International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 www.elsevier.com/locate/ijmactool Temperature measurement in high efficiency deep grinding A.D. Batako*, W.B. Rowe, M.N. Morgan Liverpool John Moores University, GERI, Advanced Manufacturing Research Laboratory, Byrom Street, Liverpool L3 3AF, UK Received 25 November 2004; accepted 13 January 2005 Available online 7 April 2005 Abstract Temperature measurements are employed for research into the mechanics of grinding and for process monitoring. Temperature measurement in grinding presents a number of challenges particularly in High-Efficiency Deep Grinding (HEDG). High work-speeds require fast response and deep cuts require measurement over a large temperature field. High-speed fluid delivery creates problems for all types of temperature measurement. Electrical noise may require zero-shift filtering. Several techniques of temperature measurement are described. The merits and problems using thermocouple techniques are discussed in detail. In particular, the effect of junction size and shape are discussed. It is shown that the shape and size of the junction have a strong effect both on the reliability of the signal and on the accuracy of the signal. Other problems discussed include improvement of signal to noise ratio, measurements under wet grinding conditions and a technique for measuring contact surface temperatures when taking deep cuts in creep grinding and high-efficiency deep grinding. q 2005 Elsevier Ltd. All rights reserved. Keywords: Grinding; Temperatures; HEDG 1. Introduction Energy in grinding is mostly converted into heat. Excessive heat leads to thermal damage, residual stress and micro-structural changes in the surface layer. These effects can be highly damaging to workpiece life and component quality. Mathematical modelling of the grinding process [1–10] has improved understanding of heat partition between the elements of the system and the mechanisms of heat transfer. Accurate analysis of temperature in grinding requires a repeatable and stable measurement technique for correlation with theory. Temperature predictions based on power or force measurement need direct temperature measurement for validation and calibration of heat convection factors. Several techniques of temperature measurement are commonly employed. Typical approaches include thermal imaging and thermocouple measurements. Other techniques involve use of heat-sensitive coatings, low melting-point * Corresponding author. Tel.: C44 151 231 2126; fax: C44 151 231 2590. E-mail address: a.d.batako@livjm.ac.uk (A.D. Batako). 0890-6955/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijmachtools.2005.01.013 coatings and fibre optics. This paper briefly reviews these techniques and presents a detailed description of the development of the single-pole thermocouple technique. A method is suggested for reliable measurement of temperatures in high speed deep grinding using the single-pole technique. 2. Non contact measurement 2.1. Optical techniques Fig. 1a shows an example of a thermal imaging approach to temperature measurement. A thermo-camera records the temperature field distribution on the side of the workpiece. In the example shown, results are compared with predictions from FEM simulation. The technique provides a graphic image of the whole temperature field including distances well removed from the grinding zone. It is less useful in accurately determining grinding contact temperatures. Thermal imaging has the advantage that sub-surface temperatures are provided in real-time. Direct readings illustrate the temperature field throughout the grinding 1232 A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 Fig. 1. Examples of non-contact temperature measurement. process. A disadvantage with this approach is that the camera records the heat radiated from a side-plane perpendicular to the contact zone. The presence of fluid can interfere with the radiation and hence affect the reading. Also, the temperatures in the plane of measurement may not be typical of temperatures central within the contact zone. Heat convection to the grinding fluid takes place on the side face of the workpiece, whereas in the central plane of the workpiece the heat transfer is two-dimensional. These factors introduce elements of uncertainty into the measurement. Fig. 1b illustrates the use of a phototransistor (4) that records infrared radiation from the grinding zone through a hole (2) made in the wheel. The current generated by the phototransistor is proportional to the intensity of infrared radiation. The signal is conditioned by an amplifier (6) and sent to a recording device (7). This method provides an almost noise free signal but has problems. The method only works well in dry grinding. In wet grinding, fluid and chips fill the measurement hole leading to loss of the output signal. The hole in the wheel is difficult to make and may require involvement of the wheel manufacturer. Control of hole-size variation and angle of inclination is difficult. The duration of the signal is a function of wheel speed and hole diameter. The third problem is that the temperature is recorded in a gap, where there is interference from swarf and grinding debris. 