6926475916

ory buffer with capacity for N samples (or N words if one sample is taken to occupy a word). The maxi-mum number of spectral lines avaiłabłe in FFT results in this case is N/2. Unfortunately. due to aliasing distortion [1j the last several spectral lines (close to the maximum frequency) are not usable. The fraction of usable spectral lines determines the resolution of FFT results. Typically. with an input memory buffer having the capacity for 1024 sampled data points, about 400 usable spectral lines (theactual number of spectral lines being 512) can be generated. Zoom analysis (4) is used to improve spectral resolution. Zoom analysis uses a powerful coordinate transfor-mation procedurę in FFT to virtually eliminate the aliasing distortion in a specified band of frequencies. In effect, the same resolution (512 lines) is obtained even at the high frequer»cy end of the spectrum. Statistical error band analysis (2) generates confi-dence intervals for spectral estimates such as fre-quency response functions, spectral densities, and coherence functions. Nonę of the systems considered provides this capability. The generał feeling is that such analyses are somewhat artificial in that an assumption is needed regarding the probability distribution of the measurement error.

The hidden-line capability for modal display is not found in typical modal analysis systems. The consensus is that this provision will require a significant effort of sophisticated programming (resulting in increased overall cost). The complexity arises when the object is viewed from variable directions in perspective; the hidden linę configuration then varies and has to be continuously taken into account in the graphics software. The task becomes morę com-plicated in animation. System B provides a color monitor that might be helpful in view contrast.

System D provides the capability to generate the structural mass and stiffness matrices. We have noted that, in order to make this computation possible, it is necessary to make the number of modes analy2ed equal to the number of degrees of freedom in the test object. For example, for a 300 d o.f. system it will be necessary to analyze all 300 modes in order to generate the 300 x 300 mass matrix and a stiffness matrix of the same size. This computational effort can often become futile for the following reasons:

a. estimated mass and stiffness matrices are sen sit i ve to measurement noise

b.    analysis of a large number of modes is an exces$ive burden on Computer resources

c.    Iow-coherent data are given the same signifi-cance as the morę accurate data in these com-putations

The Table summarizes some basie data on the capa-bilities of the four systems studied. It is possible to identify a few relative merits and drawbacks. System C is compact and easy to operate even if the user has a marginal knowledge of theory and programming. A compatible eight-pen digital plotter can be used under system control. The cartridge drive is a slow (sequential) data storage device. but the rela-tively large memory in the Computer (256 K bytes) offsets this deficiency. System D inherits the experi-ence gained through its predecessor. The system has a faster and larger (30 M bytes) Winchester hard disk storage. Also. its analyzer provides several signal-dependent weighting Windows. System B is the least expensive in a linę of modal analysis systems marketed by the same manufacturer. A color monitor provided as an option with these systems is an attrac-tive feature. System A provides an extensive data storage capability and speed. It carries a wealth of past experience in the area of digital signal analysis instrumentation. Several other options usually go in hand with modal analysis. These include the structural modification option that considers the sensitivity of modal parameters to changes in mass. stiffness. or damping in the system and the sub-structure linking option that in effect develops an overall model from test data obtained for two (or morę) subsystems of the complete system. Only the modal analysis capability has been addressed in this paper.

CONCLUSIONS

Experimental modal analysis is being extensively used in industry to evaluate the dynamie performance of Products during their design development. quality control. and qualification. Hardware and software selection for a stand-alone modal analysis system depends on the requirements of the specific application. A thorough understanding of the func-tional operation of the system and underlying theory can considerably aid this selection process.