Data Analysis for the Projection of Flexible Composite Materials to Naval Transport Scenarios

Accidental spills of oil or other types of hydrocarbons represent a problem of utmost importance, but, in the situation where a multi-criteria approach to the phenomenon is desired, it is necessary a quickly intervention for mitigate the effects of their spread, and for isolation, collection, transportation and storeage for reprocessing. In case of oil recovering, trapped oil can be pumped out to holding tanks (shuttles) for transporting to shore. The functional characteristics required for naval transport are represented by the: operational in strong sea currents oil spill recovery (min. 4bf), transport and storage (at min. 2kt); rapid response in an emergency (possibility to be used in max. 1 h in conjunction with oil spill recovery equipment: vessel, booms, skimmers etc). Although it could be considered that the tear resistance on the longitudinal and transversal system for any composite structure including a textile matrix based on woven structure, could be influenced by the physical-mechanical characteristics of the textile reinforcement (the nature of the raw material and the diameter of the threads), in the situation of usage the flexible elements (narrow fabrics) to join the panels, the correlation between the above mentioned parameters could be made only by mathematical approaches. This assumption is based on the fact that in the textile field, the mechanism of deformation of threads and implicitly of planar structures is not fully explained (as in the field of constructions). The paper presents the analysis of the data collection, with the help of multiple regression, each of the 6 dependent variables (tear resistance values assessed according to three accredited methods in longitudinal and transversal systems) being modeled with the help of 5 independent variables (resistance to maximum force breaking strength, knot resistance, loop for two types of sewing thread and the breaking resistance of composite material fabrics). For the 3500 values obtained as a result of the experiments carried out, the initial hypotheses were related to: i) the experience matrix (u observations for q variables) is fixed, it is not stochastic, and the number of experiences is greater than the number of variables and ii ) the matrix of measured values for the independent variables has linearly independent columns, so it forms a basis of a q-dimensional space. The main problems followed were related to the model parameters, measurement errors, adjustment precision and the choice of the prediction model. The built probabilistic models explain between 55-80% of the variation of the dependent variable, so it would be indicated to introduce additional variables (e.g. for composite material with 45/55% PES/p-aramida matrix) of the type: pattern of the fabric, yarn density in warp and weft, coating thickness. Additionally, the values obtained for the t test identified the importance of the predictors placed in the multivariate regression equations.