JP4367784B2 - Short-circuit diagnosis system for motor stator windings - Google Patents

Description

  The present invention relates to a motor stator winding short-circuit diagnosis system that diagnoses whether or not a motor stator winding is short-circuited.

  Conventionally, a diagnosis system (see Patent Document 1) for diagnosing whether or not an abnormality such as a short circuit has occurred in a winding of an electric motor has been proposed. According to this conventional diagnostic system, by performing pattern recognition on the waveform of the power source current flowing from the power source to the motor to be diagnosed, whether or not the stator winding of the motor has caused an abnormality such as a short circuit. Is to diagnose.

However, in general, the power supply voltage supplied to the electric motor from a commercial power supply varies. That is, the power supply voltage of a commercial power supply fluctuates in the time axis (frequency axis) direction and the amplitude axis direction. As a result, the current waveform flowing through the electric motor also varies in the time axis direction and the amplitude direction. When diagnosing the presence or absence of a short circuit in the stator winding of the motor using the conventional diagnostic system described above, it is necessary to perform pattern recognition on the waveform of the power supply current flowing through the motor, but it varies in the time axis direction and the amplitude direction. Even if the current waveform to be recognized is recognized, there is a possibility that the same waveform may be obtained when normal and abnormal (short circuit), and it may not be possible to accurately diagnose the presence or absence of a short circuit in the stator winding of the motor. There is a problem.
JP 2000-292465 A

  Therefore, the present invention provides a motor stator winding short-circuit diagnosis system that can accurately diagnose the presence or absence of a short circuit in the stator winding of the motor even if the power supply voltage supplied from the power source to the motor fluctuates. This is a problem to be solved.

The above problem can be solved by the short circuit diagnosis system for the motor stator winding described in the claims.
According to the short-circuit diagnosis system for the motor stator winding according to claim 1, the voltage supplied from the power source to the three-phase motor to be diagnosed by the detection unit and the current flowing from the power source to the stator winding of the motor The determination unit detects that the value of the current after a predetermined time starting from the time when the voltage detected by the detection unit becomes zero indicates that the stator winding is normal or the inductance component decreases. Then, it is possible to determine that the stator winding is short-circuited when a preset threshold value is exceeded in consideration of the short-circuit state in which the current phase advances.

  Here, the determination unit will be described. In order to make this explanation concrete, the motor to be diagnosed is a three-phase motor. FIG. 10 shows a current waveform for one cycle of the current flowing from the power source to the motor when the stator winding of the motor is normal, and FIG. 11 is short-circuited to the W phase of the stator winding of the motor. The waveform of the current corresponding to one cycle of the current flowing from the power source to the W phase of the motor is shown. 10 and 11 show waveforms for one cycle with the waveform starting point when the W-phase current value becomes zero. According to the current waveforms in FIG. 10 and FIG. 11, the time when the current value becomes zero after the waveform start point and the time when the maximum value and the minimum value are taken are almost the same, and the current value at each time is also almost the same. Since they match, there is no significant change in both waveforms. In addition, when there is no fluctuation in the power supply voltage, it has been confirmed by experiments that the maximum value of the current waveform is increased by short-circuiting the stator winding, and thus the waveform is the same at normal time and abnormal time. Is considered to be due to fluctuations in the power supply voltage.

  Next, the relationship between the voltage and current phases detected by the detection unit will be examined. FIG. 3 shows a W-phase current waveform Iw and a line voltage waveform Vuv between the U-V phases when the motor stator winding is normal. FIG. 4 shows a W-phase current waveform Iw and a line voltage waveform Vuv between the U and V phases when a short circuit has occurred in the W-phase winding of the stator winding for one turn. Further, FIG. 5 shows a W-phase current waveform Iw and a line voltage waveform Vuv between the U-V phases when a short-circuit has occurred in the W-phase winding of the stator winding for two turns.

