Simulation device and program

Abstract

A simulation device is provided with: a model storage unit that stores a motor physical model obtained by modeling a winding circuit unit and a rotational motion equation unit, and an abnormal state model obtained by modeling an abnormal state of a motor; and a model calculation unit configured to perform a calculation process using the motor physical model, and to set, as an input to the motor physical model, an abnormal parameter indicating a deviation amount with respect to a normal state calculated by the abnormal state model.

Description

[0001] This application is based on Japanese Patent Application No. 2023-137209 (2023.8.25) and Japanese Patent Application No. 2024-107513. Technical Field

[0002] The present disclosure relates to simulation devices. Background Art

[0003] In the past, regarding factory equipment maintenance in the field of production machinery, the application of AI (artificial intelligence) for condition management maintenance (Condition Based Maintenance) of mechanical systems has continued to develop (for example, International Publication No. 2019 / 035279). Summary of the invention

[0004] Here, when AI is introduced into motor abnormality detection, it would be very useful if the abnormality detection effect of AI could be confirmed without using an actual machine (real device) such as a motor.

[0005] In view of the above situation, an object of the present disclosure is to provide a simulation device capable of effectively realizing a structure for virtually generating an abnormal state of a motor.

[0006] A simulation device according to one aspect of the present disclosure is configured to include:

[0007] The model storage unit stores the following models:

[0008] A physical model of the motor, which is modeled by a winding circuit portion and a rotational motion equation portion; and

[0009] an abnormal state model obtained by modeling an abnormal state of the motor; and

[0010] a model calculation unit configured to perform calculation processing using the motor physical model;

[0011] An abnormality parameter representing a deviation from a normal state calculated by the abnormal state model is used as an input to the motor physical model. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a diagram showing the structure of a computer according to an exemplary embodiment of the present disclosure.

[0013] Figure 2 1 is a diagram showing a configuration of a simulation device according to an exemplary embodiment of the present disclosure.

[0014] Figure 3 The following are a side view (left side) and a front view (right side) showing a schematic configuration example of a motor.

[0015] Figure 4 It is a diagram showing the structure of a motor model.

[0016] Figure 5 It is a diagram showing the input-output relationship between the motion equation part and the winding circuit part.

[0017] Figure 6 This is a diagram showing an example of a plot of the back electromotive force constant with respect to the rotation speed.

[0018] Figure 7 This is a diagram showing an example of plotting the motor terminal current, the back electromotive force constant, and the motor torque against the rotational speed.

[0019] Figure 8 It is a schematic plan cross-sectional view showing a structural example of a brushed DC motor.

[0020] Fig. 9 Yes Figure 8 The structure shown is an expanded view along the circumferential direction.

[0021] Fig.10 It is a diagram showing model parameters in the motor physical model.

[0022] Fig.11 It is a diagram showing the gap between commutator segments.

[0023] Fig.12 This is a diagram for explaining the contact resistance between the commutator segments and the brushes.

[0024] Fig.13 It is a diagram showing the rotational movement of the sides of the winding.

[0025] Fig.14 This is a diagram showing an example of magnetic flux density distribution.

[0026] Fig.15 This is a diagram showing the positional deviation between the permanent magnet and the brush.

[0027] Fig.16 It is a diagram showing the length of the side of the winding.

[0028] Fig.17 It is a diagram showing the distance of the winding edge from the rotation axis.

[0029] Fig.18 This is a diagram showing the filter configuration near the motor terminals.

[0030] Fig.19 The diagrams are a schematic plan view, a perspective view, and a development view showing the rotational movement of the winding.

[0031] Fig. 20 It is a schematic development diagram showing the rotational movement of the winding.

[0032] Fig.21 This is a diagram showing the overall image of the winding circuit section.

[0033] Fig. 22 This is a table showing the method of calculating contact resistance.

[0034] Fig.23 This is a diagram showing the structure of a motor physical model when modeling is performed using Simscape (registered trademark) / Simulink (registered trademark).

[0035] Fig.24 This is a diagram showing a partial structure of a winding circuit model.

[0036] Fig.25 This is a diagram showing an example of simulation results.

[0037] Fig.26 This is a diagram showing an example of comparison between a simulation and an actual machine.

[0038] Fig. 27 This is a diagram showing the relationship between the motion equation part and the bearing poor lubrication model.

