- Email: [email protected]
Reliability of electrode wear compensation based on material removal per discharge in micro EDM milling
CIRP Annals - Manufacturing Technology 62 (2013) 179–182
Contents lists available at SciVerse ScienceDirect
CIRP Annals - Manufacturing Technology jou rnal homep age : ht t p: // ees .e lse vi er . com /ci r p/ def a ult . asp
Reliability of electrode wear compensation based on material removal per discharge in micro EDM milling G. Bissacco (2)a,*, G. Tristo b, H.N. Hansen (1)a, J. Valentincic c a
Department of Mechanical Engineering, Technical University of Denmark, Kgs. Lyngby, Denmark Department of Industrial Engineering, University of Padova, Padova, Italy c Faculty of Mechanical Engineering, University of Ljubljana, Ljubljana, Slovenia b
A R T I C L E I N F O
A B S T R A C T
Keywords: Electrical discharge machining (EDM) Wear Micromachining
This paper investigates the reliability of workpiece material removal per discharge (MRD) estimation for application in electrode wear compensation based on workpiece material removal. An experimental investigation involving discharge counting and automatic on the machine measurement of removed material volume was carried out in a range of process parameters settings from fine finishing to roughing. MRD showed a decreasing trend with the progress of the machining operation, reaching stabilization after a number of machined layers. Using the information on MRD and discharge counting, a material removal simulation tool was developed and validated. ß 2013 CIRP.
1. Introduction Micro EDM milling has become in the recent years an established process for the manufacturing of 2½D and 3D micro components and for the generation of micro features on larger components. In such a process part geometry is obtained by material removal in discrete units by means of controlled discharges occurring between electrically conductive tool and workpiece electrodes. Application materials are mostly metals and alloys. Recent research has demonstrated the applicability of the process to conductive ceramics [1]. In EDM milling the tool electrode consists of a cylindrical rod driven along defined paths similarly to a milling tool, and material is removed layer by layer with layer thickness as low as 0.1 mm. The minimum machinable feature size, achievable accuracy and surface roughness are largely affected by the minimum achievable material removal unit [2]. As the amount of the workpiece electrode removed by each discharge is determined by the discharge energy distributed to the workpiece [3], small material removal units are obtained with small discharge energies, which involve very short discharge pulses. Factors affecting the limits of miniaturization in micro EDM are presented in [4]. Tool electrode wear changes the relative position of tool and workpiece and must be compensated for in order to maintain high machining accuracy. Since milling EDM is performed in very thin layers compared to the tool diameter, the tool profile stabilizes rapidly and wear correction can be done by a one dimensional motion parallel to the electrode axis. Recent electrode wear compensation approaches use either an estimation of the volumetric wear ratio with a continuous downward electrode movement proportional to the relative inplane displacement (anticipated tool wear compensation) or real
* Corresponding author. 0007-8506/$ – see front matter ß 2013 CIRP. http://dx.doi.org/10.1016/j.cirp.2013.03.033
time compensation based on discharge counting. Furthermore real time compensation can be based on full discharge pulse discrimination or on discharge counting and statistical treatment of the discharge population [5]. Real time compensation was implemented by Bleys [6] in conventional (macro) EDM milling using an isoenergetic pulse generator. Aligiri et al. [7] have proposed a method for tool wear compensation in micro EDM drilling based on real time full discharge discrimination and single crater size prediction using a thermal model. However discharge pulses in micro EDM are not isoenergetic and full discharge discrimination is difficult due to the very short pulse duration which can be as low as 20 ns. Jung et al. [8] have implemented real time tool wear compensation based on discharge pulse monitoring and pulse frequency control. Bissacco et al. [5] recently proposed an approach based on discharge counting and statistical characterization of discharge population for tool wear compensation in micro EDM milling. Such an approach was applied to electrode wear compensation based on Tool Wear per Discharge (TWD) estimation. Implementation of tool wear compensation in micro EDM milling based on statistical characterization of the discharge population and workpiece Material Removal per Discharge (MRD) estimation has not been realized yet. The implementation of such an approach requires accurate and reliable estimation of MRD in various processing conditions. This paper introduces such an approach and investigates the reliability of on the machine workpiece MRD as the most critical contribution to the accuracy of the method. For this purpose an experimental investigation, involving discharge counting and automatic on the machine measurement of removed material volume, was carried out in a range of process parameters settings, from fine finishing to roughing. On the machine measurements of removed material volume were validated by means of confocal microscopy. On the basis of the obtained results, the observed trends for MRD at increasing machining depths are discussed and an analysis of the
180
G. Bissacco et al. / CIRP Annals - Manufacturing Technology 62 (2013) 179–182
