Several modeling approaches can be considered. The constant-viscosity Newtonian model is the simplest approach, and this is recommended if no rheological data are available. The use of a generalized Newtonian fluid model should not be considered for blow molding or thermoforming, since such a model will lead to unrealistically high velocities due to strain thinning.
Further modeling recommendations for 2D and 3D simulations are provided below.
In addition to the constant-viscosity Newtonian model, differential viscoelastic models (Maxwell, Oldroyd-B, PTT, Giesekus, FENE-P, DCPP, and Leonov) are available for blow molding and thermoforming. The White-Metzner model is not recommended, in view of the dependence of the rheological properties on the local kinematics. Although multi-mode simulations are available, for reasons of computational cost in Ansys Polyflow, single-mode transient viscoelastic calculations are suggested, at least in first instance.
The flow involves a typical time scale τ, corresponding to the inflation. The relaxation time should be of the same order of magnitude as this time scale.
For the viscosity factor, select a value corresponding to the shear viscosity as obtained for a shear rate of 1/τ.
For a strongly strain-hardening material (for example, LDPE), you can use the Maxwell or Oldroyd-B model when the strain rate remains moderate. Alternatively, you can use a low value of ε (typically 10–3 to 10–2) with the PTT model, a low value of α with the Giesekus model (typically 10–3 to 10–2), or a high value of q for the DCPP model. For strain-thinning or moderate strain-hardening materials (for example, LLDPE or HDPE), use the PTT or Giesekus model with a higher value—typically about 0.1 or more—for ε or α or the DCPP model with a low value of q.
If data on elongational viscosity are available, they should be used. If the resulting values for the nonlinear parameters are not in agreement with the expected behavior of the melt, they can be fixed as noted in Assigning a Value to a Parameter.
In the automatic fitting procedure, it is preferable to consider the data in the range of angular frequencies and deformation rates of interest, typically one decade above and one below the value 1/τ. If data extrapolation is necessary, it should be done over no more than one decade. Also, use appropriate weighting factors (see Weighting Measured Data) if some data are more reliable than others.
A purely Newtonian contribution can be added to the model. This corresponds to that part of the spectrum associated with very short times, and the response of which is shorter than the process time τ itself.
The shear viscosity curve for the model may differ from measurements, but this can generally be disregarded, since elongation is the main component of the flow.
For computational reasons, the use of the shell element (membrane element) is recommended for blow molding or thermoforming simulations that involve objects that are thin and have geometrically complex shapes. In addition to the constant-viscosity Newtonian model, the integral viscoelastic KBKZ model with a relaxation spectrum is also a good choice when a Lagrangian representation is used. No damping (that is, the Lodge-Maxwell model) is needed, due to the moderate deformations involved in the process. The Lodge-Maxwell model predicts strain hardening up to a level that is experimentally observed.
Based on the oscillatory properties, a spectrum of 4 to 8 relaxation times is recommended. If you use a low number of relaxation times, they should be selected around the typical process (inflation) time τ. If you use a high number of relaxation times, they can spread to values as short as 10–4 and as long as 104 s.
Oscillatory data should be considered in the range of angular frequencies as covered by the range of reciprocal relaxation times. Viscosity factors should be identified on the basis of linear properties, as well as on the basis of the shear viscosity (possibly with a zero weighting).
If data on elongational viscosity are available, they should be used. If the resulting values for the nonlinear parameters are not in agreement with the expected behavior of the melt, they can be fixed as noted in Assigning a Value to a Parameter.
A purely Newtonian contribution can be added to the model. This corresponds to that part of the spectrum associated with very short times, and the response of which is shorter than the process time τ itself.
The shear viscosity curve for the model may differ from measurements, but this can generally be disregarded, since elongation is the main component of the flow.