The Research Of Approach To Predict The Maximum Length Of One-step Drawing

Drawing is an important sheet metal forming technique. It is used extensively in many products, such as automobiles, tractors, airplanes, watches, electrics, instruments, and other consumer goods. The technical level of drawing, in a way, reflects the industry modernization level of a nation or zone. Forming limit is one of the hot questions in the drawing research field. Look-up table method is only suitable for those parts which have regular shape and specific material. Further more, the data in tables are too old for us now. Formula derivation method is also only suitable for one condition, and when calculate by these formula, we will use calculus frequently. Numerical simulation is a very useful method to analyze forming limit. But it has some defaults, for example, software is expensive, waste time, and need a lot of specialized knowledge. The most important thing is that this kind of software is not easy to learn in a short time. In resent years some intelligent methods are used extensively in drawing, but mainly focus on processing design, blank holder force optimizing, and spring-back. Until now there is not a method to predict the maximum length of drawing quickly and preciously. So, it is important to research in this field.In this thesis, the research situation and methods of sheet metal drawing are summarized. Furthermore, the author discussed the theory of drawing and sheet metal simulation in detail. DYNAFORM, a professional stamping software developed by ETA CO. US, is introduced, too. It is maybe helpful to those who work on drawing simulation.A new approach to predict the maximum length of one-step drawing is proposed. This approach combines advantages of orthogonal experiment, numerical simulation and artificial neural network together. So, the maximum length of drawing can be predicted quickly. Cup shape parts are mainly discussed in this thesis to prove the feasibility of this approach.Factors which inflect the maximum length of one-step drawing are analyzed. Related theory is researched and conclusions about how theses factors inflect the maximum length are put forward. These conclusions are proved by a group of simulative tests, and the values of these factors used in this thesis are given.Based on the orthogonal experiment theory, suitable test tables are scheduled according to the characters of this research. The key eyes are put on the choice of factors and there levels, numerical parameters. Tests are done with a professional stamping software called DYNAFORM. These test results can be used by researchers in drawing simulation field.A BP ANN to predict the maximum length of drawing is set up. Decisions are made about how many layers and neurons should be used, and what kind of inputs, outputsand train algorithm are suitable. This ANN is trained by the sample data gain from previous tests. Finally, this net's precision is examined by remained data.The approach proposed in this thesis is developed in MATLAB6.5. The result of BP network is close to the simulation result of DYNAFORM. This approach is all-purpose. It can be extended to other shape parts and be used to direct design and manufacture.