Experimental Analysis of Circular Milling for Material Identification in Aerospace Industry

In the airplane design, thousand holes should be manufactured by machining leading this process in the first place for process optimization. Recently, the stack materials used in this application are sheet layers of Titanium alloy, Aluminum alloy and CFRP, which normally are machined under different cutting conditions. There are two main solutions when drilling stacks: using the more conservative ones or changing the parameters for each layer. In orbital drilling the forces are lower, with better quality but with longer machining time, compared to drilling, allowing the measurement and control of the process. Smart machining techniques can be applied for recognising and adapting the cutting parameters in real time, depending on a database for decision-making. In this paper, a technique is proposed to identify the cutting material based on the cutting and feed components of the machining force in a circular milling process. Experiments consist of machining separately Titanium and Aluminium alloy workpieces in a similar range of cutting speed and feed, measuring forces and the position of the cutting tool (X,Y, SP) using the machine-tool data. The results show that it is possible to identify the materials using the calculated tangential and radial force components in the tool referential frame because the values of specific cutting and feed force are significantly different for each material. An specific force map is used as a signature to distinguish the materials in real time machining which can be used for orbital drilling in the next step of the research.


Introduction
Multi stack drilling is one of the major operations in aircraft assemblies and it demands lots of time and resources to achieve high-quality holes. Usually, Titanium and Aluminium stacked materials are drilled before aircraft assembly and this poses a major challenge in machining due to the different machinability of materials. Some of the drilling challenges include poor finishing, burr formation and rapid tool wear [1]. However, hole quality aspects like exit burrs can have a significant influence during assembly and also affect productivity [2].
As an alternative for axial drilling operation, orbital drilling, an helical milling process, can be applied. This process shows advantages regarding the process flexibility and ability to produce highquality holes on different materials [3,4]. Sun et al [5] data shows that the fatigue life of holes produced by helical milling showed better life compared to conventional drilling. The burr height at the hole exits was considerably lower in orbital drilling compared to conventional drilling [6]. However, helical milling alone will not completely eliminate the problems faced in stacked hole making operations.
The environmental protection initiatives, such as ISO 14001 and lean manufacturing ideology, push researchers to develop flexible and smart processes to improve efficiency and productivity to reduce costs [7]. It is a fact that thousands of holes have to be drilled for assembling an airplane, so smart drilling is a key point for performance maximization. Hence, increasing the feed rate or cutting speed on the layer that presents higher machinability is directly related to this goal [8].
Smart machining refers to real time adaptation of cutting parameters for process optimization. It involves identifying the material accurately and then adapting the proper cutting parameters in real time for process optimization. In this area, Pardo et al [9] discusses decision making strategies in order to identify the tool position in stack drilling for adaptation of cutting parameters. A force sensorless method is proposed by Fang et al [10] in stacked orbital drilling of CFRP and Titanium in order to determine the tool position across different layers. Also, Wenkler et al [11] developed a way of predicting specific cutting force by Artificial Neutral Network which can be implemented in process planning or smart manufacturing.
Mechanistic force models can provide quantitative cutting force predictions based on the uncut chip thickness principle. Once experimental tests are made, the cutting force models enable estimating the specific cutting coefficients for a different set of variables. In this area, numerous papers predict cutting forces by considering different factors to improve the model accuracy. Armarego et al [12], for example, developed a force model considering tool run out to accurately predict cutting forces. Montgomery et al [13] included tool workpiece vibrations into the model to calculate cutting forces and specific force coefficients based on the chip load. Geometry of chip formation and prediction of forces in circular milling is discussed by Banerjee et al [14]. Wu et al [15] discussed force models for circular milling taking into account the uncut chip thickness and feed rates variation at the corners. The cutting coefficients calculation in circular milling models can be further extended for helical milling adding vertical feed.
The specific cutting coefficients data can be used to identify the material being machined. They can act as material signatures to adapt proper cutting parameters in real time. These data could directly improve the process productivity and hole quality if appropriate cutting parameters are applied for the specific material. There are many articles proposing methodologies to calculate cutting forces and cutting coefficients, but rarely on a data map of specific force coefficients of different materials. Similar data maps were developed recently by our team for axial drilling [16].
This article proposes a technique for material identification in circular milling process of Aluminium and Titanium alloys used in aerospace industry. Our work also introduces integration of external sensor to identify cutting flute position in real time which can be helpful in smart machining applications.
The materials used for the experiments were Aluminum 2017A and Titanium Ti6Al4V, being subjected to a common set of cutting conditions. In the following sections, the force model and cutting coefficients are presented, along with material and methods, followed by experimental results and identification maps for both the materials.

