Investigations of surface quality and energy consumption associated with costs and material removal rate during face milling of AISI 1045 steel

Machining of AISI 1045 steel is prominent in several industries due to their good machining characteristics. In this study, the optimum conditions of fly (face) milling of parts made of AISI 1045 steel was analyzed. The generated surface quality, the cost of the cutting tool components, the energy consumption, the wearing of the cutting tool, and material removal rate are the main parameters in this study. Several cutting experiments over different cutting lengths have been conducted and analyzed statistically to determine the optimum targeted cutting conditions. A multilayer regression analysis was conducted on obtained experimental results and inducing non-linear mathematical equations with high coefficient of determination (R2 = 0.98). The influence of feed per tooth (fz), cutting speed (vc), flank wear (VB) to surface roughness (Rz), cutting power (Pc), material removal rate (MRR), sliding distance (ls), and the tool life (T/) has been considered. The overall results, estimated through Grey relational analysis (GRA), revealed that the optimum fly milling performance for a fast manufacturing (case 1) are obtained for feed per tooth fz = 0.25 mm/tooth, cutting speed vc = 392.6 m/min, and machined length l = 5 mm. While the optimum parameters for resource (tools) conservation (case 2) are feed per tooth fz = 0.125 mm/tooth, cutting speed vc = 392.6 m/min, and machined length l = 5 mm.


Introduction
Face milling of structural materials such as AISI 1045 steel is very important for modern manufacturing applications. Products made from such materials are widely used in critical applications in the automotive, shipbuilding, and many other industries. Currently, for modern production, it becomes the most urgent task in addition to ensuring the requirements for machining accuracy (surface roughness), as well as the need to ensure the sustainability of production. Environmental and social aspects must be taken into account in order to ensure the efficient use of available resources. In this growing world, the demand for energy-efficient manufacturing processes is rapidly increasing because of strict environmental concerns. However, the power consumption and cost issues in manufacturing processes are two important factors from the perspective of the machining system that directly oblige the manufacturer to search for environmentally friendly processes [1]. Therefore, the main objective of this paper is to perform an experimental study especially related to two research objectives, i.e., power consumption measurement and cost estimation of the face milling process. Also, the complex influence of flank wear of face mills on such indicators as surface quality, power consumption measurement, and cost estimation is discussed. It is important to know such trends, and it is even better to be able to control such parameters as the tool wear increases. It is important to show this not only for some average wear values that are accepted for the disastrous but also to investigate the behavior of the output parameters of face milling outside these average wear values.
Face milling is a highly efficient machining process, widely used in manufacturing of flat surfaces [2][3][4][5][6]. At present, there are studies devoted to face milling of AISI 1045 steel (an analog of steel 45 considered in this article). Srivatsan et al. [7] presents the effect of cutting depth and tool speed on the residual stresses that occur when milling AISI 1045 steel. Padma Ooh et al. [8] in their article performed studies of residual stresses in AISI 1045 steel caused by milling. However, in these works [7,8], the process of face milling is studied without taking into account tool wear. D'Errico et al. [9] investigated seven commercial inserts of metal-ceramic with dry face milling of carbon steel (AISI-SAE 1045). Richetti et al. [10] evaluated the flank wear curves for AISI 1045 and 8640 plates of steel using 1, 2, 3, and 6 inserts in a face milling cutter. Muñoz-Escalona et al. [11] developed the empirical models to predict tool wear mechanisms during milling AISI 1045. However, in these papers [9][10][11], the process of tool wear without examining the relationship with roughness is investigated. Moreover, it is important to ensure the critical characteristic of flat-surfaced workpieces, i.e., surface roughness Rz ("maximum hight of the assessed profile" according to ISO 4287:1999). Pimenov et al. [12] showed the influence of relative position of the cutter relative to the workpiece and the kinematics of milling on the components of the cutting forces, the spindle acceleration of the machine during the face milling of steel SAE 1045. Ali et al. [13] investigated the surface roughness, material removal rate, and cutting time for torch milling operations. Toledo et al. [14] analyzed the effect of the relationship between the length of the parallel surface of the secondary cutting edge and the feed per tooth (b s /f z ) on the surface roughness of AISI 1045 steel during face milling. Pimenov [15] introduced the influence of feed, cutting and cutting on the roughness of flat surfaces in the face milling of steel 45. And this criterion is important for the quality of the treated surface.
