IT202000007249A1 - METHOD AND SYSTEM FOR DIAGNOSIS OF FAULTS IN A MACHINERY DRIVEN BY AN ELECTRIC MOTOR. - Google Patents

Description

DESCRIPTION

accompanying a patent application for an industrial invention entitled:

? METHOD AND SYSTEM FOR DIAGNOSING FAULTS IN A MACHINERY DRIVEN BY AN ELECTRIC MOTOR ?.

TEXT OF THE DESCRIPTION

The present invention refers to a method and system for diagnosing faults in a machinery operated by an electric motor, in particular for types of faults linked to a transmission system connected to a rotor of the electric motor and also for types of faults linked to the power circuit (inverter) of the motor or to the motor itself.

Fully automated production plants are known in which Cyber-Physical Systems (CPSS) automatically manage industrial processes. These production plants include a plurality of of rotating machines. A rotating machine comprises an electric motor having a rotor connected to transmission systems such as drive shafts, toothed wheels, gears, pulleys, belts, etc., which transmit the rotary motion of the motor rotor.

These transmission systems could be subject to failures and not work properly due to various factors, such as wear, construction inaccuracies, assembly errors, overload, etc. The same goes for the electric motor or for the drive circuit of the electric motor, which deals with the management of the power to be supplied to the motor in order to carry out certain movements of loads.

Failures that occur in the transmission systems rather than the engine etc. reduce productivity. of the factory and this becomes particularly critical in fully automated environments. Therefore, fault detection in automation systems plays a key role in avoiding further losses. The design of a reliable and effective PdM (Predictive Maintenance) system can? help to overcome the inconveniences deriving from breakdowns. For these reasons, the PdM is receiving a very important role, due to the fact that a trivial failure of a transmission system could stop an entire production line.

Studies have been done on the analysis of a transient state of an electrical signal in an attempt to understand what happens to the current signal of an electric motor that drives a belt when the slack and slippage of the belt increases. However, these problems in industrial applications are simply avoided by using toothed belts, so that no slippage occurs.

Some studies rely on analyzing the current spectrum of an electric motor stator to detect faults in belt drive systems. Multiples of the belt's natural frequency in the electrical signal indicate problems. In this case, the motors have been tested with a speed reference? constant and fixed. However, this hypothesis of speed? constant of the motor not? suitable for industrial robotic systems, which often operate at high accelerations, in which the reference profile of the speed? of the rotor changes over time and the engine RPM changes accordingly. In this case the changes of the signal in frequency and time and therefore the classical spectral analysis by Fast Fourier Transform (FFT) give rise to an average of all the frequencies of the spectrum that make it difficult or impossible to extract precise characteristics from the signals. For similar reasons also the short-term Fourier transform (STFT)? useless with such types of signals.

However, studies have been carried out on MCSA (motor current signature analysis) techniques based on Fourier transform and STFT of the electric signal of the motor to detect motor failures. However, with these studies, the results obtained only work with speed? motor constants (stationary case) and are not applicable to most real cases.

Purpose of the present invention? to eliminate the drawbacks of the known art by providing a method and system for diagnosing faults in a machinery operated by an electric motor, which are reliable, versatile and suitable for use in the diagnosis of tastes in transmission systems, electric motors and inverters in any working condition both in permanent and transitory regime.

These objects are achieved in accordance with the invention with the characteristics of the independents.

Advantageous embodiments of the invention appear from the dependent claims.

The method and the system according to the invention are defined by the independent claims.

Further features of the invention will appear more? clear from the detailed description that follows, referring to a purely exemplary and therefore non-limiting embodiment thereof, illustrated in the attached drawings, in which:

Fig. 1? a block diagram schematically illustrating the method and the diagnosis system according to the invention;

Fig. 2? a time domain graph of a current signal absorbed by the electric motor during its operation;

Fig. 3? a time domain graph of the current signal of Fig. 2, of which? its envelope was evaluated;

Fig. 4? a table illustrating the approximation levels and the levels of detail obtained with a 10-level wavelet decomposition of the signal of Fig. 3; Fig. 5 shows graphs illustrating the signal levels of Fig. 4;

Fig. 6? a table illustrating values of an index calculated experimentally and used to determine at least one threshold index;

Fig. 7? a schematic view of a belt tensioning system; And

Fig. 8? a schematic view showing a micro tensioner applied to a belt.

