The paper deals with a fundamental problem at the core of a data-driven approach for history-dependent materials: how to compute the hidden state variables related to the material memory from experimental data. This problem, already introduced in Ladeveze (2019, 2022), is transformed here to be solved by classical numerical methods. These additional state variables allow the Experimental Constitutive Manifold, built from experimental data, to be a consistent data-driven material model. The proposed computational method is described and analyzed on 2D problems for which experimental data are simulated using classical elastic-(visco)-plastic models.

该论文处理了历史相关材料的数据驱动方法的核心基本问题：如何从实验数据中计算与材料记忆相关的隐藏状态变量。这个问题已经在 Ladeveze (2019, 2022) 中介绍过，在这里被转换为通过经典数值方法来解决。这些额外的状态变量允许从实验数据构建的实验本构歧管成为一致的数据驱动材料模型。所提出的计算方法在 2D 问题上进行了描述和分析，其实验数据使用经典的弹性 - （粘） - 塑性模型进行模拟。