Int J Adv Manuf Technol (2012) 59:1–7 DOI 10.1007/s00170-011-3472-6 ORIGINAL ARTICLE Dynamics of the guideway system founded on casting compound Bartosz Powałka & Tomasz Okulik Received: 31 December 2010 / Accepted: 13 June 2011 / Published online: 30 June 2011 # The Author(s) 2011 This article is published with open access at Springerlink.com Abstract The work presents a new technology for the assembly of ball guideway systems which involves the use of a thin layer of a casting compound The experimentally verified simulation research presented in the work indicates that the use of the casting compound between the guide rail and the bed of the machine tool positively influences the dynamics of the system The paper is concerned with the comparison between the new solution with the guide rail assembly technology presently in use on the basis of a guideway system consisting of a body and a milling table The dynamics was compared with the use of a frequency response function which had been determined in an impulse test The proposed solution is characterised by a higher dynamic stiffness, which may directly influence the precision of the machined surfaces Keywords Casting compound Ball rail system Guideways Dynamics Introduction Linear ball guideways, which are now being used more often in modern machine tools, have replaced the previously used slide guideways A considerable disadvantage of the slide guideway system was the stick–slip phenomenon B Powałka (*) : T Okulik Institute of Manufacturing Engineering, Faculty of Mechanical Engineering and Mechatronics, West Pomeranian University of Technology, Szczecin, Piastów 19, 70-310 Szczecin, Poland e-mail: bartosz.powalka@zut.edu.pl T Okulik e-mail: tomasz.okulik@zut.edu.pl which occurred while machining at a low feed rate [1, 2] This would contribute to the deterioration in the accuracy of the machine tool positioning Fortunately, the use of ball guideways eliminated this phenomenon The introduction of ball guideway systems improved the operating properties of the machine tool frame system by reducing the resistance to motion and increasing the permissible feed speed It also assisted in simplifying the assembling technology compared to the slide guideway system However, the main disadvantage of the ball guideway system is its low damping Low damping might lead to vibrations, which may, in turn, lead to the appearance of chatter marks on the machined surface Machine tool constructors have tried to improve the dissipation parameters of the machine tool body system in various ways One of the applied solutions is the use of composite materials which have high damping qualities coupled with the high specific stiffness for the construction of the machine tool body system Such properties of the composite are achieved by using a material with high Young module and a material with high damping Choi and Lee [3] proposed a spindle construction of carbon fibre–epoxy, which resulted in an increased natural frequency and damping than that of the steel spindle Suh et al [4] proposed the use of a carbon fibre composite laminate for the construction of the spindle cover Other examples of the use of composite materials for the improvement of dynamic stiffness regarded headstock [5] and machine tool columns [6] Kim et al [7] designed a three-axis ultra-precision CNC grinding machine whose bed was made of resin concrete which contributed to the increase in the damping capacity The effectiveness of the use of resin concrete for the construction of the bed was verified in the impulse test as well as while machining hard and brittle materials The article by Kim et al [8] presents a research on sandwich structures composed of fibre-reinforced composite materials, polymer foams and resin concrete in regard to their use for the construction of a micro-EDM machine structure The constructed prototype was characterised by good stiffness and dissipation properties Another interesting way of increasing the damping in machine tools is the use of viscoelastic materials to dissipate energy [9, 10] The concept of viscoelastic materials is based on the use of constrained layer damping CLD [11, 12] The vibration damping mechanism in the structures consisting of a viscoelastic layer bounded by steel sheets on both sides was examined by Chen et al [13], who used the theory devised by Ungar [14] Wakasawa et al [15] examined structures packed with balls The use of such structures allowed for a considerable increase in the damping capacity The research took into consideration the influence of the ball size, ball arrangement as well as the degree in which they were packed together and the direction of excitation on the increase and stability of the damping capacity An increase in the damping capacity might also be achieved