## 1. Introduction

In the last years, more than 53% of global electric energy consumption has been used by electric motor systems. Since more than 70% of this electric energy is employed by induction motors with low efficiencies (IE0 and IE1 classes), there is a huge potential for energy savings [

1].

Optimization of induction motors has been a field of study for manufacturers and researchers in the last decades. Improvements in the active materials, as well as the introduction of new technologies, such as copper rotor motors, synchronous reluctance motor and line start permanent magnet motors (LSPMM) are some of the techniques implemented by manufacturers to achieve higher efficiencies [

2]. However, because these technologies are still recent, it is necessary to carry out previous studies to know their strengths and weaknesses when subjected to the supply conditions of existent electrical systems.

Harmonics in electric motors have been the subject of many studies to analyze their main effects on the performance of these asynchronous machines. Continuous operation of motors on a polluted harmonic system results in higher temperatures in stator and rotor windings and core due to additional harmonic losses, torque reduction, noise and mechanical vibrations—some of the main effects found in literature [

3,

4,

5,

6,

7]. The close interaction between current harmonics, saturation and mechanical problems, such as bearing failure and static eccentricity, can result in premature failure and, consequently, reduced service life, as presented in [

8,

9,

10].

The permanent presence of non-linear loads in electrical systems results in current and voltage distortions, which can produce detrimental effects depending on the type of load and its interaction with other system components. According to [

11], harmonic distortions are commonly found in faulted power distribution systems and in most cases contains harmonic of third, fifth, seventh, ninth and 11th orders. Work in [

11] also presents a methodology to detect harmonics in power systems using wavelet transform. However, in real applications, a preprocessing step is necessary before applying the proposed harmonics wavelet method due to the high noise levels. In [

12], a systematic literature review of recent failure prognosis systems is provided, including the main approaches and some of the most prominent application domains for failure diagnosis.

A bibliographic review on fault diagnosis using signal processing techniques in induction motors is presented in [

13]. The review indicates that current signature acquisition and processing can be used to characterize the failure nature in electrical machines. The review also shows that mechanical and electrical failures of induction motors exhibits explicit harmonic component in stator current. The works in [

10,

14] also show alternatives for the fault diagnosis in electric motors. The presence of harmonics in fault conditions are also mentioned in these studies.

New higher efficiency motors are built according to the IEC 60034-7 [

15] standard that specifies the requirements with reference to classification of construction types, mounting arrangements, terminal box position, which contribute to the substitution between technologies.

An analysis of the technical and economic benefits of substitution between these technologies has been presented in [

2,

3,

16,

17,

18,

19,

20,

21,

22,

23]. In addition, the studies in [

17,

18,

21,

22,

24,

25,

26,

27,

28,

29,

30,

31,

32] include the main constructive characteristics, weaknesses and strengths related to the LSPMM. In [

16], an in-field replacement example of a squirrel-cage induction motor by a line start permanent magnet motor (LSPMM) is presented. A good performance and lower consumption are obtained with these technologies, however at no load, the harmonic losses seem to be higher in LSPMMs.

Studies have shown that the higher initial cost of LSPMMs can be paid in a short time period due to energy savings and initial financial incentive programs; higher efficiency, power factor and thermal behavior are also some of the advantages obtained with this technology. However, a cost–benefit analysis considering the type of application and supply conditions must be developed before substitution.

Due to the synchronous speed, no currents are induced inside the LSPMM rotor (neglecting spatial and time harmonics), so the rotor temperature of these motors is about 30% lower than that of the induction motors with the same output power [

25]. In [

33], the rated load winding temperature rise of IE4 IMs is lower when compared with IE3 class and that of IE3 class is lower than the IE2 class IMs. Fifth and seventh voltage harmonics are analyzed in [

34,

35], showing that fifth harmonic results in higher temperature increases when compared with the seventh harmonic, mainly due to the counter-rotating field with respect to fundamental frequency produced by the fifth negative sequence harmonic. According to [

36], temperature rise of IMs due to harmonics is approximately between 4–6 °C.

A summary of the literature review related to harmonics and electric motors, classified according to the approaches analyzed in this work, is presented in

Table 1.

