Professional Documents
Culture Documents
Health Monitoring and Prognosis of Electric Vehicl
Health Monitoring and Prognosis of Electric Vehicl
Health Monitoring and Prognosis of Electric Vehicl
Research Article
E-mail: manicavasagam2003@yahoo.com
Abstract: Electric mobility has become an essential part of the future of transportation. Detection, diagnosis and prognosis of
fault in electric drives are improving the reliability, of electric vehicles (EV). Permanent magnet synchronous motor (PMSM)
drives are used in a large variety of applications due to their dynamic performances, higher power density and higher efficiency.
In this study, health monitoring and prognosis of PMSM is developed by creating intelligent digital twin (i-DT) in MATLAB/
Simulink. An artificial neural network (ANN) and fuzzy logic are used for mapping inputs distance, time of travel of EV and
outputs casing temperature, winding temperature, time to refill the bearing lubricant, percentage deterioration of magnetic flux to
compute remaining useful life (RUL) of permanent magnet (PM). Health monitoring and prognosis of EV motor using i-DT is
developed with two approaches. Firstly, in-house health monitoring and prognosis is developed to monitor the performance of
the motor in-house. Secondly, Remote Health Monitoring and Prognosis Centre (RHMPC) is developed to monitor the
performance of the motor remotely using cloud communication by the service provider of the EV. The simulation results prove
that the RUL of PM and time to refill the bearing lubricant obtained by i-DT twins theoretical results.
IET Electr. Power Appl., 2019, Vol. 13 Iss. 9, pp. 1328-1335 1328
© The Institution of Engineering and Technology 2019
cloud communication by the service provider of the EV. The
originality of this paper includes a development of i-DT for an EV
motor using ANN and FL for the following tasks:
2.1 EV motor
Traction motors are crucial part of EVs and their health has to be
monitored. EV motors are susceptible to health degradation due to
temperature rise in stator winding of the motor. Current flows
through the EV motor continuously during movement of the
vehicle. The current drawn by the motor causes I2R loss and heat
due to resistance of the stator winding.
Fig. 1 Block diagram representation of IHMP and RHMPC
(a) Block diagram representation of in-house health monitoring and prognosis
2.2 Health monitoring and prognosis
(IHMP), (b) Block diagram representation of Remote Health Monitoring and
Health Monitoring and Prognosis is considered with two Prognosis Centre (RHMPC)
approaches, IHMP and RHMPC in this paper. Average speed and
time of run are given as inputs to IHMP which consists of EV
motor and i-DT developed with ANN/FL located inside the EV.
Remote Health Monitoring and Prognosis system consists of EV
motor, TCP/IP based communication system, RHMPC. EV motor
is selected based on type of electric vehicle system chosen for
transportation. Normally induction motors and Brushless DC
Motor (BLDC) motors are selected depending on torque
requirement. EVs are implemented with TCP/IP to enable two-way
communication from EV to cloud and cloud to health monitoring
system. The main objective is to monitor the health of EV motor
both inside the vehicle as well as remotely as shown in Figs. 1 and
2 with all the components, except the communication system used
for remote monitoring. In this paper, EV motor parameters such as
speed and time of run are considered rather than the vehicle
parameter since the main objective focuses on EV motor health
monitoring and prognosis parameters.
3.1.2 Collection of real time input vectors: A tachometer is used The measured temperature of casing is used to calculate winding
to measure the speed of the EV motor and stop watch is used to temperature, time to replace the bearing lubricant and the health of
observe the time of run of EV motor. The details of EV considered PM using mathematical model as shown in Fig. 5.
for observing the reading is given in Appendix, Fig. 3. ANN blocks
IET Electr. Power Appl., 2019, Vol. 13 Iss. 9, pp. 1328-1335 1329
© The Institution of Engineering and Technology 2019
Fig. 4 Collection of input–output vectors for training ANN
where Tc is the motor casing temperature, Tamb is the ambient 3.2.3 Estimation of RUL: The temperature of the stator winding
temperature and Rc−a represents the thermal resistance from the radiates the heat to PM rotor made up of neodynium. Temperatures
casing to the ambient (°C/W), specific for a given motor. In this beyond 100°C result in 0.15% decrease of magnetic flux density of
work, Rc−a is considered to be 1.02°C/W. Casing temperature is the PM which is used to estimate RUL of the PM rotor [21, 24].
measured and winding temperature is calculated using (2) due to The relationship of the different output parameters derived from
the limitation of measuring winding temperature. casing temperature using (2) and (3) is shown in Fig. 5. Based on
the mathematical computations, the maximum and minimum range
for the input–output vectors is given in Table 2.
