G. GANTAR et al.: OPTIMIZATION OF PRESS-FIT PROCESSES 207–212 OPTIMIZATION OF PRESS-FIT PROCESSES OPTIMIZACIJA POSTOPKOV MONTA@E Z VTISKOV ANJEM Ga{per Gantar 1,2 , Peter Göncz 1 , Miha Kova~i~ 1,3,4* 1 College of Industrial Engineering, Mariborska cesta 2, 3000 Celje, Slovenia 2 Environmental Protection College, Trg mladosti 7, 3320 Velenje, Slovenia 3 [tore Steel d.o.o., @elezarska cesta 3, 3220 [tore, Slovenia 4 Faculty Of Mechanical Engineering, University of Ljubljana, A{ker~eva cesta 6, 1000 Ljubljana, Slovenia Prejem rokopisa – received: 2020-06-05; sprejem za objavo – accepted for publication: 2020-11-12 doi:10.17222/mit.2020.100 The press-fit process is an efficient, low-cost method for joining parts. The parts that must be joined interfere with each other’s occupation of space; therefore, contact dimensions and their tolerances influence the quality of the assembly. The traditional method for the selection of contact dimensions and their tolerances is based on engineering experience. The idea of the research work presented in this paper is to optimize the press-fit process at an early stage of development process, involving prediction and optimization of the joining force and consequently the prediction and minimization of the rejection rate. Accordingly, sev- eral finite-element (FE) simulations of the press-fit process for predicting the joining forces were conducted, considering in- put-parameter variations (material properties: yield stress, hardening exponent; geometry: shaft diameter, guide diameter of the core, functional diameter of the core; friction coefficient). Based on FE simulations and 47 different input-parameter-variation results, the empirical model for predicting the joining force using the response-surface methodology (RSM) was obtained. By using RSM and a stochastic Monte Carlo simulation, the rejection rate was also determined. The predicted and the actual rejec- tion rates for selected process parameters were 1.4 % and 1.5 %, respectively. Consequently, the press-fit process can also be op- timized to reduce the rejection rate using the same Monte Carlo simulation. The results of the analysis show that the rejection rate can be reduced from 1.4 % to 0.2 %. Keywords: press fit, joining force modelling, response surface methodology, Monte Carlo simulation Monta`a z vtiskovanjem je cenovno ugoden postopek za spajanje delov. Sestavna dela zdru`imo z vzajemnim vtiskovanjem, zato dimenzije in tolerance bistveno vplivajo na kvaliteto spoja. Tradicionalno izbira dimenzij in toleranc temelji na izku{njah. V ~lanku je predstavljena optimizacija procesa monta`e z vtiskovanjem v zgodnji fazi razvoja, ki zajema napovedovanje in optimizacijo vtiskovalne sile in posledi~no zmanj{anje izmetnih kosov. Tako se je izvedlo ve~ simulacij z metodo kon~nih elementov, kjer se je spreminjalo vplivne parametre (lastnosti materiala: meja te~enja, eksponent utrjevanja; geometrija: premer gredi, premer vodila jedra, funkcionalni premer jedra; koeficient trenja). Na podlagi simulacij, izvedenih s pomo~jo metode kon~nih elementov, pri katerih smo spreminjali 47 vhodnih parametrov, smo razvili empiri~ni model za napovedovanje sile vtiskovanja s pomo~jo metode odzivnih povr{in. Z uporabo modela, pridobljenega s pomo~jo metode odzivnih povr{in in simulacije Monte Carlo se je dolo~il tudi dele` izmetnih kosov. Dejanski dele` izmetnih kosov je bil 1,4 %, napovedan pa 1,5 %. Posledi~no je bilo mo`no z uporabo Monte Carlo simulacij proces vtiskovanja optimizirati in zmanj{ati koli~ino izmeta. Rezultati analize ka`ejo, da je koli~ino izmeta mogo~e zmanj{ati iz 1,4 % na 0,2 %. Klju~ne besede: postopek monta`e z vtiskovanjem, modeliranje sile vtiskovanja, metoda odzivnih povr{in, simulacija Monte Carlo 1 INTRODUCTION A press fit is a process for the assembling of two parts. The parts are pressed together at room temperature by tools (assembly punch and assembly die) using a join- ing force, which is provided by the assembly press. The inner part (e.g., shaft) is oversized for the space in the outer part (core or housing, for example); therefore, two parts interfere with each other’s occupation of space. Both parts deform to fit together into the assembly and create a normal force. The friction force, which is caused by the normal force, holds the parts together and pre- vents disassembly during the utilization of the assembly. The selection of the contact dimensions of parts to be as- sembled determines the tightness of fit, the joining force, and subsequent the disassembly force during use, as ex- plained in 1–5 . Regardless of the simplicity of the press-fit process principle, there is a lack of generality due to the diversity of industrial possibilities in contemporary literature, al- though its outstanding potential in serial production could be well utilized. 