Evaluation of Current Signature in Bearing Defects by envelope analysis of the vibration in induction motors

Motor current signature analysis (MCSA) enables non-invasive monitoring, without interruption of machine operation in a remote and online way, allowing the identification of various types of faults of electrical and mechanical nature without the need of accessing the motor itself, but only its supply cables. Despite its advantages, it has limitations in accurately diagnosing incipient roller bearing faults. For the detection of incipient roller bearing faults, envelope analysis of vibration signals is a well-known and stablished technique used by motor condition monitoring experts for a long time, overcoming MCSA for that purpose. Thus, it is proposed in this paper, that the fault characteristic frequencies of roller bearings are identified in the current spectrum with the aid of envelope analysis on the bearing vibration signal. After this aided identification, the fault related spectral components in the current spectrum can be correctly tracked over time for trending evaluation and decision-making. This approach can represent a significant economic value in a motor condition monitoring program, since vibration envelope analysis is performed only at a first step and, after that, its results can be applied for the MCSA monitoring of all same-model motor drivers in an industrial site. This approach is even more valuable considering the concept of the Self-Supplied Wireless Current Transducer (SSWCT) also proposed in this paper. The SSWCT is an Industrial Internet of Things (IIOT) device for MCSA application in an Industry 4.0 environment. This proposed device has wireless communication interface and wireless/battery less power supply, being supplied by the energy harvested from the magnetic field of the same currents it is transducing. So, it is a completely galvanic isolated monitoring device, without batteries and without any electric connections to the industry electric system, easily installable to the motor cables, not using precious space in the electric panels of the motor control centers and not having any physical contact to the monitored asset.


Introduction
Currently in industry, fault detection and identification of causes and consequences of failures are performed by operators. If any of these tasks are not performed correctly, adverse effects can occur, such as: reduction of product quality, process safety and lifetime of machinery. In this scenario, condition-based maintenance helps reduce unplanned downtime and maintenance costs. Since the induction motor is widely used in industries, fault detection in induction motors becomes an indispensable tool in condition-based maintenance [1]. According to statistical evidence, most Those frequencies depend on the bearing geometrical characteristics, and can present an error from external factors such as: geometric variations imposed by the failure itself, incorrect bearing installation, cage locking, improper lubrication and ball slip, for instance. Thus, care must be taken to avoid the analysis of spectral components not related to the defect. In MCSA, this is a more critical concern, since it is not uncommon to have high energy supply frequency harmonics near the bearing fault frequencies, which can demand a very high-resolution spectrum to distinguish spectral components of diverse nature.
In order to correctly identify, in the current spectrum, the correct localized fault frequencies of roller bearings, this paper proposes the use of the envelope analysis of vibration signal as a first step. The envelope analysis is successfully used for decades in the detection of incipient roller bearing faults [10,16,18,19]. In this technique, the vibration signal is bandpass filtered in a frequency range corresponding to the mechanical resonance of the system, which tends to amplify the impacts produced by a localized fault in any element of the roller bearing. The filtered signal is then amplitude demodulated using the Hilbert transform, which results in the amplitude envelope of the mechanical resonance region. The envelope spectrum is then obtained by the Fourier transform. This spectrum presents the localized fault frequencies with great sensitivity [16].
So, this paper proposes, as a first step in the process of monitoring localized roller bearing faults with the MCSA technique, the use of the vibration envelope as an aiding tool for the spectral component identification. For this, current and vibration signals are acquired simultaneously and their spectra (demodulated current and vibration envelope) are compared in order to identify precisely, in the current spectrum, among all the present spectral components which are or are not the localized roller bearing fault components.
Once this first step is accomplished, vibration envelope is no more necessary, and the MCSA technique can carry on the monitoring process, keeping track of the relevant spectral components and their amplitude to support the decision-making process. So, after the first step of this process, all the benefits of MCSA technique, mentioned above, can be taken.
Once the spectral components are assuredly identified, it is easier for the MCSA algorithm to keep tracking of the spectral components of interest, even if they change their frequency over time. In case of superposition of two spectral components of different nature, the change in magnitude and/or phase of the spectral components over time can be used to identify that two of the tracked components are at the same frequency bin.
In line with the new trends related to the condition monitoring of assets in industry 4.0, the internet of things and environment energy harvesting, this paper proposes also the implementation of the presented approach in a concept device of a Self-Supplied Wireless Current Transducer (SSWCT).
The SSWCT is an Industrial Internet of Things (IIOT) device for condition monitoring of assets using MCSA in an Industry 4.0 environment. This proposed device has IEEE 802.11 [20] standard wireless communication interface (Wi-Fi) and has a wireless/battery less power supply, being supplied by the energy harvested from the magnetic field of the same currents it is transducing.
In this way, the proposed device can be considered a completely galvanic isolated monitoring device, having no electric connections to the industry electric system. The main advantage of the proposed concept is the ease of installation. Since the device is self-contained, there is no need for peripherals like: power supply modules, power cables, communication modules, etc. In the proposed concept, a single module device embraces the motor cable to be measured and is immediately in operation, capable of acquiring signals, transmitting them to the analysis software or processing them locally. Other significant advantage of the proposed concept is that there is no need for battery replacement. As long as the monitored motor is running, the device is being power supplied.
So, this paper deals with two techniques for roller bearing fault detection and trending: vibration envelope analysis and motor current signature analysis, MCSA. Vibration envelope analysis is clearly the best technique. The MCSA technique is not the best one for roller bearing fault detection but it has some advantages for long-term monitoring: for example, it allows the concept of the self-supplied wireless current transducer, SSWCT, that is a self-contained module to be installed in motor cables, not being exposed to the motor rough environment. The paper highlights the elusive nature of the roller bearing fault "constants" (BPFO, BPFI, FTF, BSF) and, so, the first use of vibration envelope to help in the setup stage of the MCSA monitoring.
In the following Section 2, background in the detection of localized bearing faults is provided. Section 3 provides brief background for bearing fault detection with MCSA. In Section 4, the experimental results are presented and the proposed methodology is justified. Section 5 presents the concept and proof-of-concept implementation of the SSWCT. And, in Section 6, the conclusions are presented. The "Appendix A-Rotor frequency measurement and Ball Pass Frequency-Outer Race (BPFO) accuracy consideration" presents detailed information about the measurement accuracy.