3.2.2 Determination of time to replace lubricant: The casing
temperature is used to compute the time to replace lubricant,
Lubtime of the motor bearing [23]. This is done by considering a 3.2.4 Training of ANN for i-DT: Two ANNs are trained in
horizontal bore with inner diameter (d) of 40 mm. MATLAB/Simulink using feedforward algorithm with data given
1330 IET Electr. Power Appl., 2019, Vol. 13 Iss. 9, pp. 1328-1335
© The Institution of Engineering and Technology 2019
Fig. 6 Comparison of theoretical, ANN and Fuzzy values for motor health
parameters
(a) Winding temperature of i-DT and theoretical computation, (b) Casing temperature
of i-DT and theoretical computation, (c) Lubricant refill time of i-DT and theoretical
computation, (d) PM health deterioration of i-DT and theoretical computation
IET Electr. Power Appl., 2019, Vol. 13 Iss. 9, pp. 1328-1335 1331
© The Institution of Engineering and Technology 2019
Table 3 Representation of rule base
Speed Time of run Casing temperature Winding temp Time to refill lubricant PM HEALTH
S1 t1 Tc1 tw1 Lub1 Det1
t2 Tc1 tw1 Lub4 Det2
t3 Tc2 tw2 Lub4 Det2
t4 Tc3 tw3 Lub4 Det4
t5 Tc1 tw1 Lub1 Det1
S2 t1 Tc1 tw1 Lub4 Det2
t2 Tc2 tw2 Lub2 Det2
t3 Tc2 tw2 Lub4 Det3
t4 Tc3 tw3 Lub4 Det4
t5 Tc4 tw3 Lub1 Det5
S3 t1 Tc2 tw3 Lub4 Det2
t2 Tc2 tw2 Lub4 Det2
t3 Tc3 tw3 Lub3 Det3
t4 Tc4 tw3 Lub1 Det4
t5 Tc5 tw4 Lub1 Det5
S4 t1 Tc3 tw3 Lub4 Det2
t2 Tc2 tw3 Lub4 Det2
t3 Tc3 tw4 Lub3 Det3
t4 Tc4 tw4 Lub4 Det4
t5 Tc5 tw4 Lub1 Det5
S5 t1 Tc3 tw4 Lub4 Det2
t2 Tc3 tw5 Lub4 Det2
t3 Tc4 tw4 Lub3 Det3
t4 Tc5 tw5 Lub1 Det5
t5 Tc5 tw5 Lub5 Det5
4.2 Rule base autonomous electric vehicle and intelligent transportation, where
data is used for decision making. For both approaches, the average
Average speed of EV and time of run have five membership speed and time of run of EV are the input parameters. The distance
functions each. A fuzzy rule base is represented by a set of travelled by EV can be obtained directly through speedometer. The
linguistic statements as follows: time of run of EV can be captured by triggering an on-board timer
IF speed is s1 and time of run is t1, THEN winding temperature like TDC7200 [26, 27]. The service provider of the cloud
is tw1, casing temperature is Tc1, time to refill lubricant is Lub1, authenticates access to the user. The two inputs are stored in cloud
PM health is Det1. and can be retrieved by RHMPC as shown in Fig. 8.
Similarly, by considering other conditions, 5 × 5 = 25 rules are In RHMPC, the distance travelled is divided by the time of run
framed. The rule base is represented in Table 3. to estimate the average speed of the EV. It subtracts the present
data from the previous data obtained through cloud. The time of
4.3 Defuzzification run and average speed of EV are provided as inputs to i-DT that
The centre of gravity method is used for the defuzzification. The computes the four health parameters of the motor. Since, i-DT
fuzzy output is converted into crisp output (u) using [25]: developed is depending on casing temperature of motor, from
which winding temperature, lubricant refill time, PM health
N deterioration are derived necessity of vehicle model is not required.
