An improved approach to tool life promotion concerning cutting edge microgeometry

The severe wear of cemented carbide tool can lead to problems such as short tool life and poor part surface integrity. Fortunately, it has been found that the tool wear may be reduced by modifying the cutting edge microgeometry. However, there are a few studies that offer specific instructions for this modification. In addition, it is costly to realize this for machining tools through merely experiments. Therefore, in this paper, an improved approach is proposed by combining the finite-element (FE) simulation method with a user-defined wear rate model. To achieve this, a new wear rate model was proposed at first, regarding the wear mechanisms of the carbide tool during machining of Inconel 718 alloy. Then, four sets of carefully designed cutting tests and FE simulations were carried out to calibrate and verify the new model. At last, extensive FE simulations were conducted to investigate the influence of tool edge microgeometry on tool life. The simulated results indicate that the tool life increases with the increase of the form factor K within a given range. This finding on optimized tool edge was further verified by performing a group of cylindrical turning experiments. The experimental measurements reveal that the tool life has increased by 24% and 50% maximum when increasing K from 0.9 to 2.8 under the two given cutting conditions.


Introduction
Nickel-based superalloys have been widely used in aerospace and automotive industries for the excellent mechanical properties like high hardness and strength at high temperatures. However, nickel-based superalloys are also considered as typical difficult-to-machine materials because of the low heat dissipation and high work hardening coefficient. Cemented carbide tools are commonly applied to processing these kinds of materials for their high toughness and good wear resistance. In the processing of nickel-based superalloys, the tool wear of cemented carbide is still severe. This not only affects the processing efficiency and quality but also relates closely to the processing cost [1,2]. Therefore, a reasonable tool wear rate model has become a key factor for the prediction of tool life and tool wear process in metal cutting.
Even nowadays, the prevailing predictive models of tool life are empirical ones. These models are generally obtained through a large number of cutting experiments. Empirical models formulate the interrelationship between cutting condition parameters and the corresponding tool life. For instance, Taylor [3] found an exponential formulation between cutting speeds and tool life. Hastings et al. [4] proposed a tool durability formula involving cutting temperature. It should be noted that the prediction accuracy of these empirical formulas can be affected by many factors including tool materials and tool geometric parameters. Moreover, these models can only predict the life of the tool instead of its wear process. To solve this, researchers have conducted many investigations on the tool wear mechanism and wear rate models. Dong et al. [5] analyzed the tool wear phenomena in machining of Inconel 718 alloy. Base on analysis using scanning electron microscope (SEM) technology, they found that the main mechanisms of tool wear are abrasion, adhesion, and diffusion. Usui et al. [6] proposed a tool wear model to characterize the adhesion wear mechanism.
Rabinowica et al. [7] developed a model to predict the rate of abrasion wear. Some other researchers considered several wear mechanisms, such as Takeyama and Murata [8] who offered a model involving adhesive wear, abrasive wear and diffusion wear. Huang et al. [9] modified Rabinowica's wear model by considering also the three predominant wear mechanisms. Li et al. [10] established an analytical model for the wear mechanism of PCD tools during metal cutting. In their model the effects of adhesive wear and varying cutting forces were considered. Elias et al. [11] predicted the tool flank wear rate during micro-turning by inputting the worn tool profile into the Usui wear model. They noticed that the modified model had better accuracy in predicting tool wear rate than the traditional one [12]. The tool wear model proposed by the researchers above can only predict the life of the tool instead of its entire wear process, namely the wear profile evolution of the tool. However, it is almost impossible to analytically realize this because the wear profile evolution involves numerous non-uniformly distributed factors like temperatures, speeds, and stresses in the vicinity of tool cutting edge.
Whereas the state variables such as cutting temperatures, stresses, and speeds are difficult to obtain, it seems promising to combine the analytical wear rate model with finite element method (FEM) for cutting process which can yield all these state variables simultaneously. In recent years, the rapid progress of computer science has facilitated the wide application of FEM in the field of metal cutting. With the help of FEM, Xie et al. [12] predicted the wear progress of cemented carbide tool. They found that the predicted values fitted well with the experimental ones. Hosseinkhani et al. [13] investigated the wear rate of tool during the cutting process using the Usui model. Yadav et al. [14] continuously updated the tool geometry in the FE simulation using DEFORM-3D. Their method was capable of calculating material removal rate and tool wear at an assigned time. Attanasio et al. [15] combined the Usui model [6] with the Takeyama model [8] in order to consider both abrasive wear and adhesive wear. Then, a new wear model was proposed for 3D wear simulation analysis. The new model provided reasonably good predictions in terms of maximum flank wear width and crater depth. Binder et al. [16] computed the adhesive wear rate by considering the thermo-mechanical loads around the tool. Accordingly, the tool geometry was modified with a user-defined subroutine. Nooraie et al. [17] obtained the average temperature of all nodes in flank wear zone by FE simulations. Then, the flank wear was estimated with the temperature and the Usui model. The attained width of flank wear agreed well with the experimental measurements. Wang et al. [18] established a new wear rate model that combined the tool wear rate equations developed by Usui [6], Takeyama [8] and Attanasio [15]. These approaches mentioned above shares the same advantage in predicting tool wear/life using simulated date. Meanwhile, the disadvantage is exposed in providing the tool wear profile evolution and considering specific tool edge microgeometry.
As has been analyzed above, most of the researches on wear model focus merely on tool wear/life prediction. However, the wear model has been applied seldom to investigating effects of the edge microgeometry on both tool wear progression and the tool life. In this study, the wear mechanism of carbide tools in cutting of Inconel 718 alloy was analyzed by orthogonal cutting experiments. As a result of this procedure, a modified wear rate model was proposed and calibrated by considering both abrasive and adhesion wear. Then, the material parameters in the model are obtained through the combination of extensive cutting experiments and FE simulations. A good agreement was found among the comparisons between the experimental results and the simulated ones. Then, the verified model was used to analyze the interrelationship between edge parameters and tool life. The optimized tool edge microgeometry design was determined and validated through extended cylindrical turning experiments. Notable promotion of tool life was revealed from comparisons between the standard tool and optimized ones. This paper is composed by five sections. Following a brief introduction of the previous studies on tool wear prediction, an improved approach was developed in the second section based on secondary development of FEM. In that approach, a new wear rate model was proposed and implemented. The corresponding calibration procedure for the presented wear rate model was conducted in the third section. Then, this improved approach was applied to the in-depth investigation of influences of edge microgeometry on tool life. The findings and contributions are drawn in the last section.

