Parametric optimization of hot forging process: a six sigma based approach

One of the major concerns for industries in the modern world is to focus efforts on producing high quality products with minimal costs. Various quality improvement philosophies have emerged in recent times, Six Sigma being one of the most practical and efficient techniques for quality improvement of processes. In this work, Six Sigma based DMAIC (Define, Measure, Analyze, Improve, Control) approach is used to enhance productivity and quality performance, and to make the hot forging process robust to quality variations. Finite element method has been employed for the simulation of hot forging of the connecting rod. The influence of design and process parameters is investigated for the response forging die load . Analysis of various critical parameters and the interaction among them has been carried out with the optimize the response and make the analysis more precise and robust, response surface methodology has been incorporated. Parameters have been optimized, leading to the accomplishment of a minimized forging die load which is verified using a confirmation experiment. Confirmatory results reveal the potential of the DMAIC approach of Six Sigma in optimizing the process parameters successfully and thereby present significant applicability in the industry.


Introduction
Forging is a traditional metal working technology that has been used by humans for several centuries. On the basis of temperature of the work-piece, forging is classified as cold, warm or hot forging. In hot forging, the metal work-piece or billet is heated at high temperatures to improve ductility and then it is plastically deformed using localized compressive forces imparted by hammers or presses. The work-piece deforms plastically to fill the cavity in the die which is designed according to the shape of the desired product. [1] In internal combustion engines, a connecting rod connects the piston to the crankshaft. The connecting rod acts as a link and converts the reciprocating motion of the piston into the rotary motion of the crankshaft. Connecting rods are manufactured in developing countries like India, through the conventional hot closed die forging that involves flash formation. Typically, a cylindrical work piece or billet is heated and deformed plastically in various stages to yield a connecting rod. These stages include blocking, finishing and trimming operations. The work piece takes the final shape of connecting rod in the finishing operation, which is the last stage of forming before the trimming operation. [2] Hot forging process is advantageous over other manufacturing processes like casting or machining, as forged products offer higher strength than cast or machined products. This is because the internal grain structure of the forged part varies continuously as it deforms with the work piece during forging. Process design for hot forging usually follows a bottom up approach, with repeated iterations to study and improve the process. Thus, cutting costs and reducing the time consumed becomes a challenging task. Hence there is a need to study the process parameters involved in hot forging operation, and to optimize them. [3] Six-Sigma has been hailed as one of the most viable techniques in research in the recent times. Many researchers have employed six-sigma based techniques to study and improve a wide variety of industrial processes [4]. It is an analytically rigorous continuous improvement process. Over the years, Six Sigma has become an effective management philosophy and has served as a model for the industry to follow. [5] This work focuses on the implementation of Six Sigma philosophy to optimize the conventional hot closed die forging process. A widely used technique to implement Six Sigma is the DMAIC approach (Define, Measure, Analyze, Improve and Control) (Koning and Mast, 2004) [6]. Using this approach, a lot of research has been done to improve various process outcomes which drastically improve quality, reduces costs and time consumed. Figure 1 illustrates the general flow of Six Sigma based DMAIC approach. In this work, we have studied hot closed-die forging through the DMAIC approach of Six-Sigma and organized our work as per the steps involved in the '0$,& DSSURDFK 7DJXFKL ¶V PHWKRG RI H[SHULPHQWDO design has been employed to study the relationship of various design and process parameters with the response variable. Computer based simulation has helped us to conduct trials in a time and cost effective manner. DEFORM 3D, a Finite Element Method (FEM) based software package has been used to run simulations. Statistical analysis has been used to establish a functional relationship between the response and factors of the experiment. A regression equation is built to generate response surface plots. Using the Response Surface Methodology (RSM), the factors of the experiment are further fine tuned in the optimum range. The predictions of RSM are finally confirmed through experimental simulation. Rest of the paper has been structured as per the steps involved in the DMAIC approach.

