Behavioral Pattern Identification for Structural Health Monitoring in Complex Systems
![open_access](/assets/open_access_icon-bc813276d7282c52345af89ac81c71bae160e2ab623e35c5c41385a25c92c3b1.png)
Open Access
- Author:
- GUPTA, SHALABH
- Graduate Program:
- Mechanical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 20, 2006
- Committee Members:
- Asok Ray, Committee Chair/Co-Chair
Anil Kamalakant Kulkarni, Committee Member
Donald Albert Koss, Committee Member
Jeffrey Scott Mayer, Committee Member
Eric Keller, Committee Member - Keywords:
- Life Extending Control
Flexible Mechanical Couplings
Pulse Combustors
Stochastic Behavior
Life Estimation
Thermodynamic Formalism
Complex Systems
Symbolic Dynamics
Signal Processing
Pattern Identification
Structural Health Monitoring
Fatigue Damage
Anomaly Detection
Fault Diagnosis and Prognosis - Abstract:
- Estimation of structural damage and quantification of structural integrity are critical for safe and reliable operation of human-engineered complex systems, such as electromechanical, thermofluid, and petrochemical systems. Damage due to fatigue crack is one of the most commonly encountered sources of structural degradation in mechanical systems. Early detection of fatigue damage is essential because the resulting structural degradation could potentially cause catastrophic failures, leading to loss of expensive equipment and human life. Therefore, for reliable operation and enhanced availability, it is necessary to develop capabilities for prognosis and estimation of impending failures, such as the onset of wide-spread fatigue crack damage in mechanical structures. This dissertation presents information-based online sensing of fatigue damage using the analytical tools of symbolic time series analysis (STSA). Anomaly detection using STSA is a pattern recognition method that has been recently developed based upon a fixed-structure, fixed-order Markov chain. The analysis procedure is built upon the principles of Symbolic Dynamics, Information Theory and Statistical Pattern Recognition. The dissertation demonstrates real-time fatigue damage monitoring based on time series data of ultrasonic signals. Statistical pattern changes are measured using STSA to monitor the evolution of fatigue damage. Real-time anomaly detection is presented as a solution to the forward (analysis) problem and the inverse (synthesis) problem. 1) the forward problem - The primary objective of the forward problem is identification of the statistical changes in the time series data of ultrasonic signals due to gradual evolution of fatigue damage. 2) the inverse problem - The objective of the inverse problem is to infer the anomalies from the observed time series data in real time based on the statistical information generated during the forward problem. A computer-controlled special-purpose fatigue test apparatus, equipped with multiple sensing devices (e.g., ultrasonics and optical microscope) for damage analysis, has been used to experimentally validate the STSA method for early detection of anomalous behavior. The sensor information is integrated with a software module consisting of the STSA algorithm for real-time monitoring of fatigue damage. Experiments have been conducted under different loading conditions on specimens constructed from the ductile aluminium alloy 7075-T6. The dissertation has also investigated the application of the STSA method for early detection of anomalies in other engineering disciplines. Two primary applications include combustion instability in a generic thermal pulse combustor model and whirling phenomenon in a typical misaligned shaft.