RESEARCH ON CONDITION MONITORING OF SPEED REDUCER OF INDUSTRIAL ROBOT WITH ACOUSTIC EMISSION Xiaoqin Liu, Xing Wu, Chang Liu and Tao Liu Faculty of Mechanical and Electrical Engineering, Kunming University of Science and Technology, Kunming, Yunnan, P.R. China E-mail: liuxqsmile@gmail.com IMETI 2015 SM5005_SCI No. 16-CSME-69, E.I.C. Accession 3955 ABSTRACT Articulated industrial robots are widely used for factory automation, e.g., the car manufacture industry. As other machines, the transmission system of industrial robot is prone to fail after long term operation. Speed reducer is the key component of the transmission system, and it is important to detect its incipient faults to avoid breaking down. However, until now, few techniques have been proposed to diagnose the reducer without disassembly. Our research on monitoring and diagnosing the speed reducer of industrial robot is reported in this paper. The technique combines inspection method of acoustic emission and signal processing of vibration signal. The characteristics of acoustic emission signal and their connections with mechanical parts of robot reducer have been studied. A defect in the rolling bearing was detected on a welding robot by this technique and confirmed in disassembly. Keywords: acoustic emission; articulated robot; condition monitoring; diagnosis. ÉTUDE DU SYSTÈME DE SURVEILLANCE DES CONDITIONS DU RÉDUCTEUR DE VITESSE DES ROBOTS INDUSTRIELS À ÉMISSION ACOUSTIQUE RÉSUMÉ Les robots articulés industriels sont largement utilisés pour l automation industrielle, i.e. l industrie automobile. À l instar des autres machines, le système de surveillance des robots industriels est susceptible de se briser après une longue période d opération. Le réducteur de vitesse est l élément clé du système de transmission, et il est important de détecter l imminence d une défaillance pour éviter le bris. Mais jusqu à maintenant peu de techniques ont été proposées pour diagnostiquer le problème du réducteur sans le démonter. Notre recherche sur la surveillance et le diagnostic pour le réducteur de vitesse de robots industriels est exposée dans cet article. La technique est une combinaison de méthode de l émission acoustique et le traitement des indices de signes de vibration. Les caractéristiques des signaux d émissions acoustiques et leur connexion avec les parties mécaniques du réducteur de vitesse ont été étudiées. Une défaillance d un roulement à billes a été détectée sur un robot de soudure par cette technique, et a été confirmé lors du démontage. Mots-clés : émission acoustique; robot articulé; surveillance des conditions; diagnostic. Transactions of the Canadian Society for Mechanical Engineering, Vol. 40, No. 5, 2016 1041
1. INTRODUCTION Articulated industrial robots are widely used for factory automation, such as vehicle and mechanical parts manufacture, packaging, painting and product inspection, all accomplished with high endurance, speed and precision. Industrial robots are essential to improve producing efficient and product quality, and to liberate employees from monotony and harmful labor works. Each joint of an articulated robot is driven by an electric motor via a speed reducer. The speed reducer is going to fail after long term operation, just like common gearboxes in other transmission systems. As industrial robots are often installed on production lines, it is widely desired to obtain the health status of the robots to avoid breaking down. For this reason, fault detection of industrial robots has been addressed by many researchers. But most of the methods proposed are aim to locate the fault from all kinds of possible faults when it happened [1, 2]. Erik Olsson proposed a Case-Based Reasoning Approach using acoustic signals to diagnosis faults in industrial robots [3]. The approach was designed to perform quality check in testing environment, which may not work well in most factories with strong and complicate background acoustic noises. Though industrial robots have served factories for decades, practical monitoring method for its reducer has long been absent. Vibration analysis is today most commonly practiced for machinery monitoring and diagnosis, and is a key component to achieve predictive maintenance. The methodology and theory of vibration have been well developed, which ensure its success on many kinds of machines (fans, motors, pumps, and gearboxes, etc.). The frequency of vibration signal ranges from several Hertz to several thousands Hertz. But for articulated robot in normal operation, step motors generate strong vibration on the arms of the robot, and the vibration shares the same frequency band with vibration signal from faults in the speed reducer. The comparably weak signal of faults is covered up by the