Computational Intelligence To Optimize The Design Of The Mechanical Structure

In the context of weight cutting of artillery structure, this thesis mainly researches the theory of computational intelligence and it's application in the structural optimization of a howitzer carriage.Finite element model of box howitzer carriage is created by coupling shell and beam elements. And then the model is simplified properly and characteristic parameters are extracted. Finally, parametric model is created by PCL programming in the environment of MSC.PATRAN.Genetic algorism integrated with technology of parameterize is developed accounting for sizing and shape optimization. Penalization function is employed to deal with constraint, high efficiency genetic operators and controls parameters are designed. Thus, the global research ability of GA is enhanced. Using genetic algorism, the weight of the howitzer carriage is cut down greatly and the stress level is reduced.After sizing and shape optimization by GA, evolutionary structure optimization (ESO) is developed for topology optimization, the material with lower stress is removed gradually and then optimal topology achieved. The singularity of stiffness matrix is prevented effectively by controlling the evolution rate. The stair-case exist in optimized structure's boundary is solved by using B-spline method.Backpropagation (BP) neural network is researched as a function approximation. A BP network for mapping from structural parameters to mechanical responses of the howitzer carriage is trained. And then the BP network is used in structural optimization as a function approximation. Because of avoiding finite element analysis, the optimization process is speeded up prominently.