∑i = 1 f i μi This is the advantage of proposed method.
u= N (4)
∑i = 1 μi
6 Simulation of digital twin
5 Cloud communication between EV to RHMPC The simulation model consists of the PMSM used in EV, IGBT
EV motor parameters are monitored in two ways, using IHMP and converter and a 48 V battery as shown in Fig. 9. In simulation, the
RHMPC. IHMP will be useful for users having knowledge and speed of motor observed is in rad/s. The radius of the EV wheel is
understanding in vehicle technology. The decision will be taken by 0.2159 m. The ratio of motor speed and radius of the wheel gives
the user by observing data. RHMPC will be useful for anyone, average speed of EV in m/s which is converted to km/h. This
1332 IET Electr. Power Appl., 2019, Vol. 13 Iss. 9, pp. 1328-1335
© The Institution of Engineering and Technology 2019
Fig. 9 Block diagram for development of an i-DT
IET Electr. Power Appl., 2019, Vol. 13 Iss. 9, pp. 1328-1335 1333
© The Institution of Engineering and Technology 2019
Fig. 12 Plot of motor health parameters vs time of run and average speed
of EV using ANN based i-DT
(a) Winding temperature vs average speed and time of run of EV, (b) Casing
temperature vs average speed and time of run of EV, (c) Lubricant refill time vs
average speed and time of run of EV, (d). PM health deterioration vs average speed Fig. 13 Plot of motor health parameters vs time of run and average speed
and time of run of EV of EV using Fuzzy based i-DT
(a) Winding temperature vs average speed and time of run of EV, (b) Casing
an opportunity to incorporate pre-emptive maintenance of EV temperature vs average speed and time of run of EV, (c) Lubricant refill time vs
motor and helping in smart asset management in future. average speed and time of run of EV, (d) PM health deterioration vs avg.speed and
Excess lubrication causes the EV motor to draw excess current time of run of EV
leads to rise in temperature, thereby deteriorating EV motor life.
The temperature rise inside the EV motor causes 5% of motor stored data of previous time of run. The accumulation of previous
failure [20]. data is used to determine the health of PMSM. The results obtained
using mathematical (1)–(3) predict temperature of the winding and
7.3 Discussions lubrication refill time at any instant of time. The time of run is not
possible to be integrated in the above equations, which is very
An i-DT developed using ANN and fuzzy logic can be compared much important to monitor the health and provide prognosis.
with theoretical approach in this work. Theoretically, average Hence, ANN and fuzzy logic are implemented with time of run for
speed is calculated as distance travelled divided by time taken to developing i-DT. For validation purpose, theoretical computations
complete the distance. Health monitoring and prognosis require are compared with i-DT. Average percentage deviation of
1334 IET Electr. Power Appl., 2019, Vol. 13 Iss. 9, pp. 1328-1335
© The Institution of Engineering and Technology 2019
[5] Djeziri, M., Merzouki, R., Ould-Bouamama, B.: ‘Robust monitoring of
electric vehicle with structured and unstructured uncertainties’, IEEE Trans.
Veh. Technol., 2009, 58, (9), pp. 4710–4719
[6] Khan, M.A.S.K., Rahman, M.A.: ‘Development and implementation of a
novel fault diagnostic and protection technique for IPM motor drives’, IEEE
Trans. Ind. Electron., 2009, 56, (1), pp. 85–92
[7] Liu, X.-Q., Zhang, H.-Y., Liu, J., et al.: ‘Fault detection and diagnosis of
permanent-magnet DC motor based on parameter estimation and neural
network’, IEEE Trans. Ind. Electron., 2000, 47, (5), pp. 1021–1030
[8] Nyanteh, Y., Edrington, C., Srivastava, S., et al.: ‘Application of artificial
intelligence to real-time fault detection in permanent-magnet synchronous
machines’, IEEE Trans. Ind. Appl., 2013, 49, (3), pp. 1205–1214
[9] Hamidizadeh, S., Alatawneh, N., Chromik, R.R., et al.: ‘Comparison of
different demagnetization models of permanent magnet in machines for
electric vehicle application’, IEEE Trans. Magn., 2016, 52, (5), pp. 1–4
Fig. 14 RUL estimation of PM rotor and bearing lubricant [10] Al-Timimy, A., Al-Ani, M., Degano, M., et al.: ‘Influence of rotor endcaps on
the electromagnetic performance of high-speed PM machine’, IET Electr.
