978-1-5090-4281-4/17/$31.00 ©2017 IEEE

Fibre Optic Monitoring of Induction Machine Frame Strain as a Diagnostic Tool

A. Mohammed1, N. Sarma1, S. Djurović1

1School of Electrical and Electronic Engineering. University of Manchester, Manchester, UK

Abstract — This paper investigates the application of Fibre Bragg sensing for monitoring of induction machine frame strain. The underlying purpose is to examine the feasibility of frame surface attached fibre optic sensing to be used for condition monitoring purposes, i.e. for recognition of signatures characteristic of healthy and/or faulty machine operating conditions. To this end an experimental study is reported that compares the fibre optic monitored circumferential strain signal spectral content with that of synchronously monitored frame acceleration and shaft torque signals on a laboratory induction machine operating with and without electrical fault. It is shown that the signature of electrical unbalance and fault, as well as that of mechanical effects on the shaft can be effectively recognised in the frame strain signal monitored by the fibre optic strain sensor.

Keywords—induction machine frame strain; embedded sensing; condition monitoring; Fibre Bragg grating sensors; winding open-circuit fault

I. INTRODUCTION

Electric machines, and induction motors in particular, are a core element of modern industrial applications and are often referred to as the workhorse of industry. With ambitious global scale plans for electrification of transport and development of renewables the use of electric motors is set to further increase in the near future. This increased use will be closely followed by a step change in availability targets for electric motors, where, in addition to improved design, increased availability is largely expected to come from a step change in the efficacy of condition monitoring (CM) and diagnostic techniques.

The present state-of-the-art in CM of electric machines is largely based on conventional CM solutions that rely on monitoring signals such as frame acceleration and motor current and/or undertaking end winding/external thermal monitoring [1]. These techniques can impose a number of significant constraints to effective diagnostics, ranging from sensing point placement to EMI immunity and cost of associated sensors [1]. Fibre optic sensing has emerged as a promising alternative for electric motor condition monitoring and its application for machine mechanical and thermal monitoring [2-6] is attracting increased interest. Fibre Bragg grating (FBG) sensors in particular offer the advantage of electromagnetic interference (EMI) immunity, low cost and are small in size and lightweight and hence have a considerable potential to be applied in electric machines to deliver improved monitoring solutions.

This work investigates the feasibility of strain signal monitoring of an induction machine frame for CM and diagnostic use; the reported work is a proof of concept study undertaken with the aim to examine the operational potential of the proposed method. For this purpose the frame of a laboratory induction machine is instrumented with an FBG strain sensor operated by a commercial interrogator device. The test machine is also equipped with commercial vibration and torque measuring platforms to enable direct correlation and validation of the FBG obtained strain spectrum with conventionally used vibration and torque signal spectra. The test motor is designed to enable experimental emulation of winding fault and used to perform a range of healthy and faulty machine operation experiments.

The frame strain and vibration and the shaft torque signal spectra were synchronously recorded in the experiments. Their contents were then examined to identify artefacts that enable recognition of known spectral signatures of electrical unbalance, winding fault and mechanical effects on the rotor shaft. It is shown that the straightforward application of a low cost strain sensing FBG sensor can provide unambiguous recognition of relevant diagnostic signatures. The proposed method could therefore be of potential interest in development of reduced cost/improved robustness monitoring alternatives to conventional vibration sensing diagnostic platforms for rotating electric machinery.

II. PRINCIPLESUSE OF FIBRE BRAGG GRATING SENSOR STRAIN MEASUREMENT

An FBG sensor head is a small segment imprinted into a single mode fibre in which the fibre core refractive index is modulated by exposure to an interference pattern of ultraviolet light. The FBG sensor size depends on the applied optical fibre design and is typically very small; standard optical fibres consist of an inner core that is 9 µm in diameter and cladded with a layer of glass to a 125 µm diameter, while FBG sensor head lengths typically range from 3 to 20 mm [7]. The basic operating principle of FBG sensing is to monitor the sensor narrowband reflected wavelength after injecting broadband light into the sensing optical fibre. The wavelengths reflected by the FBG structure change with the variation in the temperature and strain it is exposed to. The FBG head reflected wavelength is known as Bragg wavelengthλΒ, and can be defined as [7]:

λΒ=2 Λ neff (1) This work was supported in part by the UK Engineering and Physical Sciences Research Council (EPSRC) HOME-Offshore: Holistic Operation and Maintenance for Energy from Offshore Wind Farms Consortium under grant EP/P009743/1.

