**5.4. Load failure patterns**

The vast majority of the published papers about failure monitoring via current spectrum analysis presents the failure patterns related to broken bars and air gap eccentricity. This chapter presents a very meaningful contribution to the previous works since it adds new patterns related to the attached load. All the patterns have been tested, first through controlled laboratory tests and later through industrial cases. The failure patterns can be divided in three groups: motor failure, transmission system failure and attached load failure. By using the induction motor as a transducer, one can monitor the complete drive train, i.e., motor, transmission system and attached load, so as to increase the reliability of productive system.

### *5.4.1. Transmission System Failure*

508 Induction Motors – Modelling and Control

**5.3. Bearing failure patterns** 

**Figure 15.** Bearing failure modes

**5.4. Load failure patterns** 

productive system.

control.

The failures related to the stator windings present a diversified set of possible manifestations according to the Figure 14. It is possible to notice their simultaneous occurrence. There are MCSA patterns for the detection of these failures, but EPVA is the most recommended technique to detect electrical imbalance in motors without direct torque

The monitoring of bearing damages is very important in predictive maintenance program since these problems account for 40% of the total amount of failures in an induction motor (Schoen et al., 1995). Many papers have recommended current signature analysis for the diagnosis of bearing faults, although it is important to register that this is an area that can be

There are several causes for bearing damages. Since this is not the objective of this work, the chapter presents just the characteristic components of failure in the outer and inner races, and rolling elements. The pattern is given by the Figure 15; where FBPFO is the rolling element characteristic frequency, FBPFI is the inner race characteristic frequency, FBSF is the outer race characteristic frequency, FFTF is the cage characteristic frequency, PD is the bearing pitch diameter, BD is the ball bearing diameter, β is the contact angle, n is the

The vast majority of the published papers about failure monitoring via current spectrum analysis presents the failure patterns related to broken bars and air gap eccentricity. This chapter presents a very meaningful contribution to the previous works since it adds new patterns related to the attached load. All the patterns have been tested, first through controlled laboratory tests and later through industrial cases. The failure patterns can be divided in three groups: motor failure, transmission system failure and attached load failure. By using the induction motor as a transducer, one can monitor the complete drive train, i.e., motor, transmission system and attached load, so as to increase the reliability of

more explored and improved, tracking earlier fault detection.

number of rolling elements, and Fr is the rotational speed.

The MCSA monitors the frequency components related to pulleys (motor pulley and load pulley), belts and gear mesh. It has been observed that load problems can reflect in the transmission system frequency components. This characteristic is one more way of detecting mechanical load failures to be used in addition to the load characteristic frequency components.

1. **Pulleys**: by analyzing the rotational frequency one can detect problems related to the motor pulley. When there is no change in the speed, it is not possible to distinguish the damaged pulley from the healthy one since they have the same rotational frequency. But when a speed transformation is present, one can monitor the load pulley and the attached load through the pattern presented in Figure 16. In this case, *flf* is equal to *fpulley*, and *fpulley* is the load pulley characteristic frequency given by (45).

$$f\_{pullcy} = \frac{D\_{motor\\_pulley} \times f\_r}{D\_{load\\_pulley}} \tag{45}$$

Where *fr* is the rotational frequency, *Dmotor\_pulley* is the diameter of the motor pulley and *Dload\_pulley* is the diameter of the load pulley. The sideband components of the fundamental are at *f*<sup>1</sup> *fp*.

**Figure 16.** Load Pulley Pattern

The most common problems are eccentric pulley, pulley with mechanical looseness and unbalanced pulley. Problems related to the load can also reflect in the same frequencies. When this happens, the analyst himself must cross pieces of information from other spectrum regions so as to arrive at a reliable conclusion.

2. **Belts**: the first step when monitoring the belt characteristic frequency components is to calculate the belt frequency (*fb*). In this case, *flf* is equal to *fb*, and *fb* is the belt characteristic frequency given by (46).

$$f\_b = \frac{D\_{\text{motor\\_pulley}} \times \pi \times f\_r}{L\_{belt}} \tag{46}$$

Where *Lbelt* is the belt length

This way the sideband components of the fundamental are at *f*<sup>1</sup> *fb*. After calculating this frequency, it is enough to follow the pattern presented in Figure 17 and follow up the tendency curve in order to diagnose problems in this transmission system element.