Fig. 1 illustrates an electric motor (1) having a rotor (10) connected to a transmission system (2). The transmission system (2) can? comprise a belt (20) returned on pulleys (21), in which a pulley? connected to the rotor (10) of the motor.

The electric motor (1) has a stator (11) with stator windings. An electrical power system (3)? electrically connected to the stator windings (11) to power the electric motor. A current sensor (4)? arranged so as to detect an electric current supplying the electric motor (1). This current sensor (4) could be arranged in the power supply system (3), in the electric motor (1) or in the electrical cables connecting the power system to the electric motor.

In case the electric motor (1)? a three-phase alternating current (AC) motor, the electrical power supply system (3) comprises an inverter connected by three phases (U, V, W) to the stator windings (11).

During the operation of the electric motor (1), the currents in the three phases (U, V, W) are identical and out of phase with each other by 120 degrees. Therefore the current sensor (4) could be placed in any of the three phases (U, V, W).

The current sensor (4) detects a current signal s (t) in the time domain, indicative of the current absorbed by the electric motor (1).

Fig. 2 illustrates an example of the current signal s (t) detected by the current sensor (4).

The current signal s (t)? sent to a digital signal processor (DSP) (5) configured to perform envelope filtering of the current signal and obtain a filtered current signal S (t) in the time domain. The DSP (5) can? include a Hilbert filter taken in absolute value that performs a Hilbert transform in the time domain, or an equivalent envelope filter.

Fig. 3 illustrates the filtered current signal S (t) outgoing from the DSP (5). The filtered current signal S (t)? an equivalent low-pass signal of the current signal s (t) detected by the current sensor (4), which has the same bandwidth, but only positive frequencies. Such envelope filtering eliminates unnecessary redone elements of the signal and leaves only useful signal components for subsequent signal processing.

The filtered current signal S (t) is analyzed and decomposed by means of a multi-resolution analysis with wavelet transform which will later come? called wavelet decomposition (6). In this way, a decomposition of the signal into different levels and frequency sub-bands is obtained. The frequency content of the current absorbed by the electric motor (1) is then divided into different signal levels which are called approximation levels (a1,? A10) and detail levels (d1 ,? d10).

Fig. 4 illustrates a table 1 which reports 10 levels of approximation and 10 levels of detail, divided into frequency bands.

Fig. 5 illustrates the graphs of the approximation levels and the levels of detail, relative to the filtered current signal S (t).

The wavelet decomposition (6) applied on the filtered current signal S (t), returns a vector of coefficients C (C1, C2,?. Cn) which are representative of energy levels of detail Edi and energy levels of approximation Eai.

These coefficients of the coefficient vector (C) are processed by a? Unit? processing (F), in order to obtain an index (Wx) relative to the x-th level of detail (dx).

The exact total energy of the filtered current signal S (t) is calculated with the following formula:

Et = C <2> (F1)

Specifically, if we want to consider the energy content (Ed10) which corresponds to the tenth level of detail (d10), this energy content is calculated by the following formula:

Ed10 = 100 * (C10) <2> / Et (F2) Where C10 are the coefficients extracted from the vector of the coefficients C which belong, specifically, to the level of detail d10 and the total energy (Et)? calculated by the formula (F1).

Then the energy content of the signal is evaluated as a percentage of the energy of the entire signal. Considering that the entire signal has an energy content Et equal to 100% of the signal, it? represented by the sum of the energy content of each single level of detail [Ed1,? Ed10], and by the sum of the energy content of each single approximation level [Ea1,? Ean].