2.2. Fibre optic techniques Recent research in fibre optic sensing has made it possible to develop a new temperature measurement device based on fibre Bragg technique. The change in the sensing grating due to the heat causes variation of the wavelength of the reflected light. Wavelength division multiplexer (WDM), coupler and two gratings method, and thin film Fig. 2. Fibre optic technique; (a) system configuration; (b) output signal; [20]. A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 dense wavelength division demultiplexer (DWDM) are the techniques used to detect a change in the wavelength. Fig. 2 illustrates the DWDM technique, where Fig. 2a shows the system configuration and Fig. 2b compares the output voltage of the fibre optic sensor with a commercial double pole thermocouple, where both sensors were located below the grinding surface. The fibre optic produced a strong signal virtually free from noise. However, this technique requires measurement before and after cutting because the sensor is located below the cutting surface. A safety layer is required at the top of the sensor. The final temperature values are then extrapolated; consequently this method does not provide direct contact temperatures. The fibre optic technique is relatively expensive and fragile to handle. 3. Direct contact measurement 3.1. The coating technique The coating technique provides direct measurement of maximum temperature experienced at a particular position. The technique has the potential to provide better insight into the process since temperature sensing takes place precisely at the position of interest. Low melting point materials such as indium, bismuth and heat sensitive paints have been used to estimate temperatures in grinding. The method provides temperatures below the contact surface with a simple set up involving a split workpiece. Used together with metallurgical analysis the method gives an insight into the temperatures experienced within a damaged layer. Low melting point coatings provide a useful cross-check on the order of magnitude of other methods of estimation. 3.2. Thermocouple techniques There are various methods of setting up a thermocouple (TC) within a workpiece to measure temperature in the grinding contact zone. Some of these methods are 1233 reviewed in Ref. [13]. A relatively simple way of inserting a thermocouple is by use of a split workpiece. There are three main techniques. Conventional double-pole thermocouples may be inserted below the workpiece surface. Alternatively, a grindable double-pole thermocouple junction may be formed at the grinding contact. A third possibility is a grindable single-pole thermocouple formed at the grinding contact. The single-pole thermocouple uses a conducting metal dissimilar to the conducting workpiece material inserted in the workpiece. The grinding action forms a thermocouple junction at the surface. 3.2.1. Double-pole thermocouples Double-pole thermocouples inserted below the workpiece surface have standard calibrations and usually provide a good temperature signal. However, because these thermocouples measure temperature below the surface, the temperature measured is due to heat conducted through the workpiece material and through the thermocouple insulation (Fig. 4, right). The thermal properties of the insulation are usually ignored and is a potential source of inaccuracy. Since the temperature gradient near the surface may be steep and non-linear, the accuracy of measurement relies on an uncertain process of extrapolation. A further problem is that the conventional thermocouple has a junction volume that is large in relation to the temperature gradient. This means that the temperature measured is an average value over the junction volume. The junction volume lies at varying depths below the surface. The usual configuration of double-pole thermocouple is shown in Fig. 3. Use of conventional double pole thermocouples to measure temperature at the workpiece contact surface requires accurate measurement of the position of the measuring junction relative to the surface before and after grinding in order to determine the average location of temperature measurement. Usually, it is assumed that the temperature is measured at the centre of the volume of the measuring junction. It is important to know, where this lies. The physical size of the junction, therefore, plays Fig. 3. Double pole TC configuration for surface grinding. 1234 A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 increases the size of the assembly and makes the process of assembling the fragile foil elements and mica insulating strips more difficult. It was therefore, decided to use singlepole thermocouples, wherever possible. The single-pole configuration is relatively easy to assemble and provide greater reliability. Fig. 4. Configuration of grindable double pole thermocouple. an important role in the temperature measured. If the temperature gradient is high at the location of the measuring junction, the reading will be an average of the temperatures between the maximum and minimum temperature over the measuring volume. A further problem is that if the junction is not formed precisely in the plane of desired measurement, there is a need to predict the temperature at the desired location by extrapolation of an assumed temperature gradient. Theoretical modelling, therefore, plays an important role in extrapolation from a measurement within the body to a point on the surface. Use of a conventional double-pole