System and method for diagnosing open failure of switch

Abstract

Disclosed is a system for diagnosing an open failure of a switch by means of an artificial neural network model. In the system for diagnosing an open failure of a switch, data input to an input layer of the artificial neural network model includes a DC component and a root mean square value of a three-phase current.

Description

Switch open fault diagnosis system and method

[0001] The present invention relates to a system and method for diagnosing an open fault of a switch.

[0002] When an open-circuit fault occurs in a switch, the current flowing through the inverter and converter becomes distorted, and the distortion varies in form and pattern depending on the switch that failed. To diagnose an open-circuit fault, it is necessary to classify the current distortion pattern according to the faulty switch. However, manual classification can lead to errors.

[0003] Prior art document, Patent Publication No. 10-2021-0142883, discloses a method for diagnosing a multi-switch open-circuit fault of a three-phase PWM converter using artificial intelligence, including the steps of obtaining a direct current component (IV_dc) of a dq-axis current in a stationary coordinate system and estimating the total harmonic distortion (THD) (THDv) of the current, applying the direct current component (IV_dc) of the dq-axis current to a pre-trained first artificial neural network (ANN) to group open-circuit fault types according to the current vector angle into a plurality of sectors, and applying the direct current component (IV_dc) of the dq-axis current and the total harmonic distortion (THDv) of the current to a pre-trained second artificial neural network to diagnose a switch in which an open-circuit fault has occurred within the grouped plurality of sectors.

[0004] However, the method of using two artificial neural networks and using the DC component and total harmonic distortion of the current as input values ​​has the disadvantage of requiring a high-spec computer and a long learning period.

[0005] <Prior Art Literature>

[0006] Patent Document

[0007] Publication Patent No. 10-2021-0142883

[0008] The present invention aims to provide a system and method for diagnosing an open fault of a switch using an artificial neural network.

[0009] According to one aspect of the present invention for achieving the above object, a switch open fault diagnosis system is provided in which data input to an input layer of the artificial neural network model includes a DC component and an RMS value of a three-phase current in a system for diagnosing an open fault of a switch using an artificial neural network model.

[0010] The above switch open fault diagnosis system can diagnose an open fault of a switch during regenerative operation of a motor.

[0011] The number of neurons in the input layer above is 6, and the number of neurons in the output layer of the artificial neural network model above can be 'the number of switches for which an open fault is to be diagnosed + 1'.

[0012]

[0013] The above formula can be used to calculate the DC component of the above three-phase current.

[0014]

[0015] The above formula can be used to calculate the effective value of the three-phase current.

[0016] According to another aspect of the present invention for achieving the above object, a method for diagnosing an open fault of a switch by an artificial neural network model is provided, comprising a step of processing and preprocessing data to be input into an input layer of the artificial neural network model, wherein the data input into the input layer of the artificial neural network model includes a DC component and an RMS value of a three-phase current.

[0017] The above switch open fault diagnosis method may further include a step of measuring and storing the operation data of the motor to generate data to be input to the input layer, and the operation data may be measured during regenerative operation of the motor.

[0018] The above switch open fault diagnosis method may further include a step of designing the artificial neural network model, wherein the number of neurons in the input layer may be 6, and the number of neurons in the output layer of the artificial neural network model may be 'the number of switches for which an open fault is to be diagnosed + 1'.

[0019]

[0020] The above formula can be used to calculate the DC component of the above three-phase current.

[0021]

[0022] The above formula can be used to calculate the effective value of the three-phase current.

[0023] According to the present invention, an artificial neural network model can be trained even with a relatively low-spec computer, and the training time can also be reduced.

[0024] According to the present invention, calculations can be further reduced by calculating the DC component and root mean square (RMS) value by applying a simplified formula.

[0025] According to the present invention, an open fault of a switch is diagnosed during regenerative operation of a motor, and an artificial neural network learns and distinguishes the pattern of current according to each faulty switch, thereby making it possible to diagnose whether all switches are faulty.

[0026] According to the present invention, since the input values ​​of the input layer are sufficiently characteristic to distinguish between various failure modes that are the output values ​​of the output layer, there is an advantage in that the depth (number of layers) of the hidden layer and / or the number of neurons in the hidden layer can be effectively reduced.

