Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4, 1039-1054 2024 Publisher: Learning Gate DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate © 2024 by the authors; licensee Learning Gate * Correspondence: parinya.sr@mail.rmutk.ac.th Experimental analysis and machine learning with IoT monitor in two-way abrasive flow machine polishing on P20 mold components Theerapong Maneepen1, Parinya Srisattayakul2*, Narong Mungkung3, Wittawat Poonthong4, Tanapon Tamrongkunanan5 1Program in Engineering and Technology Management, Faculty of Engineering, Rajamangala University of Technology Krungthep, 10120, Thailand, 659041810089@mail.rmutk.ac.th (T.M.) 2Department of Industrial Engineering, Faculty of Engineering, Rajamangala University of Technology Krungthep, 10120, Thailand; parinya.sr@mail.rmutk.ac.th (P.S.) 3Faculty of Industrial Education and Technology, King Mongkut’s University of Technology Thonburi, 10140, Bangkok, Thailand; narong_kmutt@hotmail.com (N.M.) 4King Mongkut’s University of Technology Thonburi, Thailand; (KMUTT), 10140; poonthong.golf2538@gmail.com (W.P.) 5King Mongkut University of Technolgy Thonburi, Thailand; nathwut.kow@kmutt.ac.th (T.T.) Abstract: The purpose of this paper was to reveal the effects of pressure and time on the surface roughness (Ra) and compare the experimental design (Factorial Regression) and machine learning (ML) of steel mold workpieces. Methodology began with turning, and fine sandpaper P180 to P1200 and measured with an initial value of Ra compared with the final value of Ra at the end of the two-way prototype AFM process. The process parameters are as follows; abrasive particle size (alumina; Al2O3) 5.0 μm in Silicone Oil (concentration 50% by weight) at pressures (p) of 10 bar and 20 bar, processing time (t) 5, 15, and 25 min, specimens P20 Mold steel. The experimental results show that under these conditions. The average surface roughness of the specimens differed from the initial value by Ra 0.034 to 0.021 μm, with delta values ranging from 0.011 μm to 0.005 μm. The results showed a smoother profile between before and after polishing. The approach to the topic is DOE and ML, and the theoretical or subject scope of the paper is Statistical and AI. The original value of the paper is applied to the ESP32 Arduino to control and display critical parameters. A General Factorial Regression statistical value of 76.39% is acceptable. A pressure factor of 20 bars and a time of 25 minutes gives the best effect on surface roughness. ML assisted in predicting the Surface Roughness for optimization based on the experiment. Keywords: Abrasive flow machining (AFM), Factorial regression, Machine learning (ML), Polishing, Surface roughness (SR). 1. Introduction The systematic investigation of factors influencing the abrasive flow machining (AFM) process and its outcomes: AFM typically involves investigating the effects of two critical factors or process parameters on the performance metrics of interest. These factors can include Pressure applied to the abrasive media, Media flow rate, Abrasive particle concentration in the media, Cycle time, or number of cycles. The goal is to understand how varying these key input factors impacts the output responses such as Surface finish quality (e.g. roughness), Material removal rates, and Other performance metrics like deburring capability. By systematically studying the effects of changing pressures, flow rates, abrasive concentrations, cycle times, etc., researchers and manufacturers can optimize the AFM process to achieve the desired surface characteristics, material removal rates, and overall process efficiency on complex internal geometries and workpiece materials. https://orcid.org/0000-0003-3991-9850 https://orcid.org/0000-0002-5627-548X https://orcid.org/0000-0002-4868-4654 https://orcid.org/0000-0001-7413-8502 https://orcid.org/0000-0003-1157-4600 1040 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate The formation of a dispersive particle phase under pressure facilitates particle-surface interactions that govern the material removal mechanisms in AFM. Experiments have demonstrated widely varying removal rates based on the input factors investigated. In essence, understanding the relationship between critical AFM input parameters and output performance is crucial for process control and achieving the intended surface quality/geometry specifications (J.J. Hann, P.S. Steif, 1998) [1]. AFM was polishing the contemporary and small removal surfaces with the flow of