thermocouple for contact surface measurement is time consuming because of the number of measurements required. This is further exasperated by the calculation to extrapolate temperatures and the precision required to get the wheel close enough to the thermocouple junction while avoiding cutting through it. Grindable double-pole thermocouples as shown in Fig. 4 can be formed at the workpiece surface and overcome some of the problems of the conventional double-pole thermocouple. The approach to assemble a grindable double-pole thermocouple is similar to the assembly of a grindable single-pole thermocouple as described below. The increased number of layers involved in the double-pole arrangement 3.2.2. Single-pole thermocouples Single-pole thermocouples are grindable and this is an advantage compared to conventional double-pole thermocouples. In the single-pole technique, the measuring junction is formed with the workpiece during grinding by smearing the thermocouple material over the workpiece. Thus the grinding temperature is measured at the contact surface. Single-pole thermocouples are based on an insulated foil strip or wire inserted either in a grinding wheel or in a workpiece. Usually, the thermocouple is inserted in the workpiece rather than in the grinding wheel. This is a much simpler and less expensive arrangement for experimental purposes. 3.2.2.1. Insertion in a grinding wheel. Insertion of a thermocouple in a grinding wheel allows continuous monitoring of grinding temperature. The workpiece forms the second pole of the thermocouple and is interchangeable; consequently, this method is subject to calibration of the temperature values and could be usable in a production situation. The approach is illustrated in Fig. 5, where two single-pole thermocouples (3) are inserted in a grinding wheel. High-speed slip rings are required to extract the signals and tend to introduce noise. The aggressive action of the coolant and swarf also affect the quality of the signal. The method requires fast signal recording equipment, since each signal pulse characterising contact temperature has a short duration, which depends on the process type (surface or cylindrical grinding, shallow or deep cuts). 3.2.2.2. Insertion in a workpiece. In the single-pole configuration, constantan is used to form a J-type Fig. 5. Single-pole thermocouples inserted in a grinding wheel [12]. A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 1235 Fig. 6. View of the Leg and Thin-End sections of a J or T-type TC. thermocouple when grinding ferrous materials. Constantan foil may be taken from a commercial J- or T-type thermocouple and inserted in the workpiece. Fig. 6 illustrates a single-pole thermocouple. It shows a ‘ThinEnd’ and a ‘Leg’ of a thermocouple foil with typical dimensions. The Thin-End allows measurement with depths of cut with a range from 0 to 6 mm. The Leg can be used for a greater range of depths. The Thin-End is suitable for shallow grinding or when the increment of depth of cut is not greater than 6 mm. For a depth of cut greater than 5 mm, the cross section of the thermocouple begins to increase. This leads to uncertainty in the calibration factor. Fig. 7 illustrates the arrangement of a single-pole thermocouple within a split workpiece, where configuration is similar to the arrangement shown in Fig. 4. A typical temperature signal is shown in Fig. 7b. This technique provides good temperature readings, although problems may be met particularly in wet grinding. The problem is to maintain a good junction throughout the grinding contact. If the junction breaks down the signal becomes difficult to interpret. This problem becomes more serious in wet grinding when there is a greater tendency for the junction to break down. Problems can also arise if the insulation fails. 3.3. Tests on thermocouple performance In High Efficiency Deep Grinding (HEDG), depths of cut may exceed 5 mm. This makes the Thin-End thermocouple unsuitable for temperature measurement. Fig. 8 displays an arrangement used to investigate the performance of single-pole thermocouples. Three singlepole thermocouples of different shape and size were inserted side by side in a workpiece for comparative tests. A constantan wire thermocouple of 25 mm in diameter was installed in line with a narrow (Thin-End) thermocouple (10!15 mm) and a wide (Leg) thermocouple (50!750 mm). The sensors were assembled in a row to allow the grinding wheel to travel across the measuring junctions simultaneously. The thermocouples were therefore, tested under similar conditions. A series of grinding tests were undertaken varying depth of cut. Analysis of the results allowed the best configuration to be selected for HEDG. Under dry grinding conditions all three thermocouples provided acceptable results. Under wet grinding conditions, the results in Fig. 9 demonstrate that the wide Leg thermocouple gave the most reliable temperature profile. The narrow Thin-End thermocouple and the narrower Wire thermocouple both produced poor signals. After grinding, the junctions were examined under a microscope. Microphotographs of the junctions are shown in Fig. 10. It is observed from these pictures that the wide Leg thermocouple provides a robust junction, whereas the narrow ThinEnd and Wire thermocouples possibly suffered some distortion. Fig. 11 compares temperature signals with hardware filtering to remove noise and temperature spikes. The experiment was repeated several times and the results were found to be consistent. It can be seen that the use of a filter is very helpful for interpretation of the signal. In principle, the arrangement in Fig. 8 allows comparison of the speed of response of different thermocouples. It is observed in Fig. 11 that both the wide Leg and the narrow Thin-End have approximately the same speed of response as shown by the fast rise of the temperature signal. However, a useful measure of the quality of a thermocouple signal is the shape of the Fig. 7. (a) Single pole TC configuration in split workpiece; (b) temperature signal. 