  As apparent from FIGS. 3, 4, and 5, when a short circuit occurs in the stator winding, it can be confirmed that the phase of the current tends to advance with respect to the voltage. This phenomenon can be explained by the fact that when a short circuit occurs in the stator winding, the inductance component of the stator winding is reduced as compared with normal operation. Such a phase change appears remarkably even when the power supply voltage fluctuates. Therefore, by paying attention to this phenomenon and considering the phase of the voltage supplied from the power supply to the motor and the current flowing through the motor, the stator of the motor can be used even when the power supply voltage supplied from the power supply varies. It is possible to find the difference between the case where the winding is normal and the case where a short circuit of several turns has occurred.

  6 and 7 are the same as those in FIGS. 3 and 4 for one cycle, with the time when the line voltage waveform Vuv between the U and V phases changes from positive to negative as the start time of the W-phase current waveform Iw. The result when taking out is shown. According to FIG. 6 and FIG. 7, the current value at time zero of both waveforms has already differed by several hundred mA. Further, when attention is paid to the time when the current value switches from negative to positive, in FIG. 6, the positive and negative values are switched in the vicinity of 0.008 sec. In FIG. 7, however, the waveform is shifted at 0.075 sec and the waveform is shifted. Can be confirmed. Further, when attention is paid to the time when the current value becomes zero again, it is around 0.0165 sec in FIG. 6 and around 0.016 sec in FIG. As described above, it is difficult to clarify the difference between FIGS. 10 and 11, but the difference between the two can be further clarified by paying attention to the phase of current and voltage.

  As is clear from the above description, the voltage supplied from the power source to the motor and the current flowing through the motor are compared with the case where the current waveform is determined only from the shape of the current waveform as in the conventional method. By considering the phase, it is possible to clarify the difference between the case where the stator winding of the electric motor is normal and the case where a short circuit occurs. As a result, the determination accuracy of the short-circuit diagnosis of the stator winding is improved. Moreover, these current waveforms can reduce the influence of noise by taking an average of many measurements, and the difference in the current waveforms can be made clearer.

  Here, the hidden Markov model will be described. Hidden Markov models do not diagnose using a single point of time, but diagnose the short-circuit state of the motor stator winding from the overall shape (waveform) of the feature pattern. And a strong diagnostic system can be realized. Moreover, even with feature quantities including noise, it is possible to diagnose and determine the short-circuit state of the motor stator winding by probabilistically expressing the pattern of the feature quantities. Can be improved.

  FIG. 8 is a conceptual diagram of a hidden Markov model. In FIG. 8, a11, a12, a22, a23, a33, a34 indicate state transition probabilities, and aij is the probability of transition from state Si to state Sj. Further, bi (x) indicates the probability of outputting the feature quantity x in the state Si. S1 indicates an initial state, and S4 indicates a final state. Thus, the hidden Markov model has the state transition probability, the output probability, and the initial state probability πi (probability that the initial state is Si) as parameters, and each parameter is recognized as the normal state of the motor stator winding and the recognition. This is stored for each short-circuit state desired, and corresponds to the function of the learning unit 7 constituting the determination unit 5 of FIG. 2 described in the section of the embodiment of the invention described later.

  In the description of FIG. 9, the signal input at the time of constructing the neural network is a waveform of a feature value, but the peak value, maximum value, minimum value, correlation value, kurtosis, variance, standard deviation, It may be a feature amount such as a distortion value, a waveform rate, or an average value. In the case of using these feature amounts, a neural network is constructed by putting in the output layer values that are correlated with the input feature amounts and the state of the motor stator winding. A neural network constructed in this way may be used for the determination unit.

  According to the present invention, even if the power supply voltage supplied from the power source to the motor fluctuates, there is an effect that it is possible to accurately diagnose whether or not the stator winding of the motor is short-circuited.

Next, an embodiment of the present invention will be described.
FIG. 1 shows a motor stator winding that accurately diagnoses the presence or absence of a short circuit in the stator winding of the motor M even if the power supply voltage supplied from the power source P to the three-phase induction motor M to be diagnosed fluctuates. 1 is a system diagram showing the configuration of a line short-circuit diagnosis system 1.