At the same time, face milling consumes a significant amount of energy, which is caused by a larger tooth wear area. In addition, it is important to process to minimize machining energy. Henceforth, researchers have explored many strategies to minimize the machining energy (i.e., power) consumption. For instance, Hu et al. [16] strategically optimized the machining sequence to minimize the machine tool energy consumption. The machining scheme composed of milling and drilling operations on C45 steel. They have claimed that selecting the apt route can significantly reduce the power consumption in machining. In another paper, Hu et al. [17] attempted to reduce the energy consumption in machining for tool change and tool path; a 28.6% reduction is noticed in energy consumption caused by feature transitions. Even, that optimum change caused 27.95% reduction in machining time. Considering the extreme importance of energy consumption in machining, Li et al. [18] formulated specific energy consumption and power models for face milling operations with reliability above 96%. These models were formulated in terms of MRR and cutting speed, and authors claimed that those models are useful for estimation of power without measurement. Aramcharoen and Mativenga [19] identified the critical factors of energy modeling and influence of tool wear on energy intensity during machining. Their developed models predicted energy at 95% accuracy. They have also stressed that the reduction of tool wear can play effective role in reducing the energy consumption. An intelligent model was proposed by Garg et al. [20] for conserving the energy consumption in milling machining. They have employed the advanced evolutionary algorithm, i.e., multi-gene genetic programming technique. Another energy optimization study was reported by Albertelli et al. [21] with the novelty that they considered, in their model, the energy consumed by the auxiliary systems of a machine tool in face milling process. In fact, the authors have correlated the energy parameters with cutting parameters. Most notably, the influence of changes in tool wear in the machining process was incorporated in the model too. They have emphasized that the appropriate selection of process parameters is imperative to reduce the energy and time. On the other hand, Garg et al. [22] studied and modeled the tool life and power consumption for machining using three advanced modeling methods. The inputs of the models were cutting speed, nose radius, cutting depth, and feed rate. Then statistical comparison was conducted to suggest the best model-the genetic programming model. Shnfir [23] studies the machinability of AISI 1045 hardened steel during face milling using ceramic inserts based on SiAlON and whisker (SiCW). And it gives the influence of cutting parameters, milling configurations, edge preparation, and hardness of the processed material on machinability indicators, such as the resulting cutting force, power consumption, and wear of the side tool. However, in the above studies, the cost of machining was not considered. In Khan et al. [24], multipurpose optimization was performed by integrating the Taguchi method, Grey's relational analysis (GRA), and the non-dominant sorting genetic algorithm (NSGA-II). Where the output parameters are surface roughness and active cutting energy, taking into account the removal rate of the material, but excluding the cost of processing.
Besides, tool wear increases along with the sliding distance (friction distance), and tool life determines the tool cost. In modern manufacturing, minimizing machining cost per workpiece is very important. It is therefore important in designing face milling operations to ensure the design surface roughness and minimize the power consumption and cost of machining at the same time. This approach allows for saving resources in manufacturing the end item.