the slurry medium. The abrasive flow machining process provides a high level of surface finish. It closes tolerances with an economically acceptable rate of surface generation for a wide range of industrial components (Rajendra K. Jain, Vijay K. Jain, P.M. Dixit, 1999) [2]. AFM was the high-end cutting process with deburring, radius, polish, removal recast layer, and made compressive residual stresses. This process was widely used in the 1960s and was interesting in consistency production and prediction of results output. AMS process was developed by Extrude Hone Co., ltd. in 1996. Process Parameters mention the orbital amplitude to find the material removal rate in higher amplitudes yielding, higher material removal rates but the orbital amplitude must not be bigger than the minimum internal feature of the workpiece. To find the material removal rate must focus on both the oscillation speed and the orbital amplitude and not get the effect from the geometrical dimension of the workpiece between 400 to 1200 RPM (Jun Wang, et all, 1999) [3]. High-precision abrasive flow machining has two sub-systems: a high-viscosity media; the range of between 150- 1,000,000 centipoise a viscous-elastic-plastic media (a semisolid polymer composition) and a low- viscosity media; 1-50 centipoise was a liquid abrasive slurry involve to abrasive uspended or slurried in fluid media by cutting fluids of honing fluids consisted of a thixotropic slurry plus a rheological additive and finely divided abrasive particles incorporate therein with mixed pressure and flow between 4,000 psi. V.K. Jain, and S.G. Adsul, 2000 [4] this research study on the effects of parameters of the different processes of AFM such as the number of cycles, the concentration of abrasive, abrasive mesh size, and media flow speed. Study in material removal and surface finish. To find the dominant parameters such as medium percentage concentration, abrasive media mesh size, cycle time or machining time, and speed of media flow. Test with Brass and Aluminum by comparing experimental and theoretical of workpiece surface with Scanning electron microscopy: SEM, experiment on Lath by setup on turret steady rests so that Parameter planning is shown in Table 1 (Geoffrey Boothroyd, 1996) [5]. Neelesh K. Jain, V.K. Jain, Kalyanmoy Deb, 2007, “Optimization of process parameters of the mechanical type advanced machining processes using genetic algorithms” to study between 4 processes; USM, AJM, WJM, AWJM (Neelesh K. Jain, V.K. Jain, Kalyanmoy Deb, 2007) [6]. AFM and Stereolithography (SL), to minimize the time to develop a finished prototype, simulation, and neural network. Results indicated that media pressure, grit size, percentage concentration, reduction ratio, and build orientation were significant [7], [8], [9]. The results of Rotating Abrasive flow finishing (R-AFF) show that the rotational speed of the workpiece has a significant effect on delta Ra [10]. The material removal rate model and the maximum error between theoretical and experimental values is 13.1% Cconsistent with the experimental results of S.M. Basha, et. Al., [13, 14]. Demonstrates good production efficiency. There are many types of materials and shapes (complex holes [21, 32]) used in past experiments such as Mild steel [15, 18], AISI D2 [22], Bevel gear [25], and SLM [31, 33], ABS LM and PLA parts [35]. Simulation with models, such as NN, models, to predict polishing results [16, 24, 27, 30, 34]. Using a magnetic field to help with polishing, such as [17, 23, 28]. Use ultrasonication in experiments such as [26]. As well as using the rotation of the workpiece while experimenting, including [19, 20]. Wear [29]. The development of a precise and convenient two-way abrasive flow machining setup for achieving high surface smoothness: The goal was to develop a specialized machine for polishing workpiece surfaces to a high degree of smoothness through the unidirectional flow concept of abrasive media. This involved relying on knowledge and skills to create and optimize such a machine setup. The application differed from previous research by focusing on the use of a two-way flow approach. Experiments were conducted using an abrasive media consisting of aluminum oxide mixed with silicone oil. Key aspects 1041 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate included integrating inspection sensors for monitoring parameters like pressure or flow rate, which would be further developed. Due to the high construction costs, a prototype underwent