It should be noted that various aspects could have produced 

 incorrect results in the regression analysis. For example, during 

 the digitizing operation, points were sampled at approximately equal 

 increments of time. Since the object travels faster when it first 

 enters the soils than near the point of ultimate penetration, many 

 more points were taken near the ultimate penetration point than near 

 the point of entry into the seafloor. The regression equations, 

 therefore, exaggerate the lower phases of penetration at the expense 

 of the upper phases. The region in which the deceleration relaxes 

 from its maximum point to zero (as clearly illustrated in Figures 9 

 through 12) is particularly exaggerated. This is a difficult region 

 to analyze and it probably represents a phase in which the shearing 

 resistance of the soil is not fully mobilized. An equation developed 

 to fit both this region and the earlier regions in which the full 

 shearing resistance almost certainly is mobilized has a stroiig 

 probability of being somewhat in error. 



Another problem inherent in the regression analysis is its failure 

 to separate other phases of the penetration process. For example, 

 at different penetration depths, different portions of the 

 penetrators are embedded in the soil. Considering the cone penetrator, 

 initially only the cone is embedded in the seafloor. This is 

 followed by a long period in which the shaft following the cone becomes 

 more and more deeply embedded. Finally, the housing for the accelerometer 

 strikes the seafloor and becomes embedded. In the physical analysis, 

 each of these phases is considered individually in the formulation 

 of the area integrals. In the regression analysis all of the phases 

 are masked together in one equation. The solution to this problem 

 is to consider each phase of penetration separately and to derive 

 separate regression equations for each. This should be undertaken 

 in future penetration analyses. It is likely that more consistent, 

 realistic regression coefficients will result. 



Even if the conclusions based on the regression analyses are 

 incorrect, the soil tests also yield results which conflict with the 

 results of the physical analysis in terms of the importance of soil 

 viscosity. For example, the slopes of the lines passing through the 

 laboratory test data of Figure 7 yield values of viscosity coefficient, 

 p, ranging between 350 and 4000 Ib-sec/ft-^ depending on which data 

 are considered. The physical analysis, on the other hand, indicated 

 that values of y in the range of to 3 Ib-sec/ft-^ yielded the most 

 accurate results. It is difficult to reconcile these differences 

 on the basis of the limited tests which have been performed. 



The basic problem, therefore, is exactly what role velocity- 

 dependent terms or soil viscosity plays in affecting object penetration. 

 This is not a purely academic question. If velocity-dependent terms 

 are important, the problem of predicting penetration becomes significantly 

 more complex. Penetration prediction schemes based on easily measured 

 index properties would be difficult to develop if such a complex 

 concept as soil viscosity needed to be considered. On the other hand. 



19 