measurement error contributions is presented. Furthermore a material removal simulation tool for accurate representation of machined geometry is developed and experimentally validated. 2. Tool wear compensation based on MRD 2.1. Discharge population approach In micro EDM considerable variation of discharge pulses is observed in terms of peak current, discharge duration and discharge energy, even for constant process settings and stable machining process. Therefore the workpiece material removal associated with each single discharge differs and should be assessed individually. The population approach described in [5] is based on the considerations that for a stable process the discharge population, namely the statistical distribution of discharge characteristics such as the discharge energy, is expected to be stable, and that material removal occurs by means of trains of discharges with identical distribution as that of the entire population. Therefore a MRD of the population can be obtained and attributed to each discharge of a sufficiently numerous discharge train. On the basis of this consideration, real time material removal can be estimated by counting the discharges and multiplying by the average MRD of the population. This approach greatly simplifies the task of real time material removal estimation for ultra-short pulses (less than 100 ns) such as those used in micro EDM milling finishing operations since it does not require discharge discrimination but rather discharge counting. 2.2. Implementation of MRD based compensation The sketch in Fig. 1 shows the tool motion and material removal for an undercompensated linear trajectory segment with full radial engagement. As a consequence of tool wear, the vertical level of the machined surface increases along the tool trajectory. For the implementation of material removal based tool wear compensation, occurring discharges during a machining operation are continuously counted. The number of counted discharges Nd for each elemental tool trajectory segment is multiplied by the average MRD of the discharge population for the given tool– workpiece–dielectric fluid combination and process parameters settings, thus obtaining the actual removed volume for the considered trajectory segment (indicated as ‘‘Removed material’’ in Fig. 1). By subtracting this volume to the volume that would have been removed in the same tool path segment if no tool wear occurred, the residual volume for the elemental path segment is obtained. By choosing sufficiently small elemental tool path segments, the variation of the surface level along the elemental path segment can be linearized as shown in Fig. 1 and thereby the angle a and the corresponding part depth variation Dz can be calculated. In order to correct for the detected undercompensation, such depth variation Dz will be added as a downward displacement at the end of the elemental path segment. Furthermore an inclination a will be added to the next path segment in order to obtain the correct depth along the whole path segment. In this way, the surface for the next path segment is maintained at the
correct level without updating the information of the absolute vertical position of the frontal surface of the tool electrode. Obviously the initial vertical position of the tool electrode must be known to initiate the process. For the implementation of such compensation approach it is critical to correctly estimate the volume of material removed at each path segment and therefore the accuracy of estimation of MRD is critical. 3. Stability of MRD The above described electrode wear compensation method requires the estimation of the MRD for the specific process parameters settings and tool–workpiece–dielectric fluid combination to be used in a machining operation. Therefore, for implementation of the method in an industrial environment, standard routines must be developed for the assessment of MRD on the machine. Furthermore, accuracy of MRD determination and MRD stability over time (throughout a machining operation) is critical for the compensation accuracy using the described method. 3.1. MRD experimental determination The experimental setup for the determination of the MRD consisted of a Sarix SX-200 micro EDM milling machine equipped with a laser scan micrometer by Mitutoyo for optical measurement of electrode profile. Tungsten carbide cylindrical rods with diameter of 300 mm were used as tool electrodes, while the workpiece material was a martensitic stainless steel (Uddeholm Stavax ESR). Discharge counting was performed on the current signal by means of a current probe developed in house and a digital counter HP 53131 A with a frequency range from DC to 225 MHz. Experiments were performed for three different process parameters settings, corresponding to different pulse shapes and with average discharge energy of 0.7 mJ, 12.6 mJ and 57.9 mJ. For the experimental determination of the MRD, automatic routines have been developed by preparation of a parametric part program customizable by the user through the machine console. In the experiments described here, the program was used for the generation of cylindrical cavities by removing material layer by layer in milling mode, with layer thickness of 0.5 mm, 0.9 mm and 2.5 mm for the three process parameters settings. The cavities were 500 mm in diameter and were generated by means of circular interpolations in 10 incremental steps with actual depth of approximately 50 mm per step (total 500 mm), with discharge counting and recording of the cumulated number of discharges for each step. At the end of each incremental step, the diameter and the depth of the cavity were measured on the machine using the electrode as a touch probe in a Coordinate Measuring Machine. Diameter measurements were performed at a depth level of 3 mm above the bottom of the cavity in order to take into account the conicity of the machined hole in the estimation of MRD. The diameter measurement procedure consisted in probing 8 points on the hole circumference. For improved measurement reliability, each point was probed three times. The depth of the cavity was measured by probing its bottom at the center with the electrode. The probing was repeated three times. For each of the three process parameters settings investigated, the test was repeated three times. The results are reported in Fig. 2. 3.2. Validation of on the machine measurements
Fig. 1. Sketch showing tool motion and material removal for undercompensated linear trajectory segment.
The accuracy of on the machine measurements of diameter and depth strongly affects the accuracy of MRD. Therefore on the machine measurements of removed material volume were validated by means of confocal microscopy in dedicated experiments. The comparison was carried out on three cavities machined with the three different process parameters settings previously used for MRD determination. After machining of the three cavities, a 200 mm electrode was generated on the machine using the available WEDG module and measured by means of the laser scan
G. Bissacco et al. / CIRP Annals - Manufacturing Technology 62 (2013) 179–182
181
are in line with the scatter of the MRD measurements shown in Fig. 2. 3.3. Analysis of MRD variability
Fig. 2. Measured MRD values at increasing depths for the three process parameters settings considered. Data points represent the average of three repetitions. Error bars represent the ranges of MRD values for each step depth.
