2010 SIMULIA Customer Conference 1

Machining Part Program Optimization through an Advanced Multidisciplinary Procedure

A. Del Prete1 , A. A. De Vitis1 , M. Parodi2

1University of Salento, Dep.t of Innovation Engineering, Lecce , Italy ; 2Exemplar S.r.l. , Turin, Italy

Abstract: In the production of aerospace engine components, metal cutting processes are characterized by a strong demand for increased productivity that does not compromise the high quality of the product. The antithesis stands in the fact that it is necessary to maximize the feed rate and cutting velocity in order to reduce the processing time without compromising the quality of the worked component. In aerospace machining applications on hard-cut materials like: nickel based alloys, titanium alloys, etc, it is fundamental to keep under control the local cutting zone phenomena in order to assure the final product quality. The machining process design development can be summarized by the following steps: definition and verification of the Part Program (PP) through dedicated CAD–CAM software applications, post processing of the produced PP, CNC machine simulation and physical tryout. A further development of this procedure foresees the application of the kinematic optimization to improve the cutting process with a significant time reduction through the optimization of material removal along tool path. In this study a new multidisciplinary procedure is proposed. The aim of the authors is to modify the operation parameters set in the already kinematically optimized PP according to the constraints arising from the physical nature of the cutting process obtained by FEA. A milling operation that include the use of rough and finish tools related to an aeronautical engine component made by Inconel 718 has been chosen to test the developed methodology. The aims of the procedure is to minimize the execution time of the cutting process in compliance to physical micro-scale constraints (maximum admissible cutting edge temperature and maximum admissible Cutting Forces).This foresees the integration of the CAM softwares: Vericut for tool-path verification and Optipath for kinematic optimization of the given PP in the iSIGHT model. The procedure automatically extracts the values of feed and speed in all the blocks of the PP, which have been kinematically optimized, to verify if they respect upper limits (previously set) of: analyzed responses. In the PP blocks where the physical constraints are violated, a Pointer algorithm it has been used to automatically identify the optimal set of the process parameters within the defined design space of the approximation models in order to respect the required physical constraints. The new set of process parameters has been updated into the blocks of the analyzed PP. Keywords: Machining FEA, Tool-path Simulation, Multidisciplinary Optimization, Approximation

1. Introduction

In metal cutting application, the process development can be summarized by the following steps: definition and verification of the Part Program through dedicated software applications (CAM, Post processor, CNC machine simulator), physical tryout, part inspection on CMM and Part Program certification (Figure 1). Tool path simulation is commonly adopted in order to obtain a

2 2010 SIMULIA Customer Conference

preliminary validation of a Part Program before its physical tryout. With this type of simulation the process designer may check: possible collisions between the tool and/or the stock and fixture, the material removal during rapid motion, etc. This type of application provides specific measuring tools able to compare the machine part with its original CAD model. These are the main results of the integration between CAD–CAM technologies (process design) and the shop-floor environment. A CAM simulation software has to be able to reproduce CNC machine tool characteristics (axes configuration, workspace, fixture, tooling room) and its CN in order to obtain: simulated axes motions, tools loading and all the possible instructions of a specific control.

Figure 1: Traditional tool-path verification workflow.

A further development of this procedure provides an additional step (Figure 2). In this case, the validation of a part-program through simulation is fundamental to improve the cutting process. A significant reduction of cutting process time can be achieved with the optimization of material removal rate along tool path using proper algorithms.

Figure 2: tool-path verification and kinematic optimization workflow.

