Introduction

Martensitic stainless steels have a high strength-to-weight ratio and are used in high-strength applications such as medical instruments, shaving blades, and bearings. 420 grade martensitic stainless steel parades high resistance to corrosion as similar to type 410 but it is stronger and harder. In both the annealed and hardened states, it is magnetic. Only completely cured or fully hardened and stress relieved steels provide maximum corrosion resistance. It is hardly used in the annealed state. The microstructural composition and thickness of carburized layer produced on AISI 420 MSS may be adjusted by appropriately regulating the carburizing temperature. Micro-structural analysis reveals that minimum temperature and limited duration of heat treatment enhance the resistivity of carburized materials [1]. Because of the process, characteristics and microstructure, martensitic stainless steels deteriorate ploughing or abrasive wear. To increase the wear and corrosion resistance of martensitic stainless steels, surface treatment techniques and coatings are recommended. The most likely surface treatment method appears to be plasma nitride, which is an outstanding solution for improving the wear resistance of dry contacting surfaces [2]. Both industrial and theoretical techniques, assessment of tool life have a more scientific approach [3]. The surface integrity investigation of AISI 420 for various machining variables and machining condition states that the machining variables such as velocity of cutting, feed (F), nose radius and the tool are considered for microhardness of the machined outcome. It also discloses that the rate of tool feed is the most important output responses [4]. Carbide tools coated using TiAIN alternating layers outperformed uncoated tools in terms of tool flank wear and mean roughness [5]. Chips have a considerable impact on the machining process, especially the surface quality. With all F, a greater number of chips were created at low cutting (CS) [6]. In face turning, the consequences of WR have a direct influence on the surface quality of the machined product, whereas flank wear decreases somewhat with increasing feed and CS and cutting depth (CD) [7]. With a larger nose radius, the surface quality was noticeably improved. The results revealed that the CS is the most influential parameter on surface roughness (SR) [8]. The AM samples built at angular powder deposition against the horizontal base improved the mechanical properties and minimized the directionality challenges [9]. The results indicate that the tools costs are insignificant and considers cost effective with ecological machining condition [10].

The optimum machining setting for multi objective statistical ANOVA confirmed that F has the strongest influence on the SR [11,12,13,14,15]. The revolutionary changes in machine were appropriate choice of tool attribution to the development or invention of cemented carbide tools possessing superior properties [16, 17]. A researcher presented a new parameter-free ANFIS construction approach called KERNEL. It is capable of doing optimum architecture estimation as well as optimal SSD in a timely and effective manner. It can fine-tune the current knowledge base and in the absence of domain experience, it can uncover the ideal knowledge. A novel quick and flexible SSD approach based on hypercube sampling (HC) is given to allow optimal size determination [18].

In enterprises and academics, performing multi-objective optimization under uncertainty is a regular necessity. For scattered and discontinuous data in uncertain domain, it is difficult to get reliable data through robust optimization. An author presented a data-driven technique in which a unique fuzzy clustering process was employed alongside boundary creation to transcribe the uncertain space in order to identify particular locations of uncertainty. Following that, samples are created in the mapped uncertain zones using smart Sobol sampling [19]. A multi-objective optimization formulation under uncertainty has been presented, which can lead to simultaneous maximisation of productivity and reduction in energy consumption. Among the different strategies for dealing with uncertainty, Chance Constrained Programming (CCP) is thought to be the most efficient. DDCCP (Data-Driven CCP) is a revolutionary technique that combines machine learning algorithms with CCP, making the approach data-driven. A unique fuzzy clustering approach is applied to structure uncertain space in such a way that particular regions of uncertainty are precisely detected depending on the uncertain data for more realistic sampling, ultimately influencing the best solution accuracy [20].

