Prognostic Health Management
(PHM)

D2K Modern PHM System Architecture
PHM Systems are evolving to meet higher expectations.
What should PHM Systems do?
· Determination of Health and its impact on system functions
· Monitor early warning of incipient failures
· Predictions of Remaining Useful Life
· Leveraging of advanced “reasoners”
· Signal processing for event detection
· Algorithms for event correlation and sensor fusion
· Expert Systems and rule-based architectures
· Advanced neural and statistical classifiers
· Real-time state estimators
· Model-based Reasoning
· Supervisory-level intelligence / logic
· Estimation and understanding of system state within operational context
· Decision support to assist operators in maintaining operational availability
· Optimize scheduling of maintenance and corrective actions according to the
principals of condition-based maintenance

D2K Real-time PHM System Data Flow
How have PHM Systems performed?
1. Expensive
2. Often ill specified
3. Often an afterthought - considered very late in design cycle
4. Excruciating test and validation cycles
5. Questionable performance
How should PHM Systems optimally be designed?
· PHM System Requirements need to be vetted early in the design cycle
· RCM focused Design Methodology should be followed
· Design tools should be leveraged that produce details for optimal sensor suite and location, FMECA, expected usage, maintenance actions and scheduling
· Need to analyze fault detection early in anticipation of onboard prognostics
requirements
· Need to link failures to detectable events across subsystems, and diagnosis to maintenance and corrective actions
· Onboard hardware and software architectures should compliment and support integration with PHM data