Identification of tool flank wear land
The tool flank wear is usually recognized as the criterion of tool failure since it affects directly the dimensional accuracy of machined parts. Figure 1 shows the formation of the flank wear zone. In practical wear measurement, the main wear land is divided into three zones. The first zone C is the curved part of the cutting edge at the corner. The second zone N is the worn cut of length. The third zone B is the rest straight portion of the cutting edge. VB max is the maximum wear width which is computed in zone B. VB N is the notch wear that is measured in zone N. In zone C, the corner wear VB C is obtained. Therefore, in the actual cutting process, it is usually recommended to use the following two tool failure criteria for cemented carbide tools, namely the average width VB B or the maximum width VB max at the flank wear area [19]. Assuming that the two-dimensional (2D) tool wear area is in zone B, the threshold VB B = 0.2 mm is determined as the tool failure standard [20]. During the machining process of Inconel 718 alloy, the tool wear is not caused by a single wear mechanism. Instead, there are several mechanisms of wear existing at the same time. It should be noted that in both actual cutting tests and cutting simulations of Inconel 718 alloy with recommended parameters, the cutting temperature will not exceed 700 °C when the speed of carbide tool is ranging from 30 to 50 m/ min [21]. What is more, it was found that when the cutting temperature is lower than 700 °C, the predominant wear mechanisms are abrasive and adhesive wear [22]. If the cutting temperature exceeds 700 °C, the predominant mechanism changes to the diffusion [23]. Therefore, in this study, the wear rate model is proposed by considering both abrasive wear and adhesive wear as where ẇ is wear rate in a unit of mm 3 /s. The variables p, v s , T are the contact pressure, the relative velocity at toolworkpiece interface and the local temperature, respectively. A, B and G are constants which depend on the tool and workpiece material.
In the actual cutting process, the ultimate tool flank wear land width is easy to measure. However, it is rather difficult to measure the volume loss during tool wear process. Therefore, a formula between average flank wear width and volume loss it to be derived.
The flank wear land has been reported to be slightly inclined (see Fig. 2). If this inclination angle β is unchanged [20], the area loss at cutting time t 1 is put as where α is the clearance angle. γ is the rake angle, and VB 1 is the flank wear width at time t 1 . The area loss for an arbitrary time t 2 (t 2 > t 1 ) is determined similarly. The area loss throughout a time interval Δt = t 2t 1 can be calculated as where A VB1 and A VB2 are the tool flank wear areas at time t 1 and time t 2 , respectively. Considering VB 1 = VB 2 -ΔVB and VB 2 = VB, Eq. (3) can be simplified by ignoring the highorder terms as The variable VB in Eq. (4) stands for the flank wear width at an instant time t. In orthogonal cutting, the entire length l tot equals approximately the depth of cut a p . Thus, the volume loss ΔV within Δt is given by Then, the relationship between tool wear volume loss rate dV/dt and flank wear length growth rate dVB/dt can be expressed by Equation (6) can be used to calibrate the tool wear rate model in Eq. (1) by adopting the measurements from FE simulations and the tool wear tests with various cutting parameters, which was illustrated in third section.