operational vibration, which increases the difficulty to detect the faults. Another drawback of articulated robot monitoring is the changing speed during its swing movement. Acoustic emission (AE) is the phenomenon of radiation of acoustic (elastic) waves in solid that occurs when a material undergoes irreversible changes in its internal structure, for example, as a result of crack formation or plastic deformation. Irreversible release of energy on the defect spot of contact surfaces can also generate AE signal due to impact on the spot. AE has been applied to rotary machinery diagnosis, bearing [4, 5], gear box [6], etc., and has be regarded as an effective way to detect initial faults. The frequency band of Acoustic emission is usually between 100 khz and 1 MHz, above the frequency band of vibration signal. This prevents AE signal from coupling with the vibration of robot body, and accurate diagnosis can be achieved. Our research on applying AE technique to diagnose speed reducer of industrial robot is presented in this paper. This technique combines AE inspection and signal processing borrowed from vibration analysis. Characteristics of AE signal and their connections with mechanical parts of speed reducer have been studied. Industrial field test was carried out to verify the capability of the method. The paper is organized as follows. Section 2 describes the inner structure and rotating speed transmission of RV reducer, which is the kind of reducer used on most industrial robots. Section 3 details the procedure of diagnosing fault using AE technique, including AE measurement, rotating speed extraction, and fault identification. Finally, Section 4 concludes this paper with summary and conclusions. 2. FAULT CHARACTERISTIC FREQUENCIES OF RV REDUCER RV reducers are used on most articulated industrial robots. The reducer is a two-stage system combining epicyclic gearing and cycloidal drive. Thanks to the two-stage system, RV reducer achieves high speed reducing ratio, accurate motion transmission, and heavy load capability with limited space occupation. The excellent rigid between moving components and compact dimensions minimize the transmission loss of AE. 1042 Transactions of the Canadian Society for Mechanical Engineering, Vol. 40, No. 5, 2016
Fig. 1. Transmission scheme of RV reducer. The RV reducer on the robot inspected was produced by Nabtesco[7]. The transmission scheme of the RV reducer is shown in Fig. 1. The motion in the RV reducer includes: 1. Mesh between the input gear and two or three planetary gears. 2. Rolling of bearings on the crank shafts of planetary gears. 3. Mesh between the cycloidal disk and ring gear through pins. For articulated robots, the ring gear engaging with cycloidal disks is part of the case which is fixed to the support arm, and cycloidal disks serve as the output shaft and connect to the arm to be driven. In this configuration, cycloidal disks rotate one tooth of ring gear when planetary gear rotates one cycle around its own center line. Thus, cycloidal rotational speed n p around its own center is n p = Z c n o (1) where n o is the output shaft rotational speed, Z c represents the number of teeth of ring gear. The teeth on cycloidal disk are one less than on the ring gear. During the time that planetary gear rotates n p cycles on its axis, we assume that planetary gear center line rotates n o cycles to the axis of the sun gear (input gear). If a counter-direction rotation n o is applied to the reducer, the center line of the planetary gear will be static to the input shaft. If the input gear rotational speed is n i, we have n i n o = n p Z p Z i (2) where Z p is the number of teeth on the planetary gear, Z i is the number of teeth on the input gear, and the transmission ratio R of the reducer is given by R = n o 1 = n i Z c Z p /Z i + 1 (3) By now, if the teeth number on all the gears and input speed are known, the rotational speeds for all the resolving components can be calculated. Transactions of the Canadian Society for Mechanical Engineering, Vol. 40, No. 5, 2016 1043
