Tool Wear Monitoring by Design of Experiments (DOE) for Drilling

Tool Wear Monitoring by Design of Experiments (DOE) for Drilling Pantelis N. BOTSARIS 1, John A. TSANAKAS 1, Maria E. VOGIATZI 1 1Democritus University of Thrace, School of Engineering Department of Production Engineering and Management, Faculty of Materials, Processes and Engineering Vas. Sofias 12, University Central Campus, Building I, 67100, Xanthi, Eastern Macedonia & Thrace, Greece panmpots@pme.duth.gr, itsanaka@ee.duth.gr, marivogia@gmail.com ABSTRACT In this paper, a previous study of the current research team is investigated using the Design of Experiment (DOE) method. In particular, the investigation regards a sensor-based approach for the inprocess tool wear monitoring and detection of tool breakage in drilling processes which was deployed by Botsaris et al.; in the current work, the above approach is examined in terms of four parameters, i.e. cutting speed, feed rate, depth of cut and tool age, using the DOE method. The effects of these main cutting variables on tool life have been assessed by applying the factorial design technique. The whole set of the drilling tests was performed by HSS-Co5% twist tools under dry conditions. The factorial parameters impact to tool life have been generated and plotted based on these tests. The results are presented in terms of mean and RMS values, showing a significant influence of both feed rate and depth of cut over the tool wear monitoring parameters studied in the research team s previous work, i.e. vibration and tool temperature signatures. It was found that tool wear detection via vibration signals monitoring presents limitations with regard to the undesired influence of sensor location, workpiece mounting and external parasitic vibrations (e.g. due to the spindle motor movement). On the other hand, this DOE-based study showed that tool temperature signatures appear to be more promising for in-process tool wear monitoring due to the fact that the temperature sensor focuses on the sensitive area including the cutting tool s peak and each drilled hole within the workpiece. Keywords: tool wear monitoring, Design of Experiments (DOE), cutting speed, feed rate, depth of cut, cutting tool age 1. INTRODUCTION Toward the last decades, automated machining has lead to a significant revision in maintenance philosophy within the manufacturing environment; the oldest and most common fix it when it breaks strategy is clearly being replaced by intelligent condition monitoring (CM) systems and, consequently, optimum maintenance scheduling. Specific pivotal components of a machining system, such as the tool, often suffer from either soft or hard faults in the form of tool wear and/or breakage. Tool abrupt wear, as well as tool breakage before its scheduled replacement, are practically unpredictable during a run-on manufacturing process, resulting to a nonqualified production. The time that a quality control system requires for the identification of the first defective product, until the fault source detection is generally high. This time lag aggravates the total production time and, subsequently, production cost. On-line tool condition monitoring (TCM) emerges as a safe way to implement a manufacturing system with the ability to predict and diagnose tool wear and breakage (Botsaris et al., 2009). In principle, there are two possible TCM approaches, i.e. direct and indirect methods. Direct tool wear estimation systems are able to measure directly the tool wear via tool images, computer vision, etc. which means that these methods actually measure tool wear as such, offering ease of application and high reliability. However, on-line TCM based on direct methods is not feasible in a real manufacturing environment; the detection system should be able to detect the wear zone and measure it, requiring that either the tool be removed from the machine after a certain period of time or a measuring device be installed on the machine. Thus, any of these practices is both technically and economically inadequate, as they cause significant downtime and production loss. On the other hand, instead of wear, indirect monitoring methods measure something else, i.e. a parameter, which must be a function of wear (Jantunen, 2002). Commonly used parameters in indirect methods are cutting forces, vibration, acoustic emission, current, power and temperature. Although these methods present fair reliability, design complexity and high sensor cost,