Where: Λ is the FBG gratings interval and neff is the effective index of the optic fibre core. The response of the FBG sensor to strain arises due to the change in the grating period (the physical elongation of the sensor), and the change in the reflective index due to photo-elastic effects. The Bragg wavelength shift due to applied strain can be expressed as [8]:

Δ λΒ=λB(neffd Λdε

+ Λdneff

dε )ε (2)

Where ε is the applied strain. For FBG strain sensors imprinted in standard silica fibre the sensor sensitivity to strain is ≈ 1.2 pm/µstrain [8].

III. TEST SYSTEM DESCRIPATION

For experimental research purposes the frame of a 30kW/50Hz 4-pole induction machine was instrumented with an FBG strain sensor. The test machine is a known design whose torque and vibration spectral content under heathy and electric fault conditions was previously researched [9-11]. The test machine therefore provides an effective vehicle for examining the feasibility of diagnostic signature recognition in the frame strain spectrum, through its correlation with the known signatures in the vibration and torque signals. The machine is equipped with a Bruel&Kjaer (B&K) PULSE vibration sensing platform operating a B&K Deltatron DT439 accelerometer mounted on the top of the machine load side end plate. Wide band monitoring of the shaft torque signal is achieved by mounting the motor on a precision Kistler 9281B force table [12]. An FBG sensor with a 5mm head length imprinted in a 0.125 mm diameter polyamide coated fibre was attached to the motor drive end frame surface. A simplified schematic of the test system is shown in Fig. 1.

To achieve localised non-invasive strain measurement in a material using an FBG sensor the sensor needs to be fixed to the material’s surface using a suitable adhesive. In this work an FBG sensor was installed in the circumferential direction on the motor frame end cap following the procedure proposed in [13]: the frame surface was first stripped of paint and then

brushed, sanded and cleaned with Isopropyl alcohol before the FBG sensor was attached with Kapton tape, as illustrated in Fig. 2a.

The FBG sensor was interrogated by a commercial interrogator platform (Smart Fibres SmartScan04 [14]) in the tests. The experiments were undertaken by energising the test machine from the grid via a three phase variable transformer and loading it with a speed controlled DC motor. The induction machine stator windings were rewound in previous work to enable experimental emulation of winding fault conditions [9, 10]; for the purpose of this study the machine was run with a healthy or a stator-open circuit fault winding configuration, as shown in Fig. 2b. The shaft torque, frame radial acceleration and strain signals were synchronously measured in experiments and post processed using the Fast Fourier Transform (FFT) routine to enable spectral content analysis. The machine shaft speed was measured by means of an incremental encoder.

5 mm FBG Head

Machine drive end cap

FC/APCConnector

Kaptontape

Optical fibre

(a) FBG sensor installed on the test motor load side end plate.

A1 A2

B2

B1

C2

C1

A1 A2

A1 A2

B2

B1

C2

C1

No winding fault Open-circuit fault

(b) Machine stator winding configuration.

Fig. 2. Sensor placement and winding configuration.

DC MachineWRIM

Transformer

Grid

FORCE TABLE

B&K Pulse

Platform

Acce

lero

met

er NI DAQ

Stat

or Rotor

FBG

SmartFibreSmartScan 04

Encoder

Fig. 1. A simplified schematic of the test rig.

IV. RESULTS AND DISCUSSION

A range of experiments were undertaken on the laboratory test rig to ascertain the efficacy of FBG strain sensing in enabling recognition of diagnostic signatures observed in the torque and frame vibration signal spectra. The test machine was operated at two different operating speed points for the purpose of the study: 1590 rpm and 1530 rpm. At each operating point the machine was first operated in healthy conditions (i.e. with no winding fault) and then in faulty conditions (i.e. with a stator winding open-circuit fault).

The synchronously recorded spectra of the torque, vibration and strain signals for healthy and faulty conditions at 1590 rpm are shown in Figs. 3 and 4, respectively. The spectra are presented in a 0-500 Hz bandwidth to clearly illustrate the lower frequency spectral content of diagnostic interest [9, 10]. Previously published research enables the correlation of the relevant diagnostic information in the torque and frame vibration signals with the motor operating conditions [1, 9, 15] and the prediction of individual spectral frequencies of interest. The closed form spectral signature definitions assume machine operation with unbalanced stator supply, as is typical in practical applications, and are summarized per signature origin in Table I, where: k = 0,1,2,3... and relates to air-gap field pole

numbers, s is the rotor slip, p is the pole pair number, fs is the supply frequency and nm is the rotor mechanical speed in rpm [1, 9, 15]. The frame circumferential direction strain signal spectrum was examined for components at these frequencies, as the forces giving rise to known vibration and torque pulsations will generally give rise to strain in the machine core and housing structure at identical frequencies [1, 16].