Power Appl., 2018, 12, (8), pp. 1142–1149
Table 4 Comparison of theoretical and i-DT [11] Zeng, C., Huang, S., Yang, Y., et al.: ‘Inter-turn fault diagnosis of permanent
Parameter Average Average magnet synchronous machine based on tooth magnetic flux analysis’, IET
percentage percentage Electr. Power Appl., 2018, 12, (6), pp. 837–844
[12] Wang, L., Aleksandrov, S., Tang, Y., et al.: ‘Fault-tolerant electric drive and
deviation – ANN deviation – fuzzy space-phasor modulation of flux-switching permanent magnet machine for
logic aerospace application’, IET Electr. Power Appl., 2017, 11, (8), pp. 1416–1423
winding temperature 0.557% 0.911% [13] Hou, Z., Huang, J., Liu, H., et al.: ‘No-load losses based method to detect
demagnetisation fault in permanent magnet synchronous motors with parallel
casing temperature 0.903% 0.554% branches’, IET Electr. Power Appl., 2017, 11, (3), pp. 471–477
time until next 6.396% 2.315% [14] Qazalbash, A.A., Sharkh, S.M., Irenji, N.T., et al.: ‘Rotor eddy loss in high-
lubrication speed permanent magnet synchronous generators’, IET Electr. Power Appl.,
2015, 9, (5), pp. 370–376
PM health 3.814% 1.782% [15] Lee, K.-D., Kim, W.-H., Jin, C.-S., et al.: ‘Local demagnetisation analysis of a
permanent magnet motor’, IET Electr. Power Appl., 2015, 9, (3), pp. 280–286
[16] Farasat, M., Trzynadlowski, A.M., Fadali, M.S.: ‘Efficiency improved
sensorless control scheme for electric vehicle induction motors’, IET Electr.
theoretical computations, ANN and FL are compared and given in Syst. Transp., 2014, 4, (4), pp. 122–131
Table 4. From Table 4, it can be observed that average percentage [17] Lee, J., Jeon, Y.J., Choi, D.-C., et al.: ‘Demagnetization fault diagnosis
deviation of FL i-DT is better than ANN. method for PMSM of electric vehicle’. Proc. 39th Annu. Conf. IEEE Ind.
Electron. Soc. (IECON), Vienna, Austria, November 2013, pp. 2709–2713
[18] Fernandez, D., Hyun, D., >Park, Y., et al.: ‘Permanent magnet temperature
8 Conclusion estimation in PM synchronous motors using Low-cost hall effect sensors’,
IEEE Trans. Ind. Appl., 2017, 53, (5), pp. 4515–4525
Health monitoring and prognosis of PMSM is used to extend the [19] Sarikhani, A., Mohammed, O.A.: ‘Inter-turn fault detection in PM
lifetime of PMSM by employing i-DT. An i-DT is created by synchronous machines by physics-based back electromotive force estimation’,
IEEE Trans. Ind. Electron., 2013, 60, (8), pp. 3472–3484
training ANN between inputs of average speed, time of run of EV [20] Wang, C., Prieto, M.D., Romeral, L., et al.: ‘Detection of partial
and outputs of winding temperature, casing temperature, time to demagnetization fault in PMSMs operating under nonstationary conditions’,
refill bearing lubricant, PM health. A FL i-DT is also developed IEEE Trans. Magn., 2016, 52, (7), pp. 1–4, Art. no. 8105804
with the same inputs and outputs. The developed i-DT is used to [21] Milanfar, P., Lang, J.H.: ‘Monitoring the thermal condition of permanent-
magnet synchronous motors’, IEEE Trans. Aerosp. Electron. Syst., 1996, 32,
monitor the health and provide prognosis of EV motor to schedule (4), pp. 1421–1429
time to service EV motor. Individual EV users and car servicing [22] Antunes Cezario, C., Silva, H.P.: ‘Electric motor winding temperature
companies can implement this approach by IHMP and RHMPC, prediction using a simple two-resistance thermal circuit’, COMPEL-Int. J.
