**3. Results and Discussion**

## *3.1. Dimensional Measurements and Uncertainty Calculation*

As stated in Section 2.3, three collected parts for each experimental run were examined. In order to assess the quality of the parts and of the three pillars per measurement position, as illustrated in Figure 3, they were examined to provide data for the replication fidelity of the pillars in each area of the parts and the stability of the process.

In a preliminary analysis the average pillar height (part) and pillar height deviation per area is presented in Figures 12 and 13, respectively, with their respective part measurement uncertainties as described in Equation (1) (Upart) and Equation (3) (Udev). The uncertainty bars as illustrated on the bar graphs are associated with the combined measurement uncertainty (Udev) (Figure 12) from both mold (Umold) and parts (Upart) (Figure 10) measurements (Equation (3)) as calculated based in the ISO 15530-3 [20].

**Figure 12.** Average pillar height and Upart measurement uncertainty per position on the part for Cavities 1 and 2 parts. The *x*-axis here represents the experimental DOE runs (R1 for Run 1) as presented in Table 1.

**Figure 13.** Comparison of average pillar height deviation (mold-part) per position for Cavities 1 and 2.

Figure 12 presents the real area pillar height of the biochip, which is homogeneous for most of the experimental runs. Figure 12 illustrates the replication fidelity of the pillars in both cavities. It is evident that the less replicated pillars are originated to position PP1, and for the experimental runs 1, 3, and 9, where the treatment uses the low value of the Tm parameter, and for runs 3 and 9 where Tmld

is also at a low level. In position RP2 though, the μ-pillars positioned at the runners before Cavities 1 and 2 are better replicated for all the experimental runs where the high level of the InjSp parameter was used, as higher injection speed increased the temperature of the molten polymer through the mechanism of shear thinning. In comparison, when the lower level of InjSp was used, the replication fidelity of the μ-pillars in position RP2 was lower due to the thicker cross-section where the shrinkage was larger than the rest of the molded component.