2.1. Fundamental Fault Frequencies of Bearings Rolling element bearings are used to support the rotating of planetary gears and output shaft. There are usually four bearings on each planetary crankshaft. Those bearings are the most fragile parts in the reducer. Rolling bearing faults can be identified by fundamental fault frequencies, which are frequencies that rolling bearings generate when rollers pass over a surface anomaly on either the roller or the raceway. These frequencies are a function of the bearing geometry, including pitch diameter and roller diameter, and the relative speed between the two raceways. The fundamental fault frequencies can be calculated using the following equations: BPFI = NF 2 BPFO = NF 2 FTF = F 2 [ BSF = PF 1 2B ( 1 + Bcosθ P ( 1 Bcosθ P ( 1 Bcosθ P ) ) ) ( ) ] Bcosθ 2 where BPFI is the ball passing frequency of inner race (Hz), BPFO is the ball passing frequency of outer race (Hz), FTF is the fundamental train frequency (Hz), BSF is the ball spin frequency (Hz), N is the number of balls, F is the shaft frequency (Hz), B is the ball diameter (mm), P is the pitch diameter (mm), and θ is the contact angle. The value of FTF with respect to shaft rotational speed usually falls between 0.38 and 0.42. For this reason, fault frequencies may be estimated when bearing geometry is not known, as follows: BPFO is approximately 0.4NF; BPFI is approximately 0.6NF; and BPF is approximately 0.2NF. 2.2. Gear Mesh Frequency The gear mesh frequency is the frequency at which gear teeth mate together between the input gear and the planetary gear. Gear mesh frequency GMF can be given by P GMF = Z i F i (8) where GMF is the gear mesh frequency, Z i is the number of teeth on input gear, and F i is the shaft speed of input gear. 2.3. Cycloidal Disk Mesh Frequency The cycloidal disk mesh frequency CMF is the frequency at which cycloidal disk mesh with ring disk, and can be calculated as CMF = Z c F c (9) where CMF is the cycloidal disk mesh frequency, Z c is the number of teeth on cycloidal disk, and F c is the shaft speed of cycloidal disk. 3. FAULT DIAGNOSIS BASED ON ACOUSTIC EMISSION 3.1. Measurement Method Two wide-band AE sensors (WD, PAC, US) were attached to the case of one of the speed reducers on an industrial robot by magnetic holders to detect AE signal from the inner components, shown in Fig. 2. Signals (4) (5) (6) (7) 1044 Transactions of the Canadian Society for Mechanical Engineering, Vol. 40, No. 5, 2016
Fig. 2. Acoustic emission sensors installed on a robot. Fig. 3. Acoustic emission signal from RV reducer. from the sensors were sampled by a high-speed digitizing card (PCI-9846H, ADLink, TAIWAN) installed on a personal computer. The arm driven by the reducer was programmed to rotate forward and backward for several times, and data were acquired during the swing movement. All six articulates of the robot were measured one by one. Vibration signals were also acquired synchronously with an accelerometer. The raw AE signal from the third axis in one swing period is shown in Fig. 3. The signal strength grows when rotating speed increases, and mutes when slowing down and changing direction. This undulating of signal can also be found on other axes. Only on this axis, AE impulses can be recognized clearly which may imply a defect in the reducer. 3.2. Rotating Speed Extraction During the measurement (and in real operation), the arm of the robot swung back and forth, and the rotating speed changed all the time. The signal from the defected component also changed with the rotating speed, thus it was difficult to identify the defected component by frequency analysis. Transactions of the Canadian Society for Mechanical Engineering, Vol. 40, No. 5, 2016 1045
Fig. 4. Gabor time-frequency spectrum of vibration signal. Since the rotation is linked to the angular part, it is natural to use the angle as a sampling variable instead of time. Therefore, the signal becomes synchronized to the machine cycle. In articulated robot, all the rotating components of the transmission system are enclosed except the arm, so it is impossible to install a keyphasor to get the angular reference. The only choice is the computed order tracking method which samples the signal at a constant time interval, and then uses software to resample the data at constant angular increments [8]. The key step of this method is to extract the changing of the rotating speed from the signal. As AE signal is generated from impact on defected surface, and is usually shown as a series of impulses, from which the rotating speed is difficult to extract. So the vibration signal acquired synchronously with AE was used instead. In this paper, the speed was extracted from the time-frequency spectrum of Gabor transform of the vibration signal. Gabor transform is the short-time Fourier transform with Gaussian window. The Gabor transform of signal x(t) is given by G x (t, f ) = e π(τ t)2 e j2π f τ x(τ)dτ (10) The time-frequency spectrum of vibration signal from the third joint is shown in Fig. 4. The speed running up and down can be recognized on the ridge of the time-frequency map [9]. The speed ratio curve obtained is shown in Fig. 5. Constant angular sampling of AE signal was obtained by resampling the raw signal with time intervals proportional to the speed curve. The resampled signal is shown in Fig. 6. The length of the signal shrank significantly compared to the raw signal because lower rotating speed was elevated to top speed at 1.6 s in Fig. 5. 3.3. Fault Diagnosis Envelope spectrum is the tool to identify the source of the impulses in the AE signal. The signal was firstly filtered by a band-pass filter, then the envelope was calculated by Hilbert transform. As the accurate geometric parameters cannot be obtained from the producer of the reducer, the fault frequencies can only be calculated approximately. For the bearings on the crankshafts of planetary gears, there are 18 rollers in each bearing, and the inner races rotates at 8.1 Hz, and BPFO, BPFI, BSF are 58, 87, and 29 Hz, respectively. 1046 Transactions of the Canadian Society for Mechanical Engineering, Vol. 40, No. 5, 2016