thanks to the use of cementitious materials [16] Rahman et al [17] investigated the influence of machining on two lathes, a ferrocement bed lathe and cast iron bed lathe, on the tool life The improvement in tool life for the ferrocement bed lathe was attributed to its higher damping capacity Fig a The view of the samples used in the investigation b The characteristics of the load used in the investigation Int J Adv Manuf Technol (2012) 59:1–7 Fig The schema of the test stand for determining the deformations of the contact layer An increase in the dissipation properties might also be achieved using polymer inserts in the construction of the guide carriage [18] The use of a polymer impregnated concrete damping carriage was compared with a steel damping carriage The polymer impregnated concrete damping carriage appeared to be a better solution than the steel damping carriage due to the increase in the damping capacity within the frequency range of up to 650 Hz The application of the casting compound, presented in this paper, was motivated by the need to eliminate machine tool bed grinding required before guide rail assembly The grinding operation is expensive, especially in the case of large-size machine tools If the machine tool bed is finished by milling instead of grinding, it will increase productivity and cut production costs considerably Milled surfaces are expected to have a lower contact stiffness than ground surfaces which is due to the lower real contact area In this paper, a layer of EPY (tradename) casting compound [19] is applied between a guide rail and the machine tool bed as the damping material to compensate for the decrease in Fig The graph of the deformations registered for sample C by the sensor located inside the sample for the examined thicknesses of the EPY resin layer in comparison with the reference sample A Int J Adv Manuf Technol (2012) 59:1–7 Fig Schema of the physical model of the milling table Fig Comparison of the contact stiffness for various thicknesses of the EPY resin layer contact stiffness EPY material is used in the seating of main engines, gears, power generators, compressors, bearings, stern tubes, tanks and many other naval machinery First, we examined the effect of the thickness of the EPY layer and machining method on the contact stiffness and damping capacity The obtained results were used to build a simulation model of a machine tool guideway system presented in chapter three Model research presented in the work indicated that the use of a thin layer of EPY improves the dynamic characteristic of the machine tool The positive influence of the use of the EPY layer obtained as a result of numerical simulations was confirmed experimentally Static tests of the samples In order to verify the new method of assembling the ball guide rails with the use of a layer of EPY, it was checked how its usage influences contact stiffness and damping of the joint of the guide rail and the bed sample Therefore, the experimental research was conducted by means of the use of three bed samples of various surface quality and roughness of the assembling surface The surface of sample B (Ra =4.095 μm, Rz =21.80 μm) was precisely milled The Fig Comparison of the damping capacity for various thicknesses of the EPY resin layer surface of sample C (Ra =8.152 μm, Rz =39.50 μm) was milled with a worn cutter, while the surface quality of sample D (Ra =7.898 μm, Rz =39.28 μm) was like the surface of a billet Figure 1a presents the samples used during the research Additionally, sample A (Ra =0.126 μm, Rz =1.27 μm), whose surface was ground in accordance with the current assembling technology, was used in the research as a reference point for comparison The guide rail was fixed to the bed sample using an intermediate layer of EPY resin with the thickness of to mm The thickness of mm was assumed to be a state in which a surplus of the thin EPY layer on the sample was squeezed out by the rail The intermediate layer filled only the irregularities of the surface resulting from the machining The samples were subject to quasi static compression on the INSTRON testing machine and their force– displacement responses were measured Figure presents a schema of the research stand for determining the deformations in the contact layer on the joint of the rail, the EPY layer and the bed sample Figure 1b presents the characteristics of the quasi static load used during the research The loading force increased sinusoidally to the value of Fmax =80 kN, in time t2=80 s Time t1 and t3 were equal to s Figure presents the displacements registered for sample C by the sensor situated inside the sample for all the examined thicknesses of the EPY resin layer in comparison to the reference sample (A), which corresponded to the guide rail current