11 OnLine Monitoring of Grinding Machines Gianluca Pezzullo Sponsored by: Alfa Romeo Avio Introduction The objective of this project is the development and optimization of a sensor system for machine tool diagnosis applied to grinding machines. The grinding centers under examination belong to a group of twelve used for CBN grinding of superalloy turbine blades and nozzle guide vanes. The grinding wheels are electroplated CBN with a galvanic process on a piece of hardened and ground steel using nickel like binder. The characteristics of the grinding wheels are: weight 11 Kg, diameter mm, width 5 mm. During operation of the grinding machines, a breakdown of the spindle roller bearing was verified to occur in an unpredictable way and with unacceptable frequency. In order to diagnose and prevent this catastrophic malfunction and verify the actual working state of the grinding center, a machine tool monitoring procedure based on vibration signal detection and analysis was developed. The vibration sensor was an accelerometer installed on the spindle of every grinding machine; the vibration signal was processed and analysed on the basis of a comparison between previous and actual vibration conditions as monitored on the machine tool. Decision making on machine tool state and setting of safety limits on sensor signal parameters to prevent spindle roller bearing breakdown were carried out also with the help of references from the literature [1]. Experimental tests The testing program includes sensor signal acquisition and analysis during machine tool running with and without grinding wheel (Fig. 1). A balanced grinding wheel is used for testing at a spindle operating speed of 6 rpm. Based on this speed value and taking into account the system kinematics, the characteristic frequencies of the bearing elements and those correlated to the engine and the spindle rotating speeds are calculated (Tab. 1) []. The vibration sensor is an accelerometer with bandwidth 1 Hz and it is utilised for the online monitoring of the performance of the critical components of the grinding machine, such as the spindle, the electric motor, and the roller bearings (Fig. ). Signal detection was carried out through a portable instrument (Movilog) which is basically a spectrum analyser (Fig. ). Through dedicated software (Moviscope), the instrument provides in real time a number of spectrum features which are transferred to a computer through a serial port and stored on hard disk for postanalysis. Signal analysis was carried out in the frequency domain. The principal signal features (absolute displacement AD, vibration velocity VV, acceleration G, rolling element defect RED) were measured by the sensor system with reference to frequency ranges centered on the characteristic frequencies for the critical components of the grinding machine. The first three parameters, i.e. absolute displacement AD, vibration velocity VV, and acceleration G, are obtained through acceleration measurements. This is the most frequent type of measurement, since the accelerometer is the easiest vibration sensor to use. The acceleration parameter G is the most significant in the case of high frequency signals, whereas the displacement parameter AD better relates to low frequencies. Vibration velocity VV is an appropriate compromise, suitable in most cases. Every parameter was first analysed using a large frequency band ( Hz for absolute displacement AD, 1 Hz for vibration velocity VV, 1 Hz for acceleration G) to obtain the RMS value which provides a global measurement of the signal energy. Then, narrow frequency bands relative to the characteristic frequencies of the roller bearing elements and narrow frequency bands relative to the engine rotation and the spindle rotation frequencies were used to calculate the corresponding RMS values of the signal parameters. The fourth parameter, rolling element defect RED, is a specific measurement obtained on the basis of the acceleration parameter G. The RED parameter value is given by the difference between the peak and the RMS values of the vibrating signal coming from the running bearing. Despite the fact that a comparative measurement is always preferable for the detection of a fault, a fault factor above 5 suggests a careful follow up of the bearing. When the measured value exceeds 8, the bearing should be replaced within the shortest possible period of time. The RED parameter is particularly suited for the evaluation of the state of deterioration of the rolling bearing elements. Threshold values for the various signal features were initially selected with reference to data available in the literature [1] and are reported in Tab..