The sum of the energy content of the approximation levels? given by the following formula:

Ea = [Ea1 + Ea2 +? + Ea10] (F3) Then is calculated the index (W10) relative to the level of detail pi? high, that is? at the tenth detail level (d10). This index (W10)? equal to the ratio between the energy content (Ed10) of the tenth level of detail (given by the formula (F2)) and the sum of the energy content of the approximation levels (Ea) (given by the formula (F3)), that is? the index (W10)? given by the following formula

W10 = Ed10 / Ea (F4) Experimentally, at least one threshold frequency value (f1, f2) of the belt vibration (20) is identified. For example, two threshold frequency values (f1, f2) are identified to identify three frequency ranges (T1, T2, T3) in which the belt works safely, in almost critical conditions and in critical conditions.

When the belt (20) vibrates at a frequency below the threshold frequency value, does this mean that the belt is not? tensioned correctly and therefore c ?? danger of failure; instead, when the belt vibrates at a frequency greater than the threshold frequency value, it means that the belt? tensioned correctly and works safely.

Advantageously, two threshold frequency values (f1, f2) are identified which identify three frequency ranges (T1, T2, T3). For example the first threshold frequency value? f1 = 50 Hz and the second threshold frequency value? f2 = 32 Hz). In this case it is obtained

- a first frequency range (T1) for frequencies greater than or equal to 50Hz (ie T1> 50 Hz), which identifies safe operation of the belt;

- a second frequency range (R2) for frequencies between 32 Hz and 50 Hz excluding the extremes (ie 32Hz <T2 <50Hz), which identifies an almost critical operation of the belt; And

- a third frequency range (R3) for frequencies lower than or equal to 32Hz (ie T3 <32 Hz), which identifies a critical operation of the belt.

Experimentally (using the system of Fig. 1) have a plurality been calculated? of reference values (W10R) of the index (W10) for certain known vibrations of the belt. These reference values (W10R) are used to identify at least one threshold value (W10T) which defines a threshold between a nominal operation of the equipment and a fault.

Since they were already? the frequency ranges (T1, T2, T3) in which the belt works safely or in almost critical or critical conditions have been identified, the reference values (W10R) of the index can be divided into groups of index reference values (W10T) corresponding to the respective frequency ranges (T1, T2, T3) which identify different operating conditions.

Table 2 of Fig. 6 illustrates some reference values (W10R) of the index calculated experimentally, by varying the tensioning of the belt (20). For each frequency of vibration of the belt, six reference values of the index are reported which represent six acquisitions made; that is to say six reference values of the index calculated on six signals for each frequency value indicated in table 2. By way of example, reference values of the index at 64 Hz and 53Hz have been calculated (in the first frequency range (T1) (safe condition)), at 43 Hz and at 35 Hz (in the second frequency range (T2) (almost critical condition)) and at 23 Hz (in the third frequency range (T3) (critical condition)).

From the reference values (W10R) you can? identify at least one threshold value (W10T) that defines a threshold between a nominal operation of the equipment and a fault. By way of example, in Table 2 two threshold values W10T1 = 0.0060 and W10T2 = 0.0098 have been identified. When is the value of the index W10? less than the first threshold value W10T1 the equipment works in nominal conditions; when the value of the index W10? between the first threshold value W10T1 and the second threshold value W10T2 the equipment works in almost critical conditions. When is the value of the index W10? greater than the second threshold value W10T1 the equipment? in critical condition.

The reference values (W10R) of the index and the at least one threshold value (W10T). experimentally calculated. are stored in a memory (7).

During the operation of the motor (1) and the rotation of the belt (20), the index (W10)? calculated continuously. This index (W10)? sent to a comparator (8) which compares it with at least one threshold value (W10T) stored in the memory (7).

If the index (W10) calculated continuously? greater than or equal to at least one threshold value (W10T) an alarm is activated (9).

In particular if the index (W10)? less than the first threshold value, the belt is vibrating in the first operating frequency range (T1) and no alarm is activated.

If the index (W10)? between the first and the second threshold value, the belt is vibrating in the second range of operating frequencies (T2) and the alarm (9) is activated which indicates an almost critical operation.

If the index (W10)? the greater the second threshold value, the belt is vibrating in the second range of operating frequencies (T2) and the alarm (9) is activated which indicates a critical operation of the belt.

