Study On Abrasive Characteristics Of Diamond Tools And Its Influence On Workpiece Surface Roughness

Diamond has the characteristics of high hardness and high wear abrasion resistance and is used in the field of precision and ultra-precision machining widely.It plays an important role in the processing of hard and brittle materials such as engineering ceramics,sapphire,silicon carbide,or cemented carbide.In the process of machining workpiece material with diamond tools,rely on the interaction between the abrasive particles and the workpiece remove material to achieve the desired surface quality.Therefore,the characteristics of the diamond grit have a direct impact on the machining results.This paper starts with the experiment of milling YG8 cemented carbide with diamond tools,combines image processing technology and machine learning method to extract and analyze the characteristics of diamond abrasive particles(the height of abrasive particles,the number of abrasive particles per unit area,and the uniformity of abrasive particles distribution).The influence of machining parameters on the characteristics of diamond abrasive particles and the influence of abrasive particles characteristics on the surface roughness Ra was explored.Finally,a prediction model of surface roughness Ra was established to predict the surface roughness of the workpiece.The main conclusions of this paper are as follows:(1)Based on the grayscale features of tool images,a preprocessing method of abrasive particles images is proposed,which combines K-Means clustering and morphological processing.Based on this method,the area accuracy of the single-particle image and multi-particles image segmentation is 96.42%and 97.58%,which can be well applied to the segmentation and extraction of diamond abrasive particles.(2)According to the actual state of the abrasive particles on the tool surface,the model of the protrusion height of the abrasive particles is established,and the GBDT algorithm is used to quickly generate the protrusion height of the abrasive particles,and the average accuracy reaches 97.05%.Based on the segmentation results of single abrasive particles,the average area of single abrasive particles is calculated using statistical methods.Combined with the segmentation results of multiple abrasive particles,the statistics of the numberρ_m of abrasive particles per unit area in the image are achieved by the comparison method,and the average accuracy reaches 96.36%.A method of mesh abrasive particles quantity variance is proposed to quantitatively calculate the distribution uniformity of abrasive particles.The abrasive particles are dotted to generate grids,and the D_d value of the mesh abrasive particles quantity variance is calculated to quantitatively express the distribution of abrasive particles on a two-dimensional plane uniformity.(3)Through the analysis of the experimental results,it is found that with the increase of milling time,the change rate of the average protrusion height H of abrasive particles decreases gradually andρ_m also gradually decreases until it is stable.There is no obvious regular change in D_d but it tends to be stable eventually.With the increase of spindle speed,H,ρ_m and D_d appear to increase.With the increase of feed speed and milling depth,the change of H first increases and then decreases,and the change ofρ_m first increases and then decreases and to a steady trend.D_d has no obvious regular changes.(4)The higher the average protrusion height H of abrasive particles,the greater the surface roughness of the workpiece after machining.The larger the number of abrasive particles per unit areaρ_m,the lower the surface roughness of the workpiece after processing,the greater the variance D_d of the number of abrasive particles in the grid,the more uneven the distribution of abrasive particles,and the greater the roughness of the machined surface.(5)Based on the influence of abrasive particles characteristics on workpiece surface roughness,a roughness prediction model based on Gradient Boosting algorithm is established.The input of the model includes machining parameters and abrasive particles characteristics.By optimizing the parameters of the training process,the average error of the model on the validation set is about 5.5%,which proves that the model has high accuracy in roughness prediction.(6)In order to further realize the application of the roughness prediction model of the Gradient Boosting algorithm,the model is evaluated through verification experiments,and the average error of the model is 6.3%.It shows that the model has good applicability and can accurately predict the surface roughness.