Page 424 - Practical Design Ships and Floating Structures

399

4) Joining of the strips. A developed view of the hull plate with crevices is obtained by joining the
surface strips along the reference line in a proper sequence. These crevices are to be removed by line
heat forming.

3 QUANTITATIVE ANALYSIS OF LOCAL DEFORMATION BY FLAME BENDING M
From the study on the relation between the integral shape and the local deformation values, it is known
that complicated as the curved surface is, flame bending can be conducted on a quantitative basis if the
local deformation can be determined. So it is necessary to define the different factors influencing local
deformations, and this is one of the main technical difficulties in flame bending. Among the many
technical parameters, transverse contractions and angular deformations are of great significance. This
paper deals with the relation between transverse contraction &d some main parameters.

3.1 Establishment of mathematical model

According to the analysis of the local deformation of the hull plate formed by line heat, plate width B,
( mm ) , plate length L, ( mm ), plate thickness H, ( mm ), transverse radius of curvature R,, ( mm ),
heat line length L, ( nun 1, heat line intervals D, ( mm ) and heat velocity V, ( d s 1 are determined
to be the main influence factors, based on the sample data collected from the ship plate in Dalian New
Shipyard. Three regression models are presented with the help of the stepwise regression method:
Model I: y = boB, +b, L, + b2H, +b,R, +b4 L, +b,D, +b6Vh +b, +E (5)

Model 111: = CB; L; H," R; L? D,"v,".E
Ls = -0.000079859R, + 0.00245891, - 0.00081 7040,
The results from model I:
- 0.25506Vh + 1.9592 + E

3.2 Validity of mathematical models
3.2.1 Significance test of population linear correlation

F-statistics can be obtained by analysis of variance, with a level of significance a =0.05. and F-
distribution value of (p, n-p-1 ) degrees of freedom is found from F-distribution table. The calculation
results for model I: F = MSR I MSE = 21534, Foos(p, n - p - 1 ) = 2.4286, as
F > Foos(p, n - p - 1 ) with a level of significance a =0.05, and hence HI ( there is a nonzero at least
for bl . bz ..* 6, ) is accepted. That is to say, the linear relation significance exists between y and x, in
model I.

3.2.2 Goodness offit inspection of regression functwn