Failure mode-based partitioning of the failure envelope of a structure can make the design process for optimal structures more efficient. However, partitioning the failure envelope is challenging for structures whose failure modes are not known a priori. We present a two-step algorithm that automates the failure mode-based partition of the failure envelope for a structure and demonstrate its capability using tubes with a circular cross section as canonical structural elements. The first step of the algorithm employs non-intrusive finite element analyses (FEA) to generate the structure’s failure envelope. The von Mises stress field at the onset of failure encapsulates critical information about the failure mode. We exploit this observation by using the stress field output by the first step of the algorithm as input for the second step. The second step of the algorithm uses clustering, an unsupervised machine learning technique, to partition the failure envelope based on the von Mises stress field at the onset of failure. We use the algorithm to generate partitions of the failure envelope for tubes with circular cross sections subjected to pure bending and three-point bending. In the pure bending case, where analytical results are available in the literature, the results from our algorithm show good agreement with analytical results. We provide practical guidelines for choosing suitable values for the various parameters and hyperparameters in the algorithm.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号