Detection of Localized Faults in Roller Bearings
Currently, bearing fault detection is performed by techniques such as vibration monitoring, wear debris analysis, temperature monitoring, acoustic emission, and MCSA. With respect to cost, complexity in the installation, and invasiveness, the MCSA is presented as more advantageous.
Bearing failures are classified as localized or distributed. Distributed faults result in changes affecting the entire spectrum, and it is difficult to find some pattern in the spectral components. The localized faults, discussed in this work, present spectral components that allow classification according to the affected bearing element [4].
According to Randall [16], in bearing fault analysis, the raw vibration signal spectrum often contains insufficient information for a good diagnosis and, so, envelope analysis must be applied. As shown in Figure 1, in the envelope approach, the signal is filtered in the high frequency range, next to resonance, where the impact pulses are amplified. The resulting signal is amplitude demodulated by the modulus of Hilbert Transform, producing the envelope signal, whose amplitude spectrum is obtained by the modulus of the Fourier transform. The amplitude spectrum of the envelope signal presents the diagnostic information as spectral components at the characteristic frequencies of the bearing elements.

Detection of Localized Faults in Roller Bearings
Currently, bearing fault detection is performed by techniques such as vibration monitoring, wear debris analysis, temperature monitoring, acoustic emission, and MCSA. With respect to cost, complexity in the installation, and invasiveness, the MCSA is presented as more advantageous.
Bearing failures are classified as localized or distributed. Distributed faults result in changes affecting the entire spectrum, and it is difficult to find some pattern in the spectral components. The localized faults, discussed in this work, present spectral components that allow classification according to the affected bearing element [4].
According to Randall [16], in bearing fault analysis, the raw vibration signal spectrum often contains insufficient information for a good diagnosis and, so, envelope analysis must be applied. As shown in Figure 1, in the envelope approach, the signal is filtered in the high frequency range, next to resonance, where the impact pulses are amplified. The resulting signal is amplitude demodulated by the modulus of Hilbert Transform, producing the envelope signal, whose amplitude spectrum is obtained by the modulus of the Fourier transform. The amplitude spectrum of the envelope signal presents the diagnostic information as spectral components at the characteristic frequencies of the bearing elements.

