*6.3. Tool/Nozzle Design*

When designing an experiment within the domain of machining science, one of the most important variables to optimise is the cutting tool. Often this takes the form of optimisation across parameters such as cutting insert material, tool geometry, tool holder design or nozzle configuration. Practically however, some of these variables are more easily established during experimentation. For example, if an experiment is not being run with the express intent of establishing novel tool materials (for use in a given context), it is likely that the chemical composition of the tool material is decided upon by external actors. To give a specific case, it is often taken as a given that the current best practice for the machining of challenging materials such as titanium and nickel alloys are cemented carbide tools. Of course, there remains a significant scope to optimise the specifics of the tool chemistry (e.g., WC-(Ti,Ta,Nb)C-(Co,Ni)); however, the nuance of tool composition is often removed from the experimental process by instead following the recommendations of a tool supplier (i.e., Sandvik, Seco or Kennametal), who generally offer tool compositions that have been optimised via extensive R&D to meet a range of similar machining challenges. With this in mind, and given both the relative complexity and financial (and informative) barriers to entry associated with tool composition optimisation, this section considers composition only in passing, instead focusing upon variables such as nozzle, tool and tool holder geometry.

Given this focus, one such article that considers the problem of tool design was recently published by Shokrani and Newman [46]. As part of their research, the authors focused upon the relationship between cutting insert geometry and tool life in the context of the cryogenic end milling of Ti-6Al-4V. Specifically, Shokrani and Newman mapped the relationship between both the rake and primary clearance angle on tool life, tool wear and surface roughness. The authors utilised 12 mm diameter solid WC end mills coated with TiSN-TiN to a thickness of approximately 3 μm. They employed a constant LN2 nozzle position and flow rate (20 kg/h) throughout the trials. Within this experimental design, Shokrani and Newman noted the positive implications of both primary clearance angle and rake angle (Figure 7) on tool life. Moreover, they observed that, although the impact of increased clearance angle persisted across each rake angle, the effect was most pronounced at the highest rake angle, and likewise, the correlation between rake angle and tool life was most exaggerated at the higher clearance angle. Although these observations are of clear value, the authors do however note that (despite not being manifested in their trial) excessively increasing either rake, or clearance angle, can lead to weakening of the cutting edge and ultimately premature tool failure.

In addition to the tool life implications of manipulating clearance and rake angle, the relationship between surface roughness and insert geometry was also considered. In order to realise these goals Shokrani and Newman took readings at the start of experimentation in order to remove the impact of tool wear on average surface roughness (Ra). In doing so, the authors again noted the positive implications of both increased primary clearance and rake angles such that a tool with a sharper cutting edge corresponded to an equivalent or reduced surface roughness across the entirety of the data set (Figure 8a). Whilst it is difficult to qualitatively assess the relative effect of rake and primary clearance angle on this phenomenon, it is true that the lowest surface roughness was generated at a rake angle of 14◦ regardless of the primary clearance angle. Given this observation, it is also apparent that the impact of primary clearance angle is most significant at a rake angle of 12◦. Whilst these observations make a compelling case for the use of steeper rake and clearance angles, it is important to note that the true impact of cutting edge geometry (on the surface mechanics of the machined workpiece) is not fully captured via Ra measurement. Rather, it is often true that two surfaces with an equivalent centre line average will exhibit varying degrees of asperity, and thus vastly different tribological performance [47] (Figure 8b).

**Figure 8.** (**a**) A graph to show Ra and tool geometry during the cryogenic end milling of Ti-6Al-4V. Reprinted from reference [46]. (**b**) Schematic illustration of machined surfaces with equivalent values of Ra.

Furthermore, although the work of Shokrani and Newman provides an extremely useful case study to illustrate the importance of tool geometry in cryogenic machining, the research is not without limitation. Foremost, the authors chose to prematurely terminate machining trials (by failing to test rake angles in excess of 14◦ and clearance angles above 10◦) at a point such that the negative implications of excessive clearance and rake angle were not realised. In this sense, their research may wrongly create the inference that a sharper cutting edge invariably improves performance, or equally, it may encourage future research to more aggressively select an insert than can be reasonably supported by their data set. Equally, although the variance in the tool life data was reasonable, the magnitude of the error bars in the surface roughness plot is somewhat excessive relative to the magnitude of the recorded data. For this reason, conclusions as to the most appropriate tool geometries (to optimise topology) should be made with a degree of reservation.