micrometer to be 198.4 mm. The generated electrode was used to probe the circumference and the bottom of the machined cavities using the same procedure as described in the previous section and repeating the measurements 3 times. The cavities were also measured using a confocal microscope Sensofar PLm Neox. A 100 magnification, corresponding to an in plane resolution of 0.17 mm and a vertical resolution of less than 2 nm was used for the generation of image files for measurement of diameter. For measurement of cavity depth, a 20 magnification, was used, yielding an in plane resolution of 0.83 mm and a vertical resolution of less than 20 nm. Due to the almost vertical walls of the cavities, measurement of the diameter of the holes was possible on the images from the confocal microscope only at very low depths. Therefore diameter measurements for both on the machine and confocal microscope measurements were performed at 3 mm below the part surface. The results of the comparison are shown in Table 1. Using the confocal microscope measurements as a reference, the error on the average depth is in the order of 1 mm and on the diameter it is between 2 mm and 3 mm. Such deviations appear to be systematic errors and are likely due to the roughness of the surface of the cavities as the tool electrode probes on the peaks of the cavities lateral and bottom surfaces. The observed depth and diameter errors correspond to a maximum expected volume measurement error of 3% for a step depth of 50 mm as used in the experiments for the determination of the MRD. Such errors, although not negligible,
Table 1 Results of validation of on the machine measurements. Uncertainty at 95% coverage probability on depth and diameter measurements performed on the machine is 1.3 mm and 1.9 mm respectively. Depth (mm)
The results of the experimental tests for MRD determination are shown in Fig. 2. Clearly, higher average discharge energies yield higher MRD. From the observation of the diagrams it is also visible that for constant energy settings the measured MRD data show higher scatter between the test repetitions at lower depths. Furthermore MRD is not constant during machining, with a decreasing trend with the progress of the machining operation, reaching stabilization after a certain depth depending on the energy level. This result is in accordance with results reported in [7] for drilling, where a correction factor was introduced to account for the decreasing material removal efficiency. However in drilling EDM the flushing efficiency decreases strongly with increasing depth, affecting the observed MRD. At higher depths, where MRD reaches stabilization, the variation between repeated MRD measurements is approximately 1% for the lower energy levels and 4% for the highest energy level. While the observed scatter of the MRD measurements can be regarded as acceptable, the decreasing trend observed at the lower depths makes the implementation of tool wear compensation based on MRD difficult. The reasons for such a trend are not completely understood. The range of measurement errors for on the machine volume measurements, as reported in Table 1 are not compatible with the observed variation of MRD with increasing machining depth. An analysis of the on the machine measurement procedure yields the following considerations. The cavity step depth is measured in the center of the cavity while a rounding of the bottom occurs at the periphery. Consequently the volume for the first step is over estimated, leading to a larger apparent MRD. Such overestimation occurs only for the first step, while for the following steps it is compensated by the excess material remaining as a consequence of the rounding effect in the previous step. The rounding radius is larger for higher average discharge energies, thus the absolute error on the MRD measured on the first layer is larger at higher energies. However the relative error on the measured volumes, and thereby on the calculated MRD, decreases with the increase of the total removed volume per step. For the cavity diameter values and actual step depth used in the experiments presented above, and for the highest energy settings (showing a cavity rounding radius of 12 mm), the calculated relative error contribution on the step volume, and therefore on the MRD value, due to this effect is approximately 0.33%. In addition to the above error, the rounding of the tool electrode induces an overestimation of the cavity diameter for the first step. With the diameter values used in these experiments, the maximum corresponding error is approximately 4% on the measured volume for the first step. The discussed error contributions are not sufficient to justify quantitatively the decreasing trend of the MRD observed for the process settings with the highest average discharge energy as shown in Fig. 2, which is therefore expected to be a characteristic of the process, most probably due to the discharges that occur on the cylindrical surface of the electrode while machining at greater depths. Due to the above considerations, tool wear compensation based on MRD is expected to be most efficient in connection with process settings involving lower average discharge energies. In such cases, MRD stabilization is reached quickly with the progress of the machining operation, so that an average MRD value, corresponding to that in the stable region, can be used for compensation. 4. Material removal simulation
Diameter (mm)
Energy level
E1
E2
E3
E1
E2
E3
On machine Confocal Error
57.6 58.4 0.8
63.4 64.6 1.3
60.9 62.0 1.1
512.7 510.9 1.8
519.9 517.9 2.0
524.1 521.0 3.1
4.1. Material removal simulation tool A tool for material removal simulation has been developed in Matlab environment. The tool allows accurate representation of
182
G. Bissacco et al. / CIRP Annals - Manufacturing Technology 62 (2013) 179–182