assembling technology It might be noticed on the graph that the use of the thin (0 mm) layer of EPY resin slightly increases the deformations of the Fig Difference in the investigated models Int J Adv Manuf Technol (2012) 59:1–7 Fig 10 The view of the stand for the milling table examination Fig Frequency response function for direction Y for the analysed models contact layer, i.e there is a slight decrease in the static stiffness It was found that an increased thickness, over mm of the EPY resin layer significantly reduces the contact stiffness of the joint The contact stiffness and the damping capacity were determined as a result of the conducted quasi static compression tests on the bed samples Figures and present the experimentally obtained coefficient values The first bar in each graph corresponds to reference sample A machined in accordance with the technology used so far The second bar corresponds to the samples which were not ground (B, C, D) and where the intermediate contact layer was not used The subsequent bars correspond to the samples with an increasing thickness of the intermediate layer It might be noticed on graph that the contact stiffness of the samples without the EPY layer (13,085 N/μm—sample B) is close to the stiffness with the ‘0’ layer (11,789 N/μm—sample B) In the case of sample B, for which the value of the contact stiffness for the ‘0’ thickness is the highest, a 25% decrease in stiffness was observed in comparison to sample A Based on the comparison between Fig it might be concluded that the highest damping capacity of 0.412 appears for the Fig Frequency response function for direction Z for the analysed models ‘0’ thickness of EPY layer The damping capacity for the ‘0’ thickness of the resin layer is the highest for sample B and it is ten times higher than for sample A (0.038) Thus, the use of the thin layer of EPY on the milled surface might have a positive influence on the dynamic stiffness of the system: a slight decrease in static stiffness will be compensated by a significant increase in damping capacity Since the most promising results were obtained for the assembly of guide rails on the milled surface (sample B) with the use of the ‘0’ thickness layer of EPY (0 mm), only this solution is compared to the traditional solution in regard to its dynamic stiffness in the simulation research Dynamic response of the milling table model The contact stiffness and damping capacity obtained from the static experimental research were used to build a simulation model of the milling table mounted to the bed using guide rails The goal of the analysis was to investigate the influence of a decrease in contact stiffness with the simultaneous increase in damping capacity on the dynamic stiffness observed for the guide rail mounted to the bed via intermediate layer of EPY resin Figure presents a schema of the physical model of the milling table together with the assumed location of the machining force The simulation research was conducted for two variants of the guideway system assembly In the first simulation model, the stiffness and damping parameters corresponded to the current technology for assembling the guide rails (steel–steel contact, parameters Fig 11 Schema of the test stand used for the dynamics tests Int J Adv Manuf Technol (2012) 59:1–7 Fig 14 The location of the points used in the comparative analysis Fig 12 The location of the measurement points in the dynamics tests of sample A) The stiffness and damping parameters implemented in the second model tested in the simulation were those for assembling the guide rails on a thin layer of EPY (steel–EPY layer–steel contact, parameters for sample B for mm EPY) Figure schematically presents the models used in the numerical simulation It follows that since the machining force has a dynamic character, its dynamic characteristics play a very important role in the evaluation of machine tool performance The dynamics of the machine tool are frequently represented in terms of frequency response functions (FRFs) The amplitude of FRFs and, in turn, the level of vibrations depends on the stiffness and damping parameters of the machine tool Thus, the frequency response function can be used to evaluate the impact of a simultaneous decrease of the contact stiffness and an increase of damping capacity on the dynamic performance of the guide rail mounted via the intermediate layer of the EPY Figures and present the frequency response functions obtained for the simulation model of the milling table, for direction Y (Fig 8) and direction Z (Fig 9), respectively Fig 13 A fragment of the guideway connection with a thin layer of EPY resin—the second stage of the investigation Direction X was disregarded due to the fact that for the prepared model the stiffness in this direction