Tool speed and polarity effects in micro-EDM drilling of 316L stainless steel

Abstract This paper focuses on the issues of Resistor-Capacitor-based Electrical Discharge Micro-Machining process and investigates the effects of tool speed and polarity on the performance measures such as Tool Wear Rate, Material Removal Rate, Overcut and Taper Angle by drilling on 316L Stainless Steel. Taguchi’s L54 mixed orthogonal array design is employed to conduct experiments by varying tool polarity at two levels and voltage, capacitance, spindle speed at three levels. The cause and effect relationship between the experimental factors and responses are analysed and discussed using Factorial Analysis of Variance technique. Optimum combinations of machining parameters are also evaluated using Taguchi-based Grey Relational Analysis, by considering grey relational grade matrix and influence of process parameters on the responses. Further, microscopic analysis is done to identify the micro-voids, globular formation, and cracks present on the surface of the hole produced under various machining conditions.


Introduction
In today's industrial scenario, removal of materials in a micro-metre range is an essential requirement to make precise components (Jain, 2014). Micro-holes which are created for various automobile components, optical devices, and medical instruments must be can be machined by EDMM process (Unune, Singh, & Mali, 2016). EDMM is also known as spark erosion process in which work material is eroded by electrical sparks so that mechanical stresses and vibrations problems are very less.
In this research work, the experiments are conducted to analyse the machining characteristics of EDMM process by changing the tool and work polarity and also varying the tool rotational speed. The effects of process parameters on the machining characteristics such as TWR, MRR, taper angle and overcut is also investigated while drilling micro-holes on SS 316L material. This material has wide applications like valve spools, damper plates, aerospace nozzle guide segments, combustion liners, medical and textile needles. Experiments are conducted based on Taguchi's L54 mixed orthogonal array design by varying the tool polarity at two levels and voltage, capacitance, spindle speed at three levels. In addition, the cause-and-effect relationship between the experimental factors and responses are analysed using Factorial ANOVA technique. Also, the optimum combinations of machining parameters are evaluated using Taguchi-based Grey Relational Analysis (GRA), by considering the grey relational grade matrix and the influence of process parameters on the responses. Further, microscopic analysis is done to identify the micro-voids, globular formation and cracks present on the surface of the hole produced under various machining conditions. The rest of this paper is structured as follows. Section 2 reviews the related works in the area of EDM process. Later on, Section 3 describes the experimental set-up details to perform micro-holes on 316L Stainless Steel. The effects of tool speed and polarity on the performance measures and geometrical characteristics are discussed in Section 4. The details of ANOVA, GRA and SEM analysis is presented in Section 5. Finally, the concluding remarks are drawn in the last section together with the future research directions.

Literature review
EDMM process performance is measured in terms of TWR, MRR, overcut, taper angle and surface integrity. In order to improve these performance measures, the EDMM process needs to be studied in detail. Only a few researchers have addressed the phenomenon of drilling micro-holes by EDMM. Raju, Srinivasa, Vinod, and Chellamalai (2013) analysed the influence of electrode diameter and capacitance on the MRR, surface roughness (SR), volumetric wear ratio and spark gap. Kibria, Sarkar, Pradhan, and Bhattacharyya (2010) examined the effect of dielectrics and stated that MRR and TWR were found to be lower for Kerosene than the Deionized water. Liu et al. (2016) studied the impact of input parameters and electrode shape on various performance measures of die-sinking EDM and found that electrode geometry had a critical role in the performance. Tseng, Kao, and Chang (2016) developed a real-time monitoring system for EDMM process which is capable of producing nano-silver particles in the form of debris and compared the particle size with the conventional EDM process. Li, Hou, Xu, and Yu (2016) investigated the micro-holes drilled by three different EDM drilling methods and found that better MRR and TWR is achieved for cutting edge electrodes. Wong, Rahman, Lim, Han, and Ravi (2003) analysed the material removal mechanism of EDMM process using single pulse discharges and stated that high efficiency was obtained at lower energy levels. Lee, Kim, and Kim (2015) improved the efficiency of the EDMM process by vibrating the workpiece at low frequency and identified that the short circuits that occur during machining are reduced and proper flushing occurs due to low-frequency vibration. Guo et al. (2014) analysed the material removal mechanism of the EDMM by integrating two temperature and molecular dynamics model. They found that initially the material is removed in bulk, from the cathode due to thermal shock and later the material is removed in single atoms. Manivannan and Kumar (2017) improved the machining performance of the EDMM process by cooling the machining environment with the cryogenic coolant and found that it has significantly improved the MRR and TWR. Selvarajan, Manohar, Kumar, and Dhinakaran (2017) studied the effects of EDM process parameters on various responses by drilling holes on Si 3 N 4 -TiN ceramic composites. They have modelled the EDM process using multiple regression analysis and also found the optimum combination using grey relational analysis.
Literature review shows that there are many research efforts reported in the area of EDM-die sinking process based on the transistor-type pulse generator. But there are limited literature on EDM-drilling based on resistance-capacitance-type pulse generator where the effects of tool speed and polarities are not studied. Depending upon the circuit type used in the machine, the effect of parameters is also varied and the stochastic thermal nature of the EDM process makes it difficult to explain all of those effects fully. The debris removal from the inter-electrode gap is one of the reasons for more MRR in μEDM. If there is no proper debris removal, higher spark energy produces results in higher amount of debris. This debris sticking on the workpiece trapped in and causes unwanted spark. The un-wanted sparks erodes materials from the tool electrode, which results high tool wear. Thus, a great portion of discharge energy occupies with unwanted sparking, while the remaining erodes the work material (Ali, Mehfuz, Khan, & Ismail, 2012). Hence, in this research work investigations are done to analyse the various issues of EDMM process by changing the tool polarity and also varying the tool rotational speed. The experiments were carried out by drilling several micro-holes using different process parameters, namely tool polarity, tool speed, peak current and voltage. Some process performance indicators, such as TWR, MRR, and geometric indicators such as diametric overcut and taper angle were taken into account.