1236 A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 4. Calibration of thermocouples inserted in a workpiece Fig. 8. Arrangement of three single pole thermocouples in workpiece. temperature decay. The wide Leg displays a typical temperature decay as expected from heat transfer theory, whereas the narrow Thin-End produced a poor signal decay shape. It was observed from time-to-time that the wide Leg junction was destroyed as soon as the wheel passed over the measuring tip. It was considered that this was due to the Leg being too stiff due to its greater thickness. This leads to the Leg pulling back the smeared layer as the load is reduced. The junction is not strong enough to restrain the Leg. An example of loss of signal is shown in Fig. 12a. This problem was investigated by using a standard rolling machine to reduce the thickness of the thermocouple, as shown in Fig. 12b. Several Legs were successfully reduced in thickness. Thicknesses of 25, 38, 46 mm and the original 50 mm were chosen for test. Fig. 12b shows the configuration for rolling the thermocouples and the table indicates the dimensions. Results using these thermocouples are described below. The thermocouples were assembled into the workpieces and initial grinding was carried out to create the junctions. An AD595 monolithic thermocouple amplifier was used to amplify the signals. The thermocouples were calibrated from zero to 150 8C. The calibration arrangement is shown in Fig. 13a. Block (1) holding the workpiece is placed in the middle of an oil tank. A hot plate heats the oil and a magnetic stirrer distributes the heat uniformly within a closed tank. The closed tank limits heat exchange with the surrounding environment. A power controller uses the feedback temperature (2) to maintain constant temperature. A standard thermocouple (3) measures the temperature of the oil. In addition a mercury thermometer (5) was employed as a reference. For accurate calibration, it is imperative to allow enough time for the assembly to heat up and the oil temperature to stabilise before taking readings. The readings from the thermocouples were studied for several days at room temperature and compared with a mercury thermometer. An average discrepancy of 0.5 8C was observed. Fig. 13b depicts a view of the workpiece assembly. Fig. 14 shows typical calibration graphs for the rolled thermocouples during heating and cooling. It is seen that the test thermocouple responses were in close agreement with the standard thermocouple. 5. Grinding tests A series of grinding tests at different depths were undertaken on a surface grinder. Wheel speed and table speed were maintained constant. This allowed temperature readings to be compared for the thermocouples of different thickness. Fig. 9. Examples of temperature signals under wet grinding conditions (Leg, Thin-End, Wire). A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 1237 Fig. 10. Microphotograph of junctions. It can be concluded from Fig. 15 that thermocouple thicknesses within the range tested did not significantly affect the temperature readings. A higher level of electrical noise encountered during tests at depths of 24 and 31 mm caused some discrepancy in temperature reading. However, this problem was not caused by the thermocouple thickness. It was also considered that the fluid application and consequently the fluid convection factor might not have been as consistent as desired throughout the grinding tests. However, the results are reasonably consistent. It was found that thermocouples with a thickness between 35 and 45 mm agreed well with the reference thermometer. An average thickness of 40 mm was judged to be optimal. Fig. 16 illustrates a temperature signal at a depth of cut 10 mm in dry grinding of mild steel using an alumina wheel. 6. Analysis and validation of the results Grinding forces were measured using a Kistler 3-axis dynamometer. Table speed was measured using an acoustic emission sensor that recorded the AE signal from the grinding contact. Measured temperatures were validated with the help of a thermal model [8]. 6.1. Grinding temperature estimation Power supplied to the grinding wheel is converted into heat at the contact interface. Losses unaccounted for include additional stresses locked into the workpiece and swarf material during the deformation process. Energy locked into the material does not reappear as heat within the time-span of the grinding contact, neither is it recovered unlike elastic energy. However, the energy locked into the material is very small compared with the total energy dissipated in grinding and may therefore be considered negligible. The total heat flux is therefore equal to the grinding power per unit contact area. The total heat flux qt is partitioned into qw going into the workpiece, qs transported by the wheel, qch taken by the chip and qf carried out of the contact area by cooling fluid. The total heat flux is the net grinding power P supplied to the spindle per unit contact area blc. qt Z P Fv Z t s blc blc (1) where b-grinding width, lc contact length; Ft tangential grinding force and vs is the linear speed of the grinding Fig. 11. Effect of thermocouple width on temperature signal. 