In addition to the efforts made in optimising tool geometry, Pereira and colleagues [48] also undertook research into the nozzle design for cryogenic machining. In contrast to the work of Shokrani and Newman, Pereira et al. instead focused upon the development and optimisation of CFD (ANSYS Fluent) simulation-derived nozzle outlets for CO2 + MQL cooling, focusing primarily upon nozzle diameter. Thereafter, the authors validated their models with experimental trials, and ultimately developed a prototype nozzle in accordance with the data generated in the paper (taking as a given that a CO2 velocity of 325 m/s is necessitated to aid in the cutting process). Given these research parameters, the authors inputted three nozzle diameters into their CFD model of 0.5, 1 and 1.5 mm. Pereira and colleagues thereafter qualified the relative performance of each nozzle geometry according to normal average velocity at a distance of 20 mm from the outlet, observing that (in the simulation) the CO2 velocity was best retained with the 1.5 mm nozzle. In accordance with this observation, when experimentally verified, the 1.5 mm nozzle obtained the greatest spray distance of 40 mm (compared to 18 mm and 10 mm for the 1.0 mm and 0.5 mm nozzles, respectively). With this in mind, Pereira et al. went on to reference the research of Park and colleagues [49], who observed that the maximum area fraction coverage of MQL droplets occurs at 30 mm from the nozzle tip. Given this figure, the authors determined that the most viable nozzle diameter of the three trialled was 1.5 mm owing to the capacity to most closely resemble the optimal distance for MQL coverage and general velocity profile.

Having determined that a nozzle diameter of 1.5 mm was most appropriate in a cryogenic cooling context, the authors went on to manufacture two nozzle adaptors of equal diameter, proximity and general architecture with differing outlets, one of which featured a convergent CO2 outlet whilst the other instead employed a converging diverging nozzle. The convergent nozzle was chosen with simple volumetric continuity in mind, whilst the convergent divergent nozzle was employed with the outlook of exploring the impact, or lack thereof, of compressibility on the fluid dynamics at tool exit. Given the two designs, the authors first inputted the nozzle exit geometries into their CFD simulations followed by experimentally verifying the most suitable nozzle. As part of their simulations, they observed that the use of a convergent divergent nozzle was able to create a much greater exit velocity (475 m/s as opposed to 400 m/s) at the expense of a larger spray spread (15.6 mm as opposed to 9 mm). Consequently, they determined that the impact of increasing MQL concentration by 73% in the cut was of greater importance than increasing exit velocity by 18% (particularly given the threshold velocity of 325 m/s was met in each case); as such, they opted to utilise the convergent nozzle for experimental validation.

In order to verify the performance of the nozzle in a practical context, it was then employed during cryogenic-assisted end milling of a billet of Inconel 718, a material, which previously (Sections 5.3 and 6.1) has been shown to exhibit poor cryogenic machinability. The authors elected to use TiN-coated carbide inserts, a cutting speed of 120 m/min and a feed rate of 0.12 mm/tooth, ultimately indexing tool life performance against dry machining, MQL, CO2 in isolation and emulsion cooling (Figure 9). As part of their research, the authors observed that, of the dry and near dry coolant strategies, CO2 + MQL (as delivered through their bespoke nozzle) led to the longest tool life, such that it yielded an approximate 100% increase relative to dry machining. Of the other near dry strategies, CO2 cooling in isolation performed markedly poorer than MQL in isolation. This observation can be taken as inference of the relative heightened importance of lubricity, relative to cooling in the milling of Inconel 718. Despite the relative success of the nozzle in its ability to compete with other dry, or near dry, strategies, it is noteworthy that CO2 + MQL failed to generate comparative tool life to that of the conventional emulsion-cooled strategy.

**Figure 9.** Adapted bar chart to show relative tool life during the cryogenic end milling of Inconel 718, adapted from [48].

Although Pereira et al. were able to successfully optimise a nozzle for the delivery of CO2 and MQL, it is worthwhile to remark that emulsion, as has been shown elsewhere in the literature, remains the most suitable option for the machining of Inconel 718. Whilst this observation is clearly of great value in that it laments the conclusions made earlier in this review, it does further constrain the scope of cryogenic coolants to appropriate materials. Moreover, it is equally important to note the limitations in the experimental procedure employed by Pereira and colleagues. For one, this work is similarly limited to the work of Shokrani and Newman, in the sense that the nozzle exit diameter was shown to exhibit a wholly positive correlation with output velocity and, ultimately, performance. For this reason, it would almost certainly be valuable to explore the extent to which this phenomenon persists to assure that the upper limit of nozzle diameter when increased performance is applied. In addition to this limitation, it remains unclear as to whether the parameters with which the nozzle was optimised, namely exit velocity and droplet dispersion, are the most suitable metrics of nozzle performance. In this sense, it would undoubtedly be valuable to assess each of the nozzles in a practical context, although this of course would be cost restrictive.