Force Model and Specific Force Coefficients
The forces for a general orthogonal cutting can be described by a mechanistic force model, as described by [17], considering only cutting action of the cutting edge neglecting ploughing and chisel edge effects.
In this way, the machining force F m is calculated considering the small finite elements of the cutting flute in the cutting edge referential frame decomposed: tangential (dF c ), radial (dF r ) and axial (dF z ) components.
The local force at point P of the cutting edge is a function of the uncut chip thickness h and the specific force coefficients: K c , K r and K z , as presented in Eq. 1.
where h(P )db is the elementary uncut chip load calculated on the reference point P in db elemental length of the cutting edge.
For each rotating tool machining process, there are some assumptions in order to identify a global value for specific force coefficients: milling and circular milling, as follows in the next subsections.
Milling In milling, in order to determine the cutting coefficients of the tool-workpiece pair, the local cutting force dF c (t) and the radial force dF r (t) are analyzed considering the variation of the uncut chip thickness (h) along the tool rotation.
The angle of local position φ of the flute i can be expressed considering the tool helix angle λ, the angle between flutes φ p = 2π N , where N is the number of flutes and a p the axial depth of cut.
Martellotti equation [18] described the chip thickness h(φ) as a function of feed per tooth (f z ) and φ, calculated considering the tool rotation angle (θ 2 ), and given by the equation: When summing the force contributions of each elemental cutting part db, dF c (t) and dF r (t), the directions are not the same and a change to a fixed referential is necessary. From there, dF x (t) and dF y (t) can be calculated.The forces in axial direction (dF z ) can be neglected, if the helix angle is small.
As in any machining process, some assumptions should be made for the identification of specific cutting force and specific radial force. It is considered that the maximum force is achieved in the position of θ 2 when the average uncut chip thickness along the cutting edge is higher. The reference frame is now fixed in the rotating tool θ 2 and not in each element. Then, the approximation done lead to: where A c is chip cross section area , h max is maximum uncut chip thickness and a p is axial depth of cut.

Circular Milling
Unlike in linear milling, circular milling involves movement of tool around a centered fixed point in clockwise or anticlockwise direction to enlarge a existing hole diameter. Figure 1(b) shows the circular milling tool trajectory (at θ 1 = 0) and the associated geometric parameters. A tool of diameter D t with center C is moving around the hole center O. D 1 refers to pre hole diameter and D is the final diameter to be achieved after circular milling. R tt is the tool trajectory radius with reference to hole center O [19].
In circular milling, cutting forces at the cutting edge have different referential frames that have to be analyzed in order to determine the cutting coefficients of the tool-workpiece pair. The cutting forces exerted by the tool on the workpiece can be seen in Figure 1(a). Feed motion produces F tan and F rad forces and the cutting action gives rise to tangential (F t ), radial (F r ) and axial forces (F z ) at the cutting edge. It can be noted that the direction of feed forces in circular milling changes continuously unlike in linear milling where feed direction and feed force direction remain unchanged. In circular milling context, forces in axial direction (F z ) can be neglected. Figure 2 shows chip formation in circular milling indicating entry and exit of cutting flute with the workpiece. θ 2 is measured from fixed referential frame of rotating tool with flute 1 of the end mill. The angle ψ is the swept angle of the flute which is determined by the entry and exit of the cutting flute as shown in the Fig. 2. The uncut chip thickness varies from h min to h max as the flute rotates

