Downloaded from https://www.everyspec.com ADS-79D-HDBK experiment are (slightly) different due to the inherent system uncertainty. The exponential function is used here to give a smooth monotonically decreasing curve. The value of 0 D decreases as i D increases, and it is 1 when i D = 0, and approaches 0 when i D approaches infinity. The accuracy is the highest when the predicted value is the same as the actual value, and decreases when the predicted value deviates from the actual value. The exponential function also has higher decreasing rate when i D is closer to 0, which gives higher measurement sensitivity when ) ( i t pf is around ) ( i t af as in normal scenarios.
The measurement sensitivity is very low when the predicted value deviates too much from the actual value. Figure D-1 illustrates the fault evolution and the prognosis, the actual and predicted failure times, and the prognostic accuracy. i D e FIGURE D-1. Schematic of prognostic accuracy Three evolution curves split from the prediction time labeled t p , which represents the time the RUL was calculated, and show 3 possible evolutions of the fault dimension. There is actually a wide range of possible failure evolutions, with a statistical distribution around the actual time to failure, labeled t af as shown along the horizontal axis.
The prognostic accuracy calculation is the highest (one) when the predicted failure time is equal to the actual failure time. Note that ―failure‖ as defined for prognostics is not limited to the material failure of the item affected by the fault. Failure can be a limit imposed by engineering analysis that prevents catastrophic damage or cascading failures that affect safety or repair cost. Failure can also be defined as failure to satisfy required functionality or performance. For legacy aircraft, development of a CI can be the result of an emergent requirement, which has been identified by such actions as Accident Investigations or operational experience.
In this case, the analysis and development of the CI may be pressed for time and resources. The process of defining the fault mode of interest, the sensor and sensing strategy, algorithm development, CI validation and verification, and Army wide implementation will be a dynamic and tailored process. In some cases, abbreviating the steps associated with CI development may be necessary to meet time constraints. 76