[0027] Figure 1 is a block diagram schematically showing an example of a motor drive circuit used in an elevator.

[0028] Figure 2 illustrates an example of a fault current pattern of a switch included in a converter section when the motor is motorized.

[0029] Figure 3 illustrates an example of a fault current pattern of a switch included in a converter section during regenerative operation of a motor.

[0030] Figure 4 illustrates the structure of an artificial neural network model according to one embodiment of the present invention.

[0031] Figure 5 is a schematic flowchart of a switch open fault diagnosis method according to one embodiment of the present invention.

[0032] The configuration and operation of a preferred embodiment of the present invention are described in detail with reference to the attached drawings. The present invention can be applied in various fields such as switch fault diagnosis, elevator fault diagnosis, and artificial neural network fields. Furthermore, the following embodiments can be modified in various other forms, and the scope of the present invention is not limited to the following embodiments.

[0033]

[0034] Figure 1 is a block diagram schematically showing an example of a motor drive circuit used in an elevator.

[0035] Referring to FIG. 1, a motor driving circuit used in an elevator may include one or more of a grid section (110), a filter section (120), a converter section (130), a connection section (140), an inverter section (150), and a motor section (160).

[0036] The grid section (110) supplies power to the circuit and may include an AC power source. The grid section (110) may include a three-phase AC power source.

[0037] The filter unit (120) may play a role in allowing the converter unit (130) to smoothly adjust the power factor of the power supplied from the grid unit (110). The filter unit (120) may include one or more inductors. The filter unit (120) may include three inductors connected in series with each power source of the three-phase AC power source.

[0038] The converter unit (130) can control the active power and reactive power of the power supplied from the filter unit (120). The converter unit (130) can include a plurality of switches. The switches can be power semiconductor switches and IGBTs (Insulated Gate Bipolar Transistors).

[0039] The converter unit (130) may include a plurality of switches. The converter according to the present embodiment is a 3-level PWM converter and may include 12 switches (Ta1, Ta2, Ta3, Ta4, Tb1, Tb2, Tb3, Tb4, Tc1, Tc2, Tc3, Tc4).

[0040] Ta1 and Ta4 are connected in series, and a point between the two switches can be designated as a pole. Tb1 and Tb4, Tc1 and Tc4 are connected in the same way. Each of (Ta1, Ta4), (Tb1, Tb4), and (Tc1, Tc4) can be connected in parallel with each other. Ta2 and Ta3 are connected in series, and are connected between the pole between Ta1 and Ta4 and the Y point of the connection. Tb2 and Tb3, and Tc2 and Tc3 can also be connected in the same way.

[0041] When the grid unit (110) supplies three-phase AC power, the pole between Ta1 and Ta4 can be connected to the first power source (Va), the pole between Tb1 and Tb4 can be connected to the second power source (Vb), and the pole between Tc1 and Tc4 can be connected to the third power source (Vc).

[0042] Ta1, Tb1, and Tc1 may each be connected in parallel to the first node (X) of the connection portion (140). (Ta2 + Ta3), (Tb2 + Tb3), and (Tc2 + Tc3) may each be connected in parallel to the second node (Y) of the connection portion (140). Ta4, Tb4, and Tc4 may each be connected in parallel to the third node (Z) of the connection portion (140).

[0043] This arrangement of switches can reduce harmonics of voltage and current and increase efficiency.

[0044] The converter unit (130) can be operated so that the power factor becomes 1 by adjusting the power sent to the connection unit (140) (high-efficiency operation).

[0045] The connecting unit (140) connects the converter unit (130) and the inverter unit (150). Direct current can be converted into alternating current or alternating current can be converted into direct current by the connecting unit (140). The connecting unit (140) can include a plurality of capacitors (DC-Link) (for example, two) connected in series or in parallel. The three-phase alternating current power supplied from the grid unit (110) and passed through the filter unit (120) and the converter unit (130) can be stored as direct current power by the connecting unit (140).

