Shape-memory materials are a promising new class of smart materials with many applications such as strain sensors, artificial muscles, and smart breathing textiles. These materials are subjected to force and extension in situ. Thus, the time-dependent behavior of these materials can play an important role in their long-term performance. The present study was conducted to investigate the time-dependent behavior of a type of shape-memory fabric. Nanoclay-reinforced polylactic acid/thermoplastic polyurethane was used as the precursor. The yarn that was produced was highly twisted. The twisted yarn was then shaped into a coiled structure by mandrel annealing. This yarn was then used to produce knitted fabric. The fabric was examined under both cold (25 °C) and hot (50 °C) conditions. The fabric contracted in hot water in the course direction but did not show a significant contraction in the wale direction. It returned to its original width in cold water. This effect was observed repeatedly over several cycles. This shows that the knitted fabric composed of the precursor twisted-coiled yarn exhibited low-temperature actuation and reversible two-way shape-memory behavior. A value of %16 was calculated for the contraction stroke along the course direction. The stress relaxation behavior of the two-way shape-memory fabric was then studied and analyzed. For this purpose, four different viscoelastic models were considered: the standard linear model (a), Burgers model (b), Jeffrey model (c), and Kelvin-Voigt-Maxwell model (d). We used the curve fitting procedure to find the best fit to the experimental data based on the least-squares method. The results showed that the Kelvin-Voigt-Maxwell model (model d) exhibited a higher and more acceptable regression coefficient (R2) than the other three models. The Jeffrey model showed the lowest regression coefficient (R2), thus confirming that it is not suitable for explaining the relaxation behavior of the fabric. However, a limitation of Model (d) is that it is not in line with the experimental loading stage. To modify the model, we propose the replacement of the dashpot with a dynamic frictional element. The results indicate that the proposed dynamic friction model can eliminate the limitations of the dashpot body during the loading stage.