Numerical modeling of cutting process
The 2D FE model of cemented carbide tool in orthogonal cutting of Inconel 718 was established as depicted in Fig. 3. The rake angle α and relief angle β of the tool are both 10°. The cutting edge radius r is 0.02 mm. In order to balance the accuracy and efficiency of the simulations, the maximum mesh size, the minimum mesh size, and the meshing gradient of the tool are set as 0.3 mm, 0.01 mm, and 0.8, respectively. The mesh near tool edge is refined to 0.002 mm in order to identify the edge microgeometry of the tool. The size of the workpiece is 2 mm times 0.5 mm. It moves towards the tool with an assigned speed V. An initial temperature field of 20 °C is set to both tool and workpiece, neglecting the heat convection with air. The tool material is selected as WC-Co cemented carbide. The workpiece material is Inconel 718 which is characterized by the Power-Law material constitutive model. This model reflects the constitutive behaviors of metal materials under large strains, high strain rates, and high temperature loadings.
The constitutive models for the workpiece material are given in Eqs. (7-10). Among Eqs. (7-10), σ (ε p ,̇,T) denotes the flow stress of the workpiece. g(ε p ) represents the strain strengthening function. T ( ̇ ) is the strain rate effect function, and (T) is the thermal softening function. In Eq. (7), ε p is the strain during the material deformation process. ̇ means the strain rate during the material deformation process, and T is the temperature during the material deformation process. σ 0 in Eq. (8) is the initial yield stress. p 0 is the parametric plastic deformation strain. n is the plastic deformation coefficient. In Eq. (9), ̇0 is the reference plastic strain rate. m is the strain rate sensitivity coefficient. c 0 ~ c 5 in Eq. (10) stands for the thermal softening parameters. The mechanical properties for the workpiece material Inconel 718 are shown in Table 1. The parameters for material thermal softening model are specified in Table 2. Table 3 offers its thermal properties.
During the cutting process, there are frictional contacts at both chip-tool-rake interface and workpiece-tool-flank interface due to the presence of tool flank wear land. To describe the frictional contact condition, the Coulomb friction model is adopted by where F f is friction force. F n is the normal force, and μ is the friction coefficient which was set to 0.6 [22].
Tool wear is the result of pressure, temperature, and relative slip velocity near the tool cutting edge. For each wear time increment, the tool wear calculation is performed only when the cutting process reaches its steady state. To ensure this, the tool temperature distribution was obtained by simulating a cutting length of 3 mm as given in Fig. 4a. The results were then extracted to Fig. 4b with five points (P i , i = 1 ~ 5) along the tool flank face. Figure 4b shows that the tool surface temperature becomes stable when the cutting (11) F f = × F n length reaches 1.5 mm. Other simulated results such as stress and slip velocity are extracted and analyzed in the same way.

Design of calibration tests
The chemical composition of Inconel 718 alloy is provided by the supplier and is shown in Table 4. As depicted in Fig. 5, the workpiece is prepared into numbers of separated thin webs in a width of 3.5 mm out of a cylinder with a diameter of 130 mm. The cutting parameters of the tests are shown in Table 5. As seen, there are two groups of experiments used for calibration while the other two for verification.
During each group of cutting tests, the profile of tool flank wear is recorded by optical microscope after machining every single web. Therefore, the tool life can be calculated by the number of webs that have been cut until the flank wear length exceeds the failure criterion. Each set of cutting tests was repeated three times. The average wear length from repeated measurements was taken for calibration.