1,6,4 In that way, the pre-production analysis of influential process parameters is essential. The relevant press-fit process research comprises: • joining materials analysis done by 1,5–8 • geometry analysis done by 1,6,9,10,11 • studies of load-specific applications done by 11 • prediction of stresses and deformations during press-fit processes by 1,6,7,9–13 . For the prediction of stresses during press-fit pro- cesses, analytical methods are used by 1,6,10 and finite-el- ement methods are used by 7,9–13 . The idea of the research work presented in this paper is to optimize the press-fit process in the early stage of the development process involving prediction and opti- Materiali in tehnologije / Materials and technology 55 (2021) 2, 207–212 207 UDK 620.1:658.515:536.911 ISSN 1580-2949 Original scientific article/Izvirni znanstveni ~lanek MTAEC9, 55(2)207(2021) *Corresponding author's e-mail: miha.kovacic@store-steel.si (Miha Kova~i~) mization of the joining force and, consequently, the pre- diction and minimization of the rejection rate, which makes the approach unique. First, a case study is presented. Afterwards, the FE model for simulations of the press-fit process is ex- plained and verified by comparing the predicted joining force with the measured one at the assembly press. Next, the development of an empirical model for predicting the joining force using RSM is shown. Afterwards, input press-fit process parameter optimization and the rejec- tion-rate prediction using stochastic Monte Carlo simula- tion are addressed. In the end, conclusions are drawn, and future work is described. 2 CASE STUDY For the purpose of this research, pistons of solenoid valves are studied that are mass produced for press-fit as- sembly by many companies worldwide. The case study is presented in Figure 1. The core is machined from the material 11SMnPb30, which is widely used due to its good machinability and the easy fragmentation of chips. The shaft is produced from the material CuZn39Pb3, which also possesses ex- cellent machinability. The mechanical properties and - curves for both materials were obtained via a standard tensile test at room temperature as described in 14 and 15 and are presented in Figure 2 and Table 1. Table 1: Mechanical properties of materials 11SMnPb30 and CuZn39Pb3 Part Material E (MPa) R p (MPa) R m (MPa) Shaft CuZn39Pb3 96000 350 480 Core 11SMnPb30 211000 530 572 The shaft and the core are chamfered, and the core is designed in such a way that its inner dimension de- creases gradually from guide diameter to functional diameter (D core1 > D core2 ). The minimum force of disassembly is defined by the designer of the assembly. In the studied case, 3 a disas- sembly force higher than 200 N is required (F AM I N = 200 N). The force for disassembly can be estimated as equal to the joining force because of the characteristics of fric- tion (at a given normal force and the coefficient of fric- tion, the friction force is equal in all directions). The maximum joining force, which causes plastic de- formation and upsetting of the shaft F AM A X , can be calcu- lated by using the following Equation (1): F d R Amax shaft ps h af t = ⋅ ⋅ π 2 4 (1) where: R p shaft = yield stress of shaft material and d shaft = diameter of shaft. In general, pressing a non-guided slender shaft into the core (Figure 1) could, under certain circumstances, result in buckling deformation of the shaft and the conse- quent runout. The critical axial buckling force (F BUCKLING ) can be analytically determined with Euler’s formula. 