1 Results The results of the vibration measurements carried out during testing of grinding centers # 1 to # 1 running at 6 rpm are shown in Tabs. 6. In Tabs. and, the global vibration measurements for tests with and without grinding wheel, respectively, are reported. In Tabs. 5 and 6, the VV and G signal parameters evaluated in the selected narrow frequency ranges for the various critical components of the grinding center are summarized for testing with and without grinding wheel, respectively. Fig shows the plot of all signal features for global vibration measurements vs. test # for machine # 1 with and without grinding wheel. From Fig., it can be seen that signal features are high for test # (in many cases higher than the threshold value) and low for test # (in many cases lower than the threshold level). This decrease in value occurred for all signal features occurred after the substitution of the spindle of machine # 1. Fig. shows the plot of all signal features for selected narrow frequency ranges vs. test # for machine # 1 with and without wheel. From Fig., a decrease of signal feature values for test # is again observed but signal feature values for test # are higher than the threshold level only in some cases (features VV in frequency range 97.51.5 Hz and G in frequency range 575 Hz). In Fig. a, the RED parameter is reported for all machines in frequency range 1 Hz. From Fig. a, it is possible to verify that machines # 1,, 5 and 7 evidence a RED parameter value higher than the threshold value, except for test # and carried out on machine # 1. This behavior can be correlated with spindle problems on the machines. As a matter of fact, the spindle of machine # 1 was substituted after test # and a lower RED parameter value was verified during tests # and on the same machine, thus indicating a reduction of vibration level. Moreover, after disassembling the spindle, two broken bearings were found: the balls of the bearings were welded by friction on the bearing inner ring and the bakelite ring was destroyed. The fact that the problem in machine # 1 was related to defective conditions of the spindle bearings is also confirmed by the examination of the signal features in the narrow frequency ranges characteristic of the bearings. In Fig. b, the G parameter in frequency range 575 Hz (bearing cage) for all tests on all machines is reported. From Fig. b, it can be seen that the G value for machines # 1,, 5 and 7 is actually higher than the threshold value, except for tests # and carried out on machine # 1. An acceptable value for the G parameter in frequency range 575 Hz was obtained only upon substitution of the spindle of machine # 1 after test #. Figs. a and b also show that all other tests on all other machines evidence acceptable values of the RED and G parameters. Future work Threshold values for the principal signal features will be identified for the frequency ranges of interest for the critical components of the grinding machine in order to create a knowledge base for a decision support system to be applied to the online monitoring of grinding centers. A direct relationship between vibration levels noticed with reference to kinematic system knots and the geometrical and functional degraded state of the mechanical components will be sought for. References [1] Goldman, S., 1991, Vibration spectrum analysis, Industrial Press Inc., New York [] Buzzi, L., 199, Il controllo delle vibrazioni nella manutenzione predittiva, CEMB S.p.a., Mandello del Lario, Como [] Predictive Maintenance Tool, Framatome Diagnostic, Lyon, Feb. 1991