From the values of table 2 of Fig. 6,? It is possible to notice how as the tension of the belt decreases (and therefore the vibration frequency of the belt decreases) the energy content of the current signal detected by the current sensor (4) moves from intervals at a frequency pi? high (which are those with the lowest level of detail ie [d7, d8, d9]) up to the last level of detail d10 which? the one at frequency pi? low. So the percentage energy of the last level of detail d10 increases with respect to that of the other levels. These experimental tests show that the last level of detail d10 is the significant level to understand at which range the belt is working.

By way of example, with reference to Fig. 7, it is described how? The determination of the frequency ranges (T1, T2, T3) of the belt vibration has been carried out.

A static tension (Fp) of the belt is measured a priori. The optimal static voltage value? calculated following the references contained in the belt datasheet provided by the manufacturer. The static voltage value (Fp) must take into account for? of the system on which it will come? mounted the belt, considering the various inertia that will go? to move

Is the natural frequency (Fr) of the necessary belt calculated, so that? the belt has sufficient static tension to handle the various loads, applying the following equation:

Fp = m * (2l * Fr) <2> (1)

Where l = length of the belt in meters,? = mass of the belt per unit? of length

Pi? precisely, the driving pulley (P1)? connected directly to a speed reducer connected to the motor rotor; there are two other pulleys (P2, P3) placed at the same height, which simply guarantee correct alignment and correct stability? to the belt.

In the particular case under consideration, the belt (20)? fixed at the ends? of an axis on which a robot moves. The motor acting on the rotation of the toothed belt (P1) does s? that the robot is able to move along the entire length of the axis.

Mt represents the torque.

Mt = Fu * r

Where was it? the peripheral force.

Fp = 2 Fu

The datasheets of the belts show the close relationship between the life time of the transmission system (RUL) and the adequate tensioning of the belt, therefore consequently the reliability? of the entire transmission system. In fact, for inadequate tensioning, the belt may not be able to have the same performance and can? not being able to withstand the loads to which it is subjected, with consequent deformation or breakage of the same.

The static voltage (Fp)? the force necessary to avoid skipping of the belt on the pulley during normal operation.

The correct static voltage value? equal to Fp = 2 * Fu. The peripheral force Fu pu? be calculated in different ways, usually starting from the maximum weight that the robot can? transport and from the inertia of the axes.

For each axis, the necessary belt pretensioning values (Fp) are reported, calculated as frequency values of a free section. In other words, following equation (1), once the static tension (Fp) necessary for the belt to perform the movements in safety has been calculated, the corresponding frequency value in Hz is calculated, depending on the free length of the belt.

With reference to Fig. 8, a micro-tensioner or sound level meter (M) is used to measure the vibration frequency of the belt (20). The micro tensioner (M) has a sensitive head, which is held above the belt (20). When the belt is pinched while stationary, it begins to vibrate at its natural frequency, the vibrations are detected by the micro tensioner (M) and the vibration frequency is then displayed on the micro tensioner screen (M).

The micro-tensioner (M)? used in order to find the vibration range (T1) in which the belt works in a safe condition. In fact, when the static tension (Fp) of the belt, calculated as the vibration frequency value, falls below the value in Hz corresponding to the value of the necessary peripheral force (Fu), the static tension (Fp) does not? pi? sufficient to move all loads and the belt (20) begins to deform during movement. Just this phenomenon d? origin of various common failures (breakage, abrasions, broken belt pieces, damaged belt teeth, etc.).

In this way, by inverting the equation (1), one can? obtain the correlation between the desired static pressure value Fp (usually calculated by the mechanical department that installs the belt), which corresponds to a certain natural frequency of the belt in Hz:

Table 3 shows the values in Hz of the vibration frequency of the belt, as the length (l) of the free belt section varies.

Table 3

Following these studies, yes? why? thought of identifying three frequency ranges (T1, T2 and T3) experimentally detected on a free section l = 500 mm,

The first frequency range (T1) occurs for frequencies higher than 50 Hz in which there is a correct static pressure (Fp) of the belt.