After the simulation, the studied part-program can be optimized with a kinematic algorithm that operates to keep constant the chip removed volume per time unit. The purpose of this additional process design step is to improve the cutting process with a significant time reduction through the optimization of material removal along tool path. This target is pursued using proper algorithm that operates with the logics called Constant Removed Volume. Every milling operation can be characterized by specific values of cutting parameters: depth of cut (p) [mm]; tool radial engagement (b) [mm]; feed per tooth per revolution (az) [mm\tooth-rev]; spindle speed (n) [rpm]; tool diameter (D) [mm]; number of inserts in the tool (z). Cutting speed Va is defined by Equation 1:

1000aDnV π= [m\min] (1)

The relationships among cutting parameters isdefined by Equation 2:

a zV n a z= ⋅ ⋅ [mm\min] (2)

These parameters determine the amount of material removal during tool path; removed material volume (Q) is defined by Equation 3:

2010 SIMULIA Customer Conference 3

tQ b p V= ⋅ ⋅ [mm3\min] (3)

In the kinematic optimization method, b and p in Equation 3 are constant and assigned to obtain the best use of the tool in accordance with tool producer indications, so optimization algorithm modifies Va to respect the constancy of removed volume and to respect the fixed value of b and p. It is important to notice that the optimization procedure does not modify the tool path but, it can produce some breaks in the path depending on the local quantity of material that has to be removed. Spindle speed is considered constant (Del Prete, 2009). In this paper a new multidisciplinary procedure is proposed (Figure 3). The already illustrated workflow is integrated with a finite element cutting simulation environment where it is possible to predict the physical effects arising from the cutting process: cutting forces and cutting edge temperatures .

Figure 3: Proposed CAE-CAM integrated procedure for tool-path optimization.

The aim of the authors is to modify the operation parameters set in the already kinematically optimized Part Program according to the constraints arising from the physical nature of the cutting process obtained by FEA. In particular, tool wear is known to be strictly linked to the cutting temperature and to the exchanged forces between insert and workpiece. Being able to know the values of these forces during the process it allows to select optimal input parameters, in this way it is possible to reduce the time frame required by the process and it allows to control the tool wear. The optimization procedure uses Response Surfaces properly produced according to data calculate with FEM simulations and experimental tests carried out on the base of a DOE of the variables (input parameters) which have influence on the analyzed responses (Figure 4). The advantage of these techniques, compared to the traditional ones, is that it is required to know the exact response function of the system only in certain points of the design space and from these points it is possible to extract, thanks to the response surface, the needed values in all the other points. These results produce significant benefits in reducing the calculation time as the optimization process does not require continuous and repeated calls to the solver, but it can exploit the data provided by the approximation response surface. Optimization of machining parameters not only increases the utility for machining economics, but also the product quality to a great extent. In this study, approximation models based on RBF methodology has been developed to predicting the behaviour of analyzed responses and to create a multidisciplinary optimization procedure.

4 2010 SIMULIA Customer Conference

Figure 4: Comparison between traditional and adopted procedure.

2. Approximation model construction approach based on FEA data

The authors in this study have used data drawn from the numerical simulations and experimental tests for the construction of the approximation surface model. In this way, in particular for FE simulation that required expensive calculation, like metal cutting simulations, it becomes possible to avoid costly experimental campaigns for data acquisition. The approximation model used in this work has been developed with RBF technique Radial Basis Functions. This is a type of neural network used to approximate many types of behaviour. They employ a hidden layer of radial units and an output layer of linear units, and they are characterized by reasonably fast training and reasonably compact networks. (Weissinger, 1947) was the first to use radial basis function to calculate the flow around wings. This neural network utilizes the Gaussian curve to map values. The network has n inputs and k outputs. Radial basis network is a very efficient network when function approximation is needed because it has the ability to represent nonlinear functions (Isight user’s guide, 2009). In order to model the true functions of analyzed responses (cutting forces Fx, thrust force Fy and cutting edge temperature Tmax) through approximation surfaces based on data obtained with FEA. The data training set have been obtained through 40 simulations based on design points selected within the two-dimensional design space who has the following boundaries: feed for tooth (F); cutting velocity (S); by a version of Latin Hypercube DOE, called Optimal Latin Hypercube (Figure 5).

2010 SIMULIA Customer Conference 5

Figure 5: Data acquisition for approximation models development.

2.1 Simplified FE 2D model and data extraction procedure

The analyzed process is a down cut mill operation. In the climb mode shown in (Figure 6-A), the feed on each tooth is bigger at the initial contact between tool and workpiece and it becomes very small at the end of the engagement.

