Abrasive Waterjet Cutting of Clad Composite for Achieving Minimal Cut Quality Di ﬀ erence Between Constituent Layers

: Clad composites have emerged as a suitable choice to augment the industrial needs due to having a combination of di ﬀ erent properties. The accurate cutting is challenging due to the heterogeneous nature of the composite. Conventionally, thermal cutting (plasma / gas) techniques are commonly employed which provide poor cut quality, deeper heat a ﬀ ected zones and demand additional ﬁnishing operations. Therefore, this research evaluates the potential of abrasive water jet cutting (AWJC) as a proﬁcient substitute for the cutting of stainless-clad-steel composite in terms of surface quality. However, it is di ﬃ cult to produce a similar level of surface ﬁnish at both the layers because the constituent layers have di ﬀ erent mechanical properties. The e ﬀ ect and signiﬁcance of four important AWJC parameters on cut quality are examined through statistical analyses. Optical and scanning electron microscopic analyses are further provided as evidence of the reported results. Optimal settings are also developed using a weighted signal-to-noise ratio technique which can provide minimal roughness at each layer. Moreover, using the optimal settings, a similar level of surface ﬁnish has been achieved for both the layers with a di ﬀ erence of just 0.03 µ m between the constituent layers.


Introduction
Layered composites are being widely employed in manufacturing due to having a combination of different properties. The set of properties generated after the clad development are supposed to be not achievable by a single material of that clad. Exclusive features like higher strength to weight ratio, corrosion resistance, remarkable fatigue and endurance limit are offered by their use [1]. When a stainless steel is cladded to a low carbon steel the resultant properties include high strength and high corrosion resistance [2]. The growing use of clad composites has attracted the focus of researchers in these days. Noticeable work has been done to develop/improve and characterize the clad composites [3]. For example, Cui et al. [4] studied the tribological characterization of laser cladding coatings and Feng et al. [5] developed MoSi2/TiC/γ-Ni composite coating through plasma transferred arc welding and studied the microstructure and wear properties. Likewise, Yu et al. [6] developed NiAl intermetallic compound through laser cladding and investigated the tribological behavior of the clad. But the real challenge that limits the use of clad materials is their accurate cutting. The heterogeneous nature of the material produces variable cutting forces that seriously affect the using TOPSIS technique. A water pressure of 300 MPa, traverse rate of 120 mm/min, standoff distance of 1 mm and A MF of 360 g/min were the optimal settings.
The potential of the AWJC had been tested for a variety of materials, but its cutting capability for the machining of clad composites, specifically for stainless-clad steel, had not been thoroughly examined so far. The real challenge regarding the use of this cutting technique is to control the surface finish at both of the layers of composite. An adequate level of surface finish in both the layers is not the only requirement; it is also essential that both the layers hold a similar surface finish level, as this material must be used in composite form in its end application. This aspect, with regard to the AWJC of layered composite, has not been explicitly studied yet. Therefore, in the present research cutting performance is comprehensively investigated for the cutting of layered materials with an emphasis on minimizing the surface roughness of the individual layer. Moreover, the difference between the surface roughness values of both the layers is also minimized. The optimization of the process parameters is commonly recommended to eliminate the wastage of resources, especially materials and energy [23]. Signal to noise ratio is one of the commonly used approaches to optimize the process parameters [24]. Therefore, weighted signal-to-noise ratio method, a multi-objective optimization approach, is applied to develop an optimal parametric setting that can provide minimum surface roughness of each layer having a minimal difference from the neighboring layer.