1238 A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 Fig. 12. Arrangement for rolling a thermocouple. (a) Loss of function bond, (b) TC rolling arrangement. wheel. qch Z Tmp rcae qt Z qw C qs C qch C qf (2) A heat flux may be expressed as a product of a heat convection factor (h) and a temperature T. qw Z hw Tmax ; qs Z hs Tmax ; qf Z hf Tmax ; qch Z hch Tmp (3) The heat into the workpiece can be expressed in terms of the maximum temperature on the contact surface [7,8] as sffiffiffiffiffiffi bw lc T qw Z (4) C vw max The workpiece convection factor is therefore hw rffiffiffiffiffiffi b w vw Z C lc (5) C is a factor that depends on Peclet number [8,9] andpbffiffiffiffiffiffiffi w is the thermal property of the work material, bw Z krc, where k is the conductivity, r is the density and c is the specific heat capacity. The heat generated in the chip raises the temperature to a value somewhere between the softening temperature and the melting point; it can be no higher than the melting point. Thus taking the melting point/softening temperature Tmp as 1250 8C, the heat flux carried away by the chip is vw v or qch Z ech ae w lc lc (6) The workpiece and abrasive grain contact may be considered as a workpiece-wheel sub-system of the heat partition system. The workpiece-wheel sub-system shares a heat flux, qwsZqwCqs and the heat partition between the workpiece and wheel is estimated using the Hahn concept of a grain sliding on a workpiece [14,15]. Partition between the workpiece and the wheel is given by Rws Z   0:97kg K1 qw hw Z Z 1 C pffiffiffiffiffiffiffiffi qw C qs hw C hs bw r0 vs (7) where kg is the thermal conductivity of the abrasive grains and r0 is the effective wear flat radius of the grains. Values of Rws may be calculated from Eq. 7. Thus the wheel heat convection factor obtained from Eq. 7 is h s Z hw  1 K1 Rws  (8) The maximum temperature Tmax in the grinding zone can therefore be estimated from the values of convection factors and Eqs. 1 and 2. The maximum temperature is therefore given by Fig. 13. Thermocouple calibration arrangement. (a) calibration settings, (b) wrokpiece assembly. A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 1239 Fig. 14. Typical calibration chart of the rolled TC. Tmax Z qt K qch hw Rws C hf (9) where hf is the heat convection factor for the fluid. In dry and burnout conditions, the fluid convection factor is effectively zero. As reported in [8,9,16,17] values of fluid convection for water-based fluid vary broadly from 0 to 290,000 W/m2K depending on the effectiveness of fluid delivery and the type of fluid. Fluid convection factor can also be estimated very approximately assuming steady conduction into a ‘solid’ fluid carried on the wheel surface from rffiffiffiffiffi vs (10) hf Z 0:94bf lc Heat convection into the fluid estimated from Eq. 10 is much lower than possible with a fluid surface on the wheel since the fluid contact in practice is turbulent rather than static. This increases the rate of heat transfer. However, the Fig. 15. Temperature readings varying thermocouple thickness. 1240 A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 In general the experimental results for measured temperatures are in agreement with expected behaviour in the transition region and with thermal modelling. During trials, every single grinding test provided a good temperature measurement. On average 250 grinding tests were made and 250 good signals were recorded. This demonstrates that wide Leg thermocouples provide reliable temperature readings in agreement with thermal modelling. The most pleasing aspect was that taking temperature measurements was a stable repeatable process. 7. The thermocouple size effect in HEDG and temperature signal quality Fig. 16. Temperature signal from a 40 mm thick TC in dry grinding at 10 mm depth of cut. supply of fluid into the contact zone is often less than perfect so that the convection factor for the fluid may over-estimate the heat carried away. It is found that the heat convection factor is strongly dependent on the effectiveness of fluid delivery. In practice, maximum temperature can be calculated for two different conditions. These are for the case of effective fluid delivery and for the case of fluid burn-out. This gives two extremes for the calculation. With effective fluid delivery, temperatures measured tend towards a value predicted at the lower extreme. When burn-out occurs the temperatures increase dramatically towards the other extreme. 