their great advantage is that they can be applied to online TCM (Botsaris et al., 2008). In particular, the signals of the aforementioned indirect TCM parameters are acquired and analyzed to give an in-process diagnosis of the tool wear evolution and its impact to the estimated remaining lifetime of the cutting tool. Design of Experiments (DOE) method is regarded by several researchers, in the reported literature (Tagaras, 2001; Park 2007), as a useful tool for the optimum selection of the indirect monitoring parameters, as well as for the precise acquisition and analysis of their signals. Experimental procedures were always a basic part of processes related to the development and the melioration of innovative and existent products, respectively. A designed experiment can be utilized to reduce product designing costs through both the acceleration of the design process and the reduction of complex processes. Furthermore designed experiments are a powerful tool in striving to achieve low manufacturing costs by reducing the fault rates and, therefore the need for inspection control and repeated processes. In this study, the DOE method was adopted as an alternative approach for the assessment of two indirect TCM parameters that were examined by the same research team, in a previous work (Botsaris et al., 2009). These two parameters are the generated vibrations and temperature signatures within the cutting area. As in the previous work, the performed process was drilling, the most common process in manufacturing. The design of the experiment and its implementation were realized in the machine shop of Mechanical Design Laboratory (MeDiLab ). For the majority of mechanics and engineering applications, vibration and temperature monitoring are commonly used as reliable wear indices in operation/condition monitoring systems. A detailed analysis of a vibration signal, generated during a drilling process, can give practical information with respect to the cutting tool s condition (Elbestawi et al., 2006); if in a dynamic system such as the machine tool the cutting forces increase, the dynamic response will also increase. In particular, drift forces which can be used for monitoring drill wear, are also the cause of increasing vibration as a function of wear (Jantunen, 2006). Although vibrations are undesirably influenced by the workpiece material, cutting conditions and machine tool structure and real industrial environments present significant noise ratio that interferes with the useful vibration signals, generated during a cutting process, the specific monitoring method is widely used in indirect TCM. Vibration signatures are suggested as reliable, robust and applicable for TCM, in addition to the fact that vibration signatures require fewer peripheral instruments than AE for instance. Furthermore, vibration signals have the quick response time needed to indicate changes for online monitoring (Dimla, 2002). Moreover, accelerometers, that are mainly used to obtain vibration signatures, are simple to operate and are very suitable for wear monitoring because they offer the remarkable advantages, such as ease of implementation and mounting close to the cutting action, as well as resistance to coolants, chips, electromagnetic or thermal influences (Abu-Mahfouz, 2003). On the other hand, temperature-based TCM, involving infrared or fiber-optic pyrometers and infrared thermography imagers, is a major challenge due to numerous practical difficulties involved in cutting processes. 2. EXPERIMENTAL SET-UP 2.1 Experimental Hardware and Software Figure 1 shows the hardware set-up of the conducted experiments. The whole arrangement was set as in a previous experimental work (Botsaris et al., 2009). As mentioned, vibration and tool temperature signals were obtained during drilling operations, under dry conditions, performed on Yang SMV-1000, a three-axis computer numerical controlled (CNC) vertical-type machining center, installed in the machine shop of MeDiLab. The experiments included drilling operations performed by a HSS-Co5% twist drill of 10 mm diameter, under dry conditions, to four reinforced C-70 steel workpieces (length=170mm, width=170mm, height=20mm). Due to the high hardness of the operated workpieces, the optimum drill was selected to have a point angle of 135 o and a helix angle of 35 o -42 o. To achieve the proper stiffness for the tool, during drilling, its length remained at 133 mm. Totally 109 bottom holes, of depth=15mm, drilled per each workpiece. The drilling path started from the 1A hole, continued to 1B, 1C and so on, finishing the first line (No. 1) of the workpiece. The process was continuing to the next lines (2 to 11), until the last hole 11K according to Figure 1. The vibration signals were measured from Kistler 8702/B25M1, a single-axis K-shear accelerometer, mounted on the workpiece longitudinally to the drilling direction, i.e. the Z-axis. The accelerometer has a measuring range of ± 25 g, with a sensitivity of 200 mv/g (± 5%), while its frequency response band ranges from 1 to 8000 Hz. Referring to the non-contact tool temperature measurements were performed with the use of Eurotron s IRtec series Rayomatic 10, a compact digital infrared (IR) temperature