TABLE I. ELECTROMAGNETIC TORQUE/FRAME VIBRATION SIGNAL SIGNATURE FREQUENCIES DEFINITION [1,9,15]

The frequencies of relevant spectral components originating from mechanical effects on the shaft (in blue in Figs. 3 and 4) and the electromagnetic effects when operating in healthy winding conditions (in green in Figs. 3 and 4) and with a winding open-circuit fault (in red in Fig. 4) are identified in the data shown in Figs. 3 and 4. The experimental results demonstrate a high level of consistency between the spectral signatures observed in the three monitored signals. The FBG sensor strain measurement is seen to provide clear recognition of the relevant diagnostic signatures contained in the vibration and torque signals with a noticeable decrease in the signal to noise level at higher frequencies in the examined spectral band. ………………………………………………...

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Torq

ue [N

.m]

10-3

100

104

292.1Hz

318.6Hz

371.6Hz418.3Hz

477.9Hz26.55Hz

53.1Hz

79.65Hz

100Hz106.2Hz132.8Hz

185.9Hz

218.8Hz

(a) Measured shaft torque signal spectrum

Electromagnetic Origin(Supply assumed unbalanced) Mechanical

OriginBalanced Windings Stator Winding Fault

|6 k (1−s ) f s|

|2 ± 6 k (1−s) f s|

|kp

(1−s ) f s||2 ± k

p(1−s ) f s|

k∗nm

60

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Acc

eler

atio

n[m

/s2 ]

10-4

10-1

103

292.1Hz219Hz

319Hz

372Hz

418Hz478Hz

186Hz53Hz

79.65Hz

100Hz26.4Hz 106Hz

133Hz

(b) Measured frame vibration signal spectrum

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Stra

in [

]

10-3

10-1

101

53Hz

26.5Hz 100Hz79.5Hz 106.1Hz

132.5Hz

185.6Hz

218.4Hz

292.1Hz

318.1Hz

371.2Hz

418.2Hz477.2Hz

(c) Measured frame strain signal spectrum

Fig. 3: Signals spectra for machine operation with balanced windings and unbalanced supply, 1590 rpm.

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Torq

ue [N

.m]

10-1

101

104

418.3Hz371.6Hz

292.1Hz

100Hz

79.59Hz

59.33Hz 159.4Hz 218.6Hz259.3Hz

318.6Hz

26.61Hz

53.1Hz 106.2Hz185.9Hz 477.9Hz

132.8Hz

(a) Measured shaft torque signal spectrum

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Acc

eler

atio

n[m

/s2 ]

10-4

10-1

103

53Hz

100Hz

186Hz

219Hz 319Hz

478Hz26.7Hz59.5Hz

79.7Hz

159Hz 259Hz

373Hz

419Hz

292.1Hz133Hz

105Hz

(b) Measured frame vibration signal spectrum

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Stra

in [

]

10-3

10-1

102

100Hz

53Hz26.5Hz

59.05Hz106Hz

132.6Hz

159.1Hz 259.1Hz318.1Hz 418.3Hz

477.2Hz371.1Hz79.4Hz 185.6Hz

218Hz291.6Hz

(c) Measured frame strain signal spectrum

Fig. 4: Signals spectra for machine operation with open-circuit fault and unbalanced supply, 1590 rpm.

To further highlight this observation a detailed view of the measured strain signals for healthy and faulty machine operation in a 250 Hz bandwidth is shown in Fig. 5 where relevant spectral components are identified according to their origin. It can be seen that the shaft unbalance effects are clearly reported by the FBG strain measurement at integer multiples of the rotational speed (i.e. ≈26 Hz, ≈52 Hz etc.); the inherent spectral components of electromagnetic origin (i.e. ≈100 Hz, ≈218 Hz etc.) and those exhibiting an increase in magnitude with from presence of winding fault [9] (i.e. ≈59 Hz, ≈159 Hz etc.) in the examined motor design are also clearly recognisable in the measured data.

To study the consistency of strain sensing performance the recorded torque, vibration and strain signals spectra were also cross-examined in healthy and faulty operating conditions at an operating speed of 1530 rpm and are shown in Figs. 6 and 7, respectively. The spectral components of interest predicted by the equations shown in Table I are identified per origin in the measured spectra using an identical colour scheme to that applied in Figs. 3-5. While the measured signals spectra are generally significantly noisier with the presence of winding fault, as is expected, the strain spectrum is seen to provide a clear measurement of the relevant pronounced spectral components in these conditions, at a level comparable to that seen in the vibration signal measured by a commercial monitoring system. In general, a high level of consistency can be observed between the spectral signatures recorded in torque and vibration signals and that reported by the FBG strain measurement, suggesting that the change of operating speed does not have a perceivable effect on the consistency of diagnostic signature recognition in the strain signal.