Materials and Methods
This research is oriented in evaluating the cutting capability of AWJM for stainless-clad steel. This material has been used in a variety of different applications, such as boiler tubing, heat exchangers, etc. The selected workpiece is made up of two layers. One layer is of stainless steel (S SL ) while the other is of Mild steel (M SL ). Both of these layers are joined together by the process of weld overlay to produce the cladded specimen. The chemical composition of selected workpiece material has been validated via optical emission spectrometry using a standard method ASTM E 1676-4. The chemical composition of the selected material is provided in Table 1. Salient properties of stainless-clad steel like density, yield strength, percentage elongation etc. are extracted from literature and are tabulated in Table 2. The thickness of the specimen used for experimentation was 10 mm, out of which 4 mm is of S SL whereas the rest of the thickness is of mild steel. The workpiece is clamped into a clamping device and the flatness of the work surface is ensured with the aid of sprit level. The abrasive nozzle is set at an angle of 90 • to ensure the orthogonal cutting. A garnet abrasive of 80 mesh size was employed for the cutting of clad composite in a single pass. The schematic of cutting and actual setup used for experimentation is described in Figure 1.   The cutting performance of AWJC for machining clad composite is assessed in terms of surface roughness. During each experimental run, a 40 mm cut-length was machined on an abrasive water jet machine (model: WC3WB1212H). Upon the completion of each experiment, the surface roughness was measured for both of the layers of clad composite using surface texture meter (model: Surtronic S128). An evaluation length of 4 mm and a cut-off length of 0.8 mm was used for roughness measurement. For each layer, three readings of roughness were taken, and then the average value was reported. The measurement procedure adopted for taking the measurements of surface roughness at both the layers has been elaborated in Figure 2. The value of roughness has been taken for both the layers, i.e., stainless steel (R a(SS) ) and mild steel (R a(MS) ). Four key parameters were considered as the input variables namely; stand-off distance (S OD ), abrasive mass flow (A MF ), feed rate (F R ) and water pressure (W P ). The selection of these parameters was based on the rationale that their impact has been noted as significant on surface roughness in the reported literature [15][16][17][18]20]. Preliminary trials were also performed for selecting the appropriate levels of the selected control variables. Striation formation and jet deflection, the two problematic issues associated with the AWJC, have been given due consideration during preliminary trials. The selection of the levels is done in such a way that the chances of striation formation and jet deflection would be minimal. It has also been ensured that abrasive laden jet can produce a through cut in a single pass. All the parameters have three levels, except the stand-off distance that has two levels. It was noted during the preliminary experimentation that if the value of S OD was reduced below 1 mm, the retraction of the abrasive grains damages the abrasive nozzle.
On the other end, if the value of S OD is raised beyond 2.5 mm, the value of the effective width of the jet is enlarged before the interaction with the target surface. As a result, the kinetic energy of the abrasive laden jet has been reduced and it cannot produce a through cut. It is important to mention that the workpiece placement is also a critical consideration in this study as material comprises of two layers and both have different mechanical properties. The workpiece was placed in such a manner that S SL lies at the top, whereas M SL faces the bottom. The rationale for selecting this work orientation is to take advantage of abrasive particles' kinetic energy because S SL (55 HRA) is harder compared to M SL (48 HRA). In fact, at the start of the cut, the abrasive laden jet has more penetration power owing to its higher kinetic energy. The magnitude of this energy reduces as the jet moves downward because this energy is consistently consumed in cutting the material. Therefore, it is advantageous to place S SL at the top so that the harder layer can be machined due to the higher kinetic energy of the abrasive jet. Though the abrasive jet consumes its energy to cut the upper layer, the lower layer of the material is softer, hence it can also be machined. Parameters and their levels are tabulated in Table 3. The rest of the control variables were treated as constant factors. Experimentation has been planned according to the robust Taguchi experimental design technique. This experimental methodology provides an opportunity to have accurate experimental results with a fewer number of experiments. In a similar practice, several studies can be found using Taguchi L18 for having a low number of experimental runs to control the time and cost of experimentation [25][26][27]. The optimized setting proposed via Taguchi design makes the process so robust that it can minimize the impact of noise factors on the process output. In this experimental technique the number of experiments is governed by the number of control variables and their level values. In the present work there are four control variables, each having three levels, except stand-off distance which has two levels. Therefore, a mixed level design array is selected. According to the number of control variables and their levels taken in this work, the L18 orthogonal array was found suitable for performing the experimentation. The experimental results have been thoroughly analyzed using various statistical techniques. Optical and scanning electron microscopic (SEM) images of the machined surface have also been taken for an in-depth evaluation of the parametric effects. Finally, the optimal parametric combination has been developed using a weighted signal-to-noise ratio method that ensures minimal surface roughness of each layer with a minute difference. This technique consists of five steps which are described below: Calculation of S/N ratio for each response variable; 2.
Calculation of scaled signal-to-noise (SSN ratio) for each response; 3.
Assign weights to scaled signal-to-noise ratio of each response and find weighted signal-to-noise ratio of output variables; 4.
Find the level averages for the weighted S/N ratios; 5.
Identify the optimal parametric level combination that can simultaneously optimize all the response attributes.
As a first step, the value of S/N ratio has been calculated for the selected response using the relation described in Equation (1). S/N ratio was calculated for surface roughness of both layers (S SL & M SL ) using the same relationship.
where, η accounts for S/N ratio and Y ij is representing experimental value of jth response in ith trial. After that, the SSN ratio has been found using the Equation (2).
Here, SSN = scaled signal-to-noise ratio, η(max) = maximum value of η from entire data, η(min) = minimum value of η from entire data In the third step, weighted S/N ratio has been found for R a of both layers using the relation mentioned in Equation (3).
where, P = weightage given to each output response In this work, equal weights are assigned to all the objectives. There is an involvement of three responses in this study; surface roughness of stainless steel layer (R a(SS) ), roughness of mild steel layer (R a(MS) ) and the difference of roughness value between the two constituent layers (δ). The roughness difference between the two layers is designated as delta (δ). Keeping in view that three responses are involved in the present work and all are equally important, a weightage of 0.33 has been assigned to all the responses. In the last step, level averages have been calculated based on the WSN ratios of all the experimental runs, and finally the optimal combination has been established.