development stages. The effects of the clearance between the specimen and tooling on the surface roughness (Ra) of P20 mold steel were investigated. Factorial regression statistical principles were applied to analyze the results and achieve the desired surface roughness specifications. Additionally, experimental findings were compared with machine learning (ML) models to validate the approach. The overall objective was to obtain insights that would aid in further improving and differentiating the prototype unidirectional abrasive flow machining setup for precise surface polishing applications. Table 1. Improvement in Process parameters of Abrasive flow machining (Nitin Dixit, et. al., 2021) [11, 12] 2. Materials and Methodology Figure 1. Schematic of the Two-way prototyping AFM. 2.1. Experimental Set-Up The experimental setup of the power plant is driven by bi-directional hydraulics. As shown in Figure 2, it has been designed and developed to be able to store the abrasive in each cylinder from two cylinders so that it can work in two directions to continuously polish the workpiece. Hold the workpiece with a C- Clamp, there is a seal to withstand the pressure of the polishing system. Figure 2. Shows the prototyping AFM machine. 1042 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate 2.2. Workpiece and Medium Figure 3. The workpiece, AFM nozzle, and abrasive media (Al2O3+Silicon Oil). 2.3. Experimental Procedure Steps followed: A step-by-step guide on how might conduct experimental two-way AFM: 1. Define Objectives: Trying to optimize surface finish 2. Identify Factors: Select two key factors to study. These could be, pressure (p) and cycle/processing time (t). 3. Define Factor Levels and Experimental Design: Determine the levels for each factor. For instance, choose two levels of pressure (10, 20 bar) and three levels of cycle time (t) (5, 15, and 25 min.) Use a factorial experimental design to systematically combine the different levels of the two factors. For a full factorial design, test all possible combinations of factor levels (Minitab 19). 4. Conduct Experiments: Implement AFM experiments for each combination of factor levels. Ensure that the experiments are conducted under controlled and consistent conditions. Prepare the workpiece surface with sandpaper from P40 to P1200. To measure initial surface roughness (SR); Ra micron before AFM process. To polish by AFM prototype within cycle time ranges of 5, 15, and 25 minutes consequently. 5. Data Collection: Measure and record the responses of interest after each experiment. This could involve quantifying surface roughness, or assessing other relevant performance indicators. To measure surface roughness after being polished by AFM (Final SR). To measure the raw profile and modified profile of the workpiece with the surface roughness machine (Olympus). 6. Statistical Analysis: An analysis of variance (ANOVA), to analyze the data. This will help identify significant main effects and interactions between the two factors. Here are some common statistical analyses used in AFM: 1) Regression Analysis: Regression analysis is used to model the relationship between independent variables (e.g., process parameters) and dependent variables (e.g., surface roughness). By performing regression analysis, researchers can quantify the effect of each factor on the response variable and develop predictive models for AFM processes. Regression analysis can help optimize process parameters and predict the expected outcome of AFM based on given input variables. 2) Design of Experiments (DOE): DOE involves planning and conducting a set of well-designed experiments to evaluate the effects of different factors and their interactions. By using statistical 1043 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate analyses like ANOVA and regression analysis, researchers can identify significant factors, optimize process parameters, and understand their impact on AFM performance. Applying statistical methods in AFM research helps in data-driven decision-making and the continuous improvement of the machining process. 7. Validation: Validate your findings by conducting additional experiments or using a separate dataset. This ensures the reliability and generalizability of results. 8. Interpretation and Conclusion: Interpret the results in the context of objectives. Conclude the effects of the selected factors on the AFM process. 