[0046] The first node (X) connected in parallel with Ta1, Tb1, and Tc1 of the converter section (130) is connected to the first capacitor (V DC1 ) can be connected to the + terminal of the converter unit (130). The second node (Y) connected in parallel to (Ta2 + Ta3), (Tb2 + Tb3), (Tc2 + Tc3) of the converter unit (130) is connected to the first capacitor (V DC1 ) - terminal and second capacitor (V DC2 ) can be connected to the + terminal of the converter unit (130). The third node (Z) connected in parallel with Ta4, Tb4, and Tc4 of the converter unit (130) is connected to the second capacitor (V DC2 ) can be connected to the - terminal.

[0047] The inverter unit (150) can convert the current passing through the connection unit (140) into a current suitable for driving the motor (M). The inverter unit (150) can change the frequency and size of the current converted into direct current by the connection unit (140) and convert it into alternating current required for driving the motor (M).

[0048] The inverter unit (150) may include a plurality of switches. The switches may be power semiconductor switches and may be IGBTs (Insulated Gate Bipolar Transistors).

[0049] The inverter unit (150) may include 12 switches (Ta1, Ta2, Ta3, Ta4, Tb1, Tb2, Tb3, Tb4, Tc1, Tc2, Tc3, Tc4).

[0050] Ta1 and Ta4 are connected in series, and a point between the two switches can be designated as a pole. Tb1 and Tb4, Tc1 and Tc4 are connected in the same way. Each of (Ta1, Ta4), (Tb1, Tb4), and (Tc1, Tc4) can be connected in parallel with each other. Ta2 and Ta3 are connected in series, and are connected between the pole between Ta1 and Ta4 and the Y point of the connection. Tb2 and Tb3, and Tc2 and Tc3 can also be connected in the same way.

[0051] Ta1, Tb1, and Tc1 may each be connected in parallel with the first node (X) of the connection portion (140). (Ta3 + Ta2), (Tb3 + Tb2), and (Tc3 + Tc2) may each be connected in parallel with the second node (Y) of the connection portion (140). Ta4, Tb4, and Tc4 may each be connected in parallel with the third node (Z) of the connection portion (140).

[0052] This arrangement of switches can reduce harmonics of voltage and current and increase efficiency.

[0053] The motor unit (160) includes a motor (M). The motor (M) can be driven by receiving current converted to be suitable for driving the motor (M) by the inverter unit (150).

[0054]

[0055] An open-loop fault in a switch restricts the flow of current through the faulty switch, causing distortion of the current flowing through the circuit. As a result, the efficiency of the converter unit (130) may be reduced due to a decrease in power factor, and the control performance of the motor (M) in the motor unit (160) may be degraded.

[0056] If an open fault occurs in a switch included in the inverter unit (150), vibration and noise may occur in the motor (M). Therefore, it is possible to externally confirm that an open fault has occurred in one or more of the switches included in the inverter unit (150).

[0057] However, if an open fault occurs in a switch included in the converter unit (130), it is difficult to externally confirm the fault, and thus operation may continue with the fault unchecked. Consequently, operation efficiency may be reduced, and secondary faults such as burnout and short-circuiting may occur due to heat generation. Therefore, a technology capable of diagnosing an open fault in a switch included in the converter unit (130), which is difficult to externally confirm, is required.

[0058]

[0059] Fig. 2 illustrates an example of a fault current pattern of a switch included in a converter unit (130) during motoring of a motor (M), and Fig. 3 illustrates an example of a fault current pattern of a switch included in a converter unit (130) during regenerative operation of a motor (M). (a) of Fig. 2 and Fig. 3 represents a fault current pattern of Ta1, (b) represents Ta2, (c) represents Ta3, and (d) represents Ta4.

[0060] Referring to Fig. 2, it can be seen that when the motor (M) is motored, current distortion occurs in Ta2 and Ta3 of the converter section (130), but current distortion does not occur in Ta1 and Ta4 of the converter section (130).