Define (Hot Closed Die Forging)
In this phase of the DMAIC implementation, the chief task is identifying influential parameters and describing the whole process of hot forging operation.
'LH ¶V SHUIRUPDQFH VLJQLILFDQWO\ LQIOXHQFHV WKH quality of the product. Moreover, die life directly influences process cost. With an improper die, the product shall not be formed properly, thereby making die design a crucial stage in the process. In this work, the focus is on minimizing the die load developed during hot closed die forging of automobile connecting rods. With minimized die load, damage and wear are reduced and the die life improves. Since the product quality is influenced by die life and performance, optimizing die performance improves process economy and quality. Several factors have been found to influence die performance in the literature. These factors can be broadly divided into two groups ± the Design parameters and the Process parameters. Design parameters include Flash Thickness (Ft), Flash Width (Fw), Corner Radius (Rc), and Fillet Radius (Rf). These factors describe the geometry of the die. The second group of factors, called the process parameters include Billet Temperature (Tb), Die Temperature (Td) and Friction Coefficient (μ). [7] [8] Flash is formed during the forging operation of connecting rods. Flash is a scrap material and depending upon the complexity of the part, it can exceed even half the total volume of the formed metal. Flash formed during hot forging cools faster due to greater surface area and thus sticks to the die, acting as a pressure release valve for the almost incompressible metal. It restricts the outward flow of metal and ensures deeper die filling of the work-piece. Since the flash is a scrap material, it is later punched and removed from the formed product as waste product in the trimming operation. Thus the amount of flash produced also influences process cost and needs to be optimized.
Corner radius and fillet radius are also important parameters to be considered in die design as revealed in the literature. A sharp corner or fillet poses a risk of cracking or lapping, thereby increasing die wear, and hampering the die life and performance. A smooth corner also ensures better flow of metal. With improper metal flow, die filling is inadequate and quality of the product is compromised. This also increases the die load. Corner and fillet radius affect the grain flow. Sharp corners weaken both the dies and formed part. Thus they need to be optimized for least cracking and wear of the die. [9] Forging temperature is directly related to the workability of metal. Workability of the metal generally increases with temperature. Higher temperature ensures better die filling by improving the metal flow. This also reduces die load which reduces the risk of die wear. If the dies are too hot, the amount of heat transfer between the dies and the billet will reduce. Cooler dies will allow more rapid heat transfer and cool the surface layers of the work-piece considerably. Flash formation, which is directly related to product quality, is essentially formed by quicker cooling than the rest of the work piece.
The material of die and the lubricant are selected based on these limitations. Friction between the dies and work-piece affects the die load while forging, and hence the die life. Proper lubrication affects the coefficient of friction and thus the die load. [10,11] It has been revealed in the previous works that a deteriorating die performs poorly and thus product quality is compromised. A low die load during the forging operation improves die life. It can be seen that the die life and hence the product quality can be considerably improved with the optimization of design and process parameters for a lowest die load in forging operation. Die can fail prematurely if the die load in the forging operation exceeds the strength of die material. Low die load causes minimal damage to the dies, enhances die life and performance and hence the product quality and process economy.

Measure
Researchers in metal forming industries have conducted extensive research in this area to attain the optimization of hot forging variables. Srivastava et al. [12] have studied the effect of billet temperature and coefficient of friction on forging die load and strain rate. It was observed that the forging die load increased with increasing friction coefficient and decreased with an increasing billet temperature. Desai et al. [13]  observed that several parameters influenced the quality of forged gear. These included shear friction factor, punch velocity, aspect ratio and die temperature. Friction and die temperature has been studied in this work.
Ab-Kadir et al. [14] studied the effect of corner radius and friction while optimizing the cold forging die design. It was observed that a sharp corner (less corner radius) would introduce a high stress concentration, worsening the die performance and life. Die load required was observed to increase with larger values of friction coefficient. With a smaller value of the friction coefficient, the work-piece deformed almost uniformly. Higher friction increases wear and deteriorates the die life. With an increase in the flash thickness, the required die load was seen to decrease. Thus, the most favourable flash thickness has to be designed based on the press capacity, die life, and the number of products in order to minimize the total cost.
As discussed, it is found that there are 7 influencing parameters also known as control factors, for the underlying hot forging operation. These are the four design parameters: Flash thickness (Ft), Flash width (Fw), Corner Radius (Rc), and Fillet Radius (Rf) and three process parameters: Billet Temperature (Tb), Die temperature (Td), Friction Coefficient (μ). These seven control factors have been identified as the major influencing parameters. The material of the connecting rod is AISI±1045[1650-2200F (900-1200C)], whose properties are given in Table 1.  Figure 2 shows the cause and effect diagram drawn from the observed process conditions. Figure 3 shows the 3D rendering of upper and lower dies. Figure 4 shows the design of the product.

