Optimizing And Die Life Prediction Of Hot Forging Die Based On Wear

Hot forging die is a processing equipment in industrial production. It has beenmore and more widely used along with the continuous development of technology. Thework environment of hot forging die is poor and it will constantly bear mechanical loadand thermal load during the hot forging process. The life of hot forging die that can beachieved is lower in the domestic. It is only just3000~5000parts and still far behind theinternational level. So how to increase the service life of hot forging die is an urgentneed to be solved problem. Die wear is the main reason for the die scrap. The studyshows that hot forging die failure is caused by wear accounted for about70%of the diefailure. So, reduction of hot forging die wear is the key to improve die life.First of all, in this paper, considering the special work environment of hot forgingdie, the Archard model is modified. The paper uses Deform-2d finite element softwareto simulate the forming process and calculate wear depth. The simulation result showsthe forming effect of forging is good and the maximum wear part of the wheel hub hotforging die is the fillet region of the forming upper die mandrel. Then the paper does twoaspects of optimization about the upper die mandrel. The first aspect is optimization ofoutline shape for the upper die mandrel, including the fillet radius and the linearoptimization. In the fillet radius optimization, a group of fillet radius is selected and theresult shows that the wear value is smaller when the radius is R2~R3. On the basis ofthe optimization of fillet radius, combined with the respective advantages of the finiteelement simulation, BP neural network and sequential quadratic programming algorithm,the outline linear of the upper die mandrel is optimized. In the linear optimization, using3-spline interpolation function curve to describe the outline shape of forming part of theupper die mandrel.19control points’ horizontal coordinates of3-spline interpolationfunction curve to be as design variables. Objective function is the equal wear value.Randomly generates several outline shapes and they are simulated by Deform-2d. Thepaper builds3layers BP neural network which is trained with the date of FEMsimulation results as learning samples. So that mapping relation between designvariables and objective function is established. Then the corresponding objectivefunction value will be predicted by given design variables through using the predictiveand popularized characteristic of BP neural network. Finally, the sequential quadraticprogramming algorithm is adopted to optimize the selected design variables. On the basis of the first aspect, the second aspect optimizes the process parameter based on thewear. The die initial temperature, forming speed and friction factor on the influence ofwear has been analyzed, and a group of process parameter is obtained on the basis ofminimum wear by the analysis.Through the optimization design the result shows that wear of the upper diemandrel has reduced and it is more uniform, the maximum wear decreases by26.8%.The life of die is3887parts which is predicted by cumulative wear of the previousforging cycles, compared with the original die design, increases by25.2%.However,because of the restriction of relational project time, the contents of this paper are mainlyfocused on the part of theory research.