Page 474 - Compression Machinery for Oil and Gas

454 SECTION III Applications

First, the DAQ system must be capable of reading the sensor’s signal type;
matching whether the sensor has a current-, voltage-, or resistance-based output
signal. One must also consider the range of the signal coming from the sensor.
Strain gauges and RTDs are both essentially resistance-type signals, but can
have drastically different amplitudes. Therefore, either the DAQ must be cho-
sen appropriately or additional conditioning equipment must be used.
Next, the measuring rate of the DAQ must be able to exceed the frequency
range of interest. The typically used sampling rate is the Nyquist frequency,
which strictly speaking is 2 the frequency of interest, but it is more common
in practice to use 2.3–2.5 the frequency of interest. This may, however, be
driven by what time resolution is desired.
Finally, the DAQ system chosen needs to have adequate processing capabil-
ity for the testing that will be performed. For example, if the frequency content
of the signal is desired in real time, the DAQ should be capable of performing
spectral analysis. If all processing will be handled at a later time by another
means, then recording of the data is the only necessity.

Uncertainty and Error Sources
All measured data has some level of error. This error can come from the sensor,
the transmission of the signal from the sensor to the DAQ, or the measurement
of the signal by the DAQ. The error can also be worsened by certain environ-
mental conditions. For example, high-power electrical lines running near
unshielded data lines can introduce offsets and noise in the measurement signal
because of inductance.
Following guidelines and accepted practices can help to reduce the errors.
Such practices include the regular calibration of sensors, using shielded and
properly grounded cabling, and proper sampling/averaging techniques.
Still, some level of error from a measurement is to be expected, and is
defined as its uncertainty. The uncertainty can affect the accuracy or the preci-
sion of the measurement. Accuracy errors, also sometimes called bias errors, are
usually a result of drift in the instrument or an effect of installation faults. They
are handled best by performing calibrations and following proper installation
techniques. Precision errors are usually an effect of environment and installa-
tion faults, and can be reduced by following proper installation practices and
using averaging in the sampling process.
Often, the accuracy and precision errors may not be distinguishable, so a
single uncertainty for a sensor will be defined. Some typically acceptable uncer-
tainty ranges for measurements taken in performance testing are presented in
Table 12.1.
In addition, the individual uncertainties for each sensor and DAQ measure-
ment will all be combined to form the total measurement uncertainty. Determin-
ing the effect on the overall outcome from all of the individual uncertainties is
the purpose of an uncertainty analysis.