depended mainly on the stiffness of the lead screw and, thus, the differences in FRF for the two considered models were negligibly small It might be noticed on the graphs that the use of the thin layer of EPY reduces the amplitude of the system’s response to the dynamic excitation For direction Y, there was a decrease in the amplitude of about 58% compared to the technology for assembling guide rails used so far as well as a slight decrease in the natural frequency from the value of 553 Hz to the level of 547 Hz For direction Z, a decrease of about 54% in the amplitude was observed, while maintaining the same resonance frequency of the simulated system Numerical simulation on a simple model indicated that the use of the thin layer of EPY positively influenced the dynamic stiffness of the examined guideway system Experimental dynamic tests As numerical simulations presented in section show there is an improvement of dynamic properties of the milling table model with its guideway when a thin layer of EPY is used The authors were encouraged to perform an experimental verification of the obtained results The test stand Fig 15 FRF at point 12 due to excitation at point 23 in direction +Y Int J Adv Manuf Technol (2012) 59:1–7 Fig 17 Vibration mode at 455.4 Hz Fig 16 FRF at point 12 with excitation at point 22 in direction −Z used for this purpose, presented in Fig 10, is geometrically similar to the simulation model The stand consisted of a body element with a table supported by the linear ball bearings The body was made of grey cast iron The mass of the body element was ca 314 kg The guideway connection was made with the use of ball guideway elements of Bosch-Rexroth which consisted of two guide rails of 25 and 1,410 mm in length, on which four guide carriages of 25 in length and catalogue number 1605-213-10 were moving (two carriages per each rail) The carriages of the guideway system had the preload equal to 2% of their dynamic load capacity The dynamic load capacity of each carriage was 22,800 N A table, also made of grey cast iron, with a mass of 69.4 kg was mounted on top of the carriages The guideway elements were founded with the use of side fixing slits in accordance to the recommendations of the guideway system producer The screw connections of the guideway system were tightened up with a torque recommended by the producer In addition, a turned lead screw with an external diameter of 24 mm and the lead of mm was used to position the table Front-end Scadas III was used during the investigation of dynamics for the data acquisition The excitation was performed by means of a Kistler modal hammer Kistler and PCB accelerometers were used for measuring system response The measurement data were processed with the use of LMS Test Lab software Figure 11 presents the experimental set-up used for investigation of dynamics Thirty-three measurement points were located on the tested object including: eight points on the guide rails, eight points on each body element, four points on guide carriages (one on each carriage) and 13 measurement points on the table Triaxial accelerometers were used to measure the vibration signal in each measurement point The location of the measurement points is presented in Fig 12 During the investigation, the tested system was excited successively at two points One of the excitation points was located in the central point of the table and the direction of excitation for this point corresponded to −Z The second excitation point was located on the side surface of the table near the guide carriage In this case the direction of excitation corresponded to +Y The frequency response functions were determined for each of the tested directions based on 30 realizations of the excitation signal Investigations were conducted for the guideway with and without a thin layer of EPY (Fig 7) First, the test stand was assembled in accordance with the assembling technology recommended by the producer of the guideway systems Between the guide rail and the body element there was a steel–steel contact Then the guideway system was disassembled and a thin layer of EPY was inserted between the guide rail and the body element During the assembly, the surplus of the intermediate layer was squeezed out only to fill the irregularities on the contact surfaces of the guideway system Figure 13 presents a fragment of the guideway connection with a thin layer of EPY used during the second stage of the investigation Excitations and measurements points are shown in (Fig 14) Figure 15 shows FRF at point p12 due to excitation at point p23 (Figs 12 and 14) within the frequency range from 30 to 1,000 Hz Application of the EPY results in a decrease of amplitudes in the vicinity of dominant resonances The observed amplitude reduction varies from 