Analyze
In this stage of the DMAIC approach, the main task is validating a functional relationship between the control parameters and the response parameter, thereby identifying the critical parameters that significantly influence the response variable. 7DJXFKL ¶V PHWKRG RI experimental design is a practical methodology for this purpose. It allows studying the process with minimum number of experimental trials. Thus it greatly helps in saving time and cost. $QDO\VLV RI 7DJXFKL ¶V experimental design reveals the functional relationship between the input and output variables (Ghani et al., 2004). [15] 7DJXFKL ¶V H[SHULPHQWDO PHWKRG LV XWLOL]HG WR design the parameter settings for experimental trials, to identify the relative influence of each parameter considered in the study. [16]

Selection of Orthogonal Array
An orthogonal array is used in the 7DJXFKL ¶V PHWKRG The orthogonal array (OA) is a form of general fractional factorial design. It contains a representative set of all possible combinations of experimental conditions. Using the Taguchi method, all levels of all factors are considered equally. A balanced comparison of levels of the critical parameters is achieved. This significantly reduces the total number of experimental trials. [17]For simulation purposes, the component drawing of a threedimensional solid model of connecting rod (Fig. 4) and top and bottom die (Fig. 3) are done using CATIA V5. Finally, RSM is adopted to determine the optimum settings of design and process parameters and thereby optimize the response.
More than 2 levels of design and process parameters helps in studying the non-linear relationship between the factors and the response parameter. Thus, the design and process parameters each at three levels are considered important for the present study. The operating conditions under which experimental trials are conducted are given in Table 2. These levels have been selected after an ample literature survey. For example, Newberger and Mockel [3] have studied optimum flash thickness and flash width previously.
Product weight computed from CATIA V5 is nearly 340 g. Friction Coefficient has been chosen from inbuilt values of DEFORM 3D based upon the nature of lubrication (whether in one die, or on both dies or neither) in hot closed die forging. In hot forging, when both the dies are lubricated, the value of coefficient of friction is assigned as 0.3 for the considered material and conditions. For lubrication of neither die, the value of coefficient of friction is assigned as 0.7. And for lubrication of only one of the dies, the value of coefficient of friction is taken as 0.5. Dies are assumed as rigid bodies in the simulation software. The total degree of freedom (DOF) for seven factors each at three levels is 14. Signal to noise or S/N is a measure used in studies which is used to compare the levels of a desired signal to the level of background noise. The S/N ratio is treated as the quality characteristics evaluation index. Least variation and optimal design is obtained through analyzing the S/N ratio since stability of quality is dependent on it. Depending on the objective, there are three types of S/N ratio: the lower the better, the higher-the-better and the nominal-the-better. In present work, since the die load is of the characteristic ± smaller the better, the type of S/N ratio used is expressed as follows, S/N = -10 « Where, n = number of experiments and yi = die load. Data analysis is made using MINITAB 15 at a 95% confidence interval. [18]

Experimental Procedure
7KH '()250-3D V6.1 software package was used for the simulation and analysis of this work. Some assumptions have been made in this work, such as the plastic material model for billet and the rigid material model for die has been assumed. The velocity of the moving ram is kept constant at 1 mm/s. The strain rate at the initial deformation is equal to 1 s -1 . All the simulations started with 48000 tetrahedral elements of mesh. Table 3 shows the design of experiment using Taguchi L27 orthogonal array with S/N ratio data. Figure 5 shows the S/N ratio plots for forging die load (FL). [19] Table 4, it is observed that significant factors and their optimum levels are different for different responses. The results obtained IURP WKH 7DJXFKL ¶V experimental design can further be improved by using RSM or the response surface methodology (Box and Draper 1987) [20]. Hence, the response surface methodology is employed in this work to make the analysis more precise.

Analysis of Variance (ANOVA)
An ANOVA was performed to find the relative influence of the control parameters. In Table 5, DF stands for the degrees of freedom, Adj SS stands for the adjusted sum of squares, Adj MS stands for adjusted mean squares, Fvalue refers to the variance ratio and the P-value refers to the probability value. The F ratio is used to determine whether an effect is insignificant or not. Adj SS alone is not used to determine whether an effect is significant. The degree of freedom (DOF) is also an important measure LH DQ HIIHFW ¶V PHDQ VTXDUH 06 DOVR predicts if an effect is significant. The effect is weaker for smaller values of adj MS. Contrasting the results by using Adj SS and Adj MS to select the weak effects, Shiau (1989) [21] established that MS was the better measure to group the insignificant effects. Table 5 shows that the flash thickness and billet temperature are the most important contributing parameters affecting the forging die load under a 95% confidence level.
The analysis clearly shows the relevance of all factors taken into consideration in influencing the response, with Flash Thickness and Billet Temperature being the most influential respectively. The optimum parameter combination for forging die load (FL) is shown in Table 6.

Improve
In the Improve phase of DMAIC, the results obtained from the analysis of the 7DJXFKL ¶V H[SHULPHQWDO GHVLJQ are further optimized. To optimize the settings of critical parameters, response surface methodology (RSM) has been employed to establish a robust regression model and find the optimum set of parameters.

