Downloaded from https://www.everyspec.com ADS-79D-HDBK Collecting data under conditions of greatest stationarity minimizes the effects of loads variations on the quality of the signal. If the CI for a component requires conditions of high torque or a range of torque levels, this may affect the algorithm‘s ability to meet performance metrics related to false positive rate, detectability, and accuracy. E.3.2.2 Data acquisition interval. At a minimum, at least one data set should be acquired for all monitored components for flights of 30 minutes or longer. This data should be acquired under stabilized conditions without the need for pilot action during the flight.
21 In addition, some components, such as high speed rotating parts, may experience a rapid onset of failure, on the order of a few hours. Data for these components should be acquired at frequent enough intervals to allow for fault detection and warning with preventative actions prior to the component‘s failure. E.3.2.3 Analog to digital conversion. Range: The analog-to-digital converter (ADC) should be chosen to provide sufficient range for capturing the expected excursion in signal level without clipping. Clipping or compressing the input signal amplitude induces an artificial modulation into the measured data that can mask or alter the desired feature signature.
E.3.2.4 Resolution (Dynamic Range): The resolution of the ADC should be sufficient to detect the smallest change in the signal required by the corresponding vibration diagnostic algorithm in the presence of large amplitude background. Resolution is the smallest change in a signal that can be detected and accurately indicated. It is usually expressed as a percentage of the maximum range of the instrument. E.3.2.5 Sampling rate. To avoid aliasing of the sampled signal, the minimum sampling frequency ( s ) should be at least twice as high as the highest frequency of interest ( 1 ) in the signal.
To preclude the influence of signal content above frequencies of interest, a prefilter should be used ahead of the sampler to modify the frequency content of the signal before it is sampled so that the frequency spectrum for s 2 1 is negligible. 22 Signal aliasing is the result of higher frequencies being folded into lower frequency signals due to the sampling rate being too low. While the minimum sampling rate is required to be twice as high as the highest frequency component present in the signal, this represents the theoretical minimum required to reconstruct the continuous signal from the sampled data. In practice, the sampling frequency is frequently chosen to be 10 1 to 20 1 .
E.3.2.6 Data windowing. Digital processing is performed on a ―window‖ of measured data that is often extracted from a continuously occurring event. Windows applied to data to prevent leakage error should be defined in the system performance specification. Processing of a finite record length of data inherently induces a distortion, called leakage, which can perturb the feature signature and reduce the detected signal-to-noise ratio. Care should be taken in selecting a proper amplitude taper (window) to reduce these effects. Applying no window at all is to 21 CAP 753. ―Helicopter Vibration Health Monitoring:
Guidance Material for Operators Utilizing VHM in Rotor and Rotor Drive Systems of Helicopters.‖ UK Civil Aviation Authority, Safety Regulation Group. June 2006. 22 Ogata, K., ―Discrete-Time Control Systems,‖ Prentice Hall, Englewood Cliffs, NJ, 1987, pp. 170-177. 104