Shannon entropy and relative entropy measures are calculated by using probability distribution functions (PDF) that are formed by sampling the time series currents and voltages of a system. The system's health is monitored by, first, sampling the currents and voltages at certain time intervals, then generating the corresponding PDFs and, finally, calculating the information entropy measures. If the system dynamics are unchanged, or in other words, the system continues to be healthy, then the relative entropy measures will be consistently low or constant. But, if the system dynamics change due to damage, then the corresponding relative entropy and Shannon entropy measures will be increasing compared to the entropy of the system with less damage.
Tangirala, Ravindra Krishna (2011). Fault Analysis of Electromechanical Systems using Information Entropy Concepts. Master's thesis, Texas A & M University. Available electronically from