Bearing Fault Detection by the Current Signature Analysis
In a running induction motor with a localized roller bearing fault, every time the faulty surface touches other elements, an impulse is generated. The associate frequency of this fault is proportional to the axis speed and to the bearing constructive aspects and, when in the outer race (BPFO), for instance, it is determined by [21]: where Nb is the number of the balls, fr is the axis velocity, d is the ball diameter, D is the pitch diameter, and α is the bearing contact angle. Riley [22] reported the relationship between the vibration signal and the stator current signal of an induction motor in the frequency domain by means of their respective spectra. Two models describe bearing fault detection in the current spectrum: the radial displacement model, and the torque variation model. Schoen [21] showed that the stator currents are amplitude modulated by the radial displacement produced by the bearing fault, resulting in the appearance of lateral components around the supply frequency in the spectrum. Their frequencies are expressed by:

Bearing Fault Detection by the Current Signature Analysis
In a running induction motor with a localized roller bearing fault, every time the faulty surface touches other elements, an impulse is generated. The associate frequency of this fault is proportional to the axis speed and to the bearing constructive aspects and, when in the outer race (BPFO), for instance, it is determined by [21]: where N b is the number of the balls, f r is the axis velocity, d is the ball diameter, D is the pitch diameter, and α is the bearing contact angle.
Riley [22] reported the relationship between the vibration signal and the stator current signal of an induction motor in the frequency domain by means of their respective spectra. Two models describe bearing fault detection in the current spectrum: the radial displacement model, and the torque variation model. Schoen [21] showed that the stator currents are amplitude modulated by the radial where f e is the power supply frequency and m is a positive integer referring to the vibration component harmonic.
According to the torque variation model, Blodt [7] describes that bearing faults cause oscillations in the induction motor torque and consequently on its shaft speed. Torque oscillations result in RMS current oscillations and, consequently, instantaneous current modulation, also resulting in lateral components around the supply frequency in the spectrum.

Proposition Demonstration and Experimental Results
In this work, it is proposed to use envelope analysis as a first step for aiding the current spectrum characterization. The proposed approach is tested in a laboratory setup. A motor with healthy bearings and a motor with a localized bearing fault in the outer race of the driven end were prepared, as presented in Figure 2. The fault, of artificial origin, as addressed in other studies related to bearing faults [2,23,24], is applied to the bearing by means of a surface irregularity created in the outer race. To generate this irregularity on the surface, electric current was passed through the bearing until it could be felt slightly by turning the bearing in hands.
where fe is the power supply frequency and m is a positive integer referring to the vibration component harmonic. According to the torque variation model, Blodt [7] describes that bearing faults cause oscillations in the induction motor torque and consequently on its shaft speed. Torque oscillations result in RMS current oscillations and, consequently, instantaneous current modulation, also resulting in lateral components around the supply frequency in the spectrum.

Proposition Demonstration and Experimental Results
In this work, it is proposed to use envelope analysis as a first step for aiding the current spectrum characterization. The proposed approach is tested in a laboratory setup. A motor with healthy bearings and a motor with a localized bearing fault in the outer race of the driven end were prepared, as presented in Figure 2. The fault, of artificial origin, as addressed in other studies related to bearing faults [2,23,24], is applied to the bearing by means of a surface irregularity created in the outer race. To generate this irregularity on the surface, electric current was passed through the bearing until it could be felt slightly by turning the bearing in hands.
The bearing model is 6203zz, with nominal BPFO of 3.066. The motor used in the test is a Marathon ® 0.5 hp, three-phase, squirrel cage, induction motor. The motor is Y-closed, it has 2 poles, and it is fed by an inverter at 60.066 Hz. The load is a 6-blade fan coupled directly to the rotor shaft.
A current transducer (100mV/A), a monoaxial accelerometer (100 mV/g) and a 24bit data acquisition module were used for current and acceleration signal acquisition at a sample rate of 46.875 kSps. The signals were sampled during 60 seconds which implies a spectral resolution of 0.0167 Hz.

