Synthesizing ideal crystal structures is crucial for materials research, and X-ray diffraction (XRD) analysis has become the primary means to reveal their atomic and molecular structures by determining crystal phases, orientations, lattice parameters and grain sizes. With the support of material databases, XRD analysis is supported by large databases such as ICDD？ and ICSD in order to simulate XRD patterns to determine the crystal structure, but conventional methods are limited when faced with XRD patterns of unknown structure. Methods such as machine learning, prototype search and deep learning have emerged in recent years to work towards more accurate and efficient crystal structure identification, while techniques such as Rietveld refinement can correct XRD patterns, and the recently proposed BBO-Rietveld method automatically adjusts the parameters to improve the success rate. However, Rietveld refinement relying on the initial structure is still imperfect, so there is an urgent need for an inverse design method that can directly generate the crystal structure of the target XRD pattern without database support. A team lead by Dr. Joohwi Lee from Toyota Central R&D Labs., Japan, proposed an automated crystal structure creation method, consisting of the evolutionary algorithm and crystal morphing supported by Bayesian optimization (Evolv&Morph), for reproducing the XRD pattern. The method optimizes the similarity score of XRD patterns of the created structures and the target. The evolutionary algorithm can automatically create various structures by genetic operators without input structures. Crystal morphing, using the input structures obtained by the evolutionary algorithm, expands search space and further increases the similarity of XRD patterns creating a better input structure for the post-refinement. After the refinement, for twelve binary and ternary systems in different crystal structures where the target XRD patterns were provided by simulation from the measured structures, Evolv&Morph achieves cosine similarity of ~99%. In addition, for four different powder XRD patterns, the present method achieved cosine similarity of of ~96%, after removal of background. These results indicate that the created crystal structure successfully reproduces the target XRD pattern. Therefore, the present method can identify unknown crystal structures after getting XRD measurement without depending on DB. Furthermore, Evolv&Morph can play a role of inverse design, which indicates that the desired property is defined first and the materials with such property are automatically searched. The optimization target score, which corresponds to the cosine similarity of the XRD pattern used in this study, can be exchanged for particular functional property according to the goal of the material design; therefore, the present method has a powerful potential to be applicable to material design, as well as crystal structure determination.