6.2. Experimental validation of grinding temperatures Temperatures measured with rolled Leg thermocouples were validated against predicted values using Eq. 1–10. Fig. 17 illustrates results obtained in grinding three types of steel. Fig. 17a demonstrates a typical sub-optimal fluid delivery, where measured temperatures are close to predicted values for the dry burn-out condition. In cast iron grinding, Fig. 17b, a steady transition process appears to take place. The first result at 6 mm depth of cut illustrates an optimal fluid delivery but as the depth of cut increases temperatures move into the transition zone between the two extremes. These results suggest partial fluid burn-out. Since the temperatures measured in this case are close to the burnout value measured with water-based emulsions, it suggests a transition that commences at approximately 100 8C and is complete at 200 8C. In Fig. 17c most of the results with the hardened M50 tool steel (62HRC) indicate a dry burnout condition and measured temperatures are close to values predicted for the dry condition. It is observed that temperatures at 4 and 7 mm depth of cut are within the transition zone. In the single-pole configuration, the speed of response of a thermocouple is independent of the foil thickness. This is because the thermocouple junction is created at the grinding contact and the emf is generated equally in either a thick or a thin thermocouple. The differences in thermocouple resistance are negligible if compared to the total resistance of the workpiece and the connecting wires in the circuit. It is interesting to speculate on effects on signal quality in a single-pole thermocouple. 7.1. Causes of a poor temperature signal In temperature measurement, using a single-pole thermocouple, signal quality is strongly affected by the quality of the junction, which in turn is affected by the number of abrasive grains simultaneously in contact with the thermocouple at a given time. Coolant has a detrimental effect on the establishment of a secure bond between the thermocouple and the workpiece material. Fig. 18 illustrates the effect of thermocouple size on the number of grains in contact. In Fig. 18a it is observed that with a narrow thermocouple a smaller number of grains is in contact at the junction and there may be instants when no grain is in contact with the thermocouple. In the presence of coolant, a film can form between the smeared layer of the junction and the workpiece surface so that the junction bond tends to become intermittent and also the generation of emf. An intermittent emf is frequently observed in experiments with narrow thermocouples. This requires the experiment to be repeated. With narrow thermocouples, the junctions are fragile and it seems that often the smeared thermocouple layer does not reach the workpiece body. This is shown in Fig. 18b. A wide thermocouple has numerous grains in contact at any given time and this increases the probability of maintaining contact between the smeared layer and the workpiece surface (Fig. 18c) resulting in a continuous signal. 7.2. Measuring-source power Apart from the causes underlined above, the power of the emf source affects the temperature readings. In essence, A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 1241 Fig. 17. Comparison of experimental results with predicted values. we measure the intensity of micro-current that flows through the junction. Consequently, the electrical circuit of the measuring device influences measured values depending on the loading effect according to the Thévenin theorem [18]. If the source is not powerful enough, the dissipation due to internal resistance of the measuring circuit affects the reading. Ideally the output impedance of the thermocouple should match the input impedance of the amplifier and the output impedance of the thermocouple amplifier should match the input impedance of the data logging system etc. However, in practice, this is difficult to achieve. Fig. 19 illustrates the importance of the volume of the measuring junction. A Thin-End thermocouple has a small junction volume, Fig. 19a; consequently the current generated is small. Therefore, the drop in measured voltage due to internal dissipation is a problem. The solution to this problem is the use of a wide thermocouple (Fig. 19b) 1242 A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 Fig. 18. Size effect in HEDG. (a) Number of grains in contact, (b) Thin-End junction, (c) Leg junction. that generates sufficient current to minimise the effect of losses in the internal circuit of the measuring device. As the thermocouple stretches across the heat source (wheel width), each grain generates a micro-junction as illustrated in Fig. 19c. These micro-junctions are equivalent to spot thermocouples connected in parallel and produce an average reading over the junction. 8. Signal sampling, noise and interference 8.1. Signal sampling The thermocouple response, the temperature signal profile and signal magnitude are also affected by the sampling rate and the processing speed of the data logging equipment. The worktable speed affects the requirements for these parameters. Fig. 20 shows the duration of temperature signal as the wheel, that is the heat source travels over a thermocouple. It is seen that in shallow cut with thinner thermocouple, this time becomes very short. If the work speed is 1 m/s and the depth of cut is 1 mm, the maximum contact length is typically 15 mm. The temperature has to be measured over a period of less than 0.015 s. If the sampling rate is 1000 Hz, which is one point per millisecond there will be a maximum of 15 points to characterise the temperature signal. This is insufficient. The maximum temperature could be seriously underestimated. Consequently, a high sampling rate is required. It is prudent to over-sample the signal and filter the noise out. A fast data acquisition system is paramount for accurate temperature measurement. Unfortunately, with high sampling rates the level of high frequency noise is increased. Hardware pre-acquisition filtering or post acquisition software filtering can be used to recover the required temperature signal. 