Another paper that considers nozzle design and optimization for cryogenic machining was published by Gross et al. [50], who employed a similar nozzle optimization focus to the earlier work of Pereira and colleagues. In their research Gross and colleagues undertook preliminary CO2 + MQL nozzle optimization testing followed by cryogenic CNC milling trials on Ti-6Al-4V. In their preliminary testing, the authors first measured the temperature in the CO2 jet stream over a range of distances with four nozzle diameters: 0.5, 0.3 and 0.2 mm. In order to realise this goal, the authors utilised a type K thermocouple, which was vertically fixtured to coincide with the mid-stream of the CO2 jet. As part of their research, the authors trialled both smooth jet nozzles (SJN) and plastic tube nozzles (PT), wherein a CO2 pressure of 56 bar was generally employed, other than in one ancillary trial wherein a higher pressure of 71 bar was chosen. In addition to the temperature profiles developed as part of this research, Gross and colleagues also later went on to test the relative spread of their MQL by placing blotting paper in front of the MQL jet and qualitatively analysing the profile made by the lubricant. The experimental set-up employed in their research is outlined in Figure 10.

Having undertaken the aforementioned tests, the authors made a series of interesting observations on the ramifications of varying nozzle diameter, distance and CO2 pressure (Figure 11). Foremost, Gross and colleagues note that the enlargement of nozzle diameter from 0.2 mm to 0.3 mm (and thus, increase in CO2 mass flow rate) does not reduce the minimum recorded temperature; however, it does serve to increase the size of the low temperature area at the nozzles outlet. They also note that this observation equally applies to the use of a higher-pressure CO2 supply, wherein, again, no significant change is made to the minimum recorded temperature, despite an increased low-temperature area. The temperature data also suggest that the use of each of the SJN strategies leads to a more gradual reduction in temperature from the nozzle's tip onwards than was experienced with the PT nozzle. It is, however, difficult to ascertain with any reliability the extent to which this is a consequence of nozzle construction, as the plastic tubing nozzle featured a 0.5 mm diameter in comparison to the 0.2 and 0.3 mm smooth jet nozzles employed in the tests. Additionally, in the lubricant blot paper tests, the authors observed a significant number of lubricant splashes at the oil trace when MQL was applied with compressed air. By contrast, when the MQL was used in conjunction with CO2, no such splashing existed. The authors note this as an expected consequence of the increased flow speed of the CO2 jet (relative to the compressed air) focusing the oil droplets.

**Figure 10.** Experimental set-up employed during temperature measurement, MQL dispersion and milling trials. Reprinted with permission from Daniel Gross (2019), Copyright 2019 MM Science Journal [50].

Whilst many of the prior observations are valuable for the optimization of cryogenic machining research, one of the most important observations made during the trials of Gross et al. was that temperature reduces exponentially (rather than linearly) as the nozzle is moved further away from the tool. As the relative success of an MWF strategy is highly contingent upon the ability of the coolant to dispel the heat generated in the machining process (and thus cool the tool), this observation is incredibly important to the relative success of cryogenic MWFs. These data thus clearly infer the likely importance of the nozzle's proximity to the tool and thus provide motivation to revisit and optimise trials in which CO2 cooling has in the past proven to be ineffective. In many cases, it may be true that where the poor performance of CO2 coolants was previously assumed to be a consequence of inappropriate material selection, poor selection of feeds and speeds or inadequate tooling, the adverse results rather may have been a consequence of the nozzle's lack of proximity to the tool.

**Figure 11.** Temperature distribution for four nozzle/pressure configurations. Reprinted & adapted with permission from Daniel Gross (2019), Copyright 2019 MM Science Journal [50].

#### **7. Future Work and Conclusions**

It is clear that CO2 MWFs are one of the best candidates to replace conventional MWFs in many machining scenarios. Despite this promise, this review has clearly outlined the limitations of the technology, wherein the performance of a given MWF strategy, i.e., its ability to drive improved machinability outcomes, is highly contingent upon several factors, ranging from the material species being machined (Sections 5 and 6.1) to the nozzle set-up employed (Section 6.3). Further to this point, it is not immediately clear as to the interplay between these variables, and the dependencies that very likely would present as a greater volume of data is accumulated. This complexity creates uncertainty around the technology, which impedes uptake and creates hesitancy. It is thus the task of researchers in the field to alleviate the unknown, and thereby reduce the risk associated with early adoption. It is the opinion of the author that there are two primary forms of data that must be gathered to facilitate this:

### *7.1. Machinability Data*

The pure gathering of machinability data can be thought of as a full factorial approach to acquisition. In order to have true confidence in adopting a CO2 coolant strategy for a given context, prior experimental data supporting its use, in said context, must be available. By focusing on conducting a multitude of machining trials across a broad parameter set, we are able to produce data with a high degree of specificity; if the process outlined in a trial is replicated in industry with equivalent experimental control, the industrial adopter can reasonably expect similar performance outcomes. Unfortunately, however, whilst this approach informs us as to whether a specific process will or will not be successful, it does little to increase understanding regarding the physics of the process.