Results
Experimentation on layered composite has been successfully performed under DOE (L18). In each experimental run a cut length of 40 mm was produced. Afterwards, the surface roughness of all the samples was measured in terms of arithmetic average which is a commonly used parameter in the industry for assessing the surface finish of the machined part. As the machined sample is composed of two layers of material, the surface roughness of both the layers has been measured. The results of experimentation have been tabulated in Table 4. Experimental results obtained after experimentation and measurement are thoroughly analyzed using statistical and SEM analyses.

Parametric Significance Analysis
Analysis of variance has been carried out for examining the parametric significance with respect to the selected responses. A confidence interval of 95% has been defined as a criterion for rating a parametric effect as significant. According to this criterion, any control variable having a p-value less than 0.05 would be considered as significant for the selected output variable. It is evident from the results of ANOVA presented in Table 5, that S OD , A MF , and W P are the significant control factors for the surface roughness of the stainless-steel layer (S SL ). Whereas, the roughness of the mild steel layer (M SL ) is primarily influenced by A MF and F R as described in Table 6. The percentage contribution of control variables for both the surfaces has also been calculated as shown in Figure 3. It has been found that the percentage contribution of input parameters namely; S OD , A MF , F R and W P for controlling the surface roughness of S SL is 25%, 31%, 8%, and 24%, respectively. Whereas, in case of M SL surface roughness, A MF holds the leadingly high percentage contribution of 68.3% followed by F R which has a percentage contribution of 19%.  It is pertinent to mention that percentage contribution of control variables is different for both the layers despite the fact that the layered composite is subjected to cut via AWJM in a single pass. Moreover, the surface texture produced on both layers of material is different as highlighted in Figure 4. The heterogeneous nature of the composite is mainly responsible for this shift. It is worth-noting that A MF has proven to be the most contributing parameter for both the layers (S SL and M SL ). However, the percentage contribution of A MF for controlling surface roughness of M SL is approximately double compared to its percentage contribution towards S SL surface roughness. The surfaces finish of S SL is mainly dependent on S OD , A MF and W P , however the surface finish of M SL is predominantly controlled by A MF and F R . This shift of parametric significance happens because as the jet stream penetrates down the thickness of work surface, its kinetic energy has been reduced. This dropped magnitude of the kinetic energy has been utilized in mechanical shearing of the substrate material. Subsequently, jet dynamics observed at exit is different to that found at the jet entry. Hence, both the layers of bi-metal clad experience different magnitude of mechanical shear resulting into dissimilar surface qualities. Therefore, it is challenging to have a similar surface finish on both the layers. This requires finding an optimal parametric combination that not only provides minimum surface roughness of both the layers, but also minimizes the difference to offer a similar kind of cut quality.

