Investigation For Preparation Parameters Optimizing Of Bonded NdFeB Permanent Magnet Based On The Improved BP Neural Network

Investigation for preparation parameters optimizing of bonded NdFeB permanent magnet has important significance in technology and economy.If the relationship between material magnetic properties and preparation parameters can be ascertained,we can use it to explore the optimization parameters,and seek for the best material properties at the condition of little knowledge about its particular mechanism.In order to search the influence of multi-factors on magnetic properties of bonded Nd-Fe-B permanent magnets,four factors(binder content of bonded magnet, thermosetting temperature,thermosetting time and molding stress) which contribute mostly have been choosen for studying,and uniformity design experiments are adopted in this paper.Experimentation datum are dealed with by statistical analysis software(SAS),and regression equations between preparation parameters and magnetic properties of Nd-Fe-B permanent magnets are obtained using step by step regression.And the relationship between preparation parameters and magnetic properties is established by the improved artificial neural network(ANN) predicting model.Regression equations between preparation parameters and magnetic properties of Nd-Fe-B permanent magnets are obtained using uniformity design experiments method and step by step regression means.Analysis of variance (ANOVA) shows that the regression models are creditable,and it can be used to discussed the influences of multi-factor interaction on magnetic properties.The results indicate that B_r presents a parabola-shaped tendency with the increase of modeling stress,and the interaction between binder content and thermosetting temperature,thermosetting temperature and modeling stress both have disadvantageous effects on increasing B_r.The effect of thermosetting temperature on coercivity H_cj is similar to the effect of modeling stress on B_r, and the interaction of binder content and thermosetting temperature, thermosetting temperature and thermosetting time,thermosetting time and modeling stress all affect H_cj,the last interaction contribute mostly.Maximal energy product(BH)_m presents a reducing tendency with the interaction of binder content and thermosetting temperature,thermosetting temperature and modeling stress.The effect of modeling stress on compressive strength assumed the quadratic nonlinear progressively increasing regularity.The interaction of thermosetting temperature and modeling stress has advantageous effects on increasing compressive strength,and the interaction of thermosetting time and modeling stress has disadvantageous effects.While keeping enhance the thermosetting temperature and prolong the time,magnetic properties will be reduced,but compressive strength will increase notably.The predicting model is established by using the improved artificial neural network to simulate the relationship between preparation parameters and magnetic properties.The model not only has good precision and low relative errors,but also can predict the results steadily.It is in good practicality. 2-dimension curves and 3-dimension figures of craft-properties are obtained by the ANN model.The influences of single factor or the interaction among factors on magnets magnetic properties are respectively discussed according to the curves ploted by ANN model.Simulation result shows that magnetic properties present a parabola-shaped tendency with the increase of binder content, thermosetting time has little effect on magnetic properties,the effect of thermosetting temperature is simlar to binder content,modeling stress has advantages effect on increasing magnetic properties.The interaction among preparation-craft also play an important role in megnetic properties,specialy between thermosetting temperature and thermosetting time.While keeping enhance the thermosetting temperature and prolong the time,compressive strength will increase,but magnetic properties will be reduced notably.According to the improved ANN predicting model,the best preparing parameters and properties have been obtained as follow:binder content 2.5wt%; molding stress 20kN;thermosetting at 150℃for 1b.Obtained properties of bonded NdFeB magnet as follow:B_r=0.594T;H_cj=751kA/m;(BH)_m =55kJ/m³; compressive strength is 188MPa.