The introduction of more efficient motors through improvements in their materials and processes, as well as the introduction of new technologies, has made the replacement of old and non-efficient motors by higher efficiency motors attractive to the industry again with the aim of reducing costs and improving competitiveness. However, with the introduction of new technologies, it is necessary to know their responses to the presence of disturbances in electrical systems before substitution.

Despite the fact that comparative studies between high efficiency motors, as well as permanent magnet motors, already exist in the literature, very few are carried out by the experimental method. Similarly, regarding power quality, few studies compare the presence of different individual harmonics, including positive, negative and zero sequence harmonics, as well as combinations of all and their effects on electric motors.

In that sense, this study aims to analyze the main improvements in relation to savings and performance in these technologies, presenting through experimental tests a comparison of responses of electric motors classes IE2, IE3 and IE4, the latter being a hybrid motor with squirrel cage and permanent magnets, when subjected to harmonics present in current electrical systems, of second, third, fifth, seventh order and a combination of all in the supply voltage. In addition, a statistical study by means of correlation matrices between the temperature and the input parameters of each motor is presented to analyze behavior patterns for the temperature increase for each harmonic in study.

Then, from the results and conclusions generated from the proposed experimental tests, the main advantages and challenges related to the substitution between technologies, both under ideal supply conditions and in the presence of voltage harmonics, will be analyzed in this work with the aim of creating useful conclusions for the industry and other sectors in general.

With this objective, the second section of this paper presents a brief review of the improvements experienced by the IMs in recent years. The main differences between the SCIM and the LSPMM are also presented at the end of this section.

Section 3 presents the methodology used to analyze each electric motor in the presence of voltage harmonics. The results and discussions are presented in

Section 4, which ends with the correlation matrices and scatter plots created for the second and third order voltage harmonics.

Finally, the conclusions on the results obtained are presented in

Section 5.

## 2. Improvements in Induction Motors

As a result of studies carried out by manufacturers and researchers, electric motors have undergone optimizations in their materials and construction. Because the losses in the electric motors’ active materials represent more than 80% of the total losses [

19], the main efforts to improve the efficiency of these machines have been in this area. Techniques such as the use of amorphous materials, as well as high quality magnetic materials with lower losses in W/kg, have been used to reduce losses in the core [

25,

56]. Joule losses have been reduced with new stator winding configurations, as well as increasing the amount of copper in the windings and a greater filling of the stator grooves [

56,

57]. The introduction of copper rotors, as well as permanent magnet rotors, are other proposals to reduce losses in the rotor, due to the greater conductivity of copper compared to aluminum bars. In LSPMM, thanks to synchronous speed, no current will circulate in the rotor bars (except for harmonic currents) [

2,

25]; in addition, due to the contribution of permanent magnets, a reduction of the magnetization current is obtained in these machines, whereby the input current is decreased [

58].

Friction and windage losses have been reduced through optimizations in the fan design, as well as the use of bearings/seals with several design features, such as internal geometry of the raceways, the grease type, as well as the cage polymer material [

20,

59,

60].

In relation to the insulation class, high efficiency motors started to use insulation classes F/H (maximum temperatures of 150 and 180 °C, respectively), which offer a greater tolerance to temperature increases, as well as in terms of heat dissipation [

3].

#### SCIM’s and LSPMM’s Similarities and Differences

With the implementation of these new technologies in electric motors, greater efficiencies are obtained in relation to the old motors. However, many of the initial features are still present in these new technologies. In this section, the main constructive and operational similarities and differences between the squirrel cage induction motor (SCIM) and the line start permanent magnet motor (LSPMM) are presented.

The LSPMM has a construction like that of the induction motor, however it has a hybrid rotor that involves a squirrel cage that provides self-starting capability and enables synchronous operation at steady-state and permanent magnets inside. Different combinations of the magnets in the rotor have been used by manufacturers, and

Figure 1 presents some of them. The type of configuration depends on the manufacturers, as well as the properties of the magnets used. In [

26], different permanent magnet types and configurations are analyzed; authors conclude that rotors with interior magnet types provide higher efficiency and that rotor Type 1 and magnet’s material type NdFe35 provide better efficiencies and power density in LSPMMs.

