PSI - Issue 8

A. Grassi et al. / Procedia Structural Integrity 8 (2018) 573–593 Author name / Structural Integrity Procedia 00 (2017) 000 – 000

583 11

Viscous Criterion ( V*C ) (Lau and Viano (1986)).

2.6. Design Of Experiment (DOE)

After an initial assessment, a parametrical study was performed to evaluate the effect of design parameters on the restraint performance of the devices. In literature, it is possible to find many information about the belt anchorage points (three or four) (Rouhana et al. (2003)), about innovative seat belts with independent control of the shoulder and lap portions (López-Valdés and Juste-Lorente (2015); Pipkorn et al. (2016)). These sources highlight the great influence of the geometrical characteristics on the system behavior. Thus, geometrical parameters were included as variable in the study. In addition, since the general objective of the research was to develop a safety device installable on all PTWs, changes in the device layout will be necessary to adapt to the frame of different models. For this reason, the assessment of the device performance with different geometrical parameters is crucial. A full factorial Design of Experiment (DOE) was chosen since no information about factorial interactions was available. Five factors and two levels were considered. Three out of five factors are linked to geometrical dimensions of the device:  Longitudinal slip-ring position  Vertical slip-ring position  Belts/restraint cable vertical link position The remaining variables are binary and represent alternative options for system configuration: Although the problem was evidently non-linear, two levels were selected since: 1) the range of the geometrical variables was narrow; 2) the computational time of the finite elements model was high, and a linear approximation was accepted for a preliminary study. The limits of the variability range were determined considering the dimensions of a typical seat. Setting the initial configuration as reference (Figure 4a), the changes for each variable in Table 2 were considered. Their schematic representation can be found in clockwise order in Figure 4, starting from configuration (a).  Pretensioner presence  Vest-belts link position

Figure 4. DOE: Representation of geometrical factor changes.

The full factorial design with 5 factors (3 numerical, 2 categorical) with 2 levels consists of 32 experiments (i.e. FE simulations). Replication of the 32 configurations were not considered since the experiment was a finite element calculation and there were no external factors that may be influence the tests. The backward elimination strategy was