[0039] Fig.28 This is a diagram showing the relationship between the motion equation part and the bearing damage model.

[0040] Fig.29 Shows the front view and side sectional view of the bearing and shaft.

[0041] Fig.30 It is a diagram showing the parameters of the bearing.

[0042] Fig.31 It is a diagram for explaining the revolution angle and rotation angle of the rolling element.

[0043] Fig.32 This is a diagram showing an example of a vibration model of a support system.

[0044] Fig.33 It is a diagram showing a configuration example of a sensor model.

[0045] Fig.34 This is a diagram showing an example structure of a machine learning model.

[0046] Fig.35 It is a diagram showing the structure of a 3-layer neural network.

[0047] Fig.36 This is a diagram showing an example of a setting screen for the motor type and abnormal status.

[0048] Fig.37 This is a diagram showing an example of a motor basic setting screen.

[0049] Fig.38 This is a diagram showing an example of a support system setting screen.

[0050] Fig.39 It is a diagram showing an example of the first abnormal state setting screen.

[0051] Fig.40 It is a diagram showing an example of the second abnormal state setting screen.

[0052] Fig.41 It is a diagram showing an example of the time / degradation setting screen.

[0053] Fig.42 This is a diagram showing an example of a sensor setting screen.

[0054] Fig.43 This is a diagram showing an example of an AI setting screen.

[0055] Fig.44 It is a diagram showing an example of the abnormality determination setting screen.

[0056] Fig.45 This is a diagram showing an example of a simulation result screen. DETAILED DESCRIPTION

[0057] Hereinafter, exemplary embodiments of the present disclosure will be described with reference to the accompanying drawings.

[0058] <Computer Structure>

[0059] Figure 1 1 is a diagram showing the structure of a computer 100 according to an exemplary embodiment of the present disclosure. The computer 100 functions as a simulation device of the present disclosure described later. The computer 100 is, for example, a PC (personal computer). When the computer 100 is a PC, it may be a desktop type or a notebook type.

[0060] The computer 100 includes a CPU (Central Processing Unit) 100A, a memory 100B, an auxiliary storage device 100C, an operation input unit 100D, and a display unit 100E.

[0061] The CPU 100A includes a control device and a computing device (neither of which is shown in the figure). The control device analyzes program instructions and controls each unit of the computer 100. The computing device is a device that performs computing processing.

[0062] The memory 100B is a semiconductor storage device that temporarily stores programs or data. When the computer 100 is powered off, the information stored in the memory 100B is erased.

[0063] The auxiliary storage device 100C is composed of a HDD (Hard Disk Drive) or an SSD (Solid State Drive), etc., and stores programs and data. The program stored in the auxiliary storage device 100C is read into the memory 100B. The CPU 100A executes the program read into the memory 100B.

[0064] The operation input unit 100D is composed of a keyboard, a mouse, etc., and is a device for providing operation input to the computer 100. Information input from the operation input unit 100D is sent to the memory 100B.

[0065] The display unit 100E is constituted by, for example, a liquid crystal display, etc., and converts the information acquired from the memory 100B into an image and outputs the image.

[0066] <Structure of the Simulator>

[0067] Figure 2 1 is a diagram showing the structure of a simulation device 1 according to an exemplary embodiment of the present disclosure. The simulation device 1 is a device capable of simulating abnormality detection of machine learning (AI) with a motor system as an object. However, the physical system to which the simulation device of the present disclosure can be applied is not limited to the motor system.

[0068] The simulation device 1 includes a model storage unit 2, a model calculation unit 3, a model setting unit 4, a display control unit 5, an operation input unit 6, a display unit 7, and a data storage unit 8. The program P ( Figure 1 ) is a program for causing the computer 100 to function as the simulation device 1.

[0069] The model storage unit 2 stores the system model 21, the sensor model 22, the machine learning model 23, and the abnormality determination model 24, and is composed of the auxiliary storage device 100C of the computer 100. The system model 21, the sensor model 22, the machine learning model 23, and the abnormality determination model 24 are configured as a program P by, for example, MATLAB (registered trademark) / Simulink (registered trademark). In addition, the details of each model will be described later.