Parametric Effects Analysis
After evaluating the parametric significance through ANOVA, the main effects plot analysis has been carried out to perceive the trend of control variables for the selected response characteristics. The results of the aforesaid analysis have been shown in Figures 5 and 6. It has been noticed that the control variables portray the similar trends for the surface roughness of S SL and M SL despite their difference in percentage contribution. The accompanying discussion on the parametric trends is described in the up-coming sections.

Effect of S OD on Surface Quality
As mentioned earlier, S OD has proven to be a significant control variable for the surface roughness of S SL and statistically insignificant for the M SL surface finish according to the ANOVA results. The same has also been witnessed from the parametric effects plots described in Figures 5 and 6. The reduction in the magnitude of kinetic energy is primarily responsible for this trend as abrasive laden stream has already consumed its energy while penetrating down the workpiece. That was also the reason for placing the S SL at the top to take the advantage of this kinetic energy considering that S SL is harder compared to M SL . It has been observed that S OD demonstrates a linear relationship with respect to surface roughness, i.e., an increase in the value of S OD results into the larger value of R a . A similar trend has been reported in [28]. This relationship holds true for the surface roughness of both the layers (S SL & M SL ) of clad composite. With the increase in S OD , the distance between the abrasive jet nozzle and the target workpiece is increased which helps to raise the momentum of abrasive grains. This increase in the momentum improves the penetration power of the abrasive grain. Thus, a deeper crater is formed at the target surface. Consequently, a poor surface finish is obtained as depicted in Figure 7.

Effect of A MF on Surface Quality
The effect of A MF on surface finish is the opposite to that of S OD , i.e., a larger value of the said control variable results into smaller surface roughness. The trend of A MF is observed alike for both the S SL and M SL layers. It is worth-noting that A MF has proven to be the major contributing parameter for the surface finish of both the layers. However, its contribution is more prominent for R a(MS) in contrast to R a(SS) as depicted in Figure 3. The same has also been observed in the parametric effect plots described in Figures 5 and 6. At the larger value of A MF (100 g/min), a small magnitude of surface roughness of both the layers has been obtained. The main reason behind the trend of A MF is the change in the mode of shearing action associated with the change in value of A MF . In AWJM, cutting action has been accomplished by the abrasive laden slurry. The abrasive grains impart an impact load on the target surface that causes shearing of the work surface. As the magnitude of A MF is raised from lower to higher value, a greater number of abrasive grains are available for performing the cutting action in the machining regime. In other words, more cutting sites are available in the cutting area to mechanically shear-off the workpiece material. This increase in the number of shear sites has promoted the uniform erosion of the material from the target surface. This uniformity in cutting action yields lesser surface irregularities. Hence, the surface finish of the machined surface has been improved. The same has also been witnessed by the authors of [8] during AWJC of Carbon Fiber Reinforced Plastic (CFRP) and aluminum alloy UNS A97050 stack. This phenomenon is also visible in scanning electron microscope (SEM) images shown in Figures 8 and 9. Quality of SSL and MSL machined surfaces at 40 and 100 g/min of A MF has been examined in these micrographs. It has been noted that surfaces of both the stainless steel as well as of the mild steel are found relatively smoother (smaller surface roughness) at larger A MF . At smaller A MF , shear marks found at the machined surface demonstrate that material has been removed in a non-uniform manner from both the layers as mentioned in Figures 8a and 9a. On the other end, SEM images recorded at larger A MF depicts that shearing occurs in a relatively more uniform manner. Another important observation that has been evidenced in the SEM images is the embedment of abrasive particles of the machined surface. But this phenomenon is only observed on the surface of M SL . This is mainly attributed to the lower hardness of the M SL . Interestingly, the surface texture of both the layers has been noted to be appreciably different from each other even though both layers are machined in a single pass. This difference is prominently seen in Figures 8 and 9.

