Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.
Non-WSU users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Access Type
WSU Access
Date of Award
January 2023
Degree Type
Dissertation
Degree Name
Ph.D.
Department
Industrial and Manufacturing Engineering
First Advisor
Sara Masoud
Abstract
Oil refineries and thermal power plants normally utilize massive boilers where fuel is burned to generate heat and steam. Five-fifths of all problems in these facilities can be traced back to boilers [155]. Steam leaking from cracks in either pipes or equipment is the main problem in boilers. This inspires the industry to implement solutions such as condition-based maintenance to reduce the chances of unseen failures in boiler systems. System condition monitoring is becoming more powerful yet challenging with the advancement of sensor technology. The research on condition monitoring has been mainly targeted toward predicting machinery health conditions for the purpose of preventative maintenance, especially for manufacturing processes.Maintenance is a compensatory tool for addressing systems unreliability which grows over time due to systems’ degradation. Although some research has been done to address degradation, further efforts are needed to define it as a non-stationary process [156]. It is especially hard to find the degradation from the sensor data in systems with deterioration-alleviation or degradation-aware control, as the controllers mitigate the deviation in the process measurement and keep the system running within the optimal limits [14]. To close this gap in degradation sensing and maintenance scheduling, a condition-based maintenance framework for degradation-aware control systems is proposed, which is built upon two main objectives: degradation-independent factor-based models to estimate degradation over time and Long Short-Term Memory Autoencoder - Degradation Stage Detector (LSTMA-DSD) to define alarm and failure thresholds for condition-based maintenance. Degradation dependent factor-based modelling: Two degradation dependent factor-based models are proposed here, considering different scenarios. These models are based on different operating modes (i.e., manual and automated) for supervisory control. The first model, factor-based input model, feeds on actuator measurement in addition to the supervisory control system data, while the second model, factor-based output model, reads sensor measurement of the system’s output along with information regarding optimal system setup. LSTMA-DSD: Once the degradation behaviour is estimated using the degradation dependent factor-based modelling, an LSTMA-DSD model is developed to detect different stages of the degradation behaviour for the system. This step is necessary to provide adequate time for addressing the maintenance needs. To this end, two thresholds are defined: alarm threshold and failure threshold. The alarm threshold is automatically set via the LSTMA-DSD algorithm once anomalies are detected. Then, the failure threshold is built upon the alarm threshold by considering the stage at which the degradation has been detected and the historical mean failure threshold.
Recommended Citation
Alsaedi, Faisal Ghazi, "Condition Based Maintenance For Degradation-Aware Control System" (2023). Wayne State University Dissertations. 3814.
https://digitalcommons.wayne.edu/oa_dissertations/3814