9. Documentation and Reporting: Document the experimental setup, procedures, and results thoroughly. Prepare a comprehensive report or presentation summarizing findings. By systematically varying two key factors and observing their effects on the AFM process, can gain valuable insights into the optimization and performance characteristics of AFM for specific objectives. 3. Results and Discussion 3.1. Roughness and Profile Detection The results of the experiment were as follows: Displays Figure 4 the relationship between surface roughness; (micron) and polishing with AFM prototype at 10 bar, 20 bar, and interval time from 0 to 5, 15, and 25 minutes step by step consequently. Figure 3. The surface roughness 3D polishing. 1044 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate Table 2. Run order and results with MiniTab 19. Figure 4. Difference in surface roughness before and after polishing with the AFM prototype. 1045 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate Table 3. Results delta SR, time 5, 15, and 25 minutes. 10 bar 20 bar Time Delta SR Time Delta SR 5 0.005 5 0.006 0.006 0.007 0.007 0.008 15 0.005 15 0.007 0.006 0.008 0.008 0.010 25 0.004 25 0.007 0.005 0.009 0.007 0.011 Table 3 displays the relationship between surface roughness; (micron) and polishing with AFM prototype at 10 bar, 20 bar, and interval time 0 to 5, 15, and 25 minutes. Figure 5. Displays the Initial profile and final profile of the workpiece that was polished 25 minutes. 1046 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate Figure 6. ESP32 arduino controller (SWU). This ESP32 Arduino control kit was developed to control the operation of the control system, and sensors to measure pressure, temperature, etc., as well as collect results, analyze, monitor, and summarize the results of the prototype. 3.2. Statistical Analysis and Discussion Design of Experiments (DOE) is a systematic approach used to optimize process parameters in Abrasive Flow Machining. By conducting a DOE, researchers can identify the optimal combination of factors that will yield the desired surface finish, material removal rates, or other performance measures. 1047 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate Figure 7. Factor information and ANOVA. The results of the analysis are shown as follows. The analysis of the variance table shows that the P-value is less than 0.05, indicating that Pressure and Time significantly affect the response. The correlation coefficient (r2) can be found in the regression analysis chapter where R-Sq(adj) = 76.39%, less than 70% is considered acceptable. 1048 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate Figure 8. Pareto chart of the standard effects. The relationship equation between factor and response is. Graph Pareto, notice that the bar graph of factors A, and B is past the critical line, indicating that they all have a significant effect on the results. The error variance is uniform. Independence is a characteristic of a good control plan. Main effect graph; factors, pressure, and time affect SR. Where a pressure factor of 20 bar and a time of 25 minutes give the best effect on surface smoothness. Figure 9. Regression equation for general factorial regression. 1049 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate Figure10. Residual Plots & Main Effects Plot for ∆ SR. Relationship equation of factors and responses error analysis Check the normal distribution of the error values. The data has a normal distribution of error values. 1050 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate Figure 11. Display linear model and related graphs. Figure 11 illustrates the pressure and time data and Linear model results for SR. The delta Ra value tends to increase in line with the increase in pressure. 3.3. Machine Learning Prediction Machine Learning Prediction with the Rapid Miner program to make predictions about expected experiment results from actual experimental results to compare with predictions. To make the selection of various parameters more accurate. The linear regression function is used with a flow chart and select attributes process Figure 12. The sub-work steps that turn raw data into knowledge. It consists of the following steps: Data Cleaning, Data Integration, Data Selection, Data Transformation, Data Mining, Pattern Evaluation, and Knowledge representation. Input data from Minitab 19, and the result shows the prediction (Delta SR) Figure 13-14 and apply model, parameters; pressure, time, delta SR, and Cross-Validation as shows in Figure 15. 