The presence of permanent magnets results in a different start for the hybrid motor, where unlike the IE2 and IE3 class motors, the resulting LSPMM torque is equal to the difference between the cage torque and the braking torque (due to permanent magnets) [

27,

28]. The starting of the LSPMM is still one of the challenges for manufacturers and has been the subject of many investigations, however. These motors also suffer a sensitive dependence on the starting process on the input voltage, shaft inertia momentum and cages resistance. With a reduced input voltage, the motor starts more slowly and may even fail in synchronization [

61].

Starting of the LSPMM also presents a greater number of oscillations, as well as a greater variation in torque compared to the SCIM, which is difficult to use for applications with frequent starts, being more used for applications with fixed speeds [

21].

These variations also affect the electric current behavior at startup.

Figure 2 shows the input currents during the start of electric motors classes IE2, IE3 and IE4 with a 71% of rated voltage [

62]. For the LSPMM, in addition to presenting currents with non-uniform variations, the highest inrush current peak value is obtained, also its stabilization time until entering the synchronous state of the motor is longer for the LSPMM when compared to the other technologies.

In a permanent state, the LSPMM has a lower input current, as well as lower operating temperatures once the thermal equilibrium has been reached. However, this motor has higher harmonic current distortions, resulting in total harmonic distortions of up to 8% for the LSPMM, while for the SCIM classes IE2 and IE3 it does not exceed 3%, as shown in

Figure 3.

The better performance in the LSPMM is mainly due to the presence of permanent magnets, which contribute to the magnetization current decrease and, therefore, the input current. The stator current in IMs is the sum of two components; that is, the magnetization current, essential to create the magnetic field in the air gap and the current due to the load connected to the motor output. The magnetization current is present at all times of motor operation and in some 4-pole motors it can reach 50% of the nominal motor current [

58]. In addition, when the rotor attains the synchronous speed, the slip becomes zero and no electromagnetic field will be generated in rotor bars to produce rotor current (except for harmonic currents), which considerably reduces the temperature of the shaft and, therefore, of the other components in general [

25,

58].

## 3. Methodology

In order to compare the performance and temperature of induction motors classes IE2, IE3 and IE4 in the presence of harmonic voltage distortion of orders 2nd, third, 5th, 7th, experimental measurements were performed on a bench composed of a delta connected SCIM and an electromagnetic brake as electrical load. Tests were performed in the Amazon Energy Efficiency Excellence Center (CEAMAZON) in the Federal University of Pará (UFPA).

Figure 4 shows the general test setup.

At first, the induction motors were subjected to a perfect three-phase sine voltage of 220 V for 1 h and 10 min so that they reached their thermal equilibrium. In a second moment, the value of each harmonic voltage (2nd, 3rd, 5th and 7th) increased by 2% every 10 min until it reached 25%.

The voltage harmonics were generated using a three phase AC source (1), capable of generating a pure sine wave as well as harmonics (up to the 50th order) with different distortion magnitudes. For the study, the magnitudes of each harmonic voltage analyzed were increased every 10 min until reaching 25%. To measure the induction motor input parameters, class “A” HIOKI^{TM} power quality analyzer (2) model PW3198-90 was used, which recorded the input parameters during all the experiments at 1 s intervals.

The electric load used in this experiment consists of an electromagnetic brake or Foucault brake (3), which includes two load cells that are connected to the ends of the brake with which it is possible to measure the adjustable opposite force produced by eddy currents. When multiplied by the distance to the axis, it is possible to find the torque demanded by the load. For the test, a torque of 3.8 Nm was applied to the Foucault brake, which represents 92–95% of the nominal torque of motors (4). The nominal data of each motor are presented in

Table 2.

To measure the frame temperature, the FLIR

^{TM} infrared camera model T620 was used with a calculated emissivity of 0.94. In order to analyze the temperature variation in each motor class, the motor thermographic images were captured at two angles every 2 min, from the thermal equilibrium until the end of the experiment for each harmonic analyzed.

Figure 5a,b shows the angles photographed during the experiments.

Regarding the methodology used for the treatment of measurement data and obtaining the results,

Figure 6 presents the steps performed in the present work. At first, the tests were performed on the test bench presented in

Figure 4, for each of the motors analyzed and presented in

Table 1, then the motor input measurements were made using the Power Quality analyzer equipment as well as the thermographic images taken with the infrared camera, considering the measuring points of

Figure 5a,b. The next step was to transfer the measurement data from the equipments to the analyzer (PQA–HiVIEW PRO) and camera (FLIR ResearchIR) softwares. After data processing, they were converted to CSV format files, compatible for reading in Minitab [

55] statistical software. In Minitab, the data processed for plotting the results and the statistical analysis made on the study were analyzed.