transmitter (IR pyrometer) mounted onto the tool bracket, at a distance of d=10 cm from the drill point, with an angle of θ=45 o from the drill axis. The measuring temperature range of the specific transmitter is from 0 to 600 o C, while it offers accuracy of ± 1% rdg and repeatability of ± 0.5% rdg, with a Øtarget-todistance ratio of 25:1. The transmitter was calibrated measuring the known temperature of heat source. The data from the above three sensors was recorded, with an optimum sample rate of 8 KHz, to a portable computer, via a National Instruments Cdaq-9172 data acquisition unit, and pre-analyzed with the use of National Instruments LabVIEW and Mathworks Matlab 8.0 software. Then, Microsoft Excel and SPSS software were used for the statistical analysis and the application of DOE method to the extracted data. The whole experimental drilling process was prepared and designed with Missler TopSolid 2008, a parametric CAD/CAM software that features 3-D design of workpieces and their virtual machining processes to extract code G and M commands for CNC machining centers. FIGURE 1. The hardware experimental set-up. 2.2 Design of the Experiment The DOE method that was applied to the experiments of this study features the full factorial design, in a case of 2 k factorial experiments; the symbol k refers to the number of factors which take part to the studied case, while the number 2 represents the number of the levels that classify these factors. In other words, in a case of 2 k experiments, there are k factors which can be characterized from two possible classification levels, e.g. either high or low. This, it can be perceived that, a complete experiment by 2 k full factorial design, requires at least 2 k observations. In this study, depth of cut (in mm), feed rate (in mm/min) and tool age (in drill number) were selected as the controlled variables. Based on the type of the drilling tool and the dimensions of each workpiece, the two levels for the depth of cut variable were selected to 7.5 mm and 15 mm, while for the feed rate

variable the two levels were set to 60 mm/min and 120 mm/min, in order to represent cutting conditions similar to those in real manufacturing processes. To sum up (Table 1), the factorial designed experiment that is investigated consists of 2 factors, i.e. depth of cut and feed rate, which fluctuate between 2 levels and a third factor, i.e. tool age, which take more than 2 levels. For each under-processing workpiece, the cutting tool should be unworn (brand new) in order to ensure identical initial cutting conditions for every workpiece. As a result, the dependent variables will refer to the tool age and alternation of the two, designed variables to their levels that are mentioned above. Table 2 presents aggregate of the possible variables combinations and the order of the conducted designed experiments with cutting conditions based on these combinations. For each combination, during these experiments, the signals from the vibration and temperature sensors were acquired to give the required data for the vibration and tool temperature, respectively, variables. TABLE 1. Factorial levels for the designed experiments. Factor Level characterization Level value Depth of cut (factor A) Feed rate (factor B) Tool age (factor C) Low High Low High 7.5 mm 15 mm 60 mm/min 120 mm/min More than two levels TABLE 2. The possible variables combinations and the order of the conducted designed experiments. B low B high A low Workpiece/Experiment 1 Workpiece/Experiment 2 A high Workpiece/Experiment 3 Workpiece/Experiment 4 According to the topology of the drills within the workpiece, as it is presented in Figure 1, the acquisition and further data storage and analysis was made in the form of data lines which, also, included the dead or rapid movement time between each drill. In total, 109 drills were performed to each workpiece. Thus, that the tool age factor is considered that can be classified by 109 possible levels. Moreover, as the sampling rate was selected at 8 KHz, the data size for each drill corresponds to 83330 values. The data sum was classified according to its source (vibration or temperature sensor) and, subsequently to the related variable, and was imported to Excel tables according to its location within each workpiece. 3. RESULTS AND DISCUSSION 3.1 Tool temperature data The whole analysis of the experimental data was performed with the use of SPSS statistics software, version 17. Particularly for the tool temperature data, the specific analysis was based on an arithmetic mean assessment. It should be noticed that, the infrared tool temperature sensor that was used for this set of data, is not affected by the processed workpiece geometry or the drilled holes topology, in contrast to the accelerometers that were used for the vibrations signal acquisition. In other words any alteration to the drill s location does not influence the measured signal, keeping constant cutting conditions, as desired; thus, the drills/wholes (and the related measurements) that may be examined and analyzed can be randomly picked. It is supposable that, the total number of the drills selected for analysis indicates, de facto, the number of the tool age factor s levels.

