On the estimation of tool-wear for milling operations based on multisensorial

Transcription

1 Available online at Procedia CIRP 8 (2013 ) th CIRP Conference on Modelling of Machining Operations (CIRP CMMO) On the estimation of tool-wear for milling operations based on multisensorial data C. Doukas a, P. Stavropoulos a, A. Papacharalampopoulos a, P. Foteinopoulos a, E. Vasiliadis a & G. Chryssolouris a * Abstract a Laboratory for Manufacturing Systems and Automation, Department of Mechanical Engineering and Aeronautics, University of Patras, Patras, 26500, Greece * Corresponding author. Tel.: ; fax: address: xrisol@lms.mech.upatras.gr. This paper shows the results of a preliminary experimental investigation on tool-wear in end milling. Spindle torque and vibration signals were recorded during the process. A correlation between measured signals and tool-wear was attempted. The preliminary results showed a promising application laying the foundation for an experimental modelling approach of the toolwear phenomenon. Power consumption, as depicted from the current draw signal, can be associated with the sustainability evaluation of the milling operation, due to their directly correlation to the toolwear level The The Authors. Authors. Published Published by Elsevier by Elsevier B.V. B.V. Open access under CC BY-NC-ND license. Selection and and/or peer-review peer-review under responsibility under responsibility of The International of The International Scientific Committee Scientific of the Committee 14th CIRP of Conference the 14th on CIRP Modeling Conference of Machining on Operations Modelling in of the Machining person of Operations" the Conference in Chair the person Prof. Luca of the Settineri Conference Chair Prof. Luca Settineri Keywords: Machining; Tool-wear; Sensors; Online monitoring 1. Introduction Successful monitoring [1, 2] of the wear during machining operations may provide significant benefits in terms of cost and part quality. It is reported [3] that by using wear sensors, a tool cost saving of up to 40% may be reached. Extensive research has been performed for tool condition monitoring in milling operations. In these methods, various signals [4] such as cutting force, acoustic emission (AE), machine vibration, motor current have been used in monitoring tool condition [5-13].Cutting force measurement requires installation of a force-torque sensor. This has a negative influence on the dynamics and stiffness of the machine, while at the same time the cost of such a device is considerably high. Utilizing a reflective single-beam laser system, the tool conditions can be monitored by constantly focusing a laser beam of a 50um spot size on the cutter and at the same time, by directing the beam reflection towards a receiver. The drawback of this method is that of hidden areas; for instance, when the laser beam is focused on milling pockets or grooves, just above the tool, the measurements are false. A rather usual approach to tool wear level identifications is the monitoring of power consumption. Study [14] claims that measuring the power consumption of a motor can be easily carried out by a watt meter without additional sensing devices placed on the machine. In [16], there is a description of the approaches followed for the generation of tool condition monitoring. It is apparent that in order for an accurate interpretation of the information produced to be provided, a high level of signal processing and analysis is required. The signal delivered by the sensor has to be processed for the detection of disturbances [16]. The simplest method is the use of a rigid threshold. If the latter is crossed by the signal, due to some process change affecting the signal, collision or tool breakage can be detected. The use of a dynamic threshold may be more appropriate in some cases [17]. The most promising application is that of the of accelerometers measuring the vibrations generated by the process, in conjunction with acoustic emission (AE) measurements originated from the deformation of the work piece material and are capable of capturing the cutting operation even of the very small cutting forces The Authors. Published by Elsevier B.V. Open access under CC BY-NC-ND license. Selection and peer-review under responsibility of The International Scientific Committee of the 14th CIRP Conference on Modeling of Machining Operations in the person of the Conference Chair Prof. Luca Settineri doi: /j.procir