Vibration Analysis
The acceleration spectrum for the faulty bearing is presented in Figure 3. Based on this spectrum, the band pass filter for the envelope analysis is set from 2000 Hz to 5000 Hz. Figure 4 presents the resulting envelope spectrum for the faulty bearing, where BPFO x fr harmonics are very clear. Figure  5 presents the resulting envelope spectrum for the healthy bearing, and it is clear that no BPFO x fr harmonics are present. The bearing model is 6203zz, with nominal BPFO of 3.066. The motor used in the test is a Marathon ® 0.5 hp, three-phase, squirrel cage, induction motor. The motor is Y-closed, it has 2 poles, and it is fed by an inverter at 60.066 Hz. The load is a 6-blade fan coupled directly to the rotor shaft.
A current transducer (100mV/A), a monoaxial accelerometer (100 mV/g) and a 24bit data acquisition module were used for current and acceleration signal acquisition at a sample rate of 46.875 kSps. The signals were sampled during 60 seconds which implies a spectral resolution of 0.0167 Hz.

Vibration Analysis
The acceleration spectrum for the faulty bearing is presented in Figure 3. Based on this spectrum, the band pass filter for the envelope analysis is set from 2000 Hz to 5000 Hz. Figure 4 presents the resulting envelope spectrum for the faulty bearing, where BPFO x f r harmonics are very clear. Figure 5 presents the resulting envelope spectrum for the healthy bearing, and it is clear that no BPFO x f r harmonics are present.              Figure 7 shows the envelope spectrum with zoom in BPFO x f r in 179.483 Hz. The analysis of Figures 3 and 4 can attest, as stated by Randall [16], that the acceleration envelope spectrum is the best tool for achieving a good diagnosis regarding localized roller bearing faults. Even more, based on the determination of f r and BPFO x f r from those high-resolution spectra presented in Figures 6 and 7, it is possible to conclude that the actual value for BPFO is 3.059 (179.483 Hz/58.683 Hz) and not the nominal value of 3.066. If the nominal value would be considered, BPFO x f r frequency would be 179.922 Hz, as presented in Figure 7, a 0.439 Hz difference from the actual value. This difference can have many causes such as: geometric variations imposed by the failure itself, incorrect bearing installation, cage locking, improper lubrication and ball slip, for instance.
Energies 2019, x, y FOR PEER REVIEW 7 of 16 Figure 6 presents the acceleration spectrum with zoom in the shaft frequency component, fr, in 58.683 Hz. And Figure 7 shows the envelope spectrum with zoom in BPFO x fr in 179.483 Hz. The analysis of Figures 3 and 4 can attest, as stated by Randall [16], that the acceleration envelope spectrum is the best tool for achieving a good diagnosis regarding localized roller bearing faults. Even more, based on the determination of fr and BPFO x fr from those high-resolution spectra presented in Figures 6 and 7, it is possible to conclude that the actual value for BPFO is 3.059 (179.483 Hz / 58.683 Hz) and not the nominal value of 3.066. If the nominal value would be considered, BPFO x fr frequency would be 179.922 Hz, as presented in Figure 7, a 0.439 Hz difference from the actual value. This difference can have many causes such as: geometric variations imposed by the failure itself, incorrect bearing installation, cage locking, improper lubrication and ball slip, for instance.

