Empirical algebraic models as Generalized Sigmoidal model (GSM) and Havriliak-Negami model (HNM) are selected as the pre-smoothing models for comparison. Asphalt mixtures with raw, modified and resin blended binders are tested for original dynamic moduli and phase angles within the linear viscoelastic region by the simple performance test. Three approaches of GSM are evaluated to enhance the accuracy of constructing master curves. The results indicate that modulus master curves are asymmetric with shape parameters less than 1 for thermo-plastic mixtures, and larger than 1 for thermo-setting mixtures, while the most accurate modeling phase angles are from HNM due to its exact Kramer-Kronig relations. The discrete relaxation spectra are derived from pre-smoothed data through the Windowing methods (WMs) with Generalized Maxwell model (GMM) based Storage Algorithm and Loss Algorithm (GM-SA and GM-LA). Windows regions of algorithms are modified to be suitable for iterations. Continuous spectra are determined to verify the accuracy of obtained discrete spectra through inverse Fourier transforms. The results indicate that the discrete spectra with high spectrum points density r = 2 from HNM are more consistent with the discretized continuous spectra than those from GSM with low density r = 1. Moreover, the WMs with empirical pre-smoothing models take obvious advantages in parameters adjustments and relaxation times distribution selections. GM-LA rather than GM-SA exhibit higher accuracy and convergence rate in spectra determinations, since the kernel functions. In addition, the comparisons between the GMM recovered and GSM or HNM measured modulus with fit goodness statistics show that low density of spectrum points per decade is the main reason in the GMM loss modulus oscillations. However, the deviations are less obvious between HNM pre-smoothed moduli and GMM generated moduli than those between GMM-GSM moduli. Therefore, the linkages between empirical and mechanical models are established on relaxation spectra determination by modified WM using GM-LA with denser spectrum distribution r = 2.

选择经验代数模型（如广义Sigmoidal模型（GSM）和Havriliak-Negami模型（HNM））作为预平滑模型进行比较。通过简单的性能测试，可以测试含生，改性和树脂共混粘合剂的沥青混合料在线性粘弹性区内的原始动态模量和相角。对GSM的三种方法进行了评估，以提高构造主曲线的准确性。结果表明，模量主曲线是不对称的，对于热塑性混合物，形状参数小于1，对于热固性混合物，形状参数大于1，而最精确的建模相角来自HNM，这是由于其确切的Kramer-Kronig关系。离散弛豫谱是通过加窗方法（WM），基于通用麦克斯韦模型（GMM），基于存储算法和损耗算法（GM-SA和GM-LA）的预平滑数据得出的。修改算法的Windows区域以适合迭代。确定连续光谱以通过逆傅立叶变换来验证获得的离散光谱的准确性。结果表明，与低密度r = 1的GSM相比，来自HNM的高光谱点密度r = 2的离散光谱与离散连续光谱更加一致。此外，具有经验预平滑模型的WM在以下方面具有明显的优势：参数调整和松弛时间分布选择。GM-LA而非GM-SA在光谱测定中显示出更高的准确性和收敛速度，由于内核功能。此外，回收的GMM与GSM或HNM测量的模量以及拟合优度之间的比较表明，每十倍频点的密度较低是GMM损耗模量振荡的主要原因。但是，HNM预平滑模量和GMM生成模量之间的偏差不如GMM-GSM模量之间的偏差明显。因此，经验模型和力学模型之间的联系是通过使用改进的WM使用GM-LA（具有更密集的光谱分布r = 2）确定弛豫光谱来建立的。HNM预平滑模量和GMM生成模量之间的偏差不如GMM-GSM模量之间的偏差明显。因此，经验模型和力学模型之间的联系是通过使用改进的WM使用GM-LA和更密集的光谱分布r = 2确定弛豫光谱来建立的。HNM预平滑模量和GMM生成模量之间的偏差不如GMM-GSM模量之间的偏差明显。因此，经验模型和力学模型之间的联系是通过使用改进的WM使用GM-LA（具有更密集的光谱分布r = 2）确定弛豫光谱来建立的。