Study On The Bionics Algorithm And Its Application On Structure Optimization Design Of The Oil Hydraulic Press Crossbeam

One forging machine company in which a type of 10MN oil hydraulic press is trial producing supports the project. In this paper, the strength and rigidity of the oil hydraulic press frame is analyzed with advanced design methods, and its crossbeam optimization design is performed using a bionics algorithm-combined neural network and genetic algorithm based on Pareto optimal theory. The contribution and innovation of the paper include: 1.The 3D tetrahedron elements are used to form the space-assembled structures, the whole oil hydraulic press frame's model is built with the advanced FEM analysis software ANSYS, and its strength and rigidity is analyzed. The deformation and stress of the frame has been obtained by a series of procedures. These results provide foundation for optimization of the frame design. 2.The structures approximation analysis technology is studied based on neural network. Using the saturated multi-level table of orthogonal arrays to select the samples can disassemble the variables scales finely, so the trained neural network has extensive representations. The modified BP neural network is put forward, which can adjust learning parameters dynamically and adaptively. The non-linear mapping relations between the structural parameters of the hydraulic press' crossbeam with its displacements and stresses are built based on BP neural network. With the BP neural network model replacing the original finite element model, the calculating precision satisfies the engineering requirement and the former is much simply than the latter. 3.The implement method using the APDL of ANSYS for the finite element models of the crossbeam of the oil hydraulic press is put forward and carried out. The manual work and sampling time for the BP neural network are largely decreased, which provides an effective sampling method for BP neural network applied to the modeling of the complicated structures. 4.Because of the defects of the basis of the simple genetic algorithm in structural optimization, a set of Pareto GA for multi-objective optimization is presented. There is some innovation in this part; for example, the basic operators of SGA are revised. A ranking algorithm for fitness function in GA is built taking use of Pareto concept. Pareto set filter, niche operator are used in this part. For discrete parameters of the crossbeam, a new technique for rounding off the continuous solution is given. At last, a new method is presented to choose the most satisfied solution taking account in the extent of importance for each objective. 5.As an example to illustrate the theory of this bionics algorithm, multi-objectives optimization design to minimize the volume and the displacement of the oil hydraulic press crossbeam is preceded. The results not only overcome the randomicity of the blue print comparisons, but also get the reliable exoteric optimal variables. The bionics algorithm provides a new scientism method on designing plank thickness of the hydraulic press frame and other box structures.