Therefore, the current studies are dedicated to different machinability aspects of face milling. However, there is a scarcity of study regarding the sustainability achievement in machining by power consumption reduction and cost performance improvement. A comprehensive optimization study was conducted by Yang et al. [25] for face milling. They have optimized the cost of production, time, and rate of profit while maintaining the constraints of force, power, speed and feed and surface roughness. Yang et al. [26] employed advanced modeling technique, i.e., gene expression programming to model the energy consumption in face milling. Wang et al. [27] performed multi-response optimization for reducing the cost as well as the energy consumption using evolutionary algorithms for face milling. For instance, Sales et al. [28] investigated the performance of MQL in milling of AISI 4140 steel by considering surface roughness values, flank wear, and tool life as an input process parameter. Singh et al. [29] performed the milling experiments on Inconel-718 and the performance in terms of tool wear with respect to the milling process parameters was evaluated by an evolutionary algorithm. On the same content, Gupta et al. [30] applied the two evolutionary algorithms for optimizing the turning parameters under MQL conditions. Then, Siller et al. [31] discussed the influence of specially designed carbide tools on the performance (surface quality and tool life) of AISI D3 steel during the face milling operation. Cui and Zhao [32] examined the important machining indices in terms of tool wear mechanism, chip and surface characteristics in high-speed face milling of AISI H13 steel. In another face milling operation of Ti-10V-2Fe-3Al (Ti-1023), Houchuan et al. [33] discussed the effect of cutting speeds along with the average flank wear values on surface characteristics, defects of machining, microhardness, and microstructure variations values. Similarly, the influence of damaged inserts on surface roughness values during high-speed face milling of 17-4 PH steel was investigated by Liu et al. [34]. It is appreciable that general machinability aspects are studied by the above researchers. Some critical aspects such as cost, quality of the product, and power consumption are missing in many studied. This fact encouraged the authors to pursue this comprehensive study.
At present, many researchers use advanced modeling techniques to study surface roughness in face milling. Bruni et al. [35] implemented the artificial neural network technique to analyze the effect of lubricant-cooling techniques on surface roughness values during face milling of AISI 420 B steel. Sahu and Andhare [36] used the response surface methodology for estimation of power, productivity, tool wear, and surface roughness in high-speed milling of Ti6-AL-4V alloy. Siwawut et al. [37] investigated the machining behavior and wear properties of Co-WC coated inserts in dry face milling of cast iron. Studies use artificial intelligence to establish the correspondence between face milling parameters and the resulting surface roughness taking into account tool wear. However, no qualitative studies of the influence of face milling parameters on factors such as surface roughness, tool life, cutting power, and machining cost have been conducted in the literature. Managing cutting power and minimizing cost per workpiece enable us to promote sustainable manufacturing. It is also important to take into account tool life and flank wear that are affected by the cutting parameters.
The main objective of the technological process is to provide the required accuracy of processing. This is possible if there is dependence or model for predicting roughness on cutting conditions. Secondly, it is important to ensure the required surface roughness for the minimum material removal rate. Third, it is important to minimize processing costs by optimizing this parameter. Knowing the cutting force and cutting power, it is possible to choose the cutting conditions that reduce energy costs. In turn, this has reduced emissions from power plants for electricity generation. Thus it is important to have a complex relationship between the surface roughness, material removal rate, the processing cost, and cutting power at the outlet and inlet cutting conditions. It also requires multiparametric optimization of the entire complex of the data.
The aim of this study is to investigate patterns connecting feed per tooth (f z ), cutting speed (v c ), and the flank wear (V B ) to surface roughness (maximum hight of the assessed profile) (Rz), cutting power (P c ), material removal rate (MRR), sliding distance (l s ), and tool life (T / ) and determine the optimal cutting conditions, therefore, to provide for the design surface roughness while decreasing the cutting power and minimizing face milling costs at the same time.

Experimental procedure
The experiments were performed by considering the environemnetal aspects. The details of materials and equipment used in current work is discussed below.

Experimental conditions
For complex evaluation of surface roughness and minimizing the power consumed by the cutting operation as well as machining cost in fly milling, experimental studies have been conducted to measure the various tool flank wear values. The workpiece material used was AISI 1045 steel (the composition of high-quality structural carbon steel 45 in accordance with the Russian national state standard (GOST) 1050-99). The actual chemical compositions of test specimens are given in Table 1.
In general, the microstructure of the annealed/normalized AISI 1045 steel consists of ferrite and pearlite. As shown in Fig. 1a, the microstructure of the tempered specimen consists of ferrite (white areas) and tempered martensite (dark areas). In order to reveal the precipitates, the microstructure was examined using scanning electron microscopy (SEM). Consequently, as shown in Fig. 1b, dark and white areas represent ferrite and carbides, respectively.