[0070] The functions of the model calculation unit 3 , the model setting unit 4 , and the display control unit 5 are implemented by the CPU 100A executing the program P. The operation input unit 6 and the display unit 7 correspond to the operation input unit 100D and the display unit 100E in the computer 100 , respectively.

[0071] The model calculation unit 3 performs simulation by performing calculation processing on each model stored in the model storage unit 2. The model setting unit 4 performs settings (parameter settings, model selection settings, etc.) related to each model stored in the model storage unit 2 based on the input from the operation input unit 6. The simulation of the model calculation unit 3 is performed according to the settings of the model setting unit 4. The display control unit 5 performs the following control: based on the input from the operation input unit 6, the display unit 7 displays the model setting screen described later.

[0072] In addition, the data storage unit 8 stores the state monitoring data DT owned by the user who uses the simulation device 1. In addition, the data storage unit 8 corresponds to the auxiliary storage device 100C in the computer 100. When the user studies the introduction of abnormality detection based on machine learning (AI) to a device owned by the user, for example, the state monitoring data DT obtained by monitoring the motor in the device is stored in the data storage unit 8. For example, the state monitoring data DT is stored in the data storage unit 8 by acquiring data from the outside of the simulation device 1 via a network or a USB (Universal Serial Bus).

[0073] When executing a simulation using the state monitoring data DT, a prescribed input is performed through the operation input unit 6. Then, the model operation unit 3 performs learning and inference of the machine learning model 23 by inputting the state monitoring data DT into the machine learning model 23. After learning using the normal data contained in the state monitoring data DT, inference can be performed using abnormal data. As described later, the machine learning model 23 outputs an abnormality degree, and the abnormality determination model 24 performs an abnormality determination based on the above-mentioned abnormality degree. In this way, the abnormality detection effect of machine learning can be confirmed using the state monitoring data DT held by the user, and a simulation that matches the device environment held by the user can be performed.

[0074] <System model>

[0075] Next, the system model 21 will be described. The system model 21 is a model that represents a physical model of the motor system. The system model 21 can virtually generate physical signal waveforms in normal and abnormal states of the motor system.

[0076] The system model 21 includes a motor model 211 , a driver model 212 , and a load model 213 .

[0077] The driver model 212 is a model of a driver for driving a motor. For example, when the motor is a brushed DC motor (hereinafter referred to as a BDC motor), the driver can be set to a circuit that uses a switch to apply a DC voltage to the motor, or an H-bridge circuit, etc. The H-bridge circuit is composed of two half bridges. The half bridge is composed of two switching elements connected in series between the application end of the DC voltage and the ground end (the application end of the ground potential). In addition, for example, when the motor is a brushless DC motor (hereinafter referred to as a BLDC motor), the driver can be set to a circuit composed of three half bridges corresponding to a three-phase motor, etc.

[0078] The driver in the driver model 212 can be set to be selectable by the model setting unit 4. For example, when the motors in the motor model 211 are the same BDC motors, a circuit using one switch or an H-bridge circuit as described above can be selected. In addition, when the motor in the motor model 211 is a BLDC motor, for example, a driver can be selected according to the number of phases of the motor.

[0079] The load model 213 is a model of an object driven by the motor in the motor model 211 . The driven object is, for example, a fan in a fan device or an arm in an industrial robot. The load model 213 provides the motor model 211 with information on external torque.

[0080] <Motor Model>

[0081] Here, the motor model 211 is described. The motor model 211 is a model obtained by physically modeling a motor (an example of a physical system). The above-mentioned motor is, for example, a BDC motor or a BLDC motor. In this embodiment, the motor type of the motor model 211 can be selected by the model setting unit 4. At this time, for example, multiple types can be selected for BDC motors and BLDC motors respectively. Different types in BDC motors refer to, for example, different numbers of pole pairs, different numbers of windings, different winding methods of the windings, etc. Different types in BLDC motors refer to, for example, different numbers of phases, different numbers of poles, different numbers of slots, etc.

[0082] Figure 3 1 is a side view (left) and a front view (right) showing a schematic structural example of a motor. Figure 3 The figure shows a common structure diagram that does not depend on the type of motor (such as BDC motor, BLDC motor). Figure 3 The orthogonal coordinate system shown is a stationary coordinate system fixed to the stage 201. The X-axis, Y-axis, and Z-axis are mutually orthogonal. The Z-axis extends in the direction in which the axis 20C extends and passes through the center of the axis 20C. The Y-axis extends perpendicularly to the plane of the stage 201. The X-axis extends parallel to the above-mentioned plane and extends in the horizontal direction. Figure 3 As an example, the origin O of the orthogonal coordinate system is included in the rotor 20B.

