‘Prediction of the Transient Response Sensitivity of High Order Linear Systems Using Low Order Models’

It is shown that low order models can be used to predict adequately the transient response sensitivity of high order linear systems, thus leading to a great reduction in time and complexity during analogue or digital computation, as well as facilitating the understanding of the sensitivity functions themselves. Experimental results on an analogue model of a sixth order aircraft blind landing system confirm the adequacy of the procedure. In this instance the use of the low order model eliminates six integrators from the analysis and it is to be expected that higher order systems will lead to even greater savings. The example used to illustrate the procedure is the practical problem of setting parameter tolerances given acceptable bounds of transient performance.

A new low order model is introduced, and when compared with alternative models proposed elsewhere in the literature is found to give an encouraging accuracy of prediction. The new model is particularly useful if the sensitivity analysis is undertaken on an analogue computer.