The second frequency range (T2) occurs for frequencies between 32 Hz and 50 Hz, in which there are minimum values of the peripheral force Fu in order to avoid deformation of the belt.

The third frequency range (T3) occurs for frequencies lower than or equal to 32 Hz, in which the belt begins to show deformation during its operation.

For the detection of faults in the electric motor (1) or in the motor drive, the energy content (Ed2) which corresponds to the second level of detail (d2) is taken into consideration, this energy content is calculated by the following formula:

Ed2 = 100 * (C2) <2> / Et (F5)

Where C2 are the coefficients extracted from the vector of the coefficients C which belong, specifically, to the level of detail d2 and the total energy (Et)? calculated by the formula (F1).

The W2 index of the energy content of the second level of detail? given by the following formula

W2 = Ed2 / Ea (F6)

Where Ed2? given by the formula F5 and Ea? given by the formula F3.

In this case the W2 index of the energy content of the second level of detail serves to identify a particular fault of an electrical nature which is equivalent to a short circuit in one of the stator phases. This failure can? depend on a problem on the stator phases themselves, or on a malfunction of the circuit that supplies the power (inverter).

When an electrical fault occurs in the motor or in the inverter, variations in terms of amplitude and frequency can be found in the electrical signal coming out of the motor. These variations recur in frequency ranges depending on the output signal managed by the inverter.

An inverter? an electronic input / output device capable of converting a direct current (DC bus) at the input into an alternating current at the output and to vary the parameters of amplitude and frequency.

The current signals absorbed by the motor have a fundamental frequency component, called the power frequency which depends on the modulating signal generated by the inverter. This power frequency? strictly correlated with the speed? of the rotor, for example in the case of permanent magnet synchronous motors, the following formula applies:

Where P? the number of pole pairs of the motor.

So for motors that work in transient regime and therefore with speed values? of the rotor (?) variable over time, the power frequency (?) not? a single well-defined frequency, but? varier? over time, cos? how the modulated control signal varies.

In the event of a fault on a motor phase (for example the first phase a), the inverter generates a modulated signal with a maximum period for example equal to T = 0.0005 sec., Equivalent to a modulated signal with a frequency equal to 2 kHz. In order to follow the same speed profile, and thus maintain performance, the inverter will have to? make up for the shortcomings due to the fault on the first phase, supplying more current in the two remaining phases (b and c).

With the proposed method, therefore, by analyzing the current in the second level of detail d2 which contains the frequency components present in the signal between 2 and 4 kHz (i.e. the frequency range in which the inverter works), there is a growth in terms of signal energy in this particular sub-band. This increase in terms of energy is monitored, through the W2 index corresponding to the second level of detail.

Below is a Table 4 which indicates some reference values W2R of the index, obtained on the motor current values acquired during the working cycles of the machine both in the nominal case and in the event of a fault.

Table 4

From this table you can? note the considerable increase of the W2 index in the event of a fault, this allows you to choose a threshold value W2T that allows you to establish when the equipment works in nominal conditions and when it works in fault conditions.

In this case the threshold value e could be chosen equal to W2T = 0.1500.

If the W2 index calculated continuously? less than the threshold value W2T, the machinery operates in nominal conditions and no alarm is activated.

If the W2 index calculated continuously? greater than or equal to the threshold value W2T, the machinery operates in fault conditions and the alarm (9) is activated.

Equivalent variations and modifications can be made to the present embodiments of the invention, within the reach of a person skilled in the art, which however fall within the scope of the invention expressed by the appended claims.