1051 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate Figure 12. Flow chart of a standard SVR model and select attributes process. Figure 13. Input data from Minitab 19 and Prediction (delta SR). 1052 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate Figure 14. Input data from Minitab 19 and Prediction (Delta SR). (Cont.). Figure15. Apply model; pressure, time, delta SR, and cross-validation. Performance Vector; Root mean squared error: 0.001 ± 0.001 (micro average: 0.001 ± 0.000) 1053 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate 4. Conclusions The experiment of workpiece polishing mold steel. Steps follow to prepare the workpiece surface, measure initial surface roughness before and after the AFM process, and profile of surface. The result is that pressure is 10 bar, 20 bar consequently and time range 5, 15, and 25 minutes: Pressure 10 bar; the difference before and after of average SR, trend to decrease. The difference; The difference; is delta 0.006 to 0.011 µm. Pressure 20 bar; the difference before and after of average SR, trend to decrease. The value of before and after of average surface roughness middle and trend to decrease surface roughness value (Ra) from 0.034 to 0.021 μm. More pressure will affect the surface roughness that able to produce a difference in surface roughness that is greater as a result, a smooth surface can be obtained in a faster time and closer to the required surface roughness SR. That increases with the number of cycles and extrusion pressure, whereas it decreases with the increase in abrasive mesh size. • Statistics Factorial Regression obtained from experimental results It was found that there was a significant difference at 0.05 of the pressure, and time, especially at a pressure of 20 bar and a time of 25 minutes, gave the best results. • Machine Learning Prediction with the Rapid Miner program to make predictions about expected experiment results from actual experimental results to compare with predictions. To make the selection of various parameters more accurate. Performance Vector; Root mean squared error: 0.001 ± 0.001 • The results of AFM demonstrate its effectiveness in achieving high-quality surface finishes and the overall surface integrity of the workpiece. Surface roughness measurements provide quantitative data on the achieved surface quality and machining efficiency. It is the prototype to drive the abrasive media to achieve finer control of the variables to develop more precision in the subsequent generation development and develop a polishing control system that uses intelligence to work with a sensor system using AI technology. This can be further applied to industrial work for internal and precision processing prospects. This is to provide basic information for further development of the prototype AFM machine for increased efficiency for use in the country and region in the future. Acknowledgments: Thank you, Department of Industrial Engineering, Faculty of Engineering, Rajamangala University of Technology Krungthep, Thailand for Minitab 19, the surface roughness machine (Olympus), and other comments. Diamond Polishing Co., Ltd. And other people in the Mold and Die industry for suggestions about techniques and approaches to improve polishing machines. Copyright: © 2024 by the authors. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). References [1] Hann J.J., Steif P.S., Abrasive wear due to the slow flow of a concentrated suspension, Wear, 219 (1998), 177-183. [2] Rajendra K. Jain, Vijay K. Jain, P.M. Dixit, Modeling of material removal and surface roughness in abrasive flow machining process, Int. J. Mach. Tools Manuf., 39 (1999), 1903-1923. [3] Jun W., William S., Liangchi Z., Abrasive technology: current development and applications I, World Scientific, Singapore, (1999), 269-317. [4] Jain V.K., Adsul S.G., Experimental investigations into abrasive flow machining (AFM), Machine tools and manufacturing, 40 (2000), pp. 1003-1021. [5] Geoffrey B., Handbook of Manufacturing Engineering, Island, (1996), 769-801. [6] Neelesh K.J., Jain V.K., Kalyanmoy D., Optimization of process parameters of the mechanical type advanced machining processes using genetic algorithms, Int. J. Mach. Tools Manuf., 47 (2007), 900-919. https://creativecommons.org/licenses/by/4.0/ 1054 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 8, No. 4: 1039-1054, 2024 DOI: 10.55214/25768484.v8i4.1480 © 2024 by the authors; licensee Learning Gate [7] Robert E., Williams, Vicki L.M., Abrasive flow finishing of stereolithography prototypes, Rapid Prototyp. Rep., 