This paper relates to a recent project in which the aim was to devise a procedure that could be used to identify the linearised transfer functions of the longitudinal motion of the solar powered airship Lotte from flight test data.

Using a non-linear mathematical model of the Airship, a linearised state space model was derived numerically and this was used to derive the response transfer functions, which formed the theoretical results. A test procedure was then developed to identify the transfer functions experimentally from frequency response data obtained in simulated flight tests. Initially, the linearised model was used and the test procedure was refined until a suitable level of coherence between the experimental and theoretical results was observed. The test procedure was then applied to simulated flight test data from the non-linear model, and finally sensor noise was added in order to represent actual flight test conditions as closely as possible. Unfortunately, the results were inaccurate and in some cases, they were unacceptable. When the test procedure was applied to the non-linear model, significant errors were seen in all the results. The addition of sensor noise led to heavily distorted results and a further degradation of the resulting transfer function estimates.

It was concluded that meaningful results could not be achieved unless frequency response data with a higher resolution could be obtained and the effects of sensor noise could be reduced. Presently, no test procedure has been developed that would be of any use in a practical situation. However, a number of ways of possibly improving the test procedure or otherwise improving the quality of the results were conceived near the end of the project. It is therefore the purpose of this paper to account for the shortcomings that were encountered in the project and to introduce briefly the unexplored remedies, in the hope that these may be of some benefit to future identification studies.