8.2. Noise and interference In the single-pole configuration, the workpiece is one half of the thermocouple. This makes the single-pole thermocouple signal vulnerable to environmental electrical noise and interference. Noise is undesirable data added to the measurement signal. Noise exists in the form of Fig. 19. Effect of junction volume on source power. (a) Point resolution, (b) Measuring volume stretched across heat source, (c) Equivalent micro-junctions. A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 1243 between the TC and the workpiece to the worktable (Fig. 21b). In the single-pole configuration as shown in Fig. 21b, physical insulation of the workpiece is effective for dry grinding. Fig. 20. Duration of signal data logging. interference from external electrical and magnetic fields with the measurement system and random motion of electrons and other charge carriers from components within the system. Fig. 21a [21] illustrates a signal recorded during a test to identify adequate time and conditions for temperature measurement. It clearly shows a high level of noise when other machinery is operating nearby. This reflects the importance of identifying sources of noise and defining optimal conditions to minimise noise level. 8.2.1. Source of noise Any operating machine is a source of noise but HF noise is mainly generated by high frequency invertors and electrical drives, hydraulic pumps, old CNC controllers and signal monitoring kits (random noise and inductive coupling). Noise is also generated by the mains power supply system from nearby high-power machinery and the earth (capacitive coupling, [18]). Also multiple earths can cause noise due to difference in potential between earth points. In grinding, fluid application drastically increases noise interference, especially in the single-pole TC configuration. This is because the coolant forms a loop 8.2.2. Noise minimisation If the measurement is carried out in a workshop, where other machinery is in operation, it is important to identify the best time when the level of noise is a minimum. Attention must also be paid to how extension cable and sockets are laid out. If the measurement involves more than one signal amplifier, it is advisable to supply power to them from different main sockets, not from a single multiplesocket extension. Signal cables should not cross or lie close to main power cables. This reduces inductive and capacitive coupling. A single-point earth is important for noise reduction. Fig. 22 illustrates three signals recorded from surface grinders in the same environment but under different noise conditions. Fig. 22a shows an ideal clean temperature signal recorded from a surface grinder with a single-speed motor without speed control or an inverter. The signal was sampled at 250 kHz. In this signal the superimposed grain temperature spikes are observed above the background temperature. The signal in Fig. 22b was recorded on another surface grinder in the location with the same data logging equipment but the sampling rate was lowered to 150 kHz. The machine has a variable speed controller with a high frequency inverter. This result reflects noise generated by the controller and the inverter, as no other machinery operated at that instant. Fig. 22c illustrates a compound effect of noise from numerous machines operating at the instant when the signal was recorded at 150 kHz on the same grinding machine tools as in Fig. 22b. The drastic increase in noise amplitude makes it impossible to measure maximum temperature accurately. A filtered signal is always reduced in amplitude compared with the original Fig. 21. Source of noise in recorded signal. (a) Environmental noise due to machinery [21], (b) Coolant application looping workpiece to table. 1244 A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 Fig. 22. Temperature signal recorded from two types of surface grinder. (a) Single speed grinder no inverter, (b) Variable speed grinder & inverter on, (c) inverter & machinery on. signal. In creep-feed grinding, the workspeed is relatively low and consequently, a lower sampling rate would cut down the level of HF noise. However, in HEDG and in shallow-cut grinding the table speed may reach 1 m/s thus the signal sampling rate is critical. Hardware and software filtering techniques are used to recover the temperature signal. Hardware filtering has the advantage that it can be incorporated before the amplifier and it cuts off undesirable noise before it is amplified. Hardware filters have a preset cut-off frequency and do not have the flexibility of software filtering. Sometimes the resulting signal from a hardware filter can be noisy due to interference in the amplifiers and data acquisition devices. When the sensor has a built-in amplifier it is better to use software filtering as it offer more flexibility in the range of cut-off frequencies and the type of filter to use. Depending on the algorithm used a hardware filter may induce a shift in the resulting signal. Software filtering allows use of the zero-shift technique to reproduce the background temperature without shifting the result. Adopting the above measures it is possible to obtain a signal clean enough that may not need filtering. Fig. 23 shows a signal recorded using the listed precautions. It also shows the difference between a conventional Butterworth filter and the zero-shift technique with the same filter. It seen that conventional filtering shifts the signal to the right and also the moving average. 