4 (1998), 56-67. [8] Rajendra K.J., and Vijay K. J., Simulation of Surface generated in abrasive flow machining process, Robot. Comput. Integr. Manuf., 15 (1999), 403-412. [9] R.K. Jain, V.K. Jain, P.K. Kalra, Modeling of abrasive flow machining process: a neural network approach, Wear 231 (1999), pp. 242-248. [10] Mamilla R.S., et. al., Experimental investigations into rotating workpiece abrasive flow machining, Wear, 267(2009), 43-51. [11] Kumar S.S., Hiremath S.S., A Review on Abrasive Flow Machining (AFM), Procedia Technology, 25(2016), 1297- 1304. [12] Nitin D., Varun S., Pradeep K., Research trends in abrasive flow machining: A systematic review, J. Manuf. Process., 64 (2021), 1434-1461. [13] Basha S.M., et al., Development and performance evaluation of galactomannan polymer based abrasive medium to finish atomic diffusion additively manufactured pure copper using abrasive flow finishing, Addit. Manuf., 61 (2023), 103290. [14] Xiaoxing D., et al., Fine finishing of internal surfaces using cassava starch medium, J. Mater. Process. Technol., 315 (2023), 117918. [15] Rajendra K.J., Vijay K.J., Simulation of surface generated in abrasive flow machining process, Robotics and computer integrated manufacturing, 15 (1999), 403-412. [16] Jain R.K., Jain V.K., Optimum selection of machining conditions in abrasive flow machining using neural network, J. Mater. Process. Technol., 108 (2000), 62-67. [17] Sehijpal S., Shan H.S., Kumar P., Wear behavior of materials in magnetically assisted abrasive flow machining, J. Mater. Process. Technol., 128 (2002), 155-161. [18] Jeong-Du K., Kyung-Duk K., Deburring of burrs in spring collects by abrasive flow machining, J. Adv. Manuf. Technol., 24 (2004), 469-473. [19] Ravi Sankar M., et al., Rotational abrasive flow finishing (R-AFF) process and its effects on finished surface topography, Int. J. Mach. Tools Manuf., 50 (2010), 637-650. [20] Ravi S.M., et al., Rheological characterization of styrene- butadiene based medium and its finishing performance using rotational abrasive flow finishing process, Int. J. Mach. Tools Manuf., 51 (2011), 947-957. [21] A-Cheng W., et al., Uniform surface polished method of complex holes in abrasive flow machining, rans. Nonferrous Met. Soc. China., 19 (2009), 250-257. [22] Kenda J., et al., Surface Integrity in Abrasive Flow Machining of Hardened Tool Steel AISI D2, Procedia Engineering., 19 (2011), 172-177. [23] Kamble P.D., et al., Use of magneto abrasive flow machining to increase material removal rate and surface finish, Mechanical, Automobile & Production Engineering, 2 (2012), 249-262. [24] Uhlmann E., et al., Development of a material model for visco- elastic abrasive medium in abrasive flow machining, Procedia CIRP., 8 (2013), 351-356. [25] Venkatesh G., et al., Finishing of bevel gears using abrasive flow machining, Procedia Engineering, 97 (2014), 320- 328. [26] Venkatesh G., et al., On ultrasonic assisted abrasive flow finishing of bevel gears, Int. J. Mach. Tools Manuf., 89 (2015), 29-38. [27] Uhlmann E., et al., CFD simulation of the abrasive flow machining process, Procedia CIRP, 31 (2015), 209-214. [28] Satish K., et al., Nanofinishing of freeform surfaces (knee joint implant) by rotational-magnetorheological abrasive flow finishing (R-MRAFF) process, Precision Engineering, 42 (2015), 165-178. [29] Tina B., et al., Wear of abrasive media and its effect on abrasive flow machining results, Wear, 342-343 (2015), 44-51. [30] Uhlmann E., et al., A pragmatic modeling approach in abrasive flow machining for complex-shaped automotive components, Procedia CIRP, 46 (2016), 51-54. [31] Duval-Chaneac M.S., et al., Characterization of maraging steel 300 internal surface created by selective laser melting (SLM) after abrasive flow machining (AFM), Procedia CIRP, 77 (2018), 359-362. [32] A-Cheng W., et al., A study on the abrasive gels and the application of abrasive flow machining in complex-hole polishing, Procedia CIRP, 68 (2018), 523-528. [33] Sangil Han, et al., Surface integrity in abrasive flow machining (AFM) of internal channels created by selective laser melting (SLM) in different building directions, Procedia CIRP, 87 (2020), 315-320. [34] Vipin K.S., Modeling and analysis of a novel rotational magnetorheological abrasive flow finishing process, Int. J. Lightweight Mater. Manuf., 4 (2021), 290-301. [35] Dixit N., Sharma V., Kumar P., Experimental investigations into abrasive flow machining (AFM) of 3D printed ABS and PLA parts, Rapid Prototyp. J., 28(2022), 161-174.