A sensitivity analysis of a nulling network used on an inductance type lubricating film thickness probe
This thesis investigates the application of classical sensitivity to the drift problems encountered in inductive film thickness probes used on dynamic seals. Of primary interest were the sensitivities of the measurement system to component values of the nulling network.
The sensitivities of the circuit model were found in two different ways. The first method was to derive the sensitivity functions from the transfer function by taking derivatives. The second method was to use a numerical technique to compute the values of the sensitivities. The sensitivity functions were evaluated and compared to the ones found by numerical techniques. The sensitivities of the null network components was measured by making small changes to these parameters and observing the output change. Techniques for finding probe model parameters by fitting techniques were developed. Computer programs were written to support model parameter estimation and sensitivity evaluation.
The derived sensitivity functions were found to yield values consistent with those found by numerical techniques. Measured values of sensitivity were significantly different from the calculated values. The fitting techniques used to find model component values give reasonable results.
Derivation of sensitivity functions for circuits of this complexity was found not to be useful enough to justify the effort required to arrive at them. Numerical techniques can yield as much information and require much less effort to arrive at the values. Numerical techniques are significantly less prone to error. Although the sensitivity functions work well with the model used, evidence indicates that a more complex model must be used to account for the deviations from the measured values. The definition of classical sensitivity does not lend itself well to circuits which can have zero output. A simple modification to this definition will yield a significant improvement in the insight derived from the resulting sensitivity values.
Thesis78.N475.pdf
4.25 MB
Adobe PDF
e1d2ef408af260adeffbc37047621985