1652
Achievements and Trends in Material Forming The local position of the flute in circular milling can be expressed relating to tool trajectory position as indicated below: Equation 6 can be rewritten considering instantaneous position of the flute at angle φ calculated, in this case, as also a function of θ 1 : The forces measured during experiments by the dynamometer are in XY referential frame: F X and F Y as shown in Figure 1(a). The tool rotates around the hole center O in clockwise direction given by angular position θ 1 and around its own axis given by θ 2 , also in clockwise direction. The forces F X and F Y can be transformed using a rotation matrix (Eq. 11) to deduce F tan and F rad at the tool center. F tan and F rad are significant to understand tool deflection effects on hole quality. To calculate F t and F r at the cutting edge, the rotation matrix as a function of θ 2 and the dynamometer referential frame components (F X and F Y ) presented in Eq. 12 can be applied. The specific cutting coefficients K c and K r (N/mm 2 ) are identified by analyzing F t and F r maximum values for a particular set of tool revolutions using Eq.7 and Eq.8.

Materials and Methods
In this section, circular milling experiments in Aluminium (2017A) alloy and Titanium (Ti6Al4V) alloy workpieces measuring cutting forces are presented along with external sensor integration. The experiments were carried out for a specific range of cutting conditions, following the experimental setup and the design of experiments described below.
Experimental setup Circular milling experiments were carried out on a CNC milling center DMU85-DMG mono block machine. The tool used for the experiment is carbide end mill of 8 mm diameter from Fraisa [20]. A prehole of 11 mm diameter is already drilled before by a drill tool. The workpieces were fixed on a 9257B Kistler dynamometer, as shown in figure 3(a), using only its internal measuring region, connected to a 5070 Kistler amplifier. Figure 3(b) shows Titanium alloy sample with milled dimensions. Analogical data of force is converted to digital using a 9201 National Instruments acquisition module with 10 kHz acquisition rate. The measured forces were filtered using a low pass Butterworth filter with cutoff frequency of 700 Hz (appropriated to the maximum spindle speed). In order to detect cutting flute position, an external photoelectric sensor (reference: WL150-P132 from Sick) is connected to a DC power supply and the sensor output is connected to data acquisition card NI 9215 from National Instruments via BNC cable. Sensor data is read and written on Labview platform connected to NI DAQ through USB. The CNC milling center in use has Sinumerik controller from Siemens which is connected to Sinucom NC Trace software via ethernet connection to trace machine related data. The working principle of external sensor along with tracing of machine data is explained in the next subsection.

Cutting edge position identification
The external photoelectric sensor is utilized to detect flute position by the principle of light emission and detection on a reflective tape attached to the tool holder as shown in figure 4(b). The sensor produces a voltage high when it detects reflective tape positioned in line with the cutting flute. Also, the CNC machine has a spindle encoder reference (defined by the machine builder) which is used to identify the spindle position in rotary motion. This is illustrated in figure 4(a). It should be noted that

1654
Achievements and Trends in Material Forming The data from the external sensor and machine trace is analysed to find the angular relation (δ) between the cutting flute position and machine encoder reference point in order to deduce cutter rotation angle θ 2 as shown in the figure 4(a). Figure 4(c) shows the schematic of instrumentation setup in order to acquire the real time data from the CNC machine as well as external photoelectric sensor. In order to precisely deduce the angular relation between cutting flute position and machine encoder point, it is significant to read both of acquired data (Machine data and external sensor data) on the same platform in the same time frame. PLC from the CNC machine can be integrated to Labview via OPC UA 1 [21] which can facilitate to access machine data on Labview along with external sensor data. Our work is ongoing in order to integrate data acquired from different platforms in a common window like OPC UA for better decision making.
This article used a method to deduce δ ( Fig. 4(a)) by identifying the common time stamps from two systems and then comparing the signals obtained in the same time frame. Clock from external sensor system and machine trace system is set identical. The set of data points having common time stamps from both the systems is then super imposed as shown in the Figure 5 to identify angular difference between cutting flute and spindle encoder reference position. The sensor signal peak in the graph denotes a cutting flute detection and the corresponding value of the spindle position is found to be consistently between 153 − 158 o . Since, the above technique involves integrating two separate platforms with different time frames, the value of δ identified varies in a range but this work proposes novel ways to identify cutting edge position in real time for application in smart machining techniques. Table 1 summarizes the cutting conditions on Aluminium and Titanium along with tool details for circular milling operations. The machining responses were compared for a constant cutting velocity and a range of feed rates as indicated in the Table 1. Feed as low as 0.02 mm/tooth which is the minimum feed value for any monoblock carbide tool to perform is included in the cutting parameters range. Each test is repeated twice and care is taken that, same feed rate test is not carried out successively twice but in a random order, to eliminate chances of rapid tool wear (if any) at higher feed rate values. Circular milling was carried out on the previously drilled hole with an axial depth of 1 mm and a radial width of cut of 1.9 mm. This leads to a final milled hole diameter of 14.8 mm in order to have a cutting continuity coefficient c less than 1 [22].