The measured data was analyzed in two sets of drills. The first set included the criterion that all factors are classified by two levels, as well as the tool age factor. Thus, two drills from each workpiece were selected to be analyzed, giving a total of 8 possible variable s combinations. In the second set, 14 drills from each workpiece were selected to be analyzed; in this case the tool life factor had 14 levels. As a result, this set included a total of 56 possible variable s combinations for the temperature measurements. 3.1.a First set analysis The selected wholes/drills in the first set were the 10 th and the 100 th which are symmetrical on the workpiece plane and their (process) time distance is adequate to give a significant variation in tool s life. The specific experiment is a 2 3 factorial design experiment with no repeats. The limitation that came into question is that the calculation of any errors would be impossible, since in this problem there are null degrees of freedom. In this case, according to the analysis of factorial design experiments without repeatability (residual analysis), the high order interactions are considered as errors, since they are much minor than the rest interactions (Montgomery et al., 2007). Therefore, the whole analysis is performed including to the error- the third order interaction, as well as the interactions (depth of cut)*(tool life) and (feed rate)*(tool life) which feature small squares sum. TABLE 3. Variance analysis for the temperature (first set). Source Type III sum of squares df Mean square F Sig. Partial Eta squared Corrected model 2629.344 a 4 657.336 6.151 0.084 0.891 Intercept 67700.486 1 67700.486 633.543 0.000 0.995 Depth of cut (DoC) 860.343 1 860.343 8.051 0.066 0.729 Feed rate (FR) 321.799 1 321.799 3.011 0.181 0.501 Tool age (TA) 1067.200 1 1067.200 9.987 0.051 0.769 (FR)*(DoC) 380.002 1 380.002 3.556 0.156 0.542 Error 320.581 3 106.860 Total 70650.411 8 Corrected total 2949.925 7 a R squared=0.891 (Adjusted R squared=0.746) FIGURE 2. The (depth of cut)*(tool age) interaction.

As a result, we have the ANOVA (or variance analysis) table, in Table 3, which is exported by applying the current data to the SPSS software. As it can be noticed, both tool age and the depth of cut have a slightly significant impact to the dependent variable, due to the fact that Sig(tool age)=0.051 and Sig(depth of cut)=0.066 which are barely over 0.05, i.e. the significance level. In particular, the tool age s impact appears to be much greater than the depth of cut s one, since the former presents a Sig value closest to the significance level (e). On the other hand, neither feed rate nor the interaction (feed rate)*(depth of cut) possess substantial effect to the dependent variable of temperature. However, it cannot be asserted that these parameters do not affect the dependent variable at all; both feed rate and the interaction do not exceed sensibly the significance level. Figures 2, 3 and 4 show the (depth of cut)*(tool age), (feed rate)*(tool age) and (feed rate)*(depth of cut) interactions of 2 nd rate, respectively. FIGURE 3. The (feed rate)*(tool age) interaction. FIGURE 4. The (feed rate)*(depth of cut) interaction. TABLE 4. Parameter estimates (first set - 2 3 factorial design experiment for the temperature). Parameter Estimate (B factor) Std. error T Sig. 95% Confidence interval Lower bound Upper bound Partial Eta squared Intercept 127.47 8.172 15.558 0.001 101.139 153.155 0.988 A: Depth of cut=1-34.525 10.337-3.340 0.044-67.423-1.627 0.788 B: Depth of cut=2 0 a C: Feed rate=1-26.469 10.337-2.560 0.083-59.367 6.429 0.686 D: Feed rate=2 0 a E: Tool age=10-23.100 7.310-3.160 0.051-46.362 0.163 0.769 F: Tool age=100 0 a A*C 27.568 14.619 1.886 0.156-18.957 74.093 0.542 A*D B*C B*D 0 a 0 a 0 a a This parameter is set to zero because it is redundant.