0 50 100 150 200 25010-4

10-1

102

Frequency [Hz]

Stra

in (

)

212.1Hz mechanical

218Hz electrical

185.6Hz mechanical

159Hzfault

106.1Hzmechanical

132.5Hzmechanical

79.6Hzmechanical

100Hz electrical26.5Hz

mechanical 53Hzmechanical

59.9Hz fault

Fig. 5: A detailed view of the measured strain spectra for healthy (shown in blue) and faulty (shown in red) machine operation, 1590 rpm

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Torq

ue [N

.m]

10-3

100

104

100Hz

76.6Hz102.2Hz

127.6Hz

178.8Hz

280Hz 357.5Hz206.5Hz

306.4Hz

406.3Hz 484.5Hz

51.04Hz25.5Hz

(a) ) Measured shaft torque signal spectrum

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Acc

eler

atio

n [m

/s2 ]

10-4

10-1

102

485Hz

406Hz

307Hz

281Hz 358Hz76.6Hz102Hz

127.9Hz

179Hz206Hz100Hz50.9Hz

25.5Hz

(b) Measured frame vibration signal spectrum

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Stra

in [

]

10-3

10-1

101

404.7Hz

178.5Hz206Hz

305.9Hz25.48Hz76.45Hz

50.97Hz 100Hz

102Hz 127.5Hz280Hz 356.9.97Hz

484.5Hz

(c) Measured frame strain signal spectrum

Fig. 6: Signals spectra for machine operation with balanced windings and unbalanced supply, 1530 rpm.

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Torq

ue [N

.m]

10-2

101

104

76.6Hz

176.6Hz25.51Hz

51.03Hz102.2Hz

100Hz

406.4Hz53.22Hz

127.7Hz

153.2Hz 206.4Hz 306.4Hz

357.5Hz 485Hz

253.2Hz

280Hz

(a) Electromagnetic torque measurement

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Acc

eler

atio

n [m

/s2 ]

10-4

10-1

102

76.6Hz

485Hz102Hz

100Hz

25.5Hz

51.1Hz

128Hz

153Hz 306Hz 407Hz358Hz281Hz

253Hz206Hz177Hz53.1Hz

(b) Vibration measurement

Frequency [Hz]0 50 100 150 200 250 300 350 400 450 500

Stra

in [

]

10-3

10-1

101

25.5Hz51Hz 76.5Hz

100Hz

102.2Hz127.7Hz

153Hz

176.3Hz206.4Hz 252.2Hz 305.9Hz

280.4Hz 357.5Hz

405.7Hz

484Hz

53.22Hz

(c) FBG sensor strain measurement

Fig. 7: Signals spectra for machine operation with open-circuit fault and unbalanced supply, 1530 rpm.

A further illustration of the usability of the frame strain signal can be obtained from examination of its time domain form in faulty conditions. The measured strain signal for test machine operation with a winding fault at 1590 rpm shown in Fig. 8 clearly exhibits an envelope modulation at twice supply frequency (2fs = 100 Hz for the examined test system), as would be expected in a machine operating with stator electrical fault/unbalance that gives rise to frame vibration and hence strain at two times the fundamental supply frequency [1,16]. Relating this to the strain spectrum at the same condition shown in Fig. 4c, it can be seen that the 100 Hz component is the dominant component in the examined spectrum.

Time[Sec]50 50.01 50.02 50.03 50.04 50.05 50.06 50.07

Stra

in[

stra

in]

-10

-5

0

5

101/2fs 1/2fs

Fig. 8: Measured strain signal for machine operation with open-circuit fault and unbalanced supply, 1590 rpm.

V. CONCLUSIONS

This study examines the feasibility of FBG frame strain sensing in induction machines for CM and diagnostic purposes. The measured strain signal spectrum is correlated with the frame vibration signal spectrum that is conventionally used for diagnostic purposes in electric machinery and high levels of consistency were observed in the examined 0-500 Hz spectral band. The results show that the FBG sensor monitored frame strain signal in circumferential direction can enable clear recognition of conventional signature of shaft mechanical unbalance effects and those that are inherently electromagnetic in origin and relate to the presence of supply unbalance or winding fault.