18% for the dominating mode around 309 Hz to 40%, for Table Modal parameters of the investigated object Mode Mode Mode Mode Mode Mode Mode Mode fn [Hz] ζ [%] fn [Hz] ζ [%] fn [Hz] ζ [%] fn [Hz] ζ [%] fn [Hz] ζ [%] fn [Hz] ζ [%] fn [Hz] ζ [%] fn [Hz] ζ [%] With EPY Without EPY 62.0 62.8 3.41 3.23 77.6 80.3 1.93 1.71 86.9 88.1 1.49 1.71 305.9 308.43 0.64 0.57 329.1 332.2 0.66 0.55 452.0 455.4 0.94 0.47 488.2 495.5 0.60 0.31 606.3 621.6 0.54 0.47 Int J Adv Manuf Technol (2012) 59:1–7 the 605 Hz resonance Figure 16 presents the FRF measured at point p13 due to excitation at point p22 Similarly, an improvement of the dynamic performance of the system is observed A significant drop of FRF amplitude has been observed in the vicinity of 500 Hz (about 48%) The amplitude reduction and also a decrease of resonance frequencies can be attributed to the increased damping introduced by the application of the EPY layer Table summarizes the modal parameters of the compared configurations An improvement of damping ratios (ζ) is more significant for modes that exhibit relative table–machine tool bed motion For instance, the damping ratio of Mode has improved by 100% This mode is visualized in Fig 17 Discussion and conclusions The impulse tests conducted on the described test stand validate the statement that the use of a thin layer of EPY improves the dynamic performance of the object An increase in the dynamic stiffness of the system was obtained for two tested perpendicular directions Z and Y An increase in the dynamic stiffness resulting from the increased damping capacity in the EPY layer occurred in the areas of dominant resonances These resonances are responsible for the dynamics of the tested system This means that the decrease in stiffness resulting from the use of the thin layer of EPY is compensated with a significant increase in damping capacity The increase might be caused by the convection of the mechanical energy of vibrations into the thermal energy on the resin–steel contact [12, 13] or by an increase of the effective contact surface as the resin fills all the irregularities of the connected surfaces [19] The explanation for the complex mechanism of the damping of vibrations in the thin layer of EPY resin is beyond the range of this work The presented proposal for the assembly of guideway systems that make use of a layer of EPY resin might be very attractive from the practical point of view The attractiveness results from the reduction in the costs of preparing the assembled surfaces for the guide rails by eliminating the expensive grinding operation An undeniable advantage of the method is also its simplicity as it does not require the construction of any special equipment The solution presented in the work is the subject of patent application no P388153 in the Patent Office of the Republic of Poland Acknowledgement The work was financed from the resources for science in the years 2009–2010 as a research project no N503 174637 Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited References Bell R, Burdekin M (1969–70) A study of stick–slip motion of machine tool feed driver Proceedings of the Institution of Mechanical Engineers 184:543–560 Marui E, Endo H, Hashimoto M, Kato S (1996) Some considerations of slideway friction characteristics by observing stick–slip vibration Tribol Int 29:251–262 Choi JK, Lee DG (1997) Manufacture of a carbon fibre–epoxy composite spindle bearing system for a machine tool Compos Struct 37:241–251 Suh JD, Chang SH, Lee DG, Choi JK, Park BS (2001) Damping characteristics of composite hybrid spindle covers for high speed machine tools J Mater Process Technol 113:178–183 Chang SH, Kim PJ, Lee DG, Choi JK (2001) Steel-composite hybrid headstock for high-precision grinding machine Compos Struct 53:1–8 Lee DG, Chang SH, Kim HS (1998) Damping improvement of machine tool columns with polymer matrix fiber composite material Compos Struct 43:155–163 Kim HS, Jeong KS, Lee DG (1997) Design and manufacture of a three-axis ultra-precision CNC grinding machine J Mater Process Technol 71:258–266 Kim DI, Jung SC, Lee SH, Chang SH (2006) Parametric study on design of composite foam resin concrete sandwich structures of precision machine tool structures Compos Struct 75:408–414 Marsh ER, Slocum AH (1996) An integrated approach to structural damping Precis Eng 18:103–109 10 Bamberg E, Slocum AH (2002) Concrete-based constrained layer damping Precis Eng 26:430–441 11 Plass HJ (1957) Damping vibrations in elastic rods and sandwich structures by incorporation of additional viscoelastic material In: Proceedings of Third Midwestern Conference on Solid Mechanics, pp 388–392 12 Ross D, Ungar E, Kerwin EM (1959) Damping of plate flexural vibrations by means of viscoelastic laminae In: Ruzicka JE (ed) Structural Damping