the machined geometry which can be used for verification of the progress of the machining operation as well as for improved tool trajectory definition and tool wear compensation for the following process steps. The input to the simulation tool consists of a three columns array where the first and second columns contain the recorded x and y positions of the center of the tool electrode at defined path intervals, while the third column contains the corresponding number of counted discharges. Within the simulation module, the path interval is divided in any preferred number of segments and the amount of discharges corresponding to each segment is calculated by dividing the discharges counted for the whole path interval by the number of segments. Tool electrode and workpiece are discretized in a number of 2½D voxels with square footprint equal to the average area of the discharge craters measured for the specific process settings. It is important that the voxel size is chosen to be close to the average crater size. As the volume removed by a single spark is constant and equal to the MRD, the amount of reduction in height depends on voxel lateral dimension. If voxels are too small the voxel height reduction caused by a single discharge is exaggerated and the accuracy of the simulation is compromised. The tool electrode is considered in the simulation as having a diameter equal to the actual diameter plus two times the lateral discharge gap (which is determined experimentally). For each segment, the couple of voxels on tool and workpiece that have minimum distance are identified and the Z coordinate of the workpiece voxel is reduced by an amount equal to the MRD divided by the voxel area. Thus the volume removed from the voxel corresponds to MRD. If more than one tool– workpiece voxels couples with the same distance are found (as happens at the beginning of the operation), a random choice is operated automatically. The output of the simulation tool consists of a file of the machined part in STL and SDF formats. The STL file can be imported in a generic CAM software for the generation of the tool trajectory for the following machining operation. The SDF file can be imported in programs such as SPIP for measurement of part geometry. 4.2. Validation of material removal simulation tool Experimental validation of the material removal simulation tool was obtained by means of machining tests with comparison of machined features and simulation results. The generated features were a straight groove 500 mm in length, with full radial engagement of the tool electrode (slotting) and a square pocket with a square island. In both tests the tool electrode was dressed using the machine WEDG unit to a diameter of 200 mm and process parameters settings corresponding to the intermediate average discharge energy E2 (11.5 mJ) were used. The discharge gap was measured to be 5.5 mm by means of Confocal and SEM measurements on dedicated tests, while the crater diameter was measured to be approximately 10 mm. No tool wear compensation was applied and discharges were counted during machining. For the simulation, the actual tool path segments, counted discharges, tool diameter and lateral sparking gap were given as input. MRD was assumed constant and equal to 12.4 mm3 as results from the stabilized region in Fig. 2. The groove was machined in 10 steps, each consisting of 20 layers with nominal layer depth of 0.9 mm, for a total nominal depth of 180 mm (actual depth 77 mm). At the end of each step of 20 layers, the number of discharges was recorded and the groove floor was measured at half the path length by means of touches using the tool electrode. Machining was simulated layer by layer, with a motion resolution of 10 mm in x–y directions. Table 2 shows the comparison between the groove depth as measured on the machine and the corresponding values from the simulation for each of the steps of 20 layers each. An overall good agreement between measured and simulated material removal is observed. With the exception of the first layer, the maximum depth error of the material removal simulation is 1.2 mm for a total depth of 77 mm. The error in the first layer is
Table 2 Results of validation of material removal simulation in slotting. Step no.
1
2
3
4
5
6
7
8
9
10
Actual depth (mm) On the machine 0.7 14.3 21.8 29.5 37.6 46.5 53.7 61 69.2 77.2 Simulation 5.4 13.2 20.6 28.4 36.6 45 53 61.5 69.8 78.4 Error 4.7 1.1 1.2 1.1 1 1.5 0.7 0.5 0.6 1.2
Fig. 3. Depth profile comparison along one side of the pocket.
likely to be due to alignment errors of the workpiece on the machine fixture. The square pocket was machined in 50 layers, with a basic trajectory consisting of 4 segments and nominal layer depth of 0.9 mm. Discharges were recorded at each layer at the end of each side. Simulation was performed layer by layer, with a motion resolution of 10 mm in x–y directions. The comparison between the machined and simulated pocket profile along one of its sides is shown in Fig. 3, showing a very good agreement between measurements and simulation. These results confirm the reliability of material removal simulation based on discharge counting and MRD estimation. 5. Conclusion The reliability of workpiece material removal per discharge (MRD) estimation for application in electrode wear compensation based on material removal has been investigated. MRD showed a decreasing trend with the progress of the machining operation, reaching stabilization after a certain depth depending on the energy level. The accuracy of MRD estimation was found higher for process parameters combinations yielding lower average discharge energies. In such conditions tool wear compensation based on MRD is expected to be most efficient. Exploiting the information on MRD and counted discharges during a machining operation, a material removal simulation tool for accurate workpiece representation of the machined geometry has been developed and validated. The simulation tool can be used for verification of the progress of the machining operation as well as for improved tool trajectory definition and tool wear compensation for the following process steps.
References [1] Ferraris E, Reynaerts D, Lauwers B (2011) Micro-EDM Process Investigation and Comparison Performance of Al3O2 and ZrO2 Based Ceramic Composites. Annals of the CIRP 60(1):235–238. [2] Masuzawa T (2000) State of the Art of Micromachining. Annals of the CIRP 49(2):473–488. [3] Kojima A, Natsu W, Kunieda M (2008) Spectroscopic Measurement of Arc Plasma Diameter in EDM. Annals of the CIRP 57(1):203–207. [4] Kunieda M, Lauwers B, Rajurkar K, Schumacher BM (2005) Advancing EDM Through Fundamental Insight into Process. Annals of the CIRP 54(2):599–622. [5] Bissacco G, Hansen HN, Tristo G, Valentincic J (2011) Feasibility of Wear Compensation in Micro EDM Milling Based on Discharge Counting and Discharge Population Characterization. Annals of the CIRP 60(1):231–234. [6] Bleys P, Kruth JP, Lauwers B (2004) Sensing and Compensation of Tool Wear in Milling EDM. Journal of Materials Processing Technology 149:139–146. [7] Aligiri E, Yeo SH, Tan PC (2010) A New Tool Wear Compensation Method Based on Real-time Estimation of Material Removal Volume in Micro-EDM. Journal of Materials Processing Technology 210:2292–2303. [8] Jung JW, Jeong YH, Min BK, Lee SJ (2008) Model-based Pulse Frequency Control for Micro-EDM Milling using Real-time Discharge Pulse Monitoring. Journal of Manufacturing Science and Engineering 130(3):1–11.