Determination of coefficients in tool wear rate model
To describe the flank wear progression along cutting time, the tool with the holder is unloaded together and then put into the customized fixture to observe the wear land profile under an optical microscope (Keyence, VHX-7000) nearby. The length of flank wear land can be identified from both image and the scanned worn edge profile.   The tool wear results from the four groups of tests were summarized in Fig. 6. The fitting is performed to obtain slopes at different wear stages. A three-stage progression can be noticed from the wear results shown in Fig. 6a-c, except the curve in Fig. 6d that stopped right near the wear width of 0.2 mm. With the fitting among the measurements, the tool wear rate can be estimated using Eq. (6). At last, the calculated tool wear rates from the two calibration tests are given in Fig. 7.
On the other hand, a series of cutting simulations were performed with same cutting parameters. To represent different tool wear stage, tools with five custom flank wear widths, i.e., 50 μm, 100 μm, 150 μm, 200 μm, and 250 μm, were employed as collected in Fig. 8b-f. Extractions of the simulated data were included in the way shown in Fig. 4. Figure 9a and b show the average results of temperatures and normal stresses along the tool wear land. In the cutting simulations, the relative speed v s between the tool worn    process in practical machining. With the change of tool geometry, the temperatures and forces also change especially near the cutting edge. However, considering the calculation efficiency, the length of the 2D cutting simulation cannot be too long: otherwise, the wear simulation of the entire cutting process cannot be accomplished. Therefore, in the wear simulation module of Third Wave AdvantEdge, the tool wear simulation is realized by finite discretization of tool wear process, namely several wear increments out of total wear time. In tool wear simulation, the main factors affecting the accuracy of tool wear simulation are the increment of tool wear time and the meshing of tool and workpiece. However, the decrease of tool wear time increment will bring sharp increase to the calculation time. After a number of attempts The procedures involved in the tool wear process simulation approach is demonstrated by a flowchart in Fig. 10. It can be found that with those extractions of steady state cutting process variables, the tool wear rate is computed by calling the packaged wear subroutine. The displacement of each node is calculated according to the tool wear time increment, based on which the tool profile is finally updated. Finally, the new tool geometry participates in the next wear iteration. These iterative steps will not stop until the wear time reaches the total wear time (t tot ). With the tool failure criterion, the tool life and tool wear process can be determined from the outputs.
To further confirm the accuracy of the calibrated model, results from additional tool wear simulations are compared with those from the two verification tests (see Table 5). Figure 11 shows the simulated tool topographies during the wear process under cutting condition No. 2. The volume loss of the tool was presented by the presence of initial and worn contours.
It can be found from Fig. 11a to b that tool wear land became very clear after only 1 min. In the next 1 min, however, the wear land did not grow too much by comparing Fig. 11c with 11b. This time may correspond to the stable wear stage where tool wear becomes modest. In Fig. 11d, the tool flank wear width exceeded 0.2 mm, and the tool therefore failed. It can also be noticed that tool wear happened also on the rake face. This is actually in accordance with the observations during the cutting experiments.
To show the f lank wear progression, the obtained f lank wear results are all plotted in Fig. 12 where measurements from experiments are compared with those from the simulations. The difference between these results is demonstrated by the error bars. The average error is estimated less than 10%. It implies that the developed approach can estimate both the life and wear process of the tool, which verified the validity and accuracy of the proposed tool wear model.

Effects of cutting edge microgeometry
Even though the cutting edge microgeometry has shown notable influence on the tool life, part surface integrity, and cutting efficiency [25], it has seldom been addressed using numerical approach. On the one hand, the tool edge microgeometry is complex. The edge radius is found unable to characterize accurately the microgeometry of cutting edges with asymmetrical passivation. Therefore, an accurate characterization of tool edge microgeometry is urgently needed for studying its interrelationship with the tool life. Figure 13 provides the prevailing characterization of the edge microgeometry of different cutting edges [23]. Those two parameters, S γ and S α , denote the length of honed segment on rake and flank face, respectively. Δr stands for the minimum distance between the edge surface and the assumed tool tip, which is not adopted here since it is almost impossible to control with current passivation technologies. Instead, the two segments are usually employed in describing this kind of cutting edge. A form factor K, defined as S γ /S α , is introduced for convenience. It distinguishes the edge into two kinds, namely the waterfall edge (K < 1) and horn-shaped cutting edge (K > 1). In this section, both of the  Table 6 specifies the 10 sets of cutting edge microgeometries used in the FE simulations. Three of them are depicted in Fig. 14. Figure 15 gives a comparison of the simulated flank wear widths of the entire 10 tools after 2 min (wear time). The rounded tool (tool No. 2 in Table 6) is the standard tool edge design which is treated as a reference for tool life promotion. A difference can be noticed among these tools, which means the edge microgeometry made a difference to the tool wear process. It is also found from Fig. 15 that tool No. 9 has the lowest flank wear width, which turned out having the longest tool life as well. On the contrary, the shortest tool life was observed with tool No. 8 and tool No. 10. To signify the variation tool life, these results are presented in a diagram as shown in Fig. 16.
It is found from Fig. 16 that maximum tool life of nearly 3 min was realized by tool No. 9 (S γ = 100 μm, S α = 50 μm, K = 2). Compared to the life of reference tool, it raised over 20%. The shortest tool life was detected in tool No. 10 (S γ = 100 μm, S α = 100 μm, K = 1). This could be ascribed to the increased cutting and thrust forces due to such a large edge radius, regarding the feed rate of the same magnitude (f = 0.1 mm/rev). It is surprising that tool No. 1 (S γ = 15 μm, S α = 15 μm, K = 1) yielded slightly longer tool life than the reference tool, despite that it is sharper. In practical machining, the tool with such a sharp edge would face chipping in the cut-in stage. However, it is unlikely to predict the chipping by current wear rate models. In spite of this, the diagram of tool life implies that tool with horn-shaped edge (K > 1) performs better. From the other simulated data, it is inferred that this kind of tool benefits from the improved material flow states and ensuing distribution of temperatures and stresses around the tool edge [26].