Current Analysis
The demodulated stator current spectrum is shown in Figure 8 for the motor with a faulty bearing. The demodulated current (or the current envelope) is chosen to facilitate the comparison with the acceleration envelope spectrum. Current demodulation is performed applying the Hilbert transform modulus to the current signal without any band-pass filtering. As well, Figure 9 presents  Figure 7 shows the envelope spectrum with zoom in BPFO x fr in 179.483 Hz. The analysis of Figures 3 and 4 can attest, as stated by Randall [16], that the acceleration envelope spectrum is the best tool for achieving a good diagnosis regarding localized roller bearing faults. Even more, based on the determination of fr and BPFO x fr from those high-resolution spectra presented in Figures 6 and 7, it is possible to conclude that the actual value for BPFO is 3.059 (179.483 Hz / 58.683 Hz) and not the nominal value of 3.066. If the nominal value would be considered, BPFO x fr frequency would be 179.922 Hz, as presented in Figure 7, a 0.439 Hz difference from the actual value. This difference can have many causes such as: geometric variations imposed by the failure itself, incorrect bearing installation, cage locking, improper lubrication and ball slip, for instance.

Current Analysis
The demodulated stator current spectrum is shown in Figure 8 for the motor with a faulty bearing. The demodulated current (or the current envelope) is chosen to facilitate the comparison with the acceleration envelope spectrum. Current demodulation is performed applying the Hilbert transform modulus to the current signal without any band-pass filtering. As well, Figure 9 presents

Current Analysis
The demodulated stator current spectrum is shown in Figure 8 for the motor with a faulty bearing. The demodulated current (or the current envelope) is chosen to facilitate the comparison with the acceleration envelope spectrum. Current demodulation is performed applying the Hilbert transform modulus to the current signal without any band-pass filtering. As well, Figure 9 presents the demodulated stator current spectrum for the motor with a healthy bearing. In both spectra, the only components standing out are the supply frequency harmonics, having no relation to the bearing fault spectral components. the demodulated stator current spectrum for the motor with a healthy bearing. In both spectra, the only components standing out are the supply frequency harmonics, having no relation to the bearing fault spectral components.  On the other hand, Figure 10 presents the zoom of demodulated current for the motor with a faulty bearing in the region around BPFO x fr and it is possible to see a small peak exactly at frequency 179.483 Hz, as indicated by the acceleration envelope of the faulty bearing. Other peaks are present and can be associated to the inverter supply frequency harmonics (3 x fe), to the mains frequency harmonics (3 x f1) and even to rotor cage electric imbalance harmonics (3 x fe -s x fe, where s is the rotor slip), reinforcing the importance of envelope analysis for an effective diagnosis. The nominal BPFO x fr frequency is also pointed in figure 9, showing that it must not be considered, as it leads to misinterpretation of the current spectrum and, consequently to a wrong diagnosis.
Thus, the proposition of using the acceleration envelope as an aid tool for the characterization of the current spectrum in a first approach is very valid. After this first approach, with a well-known the demodulated stator current spectrum for the motor with a healthy bearing. In both spectra, the only components standing out are the supply frequency harmonics, having no relation to the bearing fault spectral components.  On the other hand, Figure 10 presents the zoom of demodulated current for the motor with a faulty bearing in the region around BPFO x fr and it is possible to see a small peak exactly at frequency 179.483 Hz, as indicated by the acceleration envelope of the faulty bearing. Other peaks are present and can be associated to the inverter supply frequency harmonics (3 x fe), to the mains frequency harmonics (3 x f1) and even to rotor cage electric imbalance harmonics (3 x fe -s x fe, where s is the rotor slip), reinforcing the importance of envelope analysis for an effective diagnosis. The nominal BPFO x fr frequency is also pointed in figure 9, showing that it must not be considered, as it leads to misinterpretation of the current spectrum and, consequently to a wrong diagnosis.
Thus, the proposition of using the acceleration envelope as an aid tool for the characterization of the current spectrum in a first approach is very valid. After this first approach, with a well-known On the other hand, Figure 10 presents the zoom of demodulated current for the motor with a faulty bearing in the region around BPFO x f r and it is possible to see a small peak exactly at frequency 179.483 Hz, as indicated by the acceleration envelope of the faulty bearing. Other peaks are present and can be associated to the inverter supply frequency harmonics (3 x f e ), to the mains frequency harmonics (3 x f 1 ) and even to rotor cage electric imbalance harmonics (3 x f e -s x f e , where s is the rotor slip), reinforcing the importance of envelope analysis for an effective diagnosis. The nominal BPFO x f r frequency is also pointed in Figure 9, showing that it must not be considered, as it leads to misinterpretation of the current spectrum and, consequently to a wrong diagnosis.

