5 th International & 26 th All India Manufacturing Technology, Design and Research Conference (AIMTDR 2014) December 12 th 14 th, 2014, IIT Guwahati, Assam, India CORRELATION AMONG THE CUTTING PARAMETERS, SURFACE ROUGHNESS AND CUTTING FORCES IN TURNING PROCESS BY EXPERIMENTAL STUDIES Jithin Babu.R 1, A Ramesh Babu 2 1 ME Student, PSG College of Technology, Coimbatore,641004,jithinrajanbabu@live.com 2 Associate Professor, PSG College of Technology, Coimbatore,641004,dr.rameshbabua@gmail.com Abstract In this work experimental investigations and statistical analysis are carried out to study the effect of cutting parameters (cutting speed, feed rate, and depth of cut) on surface roughness and cutting forces during dry turning of aluminium alloy. Full factorial design of experiments corresponding to 27 runs (3 3 ) was followed for the experimental design. The contribution of each factor on the output is determined by analysis of variance. During analysis it is found that feed rate is the most influencing parameter affecting surface roughness (70.35%). Depth of cut (85.37%) was the most influencing parameter on the cutting forces. Later prediction models were created using second order multiple regression method. The model provided good prediction accuracy with a mean absolute error of 3.47% for surface roughness and 6.8 % for the cutting forces. In order to validate the regression model, another prediction tool using artificial neural network (ANN) is proposed. It is clearly seen that the proposed model is capable of predicting the surface roughness and cutting forces with good accuracy. The statistical analysis, multiple regression modeling and neural network prediction were performed on MINITAB 16 and MATLAB nntoolbox. Keywords: surface roughness, ANOVA, ANN. 1 Introduction Surface Roughness is often a good predictor of performance of mechanical components since the irregularities in the surface may form nucleation signs for cracks and corrosion, reduce the fatigue life of components & increase wear. In Some cases surfaces should be rough as in case of bearings so as to hold the lubricating particles. Surface roughness is an important parameter in all machining processes such as in turning, milling, grinding etc. During machining the major factors that affect the surface roughness are cutting speed, feed rate, depth of cut, machine tool vibrations, temperature of cutting fluid, tool geometry etc., so it becomes important for the manufacturing industry to find the suitable levels of process parameters for obtaining desired surface roughness. Cutting forces also play an important role to predict machining performance for any machining operation. Estimating the cutting forces helps in structural design of machine tool system, condition monitoring and studying the machinability characteristics of work materials. Bartarya et al. [1] studied the effect of cutting parameters on surface roughness and force components on EN31 steel, ANOVA analysis was performed and the prediction models were created using regression for surface roughness and cutting forces. It was observed that depth of cut was the most influencing parameter on surface roughness and cutting forces. Iihan asilturk et al. [2] studied on the significance of cutting parameters on surface roughness in AISI 1040 Steel using ANN and regression methods; it was observed that ANN predicts surface roughness with good accuracy than regression model. Gajanana et al. [3] studied on the optimization of process parameters in End milling process, it was observed that cutting speed was the most important factor influencing surface roughness. Vikas upadhyay et al. [4] further studied the effect of cutting parameters and vibration signals on surface roughness, the prediction models were made using regression as well as Artificial Neural Network (ANN), similar works were also carried by [10, 11]. John patten et al. [5] studied the comparison between numerical simulations and experiments for a single point cutting tool in turning operation, the cutting 459-1
CORRELATION AMONG THE CUTTING PARAMETERS, SURFACE ROUGHNESS AND CUTTING FORCES IN TURNING PROCESS BY EXPERIMENTAL STUDIES forces from the experiment and the simulations showed good agreement. A study on the various works carried out in the field of prediction of roughness in machining using regression analysis, genetic algorithm, neuro fuzzy systems, artificial neural network was given by P.G.Bendaros et al. [6]. Process Optimization on various materials was carried out by [7, 8]. A detailed study on Taguchi DoE was provided by [9]. 