GENERATION AND UTILIZATION OF SELF-IMPROVING DATA-DRIVEN MODELS WITH SELECTIVE SIMULATION OF 3D OBJECT DESIGN

摘要

Methods and systems are disclosed for computer aided design using self-improving data-driven models with selective simulation of a three-dimensional object design. A simulation engine executes a three-dimensional simulation of a designed object for a design point. A classifier engine executes a first neural network to predict whether the design point for a 3D simulation will successfully be executed by the 3D simulation engine to produce a computed key performance indicator (KPI) value. A surrogate model engine executes a second neural network representation of a surrogate model to compute KPI values corresponding to a new design point while bypassing a simulation by the simulation engine.