[0061] Referring to Fig. 1, when the motor (M) is motored, Ta2, Ta3, Tb3, Tb3, Tc2, Tc3 of the converter unit (130) are power semiconductors ( ) and diode( ) are driven using both, but Ta1, Tb1, Tc1, Ta4, Tb4, Tc4 of the converter unit (130) are driven only through diodes, so even if an open circuit fault occurs in Ta1, Tb1, Tc1, Ta4, Tb4, Tc4 of the converter unit (130), normal operation is possible and no current distortion occurs. Therefore, only an open circuit fault in Ta2, Ta3, Tb2, Tb3, Tc2, Tc3 of the converter unit (130) can be diagnosed by the motoring of the motor (M), and an open circuit fault in Ta1, Tb1, Tc1, Ta4, Tb4, Tc4 of the converter unit (130) cannot be diagnosed.

[0062] On the other hand, referring to Fig. 3, it can be seen that when the motor (M) is in regenerative operation, current distortion occurs in all of Ta1, Ta2, Ta3, and Ta4 of the converter section (130).

[0063] Referring to Fig. 1, when the motor (M) is in regenerative operation, all switches (Ta1, Tb1, Tc1, Ta2, Tb2, Tc2, Ta3, Tb3, Tc3, Ta4, Tb4, Tc4) of the converter unit (130) are operated using both power semiconductors and diodes, so that when an open-circuit fault occurs in all switches, current distortion occurs. Therefore, in order to diagnose an open-circuit fault in all switches through current distortion, an open-circuit fault in the switches can be diagnosed during the regenerative operation of the motor (M).

[0064]

[0065] Figure 4 illustrates the structure of an artificial neural network model according to one embodiment of the present invention.

[0066] Referring to FIGS. 1 and 4, the artificial neural network model applied to the present invention may include an input layer (I), a hidden layer (H), and an output layer (O).

[0067] The input layer (I) is a layer into which data to be learned in an artificial neural network model is input, and may include one or more neurons. When three-phase current is supplied from the grid section (110), the number of neurons in the input layer (I) may be six (three direct current components of the three-phase current + three root mean square (RMS) values ​​of the three-phase current).

[0068] The hidden layer (H) processes data input to the input layer (I) and generates output data, and may include one or more neurons. The number of neurons in the hidden layer (H) can be optimally determined through repeated learning, taking into account factors such as computational time and accuracy.

[0069] The output layer (O) is a layer that outputs output data generated in the hidden layer (H) and may include one or more neurons. When the diagnosis target is a switch included in the converter unit (130), the number of neurons in the output layer (O) may be (the number of switches included in the converter unit (130) + 1) (each failure mode of the switch + when all switches are in a normal state).

[0070] Since the current distortion pattern varies depending on the switch in which a failure occurs, there are as many failure modes as the number of switches. If the converter unit (130) has 12 switches, there are 12 failure modes in the converter unit (130). Therefore, the number of neurons in the output layer (O) becomes 13. The design of the number of neurons in the input layer (I), the output layer (O), and the hidden layer (H) will be described later.

[0071]

[0072] Figure 5 is a schematic flowchart of a switch open fault diagnosis method according to one embodiment of the present invention.

[0073] Referring to FIG. 5, the method for diagnosing a switch open fault of the present invention may include one or more of a step of measuring and storing operating data (S110), a step of processing and preprocessing data (S120), a step of designing an artificial neural network model (S130), a step of learning data by the artificial neural network model (S140), a step of designing an algorithm for the artificial neural network model (S150), a step of determining whether the artificial neural network model is sufficiently optimized (S160), and a step of diagnosing an open fault of a switch by applying the artificial neural network model (S170).

[0074]

[0075] 1) Step of measuring and storing driving data (S110)

[0076] In order to generate data to be input into the input layer (I) of an artificial neural network, the operating data of a system (e.g., an elevator) to which a motor (M) and a switch are applied can be measured and stored. The operating data may include a three-phase AC current value supplied by a three-phase AC power source of a grid unit (110).

[0077]

[0078] 2) Data processing and preprocessing step (S120)

[0079] The stored driving data can be processed and preprocessed to be suitable for input into the input layer (I) of an artificial neural network. The data processing and preprocessing step may include calculating the DC component and root mean square (RMS) value of the three-phase AC current value.

[0080] The following formula can be used to calculate the DC component of a three-phase current.