16 However, in the presented case, plastic defor- mation was the only limiting parameter (F AM A X < F BUCKLING ), excluding the buckling in further steps of the study. In general, the stress state in the core should also be regarded as the limiting parameter for evaluation of the maximum-allowable joining force. However, in the presented case the shaft is machined from much softer material than the core and the stress state in the core is not critical. In all cases, the joining force F A is the most impor- tant process output that can be used to evaluate the feasi- bility of the press-fit assembly process. Therefore, in practice during the assembly operation, the joining force F A is measured, and all products are ejected, where the joining force F A at the final state of the press fit is not within the prescribed range F AM I N < F A < F AM A X . G. GANTAR et al.: OPTIMIZATION OF PRESS-FIT PROCESSES 208 Materiali in tehnologije / Materials and technology 55 (2021) 2, 207–212 Figure 2: - curves of materials 11SMnPb30 and CuZn39Pb3 Figure 1: Press-fit process 2.1 FE analysis An FE model was set up for the investigation of how the input parameter influences the joining force F A (Fig- ure 3). The model was defined as static (2D) axisym- metric. This model is preferred from the computational times’ point of view because the studied system is axisymmetrical and inertial forces during assembly or disassembly process can be neglected. For the discreti- zation of both parts, 3-node triangle and 4-node quadri- lateral axisymmetric linear elements were used. For both parts, an elastoplastic material model was used. A sur- face-to-surface contact was defined between the two parts. For this contact pair, the total normal contact force (F N ) was calculated at different lengths of engagement (L). The necessary assembly force was then calculated with Coulomb’s Law, where the coefficient of friction was approximated as μ = 0.075, as suggested in 17 for a steel-brass dry contact. Calculated joining force for selected combination of input dimensions (d shaft = 1.993 mm, D core1 =2m ma n d D core2 = 1.965 mm) is presented in Figure 4. The final engagement length (L MAX = 6,8 mm) is achieved when the shaft end reaches the stopping hole. From that point, the required assembly force increases rapidly. To validate the quality of the FE model, the predicted joining force F A was compared to experimental results; 30 samples were produced on the assembly press equipped with the force-measurement sensor. Their input dimensions were the same as those for numerical simula- tions. The following values joining force F A were mea- sured: average = 505 N, minimal = 475 N, maximal = 560 N. The FE model, therefore, predicted 14 % higher joining force F A , than the average value measured. Furthermore, the process window of the press-fit pro- cess was calculated by using the developed FE model and repeating the FE simulations with several different combinations of d shaft , and D core (Figure 5). Certain combinations d shaft – D core2 are leading to an insufficient joining force F, and others are leading to plastic deformation of the shaft (F A > F AM A X , which is calculated by Equation (1)). An acceptable combination of d shaft and D core2 in the middle represents the process window of the studied press-fit process. The upper part of the process window is unusable in industrial practice (it is impossible to produce the core with D core2 > D core1 = 2 mm with standard cost-effective machining processes) but this does not change/influence the approach for the robust design of press-fit processes proposed in this paper. G. GANTAR et al.: OPTIMIZATION OF PRESS-FIT PROCESSES Materiali in tehnologije / Materials and technology 55 (2021) 2, 207–212 209 Figure 5: The process window for the studied press-fit process Figure 3: FE model Figure 4: Predicted joining force Intuitively, it would be reasonable to set the men- tioned control input parameters exactly in the middle of the process window. But the question of how to evaluate and how to minimize the rejection rate (by selecting opti- mal combination of d shaft , D core1 and D core2 ) of the studied press-fit process remains. 2.2 Analysis of joining force To predict how the joining force F A varies during the press-fit process, the following steps were performed: Estimation of expected variations of input variables; Development of empirical model for predicting join- ing force F A influenced by E shaft , R p shaft , - shaft , d shaft , E core , R p core , - core , D core1 , D core2 and μ; Calculations of variations of joining force using Monte Carlo method. 