When creep recovery is predicted by the principle of superposition from the unit creep curves for various loading ages, both Branson's law (ACI 209/11) and the double power law give non-monotonic recovery curves. This has been criticized by some as thermodynamically inadmissible. Using an age-dependent Maxwell chain model, it is shown, however, that this phenomenon is thermodynamically admissible.
Creep recovery, though, is not accurately predicted by the principle of superposition, and the result should rather be interpreted as the admissibility of a divergence of creep curves for different loading ages, which is indeed observed in many measurements.
Such divergence can be modeled by Maxwell chains but not by Kelvin chains, so the latter imply an arbitrary, unjustified limitation. The same is characteristic of the "improved Dischinger formulation" (recently adopted from DIN for the new C.E.B. Model Code), which represents a special case of Kelvin chain. This is a further reason why the separation of reversible (delayed elastic) creep component is an arbitrary as ~ sumption lacking thermodynamic justification.
For aging creep only Maxwell chain rheologic models should be used. Other related shortcomings of "improved Dischinger formulations" are also summarized.