1 Component Typical frequency (Hz) Inner race 196. Outer race 177. Ball 18. Cage 7. Motor 9.6 Spindle 1. Tab.1 Typical frequencies for machine components. Parameter Frequency Low High range (Hz) threshold threshod AD [m] 15. 17. RED 1 5. 8. GA [g] 1.8.1 VV [mm/s] 1 1.5. VV [mm/s] 9.5 95.. VV [mm/s] 97.5 1.5.15 GA [g] 5 75.5 GA [g] 175 155.5 GA [g] 17 175.5 GA [g] 195 1975.5 Tab. Signal feature threshold values. Machine Id. AD (m) RED GA (g) VV (mm/s) A 1 165. 9.7.87.175 1. 9..15 1.757 71.6 1..69 1.79 1. 1.9.6 1.798 B 1 1.6 1.1 17......69.78.69.5 1.78.78 1.76 1.58 C 1 1.5.91 1.89 1.5.1 15.8 1.5.87 11.7.56 D 1 166.5 6..9.75 E 1 18. 18. 1. 9.7.9.96.77.15 1.918 1.96 F 1 16.7.9.5 1.87 G 1 17. 1.7.9.18 H 1 171. 5.7.5 1.951 L 1 169..8.5.9.59 Tab. Wide frequency band vibration measurements during testing with grinding wheel. 1.8 Machine Id AD (m) RED GA (g) VV (mm/s) A 1 16. 8.8.8 1.9 16.9 9.8.86 1.818 85.9.5 1.11 15...61 1.878 B 1 C 1 D 1 E 1 F 1 G 1 H 1 I 1 J 1 16. 17.7 17. 15.5 18.1 165.8 1.1 18.8 171.5 156. 16.1 16.9 168. 19.5 15.7 165.6 19. 1. 168. 15. 17.8 155. 15. 1..1.. 1.5 1.5 1.5 6.7 6. 5.8 1. 1.6 1.5.1.8 11.1 1. 1..1.6 5..1.9..5.1.9.1.71.79.58.59.6.95.9.5..75.55.68.5.51.81.66.6....1.77.15.87.8.71 1.7 1.7 1.8 1.618 1.95 1.77 1.8 1.965 1.71 1.658.9 1.771 19.87 1.8 1.69.99 1.97 1.69 1.9 1.7 1.6 1.967 1.7 1.6 1.86 1.7 1.665
1 K 1 L 1 156. 18.1 18.5 17. 155. 19. 6..5.5..8..78.71.9.8.8. Tab. Wide frequency band vibration measurements during testing without grinding wheel..1 1.718 1.76.11 1.81 1.656 Component Motor Spindle Cage Ball Outer race Inner race Freq. band (Hz) 9.597.5 97.51.5 575 175155 17175 1951975 Parameter VV (mm/s) VV (mm/s) GA (g) GA (g) GA (g) GA (g) Machine A 1 Machine B 1..18...19.16.7.8.61.1.7.9.6.1.97.17.1.5.9..18...7..8..8.8..9 Machine C 1.7......111 1.57.171.181.161.1.61.89.19..1 Machine D 1.6.8.6..8 Machine E 1.16.15.5..116.95.1 Machine F 1..156..1.9.16.18. Machine G 1.15.6.8 Machine H 1.5.8.8.. Machine L 1.19.6.18.. Tab. 5 VV and GA parameters in selected narrow frequency ranges during testing with grinding wheel. Component Motor Spindle Cage Ball Outer race Inner race Freq. band (Hz) 9.597.5 97.51.5 575 175155 17175 1951975 Parameter VV (mm/s) VV (mm/s) GA (g) GA (g) GA (g) GA (g) Machine A 1 Machine B 1 Machine C 1 Machine D 1 Machine E 1 Machine F 1 Machine G 1 Machine H 1 Machine I 1 Machine J 1 Machine K 1 Machine L 1.6.8.18.15.61..111.68 5.67.185.18.17.169.15.57..7.1.9.9..18.117.56.59.7.75.61.9.576.559.9..1.17.1.118..51.8.11.7.11.79.11 6.75.6.9.65.1.76.115.15.16.51.7.1.88.57.7.1.6.68.71.1.9.1.17.1.1.15.97.51.7.1..15.17.1..8.85 1.56.8.9.6.8.7.16..18.8.16.79.8..8.9.19.15.1.11.1.5..9.8..6.....168.....1..1.......1...7..9.18.9.9.9.17.7.1..9.7.58.15.15.15.7..66.7..11.9..115.8..19....5.51.75.7...
15 Tab. 6 VV and GA parameters in selected narrow frequency ranges during testing without grinding wheel. Grinding machine Accelerometer Analyser FFT Computer Fig. 1 Scheme of the monitoring system. Fig. Scheme of the monitoring system. Fig. Signal detection and analysis system.
16 AD [mm] GA [g] 15 1..15.1.5 1 (a) 1 RED 1 8 1 (b) (c) (d) Fig. Machine B. Signal features vs. test # for wide freq. band: (a) AD, (b) RED, (c) GA, (d) VV. Black circle: no wheel; white square: with wheel. Dotted lines indicate upper and lower thresholds. VV [mm/s].5.5 1.5.5 1 VVM [mm/s] GAC [g]....1.15.1.5..1 GAOR [g]. 1 (a) 1 (c) 1 VVS [mm/s] GAB [g]..15.1.5...1..1 GAIR [g]. 1 (b) 1 (d) (e) (f) Fig. Machine B. Signal features vs. test # for narrow freq. ranges: (a) VV motor, (b) VV spindle, (c) GA cage, (d) GA ball, (e) GA outer r., (f) inner r. Black circle: no wheel; white square: with wheel. Dotted line indicates threshold level. 1
17 1 1 1 8 6 RED 1 A B C D E F I H G Machine Id. J K L (a)..15.1 GA [g].5 1 A B C D E F I H G Machine Id. J K L (b) Fig. 1 Signal features (a) RED in the wide frequency band and (b) GA in frequency range 575 Hz (spindle bearing cage) for tests all machines without wheel. Continuous lines indicate thresholds.