Proposition of a Self-Supplied Wireless Current Transducer-SSWCT
In recent times, new scenarios have emerged based on technological advances in two areas mainly, which are the development of microprocessors with high processing power and high efficiency; and the development of wireless communication protocols with high transfer rates and low consumption. These advances have stimulated a revolution in the world of sensors that have been called IoT (Internet of Things), where it is seen that all devices will present, in the near future, some kind of intelligence and interconnection through the Internet. In the industrial environment, another revolution related to these technological advances has been establishing, the so-called fourth Industrial Revolution, or Industry 4.0, where the physical systems of the factory floor will have their parameters monitored and digital models of their operative and maintenance condition will be updated for decision-making and optimization purposes.
Thus, in line with these innovative trends, to implement the MCSA approach describe in the sections above, it is proposed a Self-Supplied Wireless Current Transducer-SSWCT. A proof of concept of this proposed device was implemented in a Research and Development project for MCSA monitoring of induction motors and it is being tested in Pimental Hydroelectric power plant (at Belo Monte Complex, in north region of Brazil) for monitoring motors driving pumps of the cooling and speed regulator systems of the plant.
The concept of the SSWCT eliminates the use of external power modules, cables and batteries. Along with a wireless (IEEE 802.11 standard [20]) communication interface, it constitutes a completely galvanically insulated device, since there is no electric contact to the plant electric system. It is capable of transmitting the current signals to an analysis software as a digital wireless transducer, but it also has enough processing power and mass memory for digital signal processing tasks and data-logging, in order to operate as a stand-alone device.
The SSWCT harvests its power from the magnetic field of the same currents it is transducing. For that, it uses current transformers with high permeability magnetic material cores, optimized for power extraction.
The main advantage of this concept is the ease of installation, for it is a self-contained device that can be installed in the motor cables, saving space in the motor control center panels. The implemented device can monitor one motor current with its built-in measurement current transformer or two motor currents, using an external current transformer connected to the main module. Thus, the proposition of using the acceleration envelope as an aid tool for the characterization of the current spectrum in a first approach is very valid. After this first approach, with a well-known current spectrum, the tracking of the localized bearing faults over time can be carried on only by the MCSA technique itself, saving time and resources.