2 416 C. Doukas et al. / Procedia CIRP 8 ( 2013 ) These methods have been claimed to work well for discrete events, such as tool breakage but are often more difficult to employ in tool wear monitoring. Tool-wear is considered being the major issue on the operation of intelligent manufacturu ing systems and thus, significant theoretical work of the wear behaviour has been performed, including sensor based force-tool wear models [18] and/or geometric (spindle/workpiece displacement) based algorithms [19]. The multi sensorial systems are of importance to intelligent manufacturing systems and as a result, several approaches are considered for the sensor integration [20]. Current practice in industrial environments implies the usage of the very first production parts to evaluate and fine tune the production process. This approach not only cost a significant amount of time, but also potentially contains significant amount of errors and reduces the suitability of the process in terms of Profit, People and Pollution. On the other hand machining processes constitute the major manufacturing activity and therefore, it becomes nowadays more and more important to develop technologies and techniques to achieve not only environmentally benign production, but in order to improve the sustainability of the process, to achieve the best possible result even from the first production part. The current study makes a preliminary experimental investigation on the correlation of tool-wear with spindle torque through spindle current and/or accelerometer signals from vibrations during milling operations. The overall goal is to estimate and model, in next studies, utilizing signal readings the toolwear mechanism. To achieve the above, a regression model will be developed to predict tool wear level based on experimental data. The coefficients for the regression model can be statistical processing software, in conjunction with design of experiments methods to minimize the experimental time and effort required. The process variables with the highest effect on tool wear include cutting speed, feed rate and cutting depth, while the least mean square fit method can been used for the determination of regression coefficients. 2. Experimental setup 2.1. Monitoring equipment A PC based data acquisition system has been used in the current study comprising an eight-channel dynamic signal acquisition module and a dedicated analog-todigital converter per channel, for maximum device throughput and higher multichannel accuracy. The acceleration sensor used is a tri-axial accelerometer with effective frequency range, up to 7 khz and sensitivity of 100mV/g. The current measurement is performed using a power analyser and inductive clamps (Fig.1). Fig. 1. Experimental setup 2.2. Tool and material selection The experimental setup used in this study is orientated towards the face milling operation of CGI 450 (Table 1) cast iron plates, using a 50mm diameter face mill head with indexable carbide inserts. Pre-drilled holes on the surface of the stock material are to accelerate the tool wear. Estimated tool-life, according to the tool manufacturer s data sheets, is calculated approximately 2 hours, under proposed machining variables. The workpiece used was 250x250x50mm, as shown in Fig. 2 Table 1. CGI450 chemical composition and mechanical properties CGI 450 Pearlite C SI CE Mn > S Mg CeMM Cu Sn <0.022 < Ultimate Tens. Strength Yield strength Elastic Modulus Mpa Mpa Gpa Elongation % Density Brinell Hardness g/cc BHN Pre-drilled holes on the surface of the stock material are to accelerate the tool wear. Estimated tool-life, according to the tool manufacturer s data sheets, is calculated approximately 2 hours, under proposed machining variables. The workpiece used was 250x250x50mm, as shown in Fig. 2

3 C. Doukas et al. / Procedia CIRP 8 ( 2013 ) Machining strategy A 3axis CNC knee mill, operating a spindle of 5Hp is being used for machining and it is capable of reaching approximately 3800 RPM. In order for the workpiece to be securely positioned on the machine table, an adaptor plate has been designed and manufactured, while also allowing the positioning of the acceleration sensor as close to the machining area as possible. Straight cutting passes have been performed, alongside the Y-axis of the machine, to minimize effects of feed direction changes. A cutting depth of 0.5mm has been used along with a feedrate of 1000mm/min. The experiment has been repeated at the spindle speed of 1350rpm and 2700rpm, to investigate the effect of cutting speed on the wear level (Table 2). Every 15,5 min the process is paused and the inserts are removed and inspected under an optical microscope for the measurement of the tool wear level. 3. Results Machining of the CGI plate has been performed, taking 6 overlapping by 50% (25mm) passes along the Y-axis of the workpiece. The target of the toolwear level was set to 150 microns for flank wear. Fig. 3. Flank wear (5x magnification) Microscope measurements Table 2. Setup variables Variable S=1350 S=2700 Cutting Speed 210m/min 420m/min Feed Rate 1000mm/min Depth of cut 0,5mm Feed per tooth 0,15 0,15 The Accelerometer Signals and Spindle current are monitored at the beginning, middle and end of each pass for a period of 10sec to minimize the amount of data collected and enable faster post processing. The tool-wear level is assessed by optical microscopy measurements of each full set of inserts (5 inserts) and calculating the arithmetic mean of flank wear value. Each full experiment was contacted twice, resulting in identical results both for toolwear and signal values. Worn inserts were removed from the tool head and examined under a camera, equipped with an optical microscope. Pictures from the sides of the tool were acquired using the same magnification level. A reference mark was added to each picture for measurement purposes. In Fig. 3, a microscopy photo of the flank wear is presented. In Fig.4 and 5 the tool wear development over time is presented for both high and low cutting speeds. flank wear [ m] Toolwear vs. time, cutting speed: 210 m/min Fig. 4. Tool wear versus time (Cutting Speed=210m/min)