Contents lists available at SciVerse ScienceDirect
CIRP Annals - Manufacturing Technology jou rnal homep age : ht t p: // ees .e lse vi er . com /ci r p/ def a ult . asp
Reliability of electrode wear compensation based on material removal per discharge in micro EDM milling G. Bissacco (2)a,*, G. Tristo b, H.N. Hansen (1)a, J. Valentincic c a
Department of Mechanical Engineering, Technical University of Denmark, Kgs. Lyngby, Denmark Department of Industrial Engineering, University of Padova, Padova, Italy c Faculty of Mechanical Engineering, University of Ljubljana, Ljubljana, Slovenia b
A R T I C L E I N F O
A B S T R A C T
Keywords: Electrical discharge machining (EDM) Wear Micromachining
This paper investigates the reliability of workpiece material removal per discharge (MRD) estimation for application in electrode wear compensation based on workpiece material removal. An experimental investigation involving discharge counting and automatic on the machine measurement of removed material volume was carried out in a range of process parameters settings from fine finishing to roughing. MRD showed a decreasing trend with the progress of the machining operation, reaching stabilization after a number of machined layers. Using the information on MRD and discharge counting, a material removal simulation tool was developed and validated. ß 2013 CIRP.
1. Introduction Micro EDM milling has become in the recent years an established process for the manufacturing of 2½D and 3D micro components and for the generation of micro features on larger components. In such a process part geometry is obtained by material removal in discrete units by means of controlled discharges occurring between electrically conductive tool and workpiece electrodes. Application materials are mostly metals and alloys. Recent research has demonstrated the applicability of the process to conductive ceramics [1]. In EDM milling the tool electrode consists of a cylindrical rod driven along defined paths similarly to a milling tool, and material is removed layer by layer with layer thickness as low as 0.1 mm. The minimum machinable feature size, achievable accuracy and surface roughness are largely affected by the minimum achievable material removal unit [2]. As the amount of the workpiece electrode removed by each discharge is determined by the discharge energy distributed to the workpiece [3], small material removal units are obtained with small discharge energies, which involve very short discharge pulses. Factors affecting the limits of miniaturization in micro EDM are presented in [4]. Tool electrode wear changes the relative position of tool and workpiece and must be compensated for in order to maintain high machining accuracy. Since milling EDM is performed in very thin layers compared to the tool diameter, the tool profile stabilizes rapidly and wear correction can be done by a one dimensional motion parallel to the electrode axis. Recent electrode wear compensation approaches use either an estimation of the volumetric wear ratio with a continuous downward electrode movement proportional to the relative inplane displacement (anticipated tool wear compensation) or real
* Corresponding author. 0007-8506/$ – see front matter ß 2013 CIRP. http://dx.doi.org/10.1016/j.cirp.2013.03.033
time compensation based on discharge counting. Furthermore real time compensation can be based on full discharge pulse discrimination or on discharge counting and statistical treatment of the discharge population [5]. Real time compensation was implemented by Bleys [6] in conventional (macro) EDM milling using an isoenergetic pulse generator. Aligiri et al. [7] have proposed a method for tool wear compensation in micro EDM drilling based on real time full discharge discrimination and single crater size prediction using a thermal model. However discharge pulses in micro EDM are not isoenergetic and full discharge discrimination is difficult due to the very short pulse duration which can be as low as 20 ns. Jung et al. [8] have implemented real time tool wear compensation based on discharge pulse monitoring and pulse frequency control. Bissacco et al. [5] recently proposed an approach based on discharge counting and statistical characterization of discharge population for tool wear compensation in micro EDM milling. Such an approach was applied to electrode wear compensation based on Tool Wear per Discharge (TWD) estimation. Implementation of tool wear compensation in micro EDM milling based on statistical characterization of the discharge population and workpiece Material Removal per Discharge (MRD) estimation has not been realized yet. The implementation of such an approach requires accurate and reliable estimation of MRD in various processing conditions. This paper introduces such an approach and investigates the reliability of on the machine workpiece MRD as the most critical contribution to the accuracy of the method. For this purpose an experimental investigation, involving discharge counting and automatic on the machine measurement of removed material volume, was carried out in a range of process parameters settings, from fine finishing to roughing. On the machine measurements of removed material volume were validated by means of confocal microscopy. On the basis of the obtained results, the observed trends for MRD at increasing machining depths are discussed and an analysis of the
180
G. Bissacco et al. / CIRP Annals - Manufacturing Technology 62 (2013) 179–182
measurement error contributions is presented. Furthermore a material removal simulation tool for accurate representation of machined geometry is developed and experimentally validated. 2. Tool wear compensation based on MRD 2.1. Discharge population approach In micro EDM considerable variation of discharge pulses is observed in terms of peak current, discharge duration and discharge energy, even for constant process settings and stable machining process. Therefore the workpiece material removal associated with each single discharge differs and should be assessed individually. The population approach described in [5] is based on the considerations that for a stable process the discharge population, namely the statistical distribution of discharge characteristics such as the discharge energy, is expected to be stable, and that material removal occurs by means of trains of discharges with identical distribution as that of the entire population. Therefore a MRD of the population can be obtained and attributed to each discharge of a sufficiently numerous discharge train. On the basis of this consideration, real time material removal can be estimated by counting the discharges and multiplying by the average MRD of the population. This approach greatly simplifies the task of real time material removal estimation for ultra-short pulses (less than 100 ns) such as those used in micro EDM milling finishing operations since it does not require discharge discrimination but rather discharge counting. 