The specimen cooled with chilled air has a higher resistance to corrosion behaviour after a longer period of time [21]. The tools with CVD coating presented a better performance than the PVD for the cutting conditions used [22, 23]. Undissolved carbides and tempering carbides were discovered in the martensitic matrix by using optical microscopy, scanning electron microscopy (SEM), and atomic force microscopy (AFM). When the spindle speed is increased, the SR decreases, which is important during turning operation [24]. Texturing is thought to alter tribological activity in the sliding region, as the results showed that the texture model that starts from a distance of 1 m performed better [25]. Untreated coated tools display obvious delamination of the coating, which has been successfully limited by using microblasting [26].

The use of the desirability function (DF) for multi objective performance parameter optimization reveals that there is no single best solution [11, 27]. The optimal combination of cutting speed, feed, cutting depth, and cooling mechanism for decreasing SR and maximising material removal rate (MRR) while maximizing tool life was determined using Taguchi-based entropy weighted grey relational analysis [28]. The machining characteristics (SR, cutting force) are determined by changing the range of cutting parameters [29]. Cold nitrogen gas and oil mist conditions boost machining responses as compared to dry machining. Avoiding emulsion cooling, lower waste costs, and longer tool life minimize unit production costs as compared to wet machining [30]. Abrasion and diffusion were the dominant wear mechanisms in martensitic stainless steel machining, while attrition and abrasion were dominant in super martensitic stainless steel machining [31]. TiAlN-coated tool steel has a far higher hardness than the uncoated tool steel [32]. As wear rate (WR) increases, the roughness characteristics change. The mean and total roughness metrics are essential for determining tool life based on surface quality [33, 34]. This paper mainly focuses on investigating the effect of TiAlN, AlTiN, and TiAlSiN coated inserts on the performance of AISI 420 steel by varying CD, F and CS. This paper also identifies the optimised process parameter which majorly influences SR and WR.

Materials and Methods

In this experimental study, uncoated CNC inserts obtained from Usman Tools, Coimbatore were utilized. The insert is finished into 40 mm diamond bar with length of 1.2 cm and nose radius 0.8 mm. The inserts with negative rake angle are aligned to the specification of CNMG-120408. Hard coatings such as titanium aluminium nitride (TiAlN), aluminium titanium nitride (AlTiN) and titanium aluminium silicon nitride (TiAlSiN) were deposited on cutting tool through PVD technique in Famex Coating Private Limited at Coimbatore. The photographical images of coated inserts are depicted in Fig. 1a–c.

Fig. 1
figure 1

a TiAlN b AlTiN c TiAlSiN

Full size image

The microhardness of all inserts were determined by using (Model: Wilson400 series) by applying 1 kg load for 10 s. For these tests, a diamond indenter was used. The tests were conducted in accordance with ASTM E- 384–17. The hardness was reported as an average of three measurements to minimize the irregularity effects of coating. The hardness of the work piece (AISI 420 martensitic stainless steel) was determined using the Rockwell Hardness Tester. The test was carried out in C-scale with a diamond indenter and a 150 kg load.

AISI 420 martensitic stainless steels were utilized for this research study. The steel rod was 36 mm in diameter. The long rod was cut into 9 pieces of length 100 mm using cutting machine and its ends were faced using manual lathe. The JOBBER XL numerical controlled turning centre was employed for the research experiments as shown in Fig. 2. Table 1 indicates the different parameters and their levels, which are used to carry out the experiments.

Fig. 2
figure 2

Turning of AISI 420 Stainless steel

Full size image
Table 1 Machining parameters
Full size table

With various combinations of input parameters, Taguchi optimization was employed to produce better order of machining sequence. This research experimentation used a L9 orthogonal array as shown in Table 2. All nine experiments were carried out with the different input setting. SR, MRR and WR were calculated as output responses during machining using an orthogonal array. For each trial, three trails were completed, and the average value was recorded. SR was found through surface tester with 1-inch travel length in X-axis with the 0.5 mm/s as average run speed. The mass variation of coated CNC tool inserts before and after machining was used to calculate WR. The mass variation of work piece before and after turning was used to determine MRR. Coating, CS, F and CD were the variables in this experiment. SR (Ra and Rz), WR and MRR were all optimized using an optimization approach that included GRA and fuzzy logic approach. Grey analysis and fuzzy logic values were matched, and the best optimized result was discovered.