Figure 6: (A) Climb mill operation ; (B) FE simplified 2D model.

The maximum force calculated at the interface tool- workpiece it occurs when the tooth surface impacts on the workpiece. To detect the force components, Fx and Fy (Figure 6-A), a simplified 2D model has been chosen (Figure 6-B). In this case the real feed per tooth is the DOC on the

6 2010 SIMULIA Customer Conference

finite element model (az, max) and the cutting speed equal to the tangential velocity of cutting profile (Vt). The calculated values of the exchanged forces have been detected when their value it has been considered stable.Cutting temperatures data, used to create the wanted RSM, are computed as an average of temperature values extracted by five fixed nodes on the cutting edge mesh when thermal steady state condition on tool-workpiece interface is reached, after 2 mm of cutting length (Figure 7).

Figure 7: Cutting edge temperature measurement points.

2.2. FE model set up

FE analysis of the dry milling operations with the circular insert in WC and no wear have been carried out. A two-dimensional plane-strain thermo-mechanical analysis, based on the update Lagrangian formulation was performed using an implicit finite element commercial code specifically dedicated to machining FEA, SFTC-DEFORM 2D V9.0. For this purpose the considered workpiece (dimensions: 4x1.5 mm) was initially modelled with 6500 bilinear four-nodes quadrilateral elements, with dimensions respectively: of 0.007 mm along the cutting edge, 0.01 mm on the first 0.3 mm of the machined surface and 0.15 mm in the remaining area. The inserts have been modeled in tungsten carbide (WC) with 8% in Cobalt has been characterized using the software default material library. They have been modeled as rigid and meshed with 1500 elements having as small dimension of 0.004 mm in the nose zone in order to well approximate the small radius of curvature in this area. A constant frictional stress law on rake face is assumed equal to a fixed percentage of the shear flow stress of the machined material, Equation 4.

m kτ = ⋅ (4)

Where k is the shear flow stress of the workpiece material and m is a friction factor, assumed equal to 0.5. A value of 100 kW/m2K has been adopted to model the interface heat transfer coefficient, h, between tool and burr. The workpiece material is Inconel 718 with chemical composition that is in compliance with regulation AMS 5662 and with a given standard solution heat treatment and a two-step precipitation or aging heat treatment. For this type of material, proper values were defined for the following characteristics: Young’s module; Poisson’s ratio; Thermal expansion; Heat capacity; Emissivity; Thermal conductivity. Corresponding values are experimentally defined in the range of temperature [20°C; 1200°C]. For workpiece material

2010 SIMULIA Customer Conference 7

characterization in plastic field was adopted a Johnson Cook constitutive model, Equation 5 (Del Prete, 2007).

( )0

1 ln 1m

n roomeq

melt Room

T TA B C

T Tεσ εε

− = + + − −

(5)

where ε is the plastic strain, ε is the strain rate (s-1), 0ε is the reference plastic strain rate (s-1). T is the material temperature (°C), Tmelt is the melting material temperature (1400 °C) and Troom is the room temperature (20 °C). Coefficient A is the yield strength (MPa), B is the hardening modulus (MPa), C is the strain rate sensitivity coefficient, n is the hardening coefficient and m the thermal softening coefficient (Uhulmann,2007).

3. Case Study

A CCOC (Combustor Chamber Outer Casing) of an aeronautical engine in Inconel 718 has been chosen to test the developed methodology. A part program that involves the use of milling operations has been selected; this operation involves the use of two rough and seven finish tools. The related Part Program which has about 15000 instruction blocks has been simulated and the machining time has been computed. In Figure , it is reported the machining operation sequence: 1) Pocket opening; 2) Chamfer roughing; 3) Pocket milling.

Figure 8: Tool-path simulation of examined PP (by courtesy of AVIO S.p.A).