2.2. Implementation of MRD based compensation The sketch in Fig. 1 shows the tool motion and material removal for an undercompensated linear trajectory segment with full radial engagement. As a consequence of tool wear, the vertical level of the machined surface increases along the tool trajectory. For the implementation of material removal based tool wear compensation, occurring discharges during a machining operation are continuously counted. The number of counted discharges Nd for each elemental tool trajectory segment is multiplied by the average MRD of the discharge population for the given tool– workpiece–dielectric fluid combination and process parameters settings, thus obtaining the actual removed volume for the considered trajectory segment (indicated as ‘‘Removed material’’ in Fig. 1). By subtracting this volume to the volume that would have been removed in the same tool path segment if no tool wear occurred, the residual volume for the elemental path segment is obtained. By choosing sufficiently small elemental tool path segments, the variation of the surface level along the elemental path segment can be linearized as shown in Fig. 1 and thereby the angle a and the corresponding part depth variation Dz can be calculated. In order to correct for the detected undercompensation, such depth variation Dz will be added as a downward displacement at the end of the elemental path segment. Furthermore an inclination a will be added to the next path segment in order to obtain the correct depth along the whole path segment. In this way, the surface for the next path segment is maintained at the
correct level without updating the information of the absolute vertical position of the frontal surface of the tool electrode. Obviously the initial vertical position of the tool electrode must be known to initiate the process. For the implementation of such compensation approach it is critical to correctly estimate the volume of material removed at each path segment and therefore the accuracy of estimation of MRD is critical. 3. Stability of MRD The above described electrode wear compensation method requires the estimation of the MRD for the specific process parameters settings and tool–workpiece–dielectric fluid combination to be used in a machining operation. Therefore, for implementation of the method in an industrial environment, standard routines must be developed for the assessment of MRD on the machine. Furthermore, accuracy of MRD determination and MRD stability over time (throughout a machining operation) is critical for the compensation accuracy using the described method. 3.1. MRD experimental determination The experimental setup for the determination of the MRD consisted of a Sarix SX-200 micro EDM milling machine equipped with a laser scan micrometer by Mitutoyo for optical measurement of electrode profile. Tungsten carbide cylindrical rods with diameter of 300 mm were used as tool electrodes, while the workpiece material was a martensitic stainless steel (Uddeholm Stavax ESR). Discharge counting was performed on the current signal by means of a current probe developed in house and a digital counter HP 53131 A with a frequency range from DC to 225 MHz. Experiments were performed for three different process parameters settings, corresponding to different pulse shapes and with average discharge energy of 0.7 mJ, 12.6 mJ and 57.9 mJ. For the experimental determination of the MRD, automatic routines have been developed by preparation of a parametric part program customizable by the user through the machine console. In the experiments described here, the program was used for the generation of cylindrical cavities by removing material layer by layer in milling mode, with layer thickness of 0.5 mm, 0.9 mm and 2.5 mm for the three process parameters settings. The cavities were 500 mm in diameter and were generated by means of circular interpolations in 10 incremental steps with actual depth of approximately 50 mm per step (total 500 mm), with discharge counting and recording of the cumulated number of discharges for each step. At the end of each incremental step, the diameter and the depth of the cavity were measured on the machine using the electrode as a touch probe in a Coordinate Measuring Machine. Diameter measurements were performed at a depth level of 3 mm above the bottom of the cavity in order to take into account the conicity of the machined hole in the estimation of MRD. The diameter measurement procedure consisted in probing 8 points on the hole circumference. For improved measurement reliability, each point was probed three times. The depth of the cavity was measured by probing its bottom at the center with the electrode. The probing was repeated three times. For each of the three process parameters settings investigated, the test was repeated three times. The results are reported in Fig. 2. 3.2. Validation of on the machine measurements
Fig. 1. Sketch showing tool motion and material removal for undercompensated linear trajectory segment.
The accuracy of on the machine measurements of diameter and depth strongly affects the accuracy of MRD. Therefore on the machine measurements of removed material volume were validated by means of confocal microscopy in dedicated experiments. The comparison was carried out on three cavities machined with the three different process parameters settings previously used for MRD determination. After machining of the three cavities, a 200 mm electrode was generated on the machine using the available WEDG module and measured by means of the laser scan
G. Bissacco et al. / CIRP Annals - Manufacturing Technology 62 (2013) 179–182
181
are in line with the scatter of the MRD measurements shown in Fig. 2. 3.3. Analysis of MRD variability
Fig. 2. Measured MRD values at increasing depths for the three process parameters settings considered. Data points represent the average of three repetitions. Error bars represent the ranges of MRD values for each step depth.