Response Surface Methodology
A lot of researchers have used this methodology to tackle several problems in the recent years. Tiwari et. al [22] have used response surface methodology in the improve phase (of DMAIC) to optimize hot radial forging variables. It is a sequential method for a series of regression model fits to the output variables (Shang et al., 2004). [23] Implementing the response surface methodology starts with a linear regression model: A low order (linear) polynomial is selected to approximate the true function in some regions of the independent variable (Shang et al., 2004) [23] . Thereafter, a higher-order (quadratic or higher) polynomial is employed to search for the general vicinity of the optimum region. In this work, the second-order model was directly employed to further augment the optimization process. This significantly saved analysis time. A typical fitted second-order model is of the form Least square estimate method is used to construe estimated regression coefficients given as: Where x is the process parameters and can be represented as On comparing the partial derivatives of the regression equation to zero, the optimum levels of process parameters can be found.
The point x0 = (x10, x20, . . ., xk0) is called the stationary point and is expressed as: In this work, 27 experimental trials were conducted as shown earlier. Based on the analysis of these trials, the significant contribution of the seven process parameters was analyzed using the F-value as shown in Table 7. ANOVA analysis also illustrates that the linear effect and square effect of each parameter and two-way interactions (among flash width and billet temperature) should be taken into account. Hence, the second-order model is considered to develop the regression equation between the input variables and the response i.e. die load. Table 8 depicts the X matrix. The coding is as x1 for factor A, x2 for B, x3 for C, x4 for D, x5 for E, x6 for F and x7 for factor G.  (12)

Control
Control, which is the final phase of the DMAIC approach, is a crucial stage of Six Sigma implementation. The entire work done is updated and preserved. Recording the data in the optimization process is crucial, as the lack of continuous improvement impedes six sigma efforts. With appropriate documentation, the process is constantly monitored and the documentation is maintained and updated at regular periods. Statistical quality control tools like control charts assist the monitoring of the process and alert personnel if the process has gone out of control at a point of time. [22] Table 7 shows the results obtained from the taguchi design and provides an insight into the process parameters affecting the forging process. The ANOVA reveals that the flash thickness is one of the most crucial and influential parameter; the billet temperature is the second most significant parameter; the interaction effect between flash thickness and billet temperature is also quite significant and must be considered in the die design stage of production. Response surface methodology was employed to optimize the set of parameters to attain a minimum die load. Table 10 gives the optimum conditions found by RSM for the hot forging of the connecting rod. The optimum parameter values were then applied to the process and confirmation experiments were done to check the results.

Confirmation Experiments
A confirmation test was conducted through computer based simulation with the obtained optimal combination of parameters at appropriate levels. Thus the quality characteristic ± die load, for hot forging of connecting rod was assessed. The conformation simulation was done at levels predicted by the Taguchi method and RSM. The optimum levels for forging die load EDVHG RQ 7DJXFKL ¶V experiment are Ft3, Fw1, Rc3, Rf1, Tb3, Td3, and μ1. The response values obtained from the confirmation experiment is a die load of 1.29 MN as shown in figure  8. The response values obtained from the confirmation experiment is a die load of 0.68 MN as shown in figure  9. The forging die load shows a reduced value from 1.29 MN to 0.68 MN respectively. The corresponding improvement was 47.3%.

CONCLUSION
In this work, the conventional hot closed die forging of connecting rod involving flash formation has been studied using the Six Sigma based DMAIC approach. FEM based simulation software package, DEFORM TM 3D V6.1 was used to simulate the hot forging process and study the effect of design and process parameters on the forging die load. Taguchi method as well as Response Surface Methodology (RSM) was employed to optimize the parameters for a minimum forging die load (FL). The confirmation test value for the forging die load is 1.29 MN. After performing ANOVA on the experimental results, four parameters namely flash thickness, billet temperature, friction coefficient, and corner radii were found to be the most significant parameters affecting the forging die load under a 95% confidence level. The conformation test value for forging die load, 1.29 MN obtained from 7DJXFKL ¶V PHWKRG is further improved from results obtained by response surface method. RSM has been employed successfully to obtain the minimized die load of 0.68 MN for the optimum setting of parameters. Future scope of work may involve extensive experimentation which may help to analyze the parameters in the real time working environment. Other techniques for analysis and parameter optimization maybe adopted. More parameters maybe included in future works. Same parameters with different and/or more levels maybe incorporated. Apart from the die load, other response parameters maybe included in studies involving die life.