respectively. The simulation results indicate i-DT can be used Comput. Math. Electri. Electron.Eng., 2010, 29, (5), pp. 1325–1330
[23] Guide to Electric Motor Bearing Lubrication, ExxonMobil Lubricants and
effectively by the EV user and automotive industries to predict the specialities, February 2002
health and RUL. The developed i-DT twins the present health of [24] Ding, X., Liu, J., Mi, C.: ‘Online temperature estimation of IPMSM
EV motor remotely and in-house and can be implemented in EVs permanent magnets in hybrid electric vehicles’. 2011 6th IEEE Conf. on
and autonomous vehicles. Industrial Electronics and Applications (ICIEA), Beijing, China, 2011, pp.
179–183
[25] Manickavasagam, K.: ‘Intelligent energy control center for distributed
9 Acknowledgment generators using multi-agent system’, IEEE Trans. Power Syst., 2015, 30, (5),
pp. 2442–2449
The authors would like to express their profound gratitude to Dr. [26] Morinigo-Sotelo, D., Duque-Perez, O., Perez-Alonso, M.: ‘Assessment of the
Govind R Kadambi, Pro-Vice Chancellor, Dr. M. Arulantham, lubrication condition of a bearing using spectral analysis of the stator current’.
2010 Int. Symp. on Power Electronics Electrical Drives Automation and
Dean, Faculty of Engineering and Technology, Dr. K.M. Sharath Motion (SPEEDAM), Naples, Italy, 2010, pp. 1160–1165
Kumar, Deputy Director, Directorate of sponsor research, M.S. [27] TDC7200 Time-to-Digital Converter for Time-of-Flight Applications in.
Ramaiah University of Applied Sciences for their continuous Available at www.ti.com/lit/ds/symlink/tdc7200.pdf
encouragement and kind support for providing the facility to carry [28] Armstrong, R.W.Jr: ‘Load to motor inertia mismatch: unveiling the truth’,
Drives and Controls Conference, Telford, England, 1998. Available at http://
out the research. www.motiontech.com.au/assets/pdf/Kollmorgen%20Load%20to%20Motor
%20Inertia%20Mismatch%20Unveiling%20The%20Truth%20230812.pdf
[29] A guide to preventing failure Motors don't just fail...do they?. Available at
10 References www.okyanusbilgiambari.com/bilgiambari/TPM/motors_ebook- ABB.pdf
[1] Loureiro, R., Benmoussa, S., Touati, Y., et al.: ‘Integration of fault diagnosis
and fault-tolerant control for health monitoring of a class of MIMO intelligent
autonomous vehicles’, IEEE Trans. Veh. Technol., 2014, 63, (1), pp. 30–39, 11 Appendix
doi: 10.1109/TVT.2013.2274289
[2] Gritli, Y., Rossi, C., Casadei, D., et al.: ‘Demagnetizations diagnosis for
permanent magnet synchronous motors based on advanced wavelet analysis’.
Proc. 20th Int. Conf. Electr. Mach., Marseille, France, September 2012, pp.
EV and Motor considered for analysis is given in Fig. 3a.The
2397–2403 mounting of motor in vehicle is given in Fig. 3b. The measurement
[3] Weng, C., Sun, J., Peng, H.: ‘Model parametrization and adaptation based on of casing temperature of motor is given in Fig. 3c. Speed
the invariance of support vectors with applications to battery state-of-health measurement of vehicle is shown in Fig. 3d. The specification of
monitoring’, IEEE Trans. Veh. Technol., 2015, 64, (9), pp. 3908–3917
[4] Loureiro, R., Merzouki, R., Bouamama, B.O.: ‘Bond graph model based on
EV motor: make-Rock star, 48 V, 3 kW, 3000 rpm. Sl. No
structural diagnosability and recoverability analysis: application to intelligent G1604022350 is indicated in Fig. 3e.
autonomous vehicles’, IEEE Trans. Veh. Technol., 2012, 61, (3), pp. 986–997
IET Electr. Power Appl., 2019, Vol. 13 Iss. 9, pp. 1328-1335 1335
© The Institution of Engineering and Technology 2019