4 418 C. Doukas et al. / Procedia CIRP 8 ( 2013 ) Toolwear vs. time, cutting speed: 420 m/min Spindle electrical current measurement, time domain, cutting speed: 420 m/min flank wear [ m] Elec. Current electr. current [A] Fig. 5. Tool wear versus time (Cutting Speed=420m/min) 3.2. Spindle Measurements The plot of the electrical current versus time (Fig. 6) during the full length of the experiment is presented. The signals increase over the course of the experiments, by a small rate till the middle of them indicating a plateau in the toolwear development and then increase rapidly till the end of the experiment, during which wear level of 150microns is detected. The same tendency is observed when the cutting speed is increased but significantly amplified (Fig. 7). Elec. Current Spindle electrical current measurement, time domain, cutting speed: 210 m/min 2.9 electr. current [A] Fig. 7. Electrical current draw versus time (CS=420m/min) 3.3. Accelerometer Measurements An accelerometer signal, due to the machine tool vibrations, is recorded during the cutting process. It has been analysed in the frequency domain. For the frequency domain analysis (Fig. 8 and 9), two bands were defined: the high and the low ones. The indicator in the graphs is the high band max value over that of the low band (HoL index). The threshold separating the Low band from the High band is set to 2 khz for low cutting speed and 4 khz for high cutting speed. The signal graphs (Fig. 8 and 9) show a rapid increase as the toolwear reaches higher levels. High Band max over Low Band max Acceleration Measurement, Frequency Domain, Cutting Speed: 210 m/min HoL index Fig. 6. Electrical current draw versus time (CS=210m/min) Fig. 8. Frequency domain: Primary Filter-Bank (wavelet) Analysis. (Threshold separating LB from HB: 2000 Hz)

5 C. Doukas et al. / Procedia CIRP 8 ( 2013 ) High Band max over Low Band max Acceleration Measurement, Frequency Domain, Cutting Speed: 420 m/min HoL index Fig. 9. Frequency domain: Primary Filter-Bank (wavelet) Analysis. (Threshold separating LB from HB: 4000 Hz) 4. Discussion The acquired signal could somehow be correlated with the tool wear levels over the course of the experiment (Fig. 10). The electrical power consumption rises initially, reaches a plateau till the tool wear has higher values and from there on, the consumption rises again rather rapidly. The accelerometer signals follow somewhat the findings of the electrical current measurement. The frequency domain analysis clearly depicts the increase of the signal value when the toolwear reaches higher levels. flank wear [ m] Toolwear vs. filtered current, cutting speed: 210 m/min filtered current Fig. 10. Tool wear versus electrical current. (A low-pass filter used to cut off sudden peaks in the electrical current data) In order to achieve real adaptive manufacturing, a clear description of the manufacturing process is needed, including the tool and workpiece interaction. Similarly to enable real-time condition monitoring and control during the actual manufacturing process, smooth integration of tool wear modelling approach to the real machine tool and machining process is needed. Both acceleration and current draw signals can be processed in real time, enabling efficient usage of cutting tool and improving production performance. 5. Conclusions Microscopy measurements and photos of tool worn inserts have been taken during face milling operations. Machine tool motor current and signals from an accelerometer mounted on the machine tool have been also recorded during machining. The RMS value of the electrical current signal could be used for the assessment of the wear level, whilst for the accelerometer signals; a frequency domain analysis is required for a useful indication of tool wear levels. Further investigations are needed to verify these initially promising results and accurately model the tool wear mechanism. Acknowledgements The work reported in this paper was partially supported by CEC / FP7 NMP-ICT Programme, "The Foundation for the Smart Factory of the Future-FoFdation, (FP NMP-ICT-FoF ). References [1] Chryssolouris, G., 2006, Manufacturing systems theory and practice 2nd ed. New York: Springer-Verlag. [2] Kurada S., Bradley C., 1997, A review of machine vision sensors for tool condition monitoring, Computers in Industry 34, p. 55. [3] Lim G., 1995, Tool-wear monitoring in machine turning, Journal Materials Processing Technology 51, p. 25. [4] Lee, D. E., Hwang, I., Valente, C. M. O., Oliveira, J. F. G., Dornfeld, D. A., 2006, Precision manufacturing process monitoring with acoustic emission, International Journal of Machine Tools & Manufacture 46, p [5] Chung K.T., Geddam A., 2003, A multi-sensor approach to the monitoring of end milling operations. Journal of Materials Processing Technology 139, p. 15. [6] Jemielniak K., Arrazola P.J., 2008, Application of AE and cutting force signals in tool condition monitoring in micromilling. CIRP Journal of Manufacturing Science and Technology 1, p. 97. [7] Marinescu I., Axinte D, A, 2008, A critical analysis of effectiveness of acoustic emission signals to detect tool and workpiece malfunctions in milling operations. International Journal of Machine Tools & Manufacture 48, p

