Inferential Sensing Techniques to Enable Condition Based Maintenance 2009-01-2912

Inferential sensing, as it relates to the equipment operator, can be viewed as human intuition [1]. The person operating the equipment can sense there is something wrong while their intuition tells them when and what needs troubleshooting and repair. Attempts have been made to implement this human intuition model to monitor a vehicle operation and detect abnormalities. In many approaches traditional sensors are added to the vehicle which increases cost, complexity, and another failure point. After years of developments and techniques, there are few highly reliable on-board systems that can duplicate the human intuition model since the specific failure cannot be directly measured but must be inferred from a variety of symptoms. This paper describes an engineering approach using Physics of Failure (PoF) for specific subsystems, developing the applicable fatigue models, and then collecting, monitoring, and manipulating the real-time on-vehicle data to complement the “operator intuition”.

Conditioned Based Maintenance (CBM) delivers value through cost avoidance of unexpected down time and reduction in life cycle maintenance costs by scheduling repairs on the correct equipment subsystems at the correct times. CBM Solutions can also provide Condition Monitoring, Maintenance Scheduling, Service & Warranty Cost Reduction and Design Life Optimization. These values are delivered by improving the understanding of the severity of usage for equipment operations. The basic elements to achieve the aforementioned benefits will be reviewed in the context of process steps so the reader's perspective is aligned with the sequence of development activity needed to deliver value through this PoF and fatigue modeling approach.

An on-board electronics hardware and software architecture that can deliver CBM Solutions by using inferential sensing methodologies in a low cost platform is outlined. Through real-time statistical data techniques several months of data can be collected for several subsystems on a platform that is no more complex than a typical Electronic Control Unit.