Experimental set-up details
In this work, DT110-Multiprocess Micro-machining centre (Figure 1) (Make: Mikrotools Pvt. Ltd., Singapore) was used to drill micro-holes on SS 316L workpiece by EDMM process. It is a 3-axes CNC servo controlled machine which is energized by RC pulse generator. The maximum travel range of this machine is 200 mm in X-axis, 100 mm in Y-axis and 100 mm in Z-axis with a resolution of .1 μm in all directions and also has a full closed-feedback control that ensures sub-micron accuracy. The EDMM set-up consists of a cylindrical electrode rigidly fixed in the spindle and the workpiece material is clamped on a fixture. The spindle can rotate up to 5000 rpm with the help of variable speed spindle drive. During machining, both the workpiece and the tool are immersed in the dielectric fluid. An external flushing system is used to circulate the dielectric and to provide proper flushing. The work material, electrode material and dielectric used in this study were 316L stainless steel of size 25 × 25 × .5 mm, .3 mm tungsten rod and DCO-1000i EDM oil, respectively, and their corresponding properties are listed in Tables 1-3. The experimental conditions are listed in Table 4.
The performance of the EDMM process depends on various factors like voltage, capacitance, speed, feed rate, dielectric, flushing pressure, polarity, etc. Among these tool polarity,   spindle speed, voltage and capacitance are the primary factors which affect the performance of EDMM. Hence, these four factors are taken as inputs for the experimental design. To study the effects of each factor and to find its ranges, the experiments are conducted based on full factorial design (Montgomery, 2013) with two replications. Totally 54 (L54) experiments have been conducted with three factors (speed, voltage and capacitance) at three levels and one factor (tool polarity) at two levels for each replication as shown in Table 5. The workpiece and tool mass is measured using an analytical balance having a repeatability of .00002 g (Make: Contech CA184), before and after machining. The entry and exit hole diameters are assessed by Non-contact video measurement system (ARCS KIM1510E). The TWR, MRR, Overcut and Taper angle is selected as output responses to measure the performance characteristics. The equations used to calculate those output responses are given below (Karthikeyan, Ramkumar, Dhamodaran, & Aravindan, 2010;Mathew, Somashekhar, Sooraj, Subbarao, & Ramachandran, 2009). In this study, three replicated tests were conducted. The mean value of three tests was taken as the final value which is used for calculating the response values (MMR, TWR, overcut and taper angle) and for ANOVA analysis.
The mean values of these output responses are shown in Table 5.
(1) MRR mm 3 / min = Amount of material removal from workpiece drilling time (2) TWR mm 3 /min = Volume of material removal from electrode drilling time

Results and discussion
In this section, the influence of various EDMM process parameters on output responses like MRR, TWR, Taper angle and Overcut are discussed. The discussion is based on the experimental results obtained by drilling micro-holes on SS 316L material and Factorial ANOVA analysis performed on it. Figure 2 indicates the results related to MRR, TWR and Taper angle response changes due to polarity effects. It can be noticed that the MRR is increases for tool negative than the tool positive, because of more amount of electrons are discharged from the tool. These electrons strike the workpiece at high velocity and produce more amount of heat at the IEG. Hence, work material is eroded from the top substrate due to melting and evaporation (Jerald, Kumanan, Kumar, & Chandrakar, 2013). It can also be inferred from the Figure 3 that the melting volume of tool negative is larger than the tool positive. This volume difference is due to the variation in amount of heat generated (Nakajima, Okada, & Uno, 1991). It can also be seen from the Figure 4 that the tool shape differs abnormally in negative polarity this is due to crack propagation resulting from rapid heat cycle and deposition of metal on the tool surface. Hence, the tool wear for negative polarity is much higher than the positive polarity. When the TWR is more, the shape of the tool also gets altered; hence, the machined micro-hole becomes tapered. For tool positive, the TWR and the taper angle is small and for tool negative the TWR is more, so taper angle is also higher.