2.2.1 Input parameters Each input parameter of the press-fit process should be considered as a probabilistic variable. In our study, the variations of the input parameters were not actually determined by measurements and experiments, but esti- mated according to prior experiences. In Table 2, the most relevant input parameters, their nominal values and expected variations are gathered. Table 2: Nominal values and expected variations of input variables. Input variable Mean value and expected variation SHAFT Yield stress (MPa) R p shaft = 350±40 Hardening exponent (1) n shaft = 0.16±0.02 Diameter of the shaft (mm) d shaft = 1.993±0.007 CORE Yield stress (MPa) R p core = 530±50 Hardening exponent (1) n core = 0.16±0.02 Guide diameter of the core (mm) D core 1 = 2±0.012 Functional diameter of the core (mm) D core 2 = 1.965±0.012 OTHER Coefficient of friction μ = 0.075±0.02 The slopes of - curves in the plastic region were approximated using the hardening law f = plastic n .I nTa- ble 2, hardening exponents for both materials are pre- sented. Expected variations of the material properties (R p shaft , n shaft , R p core , n core ) are based on the data previously gath- ered in different forming processes. 18 Experimental work, presented in 19 and 20 reports comparable variations of material properties. The expected variation of the diameter of the shaft d shaft was selected since the wires with tolerances h8 are commercially available and widely used in various in- dustrial applications. The expected variations of the di- ameters of the core D core 1 and D core 2 were selected due to the fact that such tolerances are achievable in state- of-the-art machining operations with reasonable costs. The expected variation of the friction coefficient μ is also based on previously gathered data in 18 . In the presented work, it was assumed that the varia- tions of all the input variables are normally distributed with standard deviations equal to one quarter of the ex- pected variations specified in Table 2. A part of the experimental matrix (6 out of 47 runs) can be seen in Table 3. According to the selected design of experiments, FE simulations were run for the predic- tion of joining force F A for different setting of input vari- ables (right column of Table 3). 2.2.2 Development of empirical model for predicting joining force F A using response surface methodology RSM is a method for the determination of the rela- tionships between several input parameters and one or more output parameters (also termed responses of the studied system) and is further described in 27 . Different designs of experiments can be used. We used a three- level Box-Behnken Design. The low and high levels of input variables were selected in such a way as to cover the area of input parameters, which was later used for optimization. The response function coefficients were determined by a standard method of least squares, which minimizes the sum of the squared deviations of fitted values. It was expected that the behaviour of the forming system is non-linear; therefore, a second-order polynomial func- tion was used. The fitness of the response function has been estimated using the Analysis Of Variance (ANOV A) technique as described in 21 . The R-squared G. GANTAR et al.: OPTIMIZATION OF PRESS-FIT PROCESSES 210 Materiali in tehnologije / Materials and technology 55 (2021) 2, 207–212 Table 3: Experimental design matrix and results of FE simulations Run Input variables Response R p core n core D core 1 D core 2 R p shaft n shaft d shaft F A 1 480 0.14 1.98 1.945 310 0.18 2.000 0.055 1003 2 480 0.14 2.02 1.985 390 0.18 1.986 0.055 35 3 480 0.18 2.02 1.945 310 0.14 2.000 0.095 1385 4 480 0.18 1.98 1.945 390 0.18 2.000 0.055 997 5 580 0.14 2.02 1.985 310 0.14 2.000 0.055 1142 47 530 0.16 2.00 1.93136 350 0.16 1.993 0.075 777 value of the model is 0.9991 and average relative devia- tion between predicted and calculated values is 2.54 %. 2.2.3 Calculations of variations of joining force using developed empirical model and Monte Carlo method A Monte Carlo simulation is a method to determine the probabilistic response of complex systems. The prin- ciple of this method is to use a random number generator to simulate the variations of the input variables. 22 Once the empirical model for joining force F A was obtained, using a RSM model, it was possible to use the Monte Carlo techniques to evaluate the variation of the response of the system (joining force F A ) due to varia- tions of the input parameters. The predicted variations of joining force F A for the nominal average values of all in- put parameters and their expected variations (gathered in Table 2) is presented in the upper part of Figure 6. In the lower part of Figure 6, the actual distribution of the measured joining force F A is presented. The mea- surements were performed for 5 months; 21800 test pieces were produced at this time. The actual dispersion of the joining force F