In the past decades, numerous efforts have been placed on the development of prognostic algorithms as well as overall design of Prognostic and Health Management (PHM) systems. This is in an effort to achieve increased safety, reliability and sustainability for engineering systems today. However, it has been observed that a lot of research efforts in the PHM community have concentrated on applying similar methodologies and algorithms in solving similar problems. There is lack of a systematic approach for prognostic modeling designs and inadequate practical instructions on identifying the critical system elements, designing experiment plans, setting research goals and determining the right PHM tools for general applications, ranging from components to machines to large-scale processes and systems. In addition, there is a gap in PHM literature today that focuses on the practical implementation of prognostic solutions and such existing studies are still in the research phase.
In this thesis, the major focus is to provide a systematic reconfigurable platform to accelerate the Prognostics and Health Management (PHM) process through simplification of PHM design, modeling, and evaluation. Several prognostic frameworks found in the literature were compared and discussed, and an enhanced practical framework for PHM applications has been introduced with comprehensive detail. In addition, options were presented for hardware and software deployment within the reconfigurable platform. National Instruments (NI) hardware and LabVIEW software toolkit was finally selected to develop the proposed reconfiguration prognostic platform, given their great compatibility and comprehensive functions of data collection, synchronization, management, analysis and result reporting.
Due to the lack of a PHM toolkit in NI LabVIEW, this thesis develops a LabVIEW-based toolkit consisting of several key PHM algorithms for health assessment, diagnosis, and prediction. Specifically, the integrated toolkit includes most frequently-used algorithms such as Logistic Regression, Self-Organizing Maps, Neural Network, Support Vector Machine and others. To facilitate the use of the toolkit, some integrated (express VI) functions and graphical user interface were also developed to accelerate the process of signal processing, sensor fusion, health diagnosis and prognosis with minimum analyst manipulation and less external knowledge input. The prognostic platform can be easily configured and integrated into the enterprise asset management system to deliver information for maintaining optimal performance of the system.
The performance and application of the developed prognostic platform is validated for real-world industrial case studies in motor, shaft and bearing systems. The platform was successfully applied for different objectives and application requirements, which demonstrates the toolkit’s reconfigurable scheme and effectiveness for solving real-world problems.