Table 2 L9 orthogonal array
Full size table

Since the Taguchi design strategy is meant to optimize just one solution characteristics, the GRA optimizes a variety of outcomes. GRA was chosen based on literature survey to optimize multiple responses: SR (Ra and Rz), MRR and WR. The optimization process in GRA is divided into three phases.

When the experimental data have minimum-the-better characteristic, the measured data can be normalised by using Eq. (1).

$${\text{Xi}}\left( {\text{p}} \right) = \frac{{\max {\text{xi}}\left( {\text{p}} \right) - {\text{xi}}\left( {\text{p}} \right)}}{{\max {\text{xi}}\left( {\text{p}} \right) - \min {\text{xi}}\left( {\text{p}} \right)}}$$
(1)

where Xi(p) = normalized value.xi(p) = experimental output value.max xi(p) = maximum of pth response.min xi(p) = minimum of pth response.

When the experimental data have higher-the-better characteristic, the measured data can be normalised by using Eq. (2).

$${\text{Xi}}\left( {\text{p}} \right) = \frac{{{\text{yi}}\left( {\text{p}} \right) - \min {\text{yi}}\left( {\text{p}} \right)}}{{\max {\text{yi}}\left( {\text{p}} \right) - \min {\text{y}}\left( {\text{p}} \right)}}$$
(2)

Then, the grey relational coefficient can be calculated by using Eq. (3)

$$\xi i\left( p \right) = \frac{\Delta \min + \Psi \Delta \max }{{\Delta 0{\text{i}}\left( {\text{p}} \right) + \Psi \Delta \max }}$$
(3)

where ξi(p) = grey relation coefficient.Δ0i(p) = ‖X0(p)-Xi(p)‖Ψ = distinguishing coefficient (0.5).Δmin = minimum of Δ0i.Δmax = maximum of Δ0i.

Databases are used to extract membership functions, which are then used to generate fuzzy rules. The decision-making unit is used to infer from the developed laws. Based on the degree of matching, the fuzzification interface translates the input into linguistic words. The fuzzy effects are then converted into crisp production by the defuzzification tool. Then, the obtained grades were fuzzified using fuzzy approach.

Results and Discussion

TiAlN, AlTiN, TiAlSiN coating applied on the tungsten carbide CNC insert had increased hardness as depicted in Table 3. The hardness of TiAlN, AlTiN, TiAlSiN coated inserts was determined and better hardness (1916 HV) was apparent for AlTiN as compared to TiAlN and TiAlSiN coated inserts. WR, SR, and MRR are the machining data that were obtained from experimentation as listed in Table 4.

Table 3 Hardness values of different coated tools in HV
Full size table
Table 4 Machining data
Full size table

The machined samples are shown in Fig. 3. Analysis of variance has been performed for all the output responses individually, and the data were depicted in Tables 5, 6, 7 and 8.

Fig. 3
figure 3

Machined work pieces

Full size image
Table 5 Analysis of variance for Ra
Full size table
Table 6 Analysis of variance for Rz
Full size table
Table 7 Analysis of Variance for WR
Full size table
Table 8 Analysis of variance for MRR
Full size table

GRA has been selected in order to optimize numerous responses. The output responses were normalized according to a method based on the GRA technique and the absolute values were recorded. The grey ratio coefficients for the responses from the normalized values were calculated, and the values were described in Table 9. Weights were allocated as 0.5 for each grey coefficient. The grey relationship coefficients of the individual response were computed from GRG average. The results were ranked based on final GRG and the experiment number 1 (CS: 120 m/min, F: 0.1 mm/rev, CD: 0.5 mm and TiAlN coating) was ranked with highest grade.