Colloquium ASME, Atlantic City 13 Chen YS, Hsu TJ, Chen SI (1991) Vibration damping characteristics of laminated steel sheet Metallurgical Trans A 22A:653–656 14 Ungar E (1979) In: L.L Beranek (ed.) Noise and vibrations control McGraw-Hill, Inc., New York 15 Wakasawa Y, Hashimoto M, Marui E (2004) The damping capacity improvement of machine tool structures by balls packing Intern J Machine Tools and Manufacture 44:1527–1536 16 Rahman M, Mansur MA (1993) Evaluation of a lathe with ferrocement bed Annals of the CIRP 42:437–440 17 Rahman M, Mansur MA, Lau SH (2001) Tool wear study in a lathe made of cementitious material J Mater Process Technol 113:317–321 18 Rahman M, Mansur MA, Lee LK, Lum JK (2001) Development of polymer impregnated concrete damping carriage for linear guideways for machine tools Intern J Machine Tools and Manufacture 41:431–441 19 Grudziński K, Jaroszewicz W (2004) Seating of machines and devices on foundation chocks cast of EPY resin compound Zapol, Szczecin Int J Adv Manuf Technol (2012) 59:9–19 DOI 10.1007/s00170-011-3469-1 ORIGINAL ARTICLE Drilling performance of green austempered ductile iron (ADI) grade produced by novel manufacturing technology Anil Meena & M El Mansori Received: 17 January 2011 / Accepted: 13 June 2011 / Published online: July 2011 # Springer-Verlag London Limited 2011 Abstract Machinability study on drilling of green austempered ductile iron (ADI) grade was conducted using a TiAlN-coated tungsten carbide drill The green ADI grade was produced by a novel manufacturing technology known as continuous casting-heat treatment technology to save energy and time in foundry However, in spite of good combination of strength, toughness and enhanced wear resistance, the microstructural properties of ADI sometimes lead to machinability issues The effect of cutting parameters on cutting force coefficients, chip morphology, and surface integrity of the drilled surface were discussed Results showed that the strength properties of novel ADI are comparable to that of ASTM grade ADI, whereas percent elongation is comparable to that of ASTM grade ADI Results obtained also showed that the combined effect of cutting speed at its higher values and feed rate at its lower values can result in increasing cutting force coefficients and specific cutting energy At higher cutting speed, hardness values increases at the subsurface layer of the drilled surface due to plastic deformation Keywords Austempered ductile iron Novel manufacturing technology Drilling Cutting force coefficients Surface integrity Introduction Austempered ductile iron (ADI) is an alloyed heat-treated ductile cast iron [1] In recent years, ADI has emerged as a A Meena (*) : M El Mansori Arts et Métiers ParisTech, LMPF-EA 4106, Rue Saint Dominique, BP 508, 51006 Châlons-en-Champagne, Cedex, France e-mail: Anil.MEENA-0@etudiants.ensam.eu major engineering material due to its high strength and hardness, coupled with substantial ductility and toughness [2] The attractive properties offered by ADI are attributed to a unique “ausferrite” microstructure that is induced by austempering heat treatment process [3] Ausferrite consists of graphite nodules embedded in a matrix of acicular ferrite and carbon enriched austenite [4] However, this unique microstructure significantly affects mechanical and thermal machining properties due to its high strength, hardness, and the inclination of its retained austenite to strain hardening, which leads to short contact length and higher mechanical loads on the cutting tool’s edge [5] From the machinability point of view, the characteristics of ADI derived from the austenite are a low thermal conductivity and high workhardening coefficient [6] The austenite lattice has a higher tendency to deform due to the greater number of sliding planes, while the increase in strength and hardness during the deformation also results as a transformation of retained austenite to martensite During machining, the chips are formed on the basis of catastrophic failure in narrow shear surface due to low thermal conductivity of austenite In such a way, unfavorable, segmental chips are formed [7] This limits the use of the material in the various industrial sectors and causes: the formation of built up edges (BUEs) when carbide tools are used, low tool life, increased cutting forces, and the appearance of unfavorable tough chips during machining From the open literature [8–12], it was found that very few studies have dealt with the drilling of ADI specifying the effects of cutting parameters on chip morphology and the resultant chip formation process Drilling is, however, a major machining process for many applications Indeed ADI material, thanks to its high strength to weight ratio and enhanced mechanical properties, predestine this