  As shown in FIG. 1, a current detector 2 that detects a power source current flowing from the power source P to the motor M and a voltage detection wire 3 that detects a power source voltage supplied from the power source P to the motor M are provided. The current detection signal output from the current detector 2 and the voltage detection signal via the voltage detection wire 3 are input to the detection unit 4. In general, since the current detection signal and the voltage detection signal are analog signals, the analog current detection signal and the voltage detection signal are converted into digital signals by an A / D conversion circuit provided in the detection unit 4. And output to the determination unit 5.

  Based on the digital signal output from the detection unit 4, the determination unit 5 uses the current in consideration of the phase of the power supply voltage and the power supply current as described above (described in the columns 0008 to 0010) as a feature amount. Diagnose and determine whether the stator winding is short-circuited. Note that, as described above (described in the column 0011), the determination unit 5 includes, for example, a peak value, an effective value, an average value, a correlation value, a kurtosis, a variance, a standard deviation, a distortion value, and a waveform rate. Whether or not the stator winding of the electric motor M is short-circuited depending on whether or not the current value at a certain time is obtained, threshold values are provided for those values, and the value exceeds a preset threshold value. In the case of short-circuiting, the degree of short-circuiting may be diagnosed and determined. In the current waveform, the time when the current value is used as a reference, such as the time when the current value becomes zero, the time when the current value becomes the minimum value, or the like, may be used as the feature amount. Further, a combination of these values may be used as the feature amount. Or you may use the secondary quantity calculated | required based on the above values as a feature-value. Furthermore, a power supply voltage having a relative relationship with respect to the power supply current may be used as the feature amount.

  The determination unit 5 diagnoses whether the stator winding of the electric motor M is short-circuited as described above, and displays the determined result on the display unit 6.

  FIG. 2 shows that even if the power supply voltage supplied from the power supply P to the three-phase induction motor M to be diagnosed fluctuates, the motor M using the hidden Markov model described above (described in columns 0012 to 0014) is used. 1 is a system diagram showing the configuration of a motor stator winding short-circuit diagnosis system 1 for diagnosing the presence or absence of a stator winding short circuit; FIG.

  As shown in FIG. 2, a current detector 2 that detects a power source current flowing from the power source P to the motor M and a voltage detection wire 3 that detects a power source voltage supplied from the power source P to the motor M are provided. The current detection signal output from the current detector 2 and the voltage detection signal via the voltage detection wire 3 are input to the detection unit 4. As in FIG. 1, the detection unit 4 is provided with an A / D conversion circuit, and the current detection signal and the voltage detection signal are converted into digital signals and output to the determination unit 5.

  The determination unit 5 includes a learning unit 7 and a diagnosis unit 8. The learning unit 7 inputs feature amounts obtained from the current detection signal and voltage detection signal in a plurality of predetermined states (normal state and short-circuit state) of the motor stator winding, and each feature amount Are parameterized based on the hidden Markov model, and each parameter is stored as diagnosis determination information.

  The diagnosis unit 8 calculates the probability (likelihood) that the motor M generates a waveform of the feature amount detected in the operating state based on each parameter stored in the learning unit 7 in advance, and the generation probability thereof Is determined, and the state of the motor stator winding is diagnosed based on diagnosis determination information corresponding to the parameter. If the state is a short circuit state, the degree is specified.

  When the learning unit 7 parametrizes the waveform of each feature amount based on the hidden Markov model, the learning unit 7 performs calculation by applying a calculation algorithm including a forward algorithm and a backward algorithm. Further, in the diagnosis unit 8, when calculating the probability (likelihood) of generating the feature amount waveform using the parameters stored in the learning unit 7, a calculation algorithm including a forward algorithm is applied. calculate.

  The display unit 6 displays the state of the stator winding of the electric motor M diagnosed by the diagnosis unit 8 and is a display unit of a computer that constitutes the learning unit 7 and the diagnosis unit 8. The state of the stator winding diagnosed by the diagnosis unit 8 is displayed on the display unit of the computer and can be transmitted to, for example, a central monitoring device of a factory or a building by communication means.