micrometer to be 198.4 mm. The generated electrode was used to probe the circumference and the bottom of the machined cavities using the same procedure as described in the previous section and repeating the measurements 3 times. The cavities were also measured using a confocal microscope Sensofar PLm Neox. A 100 magnification, corresponding to an in plane resolution of 0.17 mm and a vertical resolution of less than 2 nm was used for the generation of image files for measurement of diameter. For measurement of cavity depth, a 20 magnification, was used, yielding an in plane resolution of 0.83 mm and a vertical resolution of less than 20 nm. Due to the almost vertical walls of the cavities, measurement of the diameter of the holes was possible on the images from the confocal microscope only at very low depths. Therefore diameter measurements for both on the machine and confocal microscope measurements were performed at 3 mm below the part surface. The results of the comparison are shown in Table 1. Using the confocal microscope measurements as a reference, the error on the average depth is in the order of 1 mm and on the diameter it is between 2 mm and 3 mm. Such deviations appear to be systematic errors and are likely due to the roughness of the surface of the cavities as the tool electrode probes on the peaks of the cavities lateral and bottom surfaces. The observed depth and diameter errors correspond to a maximum expected volume measurement error of 3% for a step depth of 50 mm as used in the experiments for the determination of the MRD. Such errors, although not negligible,
Table 1 Results of validation of on the machine measurements. Uncertainty at 95% coverage probability on depth and diameter measurements performed on the machine is 1.3 mm and 1.9 mm respectively. Depth (mm)
The results of the experimental tests for MRD determination are shown in Fig. 2. Clearly, higher average discharge energies yield higher MRD. From the observation of the diagrams it is also visible that for constant energy settings the measured MRD data show higher scatter between the test repetitions at lower depths. Furthermore MRD is not constant during machining, with a decreasing trend with the progress of the machining operation, reaching stabilization after a certain depth depending on the energy level. This result is in accordance with results reported in [7] for drilling, where a correction factor was introduced to account for the decreasing material removal efficiency. However in drilling EDM the flushing efficiency decreases strongly with increasing depth, affecting the observed MRD. At higher depths, where MRD reaches stabilization, the variation between repeated MRD measurements is approximately 1% for the lower energy levels and 4% for the highest energy level. While the observed scatter of the MRD measurements can be regarded as acceptable, the decreasing trend observed at the lower depths makes the implementation of tool wear compensation based on MRD difficult. The reasons for such a trend are not completely understood. The range of measurement errors for on the machine volume measurements, as reported in Table 1 are not compatible with the observed variation of MRD with increasing machining depth. An analysis of the on the machine measurement procedure yields the following considerations. The cavity step depth is measured in the center of the cavity while a rounding of the bottom occurs at the periphery. Consequently the volume for the first step is over estimated, leading to a larger apparent MRD. Such overestimation occurs only for the first step, while for the following steps it is compensated by the excess material remaining as a consequence of the rounding effect in the previous step. The rounding radius is larger for higher average discharge energies, thus the absolute error on the MRD measured on the first layer is larger at higher energies. However the relative error on the measured volumes, and thereby on the calculated MRD, decreases with the increase of the total removed volume per step. For the cavity diameter values and actual step depth used in the experiments presented above, and for the highest energy settings (showing a cavity rounding radius of 12 mm), the calculated relative error contribution on the step volume, and therefore on the MRD value, due to this effect is approximately 0.33%. In addition to the above error, the rounding of the tool electrode induces an overestimation of the cavity diameter for the first step. With the diameter values used in these experiments, the maximum corresponding error is approximately 4% on the measured volume for the first step. The discussed error contributions are not sufficient to justify quantitatively the decreasing trend of the MRD observed for the process settings with the highest average discharge energy as shown in Fig. 2, which is therefore expected to be a characteristic of the process, most probably due to the discharges that occur on the cylindrical surface of the electrode while machining at greater depths. Due to the above considerations, tool wear compensation based on MRD is expected to be most efficient in connection with process settings involving lower average discharge energies. In such cases, MRD stabilization is reached quickly with the progress of the machining operation, so that an average MRD value, corresponding to that in the stable region, can be used for compensation. 4. Material removal simulation
Diameter (mm)
Energy level
E1
E2
E3
E1
E2
E3
On machine Confocal Error
57.6 58.4 0.8
63.4 64.6 1.3
60.9 62.0 1.1
512.7 510.9 1.8
519.9 517.9 2.0
524.1 521.0 3.1
4.1. Material removal simulation tool A tool for material removal simulation has been developed in Matlab environment. The tool allows accurate representation of
182
G. Bissacco et al. / CIRP Annals - Manufacturing Technology 62 (2013) 179–182