Fig. 5. Speed ratio extracted from vibration signal. Fig. 6. Resampled acoustic emission signal. In the envelope spectrum of the resampled signal, the first frequency band can be found centering at 56 Hz, which is close to BPFO of 58Hz. The second frequency band centering at 107 Hz is about twice of the first frequency band and should be the second harmonic of the BPFO. Thus, the presence of a bearing fault can be concluded. After disassembling the RV reducer, a large pit was found on the outer race of one of the rolling bearings. This confirmed the diagnosis from the analysis. The large area of the pit also damaged the regularity of the signal, and flattened the BPFO on the envelope spectrum. 4. CONCLUSIONS Acoustic emission technology was applied to detect mechanical faults in the speed reducer of articulated industrial robot. The result shows that good signal quality ensures the success of detecting and locating the defect on the rolling bearing. One reason of the success of AE technology is the compact size and high stiffness of RV reducer, which ensure the effective transmission of AE signal from defect spot to the sensor. The other reason is the high Transactions of the Canadian Society for Mechanical Engineering, Vol. 40, No. 5, 2016 1047
Fig. 7. Envelope spectrum of resampled AE signal. Fig. 8. The defect on the outer race of one of the bearings. frequency band of AE signal, which is high above the frequency band of vibration signal, and avoids the coupling of vibration from the robot arms. The acoustic emission is a well-developed technology with off-the-shelf instruments and abundant analysis methodologies. Its ability of diagnosing faults on other components of robot speed reducers will be studied in our future research. ACKNOWLEDGEMENTS This research was financially supported by research grants from National Natural Science Foundation of China (No. 51465022, and No. 51265018) and a research grant from Yunnan Provincial Department of Education (No. ZD2013004). 1048 Transactions of the Canadian Society for Mechanical Engineering, Vol. 40, No. 5, 2016
REFERENCES 1. Secchi, C. and Visioli, A., On the fault detection and isolation of industrial robot manipulators, in Proceedings of 7th Symposium on Robot Control, Wroclaw, Poland, pp. 473 478, September 1 3, 2009. 2. Van, M., Kang, H.J. and Ro, Y.S., A robust fault detection and isolation scheme for robot manipulators based on neural networks, in Proceedings of the 7th International Conference on Advanced Intelligent Computing, Zhengzhou, China, pp. 25 32, August 11 14, 2011. 3. Olsson, E., Funk, P. and Bengtsson, M., Fault diagnosis of industrial robots using acoustic signals and casebased reasoning, in Proceedings of Advances in Case-Based Reasoning, Madrid, Spain, pp. 686 701, August 30 September 2, 2004. 4. Mba, D. and Rao, R.B., Development of acoustic emission technology for condition monitoring and diagnosis of rotating machines: Bearings, pumps, gearboxes, engines, and rotating structures, The Shock and Vibration Digest, Vol. 38, No. 1, pp. 3 16, 2006. 5. Al-Ghamd, A.M. and Mba, D., A comparative experimental study on the use of acoustic emission and vibration analysis for bearing defect identification and estimation of defect size, Mechanical Systems and Signal Processing, Vol. 20, No. 7, pp. 1537 1571, 2006. 6. Toutountzakis T., Tan C.K. and Mba, D., Application of acoustic emission to seeded gear fault detection, NDT & E International, Vol. 38, No. 1, pp. 27 36, 2005. 7. Nabtesco Motion Control Inc. RV-C series gearbox: Features. http://www.nabtescomotioncontrol.com/products/rvc-series-gearbox/, accessed September 28, 2015. 8. Fyfe, K.R. and Munck, E.D.S., Analysis of computed order tracking, Mechanical Systems and Signal Processing, Vol. 11, No. 2, pp. 187 205, 1997. 9. Guo, Y., Qin, S, Tang, B. and Ji, Y., Order tracking of rotating machinery based on instantaneous frequency estimation, Journal of Mechanical Engineering, Vol. 39, No. 3, pp. 32 36, 2003. Transactions of the Canadian Society for Mechanical Engineering, Vol. 40, No. 5, 2016 1049