[0083] like Figure 3 As shown, the motor 20 is fixed on the table 201. The motor 20 has: a housing 20A, a rotor 20B, a shaft 20C, a stator 20D and a bearing 20E. The rotor 20B, the shaft 20C, the stator 20D and the bearing 20E are accommodated inside the housing 20A. The stator 20D is fixed to the housing 20A. The rotor 20B is arranged on the inner circumference of the stator 20D. The shaft 20C protrudes from the rotor 20B on both sides in the direction of the rotation axis (Z-axis direction). The shaft 20C is rotatably supported by bearings 20E on both sides in the direction of the rotation axis. The bearing 20E is fixed to the housing 20A. In addition, Figure 3 In the embodiment, the rotor 20B is an inner rotor type located on the inner circumference side of the stator 20D, but the rotor may be an outer rotor type located on the outer circumference side of the stator.

[0084] For example, when the motor 20 is a BDC motor, the stator 20D includes a magnet, and the rotor 20B includes an iron core, a winding, and a commutator. The brushes included in the BDC motor can contact the commutator. The current flows from the brushes to the windings via the commutator, and thus the rotor 20B rotates due to the action of the magnetic lines of force generated by the windings and the magnetic lines of force of the magnets.

[0085] For example, when the motor 20 is a BLDC motor, the stator 20D includes an iron core and a winding, and the rotor 20B includes a magnet. When current flows through the winding, the rotor 20B rotates.

[0086] The rotor 20B and the shaft 20C rotate about the rotation axis, thereby driving a load connected to the shaft 20C.

[0087] Figure 4 2 is a diagram showing the structure of the motor model 211. The motor model 211 includes a motor physical model 2111. The motor physical model 2111 includes a motion equation part 2111A and a winding circuit part 2111B.

[0088] Figure 5 2111A and the winding circuit section 2111B. in is input to the winding circuit unit 2111B, and the motor terminal current i is calculated and output. m In addition, as described below Fig. 9 As shown, the input voltage V in Applied between the motor positive terminal Tp and the motor negative terminal Tn, the motor terminal current i m is the current flowing through the motor terminals. Motor terminal current i m The equation of motion is input to the motion equation unit 2111A, and the mechanical angular velocity ω of the shaft 20C is calculated and output. m and mechanical angle θm . Mechanical angle angular velocity ω m and mechanical angle θ m It is fed back to the winding circuit section 2111B.

[0089] The motion equation unit 2111A has the following equation (1) as the motion equation.

[0090]

[0091] J m : Inertia of the rotating part of the motor 20 (rotor 20B and shaft 20C), T m : Motor torque, T ex : External torque.

[0092] Motor torque T m It is represented by the following formula (2).

[0093] T m =K t ·i m …(2)

[0094] K t : Torque constant.

[0095] In the motor, torque is generated by the effect of the magnetic flux distribution of the permanent magnet and the magnetic flux distribution of the winding current. The contribution of the magnetic flux distribution of the permanent magnet to the torque is determined by the shape and configuration of the magnetic poles and the geometric configuration relationship of the winding. It is independent of the speed and the value of the motor terminal current and is a constant gain contribution. Therefore, as shown in the above formula (2), the motor torque T m The torque constant K is the constant coefficient. t and motor terminal current i m The product of.

[0096] In addition, the torque constant K t and the back electromotive force constant K e With K t =K e Relationship between back EMF V bemf and mechanical angular velocity ω m With V bemf =K e ·ω m The back electromotive force is the voltage generated across the motor terminals of the modeling target motor when the shaft of the modeling target motor is connected to another motor and the shaft is rotated at a constant speed by the other motor. The back electromotive force V is measured by changing the rotation speed. bemf , calculate K e =V bemf / ω m As a result, Figure 6As shown in FIG. 1 , the back electromotive force constant is approximately constant regardless of the rotation speed. Therefore, the back electromotive force constant can be regarded as a constant, and the average value of the plotted back electromotive force constant is taken as the back electromotive force constant K. e This value can be set as the torque constant K t Thus, the back electromotive force constant K e is a constant, and therefore becomes the basis of the above formula (2).