## 4. Results and Discussion

#### 4.1. Current Increase Due to Harmonics

Of the three analyzed technologies, the LSPMM class IE4 had the lowest input current consumption for the same load percentage. However, the presence of voltage harmonics makes this scenario change. In

Figure 7a–c, the increase in input current of each motor is presented in the presence of voltage harmonics. In general, it can be seen how the second voltage harmonic turns out to be the most critical of the individual harmonics, resulting in the greatest increases in the line current, then, the combination of all harmonics results in the highest current demanded, which affects strongly the IE4 class LSPMM, which reaches currents up to two times its initial value. The fifth negative sequence harmonic results in a greater increase when compared to the seventh positive sequence harmonic for the three technologies. Third voltage harmonic did not result in any impact for the IE2 and IE3 class motors, however for the IE4 class motor it showed a slight increase, showing similar values with seventh voltage harmonic.

#### 4.2. Total Current Harmonic Distortion

Harmonic voltages produce harmonic currents, which, according to the order, percentage and motor technology, can result in negative impacts on the operation, as well as a reduction in its useful life. In addition to the electric current, it was commented that the LSPMM total current harmonic distortion (THDI) presented values of up to four times the THDI of the other technologies. To analyze the variation of this parameter,

Figure 8a–c presents THDI for the IE2, IE3 and IE4 class motors. It can be seen that the fifth harmonic does not produce a considerable variation of THDI in relation to its initial value, fifth and seventh harmonics result in uniform increases for the three technologies, reaching values around 50% and 40% for 25% distortion, respectively. The second voltage harmonic turns out to be much more damaging to the LSPMM, where THDI reaches over 150% and the combination of all results in values close to 175%, well above the IE2 and IE3 class motors, which show similar increases and do not exceed 150% of THDI.

This increase is mainly due to the increase in voltage distortion for each motor, but also to the appearance of new harmonics within the waveform. It was observed that with harmonic voltage distortion percentages higher than 8%, new harmonic currents appeared. It was observed that from percentages higher than 8% of voltage distortion, new harmonic currents appeared. In this way, with the presence of the fifth harmonic voltage, a seventh order harmonic current component appeared; while with the presence of seventh harmonic voltage, a fifth order harmonic current component appeared. With the presence of second harmonic voltage, a 4th order harmonic current component also appeared. All this contributes to the increase in THDI, occurring for all motors under study and being higher in the LSPMM due to the presence of permanent magnets.

#### 4.3. Reactive Power and Power Factor with Voltage Harmonics

The presence of permanent magnets contributes to the reduction of the magnetization current due to the magnetic fields generated in the air gap, with which a lower reactive power consumption is expected for the LSPMM. This can be observed in

Figure 9c, where for 0% harmonic distortion this motor has lower consumption than the IE2 and IE3 class motors (

Figure 9a,b). The presence of voltage harmonics results in a greater reactive power consumed, which varies according to the harmonic content. It can be seen that the fifth harmonic does not represent any considerable increase in this variable then the seventh harmonic, that despite being of positive sequence results in a slight increase of reactive power, while the fifth harmonic being negative sequence results in a higher consumption. Within these harmonics the hybrid motor has lower reactive consumption, followed by the IE3 class motor, being the high efficiency motor (IE2 class) the one that consume the most reactive power from the network.

The reactive power consumption is considerably increased with the presence of a second voltage harmonic, resulting in increases of up to 10 times that experienced with the aforementioned harmonics, the LSPMM being the most affected with this voltage harmonic.

The combination of all harmonics turns out to be the most damaging, reaching 2 kvar values for the LSPMM, which will result in a low power factor for this technology, as will be presented in the following figure.