Surface roughness Rz was measured using a profilometer Abris-PM7.0, which is a stylus-instrument (GCI SI VNIIMS, Moscow, Russia). The readings were taken for the base length L / = 0.4 mm at the start, the middle, and the end of the pass of the mill. Every experiment had, therefore, 3 × 5 iterations (k = 15).
The flank wear values, as well as machined surface values, were noted after each pass of the mill. This way, the experimental points of surface roughness were obtained for various flank wear areas and fly milling parameters. Statistical processing was then carried out on the experimental data to achieve the statistical design reliability of 0.95. The average values of the measured parameter were established based on the data from 5 experiments. Homogeneity of the sampling variance was checked using Cochran's Q test. Figure 2 presents the flank tooth of the fly mill. Table 4 lists the experimental and estimated data for the cutting parameters presented in Table 3.

Calculations for various stages of fly milling
Sliding distance, l s , is determined by Eq. (1): where D Sl is the length of the tooth mill sliding trajectory (D Sl = 80.4 mm).
Cutting power, P c , is determined by Eq. (4) [38]: where K 1 ¼ ffiffiffiffiffiffi 3:25 p ffiffi 3 p ¼ 1:08 is a coefficient that represents the ratio of normal cutting force and shear force components [38]; K 2 = 0.41 and K 3 = 0.59 described as the coefficient of horizontal asymptote and the degree of the damping exponents [39], a / is equal to the absence of a portion of the radius section on the top of the tooth and is equal to R mi (1 − cos k r ) in the presence of that portion [39]; b / is equal to a p [39]; k r is the angle of cutting edge; a p is the depth of cut; f z is the feed per tooth; f is the friction worked stock ratio on the flank surface of the mill tooth cutting point; β is the angle of action; Φ is the angle of shear; σ i is the stress intensity [40] (see physical and mechanical properties of the steel 45 (AISI 1045 steel) in Table 6 [40]) (the intensity of stress is a function of the intensity of strain, ε, the strain rate,ε, and the temperature, T 0 , of the material: σ i = f(ε;ε; T 0 ); dl is the elemental length of the cutting edge; V B is the flank wear on the tool [40,41]; ψ i is the angular coordinate of the ith tooth; and, i = x, y, z represents each axis of the coordinate tool system).
Material removal rate, MRR, is determined by Eq. (5): Equations (1-5) were used to determine the values of the parameters listed in Table 4.

Optimization by Grey relational analysis
The optimization by the Grey relational analysis (GRA) is performed by considering the optimization problem as a "multi-objective optimization." Whenever more than one response is optimized, GRA stands out as an effective method to solve the optimization. In the current study, the surface roughness, part processing cost, cutting power, and material removal rate are granted as the responses-in  I. Preprocessing of data: The surface roughness, part processing cost, cutting power, and material removal rate have a different scale of magnitudes. Before proceeding ahead, it is imperative to convert different scales into a single scale, from 0 to 1. This is done by normalization following Eq. 6 (minimization is the target) and Eq. 7 (maximization is the target).
II. Here, the experimental data (original) is indicated by x 1 (k); the normalized preprocessed data is represented by y 1 (k); also, the maximum and minimum values are presented by maxx 1 (k) and minx 1 (k), respectively. III. Grey relational coefficient: Next, the grey relational coefficient, which defines the relation of experimental value and ideal value, is calculated using Eq. 8.
IV. Here, the deviation sequence is presented △ 0i (k). The V. Grey relational grade: As mentioned earlier, the multiple responses are combined to a single function, i.e., Grey relational grade (GRG). The grey relational coefficients are merged into GRG. The conversion is associated with particular weight values for each response. The typical calculation scheme for GRG is shown in Eq. 9. Depending on the manufacturer's requirements, the weight value changes. In fact, controlling of weights to responses controls the ultimate optimum levels of parameters. For instance, in our current research, two different sets of weight values are accounted to address technological performance as well as the conservation of resources. The details are discussed later.