**Figure 17.** Belt Failure Pattern

Besides diagnosing problems such as loosen belt, broken belt or too taut belt, one can analyze problems originating in the load. In case of load failure, the vibration levels in the belts increase considerably and result in higher amplitudes for the belt characteristic frequencies.

3. **Gear Mesh**: in this case, two spectrum regions must be monitored. The first one, in a lower frequency band, shows punctual failure in the gear (for instance, a broken tooth). These frequencies are related to the rotational frequencies before and after the speed transformation. This way the sideband components of the fundamental are at *f*<sup>1</sup> *fr*1 and *f*<sup>1</sup> *fr*2 respectively. Where *fr*1 is the rotational frequency before the speed transformation and *fr*2 is the rotational frequency after the speed transformation. The second spectrum region of interest shows distributed failures in the gear. They are known as gear mesh frequency (*fg*) and can be calculated by multiplying the rotational shaft speed by the gear teeth number Figure 18a illustrates this situation, and Figure 18b shows the sideband components of the fundamental are at *f*<sup>1</sup> *fg*.

$$f\_{\mathcal{g}} = \mathbf{n} \cdot f\_{r1} = \mathbf{N} \cdot f\_{r2} \tag{47}$$

Predictive Maintenance by Electrical Signature Analysis to Induction Motors 511

*vp r pump* \_ *f nf* (48)

As seen previously, a load fault reflects in the motor stator current by means of torque oscillations. This chapter presents in this section three different kinds of loads and their respective patterns. Other load types result in different patterns but the fundamental sequence is always the same: define the characteristic frequencies from the constructive data, find their presence in the motor current signature due to torque oscillations from load

1. **Centrifugal Pumps:** for the analysis of centrifugal pumps one has to consider the pump rotational frequency (*fr\_pump*) and the vane pass frequency (*fvp*) that is given by:

The analysis of the pump rotational frequency (*fr\_pump*) indicates problems related to misalignment or pump imbalance. In this case, *flf* is equal to *fr\_pump* and the sideband components of the fundamental are at *f*<sup>1</sup> *fr\_pump*. On the other hand, the increase of the amplitudes of vane passing frequency indicates problems inside the pump, such as vane deterioration. Now the sideband components of the fundamental are at *f*<sup>1</sup> *fvp*. Figure 19

In addition to those frequencies one has to monitor the increase of saliencies close to the supply frequency. These frequencies are characteristic of pump signature and also can

2. **Screw Compressor:** the complete set motor, gear mesh and screw compressor can be monitored by means of MCSA satisfactorily. The motor and the gear mesh can be analyzed according to the patterns presented previously. Figure 20a shows the scheme of a screw compressor. Where N is the motor gear teeth number, n is the compressor gear teeth number, Lm is the male screw lobules number, Lf is the female screw lobules number, Fr is the motor rotational frequency, Fr1 is the male screw rotational frequency, Fr2 is the female screw rotational frequency and Fp is the pulsation frequency. The screw

compressor failure spectral pattern is presented in Figure 20b.

*5.4.2. Attached load failure* 

faults, analyze the tendency curve and diagnose the fault.

Where *n* is the number of pump vanes.

shows the pattern for these frequencies.

**Figure 19.** Centrifugal pump failure pattern

indicate pump problems.

**Figure 18.** Gear features: (a) gear mesh, and (b) gear mesh failure pattern

### *5.4.2. Attached load failure*

510 Induction Motors – Modelling and Control

This way the sideband components of the fundamental are at *f*<sup>1</sup> *fb*. After calculating this frequency, it is enough to follow the pattern presented in Figure 17 and follow up the

Besides diagnosing problems such as loosen belt, broken belt or too taut belt, one can analyze problems originating in the load. In case of load failure, the vibration levels in the belts increase considerably and result in higher amplitudes for the belt characteristic frequencies.