Because the active power did not increase at the same rate as the reactive power, the motors suffered a decrease in their power factor. It can be seen in

Figure 10a–c how fifth and seventh harmonics result in slight decreases in the power factor, the fifth harmonic being more damaging, while the fifth harmonic remains varying over its initial value, except for the IE4 class hybrid motor, where it experiences a slight increase. For the second harmonic, it was already observed it produced large increases in current and reactive power, the power factor was also impacted with the presence of this harmonic, falling to values of down to 0.45 for the LSPMM, while the motor classes IE2 and IE3 have similar decreases with values close to 0.54, as presented. The presence of different combined harmonics in the supply voltage results in greatest decreases in power factor, it is observed based on the results that the presence of the second harmonic with negative sequence produces the greatest contribution in relation to the other harmonics present.

#### 4.4. Temperature Increase Due to Harmonics

Harmonics result in increases in the losses experienced by each motor and these losses vary according to the percentage of load, the level and type of harmonic content in the waveform, as well as the present technology. Because these losses are manifested primarily in the form of heat, the temperature is an indication of their increase with each harmonic analyzed. This increase is presented in

Figure 11a–c.

Initially, the motors have different operating temperatures with sinusoidal voltage without distortion, the IE3 class motor being the one with the highest operating temperature and the IE4 class motor having the lowest due to the lower operational current. The third zero sequence harmonic does not produce considerable increases in the temperature of the three IMs. Fifth and seventh harmonics result in similar increases for the three motors, however due to the higher initial temperature, the IE3 class motor reaches values close to 48 °C. With the second voltage harmonic, the hybrid motor experiments the greatest temperature increase, reaching values of 60 °C, while the IE3 class motor has the lowest temperature increase with this harmonic. In general, the IE4 class hybrid motor is the one that is most affected by the presence of harmonics in the supply voltage, while the IE3 class motor shows a greater tolerance for this type of disturbance.

#### 4.5. Correlation Matrix for Temperature

Harmonic voltages cause an increase in the line current, which results in an increase in losses and, consequently, in the motor temperature. To analyze the harmonic influence in the motor temperature, a Spearman’s correlation analysis was developed in Minitab 18 [

63], between the thermographic images data and the motor input parameters, in order to verify the relationship between these variables. Spearman’s correlation assesses the monotonic relationship between two variables. This correlation coefficient uses only the ranks of the values and not the values themselves. Thus, this measure is suitable for both ordinal and continuous variables. It is a useful test when Pearson’s correlation cannot be performed due to violations of normality, a non-linear relationship or when ordinal variables are being used [

64,

65,

66]. For this case and after finding a non-linear relationship between some variables, Spearman’s correlation method was used [

67]. The development of the Spearman’s rank correlation coefficient is presented in (1):

where

n is the number of value pairs and

${D}_{i}={X}_{i}-{Y}_{i}$ is the difference between each corresponding

${X}_{i}$ and

${Y}_{i}$ value rank.

In general, correlation analysis results in a number between −1 and +1, called the correlation coefficient. The higher the coefficient, the closer the relationship between the variables.

The analysis was performed for each harmonic considered in this study.

Figure 12,

Figure 13,

Figure 14,

Figure 15,

Figure 16 and

Figure 17 show the correlation matrices and the graphical representation between these variables for the second and third harmonic voltage in the IE2, IE3 and IE4 class motors, respectively.

In the correlation matrix, the upper cell shows the Spearman coefficient while the lower cell shows the p-value, useful for rejecting the null hypothesis when compared to the significance level (0.05 assumed). In the graphical representation, the temperature variation (Y axis) versus the second and third order harmonic voltages, line current, THDI, power factor and active power (X axis) is presented.

Regression (red) and smoother lowess (green) lines are also included within the graphics to better see and explore the potential relationships between the analyzed variables. In this way, where high correlation coefficients are obtained, the variables show similar variation patterns, while where the coefficients have values close to zero, no similar variation patterns are observed, as in the case of the third harmonic voltage.

For the IE2 class motor, the second harmonic voltage is presented in

Figure 12a and

Figure 13a, Spearman coefficients are observed quite close to ±1, which indicates a high correlation between the variables present in the matrix. In addition, the

p-value is zero for each second harmonic correlation in the motors, presenting lower values, compared with the level of significance (α = 0.05). This behavior is also observed for the IE3 and IE4 class motors, for which a non-linear initial growth is also observed.

A different scenario is observed for the fifth harmonic voltage, for which the electric motor delta-connected is an open circuit and not considerable effects are expected. In

Figure 12b, it is observed that all parameters have low correlation values between them, especially in relation to temperature. There is practically no solid relationship between the variables and it occurs when the relationship is random or non-existent, showing low correlation coefficients. This behavior is observed in

Figure 13b.