While further research on a larger number of machine designs and operating conditions is required to ascertain the wide generality of the observed phenomena, the presented method is shown to have a considerable potential to provide a straightforward, low cost and EMI immune alternative to conventional accelerometer based frame vibration monitoring for CM and diagnostic purposes. Future work will report results of further studies examining the long term effects of sensors thermal cross sensitivity as well as optimal location of the strain sensing point and sensor bonding.

ACKNOWLEDGMENT

This work was supported in part by the UK Engineering and Physical Sciences Research Council (EPSRC) HOME-Offshore: Holistic Operation and Maintenance for Energy from Offshore Wind Farms Consortium under grant EP/P009743/1.

REFERENCES

[1] P. Tavner, "Review of condition monitoring of rotating electrical machines," Electric Power Applications, IET, vol. 2, pp. 215-247, 2008.

[2] A. Mohammed, S. Djurović, A. C. Smith, and K. Tshiloz, "FBG sensing for hot spot thermal monitoring in electric machinery random wound components," in 2016 XXII International Conference on Electrical Machines (ICEM), 2016, pp. 2266-2272.

[3] A. Mohammad and S. Djurović, "Evaluation of fiber-optic sensing performance for embedded thermal monitoring of electric machinery wound components," in 2016 5th Mediterranean Conference on Embedded Computing (MECO), 2016, pp. 72-76.

[4] Fabian, M., Ams, M., Gerada, C., Sun, T., Grattan, K. T. V, "Vibration measurement of electrical machines using integrated fibre Bragg gratings." Proc. SPIE 9634, Int. Conf. on Optical Fibre Sensors, 2015.

[5] K. M. Sousa, U. J. Dreyer, C. Martelli and J. C. Cardozo da Silva, "Dynamic Eccentricity Induced in Induction Motor Detected by Optical Fiber Bragg Grating Strain Sensors," in IEEE Sensors Journal, vol. 16, no. 12, pp. 4786-4792, June15, 2016.

[6] A. Mohammed, S. Djurović, ’ FBG Thermal Sensing Features for Hot Spot Monitoring in Random Wound Electric Machine Coils’, in IEEE Sensors Journal, vol. PP, no. PP, pp. 1-11, 2017, in press.

[7] A. D. Kersey, M. A. Davis, H. J. Patrick, M. LeBlanc, K. P. Koo, C. G. Askins, M. A. Putnam, and E. J. Friebele, “Fiber grating sensors.” Journal of lightwave technology 15, no. 8 (1997): 1442-1463.

[8] Rao, Yun-Jiang, “In-fibre Bragg grating sensors,” Measurement science and technology 8.4 (1997): 355.

[9] D. S. V.-R. S. Djurović, A.C. Smith, "Investigation of wound rotor induction machine vibration signal under stator electrical fault conditions," The IET Journal of Engineering, vol. 4, pp. 719,729, 2014.

[10] S. Djurović, D. Vilchis-Rodriguez, and A. C. Smith, "Vibration monitoring for wound rotor induction machine winding fault detection," in Electrical Machines (ICEM), 2012 XXth International Conference on, 2012, pp. 1906-1912.

[11] D. S. Vilchis-Rodriguez, S. Djurović, N. Sarma, K. Tshiloz, A. C. Smith, and Y. Wei, "On wound rotor induction machine rotor electrical unbalance recognition using stator frame vibration spectral analysis," in 8th IET International Conference on Power Electronics, Machines and Drives (PEMD 2016), 2016, pp. 1-7.

[12] R. C. Healey, S. Lesley, S. Williamson, and P. R. Palmer, "The measurement of transient electromagnetic torque in high-performance electrical drives," in Power Electronics and Variable Speed Drives, 1996. Sixth International Conference on (Conf. Publ. No. 429), 1996, pp. 226-229.

[13] I. Grabovac, T. Nuyens, C. Davis, Packaging and Mounting of In-Fibre Bragg Grating Arrays for Structural Health Monitoring of Large Structures, Defence Science and Technology Organisation, DSTO-TR-2490, 2010.

[14] www.smartfibres.com/products/smartscan.[15] S. Djurović, D. S. Vilchis-Rodriguez and A. C. Smith, "Supply Induced

Interharmonic Effects in Wound Rotor and Doubly-Fed Induction Generators," in IEEE Transactions on Energy Conversion, vol. 30, no. 4, pp. 1397-1408, Dec. 2015.

[16] Finley, W.R., Hodowanec, M.M., W. G. Holter: ‘An analytical approach to solving motor vibration problems’. Proc. Industry Applications Society 46th Annual Petroleum and Chemical Technical Conference (Cat.No. 99CH37000), San Diego, CA, 1999, pp. 217-232.