[0097] Here, in the above equation of motion (1), the left side represents the inertia J m The product of the mechanical angle angular acceleration and the right side is the input voltage V in The motor torque T generated by applying the voltage to the motor terminals m 、The loss torque T that summarizes various losses loss and external torque T ex The resultant torque of the external torque T ex It corresponds to the torque output from the load model 213 and the torque output from a person or the environment.

[0098] Loss torque T loss Ingredients

[0099]

[0100] is the loss torque under normal conditions. In addition, B m0 , B m1 , B m2 Represents the loss coefficient.

[0101] As mentioned above, the loss torque is assumed to be the mechanical angular velocity ω m The assumption about the loss torque is that the motor torque T is constant when the speed is fixed and no external torque is applied in the normal state. m —Loss torque = 0, that is

[0102]

[0103] Established and obtained.

[0104] First, the rotation speed is changed, and the input voltage V is measured at each rotation speed. in The average value of the motor terminal current when the current changes. The measurement results are shown in Figure 7 Use the average value of the motor terminal current and the above Figure 6 The back EMF constant K shown e (=Torque constant K t ), calculate the motor torque T according to the above formula (2) m .like Figure 7 As shown, the motor torque T is obtained according to the speedm When the motor rotates at a constant speed, the torque T m Balanced with the loss torque, therefore, Figure 7 The motor torque T shown m The plot of can be regarded as the plot of loss torque. Figure 7 The plot shown in the figure is regressed, and the determination coefficient is good. Therefore, as mentioned above, the loss torque is set as the mechanical angle angular velocity ω m The quadratic form of .

[0105] <BDC Motor>

[0106] Here, before explaining the winding circuit portion 2111B, a BDC motor as an example of a modeling object of the winding circuit portion 2111B is explained in more detail. Figure 8 It is a schematic plan cross-sectional view showing a configuration example of the motor 20 when it is set as a BDC motor. Figure 8 1 is a diagram viewed in the direction of the rotation axis J. The rotation axis J coincides with the above-mentioned Z axis. In addition, hereinafter, the direction in which the rotation axis J extends is referred to as the axial direction, the direction around the rotation axis J is referred to as the circumferential direction, and the direction perpendicular to the rotation axis J is referred to as the radial direction.

[0107] The stator 20D includes permanent magnets Mg and brushes BR. Figure 8 In the structure, the permanent magnet Mg includes magnets MgS1, MgN1, MgS2, and MgN2. The magnets MgS1, MgN1, MgS2, and MgN2 are respectively arranged with S poles, N poles, S poles, and N poles in the radial inner side along the circumferential direction. That is, on the radial inner side, the magnetic poles are alternately arranged with S poles and N poles in the circumferential direction.

[0108] The brushes BR include a positive brush and a negative brush as described below. The polarities of the brushes are alternately arranged in the circumferential direction.

[0109] The rotor 20B includes an iron core 202, a winding WR, and a commutator segment CM. The iron core 202 is formed by stacking electromagnetic steel plates in the axial direction, for example. The iron core 202 is arranged radially inside the permanent magnet Mg. The iron core 202 has an annular portion 202A and teeth 202B. The annular portion 202A extends in the axial direction and is formed into a circular ring shape in the circumferential direction. The teeth 202B protrude radially outward from the outer peripheral surface of the annular portion 202A. A plurality of teeth 202B are arranged in the circumferential direction.

[0110] exist Figure 8 In the structure, the winding WR includes windings WR1 to WR16, that is, 16 windings. Figure 8In the diagram, only windings WR1 to WR4 are shown as representatives. Each winding WR is wound around a tooth 202B in a manner that passes through one circumferential side of a tooth 202B (shown as θ1) and from the tooth 202B to the other circumferential side of the second tooth 202B. The teeth 202B wound by each winding WR are staggered one by one in the circumferential direction. In this way, Figure 8 The structure is a concentrated winding structure.

[0111] The commutator segments CM are arranged radially inward of the core 202 and radially outward of the brushes BR. Figure 8 In the structure of the commutator segment CM, the commutator segments CM1 to CM16 (in Figure 8 The commutator segments CM1 to CM16 are arranged in a circular ring shape in the circumferential direction. One side and the other side of the lead wires drawn from each winding WR are connected to each of the commutator segments CM adjacent to each other in the circumferential direction. The commutator segments CM connected by each winding WR are staggered one by one in the circumferential direction.