[0081]

[0082] When diagnosing open-circuit faults in switches using an artificial neural network model, the magnitude of the values ​​derived from the formula may be more important than the magnitude of the values ​​themselves, depending on the fault condition. Omitting the division operation does not change the magnitude relationship, and since division is a significant computational burden, the DC component of the three-phase current can be calculated more simply by excluding the division operation, as shown below.

[0083]

[0084] The present invention has the advantage of reducing the computational burden on a computer by using a formula excluding division operations to calculate the DC component of a three-phase current, thereby enabling training of an artificial neural network model even with a relatively low-spec computer, and reducing the training time.

[0085] The following formula can be used to calculate the root mean square (RMS) value of three-phase current.

[0086]

[0087] Likewise, the magnitude relationship does not change even if division and square root operations are omitted, and since division and square root operations are a large burden on the computer, the root mean square (RMS) value of the three-phase current can be calculated more simply by excluding division and square root operations as shown below.

[0088]

[0089] The present invention has the advantage of reducing the computational burden on a computer by using a formula excluding division and square root operations to calculate the root mean square (RMS) value of a three-phase current, thereby enabling training of an artificial neural network model even with a relatively low-spec computer, and reducing the training time.

[0090]

[0091] 3) Step for designing an artificial neural network model (S130)

[0092] An artificial neural network model can be designed by determining the type of input data and the number of neurons in the input layer (I), the type of output data and the number of neurons in the output layer (O), etc.

[0093] The number of neurons in the input layer (I) can be determined by the number of inputs used for diagnosis, and the number of neurons in the output layer (O) can be determined by the number of faults to be diagnosed.

[0094] When a three-phase current is supplied from the grid section (110), the number of neurons in the input layer (I) may be 6 (3 DC components of the three-phase current + 3 root-mean-square (RMS) values ​​of the three-phase current). When the diagnosis target is a switch included in the converter section (130), the number of neurons in the output layer (O) may be (the number of switches included in the converter section (130) + 1) (each failure mode of the switch + when all switches are in a normal state).

[0095]

[0096] 4) Step in which the artificial neural network model learns data (S140)

[0097] An artificial neural network model can learn input data entered into the input layer (I) and generate output data.

[0098] Learning can be viewed as the process of finding the weights (W) that exist between layers. These weights (W) act as synapses between layers and determine the propagation ratio of values ​​output from the activation function. Weights (W) can be optimized to minimize the error between the artificial neural network model's predicted values ​​and the correct answer.

[0099] To optimize the weight (W) values, the backpropagation algorithm can be used. The error can be calculated using the cross-entropy error function.

[0100] The number of layers and the number of neurons in the hidden layer (H) can be initially set to approximate values ​​(empirical values). Afterwards, the weights are obtained through learning, and the diagnostic performance (error, cross-entropy error) is checked. If the diagnostic performance deteriorates, the diagnostic performance can be improved by increasing the number of neurons in the hidden layer (H). Conversely, if the diagnostic performance is excellent when trained with the initial number of hidden layers (H) and the number of neurons, the performance can be maintained while reducing the number of hidden layers and the number of neurons to reduce the computational burden. Through this iterative learning, the optimal number of layers, number of neurons, and weights in the hidden layer (H) that can satisfy both the diagnostic performance and the computational burden can be determined. The results of the iterative learning can be used to determine the optimal number of neurons in the hidden layer (H).

[0101] According to the present invention, since the input value (DC component, root mean square value) of the input layer (I) is sufficiently characteristic to distinguish between various failure modes (each failure mode of the switch + the case where all switches are in a normal state) which are the output values ​​of the output layer (O), there is an advantage in that the depth (number of layers) of the hidden layer (H) and / or the number of neurons in the hidden layer (H) can be effectively reduced. The depth of the hidden layer (H) of the present invention can be 1 layer, and the number of neurons in the hidden layer can be 5, which can be seen as a considerably simpler configuration compared to a general artificial neural network.

[0102]

[0103] 5) Step of designing the algorithm of the artificial neural network model (S150)

[0104] An artificial neural network model can be applied to design algorithms necessary for diagnosing open-circuit faults in switch systems. For example, C-based program code can be used to design the algorithm.