Proposition of a Self-Supplied Wireless Current Transducer-SSWCT
In recent times, new scenarios have emerged based on technological advances in two areas mainly, which are the development of microprocessors with high processing power and high efficiency; and the development of wireless communication protocols with high transfer rates and low consumption. These advances have stimulated a revolution in the world of sensors that have been called IoT (Internet of Things), where it is seen that all devices will present, in the near future, some kind of intelligence and interconnection through the Internet. In the industrial environment, another revolution related to these technological advances has been establishing, the so-called fourth Industrial Revolution, or Industry 4.0, where the physical systems of the factory floor will have their parameters monitored and digital models of their operative and maintenance condition will be updated for decision-making and optimization purposes.
Thus, in line with these innovative trends, to implement the MCSA approach describe in the sections above, it is proposed a Self-Supplied Wireless Current Transducer-SSWCT. A proof of concept of this proposed device was implemented in a Research and Development project for MCSA monitoring of induction motors and it is being tested in Pimental Hydroelectric power plant (at Belo Monte Complex, in north region of Brazil) for monitoring motors driving pumps of the cooling and speed regulator systems of the plant.
The concept of the SSWCT eliminates the use of external power modules, cables and batteries. Along with a wireless (IEEE 802.11 standard [20]) communication interface, it constitutes a completely galvanically insulated device, since there is no electric contact to the plant electric system. It is capable of transmitting the current signals to an analysis software as a digital wireless transducer, but it also has enough processing power and mass memory for digital signal processing tasks and data-logging, in order to operate as a stand-alone device.
The SSWCT harvests its power from the magnetic field of the same currents it is transducing. For that, it uses current transformers with high permeability magnetic material cores, optimized for power extraction.
The main advantage of this concept is the ease of installation, for it is a self-contained device that can be installed in the motor cables, saving space in the motor control center panels. The implemented device can monitor one motor current with its built-in measurement current transformer or two motor currents, using an external current transformer connected to the main module. Figure 11 presents the concept device developed by the authors. In the block diagram, presented in Figure 11a, the Power Harvester Current Transform (CT) outputs the power in AC form, the Full-Wave Rectifier transforms it to DC form, the Overvoltage Protection clamps the voltage to the limit of the Buck Converter input and the Buck Converter steps down and regulate the voltage to a Timer Relay, that will power the application only after a preset delay to avoid transients. The application is a Single Board Computer with a Wi-Fi module and a 24-bit ADC. The application digitalizes the current measured by the Measurement CT and processes, stores and transmits it to the software application for analysis and log.
Energies 2019, x, y FOR PEER REVIEW 10 of 16 Figure 11 presents the concept device developed by the authors. In the block diagram, presented in Figure 11a, the Power Harvester Current Transform (CT) outputs the power in AC form, the Full-Wave Rectifier transforms it to DC form, the Overvoltage Protection clamps the voltage to the limit of the Buck Converter input and the Buck Converter steps down and regulate the voltage to a Timer Relay, that will power the application only after a preset delay to avoid transients. The application is a Single Board Computer with a Wi-Fi module and a 24-bit ADC. The application digitalizes the current measured by the Measurement CT and processes, stores and transmits it to the software application for analysis and log.
(b) (c) The block diagram in Figure 11a presents only one power harvester CT, but two CTs ( Figure  11b) are used in order to achieve the minimum operating current of 35 A, stablished as a design constraint. As well, two measurement CTs can be used: an internal one, shown in Figure 11b and an external one, not shown.
In the design process of the Power Harvester CTs, the main constraints were: the magnetic core size (outer diameter 55 mm, inner diameter 31 mm and height 30 mm), the minimum current for The block diagram in Figure 11a presents only one power harvester CT, but two CTs (Figure 11b) are used in order to achieve the minimum operating current of 35 A, stablished as a design constraint. As well, two measurement CTs can be used: an internal one, shown in Figure 11b and an external one, not shown.
In the design process of the Power Harvester CTs, the main constraints were: the magnetic core size (outer diameter 55 mm, inner diameter 31 mm and height 30 mm), the minimum current for operation (~35 A), and the maximum load power (~2.5 W). The magnetization curve for the chosen core were determined and used in a simulation model to optimize the number of secondary winding turns to provide the maximum load power in the minimum current condition. Figure 12 presents the SSWCT installed in a motor cable at Pimental power plant. The second and external CT, mentioned above, can be seen in the picture. Once the installation at Pimental is very recent and interesting results regarding roller bearing faults have not been collected yet, a test setup for the SSWCT is presented in Figure 13. The SSWCT is presented open for test measurements. The supply current of the motor-fan assembly is about 1.5 A in 220V and 60Hz. The SSWCT was designed to operate with a minimum of 35 A. So, to operate the SSWCT with such a small laboratory motor, 24 turns were passed through the CT, producing a magnetomotive force of~36 A.turn (~1.5 A x 24 turns). Of course, the current measured is divided by 24 in the software to yield the correct value. In the figure, the primary current, 35.7 A, is shown in the orange multi-meter and the rectified voltage, 17.6 Vdc, in the output of the power harvester is presented in the blue multi-meter. The motor current is measured by the internal current transformer of the SSWCT and the accelerometer output is measured by the data acquisition module presented in Figure 2. operation (~ 35 A), and the maximum load power (~ 2.5 W). The magnetization curve for the chosen core were determined and used in a simulation model to optimize the number of secondary winding turns to provide the maximum load power in the minimum current condition. Figure 12 presents the SSWCT installed in a motor cable at Pimental power plant. The second and external CT, mentioned above, can be seen in the picture. Once the installation at Pimental is very recent and interesting results regarding roller bearing faults have not been collected yet, a test setup for the SSWCT is presented in Figure 13. The SSWCT is presented open for test measurements. The supply current of the motor-fan assembly is about 1.5 A in 220V and 60Hz. The SSWCT was designed to operate with a minimum of 35 A. So, to operate the SSWCT with such a small laboratory motor, 24 turns were passed through the CT, producing a magnetomotive force of ~36 A.turn (~1.5 A x 24 turns). Of course, the current measured is divided by 24 in the software to yield the correct value. In the figure, the primary current, 35.7 A, is shown in the orange multi-meter and the rectified voltage, 17.6 Vdc, in the output of the power harvester is presented in the blue multi-meter. The motor current is measured by the internal current transformer of the SSWCT and the accelerometer output is measured by the data acquisition module presented in Figure 2.          Figure 15 presents the demodulated current spectrum produced using the SSWCT, where a small peak at the same frequency of the dominant peak, shown in the acceleration envelope spectrum, can be seen. This test, and others performed in the laboratory, shows that the SSWCT presents the same functionality of a traditional data acquisition system used for motor current signature analysis. For this run, the rotor frequency, fr, is 59.117 Hz. It is curious to notice that, after many hours running, the actual BPFO for this run is 3.084 (182.333 / 59.117), against the nominal value 3.066, showing once more the elusive nature of this "constant" and highlighting the importance of the use of acceleration envelope spectrum as a reference for bearing fault studies in the motor current signature analysis.  Figure 15 presents the demodulated current spectrum produced using the SSWCT, where a small peak at the same frequency of the dominant peak, shown in the acceleration envelope spectrum, can be seen. This test, and others performed in the laboratory, shows that the SSWCT presents the same functionality of a traditional data acquisition system used for motor current signature analysis. For this run, the rotor frequency, f r , is 59.117 Hz. It is curious to notice that, after many hours running, the actual BPFO for this run is 3.084 (182.333/59.117), against the nominal value 3.066, showing once more the elusive nature of this "constant" and highlighting the importance of the use of acceleration envelope spectrum as a reference for bearing fault studies in the motor current signature analysis.