[0112] The commutator segments CM can be in contact with the brushes BR. As the rotor 20B rotates, the commutator segments CM rotate, and which commutator segments CM are in contact with the brushes BR and the contact resistance thereof change with time.

[0113] Fig. 9 Yes Figure 8 The structure shown is an expanded view along the circumferential direction. Fig. 9 represents the rotation direction θ of the rotor 20B rt . Rotation direction θ rt is with Figure 8 The same direction as the circumferential side θ1 shown. Along the rotation direction θ rt The S pole and N pole are configured. In addition, along the rotation direction θ rt The commutator segments CM1 to CM16 are arranged. Specifically, the commutator segments CM16 → CM15 → . . . are arranged in the order of the rotation direction θ. rt When the commutator segment CM1 is arranged, it returns to the commutator segment CM16. That is, the commutator segment CM is arranged along the rotation direction θ rt Configured in a ring shape.

[0114] As described above, one and the other of the lead wires led out from the winding WR are connected to each of the commutator segments CM adjacent to each other in the circumferential direction. Fig. 9As shown, one side of the lead wire from winding WR1 is connected to commutator segment CM16, and the other side is connected to commutator segment CM1. One side of the lead wire from winding WR2 is connected to commutator segment CM1, and the other side is connected to commutator segment CM2. One side of the lead wire from winding WR3 is connected to commutator segment CM2, and the other side is connected to commutator segment CM3. One side of the lead wire from winding WR4 is connected to commutator segment CM3, and the other side is connected to commutator segment CM4. Similarly, after connecting to winding WR16, it is connected to commutator segment CM. In addition, Fig. 9 In FIG. 1 , only windings WR1 to WR4 are shown as representatives. Thus, the windings WR are connected in series via the commutator segments CM to form a loop.

[0115] The brush BR includes positive brushes BR_P1 and BR_P2 and negative brushes BR_N1 and BR_N2. The negative brush BR_N1, the positive brush BR_P1, the negative brush BR_N2, and the positive brush BR_P2 are sequentially connected along the rotation direction θ. rt Configuration.

[0116] As the rotor 20B rotates, the commutator segments CM1 to CM16 rotate in the direction θ. rt The commutator segments CM that are in contact with the positive brushes BR_P1, BR_P2 and the negative brushes BR_N1, BR_N2 are switched in sequence. Fig. 9 The following states are shown: commutator segments CM4 and CM3 are in contact with the negative brush BR_N1, commutator segments CM16 and CM15 are in contact with the positive brush BR_P1, commutator segments CM12 and CM11 are in contact with the negative brush BR_N2, and commutator segments CM8 and CM7 are in contact with the positive brush BR_P2. The winding WR rotates along the direction of rotation θ with the commutator segments CM. rt Moving, thereby, the magnetic flux that crosses the magnetic poles.

[0117] <About the winding circuit section>

[0118] The winding circuit unit 2111B is a model obtained by directly modeling the geometrical arrangement of the winding WR, the permanent magnet Mg, the brush BR, and the commutator segment CM in the motor 20 (BDC motor). Fig.10 represents the model parameters in the motor physical model 2111. Fig.10 The parameters in the winding circuit section 2111B shown in FIG. 21 are used to explain the details of the winding circuit section 2111B. Fig.10 2111A and the input variables and internal variables in the motor physical model 2111. ex It can be processed as a parameter or as an input variable.

[0119] The pole pair number p is the number of poles of the permanent magnet Mg. Fig. 9 ( Figure 8 ) structure, there are 2 pairs of N poles and S poles, so the number of pole pairs p = 2. The total number of windings Ncoil is the total number of windings WR. Fig. 9 In the structure, the total number of windings Ncoil=16.

[0120] Regarding the inductance L_1 of each winding WR and the resistance R_1 of each winding WR, Fig. 9 In the structure, for example, the value can be measured as two parallel circuits in which eight windings are connected in series at the midpoint of the 180-degree symmetry of the loop circuit of the winding WR, and the value of each winding can be calculated as an average value for setting.