Effect of F R on Surface Quality
Based on the parametric effects demonstrated in Figures 5 and 6, it has been envisaged that lower F R has a positive impact on the surface finish of the clad specimen. At a higher value of F R , surface roughness of both the layers has been reduced. The main reason behind this trend is the contact time between the abrasive laden with the target surface. When the magnitude of F R is higher, abrasive laden traverses at a faster rate along the cut dimension which in turn reduces the contact time between the abrasive laden waterjet and the target surface. This reduction in the contact time upsurges the cutting rate during AWJC. However, this increase in the cutting rate is associated with a prominent drawback of generating non-uniform shearing of material from the target surface. Thus, non-uniform occurrence of shearing provides poor surface quality. The same has also been witnessed by the SEM micrograph shown in Figure 10. A similar trend of F R with respect to surface roughness has also been cited in another work [8].

Effect of W P on Surface Quality
The trend of W P with respect to the surface finish of both the layers has been described in Figures 5  and 6. Experimental results revealed that the value of surface roughness is increased with the increase in W P from 250 MPa to 285 MPa. The reason behind this trend is the increase in the magnitude of kinetic energy of the abrasive jet with the rise of the value of W P . This increment in the kinetic energy of the jet accelerates the abrasive particles more energetically. When these accelerated abrasive particles encounter the target surface, they impart a greater amount of shear load during their collision due to having larger kinetic energy. Consequently, deeper crates are formed at the machined surface which promotes the surface roughness. The same has also been validated in the SEM micrographs of the machined area presented in Figure 11. At larger W P (285 MPa) deeper craters are formed, while at lower W P (250 MPa) shallow craters are produced. Owing to the decrease in the craters' depth the surface finish is improved as highlighted in Figure 11. However, an increased trend has been observed for the surface roughness of both the layers with the rise in W P from 250 MPa to 285 MPa. However, this impact is statistically significant for S SL only as mentioned in Tables 5 and 6. This is because the abrasive jet loses its kinetic energy as it moves down along the workpiece thickness. The kinetic energy of the abrasive laden jet is successively dissipated due to its collisions with the work surface. This is why the change in water pressure is unable to produce a noticeable change in the surface roughness of M SL as the jet's kinetic energy has already been dissipated by the upper layer of stainless steel (having higher hardness of 55 HRA). Therefore, the effect of change of W P turns out to be insignificant. It is interesting to note that the surface roughness of both the S SL and the M SL layers was reduced as the value of W P rose from 285 MPa to 320 MPa. This was attributed to the reason that at a higher W P the shear load on the abrasive grains is increased. Bouad et al. [29] claimed that if the shear load experienced by the abrasive grain is more than its crushing load, the abrasive particle gets fractured. The cutting capability of the crushed particles is reduced, and thus shallow craters are formed. Subsequently, the magnitude of the surface roughness is reduced.