It is observed for the second harmonic voltage that, in relation to the temperature, all variables have high correlation values, which indicates that when one of them increases, the temperature also does. Among the other variables analyzed, it is observed that high correlation values are also observed, with which the co-linearity must be analyzed when creating models involving these variables. The only variable that varied during the experiment was the voltage distortion, which as mentioned increased in percentages of two until reaching 25%.

For the IE3 class motor,

Figure 14b and

Figure 15b, the third harmonic has similar results compared to the IE2 class motor, presenting low correlation values, and in this case the

p-value is greater than the significance value for some correlations, with which it is not possible to reject the null hypothesis, so it is not possible to state that there is a relationship between the variables in question.

Third harmonic voltage shows different results for the IE4 LSPMM, shown in

Figure 16b and

Figure 17b. Higher correlation coefficients can be observed for the LSPMM, but not as strong as those obtained for the second harmonic voltage. The

p-value also remains below 0.05 among most of the variables, as shown. In general, with the second negative sequence harmonic, the temperature has a defined growth pattern from its initial temperature for the three analyzed induction motors, being lower for the LSPMM, to a temperature close to 60 °C for the three technologies. For the third zero sequence harmonic, the temperature varies around its initial value for the entire experiment. This scenario is different for the LSPMM where, despite showing a lower correlation, a growth pattern is observed in

Figure 17b.

## 5. Conclusions

Considering the electric motors classes IE3 and IE4 as strong candidates for replacing old motors, this study analyzed the impact that different orders harmonic voltages have on the electric motors’ classes IE2, IE3 and IE4, the latter of permanent magnets and squirrel cage. A summary of the results is listed below:

Second negative sequence harmonic voltage proved to be the most damaging for electric motors, of which the LSPMM shows the worst performance, while the IE3 class motor presented the smallest variations.

The seventh harmonic also resulted in uniform increases in all motors, however smaller than those found for the second and fifth harmonics of negative sequence.

The third zero sequence harmonic did not produce considerable variations in electric motors, where the parameters showed variations around their initial values.

The combination of all harmonics proved to be more damaging than each individual harmonic analyzed, of which the second harmonic had the greatest contribution.

This work also sought to analyze, through a technical comparison, the feasibility of substitution between technologies, based on the results, for substitution between older and/or non-efficient motors, with higher efficiency motors, some aspects should be considered:

More efficient motors can result in greater savings in energy and economic terms, mainly in systems with good power quality.

An analysis of the electricity supply quality at the installation site must be carried out before replacement. Poor power quality reduces electric motor efficiency. Regarding the LSPMM, although in ideal operating conditions, it presented lower current, reactive power and operational temperature, with the presence of harmonics in the supply voltage it presents the worst performance of the three motors analyzed, being more affected by the second harmonic of negative sequence. Another factor to consider is the distortion presented by the LSPMM, which initially already presents values superior to that of the other technologies, and with the presence of harmonics, due to the presence of permanent magnets, higher percentages of THDI are found for this technology. Therefore, in large-scale applications, studies on the quality of the supply before and after installation must be carried out.

The application type must also be considered. For the LSPMM, it was observed at the moment of starting a difficulty to start with a load, this can be critical mainly for applications with frequent start/stop cycles.

For future replacements, the economical part is fundamental, the operating time of the old motor in order to verify the payback due to the higher costs of higher efficiency motors. Currently in the Brazilian market, the cost of the IE4 class motor is approximately 1.3 times the cost of the IE3 class motor, while the IE3 class motor is 1.3 times the cost of the IE2 class motor.

Finally, a correlation analysis between the variables recorded during the experiments was developed, as well as regression graphs, which aimed to analyze the relationship between the variables recorded in the experiments. The verification of the relationship between the variables was performed using the significance level (p-value). There was a great multicollinearity between them, which should be considered when making models for these parameters.

In relation to the second harmonic, high correlation coefficients were obtained, both from the temperature and from the other variables, confirming the high collinearity that exists between them. It was commented that the third harmonic voltage did not produce any considerable impact on the electric motors. This was confirmed with the regression graphs that show how the values do not exhibit a growth pattern as for the second harmonic analyzed.