Following this, the expected tool temperature values are estimated using the linear regression factors that are calculated via the SPSS software. Table 4 presents the parameters estimates including the linear regression factors with tool temperature as dependent variable and tool age, depth of cut and feed rate as independent variables. The estimation of the tool temperature Y i,j (where i is the workpiece number and j is the drill number) is based on the linear regression analysis. For instance, the tool temperature for the 10 th drilled hole of the first workpiece, is estimated as: Y 1,10=127.147-34.525-26.469-23.100+27.621=70.621 ο C In Table 5, the tool temperatures Y 1,10 of the above drilled hole estimated by linear regression, using all the possible variable level s combinations, are compared to the temperatures, for the same drill, that were measured experimentally. The slight declinations between the measured and the estimated values are the result of the approximate error values. The most evident declination is detected to the 5 th combination and is calculated to 10.23%. TABLE 5. Estimated versus measured tool temperatures for the 8 combinations linear regression. Combination Estimated value ( o C) Measured value ( o C) DoC=Low, FR=Low, TA=10 70.621 64.79 DoC=Low, FR=Low, TA=100 93.721 99.56 DoC=Low, FR=High, TA=10 69.522 72.68 DoC=Low, FR=High TA=100 92.622 89.46 DoC=High, FR=Low, TA=10 77.578 86.42 DoC=High, FR=Low, TA=100 100.678 91.83 DoC=High, FR=High, TA=10 104.047 97.88 DoC=High, FR=High, TA=100 133.31 133.31 3.1.b Second set analysis Figure 5 shows a grid of 109 cells which represent the drilled holes and their topology in each workpiece. The numbers, of course, indicate the order of the drilled holes during the processing of the workpiece. The red marked drills are the ones that were selected for the second set analysis. FIGURE 5. The selected drills for the tool temperature s data (second set) analysis.

In this case, there are more measurements that are analyzed than in the first set. Thus, the 3 rd order interaction can be unconditionally included in the error estimation. From the variance analysis, it was noticed that the interaction (depth of cut)*(tool age) does not affect the tool temperature, since it is Sig(depth of cut)*(tool age)=0.992. Hence, the analysis was repeated with this interaction included in the error. Table 6 is the ANOVA Table from which interesting remarks can be concluded. TABLE 6. Variance analysis for the temperature (second set). Source Type III sum of squares Df Mean square F Sig. Partial Eta squared Corrected model 25623.569 a 29 883.571 8.144 0.000 0.901 Intercept 544970.568 1 544970.568 5023.254 0.000 0.995 Depth of cut (DoC) 9360.206 1 9360.206 86.277 0.000 0.768 Feed rate (FR) 7901.482 1 7901.482 72.832 0.000 0.737 Tool age (TA) 4933.006 13 379.462 3.498 0.003 0.636 (FR)*(DoC) 1962.354 1 1962.354 18.088 0.000 0.410 (FR)*(TA) 1466.521 13 112.809 1.040 0.447 0.342 Error 2820.728 26 108.490 Total 573414.864 56 Corrected total 28444.297 55 a R squared=0.901 (Adjusted R squared=0.790) The most significant impact rates are, in higher-to-lower order, presented from the depth of cut, feed rate and tool age; the greater the F value is, the more significant impact each parameter has. Figures 6, 7 and 8 present graphically the interactions of the factors for this analysis set. FIGURE 6. The (depth of cut)*(tool age) interaction. FIGURE 7. The (feed rate)*(tool age) interaction. Suggestively, the tool temperature values for the drills 10 and 89 will be estimated using linear regression analysis for all the possible combinations, in the same way as in first set. In particular, there will be four combinations for each drill.

FIGURE 8. The (feed rate)*(depth of cut) interaction. In Table 7, the tool temperatures Y i,j,k of the two aforementioned drilled holes (k is either 10 or 89, according to the respective drilled hole) estimated by linear regression, using all the possible variable combinations (i,j), are compared to the temperatures, for the same drill, that were measured experimentally. As in Table 5, the slight declinations between the measured and the estimated values are the result of the approximate error values. The most evident declination is detected to the last combination and is calculated to 17.85%. TABLE 7. Estimated versus measured tool temperatures for the 8 combinations of linear regression. Combination i,j,k Estimated value ( o C) Measured value ( o C) 1,1,10 62.612 64.78 1,1,89 101.892 89.67 1,2,10 66.468 72.68 1,2,89 105.748 94.87 2,1,10 104.164 97.88 2,1,89 143.444 154.3 2,2,10 76.629 86.41 2,2,89 115.909 95.22 3.2 Vibration signatures data It should be clarified that, during the data acquisition and analysis and the statistical parameters extraction, no filtering for the noise cut-off was applied to the raw data. Furthermore, the vibration signals are symmetrical to the X axis; thus, the arithmetic mean value is close to zero and, subsequently, this statistical index cannot be correlated to the tool s wear in contrast to the quadratic mean (root mean square, RMS), which consists a quite reliable statistical parameter (Liu et al., 1996; Botsaris et al., 2009).