Experimental Results
The following subsections present the results of experimental forces for circular milling. Circular milling forces are presented in terms of tool angular position θ 1 (around O) which is taken as a reference for all the feed rates and cutting speed values tested. It should be noted that the trajectory rate (θ 1 /time) changes based on variation in feed rate values.

Forces in circular milling
The forces components F X and F Y obtained with the dynamometer during circular milling can be seen in the Figure 6. The graph shows the variation of force components (XY reference on the dynamometer) along time, represented by the tool trajectory angle θ 1 . shows tangential (F t ) and radial (F r ) forces components (aligned to the cutting edge) and its variation along θ 1 evolution. A window of two tool revolutions is presented (one revolution : . Eight peaks can be observed in the cutting force components for a 4 flute end mill for

Achievements and Trends in Material Forming
Aluminium and Titanium alloys experiments. The cutting forces from dynamometer were transformed as previously indicated in Eq. 12. Each peak corresponds to cutting flute (F l i ) and swept angle (ψ i ) as shown in the figure 7(a).

Identification map
After acquiring the experimental forces for circular milling, the specific cutting force coefficients were estimated. They are presented in the form of identification map in this section. It is important to claim that the circular milling experiments were done on pre drilled hole. As one of the the goal of this paper is the material identification based on the data, the range of feed per tooth was narrowed to 0.10 and 0.18 mm/th, where the values of forces were closer. Also, low feed rate values as 0.02 mm/tooth are very rarely used in the industry for these materials.
K c and K r values are estimated by considering maximum values of F t and F r values respectively, for a specific set of tool revolutions. Figure 8 shows the identification map of specific cutting coefficients estimated for Aluminium and Titanium alloys. The data includes values calculated at several tool angle positions θ 1 at different feed rates and constant cutting velocity with the objective of having a representative value with larger spectrum of experimental variability and noise. A clear distinction between both materials and force coefficient values can be identified in different regions of the map presented in Fig.8. Thus, the clear distinction between Titanium and Aluminium regions presented in identification map, can be further applied for material identification in smart machining processes.

Conclusions
The article focuses on kinematics of circular milling and introduces a model to identify cutting coefficients for developing an identification map in smart machining applications. This work also proposes an idea of detecting cutting flute position in real time for instantaneous calculation of cutting forces and specific coefficients. Research is ongoing in order to improve the accuracy of this model since it involves integrating two different data acquisition platforms with its own time frame. However, proposed method could still identify approximately angular difference between cutting flute and spindle reference position which can be interesting for online monitoring applications.
The specific force coefficients are identified in different regions of the map for Titanium and Aluminium due to the materials distinct machinability. Hence, the gap between the cutting force coefficients for both materials makes it suitable for applying smart machining techniques using data monitoring in hole making processes of stacked materials made out of Aluminium and Titanium.
Similar data maps can be developed in orbital drilling by online monitoring of cutting force and cutting flute position for material identification in stacked aerospace applications of Aluminium and Titanium alloys.