9. Conclusions on temperature measurement The merits of different methods of temperature measurement have been thoroughly explored. The single-pole grindable thermocouple technique was found to be the simplest and most reliable technique. It has been shown that it is important to optimise the thermocouple geometry to achieve reliable signals under HEDG conditions. Other factors that affect the accuracy of a measuring system have been investigated and it has been shown that it is possible to achieve reliable temperature measurements under dry and wet grinding conditions. High-speed sampling and zero-shift filtering reduce the problem of interpreting the temperature signal and measuring maximum temperature. Using this technique, temperatures measured have demonstrated good agreement with predicted temperatures. The work has highlighted the need for further work on the effectiveness of fluid delivery. Acknowledgements The authors would like to express their gratitude to the EPSRC for funding this project. (Grant GR/R68795) and to the project partners: Cranfield University and to industrial collaborators: Castrol Industrial, Element Six, Landis Lund, Renold, Saint Gobain Abrasives, Stresstech–AST, Weston EU. The authors would also like to thank Prof. O. Mgaloblishvili for his valuable support and advice. References Fig. 23. Resulting signal from adopted precautions. [1] W.B. Rowe, M.N. Morgan, J.A. Pettit, A.S. Lavine, A discussion of thermal models in grinding. Society of manufacturing engineers, paper MR90-516, Proceedings of the 4th International Grinding Conference, Dearborn, 1990 pp 1–15. A.D. Batako et al. / International Journal of Machine Tools & Manufacture 45 (2005) 1231–1245 [2] W.B. Rowe, M.N. Morgan, D.R. Allanson, An advance in the thermal modelling of grinding processes, Ann. CIRP 40/1 (1991) 339–342. [3] S.C.E. Black, W.B. Rowe, B. Mills, H.S. Qi, Experimental energy partitioning in grinding, Proceedings of Eurometalworking held at University of Udine, Italy, 94 1994. [4] S. Black, W.B. Rowe, H.S. Qi, B. Mills, Temperature measurements in grinding, Proceedings of 31st MATADOR Conference held at UMIST 20–21 April, Macmillan Press, 1995., ISBN: 0-333-64086-1 pp. 409–413. [5] W.B. Rowe, S.C. Black, M.N. Morgan, B. Mills, Experimental investigation of temperature in grinding, Ann. CIRP 44 (1) (1995) 329–332. [6] W.B. Rowe, S.C. Black, B. Mills, H.S. Qi, M.N. Morgan, Analysis of grinding temperatures by energy partitioning, Proc. IMechE Part B, J. Eng. Manuf. 210 (1996) 579–588 (ISSN. A M Strickland Prize from I Mech E). [7] W.B. Rowe, Thermal Analysis of High Efficiency Deep Grinding, Int. J. Mach. Tools Manuf. 41 (2001) 1–19. [8] W.B. Rowe, M.N. Morgan, A. Batako, T. Jin, Temperatures in High Efficiency Deep Grinding (HEDG), Ann. CIRP 50 (1) (2001) 205–208 (ISBN 1-85312-990-9). [9] W.B. Rowe, M.N. Morgan, A. Batako, T. Jin, Energy and Temperature Analysis in Grinding Keynote Paper, Sixth International Conference And Exhibition on Laser Metrology, Machine Tool CMM and Robot Performance-LAMDAMAP, in: D.G. Ford (Ed.),, Huddersfield University. Published in Laser Metrology and Machine Performance VI, WIT Press, Southampton, UK, ISBN: 1-85312-9909, 2003, pp. 3–23. 1245 [10] M.N. Morgan, W.B. Rowe, A. Batako, ‘Energy limitations in hedg and conventional grinding’, International symposium on advanced abrasive technology, Bristol Key Eng. Mater. 257–258 (2004) 63–68. [11] CIRP work group, Technical University of Braunschweig, Institute of Machine tools and Production Technology 2004. [12] V.I. Davidov, ‘Investigation of heat process in the grinding of alloy with diamond wheel’, Izveestiya Vuzov 11, Mashinostroenie, Moscow, 1967, pp. 141–146. [13] H.K. Tonshoff, T. Freimuth, J.C. Becker, Process monitoring in grinding, Ann. CIRP 51 (2) (2002) 1–21. [14] R.S. Hahn, On the mechanics of grinding process under plunge cut conditions, Trans. ASME, J. Eng. Ind. 94 (1972) 833–842. [15] W.B. Rowe, M.N. Morgan, S.C.E. Black, B. Mills, A. Simplified, Approach to thermal damage in grinding, Ann. CIRP 45 (1) (1996) 299–302 (ISBN 3-905277-25-5). [16] N.K. Kim, S. Malkin, Heat flux and partition in creep-feed grinding, Ann. CIRP 46 (1) (1997) 227–232. [17] S. Ohishi, Y. Furukawa, Analysis of workpiece temperature and grinding burn in creep-feed grinding, Bull. JSME 28 (242) (1985) 1775–1781. [18] W. Bolton, Instrumentation and Measurement Pocket Book, 2nd ed., Newness Press, 1996. [19] Iotech Inc., Signal conditioning and PC-based data acquisition handbook, 1998. [20] F.A. Bezombes, D.R. Allanson, D.R. Burton, M.J. Lalor, Hi-speed temperature measurement using novel optic fibre-based system, SPIE 5145 (2003) PP117–PP127. [21] F.A. Bezombes, Fibre Bragg grating temperature sensors for high speed machining application, PhD Thesis, Liverpool John Moores University, 2004.