A serious drawback of vibration monitoring for indirect TCM is the fact that the measured signals are influenced from the sensors location to the cutting region and subject to a high noise ratio in a typical environment of a machine shop. In particular, due to the fixed mounting of the accelerometers to the workpiece, which is also mounted to the chocks, on the machining centre s bench, there is a significant fluctuation to the distance between the sensors and each performed drilling. This fluctuation influences the measured signals due to their attenuation that is caused by their transmission through the workpiece s material. As a result, the measurements that were selected for this analysis correspond to drills that are as much exempt from this influence as possible. Similar to Figure 5, Figure 9 shows the same grid of the 109 cells which represent the drilled holes according their topology in each workpiece and the order of each drilling. The red marked drills are the ones that were selected for this analysis. FIGURE 9. The selected drills for the vibration s data analysis. Hence, the tool age factor has 12 levels each one with respect to each drill, i.e. level 1, 2, 3, 9, 10, 18, 92, 100, 101, 107, 108 and 109. In this case, in this experiment a total number of 2 2 12=48 vibration signal measurements were analyzed. Table 8 is the vibrations ANOVA Table for the above experiment. TABLE 8. Variance analysis for the vibration signals. Source Type III sum of squares df Mean square F Sig. Partial Eta squared Corrected model 65.527 a 36 1.820 3.076 0.025 0.910 Intercept 22.393 1 22.393 37.840 0.000 0.775 Depth of cut (DoC) 15.303 1 15.303 25.860 0.000 0.702 Feed rate (FR) 4.342 1 4.342 7.338 0.020 0.400 Tool age (TA) 17.627 11 1.602 2.708 0.057 0.730 (DoC)*(TA) 16.276 11 1.480 2.500 0.072 0.714 (FR)*(DoC) 3.376 1 3.376 5.704 0.036 0.341 (FR)*(TA) 8.603 11 0.782 1.322 0.326 0.569 Error 6.509 11 0.592 Total 94.429 48 Corrected total 72.036 47 a R squared=0.901 (Adjusted R squared=0.614)

It is noticed that, mainly, the depth of cut parameter and, secondarily, the feed rate and the (feed rate)*(depth of cut) interaction, have a respectable impact to the vibration signal. Moreover, by observing the Sig values, the tool age parameter appears to be slightly influential (Sig=0.057, just over the significance level), while the interaction (feed rate)*(tool age) is considered of negligible impact to the vibration signal (Sig=0.326>>0.05). Figures 10, 11 and 12 give graphically the interactions of the factors for vibrations. FIGURE 10. The (depth of cut)*(tool age) interaction. FIGURE 11. The (feed rate)*(tool age) interaction. FIGURE 12. The (feed rate)*(depth of cut) interaction. With an observant look to both the above three Figures, it can be reckoned that both interactions give an interesting correlation with the vibration signals, even in the case of (feed rate)*(tool age) interaction. Generally, though, an interaction where the P-value up to 0.326 cannot be a priori completely indifferent, especially if the error is calculated approximately, like in these analyses in which the error included the highest order interaction due to the lack of repeat to the experiments. While in variance analysis the results and the respective conclusions are regarded as reliably precise, in regression analysis the assessment that is attempted is based on quite insecure results, as it is witnessed by the significant declinations in Tables 5 and 7. 4. CONCLUDING REMARKS In the presented study, a sensor-based TCM approach, which was developed by the current research team, is examined in terms of four parameters, i.e. cutting speed, feed rate, depth of cut and tool age, using