  Note that when the determination unit 5 applies the hidden Markov model as described above to diagnose and determine the state of the motor stator winding, the current considering the phase is used as the feature amount. The state of the motor stator winding may be diagnosed and determined by applying a hidden Markov model to the waveform with the spectrum of. Furthermore, the state of the motor stator winding may be diagnosed and determined by applying a hidden Markov model to the time-varying waveform of the frequency, that is, the shape of the frequency-time waveform. Conversely, the state of the motor stator winding may be diagnosed and determined by applying a hidden Markov model to the waveform of the relative power supply voltage with reference to the power supply current. Furthermore, a secondary quantity calculated based on these values is used as a feature quantity, and a hidden Markov model is applied to the waveform of the feature quantity to diagnose and judge the state of the motor stator winding. May be.

  Further, when the state of the motor stator winding is diagnosed and determined by applying the neural network NN shown in FIG. 9 to the determination unit 5 in FIG. 1, as described above, the waveform of the characteristic amount (for example, current waveform) Is input to the input layer of the neural network. Further, a value representing the state of the motor stator winding when the waveform is input to the input layer is input to the output layer. In this way, first, as a learning stage, a neural network is constructed using the waveform of the feature amount in the normal state and short circuit state of the motor stator winding. Next, in the diagnosis determination stage, the waveform obtained by measurement is sampled at a time interval determined at the time of learning, and these values are input to the input layer of the neural network constructed in the learning stage. Then, the current state of the motor stator winding is determined from the value output from the output layer.

  Signals input when constructing a neural network are the peak value, maximum value, minimum value, correlation value, kurtosis, variance, standard deviation, distortion value, waveform rate, average value, etc. obtained from the waveform of the feature value. Also good. In the case of using these feature amounts, a neural network is constructed by putting in the output layer values that are correlated with the input feature amounts and the state of the motor stator winding. A neural network constructed in this way may be used for the determination unit. Conversely, the state of the motor stator winding may be diagnosed and determined by applying a neural network to the waveform of the relative power supply voltage with reference to the power supply current.

  In the above description, the main feature amount is the voltage that is supplied from the power source to the motor to be diagnosed and the phase of the current flowing through the motor. However, the current and the voltage are multiplied. Electric power may be used as a feature amount, and whether or not the stator winding of the motor is short-circuited may be diagnosed and determined based on the waveform or value of the feature amount.

The system system figure which showed the structure of the short circuit diagnostic system of an electric motor stator winding. The system diagram which showed the structure of the short circuit diagnosis system of the motor stator winding | coil using a hidden Markov model. It is a wave form diagram which considered the phase of the power supply voltage waveform and the current waveform which flows into an electric motor. It is a wave form diagram which considered the phase of the power supply voltage waveform and the current waveform which flows into an electric motor. It is a wave form diagram which considered the phase of the power supply voltage waveform and the current waveform which flows into an electric motor. FIG. 4 is a waveform diagram showing current waveforms in FIG. 3. FIG. 5 is a waveform diagram showing current waveforms in FIG. 4. It is a conceptual diagram of a hidden Markov model. It is a conceptual diagram of a neural network. It is a wave form diagram in case the phase of the current waveform which flows into an electric motor is not considered. It is a wave form diagram in case the phase of the current waveform which flows into an electric motor is not considered.

DESCRIPTION OF SYMBOLS 1 Short circuit diagnosis system of motor stator winding 2 Current detector 3 Voltage detection wire 4 Detection unit 5 Judgment unit 6 Display unit 7 Learning unit 8 Diagnosis unit

Claims (1)

A detection unit for detecting a voltage supplied from a power source to a three-phase AC motor to be diagnosed and a current flowing from the power source to a stator winding of the motor, and the voltage detected by the detection unit is zero The value of the current after a predetermined time with the start time as the start time is set in advance in consideration of whether the stator winding is normal or a short-circuit state in which the phase of the current advances because the inductance component decreases. An electric motor stator winding short-circuit diagnosis system, comprising: a determination unit that determines that the stator winding is short-circuited when a threshold value is exceeded.

Images