material to act as a substitute for forged steel and cast iron components As such, ADI with current applications in connecting rod and crankshaft is of increasing interest in automobile 10 Int J Adv Manuf Technol (2012) 59:9–19 industries for which drilling is one of the most critical machining processes An insightful understanding of the machinability of ADI material can lead to a better process economics, increased process stability, improved tool life, and reduced tooling cost Due to its high strength and hardness properties, the cutting tools for machining ADI should fundamentally yield at the same time; have high temperature hardness and strength, show excellent hot chemical inertness as well as high toughness at the higher temperatures [13] To meet all these requirements, TiAlN-coated tungsten carbide tools have been used for all experiments Titanium-based coatings, especially TiAlN (titanium aluminum nitride), are used in a broad range of machining operations TiAlN coatings are well known for their excellent wear and oxidation resistances, which enable improved machining process at high material removal rates During high temperature applications of this coating, a very dense and strongly adhesive layer of aluminum oxide (Al2O3) is formed by aluminum atoms diffusing to the surface preventing further oxidation Because of its super saturated metastable phase, the TiAlN-coatings also show age hardening effect, which increases its hardness at the higher temperatures [14] This paper thus focuses on the feasibility of novel manufacturing technology to produce ADI and its impacts on the material properties of ADI It also focuses on the Fig Schematic representation of conventional and novel heat treatment process for ADI experimental studies of cutting force coefficients, chip morphology, and surface integrity of drilled surface while drilling green grade of ADI with TiAlN-coated tungsten carbide tools under flooded conditions for different speed– feed rate combinations Experimental procedure and sample preparation 2.1 Workpiece material Specimens were produced by a novel manufacturing technology known as continuous casting-heat treatment technology developed by the integration of the casting (metallic mold) and heat treatment process in the foundry [15] In this process, spheroidization and inoculation were performed when the molten material temperature reached 1,450°C Spheroidization was done in tundish ladle and subsequently inoculation was done in a pouring ladle For spheroidization and inoculation FeSi–Mg (ferrosilicon magnesium) and FeSi (ferrosilicon) were used, respectively The molten material was then poured in a metallic mold to make the specimens of size 180 mm ×30 mm×15 mm rectangular blocks In the temperature range of 1,000– 1,100°C, the casting was shaken free of the metallic mold Pouring Casting shake-out Austenitization Temperature (°C) Novel process Quenching Austenite Pearlite Austempering Ausferrite Bainite Conventional process Ms Austenitizing Austempering Time Int J Adv Manuf Technol (2012) 59:377–395 389 Table Detail coefficients for DWT of X1–X3 using Haar wavelet: —detail coefficients equal to zero and 1—detail coefficients not equal to zero x1 x2 x3 CSc CS1+ CS2 CS1− CSL 0 1 1 0 0 0 sequence of motion given by the planner can be used in both cases Moreover, CSL rarely appears in practice It is worth noting here that Table has been derived from the analytical model For practical use, we have to find a way to extract the relation/signal type from nonideal sensory signals Furthermore, the extraction process has to be performed in real time Thus, for these purposes, we have used DWT proposed for the CS recognition machine offline training (Fig 4) For selection of appropriate wavelet, we exploit the fact that the wavelets from Daubechies family (db wavelets) have the first N vanishing moments [40]: Z xn y xịdx ẳ 0; n ẳ 0; 1; :::N ð16Þ where N is the order of wavelet = N vanishing moments means that db wavelets of order N are orthogonal to polynomials 1, x, x2,… xN−1 and that the signal DWT is applied so that it will be approximated by polynomials of order N−1 In other words, in all regions in which signal is well approximated by a polynomial of order N−1, the wavelet coefficients (i.e., detail coefficients) at DWT representation will be equal to zero To illustrate this property, Fig shows features X1–X3 during one part mating cycle obtained from the quasi-static model along with the detail coefficients of the first level of DWT We have used the first two wavelets from Daubechies family (db1 and db2) Detail coefficients for db1 (also known as Haar wavelet) are equal to zero in all areas in which signal is constant For db2 wavelet, detail coefficients are equal to zero in all segments in which the signal is linear