Kaplan, Sargent and Goodman 
When considering the utility of system identification tech- 
niques for analyzing data from full scale maneuvers or from towing 
tank tests for naval craft, there are various ways in which it could 
be most efficiently used for such analyses. This is especially true in 
the case of model testing where the ability to constrain motions 
enables selected coefficients to be sought independently from the 
others. The system identification program has the ability to identify 
simpler constrained maneuvers first and then incorporate the resul- 
tant or otherwise known coefficients into the more complete model 
when a complex maneuver is analyzed. This might also relax the 
indicated measuring requirements as less demand would be placed on 
the system identification program, In the case of full scale sea trials, 
where the motions are naturally unconstrained, system identification 
techniques offera useful method of directly analyzing ship motions 
when seeking knowledge of a large number of unknown coefficients 
simultaneously. 
Another possible use of system identification for surface 
ship problems is an application to the case of a ship in a restricted 
waterway, such asa canal, when applied to model testing. Various 
static force and moment derivatives and related hydrodynamic data 
can be obtained from captive model tests in a towing tank with spe- 
cially configured restrictions simulating the canal. However the im- 
portant dynamic derivatives due to angular velocity, and angular 
velocity effects combined with lateral velocity and forward velocity, 
cannot be obtained with ease or without serious questions as to data 
validity (for oscillator experiments) with ordinary towing tank test 
techniques. In that case the use of system identification applied to 
trajectory data from model experiments would allow determination of 
basic stability derivatives by that method, when normal test proce- 
dures have basic limitations. Thus it serves as an adjunct to model 
testing that would allow more complete determination of pertinent 
parameters, thereby resulting in more reliable prediction of full 
scale ship performance, 
The major problem exhibited in the application of this method 
is demonstrated when noise is artificially added to the observed data, 
as illustrated in the case of the surface ship. However, the low level 
noise associated with analog computer output data, for a lower order 
equation system, did not seem to influence the results. Similarly the 
full scale data of the aircraft analyzed in [5] also contained some 
noise in the records, and the influence of this noise was reduced by 
means of simple smoothing operations applied to each data point 
(obtaining average values in terms of data points on either side ofa 
particular data point at each instant of time). While the generated 
1646 