VI. Grey relational order: Once the GRG is computed, the highest value of the GRG is ranked as 1. The rest are ordered as in descending order. The experiment number, ranked 1, is the optimum run.

Multiple non-linear regression analysis
The multiple non-linear regression analysis was applied on the experimental data as presented in Table 4. In our case, it is advisable to use the mathematical apparatus of multiple classical correlations and regression analysis [42].  A preliminary assessment of the tabular data has established that an adequate mapping of the interrelations studied will provide non-linear regression equations. Given the combined nature of the non-linear relationship (positive and negative, increasing and decreasing, as well as equally accelerated and equally slow regressions) and successful experience of data approximation in work [43] are observed, the relationship between the four phenomena can be described by a second-order polynomial in four-dimensional space: where the sum of the coefficients with zero indices is denoted by d 0 = a 0 + b 0 + c 0 .
Since there is no universal method for selecting and rational regression curve, then we will judge the reliability of the approximation by the coefficient of determination R 2 , the value of which must be higher than 0.8 units.
As a result, a non-linear constrained optimization were selected coefficients to five equations for the normalized response function Rz*, cutting power P c *, material removal rate (MRR*), sliding distance l s *, and tool life T / * reliability criterion R 2 = 0.98: Using the obtained equations, a quantitative analysis was conducted to study the influence of the input parameters on each of the specified response functions. For surface roughness Rz*, it was found that with an increase of 0.    Fig. 5. Figure 5 shows that the flank wear V B on the material removal rate is not affected.
For sliding distance l s *, it is established that with an increase of 0.1 units (0.032 mm/tooth) in a coded form, the feed parameter f z * sliding distance decreases by 0.075 units     Table 9 Grey relational coefficient Exp. No.

Roughness, Rz
The cost price of processing one part, C  Fig. 7. Approximate the response function C* in the fourdimensional space, a second-order polynomial is not possible due to the harmonic nature of the dependence. Therefore, Fig. 8 shows the graphs of the cost price of processing one part C calculated by Eq. (3) points. Figure 8 shows that the cost price of processing one part C decreases as the flank wear increases V B decreases along a broken linear curve. In the characteristic inflection point of the curve, the rate of change of the parameter C decreases to almost zero. A further increase in tool wear above the characteristic value makes sense if the material removal rate is important with reduced requirements for roughness Rz and power P c .
After establishing the regularities of a complex multifactor process of milling, we proceed to find the optimal cutting conditions with a fixed flank wear V B = const = 0. Since only, in this case, the minimum values of the optimum in a multicriteria search can be ensured. Italicized numbers indicate the optimum runs Fig. 9 Visualization of optimal surface roughness Rz values found using optimization by Grey relational analysis Fig. 10 Visualization of optimal cutting power P c values found using optimization by Grey relational analysis Also as a result of studies of machining by milling AISI 1045 steel, it was possible to establish the following. The low plastic flow characteristics of AISI 1045 steel combined with its relatively high hardness made it easy to process and enabled good surface finish during its machining activities. The tool-workpiece brittle interaction leads to hard material separation resulting in better surface finish. Machined surface quality was degraded by increasing the feed rate and depth of cut. Higher cutting forces, friction, and worked incremental areas are, in fact, induced causing the poor surface finish manifesting in larger horizontal markings spacing. An increase of the depth of cut is also responsible for the growth of the vertical spacing separating peaks and troughs of the machined surface. Hence, increased feed rates and depth of cuts generally degrade the surface finish although they improve cutting performance. Consequently, it is highly important to find an optimum combination of the cutting parameters settings to reduce machining time and keep a high quality of surface finish. Optimization techniques (such as Grey relational analysis) or a "dense sampling" factorial design of experiment could be of great help to achieve the mentioned research motivation.

Optimization
Following the methodology of Section 2.2, multiple responses, i.e., roughness parameter, cost of the part, power consumption, and material removal rate, are simultaneously optimized using Grey relational analysis. The experimental/ computed data of Table 7 are used in the optimization. The preprocessed sequence is developed by using Eqs. 3, 4 (for Rz, C, P c ) and using Eq. 5 (for MRR), and listed in Table 4.