Effects of tool speed
Generally in EDMM process, the tool remains stationary, but in this work tool speed is also considered with various speeds. The tool speed produces centrifugal effect on the dielectric and it causes the movement of debris particles that are present in the IEG. When the spindle speed is increased, proper flushing of debris occurs due to centrifugal effect and hence there is a reduction in machining time (Cyril, Paravasu, Jerald, Sumit, & Kanagaraj, 2017). From the Figure 5, it is observed that MRR is high during better flushing conditions. Beyond a certain limit of tool speed, the movement of debris becomes faster and it causes instability by producing secondary sparks in the IEG, thus MRR is decreased but TWR, Taper angle and overcut are increased ( Figure 6).  At zero rpm, the tool remains stationary and there is no centrifugal effect. Hence, the debris doesn't have any movements which cause instability and improper flushing. Due to this, more secondary sparks and short circuits occur; hence, machining time is also increased.

Effects of voltage and capacitance
The voltage and capacitance directly contribute to discharge energy of the RC-based EDMM process. The following equation is used to calculate discharge energy (Jain, 2014).  when there is an increase in capacitance and voltage from lower levels, the discharge energy also gets increased. Hence, more heat is generated at the IEG. At high levels of voltage and capacitance the MRR and TWR are higher (Figures 7 and 8) due to high discharge energy.

Factorial ANOVA analysis
The factorial ANOVA computes the degree to which combination of independent variables used to predict the values of dependent variables (English, 2006). It also analyse the cause-and-effect relationship between the experimental factors. The ANOVA analysis on MRR, TWR, Overcut and Taper angle is done and the results are listed in Table  6. It shows the influence of input parameters on output responses. The ratio between explained and total variation is defined as coefficient of determination (R 2 ) and it is also a measure of degree of fitness (Torres, Puertas, & Luis, 2016). For MRR, the R 2 value is 96.05%, which indicates the model variability. Whereas R 2 adj value is 95.45%, which indicates the correlation between the MRR and process parameters are well characterized by the developed model. When P-value of a parameter is less than .05 at 95% confidence level, the corresponding parameter is significant and has individual effect on the output responses (Khan, Khan, Siddiquee, Chanda, & Arindam, 2014). It is seen from the Table  6 that the P-value of all the parameters are less than .05 at 95% confidence interval for output responses. Thus, all individual parameters are significant on MRR, TWR, Overcut and Taper angle.

Regression analysis
In general, a true functional relationship between dependent variables (responses) and independent variables (input factors) are unknown. This relationship between variables can be established by a mathematical model called regression model. It is an empirical model which expresses the results of the experiments quantitatively, to facilitate understanding, implementation and interpretation (Montgomery, 2013). Hence, this section focuses on establishing a truly functional relationship by fitting linear regression models. In general, the dependent variable y may be related to k independent variables as given below.
(6) Y = 0 + 1 X 1 + 2 X 2 + 3 X 3 + … … + p X p + This is a multiple linear regression model with 'p' independent variables. These independent variables are also called as predictor variables. Where β i , i = 0, 1, 2… p, are called regression coefficients. This model describes a hyper plane in the p-dimensional space of the independent variables (Li et al., 2016). The parameter β i represents the expected change in response y per unit change in X i when all the remaining independent variables X j (j ≠ i) are held constant. The derived multi-linear regression model for all output responses are as follows:

Grey relational analysis
Taguchi's design with ANOVA can only predict the effects and relationship of the process parameters over the output responses. Since the data obtained from the above design are discrete in nature, uncertainties are more which will have a complex interaction effect on the performance characteristics. These uncertainties are solved using GRA which has certain advantages over the other statistical techniques (Kundu & Singh, 2016). In GRA, multi-objective problems are transformed into a single objective problem. Thus, Taguchi design with GRA remains most potent method to solve multi-objective optimization problems (Yadav, Yadava, & Singh, 2014). The GRA processing steps are listed below (Pannerselvam, 2012).
Step 1: Initially, the raw data are pre-processed. The normalized values of the responses are calculated based on the following equations. The MRR which is higher the better performance variable thus the normalization equation (Sarkar, Panja, Das, & Sarkar, 2015) is expressed as: The TWR, Overcut and Taper angle which are lower the better performance variable thus the normalization equation is expressed as: where X pq = measured responses, min (X pq ) = minimum of X pq and max (X pq ) = maximum of X pq , p = response variables and q = trial number (number of the experiments). Fundamentally larger the normalized values, better the performance characteristics.
Step 2: The maximum of the Y pq regardless of responses and trials are computed by the following equation.
Step 3: The absolute variance between the reference value R and each normalized value is computed as follows: where R is the expected sequence, Y pq is the comparability sequence and ∆ pq is the deviation sequence of R and Y pq .
Step 4: The Grey Relation Coefficient (GRC) ξ pq for each of the normalized values is computed using the following equation.
where is the differentiating coefficient ∈ [0, 1] and Deng (2002) has stated that .5 is the most widely accepted value. The amount of relational degree between the actual and desired performance characteristics can be obtained through GRC values, ranging from 0 to 1. Higher the GRC value, the more intense is the relational degree.
Step 5: The Grey Relation Grade (GRG) for each trial is computed as follows: where n denotes number of response variables. The computed GRC and GRG values are summarized in Table 7. The level of relationship between comparability and reference sequence is shown by the GRG. Combinations of machining parameters with higher GRG values are the desired optimum multi performance characteristics (Chakravorty, Gauri, & Chakraborty, 2013). The GRG value for trail no. 29 has the largest values among the others GRG values. The optimum combination of the process parameters obtained from the mean GRG table is polarity: positive, voltage: 100 v, capacitance: 0.01 nF, tool speed: 250 rpm as highlighted in Table 8. After obtaining the optimum combination of process parameters from GRA, the final step is to run a validation experiment for the results attained by the Taguchi' s design. A confirmation experiment is necessary only when the obtained optimum combination is not one of the trail runs. In this case, the obtained optimum combination is one of the trail runs of the full factorial experiments, so a confirmation experiment is not conducted. A significant improvement in experimental GRG implies that this method can be utilized for multi-objective optimization of EDMM process. The ANOVA n results for GRG are listed in Table 9. The polarity, voltage, capacitance and speed have a significant effect on the multi-performance characteristics. The capacitance (34.85%) influences more on the multi-performance characteristics followed by polarity (31.56%), speed (14.30%) and voltage (9.23%).

SEM analysis
The machined surfaces have been analysed using HITACHI S3000H Scanning Electron Microscope (SEM) to identify the influence of process parameters on surface characteristics. From the Figure 9(a) and 9(d), it is clear that at low discharge energy the holes produced   are smooth and the circularity of the hole is better. At high voltage and capacitance ( Figure  9(b) and 9(c)), the discharge energy is more, so that heat generated is also more and the heat dissipation rate is inadequate. Thus, the gases formed at the machining area are unable to escape which leads to the formation of voids (Figure 10) on the machined surfaces. Due to the implosion of some gas bubbles, a few amount of material from the dielectric and tool electrode are deposited on the machined surfaces. Also the machined surfaces are cooled suddenly due to recharging of the dielectric during pulse off time, which produces a temperature difference between the substrates of the machined surfaces. Hence, cracks are formed due to this temperature gradient ( Figure 11).

Conclusive remarks
In this study, the effect of straight and reverse tool polarities and tool speed on the machining performance of 316L Stainless steel using micro-electrical discharge drilling process was analysed. The voltage, capacitance, speed and feed rate were the varied machining parameters. The performance was evaluated in terms of MRR, TWR, Overcut and Taper angle. Based on results and discussion of the conducted experimental study, the following conclusions are derived: • In tool negative polarity, the number of electrons hitting the workpiece is more than the electrons hitting the tool electrode. Thus, MRR is high and TWR is low. • At zero rpm, the machining conditions are unstable due to improper flushing and evacuation of debris. Hence, MRR is low and TWR is high compared to other machining conditions. Due to proper evacuation of debris and flushing of dielectric, the overcut is smaller at the speed of 250 rpm. • The cause and effect relationship of process parameters on performance characteristics are analysed using ANOVA technique. The ANOVA analysis of GRG shows that capacitance (34.85%) influences more on the multi-performance characteristics followed by polarity (31.56%), speed (14.30%) and voltage (9.23%). Hence, proper selection of process parameters is very important for EDMM. • From the average GRG, it is found that the highest value of GRG is for the Positive Polarity, Voltage 100 V, Capacitance .01nf and Spindle Speed 250 rpm. It is the optimum combination of process parameters for the EDMM process with maximum MRR and minimum TWR, overcut and taper angle.
This research study can be extent further to analyse the various process parameters of different pulse generators on MRR, surface integrity, TWR, heat-affected zone (HAZ), taper angle, surface crack density (SCD), overcut and white layer thickness (WLT). Development of a comprehensive thermal model based on the superheating theory which uses realistic boundary conditions such as Gaussian distributed heat flux, temperature dependent thermal properties and expending plasma radius is another interesting area for research. Finite element model can be created to study the material removal, tool wear procedure and to analyse the cause and effects of HAZ, SCD and WLT.