Part Program obtained whit standard CAD-CAM procedure has been subjected to kinematic optimization: this operation has as output a new Part Program. The new simulation predicted a reduction in machining time of 48% compared to the not optimized Part Program. This reduction is obtained increasing the F values but, the tool path trajectory is the same of the original Part Program. The performed kinematic optimization does not take into account the physical interaction between tool and workpiece. In the proposed multidisciplinary procedure the feed and speed indicated in the obtained Part Program thanks to the kinematic optimization have to be extracted and compared with the response surface in terms of temperature and forces generated (Figure 9) (Del Prete, 2009).

8 2010 SIMULIA Customer Conference

Figure 9: workflow of the proposed procedure.

The multidisciplinary optimization tool draws F and S values from the kinematic optimized Part Program and converts them into the required measure units of the response surface. In the Part Program, F and S are respectively expressed in [m\min] and [rpm] and it is necessary to convert them into [mm\tooth-rev] and [mm\s]. So proper equivalent relationships, Equation 6 and Equation 7 have been used to obtain the correct parameters expressions:

FfS z

=⋅

[mm\tooth-rev] (6) ; 60t

S DV π⋅ ⋅= [mm\s] (7)

The obtained values of az and Vt have been introduced in the response surfaces. The correspondent temperature, forces and surface roughness have been detected. This outputs must be less then the given fixed physical constraints. If the correspondent values of cutting edge temperature Tmax , Fx , Fy for a single az exceeded the established limits an inferior value has been considered in compliance with the fixed limits of Tmax , Fx , Fy. The optimized f values, have been reconverted in the unit of measure adopted in the Part Program and rewritten in the correspondent block of the optimized Part Program (Figure 10).

2010 SIMULIA Customer Conference 9

Figure 10: Flowchart of input-output process parameters optimization.

3.1. Multidisciplinary optimization procedure: introduction

The aim of presented work is to develop a process with Isight in order to optimize the process parameters: feed rate (F) and cutting speed (S) (minimize the time to work through the maximization of values used in the process) present in a part program while respecting the imposed constraints (maximum allowable cutting forces, maximum allowable cutting temperature and maximum allowable surface roughness). The relationship among the parameters F and S for a given cutting tool and the responses that we want to control are obtained through interpolation with approximation technique of data on cutting forces and cutting edge temperatures obtained by FEM simulations. The procedure introduces the instructions of tool substitution (cambut) in the part program based on the tool life defined by the user for each tool. Finally, the procedure creates a post-processing report that includes table of modified parameters in the PP and graphs that comparing the time execution of different examined part programs (original PP, Kinematic optimized PP and part program obtained respecting the physical constraints).

3.2. Multidisciplinary optimization procedure: description

In this section it is reported the procedure that characterize the automatic procedure for part program optimization with imposed physical constraints. The flow of process activities takes place from left to right, there is also a branch of the procedure dependent on the possibility of launching an analysis of OptiPATH (Figure 11). The macro steps of developed procedure are listed below: 1. Reading and loading of the PP and VERICUT project in the procedure. 2. Simulation of the original part program in VERICUT and extraction of tools working time information. 3. If the project Vericut has appropriate settings for the OptiPATH analysis, the process of creating the kinematic optimized part-program is enabled in VERICUT and the new cutting time is logged. 4. In the task called "Vericut CAE-CAM Optimization" are performed the operations below indicated:

a. For each tool used in the part program, the cutting parameters feed (F) and speed (S) of each blocks are extracted and optimized (the objective is to maximize the feed rate) respect the imposed physical constraints. b. For each tool used in the examined part program, the parameters F and S originally indicated are replaced with those optimized

10 2010 SIMULIA Customer Conference

c. Then the new part program is executed and registered with VERICUT cutting times of each step of the part program. d. Depending on the working time indicated for each tool in the tools wear life database, all preexisting cambut are removed and new cambut are inserted in the optimized part program.

5. Finally, the report text file created during the procedure (time, speed of cut set, tool changes, etc.) are read and processed in Excel charts.

Figure 11: Optimization procedure layout in Isight gateway.