3. **Gear Mesh**: in this case, two spectrum regions must be monitored. The first one, in a lower frequency band, shows punctual failure in the gear (for instance, a broken tooth). These frequencies are related to the rotational frequencies before and after the speed transformation. This way the sideband components of the fundamental are at *f*<sup>1</sup> *fr*1 and *f*<sup>1</sup> *fr*2 respectively. Where *fr*1 is the rotational frequency before the speed transformation and *fr*2 is the rotational frequency after the speed transformation. The second spectrum region of interest shows distributed failures in the gear. They are known as gear mesh frequency (*fg*) and can be calculated by multiplying the rotational shaft speed by the gear teeth number Figure 18a illustrates this situation, and Figure 18b shows the

*<sup>g</sup> r r* 1 2 *f nf Nf* (47)

sideband components of the fundamental are at *f*<sup>1</sup> *fg*.

**Figure 18.** Gear features: (a) gear mesh, and (b) gear mesh failure pattern

tendency curve in order to diagnose problems in this transmission system element.

Where *Lbelt* is the belt length

**Figure 17.** Belt Failure Pattern

As seen previously, a load fault reflects in the motor stator current by means of torque oscillations. This chapter presents in this section three different kinds of loads and their respective patterns. Other load types result in different patterns but the fundamental sequence is always the same: define the characteristic frequencies from the constructive data, find their presence in the motor current signature due to torque oscillations from load faults, analyze the tendency curve and diagnose the fault.

1. **Centrifugal Pumps:** for the analysis of centrifugal pumps one has to consider the pump rotational frequency (*fr\_pump*) and the vane pass frequency (*fvp*) that is given by:

$$f\_{vp} = \mathbf{n} \cdot f\_{r\\_pump} \tag{48}$$

Where *n* is the number of pump vanes.

The analysis of the pump rotational frequency (*fr\_pump*) indicates problems related to misalignment or pump imbalance. In this case, *flf* is equal to *fr\_pump* and the sideband components of the fundamental are at *f*<sup>1</sup> *fr\_pump*. On the other hand, the increase of the amplitudes of vane passing frequency indicates problems inside the pump, such as vane deterioration. Now the sideband components of the fundamental are at *f*<sup>1</sup> *fvp*. Figure 19 shows the pattern for these frequencies.

**Figure 19.** Centrifugal pump failure pattern

In addition to those frequencies one has to monitor the increase of saliencies close to the supply frequency. These frequencies are characteristic of pump signature and also can indicate pump problems.

2. **Screw Compressor:** the complete set motor, gear mesh and screw compressor can be monitored by means of MCSA satisfactorily. The motor and the gear mesh can be analyzed according to the patterns presented previously. Figure 20a shows the scheme of a screw compressor. Where N is the motor gear teeth number, n is the compressor gear teeth number, Lm is the male screw lobules number, Lf is the female screw lobules number, Fr is the motor rotational frequency, Fr1 is the male screw rotational frequency, Fr2 is the female screw rotational frequency and Fp is the pulsation frequency. The screw compressor failure spectral pattern is presented in Figure 20b.

The screw compressor analysis takes into consideration three characteristic frequencies:

a. Male screw rotational frequency: in this case, *f*l*<sup>f</sup>* = *fr*1 and *fr*1 is the male screw rotational frequency given by (51). The sideband components of the fundamental are at *f*<sup>1</sup> *fr*1.

**Figure 20.** Screw compressor: (a) schematic and (b) failure spectral pattern

$$f\_{r1} = \frac{N}{n} \cdot f\_r \tag{49}$$

Predictive Maintenance by Electrical Signature Analysis to Induction Motors 513

indicates problems like blade deterioration or break. The sideband components of the

fundamental are given by *f*<sup>1</sup> *fbp*. Figure 21 shows the fan failure patterns.

**6. Elements of a monitoring system for predictive maintenance** 

most of the time, using artificial intelligence techniques (Lambert-Torres et al. 2009).

A sophisticated monitoring system can read the entrances of hundreds of sensors and execute mathematical operations and process a diagnosis. Currently, the diagnosis is gotten,

Considering the previous statements, a monitoring system can be divided in four main stages: (a) transduction of the interest signals; (b) acquisition of the data; (c) processing of the acquired data; and (d) diagnosis. Figure 22 presents a pictorial form of this process.

