Solving the Minimum Distance Problem for the Synthesis of Mechanisms

One of the most successful general methods for the synthesis of mechanisms is that based in the deformed position problem, which has seen uninterrupted development on the last 20 years. The advantages of this formulation are its versatility (can solve prescribed and unprescribed timing problems, different kind of requirements, works with any mechanism, being only limited by computational cost); its great convergence; and its efficiency not only when optimizing simple mechanisms, but also for complex mechanisms when starting from a good initial guess. This allows one to adapt machinery for new tasks in an easy way like is needed, for example, in packaging industry. The method (as many others) is based on the synthesis point concept, which is related to a set of requirements to be accomplished in a possible configuration of the mechanism. The main caveat of the deformed position problem is derived from the fact that, when used for mechanism synthesis, it is an approximation of the real task to optimize, and this leads to the problem that it can converge to mathematical solutions which are useless. This happens when optimizing complex mechanisms via sequential quadratic programming algorithms without a good initial guess, or when using explorative methods such as genetic algorithms. In both cases, the algorithm will favour low stiffness mechanisms, due to the fact that they can reach any requirement with high deformation, but low deformation energy. To solve this problem, here the use of a minimum distance problem is proposed, where one tries to reach the minimum distance that the mechanisms is able to achieve to the considered requirement. In order for this idea to be suitable for multiple requirements in a single synthesis point, a weighted square distance summation is used. This should solve the stiffness issue while keeping all the advantages of the deformed position problem, at the cost of an increase on the convergence cost. Although developed for bidimensional problems, a three dimensional generalization should be easy to formulate. In this presentation this function is presented, along with some considerations on its resolution via a sqp algorithm using Lagrange multipliers. Furthermore, some initial experimental results are presented which allows one to figure out the future performance of the function when used as a basis for mechanism synthesis.