[0105]

[0106] 6) Step for determining whether the artificial neural network model is sufficiently optimized (S160), and step for diagnosing an open fault of the switch by applying the artificial neural network model (S170).

[0107] It is possible to determine whether the artificial neural network model is sufficiently optimized to diagnose a switch open fault. If the artificial neural network model is determined to be sufficiently optimized ('Yes' in S160), the artificial neural network model can be applied to diagnose a switch open fault (S170). If the artificial neural network model is determined to be insufficiently optimized ('No' in S160), steps 1) to 5) can be repeated to optimize the artificial neural network model.

[0108]

[0109] The computer specifications and training time required for an artificial neural network model vary significantly depending on the type of input data and number of neurons in the input layer (I), as well as the type and number of output data. For practical application, realistic computer specifications and training times are required. A well-designed artificial neural network model that can produce more efficient and appropriate output data even with a low-spec computer and a shorter training time is considered a better design.

[0110] According to the present invention, when three-phase AC power is supplied from the grid (110), by using the DC component and root mean square (RMS) value as input data of the artificial neural network model, the number of hidden layers (H) of the artificial neural network model and the number of neurons in the input layer (I) and the hidden layer (H) can be drastically reduced. By reducing the number of hidden layers (H) of the artificial neural network model and the number of neurons in the input layer (I) and the hidden layer (H), the artificial neural network model can be trained even with a relatively low-spec computer, and the training time can also be reduced.

[0111] According to the present invention, calculations can be further reduced by calculating the DC component and root mean square (RMS) value by applying a simplified formula.

[0112] According to the present invention, an open fault of a switch is diagnosed not during motoring of a motor (M) but during regenerative operation, and an artificial neural network learns and distinguishes the pattern of current according to each faulty switch on its own, so that it is possible to diagnose whether all switches are faulty.

[0113]

[0114] It is obvious to those skilled in the art that the present invention is not limited to the above embodiments, and that various modifications or variations can be made without departing from the technical spirit of the present invention.

[0115] <Explanation of symbols>

[0116] 110: Grid section 120: Filter section

[0117] 130: Converter section 140: Connection section

[0118] 150: Inverter section 160: Motor section

[0119] I: Input layer H: Hidden layer

[0120] O: Output layer

Claims

1. In a system for diagnosing open faults of switches using an artificial neural network model, A switch open fault diagnosis system, wherein data input to the input layer of the artificial neural network model includes the DC component and root mean square value of the three-phase current.

2. In claim 1, A switch open fault diagnosis system that diagnoses switch open fault during regenerative operation of the motor.

3. In claim 1 or claim 2, The number of neurons in the above input layer is 6, The number of neurons in the output layer of the above artificial neural network model is 'the number of switches for which open faults are to be diagnosed + 1', a switch open fault diagnosis system.

4. In claim 1 or claim 2, A switch open fault diagnosis system using the above formula to calculate the DC component of the above three-phase current.

5. In claim 1 or claim 2, A switch open fault diagnosis system that uses the above formula to calculate the effective value of the above three-phase current.

6. A method for diagnosing an open fault of a switch using an artificial neural network model, It includes a step of processing and preprocessing data to be input to the input layer of the artificial neural network model, A method for diagnosing a switch open fault, wherein the data input to the input layer of the artificial neural network model includes the DC component and the root mean square value of the three-phase current.

7. In claim 6, Further comprising a step of measuring and storing the driving data of the motor to generate data to be input to the above input layer, The above driving data is a method for diagnosing a switch open failure, which is measured during regenerative driving of the motor.

8. In claim 6 or claim 7, Further comprising a step of designing the artificial neural network model, The number of neurons in the above input layer is 6, The number of neurons in the output layer of the above artificial neural network model is 'the number of switches for which open faults are to be diagnosed + 1', a method for diagnosing open faults in a switch.

9. In claim 6 or claim 7, A method for diagnosing a switch open fault, using the above formula to calculate the DC component of the above three-phase current.

10. In claim 6 or claim 7, A method for diagnosing a switch open fault, using the above formula to calculate the effective value of the above three-phase current.

Images