Material Removal Modeling And Surface Properties Of Inconel 718 Superalloy Based On Sound Signal

With the rapid development of Industry 4.0 and intelligent manufacturing,the development of intelligence in the field of grinding and processing is getting better and faster.Robot belt grinding is gradually replacing traditional manual grinding because of its high material removal rate,stable molding quality and high processing accuracy,especially for difficult-to-machine materials like Inconel 718.Inconel 718 nickel-based superalloy is widely used in aerospace and other fields due to its excellent high temperature strength,mechanical properties and corrosion resistance.Robot grinding is a complex process with non-linear and strong coupling,and is increasingly demanding material removal rate and surface quality.Therefore,how to establish an accurate grinding removal model has become the key to intelligent grinding.Based on the widely used Inconel 718 nickel-based superalloy in the aviation field,this thesis first builds an intelligent and flexible robotic grinding and polishing system.A series of grinding experiments is then conducted to study the influence of grinding parameters on the material removal rate.Sound signals during the test were collected to study the correlation between the grinding sound and the material remove amount.Time-domain and frequency-domain analysis are applied on grinding sound signals.It is found that due to the reduction of the belt grinding capacity and the accumulation of grinding heat,characteristic signals such as amplitude,average energy,standard deviation,root mean square,and kurtosis factor are gradually weakened as a whole.Besides,the noise during the grinding process is mainly concentrated in the low frequency band(frequency lower than 9KHz),while the stable grinding sound frequency is between 9KHz to 14 KHz.In addition,a six-layer db3 wavelet decomposition is used to further extract the features of the original sound signal.The results show that the seven signal components D1(10000 ?20000Hz),D2(5000 ? 10000Hz),D3(2500 ? 5000Hz),D4(1250 ?2500Hz),D5(625 ? 1250Hz),D6(312.5 ? 625Hz),A6(0 ? 312.5Hz),except for the A6 and D4 frequency bands,the remaining signals have a great correlation with the original signal.Based on the above analysis,this thesis calculates the root mean square values of the five components D1,D2,D3,D5,and D6 after wavelet decomposition as input features for the prediction of the amount of abrasive belt grinding material removal.Several material removal prediction models are established based on multiple machine learning algorithms such as Support Vector Regression(SVR),Optimal Pruning Extreme Learning Machine(OP-ELM),Random Forest(RF),XGBoost,etc.The Mean Absolute Percentage Error(MAPE)is used to evaluate each prediction model.The results show that the prediction accuracy varies from high to low are: XGBoost,RF,OP-ELM,and SVR,and the corresponding MAPEs are 4.37%,4.87%,7.47%,and 19.24%.The results show that the XGBoost model proposed in this paper has the smallest error and the best effect in predicting material removal rate of Inconel 718nickel-based superalloy robot grindingFinally,the influence of grinding parameters on the surface properties of workpieces is analyzed quantitatively,including roughness,hardness,residual stress,and surface burn.It is found that robot grinding is a complex process with multiple parameters and high nonlinearity.The larger the belt size,the better the surface roughness of the workpiece.Under a certain number of belts,the surface roughness is positively related to the grinding force,and has nothing to do with the speed of the belt.With the increase of belt size,the surface residual stress gradually changes from tensile stress to compressive stress.With the increase of the belt speed,residual stress first increased and then decreased.The surface temperature of the workpiece contact area during grinding ranges from 374? to 664?,which is positively correlated with the belt speed.With the increase of the grinding force,surface temperature first increases and then decreases,and the workpiece temperature has a strong correlation with the surface burn.After grinding,a work hardening phenomenon occurs.The hardened layer reaches 125 ?m,and the hardened strength reaches 125% of the matrix strength.