(a) (b)

Figure 7。 The values of responses at the initial values: (a) clamping force diagram, (b) warpage value。

Table 3。 Coefficients of responses。 Table 5。 ANOVA table for clamping force。

Source SQ MS F P

Model 1。497 × 107 7。486 × 105 396。31 < 0。0001

A 13292。36 13292。36 7。04 0。0137

B 4。449 × 105 4。449 × 105 235。53 < 0。0001

C 14975。6 14975。6 7。65 0。0115

D 1。211 × 107 1。211 × 107 6408。93 < 0。0001

BD 14360。43 14360。43 7。6 0。0107

A2 18064。49 18064。49 9。56 0。0048

B2 1。019 × 106 1。019 × 106 539。29 < 0。0001

C2 13849。4 13849。4 7。33 0。012

D2 7。072 × 105 7。072 × 105 374。43 < 0。0001

Residual 47219。81 1888。79

Total 1。502 × 107

melt temperature, packing time, packing pressure and cooling time, respectively。 Approximate equations of two responses are presented as Eqs。 (3) and (4), respectively。 Table 3 de- scribes values for coefficients of equations as determined by regression method。

shown in the Table 2。

According to simulation results, regression response sur- face models for the two objective functions of evaluating clamping force and warpage are derived。 Clamping force and

warpage initial values are shown in the Figure 7。

Second-order polynomial regression is employed to estab-

lish non-linear relationships among design variables and res- ponses。  The  responses  are  functions  of  mold temperature,

Table 4。 ANOVA table for warpage。

Based  on  computational  cost  and  time-of  simulation, as

compared with the MoldFlow, the  developed predictive model is a much simpler and more efficient in predicting outputs with change-of-design variables。 The adequacy of the developed models, including warpage and clamping force via ANOVA analysis with sums of squares (SQ), mean squares (MS), F-value (F), P-value (P) is shown in table 4 and table 5, respectively。 The backward process eliminated the insignifi- cant terms to adjust the quadratic models。