Table 9 Grey analysis results
Full size table

The inputs to fuzzy logic system included CS, feed, CD, and coating. Input parameters can be described for instance in linguistic membership, which is low, medium and high. The average grey grades of separate outputs have also been determined as the input for fuzzy logic interface and the attainment of relational grey grade is represented by very small, very medium small, very small, very large, very medium high and very high attributes. Table 10 shows the interval of sorting. Depending on the technique, a combination of 9 fuzzy guidelines was allocated. By adjusting the fuzzy logic, maximum – minimum combination action provides results as shown in Fig. 4. Finally, the defuzzifier changes predicted outputs to GFRG, as illustrated in Fig. 5, by utilizing MATLAB (R2016b) fuzzy logic toolbox. Table 6 summarizes the outcomes of the GFRG.

Table 10 Fuzzy analysis results
Full size table
Fig. 4
figure 4

Fuzzy logic designer

Full size image
Fig. 5
figure 5

Fuzzy logic designer rule viewer

Full size image

As described in Table 10 and Fig. 6, settings for experiment number 1 were optimized: 120 m/min CS, 0.1 mm/rev F, 0.5 mm CD, coating: TiAlN). For all GRG and associated GFRG trials, the relative assessment is shown in Fig. 6. From Fig. 6, the highest values of GRA and GFRG were observed for experiment 1 based on L9 orthogonal array. Hence, the experiment 1 was the final optimized setting. The GFRG has higher grade than GRA. It was improved by 2.4% compared with GRA.

Fig. 6
figure 6

Comparison of GRA and GFRG

Full size image

After obtaining GFRG, the Taguchi method was used for optimizing the results of GFRG. GFRG was considered as output for 9 experiments. The optimized setting was found to be Coating: TiAlN, CS: 120 m/min, F: 0.10 mm/rev and CD: 1.5 mm for achieving higher GFRG as depicted in Fig. 7. To validate the results of this optimization technique, an experiment was performed with an optimized combination (A2B1C1D3) and the outputs were measured and presented as shown in Table 11.

Fig. 7
figure 7

Optimized setting for GFRG

Full size image
Table 11 Confirmation experiment
Full size table

After machining, the tool inserts were examined through microscope with magnification 100 X. The WR was observed as shown in below Fig. 8 a–i.

Fig. 8
figure 8

a–i Microscopic images of tool for experiments 1–9

Full size image

During machining, the chip formations were observed. The chip formation of various experiments is shown in below Fig. 9 a–i. In general, Golden colour chips denotes less WR, whereas blue burnt denotes more WR. The creation of the chip, which has a substantial effect on tool life, has been largely affected by CS, CD and F. It can be ascribed to changes in the material inertia among shear planes during machining. By increasing CS, the cutting force rises, and therefore the bending moment of the chip increases. This chip might damage the tool further by generating heat. The chip thickness has decreased for each feed with decreasing CS. Though, the average chip thickness similarly dropped with reduction of F. Hence, as the CS, the F considerably impacts chip thickness.

Fig. 9
figure 9

a–i Images of cutting chips for experiments 1–9

Full size image

Conclusions

TiAlN, AlTiN, TiAlSiN coated inserts were effectively deposited on the outer surface of cutting tools. Experimentation was planned based on L9 orthogonal array. CS, F, CD and coating material were taken as the operating parameters. All the machining tests were performed based on run order and the output responses such as Ra, Rz, WR and MRR were measured. These output responses have been optimized simultaneously through grey fuzzy optimization technique. Initially, the several responses were normalized and converted into single grade through GRA. Later, these grades were fuzzied using fuzzy logic tool box. From the results of GRA and GFRG, experiment number 1 was found to be the optimized setting (120 m/min CS, 0.1 mm/rev F, 0.5 mm CD, and coating: TiAlN). GFRG is then optimized through Taguchi experiment and it was found that Coating: TiAlN, CS: 120 m/min, F: 0.10 mm/rev and CD: 1.5 mm were the optimized setting for turning AISI 420 martensitic stainless steel.