Claims (10)

CLAIMS 1. Method for diagnosing faults in machinery driven by an electric motor (1), the method comprising the following steps: - detection of a current signal (s (t)) absorbed by the electric motor (1) during its operation, - processing of the current signal (s (t)) with an envelope filtering, in order to obtain a filtered current signal (S (t)), - decomposition of the filtered current signal (S (t)) by means of a wavelet decomposition (6) in order to obtain a vector of coefficients (C) representing detailed energy levels (Edi) and approximation energy levels (Eai) , - processing of said coefficients of the coefficient vector (C) so as to obtain an index (Wx) relative to the x-th level of detail, - experimental calculation of a plurality? of reference values (WxR) of said index (Wx) in nominal conditions of the machinery and in conditions of failure of the machinery, - identification of at least one threshold value (WxT) between said reference values of the index (WxR) which defines a threshold between the operation of the machinery under nominal conditions and the operation of the machinery under fault conditions; - comparison of said index (Wx) calculated continuously during the operation of the apparatus with said at least one threshold value (WxT); - activation of an alarm (9) if the value of said index (Wx)? greater than or equal to said at least one threshold value (WxT). Method according to claim 1, wherein said index (Wx)? equal to the ratio between the energy content (Edx) of a level of detail (dx) chosen to identify the presence of a fault in the equipment and the sum of the energy content of all levels of approximation (Ea). Method according to any one of the preceding claims, wherein said current signal (s (t))? obtained by means of a current detector (4) arranged in the electric motor (1) or in an electric power supply system (3) of the electric motor. Method according to claim 3, wherein said electric motor (1)? a three-phase alternating current (AC) motor powered by an electric power supply system (3) comprising an inverter connected by three phases (U, V, W) to the windings of a stator (11) of the electric motor and said current detector ( 4)? arranged on one of the three phases (U, V, W) of the electric motor. Method according to any one of the preceding claims, wherein said apparatus comprises a transmission system (2) connected to a rotor (10) of the electric motor (1); dictates plurality? the reference values (WxR) of the index are experimentally calculated at certain vibration frequencies of the transmission system (2); and said index (Wx) corresponds to the index (W10) of the level of detail pi? high (d10) of the wavelet decomposition. 6. A method according to claim 5, wherein said plurality? of reference values (WxT) include reference values (W10T) calculated with the index of the level of detail pi? high (d10) at vibration frequencies of the transmission system (2) divided into three frequency ranges: - a first range of frequencies (T1) in which the transmission system operates in safe conditions; - a second range of frequencies (T2) in which the transmission system works in almost critical conditions; And - a third frequency range (T3) in which the transmission system works in citric conditions. 7. Method according to claims 6, wherein said first frequency range (T1) comprises frequencies greater than or equal to 50 Hz; said second frequency range (T2) includes frequencies between 32 Hz and 50 Hz, excluding the extremes of the range; said third frequency range (T3) includes frequencies lower than or equal to 32 Hz. Method according to any one of claims 1 to 4, wherein said index reference values (WxT) are calculated experimentally by selecting a frequency sub-band of the motor current signal which exhibits greater alterations in energy content in the event of a fault , compared to the case of nominal operation. 9. Failure diagnosis system of equipment driven by an electric motor, said system comprising: - a current detector (4) arranged in the electric motor (1) or in an electric power supply system (3) of the motor to detect a signal of current (s (t)) absorbed by the electric motor (1) during its operation , - a digital signal processor DSP (5) configured to carry out an envelope filtering of the current signal (s (t)) in order to obtain a filtered current signal (S (t)), - a wavelet decomposition (6 ) configured to decompose the filtered current signal (S (t)) in order to obtain a vector of coefficients (C) representing detailed energy levels (Edi) and approximation energy levels (Eai), - a? unit? processing (F) configured to process said coefficients of the coefficient vector (C) so as to obtain an index (Wx) relative to a xth level of detail, - a memory (7) in which? memorized at least one threshold value (WxT) identified among a plurality of of reference values (WxR) of said index (Wx) calculated experimentally in nominal conditions of the machinery and in conditions of failure of the machinery; wherein said threshold value (WxT) defines a threshold between an operation of the machinery in nominal conditions and an operation of the machinery in fault conditions; - a comparator (8) configured to compare said index (Wx) calculated continuously during the operation of the equipment with said at least one threshold value (WxT), - an alarm (9) configured to be activated if the value of said index (Wx)? greater than or equal to said at least one threshold value (WxT). 10. A system according to claim 9, wherein said DSP comprises a Hilbert envelope filter.