Flow Instabilities in Rheotens Experiments: Analysis of the Impacts of the Process Conditions through Neural Network Modeling Stefania Tronci, 1 Salvatore Coppola, 2 Fabio Bacchelli, 2 Massimiliano Grosso 1 1 Dipartimento di Ingegneria Meccanica, Chimica e dei Materiali, Universita ` degli Studi di Cagliari, I-09123, Cagliari, Italy 2 Elastomers Research and Development Center, ENI-Versalis S.p.A., I-48100, Ravenna, Italy Fiber spinning experiments are conducted with a capil- lary rheometer and a Rheotens tester on linear sty- rene-isoprene-styrene copolymer samples by varying extrusion temperature and drawdown velocity in a wide range of values, also covering the occurrence of insta- bility phenomena. Tensile stress is measured during the experiences, and the experimental time series are then analyzed by means of a new methodology. The proposed approach is based on Neural Network modeling of the time series, coupled with Principal Component Analysis postprocessing of the results. The methodology is able to identify and quantify the effects of process condition on the dynamical behavior of the system. POLYM. ENG. SCI., 53:1241–1252, 2013. ª 2012 Society of Plastics Engineers INTRODUCTION Extensional properties of the polymers strongly affect the performance of many industrial processes like blow molding, fiber spinning, film casting, etc. For this pur- pose, a tensile tester, the Rheotens test [1] has been rec- ognized as a powerful tool for the characterization of the materials subjected to spinning processes. In a Rheotens experiment, uniaxial elongation of the polymer melt is performed under the action of a tensile drawdown force: the material is extruded with a constant extrusion velocity V 0 through a die and the extruded melt is then stretched by toothed wheels at a given draw-down speed V [ V 0 . Usually, ramp experiments are realized with this experi- mental fixture: draw-down speed is started by slowly increasing the take-up speed of the Rheotens wheels, until there is the occurrence of (i) polymer filament breakup or (ii) flow instabilities. The wheels acceleration is used to be small enough to assume quasi-steady state conditions for the tensile stress measurements. Rheotens experiments are demonstrated to be successful to infer material proper- ties. In particular, with an adequate choice of shift factors, it is possible to obtain a Rheotens mastercurve, which is invariant to changes of the polymer characteristics (e.g. melt temperature, average molar mass, etc.), and it even- tually gives useful insights on the characterization of the material properties (see e.g. Refs. 2 and 3). Although the procedure is well established, it shows some (minor) limitations, at least in its current form. First, for thermo-rheologically complex materials it is not possible to reduce data obtained at different temperature and velocities to a single mastercurve. As a consequence, the superposition of the Rheotens curves is not possible and a clear dependence on the process parameters cannot be devised. Second, the procedure leads to exclude from the analysis the flow instabilities occurring at high draw down velocities. This phenomenon, which is known as draw resonance, is manifested by sustained periodic varia- tions in spinline variables such as cross-sectional area and tension [4–7]. Mathematically [8], the onset of the draw resonance corresponds to a supercritical Hopf bifurcation [9, 10], it is a hydrodynamic instability, not a viscoelastic one, albeit altered by viscoelasticity [11, 12] and thus, even Newtonian fluids can exhibit it. Recently, chaotic regimes were also observed in similar experimental fix- tures [13]. It should be remarked that other authors, with a different approach, have focused their attention on these flow instabilities and they have already demonstrated that their characterization could give useful insights on the material properties [14]. In this work, we propose an alternative, new methodol- ogy to characterize the material properties by exploiting the Rheotens experimental fixture. The protocol is able to explore even the draw resonance conditions, which are in general not considered in the traditional Rheotens analy- sis. The methodology is based on techniques introduced Correspondence to: Massimiliano Grosso; e-mail: massimiliano.grosso@ dimcm.unica.it DOI 10.1002/pen.23387 Published online in Wiley Online Library (wileyonlinelibrary.com). V V C 2012 Society of Plastics Engineers POLYMER ENGINEERING AND SCIENCE—-2013