The deviation sequence, computed using "1 -Preprocessing sequence," is tabulated in Table 8 Later, the grey relational coefficient is computed using Eq. 3 and listed in Table 9.
For GRC, the distinguishing coefficient is taken as ς = 0.5. At last, the Grey relational grade (GRG) is calculated by combining the GRCs of all the responses. The respective weights for the responses were determined from two perspectives. First, if the technologist is faced with the task of making the part with the greatest performance, the logical deduction is to emphasize on the product quality and on the material removal. That means the surface roughness and material removal rate are assigned higher weights than the other two responses. As such, in the current study, case 1 has considered w Rz = 1.0, w C = 0.5, w Pc = 0.5, and w MRR = 1.0. In the second case, if the main objective is to conserve resources maintaining product quality acceptable, then more importance is delegated to surface quality, cost of producing single part, and the consumption of power, and less weight is exerted on material removal rate. As such, the weights for case 2 are w Rz = 1.0, w C = 1.0, w Pc = 1.0, and w MRR = 0.5. Note that in both cases the product quality was not compromised. Table 10 shows the GRG and respective rank for both cases. For case 1, experiment number 7 is found as the optimum run; the corresponding input parameters for this case are feed per tooth f z = 0.25 mm/tooth, cutting speed v c = 392.6 m/min, and machined length l = 5 mm. For case 2, the optimum parameters values are feed per tooth f z = 0.125 mm/tooth, cutting speed v c = 392.6 m/min, and machined length l = 5 mm. Therefore, it is visible that the change in the requirements from "performance" to "resource conservation" has entailed a different optimum result; here, the cutting speed and machined length though are the same, the feed per tooth is reduced from 0.25 to 0.125 mm/tooth. Fig. 11 Visualization of optimal material removal rate (MRR) values found using optimization by Grey relational analysis Fig. 12 Visualization of the optimal cost price of processing one part C values found using optimization by Grey relational analysis For fixed flank wear V B = 0, the surfaces are shown, showing the surface roughness Rz (Fig. 9), cutting power P c (Fig. 10), material removal rate MRR (Fig. 11), the cost price of processing one part C (Fig. 12) depending on cutting speed (v c ), and feed per tooth (fz). При Points "1" and "2" denote optima for case 1 and case 2 (see Table 10), respectively.

Conclusions
The research conducted has shown that: & The tabular dependence of input parameters (cutting speed v c , feed per tooth f z , flank wear V В ) and output parameters (surface roughness Rz, cutting power P c , material removal rate MRR), cost of producing single part C is obtained for fly milling of carbon AISI 1045 steel. In such an integrated formulation for milling AISI 1045 steel, the task was posed for the first time, which made it possible to evaluate the accuracy and resource-saving characteristics, including at high values of tool wear providing its cutting ability. & For the first time with high accuracy (R 2 = 0.98) for face milling of AISI 1045 steel by multilayer regression analyses identified non-linear laws of surface roughness (Rz), cutting power (P c ), material removal rate (MRR), sliding distance (l s ), and tool life (T / ) depending on feed per tooth (f z ), cutting speed (v c ), and flank wear (V В ), which can be easily integrated into CNC machines. & Multi-objective optimization for the product obtained by the fly milling of carbon AISI 1045 steel using Grey relational analysis (GRA) showed that the optimum parameters for improving the manufacturing efficiency and reduce the machining time (case 1) are as follows: feed per tooth f z = 0.25 mm/tooth, cutting speed v c = 392.6 m/min, and machined length l = 5 mm; for resource-saving (case 2), the optimum parameters are feed per tooth f z = 0.125 mm/tooth, cutting speed v c = 392.6 m/min, and machined length l = 5 mm.
Funding information This work was financially supported by the Deanship of Scientific Research at King Saud University through research group no. RGP-1439-020. The research was also supported through Act 211 Government of the Russian Federation, contract no. 02.A03.21.0011.