Open AccessIn design and manufacturing there are a lot of situations to make engineers to deal with products, which have redundant mechanical structures. This is a harmful effect that can produce non-solvable dimension chains at designing preproduction stage, and lead to rebasing in the course of assembly. Traditional means for mathematical description of mechanical structures, such as directed and undirected graphs, matrices, and etc., have no appropriate tools to identify redundancy and eliminate it. This can be achieved by using a relatively new mathematical model, i.e. a hyper-graph of mechanical linkages. It is shown that a necessary condition of redundancy is simple inequality X Eliminating the unnecessary mechanical linkages generates mechanical assembly structures with different assembly properties. The paper considers generation of non-redundant structures, which possess a maximum fractionalizing property. So-organized products can be broken up into the largest number of independently assembled fragments - assembly units. This simplifies the preproduction engineering and disassembly procedures in case of full or selective repairs during operation. It is shown that the normal sequence of assemblies enables us to reduce a mechanical structure of such products to the chain of a maximum length. We propose an algorithm to eliminate redundancy. It generates the maximum fractionalizing mechanical structures. The algorithm is based on three heuristic rules that recommend eliminating edges, which disarrange a chain structure of the hyper-graph. The paper provides a rationale heuristics and considers a test case. To test the algorithm using a large array of examples, the experimental software has been developed. Computational experiments have been conducted using a sample of thirty redundant hyper-graphs randomly generated. The experiments have shown the highest relevance of the heuristic rules and algorithm reliability.