2 Experimental set up and Details Experimental design techniques are a powerful approach in product and process development, and they have an extensive application in engineering areas. Potential applications include product design optimization, process design optimization, material selection and many others. In the current scenario, our area of interest is to minimize the surface roughness and resultant forces after turning of Aluminium alloy. The factors and their various levels were selected based on the literature survey, design data book and machine tool specifications. Full Factorial design for three levels and three factors corresponding to 27 experimental run was performed for the experimental analysis. Table 1 shows the various control factors that are considered for the experiments and their levels. 27 combinations of input parameters were used to study the effect of surface roughness and cutting forces. Table I Control factors and their chosen levels Control factors Cutting speed (m/min) Feed rate (mm/rev) 0.25 Level 1 Level 2 Level 3 35 58 76 0.35 0.4 various design parameters was obtained using SYSCON Lathe tool dynamometer (strain gauge type) which gives us the thrust force, feed force and the radial force. Machining time and the force components on three directions were measured during the experimentation; the resultant force was then calculated and was taken for analysis. The machined aluminium alloys were later examined for their surface quality using SURFCODER surface roughness tester. Three measurements of surface roughness were taken at different locations and their average was taken for analysis. The experimental conditions are shown on Table 2. The final experimental set up is shown in figure 1. Machine Tool Work material composition Tool material Environment Size Table 2 Experimental Conditions PSG Trainer Lathe (40-1600 rpm) Spindle motor power: 3 hp Aluminium Alloy 6063, Aluminium (98.7%) + Si (0.306%) +Zn (0.0002%) +Fe (0.20%) + Mn (0.008%) + Mg (0.705%) +Cr (0.005%) +Ti (0.01 % Uncoated carbide (SNUN 120408, ISO K 20). Dry turning Diameter = 30mm and length = 50mm Depth of cut (mm) 0.3 0.6 1 The experiment was conducted to analyze the effect of design parameters such as feed rate, cutting speed and depth of cut on surface roughness and cutting forces. a 3 factor, 3 level experimental runs were carried out on aluminum alloy using uncoated carbide flat top insert. The experiments were carried out in dry condition on PSG Trainer Lathe (A141) which had a maximum spindle speed range of 1600 rpm. The cutting forces induced during turning for Figure 1 Experimental Set up 459-2
5 th International & 26 th All India Manufacturing Technology, Design and Research Conference (AIMTDR 2014) December 12 th 14 th, 2014, IIT Guwahati, Assam, India 2.1 Experimental Result analysis The experimental results of 27 runs are given in graphical form. The main effect plot of influence of design parameters on surface roughness and cutting forces is given in Figure 3 and 4. Figure 5 shows the surface roughness being measured using surface roughness tester after the experimentation. The relative importance amongst the cutting parameter levels is determined more accurately in ANOVA analysis. ANOVA analysis is performed both for the cutting forces and the surface roughness in order to find out the most influencing parameter affecting these responses. It is clearly seen from the graphs that surface roughness increases as feed rate increases, this may be because when feed rate is large compared to the smaller nose radius, the surface roughness depends mainly on feed rate compared to nose radius. The plastic flow is opposite to feed direction with higher height at low feed rate which can also lead to higher roughness at low feed rate. Though the theory suggests roughness to be a function of square of feed rate, practically it is more directly related to feed rate, this can be due to tool work relative vibrations. From plot 2 it is evident that the surface roughness decreases as cutting speed increases, this may because of the fact that there is continuous reduction in the buildup edge formation for higher cutting speeds. Figure 3 Effects of cutting parameters on Ra a. Plot 1: effect of feed rate on surface roughness when v=d=c b. Plot 2: effect of cutting speed on surface roughness when d=f=c. c. Plot 3: effect of doc on surface roughness when v=f=c Figure 4 Effects of cutting parameters on Forces a. Plot 1: effect of depth of cut on forces when v=f=c b. Plot 2: effect of feed rate on forces when d=v=c. c. Plot 3: effect of cutting speed on surface roughness when f=d=c 459-3