Conclusions
This paper proposes the use of vibration envelope analysis as an aid tool for the accurate determination of the characteristic frequencies of localized bearing faults in the stator current spectrum of three-phase induction motors. The methodology is described and demonstrated with real current and acceleration signals. Motors with healthy and faulty roller bearings are prepared for laboratory experiments. After this first current spectrum characterization with the aid of envelope analysis, the MCSA technique can carry on with the analysis over time in order to exploit its advantages.
The complexity of the current spectrum is presented in order to justify the proposed approach. It is shown that many spectrum components of diverse nature can coexist in a very narrow region in Figure 15. Demodulated current signal spectrum with zoom around the actual BPFO x f r frequency.

Conclusions
This paper proposes the use of vibration envelope analysis as an aid tool for the accurate determination of the characteristic frequencies of localized bearing faults in the stator current spectrum of three-phase induction motors. The methodology is described and demonstrated with real current and acceleration signals. Motors with healthy and faulty roller bearings are prepared for laboratory experiments. After this first current spectrum characterization with the aid of envelope analysis, the MCSA technique can carry on with the analysis over time in order to exploit its advantages.
The complexity of the current spectrum is presented in order to justify the proposed approach. It is shown that many spectrum components of diverse nature can coexist in a very narrow region in the spectrum and their characterization can be tricky. This way, at least in a first stage, the use of an aid tool like vibration and envelope analysis is necessary for an assertive bearing fault diagnosis.
Finally, in line with the recent trends in IIOT, Industry 4.0 and energy harvesting, the concept of a Self-Supplied Wireless Current Transducer-SSWCT is proposed as a means for implementing MCSA monitoring in a modern way, including the proposed methodology described in the first part of the paper. A proof-of-concept implementation of the SSWCT is presented as well as its experimental use in a power plant in north of Brazil.