The procedure requires two input files: the part program created by the user (of course consistent with the Vericut project file) and the Vericut project file that defines the simulation configuration. in Vericut environment . The procedure returns as output Excel files containing information of a post-processing and the optimize part program according to the imposed constraints about cutting parameters and tools wear life (Figure 12). The files returned from the procedure are:

• Template_cambut: the Excel report on the changes included in the tool part-program

• Total_Time: table containing the total time of different analyzed part-program

• PPfile_optimized_RSM: optimized part-program.

2010 SIMULIA Customer Conference 11

Figure 12: Scheme of the main input/output parameters of the procedure.

The data about micro-scale physical cutting conditions for each tool, obtained through FEM simulations, are contained in an Excel file, which has a name like: RSM_UT_ <Tool id>. Xls, for example RSM_UT_55.xls In this study the authors have used the RSM developed on the base of data obtained by FEM calculations to represent the behaviour for the analyzed responses for all the tools used in the analyzed PP. For each tool the process parameters design space has been adequate to its operative range. All Excel files relating to the data obtained for each tools have been grouped in a zip file and stored within the model.

3.3. Process parameters optimization

This activity aims to obtain a procedure that allows to minimize the execution time of the cutting process in compliance to physical micro-scale limits (Max cutting edge temperature and Cutting Forces). In this work, the optimization process is based on approximation methodology. This methodology was used by lot of researchers for modelling machining processes. It has been also successfully used for application in surface roughness analysis (Del Prete, 2010). In the presented study, the authors have used the data obtained by FEA to build mathematical approximation models by RBF technique (Radial Basis Function). These mathematical models have been coupled with a Pointer Algorithmic to obtain the optimal set of machining parameters in according with the constraints and objective function are indicated in the optimization problem of follow described :

Variables : Vt within range of definition f within range of definition Constraints : max admissible cutting edge temperature Tmax max admissible cutting force Fx, max max admissible cutting force Fy, max Objective Function : maximize feed rate f

12 2010 SIMULIA Customer Conference

The parameters: Feed (F) and speed (S) contained in the PP blocks, that previously have underwent kinematic optimization, are optimized by optimization algorithm that queries the approximation models, trained on the basis of numerical data, to meet the requirements of physical constraints (Tmax, Fx and Fy) within the upper limits, imposed by user, and to maximize the feed rate. The new set of process parameters have been converted into the units of measure used in the Part Program and then they have been updated the respective blocks. The optimization algorithm used in this study is a generic algorithm called Pointer: this algorithm has the peculiarity of using different optimization techniques depending on the behavior of the objective function, because the optimization algorithm initially interrogates the function trying to understand its nature (e.g. if is continuous, or nonlinear or discontinuous) and therefore using the optimization technique most appropriate. The Pointer technique consists of a complementary set of optimization algorithms: linear simplex, sequential quadratic programming, downhill simplex, and genetic algorithms. Since all the optimizer control parameters are automatically set with a special control algorithm, Pointer can efficiently solve a wide range of problems in a fully automatic manner (Isight user’s guide, 2009). After the rewriting of the optimized parameters, the new PP is simulated again in Vericut and the execution time is recorded. Based on each tool maximum working time defined by the user, new cambut instructions are included in the optimized part program. These new cambut instructions replace the originals, that are deleted.

3.4. Optimization results Parameter File iSight-FD called PPfile_optimized_RSM contains the part program with the optimized F, S optimized and new Cambut instructions . The kinematic optimization does not modify the tool path but, it produce some breaks in the path depending on the local quantity of material that has to be removed. Spindle speed remains constant. The numbering of kinematic optimized PP changes with a zero that is added respect to the numbering used in CAM part program. Another characteristic of Optipath PP is that the tools displacements, when they are not engaged are characterized by very high feed rate. This value is not subjected to optimization by CAE CAM procedure. At the same time the CAMBUT instructions present in the Optipath PP have been removed through the insertion and their repositioning in the CAE-CAM Part Program in according with the imposed tool life time.