CORRELATION AMONG THE CUTTING PARAMETERS, SURFACE ROUGHNESS AND CUTTING FORCES IN TURNING PROCESS BY EXPERIMENTAL STUDIES From figure 4, it is evident that while turning Aluminium alloy 6063 cutting force increases as feed rate increases; this may be due to the fact that when feed rate increases the amount of material coming into contact with the cutting tool increases, therefore the load on the tool also increases which in turn increases the cutting force. The effect of depth of cut on cutting forces is because as the depth of cut increases the volume of the uncut chip also increases which produces more resistance on the cutter, thus increasing the cutting force 2.2 ANOVA of Surface Roughness and Cutting force ANOVA analysis is done in order to find out the significance of each parameter influencing surface roughness and cutting force, the interactions between various factors of an experiment can be quantitatively determined by using the analysis of variance (ANOVA). Table 4 and 5 summarizes the ANOVA analysis performed using MINITAB16 software. Sourc e Table 3: Analysis of Variance of Surface Roughness Sum of squar es Dof Mean squares P value P.C (%) v 0.152 2 0.076 0.03 19.19 f 0.560 2 0.280 0.002 70.3 d 0.016 2 0.008 0.264 2.01 vd 0.006 4 0.0017 0.573 0.86 Fd 0.014 4 0.0036 0.265 1.79 vf 0.028 4 0.007 0.079 3.51 Error 0.017 8 0.0022 2.24 Total 0.796 26 From the ANOVA analysis it is seen that surface roughness was most affected by feed rate (70.35 %) followed by cutting speed (19.19%). The interactions of design parameters do not have a major effect on surface roughness. Table 4 Analysis of variance of Cutting Forces Sourc e Sum of squares D of Mean squares P value P.C (%) v 1235 2 618 0.551 0.64 f 8811 2 4406 0.193 4.57 d 168477 2 84238 0.037 87.3 fd 2801 4 700 0.490 1.45 vf 2924 4 731 0.472 1.51 Error 5992 8 749 3.10 Total 192827 26 The resultant cutting force was most affected by depth of cut (87.37%) followed by feed rate (4.57%). Here also, the interaction of the design parameters do not had much influence on the cutting forces. 3 Prediction Models Based on the experimental results, the statistical analysis software system MINITAB 16 is used for multiple regression analysis. Another prediction using artificial neural network in MATLAB is also performed in order to check the prediction accuracy. 3.1 construct the prediction model with regression analysis Multiple regression analysis is a statistical technique to assess the association between the dependent variable and more than one independent variable. The process parameters, resultant cutting forces and the surface roughness values obtained during the experimentation where given as input to the software. The data presented in Figure 3 and 4 have been used to build the multiple regression models. A regression equation was developed for each desired output. The regression coefficients are estimated with the least square method using MINITAB 16. Accordingly, the second order fitted model for surface roughness and cutting force is given as follows: Ra = -0.86 + 0.00715v + 1.73f + 0.536d 0.0182vf + 0.333fd + 0.00171vd 0.000063v 2 + 1.99f 2-0.539d 2. (1) The feed rate and the interaction of cutting speed and feed were the most dominant factors on surface roughness. There is a good correlation between surface roughness and the cutting parameters. The significance of multiple regression coefficients for second order model is 0.967, and therefore the second order model can explain the variation with an accuracy of 96.7 %. ANOVA test was again used to determine the dependency of surface roughness to the selected machining parameters. It is clearly seen from Figure 6.1 that there is a strong relationship between the predicted variable (MR model) and the response variable (experimental value). The average absolute error between the predicted and experimental values of surface roughness (Ra) is 3.347 %. 459-4
5 th International & 26 th All India Manufacturing Technology, Design and Research Conference (AIMTDR 2014) December 12 th 14 th, 2014, IIT Guwahati, Assam, India Figure 5 Experimental and predicted data of Surface Roughness for 27 runs. F Resultant = -463 + 5.26v + 2436f + 40d 0.033v 2 3205 f 2 + 57.5d 2 6.23vf + 379fd + 0.59vd. (2) The depth of cut and feed rate were the most dominant factors on the resultant cutting force. The significance of multiple regression coefficients for second order model is 0.939, and therefore the second order model can explain the variation of the resultant cutting forces with an accuracy of 93.9 %. The average absolute error between the predicted and experimental values of the resultant cutting force is 6.819 %. ANNs are widely used in forecasting, pattern recognition, vision system etc. Back Propagation algorithm (BPA) is one of the most frequently used training algorithms in neural networks, but it suffers from slower convergence. Lavenberg Marquardt (LM) is a relatively faster algorithm and therefore BPA with LM is used for training the network. The cutting speed, feed rate and depth of cut are the design parameters. The input layer of the network houses the design parameters in 3 nodes and the output layer houses the response factor in 1 node. A trial and error approach was adopted to determine the number of neurons in hidden layer. For the surface roughness and cutting forces, the best approach having minimum mean squared error is achieved with five neurons in hidden layer. The neural network architecture was modeled in MATALB nntoolbox. Neural