COMPUTER BASED DESIGN METHOD AND DESIGN SYSTEM

摘要

For a large number of design variants (KV) of a technical product (TP), a training structure data record (SDT) specifying the respective design variant (KV) and a training quality value (QT1, QT2 , QT3) read in as training data (TD). Using the training data (TD), a Bayesian neural network (BNN) is trained to determine an associated quality value (Q) together with an associated uncertainty specification (UC) using a structural data set (SD, SSD). Furthermore, a large number of synthetic structure data sets (SSD) are generated and fed into the trained Bayesian neural network (BNN), which generates a quality value (Q) with the associated uncertainty information (UC) for each of the synthetic structure data sets (SSD). The generated uncertainty information (UC) is compared with a specified reliability information (REL), and depending on this, one of the synthetic structure data sets (SD) is selected. The selected structure data record (SD) is then output for the manufacture of the technical product (TP).