3.5. Post-Processing report

Post processing consist on a summary table that indicates which part program blocks have been modified in terms of F , S and cambut added/ deleted. Table and and graphs on the number of instructions for tool change (cambut) included in the optimized part program (Figure 13). For each mill used in the PP, the number of the tool substitution and the total number of insert necessary to execute the operation are reported. Finally, post processing report shows total work execution time in the three cases examined by the procedure: original CAM part program (CAM), kinematic optimized part program (OPTIPATH) and optimized with physical constraints (OPTI-CAE)

2010 SIMULIA Customer Conference 13

Figure 13: Table and graphs summary report about cambut insertions and working time of analyzed part programs.

4. Conclusion

The presented application can be considered an effective procedure for the introduction, in the Part Program previously optimized in terms of kinematic information, of the physics of the cutting process (cutting edge temperature and forces exchanged between tool and workpiece). The physical quantities were detected using response surfaces generated from data extracted from FEM simulations performed on the basis of a DOE study. Three tool paths simulations have been ran and the execution times have been compared in the case of: 1) Non optimized Part Program; 2) Kinematic Optimized Part Program; 3) Optimized Part Program based on process parameters selected in the kinematic optimization but updated with values for F and S parameters value respecting the given physical constraints (Figure 13). The obtained results showed that the kinematic optimization dramatically decreases the execution time of the analyzed Part Program (reduction of 52% respect CAM PP). Moreover, the introduction of physical constraints reduces this percentage to 18% but at the same time it allows to meet the technological constraints set to control the tool wear. The CAE-CAM optimization (OPTI-CAE in Figure 13) has lowered many of values (F and S ) present in the PP optimized kinematically (OPTIPATH in Figure 13). This has caused an increase in processing time by 42% compared to the one required by the kinematically optimized PP. As next step the authors will focus their efforts on the search of the numerical - experimental correlation of the FEM models to improve the reliability of the RS models. Another

14 2010 SIMULIA Customer Conference

aim of future research is direct to the introduction in the multidisciplinary procedure of additional information about surface roughness, dynamic analysis (chatter) and so on. In the future, the intent of the authors is to focus their efforts to get, thanks to the development and the integration of the tools used software, a fully automatic multidisciplinary procedure.

5. References

A. Del Prete, A. Spagnolo, A. A. De Vitis, A. Anglani "Experimental evaluation of the influence of part program optimization algorithms on surface roughness in milling operation" - 9st AITeM Conference – Torino, Italy, 2009

A. Weissinger. "Lift distribution of swept-back wings". NACA pp.1120, 1947.

ENGINEOUS SOFTWARE - iSIGHT Version 3.0 User's Guide - 2008.

A A Del Prete, A.A. De Vitis, D. Mazzotta, “Design space investigation by RSM Techniques in Aeronautical Metal cutting Applications” - OPTI 09 Algarve, Portugal , 2009

E. Uhlmann, Finite Element Modeling and Cutting Simulation of Inconel 718. CIRP Annals - Manufacturing Technology, 56(1), pp. 61-64, 2007.

A. Del Prete, A.A. De Vitis, D. Mazzotta, M. Cherubini “Metal Cutting simulation as support tool to Product and Process development of aeronautical components in Inconel 718” - 10th CIRP, Scilla, Italy, 2007.

A. Del Prete, A.A. De Vitis, D. Mazzotta, A. Anglani “Numerical Simulation of Broaching Process in Aeronautical Applications” - AMST'08 8th – Udine, Italy, 2008

A Del Prete, A.A. De Vitis, A. Spagnolo, D. Mazzotta, “Cutting Parameters Optimization through an advanced CAE-CAM procedure” - NAFEMS WC 09 Crete, Greece, 2009

A Del Prete, A.A. De Vitis, A. Anglani, “Roughness inprovement in machining operations through coupled metamodel and genetic algorithms technique” - 13 th Esaform Conference - Brescia, Italy, 2010

A Del Prete, A.A. De Vitis, A. Spagnolo, “Experimental development of rsm techniques for surface quality prediction in metal cutting applications” - 13 th Esaform Conference - Brescia, Italy, 2010.