network architecture which provides the best prediction accuracy is shown in Fig. 8.The learning parameters of the proposed ANN structure are presented in Table 5. Figure 8 Neural Network Architecture Figure 6 Experimental and predicted data of cutting forces for 27 runs. 3.2 construct the prediction model with artificial neural network (ANN) Artificial neural networks (ANN) are information processing systems which has the capacity to reproduce human behavior. It has the capability to analyze complex relationship between various factors that affect a particular response. The learning parameters of the proposed ANN structure include Log sigmoid as the activation function, 0.05 learning rate, 0.95 as momentum constant. To test the prediction accuracy of the developed neural network model, the model was tested with the validation data selected as the last 6 experimental runs. The experimental data contained 27 runs, in which 21 runs were utilized for training the network and 6 runs were utilized for testing the performance of the trained network. The performance criteria considered are the mean absolute percentage error and correlation coefficient (R 2 ). Figure 8 and 9 shows the comparison of measured and predicted data of surface roughness & cutting forces for the training and the testing stages. The ANN results showed that the proposed model in this study is suitable for predicting the surface roughness and the cutting forces. The performance characteristics such as the mean absolute percentage error and the correlation coefficient are in acceptable ranges. 459-5
CORRELATION AMONG THE CUTTING PARAMETERS, SURFACE ROUGHNESS AND CUTTING FORCES IN TURNING PROCESS BY EXPERIMENTAL STUDIES experimental values of Surface roughness is 2.868 % and for cutting forces is around 3.807 %. It is well established that ANN models predict surface roughness and cutting forces with high accuracy than multiple regression models. Figure 8 Comparison b/w measured and predicted data of Ra in a. training stage b. testing stage Figure 9 Comparison b/w measured and predicted data of cutting forces in a. training stage b. testing stage 4 Conclusions In turning Aluminium alloy, use of lower feed rate (0.25 mm/rev), higher cutting speed (76 m/min) and lower depth of cut (0.3 mm) are recommended to obtain better surface finish & minimum cutting forces for the specified test range. For surface roughness, feed rate is the main influencing parameter with a percentage contribution of 70.35 % followed by cutting speed and depth of cut. For resultant cutting forces, depth of cut is the most influencing parameter with a percentage contribution of 87.37 % followed by feed rate and cutting speed. The average absolute error between the predicted (MR) and experimental values of surface roughness is around 3.347 % and for cutting forces is around 6.819 %. Average absolute error between predicted (ANN) and References Gaurav bartarya, S.K.Choudhury, (2012), Effect of cutting parameters on cutting force and surface roughness during finish hard turning AISI52100 grade steel, CIRP,Procedia, pp.651-656. Iihan asilturk, Mehmet Cunkas,(2011), Modeling and prediction of surface roughness in turning operations using ANN and multiple regression method, Expert systems with Applications,pp. 5826-5832. S.Gajanana, L.Shiva Rama Krishna, Srinivasa Rao Nandam, B.K Mohan, (2012),Optimization of process parameters of end milling process using factorial DoE, Manufacturing Technology Today,pp.5-13. Vikas Upadhyay, P.K.Jain, N.K.Mehta, (2013), In Process prediction of surface roughness in turning of Ti-6Al-4V alloy using cutting parameters and vibration signals. Measurement, pp.154-160. John.A.Patten,Jerry Jacob, (2008), Comparison between numerical simulation and experiments for single point diamond turning of single crystal silicon carbide, Journal of Manufacturing Process, pp.28-33. P.G.Bendaros, G.C.Vosniakos,(2003), Predicting surface roughness in machining: a review, International Journal of Machine Tools and Manufacture, pp.833-844. Santosh Tamang, M Chandrasekaran,(2013) Multiresponse optimization of surface roughness and tool wear in turning Al/SiC particulate metal matrix composite using taguchi gray relational analysis,manufacturing Technology Today,pp.4-21. L.b abhang, M Hameedullah, (2012),Optimization of machining parameters in steel turning operation by taguchi method, Procedia engineering, pp. 40-48. Ranjit.K.Roy, (1999), A Primer on the Taguchi method, Society of manufacturing engineers, Dearborn, Michigan. E.Daniel Kirby, Zhe Zhang, Joseph C Chen,(2004), Development of an accelerometer based surface roughness prediction system in turning operations using multiple regression techniques, Journal of industrial technology,pp.2-8. P.K jain, NK Mehta, (2013), In process prediction of surface roughness in turning Ti-6Al-4V using cutting parameters and vibration signals, Measurement, pp154-160. 459-6