6 420 C. Doukas et al. / Procedia CIRP 8 ( 2013 ) [8] Inasaki I., 1998, Application of acoustic emission sensor for monitoring machining processes. Ultrasonics 36, p [9] Cao H., Chena, X., Zia Y., 2008, End milling tool breakage detection using lifting scheme and Mahalanobis distance. International Journal of Machine Tools & Manufacture 48, p [10] Ertekin Y. M., Kwon Y., Tseng B., 2003, Identification of common sensory features for the control of CNC milling operations under varying cutting conditions. International Journal of Machine Tools & Manufacture 43, p [11] Hillaire P., Wright P. K., 2005, A Process Monitoring System for Process-Based Product Validation. Journal of Manufacturing Processes 7, p. 69. [12] Stavropoulos P., Salonitis K., Stournaras A., Pandremenos J., Paralikas J., Chryssolouris G., 2007, Advances and Challenges for tool Condition Monitoring in Micro-Milling, IFAC Workshop on Manufacturing Modelling, Management and Contro l, p. 157 [13] Shiraishi M., 1988, Scope of in-process measurement, monitoring and control techniques in machining processes Part 1: In-process techniques for tools, Precision Engineering 10, p [14] Prickett P.W., Johns C., 1999, An overview of approaches to end milling tool monitoring, International Journal of Machine Tools and Manufacture 39, p [15] Byrne G., Dornfeld D., Inasaki I., Ketteler G., König W., Teti R., 1995, Tool Condition Monitoring (TCM) The Status of Research and Industrial Application, CIRP Annals - Manufacturing Technology 44, p [16] Kluft, W., 1983, Werkzeuguberwachungssysteme furs die Drehbearbeitung, doctoral thesis, RWTH Aachen. [17] Chryssolouris G., Domroese M., 1988, Sensor Integration for Tool Wear Estimation in Machining, Symposium on Sensors and Controls for Manufacturing, ASME Winter Annual Meeting, Chicago, Illinois, USA, p. 115 [18] Chryssolouris G., Guillot M., Domroese M., 1987, Tool Wear Estimation for Intelligent Machining, Symposium on Intelligent Control, ASME Winter Annual Meeting, Boston, Massachusetts, p. 35 [19] Principe, C., Yoon, T., 1991, A new algorithm for the detection of tool breakage in milling Int. J. Mach. Tools Manufact. 31, p. 443 [20] Stavropoulos P., Salonitis K., Stournaras A., Pandremenos J., Paralikas J., Chryssolouris G., 2007, Experimental Investigation of Micro-milling Process Quality, 40th CIRP International Seminar on Manufacturing Systems, Liverpool.