or constant Summarizing properties defined by relation 16, as well as the relations from Table 3, the fourth level detail coefficients obtained by DWT of X1, X2, and X3 using the db1 wavelet are chosen as representative and discriminative features Thus, we transform the feature space X=[X1 X2 X3] into a new feature space x=[x1 x2 x3]: x ¼ D W T 4db1 ðXÞ ð17Þ The level of DWT determines the signal history (i.e., number of samples) that will be involved in the current detail coefficient generation (relations 2–4) The fourth level DWT using db1 wavelet requires 16 samples to generate a detail coefficient Table summarizes the idealized values of features x1, x2, and x3 for all CSs These features should be equal or close to zero in the areas where the trend of the signal is constant Thus, they are coded by For all other signal trends, the corresponding features are different from zero, and they are coded by one in Table Practically, this set of extracted features can be considered as a code, which can be used to uniquely code the set of CSs in cylindrical part mating with chamfer crossing Before features generation, the signal is de-noised to eliminate not only the noise but also a highly dynamical and stochastic phenomena disregarded in the quasi-static model Once again, we have utilized DWT for signal denoising Conventional de-noising techniques based on IIR or FIR low-pass filters introduce signal distortion On the other hand, DWT-based de-noising techniques are characterized by phase correctness and good time localization While choosing the wavelet for de-noising, it is important to have in mind that further signal analysis (extraction of features) will be based on db1 wavelet Consequently, the wavelet used for de-noising has to be a wavelet of the order higher than one from Daubechies family, or a wavelet from another family In order to keep the computation time required by de-noising as low as possible, we have chosen wavelet db2 Figure shows the feature vectors [x1 x2 x3] generated using the described procedure from the quasi-static analytical model with aggregated noise A total of 450,000 patterns have been generated for different part mating cycles with initial errors in the range e0∈[0, mm] and θ0∈[0, 1°] Since we start from the quasi-static model with aggregated noise, the generated features are not equal but close to zero in the areas where the trend of the signal dominantly behaves as constant From the described procedure, it follows that transformations ℑ used to generate features from the insertion force have three steps: (1) generation of X1–X3 using relations 13–15, (2) DWT-based de-noising using db2, and (3) extraction of features x1–x3 using db1 DWT These transformations are handed over to the online contact states recognition machine (as shown in Fig 3) Wavelets used in ℑ have not been chosen arbitrarily, but in mathematically consequent manner We have shown that vanishing moments are an enabling property of Daubechies wavelets This property makes DWT an extremely powerful tool for feature extraction Another very important issue is the real-time applicability of DWT Exploiting the orthogonality of db wavelets and polynomial functions, we have managed to extract a truly qualitative set of features [x1 x2 x3] from the force signal 390 Fig Representation of the generated patterns in three dimensional feature space; a total of 450,000 patterns for initial error in the range e0∈[0, mm] and θ0∈[0, 1°] are shown along with CSs Int J Adv Manuf Technol (2012) 59:377–395 0.04 1.5 0.02 0.5 x3 x2 -0.02 -0.5 -0.04 -1 -0.06 -1.5 -15 -10 -5 -15 -10 x1 -5 x1 1.5 CSc 0.5 x3 CS1+ CS2 -0.5 CSL, CS1- -1 -1.5 -0.06 -0.04 -0.02 0.02 0.04 x2 These features are independent of the insertion force intensity level, i.e., variation of system’s rigidity (kx, kθ), friction coefficient (μ), initial pose error (e0, θ0) between objects to be mated, chamfer parameters (a, α), etc They contain the type of relation between force derived features X1–X3 and time 4.3 Feature space partitioning Following the proposed offline training for CS recognition machine (Fig 4), we partition the adopted feature space (Eq 17) using SVM For simplicity and real-time applicability, we have opted to use a linear kernel function Since SVMs are binary classifiers, we propose a hierarchical approach to classify the patterns into multiple classes At each level of classification, one contact state is separated from the remaining contact states From Table 4, it follows that the classification can be monothetic, meaning that only two out of three generated Table Classification steps features can be used for separating each of the contact states At step 0, based on intensity of ZV, force component CS0 is separated, as a contact state in which the value of this force is ZV