Multi-Response Optimization
Optimal parametric combination has been identified in this section after comprehensively discussing the parametric effects and their relative significance for the surface roughness of each layer of clad composite. The development of the optimal setting is challenging because a smaller R a value at each layer is not the only objective, but it is also required that the difference between the constituent layers' roughness should also be minimal. Because this clad composite has to be employed as a single material in its end applications, therefore multi-response optimization (MRO) has been carried out in this research to achieve the said objective. A variety of multi-response optimization techniques are available like Vikor, Multi-response signal-to-noise (MSN), Weighted signal-to-noise (WSN) and Grey relational etc. however, WSN has proven to be most effective [30]. That is why in the present study WSN methodology is adopted. This methodology helps to identify those levels of control parameters that can provide the desired outcome of all the selected responses. The optimal combination iterated via said optimization method can be within the experimental design table or might be some other combination based on the selected parametric levels that is not a part of the design table. The steps associated with this optimization technique have already been discussed in the materials and methods section. As there are three responses which are under consideration in this research and it is desired to have their magnitude at the minimum level, equal weights have been assigned to each response characteristic. The results of weighted signal-to-noise ratio analysis are tabulated in Tables 7 and 8.    A confirmatory experiment has been performed after finding the optimal parametric combination through WSN to authenticate the developed optimal settings. The results of the experiment are tabulated in Table 9. It has been revealed that the optimal parametric combination provides a reduction of 34% in S SL surface roughness and 11% in M SL surface roughness. Moreover, the difference between the roughness values of the constituent layers is also observed to be quite small of just 0.03 µm magnitude. These results have also been validated from the roughness profiles recorded for the samples machined at the optimal and non-optimal parametric settings, as shown in Figure 12. At the non-optimal parametric settings of control variables larger surface irregularities are prominent at both the surfaces. On the other end, surface irregularities at both the layers and their respective difference is found reasonably small as highlighted in Figure 12.  It is pertinent to mention that the present study would be quite useful for the industries dealing with the cutting of cladded materials. Issues such as poor cut quality and deeper heat affected zones pertaining to the conventional thermal cutting processes can be eliminated through the use of AWJC. Furthermore, the requirement of subsequent finishing process has also been eradicated. The proposed optimal combination developed herein would be guaranteed to provide minimal surface roughness at each layer with a minute difference of just 0.03 µm in surface roughness. This work will hopefully open new horizons of the application and cutting of cladded materials.

Conclusions
Clad composites have emerged as a suitable choice to augment the industrial needs, but its accurate cutting is challenging because of heterogeneous properties of the constituent layers. Abrasive water jet cutting (AWJC) is a proficient alternate but it is difficult to produce a similar level of surface finish at both the layers of material in addition to minimizing the roughness of individual layer. Therefore, the present study aims to investigate the cutting performance of AWJC for cutting of clad composite with a dedicated focus on analyzing the surface quality of the machined surfaces. Experimentation has been performed according to L18 orthogonal array. The results are thoroughly examined using statistical and SEM analyses. The following conclusion may possibly be extracted based on the experimental findings and their discussion: 1.
Clad composite can be successfully machined using AWJC with adequate surface quality. However, the parametric significance for roughness of both the constituent layers is found to be different.

2.
Roughness of stainless-steel layer (S SL ) is primarily influenced by A MF , S OD and W P . Whereas in case of mild steel layer (M SL ) surface roughness, only two control variables i.e., A MF and F R are found significant.
A MF has been observed to be the most contributing variable for both the S SL and M SL . However, its contribution for M SL roughness (68.3%) is more than two-fold its contribution for roughness of S SL (31%). Moreover, a larger magnitude of A MF yields better surface finish at both the layers because of the occurrence of uniform shearing action. SEM analysis depicts that few abrasive particles have been imbedded on the surface of M SL due to the lower hardness of M SL . This particle embedment is not observed at the surface of S SL . Smaller S OD provides better surface finish, but this effect is more prominent for stainless steel layer (S SL ) as abrasive laden jet has consumed its energy while moving down the workpiece thickness. The effect of F R is found to have a direct relationship with the surface roughness of clad composite in AWJC. Higher value of F R results in inferior surface quality because inappropriate shearing action takes place due to shorter contact time between abrasive laden slurry with the target material. However, this effect is only significant for the lower layer (M SL ) of clad composite. Increase in the magnitude of W P from 250 MPa to 285 MPa increases the roughness of both the S SL and M SL because at higher W P abrasive particles have larger kinetic energy and produce deeper craters upon striking the work surface. However, a further increase in W P results in a reduction of surface roughness.

3.
Optimal settings of control variables are developed using WSN methodology which ensures that not only roughness of both the constituent layers is smaller, but the roughness difference is just of 0.03 µm. Furthermore, an improvement of 34% in the stainless steel layer's surface roughness and 11% in mild steel surface roughness has also been realized using the developed optimal parametric combination.