[0121] There is actually a gap between adjacent commutator segments CM. Fig.11 As shown in FIG. 1 , the gap between the commutator segments CM is emphasized, and the commutator segment gap gap is set as the gap distance. The commutator segment gap gap is expressed by the ripple angle θ described later. r The commutator segment gap is set based on the observation results of the actual machine.

[0122] like Fig.12 As shown in the upper part of , the brush resistance R_B is the contact resistance value when the brush BR and the commutator segment CM are in contact in such a way that the width of the brush BR and the width of the commutator segment CM just overlap. Fig.12 As an example, the case where the brush BR is in contact with the commutator segment CM7 is shown. In this case, the contact resistance between the commutator segment CM and the brush BR is the smallest. The brush resistance R_B is adjusted to the motor terminal current i m The waveform (ripple waveform) is consistent in the actual machine and the model.

[0123] When the contact width between the brush BR and the commutator segment CM is very small, the limit contact resistance is infinite. Fig.12 The lower part of the figure shows, as an example, a state where the width of contact between the brush BR and the commutator segment CM7 is very small. However, it is considered difficult to perform calculations including infinity in simulation, and such a limit state will not occur in reality. Therefore, as the upper limit of the contact resistance, a saturation value Sat is set as a zero division countermeasure. The saturation value Sat is adjusted to the motor terminal current i m The waveform (ripple waveform) is consistent in the actual machine and the model.

[0124] Fig.131 is a top cross-sectional view schematically showing the movement of the side WR1_H around the rotation axis J and the magnetic flux density B of the permanent magnet MgN1 (the N pole on the radially inner side) when focusing on the side WR1_H of the winding WR1. The side WR1_H is the rotation direction θ of the winding WR1. rt The side extending in the axial direction at the end on the opposite side. Fig.13 As shown, the side WR1_H is located in the middle of the connecting permanent magnet MgN1 and in the direction of rotation θ. rt The state on the line segment between the opposite end and the rotation axis J is defined as a mechanical angle θ. m =0, according to the mechanical angle θ m , determines the magnetic flux density B of the magnetic flux passing through side WR1_H.

[0125] Therefore, if Fig.14 As shown, the magnetic flux density B is set to the mechanical angle θ m The function B(θ m ) to set the magnetic flux density distribution. At this time, set the maximum magnetic flux density B m As a parameter, within the range where the N pole is arranged radially inward, B = +B m , within the radial inner side of the S pole configuration range B = -B m In addition, if Fig.13 As shown in the figure, the polarity of the magnetic flux density B is positive in the direction toward the rotation axis J. In the range between the N pole and the S pole, the magnetic flux density B is set to +B m with -B m Varies between. Fig.14 In the setting, B is at +B m with -B m The magnetic flux density distribution can be linearly changed between the maximum magnetic flux density B and the maximum magnetic flux density B. m In addition, the relative positions of the windings WR2 to WR16 with respect to the winding WR1 are determined, so the magnetic flux density distribution with respect to the side corresponding to each side WR1_H is determined according to the mechanical angle θ. m to be sure.

[0126] Position offset S gap (unit: [rad]) is a parameter indicating the positional relationship between the magnetic poles of the permanent magnet Mg and the brush BR. Fig.15 In the example shown in the figure, the brush BR is defined as a reference state (with respect to the reference state) in the positional relationship between the magnetic poles of the permanent magnet Mg and the brush BR shown from above. Fig.14 The position shift S when the position shifts to the state shown below gap As the position offset S gap The processing in the model can be based on the position offset S gapThis is because even if the mechanical angle θ is the same, m state, that is, the same position state of the commutator segment CM relative to the brush BR, if a position deviation S gap , the magnetic flux density relative to the winding WR will also change. Position offset S gap Set based on actual machine observation results.

[0127] like Fig.16 As shown, the side length l of the winding is the length of the winding WR in the direction of rotation θ rt The length of the side extending in the axial direction at the end of the winding WR1 is a parameter. In addition, in the case of winding WR1, the length l of the side of the winding is equivalent to the length of the side WR1_H mentioned above. The length l is set based on the observation results of the actual machine.

[0128] The distance r from the rotation axis to the side of the winding is a parameter representing the radial distance from the rotation axis J to the side WR_H. Fig.17 (and Fig.13 In the same figure, as an example, the distance r from the side WR1_H of the winding WR1 is shown. The distance r is set based on the observation results of the actual machine.