the machined geometry which can be used for verification of the progress of the machining operation as well as for improved tool trajectory definition and tool wear compensation for the following process steps. The input to the simulation tool consists of a three columns array where the first and second columns contain the recorded x and y positions of the center of the tool electrode at defined path intervals, while the third column contains the corresponding number of counted discharges. Within the simulation module, the path interval is divided in any preferred number of segments and the amount of discharges corresponding to each segment is calculated by dividing the discharges counted for the whole path interval by the number of segments. Tool electrode and workpiece are discretized in a number of 2½D voxels with square footprint equal to the average area of the discharge craters measured for the specific process settings. It is important that the voxel size is chosen to be close to the average crater size. As the volume removed by a single spark is constant and equal to the MRD, the amount of reduction in height depends on voxel lateral dimension. If voxels are too small the voxel height reduction caused by a single discharge is exaggerated and the accuracy of the simulation is compromised. The tool electrode is considered in the simulation as having a diameter equal to the actual diameter plus two times the lateral discharge gap (which is determined experimentally). For each segment, the couple of voxels on tool and workpiece that have minimum distance are identified and the Z coordinate of the workpiece voxel is reduced by an amount equal to the MRD divided by the voxel area. Thus the volume removed from the voxel corresponds to MRD. If more than one tool– workpiece voxels couples with the same distance are found (as happens at the beginning of the operation), a random choice is operated automatically. The output of the simulation tool consists of a file of the machined part in STL and SDF formats. The STL file can be imported in a generic CAM software for the generation of the tool trajectory for the following machining operation. The SDF file can be imported in programs such as SPIP for measurement of part geometry. 4.2. Validation of material removal simulation tool Experimental validation of the material removal simulation tool was obtained by means of machining tests with comparison of machined features and simulation results. The generated features were a straight groove 500 mm in length, with full radial engagement of the tool electrode (slotting) and a square pocket with a square island. In both tests the tool electrode was dressed using the machine WEDG unit to a diameter of 200 mm and process parameters settings corresponding to the intermediate average discharge energy E2 (11.5 mJ) were used. The discharge gap was measured to be 5.5 mm by means of Confocal and SEM measurements on dedicated tests, while the crater diameter was measured to be approximately 10 mm. No tool wear compensation was applied and discharges were counted during machining. For the simulation, the actual tool path segments, counted discharges, tool diameter and lateral sparking gap were given as input. MRD was assumed constant and equal to 12.4 mm3 as results from the stabilized region in Fig. 2. The groove was machined in 10 steps, each consisting of 20 layers with nominal layer depth of 0.9 mm, for a total nominal depth of 180 mm (actual depth 77 mm). At the end of each step of 20 layers, the number of discharges was recorded and the groove floor was measured at half the path length by means of touches using the tool electrode. Machining was simulated layer by layer, with a motion resolution of 10 mm in x–y directions. Table 2 shows the comparison between the groove depth as measured on the machine and the corresponding values from the simulation for each of the steps of 20 layers each. An overall good agreement between measured and simulated material removal is observed. With the exception of the first layer, the maximum depth error of the material removal simulation is 1.2 mm for a total depth of 77 mm. The error in the first layer is
Table 2 Results of validation of material removal simulation in slotting. Step no.
1
2
3
4
5
6
7
8
9
10
Actual depth (mm) On the machine 0.7 14.3 21.8 29.5 37.6 46.5 53.7 61 69.2 77.2 Simulation 5.4 13.2 20.6 28.4 36.6 45 53 61.5 69.8 78.4 Error 4.7 1.1 1.2 1.1 1 1.5 0.7 0.5 0.6 1.2
Fig. 3. Depth profile comparison along one side of the pocket.
likely to be due to alignment errors of the workpiece on the machine fixture. The square pocket was machined in 50 layers, with a basic trajectory consisting of 4 segments and nominal layer depth of 0.9 mm. Discharges were recorded at each layer at the end of each side. Simulation was performed layer by layer, with a motion resolution of 10 mm in x–y directions. The comparison between the machined and simulated pocket profile along one of its sides is shown in Fig. 3, showing a very good agreement between measurements and simulation. These results confirm the reliability of material removal simulation based on discharge counting and MRD estimation. 5. Conclusion The reliability of workpiece material removal per discharge (MRD) estimation for application in electrode wear compensation based on material removal has been investigated. MRD showed a decreasing trend with the progress of the machining operation, reaching stabilization after a certain depth depending on the energy level. The accuracy of MRD estimation was found higher for process parameters combinations yielding lower average discharge energies. In such conditions tool wear compensation based on MRD is expected to be most efficient. Exploiting the information on MRD and counted discharges during a machining operation, a material removal simulation tool for accurate workpiece representation of the machined geometry has been developed and validated. The simulation tool can be used for verification of the progress of the machining operation as well as for improved tool trajectory definition and tool wear compensation for the following process steps.
References [1] Ferraris E, Reynaerts D, Lauwers B (2011) Micro-EDM Process Investigation and Comparison Performance of Al3O2 and ZrO2 Based Ceramic Composites. Annals of the CIRP 60(1):235–238. [2] Masuzawa T (2000) State of the Art of Micromachining. Annals of the CIRP 49(2):473–488. [3] Kojima A, Natsu W, Kunieda M (2008) Spectroscopic Measurement of Arc Plasma Diameter in EDM. Annals of the CIRP 57(1):203–207. [4] Kunieda M, Lauwers B, Rajurkar K, Schumacher BM (2005) Advancing EDM Through Fundamental Insight into Process. Annals of the CIRP 54(2):599–622. [5] Bissacco G, Hansen HN, Tristo G, Valentincic J (2011) Feasibility of Wear Compensation in Micro EDM Milling Based on Discharge Counting and Discharge Population Characterization. Annals of the CIRP 60(1):231–234. [6] Bleys P, Kruth JP, Lauwers B (2004) Sensing and Compensation of Tool Wear in Milling EDM. Journal of Materials Processing Technology 149:139–146. [7] Aligiri E, Yeo SH, Tan PC (2010) A New Tool Wear Compensation Method Based on Real-time Estimation of Material Removal Volume in Micro-EDM. Journal of Materials Processing Technology 210:2292–2303. [8] Jung JW, Jeong YH, Min BK, Lee SJ (2008) Model-based Pulse Frequency Control for Micro-EDM Milling using Real-time Discharge Pulse Monitoring. Journal of Manufacturing Science and Engineering 130(3):1–11.