the DOE method. The effects of these main cutting variables on tool life have been assessed by applying the factorial design technique. Based on the results, it can be concluded that the selection of both depth of cut and feed rate parameters was proved correct; the analysis of the experimental results indicated that these two variables have a substantial impact to the tool wear indices of vibration and thermal (tool temperature) signatures, and, thus, to the tool wear itself. Moreover, by applying a number of possible alternations to the level of the above variables and by controlling these possible combinations, an approximate estimation of the tool temperature and the vibrations was realized; then, the estimated values were compared with the measured ones, in order to investigate the potential of this approach to foresee the trend of any indirect monitoring parameter towards tool wear. Furthermore, the results of the presented analysis showed that, from a DOE based point of view and for the specific dry cutting conditions, tool temperature data appears to be more efficient as a tool wear index; the designed experiments were able to estimate the tool temperature of each drill, in a more precise way than in the case of the vibration data. On the other hand, vibration monitoring has specific, unconquerable yet, limitations; these signals are undesirably influenced by the workpiece material, cutting conditions and machine tool structure and real industrial environments present significant noise ratio that interferes with the useful vibration signal, generated during a cutting process. In contrast, the tool temperature measurements seem to not be influenced by the ambient, environmental temperature, since the infrared sensor is focused close enough to the cutting field. Further ambition of the current research team is to perform the presented analysis, involving a more reliable data sample from a greater number of processed workpieces that, consequently, will offer the needed repeatability for this method. The current researchers also aim to investigate the potential of alternative sensor approaches, i.e. infrared thermal imaging systems, as well as classification tools based on fuzzy logic or neural networks. REFERENCES Botsaris, P.N. and Tsanakas, J.A., 2009. Mechanical and thermal signatures as indirect tool wear monitoring indices case study: drilling. In Proceedings of the 22 nd International Congresss on Condition Monitoring and Diagnostic Engineering Management (COMADEM), San Sebastian, Spain, June 9-11 2009, ISBN 9788493206468. Jantunen, E., 2002. A summary of methods applied to tool condition monitoring in drilling. International Journal of Machine Tools & Manufacture, vol. 42, pp. 997-1010. Botsaris, P.N. and Tsanakas, J.A., 2008. State-of-the-art in methods applied to tool condition monitoring (TCM) in unmanned machining operations: a review. In Proceedings of the 21st International Congress on Condition Monitoring and Diagnostic Engineering Management (COMADEM), Prague, Czech Republic, 2008, pp. 73-87, ISBN 978-80-254-2276-2. Tagaras, G.N., 2001. Stastical quality control. Ziti Publications, Athens, Greece, ISBN: 960-431-706-7. Park G.J., 2007. Analytic Methods for Design Practice. Springer Publications, London, United Kingdom, ISBN: 978-1-84628-472-4. Elbestawi, M.A., Dumitrescu, M. and Ng, E.G., 2006. Tool Condition Monitoring in Machining. Springer Series in Advanced Manufacturing: Condition Monitoring and Control for Intelligent Manufacturing, Springer Publications, London, United Kingdom, pp. 55-82, ISBN: 978-1-84628-268-3. Jantunen, E., 2006. Indirect multisignal monitoring and diagnosis of drill wear. VTT publications, Espoo, Finland, ISBN: 951-38-6692-0. Dimla, D.E., 2002. The correlation of vibration signal features to cutting tool wear in a metal turning operation. The International Journal of Advanced Manufacturing Technology, vol. 19, pp. 705-713. Abu-Mahfouz, I., 2003. Drilling wear detection and classification using vibration signals and artificial neural network. International Journal of Machine Tools and Manufacture, vol. 43, pp. 707-720. Montgomery, D.C., Runger, G.C. and Hubele, N.F., 2007. Engineering Statistics. John Wiley & Sons, Inc. Publications, United States, ISBN: 0-471-17026-7. Liu T.I. and Anantharaman, K.S., 1996. Intelligent classification and measurement of drill wear. Journal of engineering for industry, vol. 116 (3), pp. 392-397.