**Motor Transduction Acquisition Processing Diagnosis** 

A transducer is a piece of equipment that has in its entrance an input value to be monitored (current, voltage, acceleration, temperature, etc), whereas in its output it has a signal that is conditioned and envoy to the acquisition system and processing. The main transducers used

 For measurement of temperature: they are the three main methods of measurement of temperature: thermocouple, thermister, and RTD (Resistance Temperature Detection). For the measurement of vibration: two types of transducers for the vibration analysis exist: the absolute transducers or with contact and the relative ones or without contact. The absolute transducers measure the real movement of the machine, whereas the relative ones measure the movement of an element of the machine in relation to the other element. The accelerometer is the main and more used existing absolute sensor in

**Figure 21.** Fan failure pattern

**Figure 22.** Steps of the Monitoring Process

in the monitoring processing of electric machines are:

**6.1. Transduction** 

the market.

b. Female screw rotational frequency: in this case, *f*l*<sup>f</sup>* is equal to *fr*2 and *fr*2 is the female screw rotational frequency given by (50). The sideband components of the fundamental are at *f*<sup>1</sup> *fr*2.

$$f\_{r2} = \frac{L\_m}{L\_f} \cdot f\_{r1} \tag{50}$$

c. Pulsation frequency: in this case, *f*l*<sup>f</sup>* is equal to *fp* and *fp* is the pulsation frequency given by (51). The sideband components of the fundamental are at *f*<sup>1</sup> *fp*.

$$f\_p = L\_m \cdot f\_{r1} = L\_f \cdot f\_{r2} \tag{51}$$

When the screw compressor has two stages, it is enough to apply the same reasoning for the second stage of compression. Since the speed transformations are different, the characteristic component of each stage can be separated in the spectrum.

3. **Fans:** in the same way of pumps, fan failure analysis considers the fan rotational frequency and the blade passing frequency (*fbp*):

$$f\_{bp} = \mathcal{N}\_b \times f\_{r\_{-f\_{\text{int}}}} \tag{52}$$

Where *Nb* is the number of blades and *fr\_fan* is the fan rotational frequency.

Analyzing the rotational frequency (*fr\_fan*), problems related to misalignment or fan imbalance can be detected. When, *flf* is equal to *fr\_fan* and the sideband components of the fundamental are at *f*<sup>1</sup> *fr\_fan*. Also, the increase of the amplitudes of blade passing frequency indicates problems like blade deterioration or break. The sideband components of the fundamental are given by *f*<sup>1</sup> *fbp*. Figure 21 shows the fan failure patterns.

**Figure 21.** Fan failure pattern

512 Induction Motors – Modelling and Control

are at *f*<sup>1</sup> *fr*2.

The screw compressor analysis takes into consideration three characteristic frequencies:

**Figure 20.** Screw compressor: (a) schematic and (b) failure spectral pattern

a. Male screw rotational frequency: in this case, *f*l*<sup>f</sup>* = *fr*1 and *fr*1 is the male screw rotational frequency given by (51). The sideband components of the fundamental are at *f*<sup>1</sup> *fr*1.

> *r r* 1 *<sup>N</sup> <sup>f</sup> <sup>f</sup>*

(a) (b)

b. Female screw rotational frequency: in this case, *f*l*<sup>f</sup>* is equal to *fr*2 and *fr*2 is the female screw rotational frequency given by (50). The sideband components of the fundamental

> 2 1 *m r r f*

c. Pulsation frequency: in this case, *f*l*<sup>f</sup>* is equal to *fp* and *fp* is the pulsation frequency given

When the screw compressor has two stages, it is enough to apply the same reasoning for the second stage of compression. Since the speed transformations are different, the characteristic

3. **Fans:** in the same way of pumps, fan failure analysis considers the fan rotational

Analyzing the rotational frequency (*fr\_fan*), problems related to misalignment or fan imbalance can be detected. When, *flf* is equal to *fr\_fan* and the sideband components of the fundamental are at *f*<sup>1</sup> *fr\_fan*. Also, the increase of the amplitudes of blade passing frequency

by (51). The sideband components of the fundamental are at *f*<sup>1</sup> *fp*.

Where *Nb* is the number of blades and *fr\_fan* is the fan rotational frequency.

component of each stage can be separated in the spectrum.

frequency and the blade passing frequency (*fbp*):

*<sup>n</sup>* (49)

*<sup>L</sup> f f <sup>L</sup>* (50)

*<sup>p</sup> mr f r* 1 2 *f Lf Lf* (51)

*bp b r fan* \_ *f Nf* (52)