A is similar to the predicted one. 2.3 Prediction of rejection rate and optimization of press-fit process The rejection rate can be evaluated from the probabil- ity chart for joining force F A . From the upper part of Fig- ure 6, it can be predicted that 98.6 % of the parts pro- duced would be within the required tolerance and the rejection rate would be 1.4 % (due to the plastic defor- mation of the shaft). As can be seen from the lower part of Figure 6, the measured level of the rejection rate was 1.5 %. Finally, the developed approach can be used for an optimization of the press-fit process. Assume that the material of shaft and core are selected and that wire for the shaft must be purchased in a standard dimension (d shaft = 2 mm h8). In this case, the two major input vari- ables that can be influenced and optimized are D core1 and D core2 (which are produced in the machining department of the company). While using an empirical model and Monte Carlo simulations varying of D core1 and D core2 (D core1 varies from 2 mm to 2.01 mm and D core2 from 1.965 mm to 1.97 mm), the calculated predicted rejec- tion rate can be easily presented in the 3D graph (Fig- ure 7). It is shown that by using the optimal combination of input parameters (D core1 = 2.01 mm and D core2 = 1.97 mm) the rejection rate can be reduced from 1.4 % to 0.2 %. 3 CONCLUSIONS In the research, the pistons of solenoid valves are studied, which are mass produced for press-fit assembly by many companies worldwide. An FE model was set up for the investigation of how the input parameters influ- ence the joining force F A (Figure 3). The necessary as- sembly force was then calculated by Coulomb’s Law, where the coefficient of friction was approximated as μ = 0.075 for a steel-brass dry contact. Our approach, which was a combination of FE calcu- lation for prediction of normal contact force and empiri- cal calculation of friction force with Coulomb’s Law, predicted a 14 % higher joining force than the average measured value. The results could be further improved G. GANTAR et al.: OPTIMIZATION OF PRESS-FIT PROCESSES Materiali in tehnologije / Materials and technology 55 (2021) 2, 207–212 211 Figure 7: Prediction of reject rate of studied assembly process Figure 6: Probability chart for joining force F A by repeating the calculations with a 14 % lower coeffi- cient of friction μ. Furthermore, the process window of the press-fit pro- cess was determined by repeating the FE simulations with several different combinations of core and shaft di- ameters. Afterwards, the variations of the shaft parameters (di- ameter, yield stress, hardening exponent), core parame- ters (guide and functional diameter, yield stress, harden- ing exponent) and friction coefficient were analysed. Based on the FE simulations, using 47 different input parameter variation results, the empirical model for pre- dicting joining force using RSM (Response Surface Methodology) method was obtained. The average rela- tive deviations between the predicted and calculated val- ues of the joining forces were 2.54 %. The model and Monte Carlo technique was used to evaluate the varia- tions of the joining force due to variations of the input parameters. A Monte Carlo simulation predicted that 98.6 % of the parts produced would be within the required toler- ance. Consequently, the rejection rate is 1.4 % (due to the plastic deformation of the shaft). The actual reject rate (obtained from the testing) was 1.5 %. In the study, only a rejection caused by a variation of input process parameters is evaluated. Rejections resulting for other reasons (failure of the tool, the wrong setting of the ma- chine, etc.) were not the subject of the presented study. Finally, the developed approach was used for the op- timization of the press-fit process. It was shown that by using the optimal combination of input parameters, the predicted reject rate can be reduced from 1.4 % to 0.2 %. In the future, the cost function should be integrated into the optimization procedure in order to optimize the studied press-fitting processes also from the economic point of view. In some cases, it is reasonable to increase the machining tolerances or use low-cost raw material with higher variations of properties, although the press-fitting process results in a higher rejection rate. Acknowledgement This work was supported by the Slovenian Research Agency – call title Promotion of employment of young PhDs in 2015, grant number 30955MD. 4 REFERENCES 1 S. Kleditzsch, B. Awiszus, M. Lätzer, E. Leidich, Steel-aluminum Knurled Interference Fits: Joining Process and Load Characteristics, Procedia Eng. (2014) 81, 1982–1987, doi:10.1016/j.proeng.2014. 