Validation of optimized edge microgeometry combination
To verify the effectiveness and applicability of the finding about tool life promotion, extensive turning experiments were conducted in this section. Three cutting tools (mode: CNMG120408) prepared with different K values were used in this cylindrical turning experiments. The tool with shape factor K = 1 was the reference tool. The microgeometry parameters of the cutting edge of the turning tool were    Fig. 17. Measurement of each tool was repeated three times, and the average value is adopted and summarized in Table 7. Two sets of cutting parameters are included in the turning experiments. One group is with a cutting speed V = 30 m/min, feed speed f = 0.12 mm/rev, and cutting depth a p = 1 mm. The other group is with a cutting speed V = 25 m/min, feed speed f = 0.25 mm/rev, and cutting depth a p = 2.5 mm. The experiments are conducted involving water-based coolant. Recorded morphologies of tool flank wear progression from the first group are presented in Fig. 18. Figure 18 shows the variation of flank wear profiles of three different cutting edges with evenly increased time. Compared with the other two improved tools (B, C), the standard tool (A) went failure early in 30 min (see Fig. 18c). In the time of 40 min, a large breakage can even be noticed at the tool corner. Meanwhile, the tool with optimized edge (tool C) was still serving, and no apparent breakage was observed on it.
The results of tool life from those two groups of experiments are shown in Fig. 19. Compared with the standard tool (marked by A), the tool lives with optimal edge microgeometries (marked by C) have increased by 24% and 50% maximum when increasing K from 0.9 to 2.8 from the two cutting conditions, respectively. Therefore, the influence of the tool edge microgeometry on tool life is verified. Besides, the approach presented in this paper managed to promote the tool life through edge microgeometry optimization.

Conclusions
This study presented a new approach to visualizing tool flank wear progression and optimizing serving life for cemented carbide tools during machining process of Inconel 718 alloy. Extensive cutting simulations and experiments were performed to validate effectiveness and reasonableness of   (1) A predictive approach to tool flank wear prediction was established based on a modified wear rate model that considers both abrasive and adhesive wear. It was calibrated and implemented by a user defined subroutine.   With the presented approach, the tool wear progression was visualized and tool life was accurately predicted. (2) The influence of cutting edge microgeometry on the tool life was investigated and diagramed using the proposed approach, which indicates that the microgeometry parameters could have significant effects on tool life. It is further found that the tool life increases with the increase of the shape factor K within certain range given in this paper. (3) The applicability of the proposed approach to tool edge microgeometry optimization and tool life promotion was validated through additional cylindrical turning experiments of Inconel 718 alloy. The results show that compared with the reference tool, the life has increased by 24% and 50% maximum when increasing K from 0.9 to 2.8 for the two cutting conditions, respectively.
Even though the presented approach finally managed to promote noticeable tool life as a result of optimizing tool microgeometry, it should still be worth noticing that it was tested under limited tool-workpiece combination, and the proposed wear rate model is not sufficient in incorporating all the wear mechanisms involved during cutting process. Moreover, the wear on tool rake face parallels that on flank face, which is not thoroughly demonstrated in the developed approach. Therefore, there are more improvements undergoing from these aspects.