10.268 2 F. Mahi, U. Dilthey, Joining of Metals, in Reference Module in Ma- terials Science and Materials Engineering, Elsevier 2015, doi:10.1016/B978-0-12-803581-8.03785-1 3 A. G. Razzell, S. B. Venkata Siva, P. S. Rama Sreekanth, Joining and Machining of Ceramic Matrix Composites, in Reference Module in Materials Science and Materials Engineering, Elsevier 2016, doi:10.1016/B978-0-12-803581-8.03915-1 4 P. Groche, S. Wohletz, M. Brenneis, C. Pabst, F. Resch, Joining by forming–Ar e v i e wo njoint mechanisms, applications and future trends, J. Mater. Process. Technol. 214 (2014) 10, 1972–1994, doi:10.1016/j.jmatprotec.2013.12.022 5 K. Martinsen, S. J. Hu, B. E. Carlson, Joining of dissimilar materials, CIRP Ann. – Manuf. Technol. 64 (2015) 2, 679–699, doi:10.1016/ j.cirp.2015.05.006 6 S. Kleditzsch, B. Awiszus, M. Lätzer, E. Leidich, Numerical and ana- lytical investigation of steel–aluminum knurled interference fits: Joining process and load characteristics, J. Mater. Process. Technol. 219 (2015), 286–294, doi:10.1016/j.jmatprotec.2014.12.019 7 R. Kiebach, K. Engelbrecht, K. Kwok, S. Molin, M. Søgaard, P. Niehoff, F. Schulze-Küppers, R. Kriegel, J. Kluge, P. Vang Hendriksen, Joining of ceramic Ba0.5Sr0.5Co0.8Fe0.2O3 mem- branes for oxygen production to high temperature alloys, J. Memb. Sci. 506 (2016), 11–21, doi:10.1016/j.memsci.2016.01.050 8 M. Pawlicki, T. Drenger, M. Pieszak, J. Borowski, Cold upset forg- ing joining of ultra-fine-grained aluminium and copper, J. Mater. Process. Technol. 223 (2015), 193–202, doi:10.1016/j.jmatprotec. 2015.04.004 9 T. N. Chakherlou, B. Abazadeh, Investigating clamping force varia- tions in Al2024-T3 interference fitted bolted joints under static and cyclic loading, Mater. Des. 37 (2012), 128–136, doi:10.1016/ j.matdes.2011.12.037 10 J. Mucha, Finite element modeling and simulating of thermo- mechanic stress in thermocompression bondings, Mater. Des. 30 (2009) 4, 1174–1182, doi:10.1016/j.matdes.2008.06.026 11 S. Salehghaffari, M. Panahipoor, M. Tajdari, Controlling the axial crushing of circular metal tubes using an expanding rigid ring press fitted on top of the structure, Int. J. Crashworthiness 15 (2010)3 , 251–264, doi:10.1080/13588260903209099 12 M. Bambach, A finite element framework for the evolution of bond strength in joining-by-forming processes, J. Mater. Process. Technol. 214 (2014) 10, 2156–2168, doi:10.1016/j.jmatprotec.2014. 03.015 13 M. Lorenzo, C. Blanco, J. C. P. Cerdán, Numerical Simulation and Analysis via FEM of the Assembly Process of a Press Fit by Shaft Axial Insertion, Springer 2013, 787–795, doi:10.1007/978-94-007- 4902-3_82 14 G. A. Pantazopoulos, A. I. Toulfatzis, Fracture Modes and Mechani- cal Characteristics of Machinable Brass Rods, Metallogr. Micro- struct. Anal. 1 (2012) 2, 106–114, doi:10.1007/s13632-012-0019-7 15 S. Bo`i~, D. [ircelj, Measuring of stress-strain behaviour of steel 1.0718 and aluminium alloy at different temperature range, Mach. Technol. Mater. 6 (2011), 36–40 16 R. C. Hibbeler, Statics and Mechanics of Materials, 4 th Edition, Pearson 2014 17 A. Van Beek, Advanced engineering design: Lifetime performance and reliability, TU Delft 2006 18 G. Gantar, K. Kuzman, Optimization of stamping processes aiming at maximal process stability, J. Mater. Process. Technol. 167 (2005) 2–3, 237–243, doi:10.1016/j.jmatprotec.2005.05.027 19 T. de Souza, B. F. Rolfe, Characterising material and process varia- tion effects on springback robustness for a semi-cylindrical sheet metal forming process, Int. J. Mech. Sci. 52 (2010) 12, 1756–1766, doi:10.1016/j.ijmecsci.2010.09.009 20 A. Michael, M. Scholting, E. Atzema, Characterisation and model- ling of the stochastic behaviour of deep drawing steels, VII Interna- tional Conference on Computational Plasticity COMPLAS VII E. Oñate and D. R. J. Owen (Eds) CIMNE, Barcelona, 2003, 1–20 21 R. H. Myers, D. C. Montgomery, C. M. Anderson-Cook, Response Surface Methodology: Process and Product Optimization Using De- signed Experiments, 3rd Editio, John Wiley & Sons 2009 22 R. Y. Rubinstein, D. P. Kroese, Simulation and the Monte Carlo Method, John Wiley & Sons, Inc., Hoboken 2007, doi:10.1002/ 9780470230381 G. GANTAR et al.: OPTIMIZATION OF PRESS-FIT PROCESSES 212 Materiali in tehnologije / Materials and technology 55 (2021) 2, 207–212