[0129] Fig.18 This is a diagram showing the circuit structure around the motor terminal (in Fig. 9 etc.). A capacitor C0 is connected between the motor positive terminal Tp and the motor negative terminal Tn. An inductor L0 is connected to the motor positive terminal Tp and the motor negative terminal Tn, respectively. The motor positive terminal Tp is connected to one end of the inductor L0 on one side, and the positive brushes BR_P1 and BR_P2 are connected to the other end. The motor negative terminal Tn is connected to one end of the inductor L0 on the other side, and the negative brushes BR_N1 and BR_N2 are connected to the other end. An LC filter is formed by the capacitor C0 and the inductor L0. The LC filter is used to suppress the pulse-shaped noise generated when the commutator segment CM and the brush BR are switched. The capacitance value of the capacitor C0 and the inductance value of the inductor L0 are set as parameters, respectively.

[0130] Next, the induced electromotive force generated in the winding WR is described. The windings WR1 to WR16 respectively cross the magnetic flux of the permanent magnet Mg, thereby generating induced electromotive force in the windings WR1 to WR16 respectively. As described above, the magnetic flux density distribution B is set to the mechanical angle θ m Using the set magnetic flux density distribution, the induced electromotive force is calculated from the time change of the magnetic flux interlinked with the windings WR1 to WR16 respectively.

[0131] Fig.19The schematic plan view, perspective view, and development view show a state where one winding WR moves due to the rotation of the rotor 20B. The winding WR moves from the state represented by ABCD to the state represented by A'B'C'D'.

[0132] Here, refer to Fig. 20 The front side CD is shown in the unfolded diagram relative to the rotation direction θ. rt The scanned area DCC'D' is the area increased by the rotational movement of the winding WR and is expressed as a positive area. On the other hand, the rear side BA is rt The scanned area ABB'A' is the area that decreases due to the rotation of the winding WR and is expressed as a negative area. In addition, the area A'B'CD is the area that does not change before and after the rotation, and the magnetic flux passing through the area does not change either. Therefore, the change in magnetic flux is not affected by the change in magnetic flux. No contribution. According to the above, the flux changes It is obtained by the product of the area with a sign and the magnetic flux density passing through the area. Therefore, the induced electromotive force e coil，No. It is expressed by the following formula.

[0133]

[0134] In addition, coil，No. It is the induced electromotive force generated by the winding represented by No. (winding number) (for example, WR1 if No.=1).

[0135] In addition, using the length of the side of the winding l and the distance r from the rotation axis to the side of the winding, the above positive area = l·rω is calculated m , negative area = -l·rω m .

[0136] Fig.21 2111B is a diagram showing the overall image of the winding circuit section 2111B. coil，No. Each of the windings WR1 to WR16 is modeled so as to output from a voltage source inserted in series with the inductor L_1 and the resistor R_1 .

[0137] In addition, Fig.21 The ripple angle θ shown r If the brush BR of interest is referred to as a specified brush ( Fig.21 The positive brush BR_P1 in the figure is the positive brush BR_P1), and the commutator segment CM of interest is referred to as the specified commutator segment ( Fig.21 Where CM2 is the commutator segment, the contact resistance Rc between the specified brush and the specified commutator segment is calculated as follows.

[0138] If the rotation direction of the commutator segment is specified, θ rtThe commutator segment CM adjacent to the front commutator segment ( Fig.21 In the figure, the commutator segment CM1 is set to the rotation direction θ of the specified commutator segment. rt The commutator segment CM on the opposite side is called the rear commutator segment ( Fig.21 In the example, the commutator segment CM3 is defined as the ripple angle θ, and the state where the rear end of the specified brush is consistent with the rear end of the front commutator segment (the state where the width of the specified brush is exactly consistent with the width of the front commutator segment) is defined as the ripple angle θ. r = 0, the distance from the rear end of the specified brush to the rear end of the front commutator segment is expressed as the ripple angle θ r [rad] When the rear end of the specified brush is aligned with the rear end of the specified commutator segment (when the width of the specified brush is exactly aligned with the width of the specified commutator segment), the ripple angle θ r =2π, when the rear end of the specified brush is consistent with the rear end of the rear commutator segment (the width of the specified brush is exactly consistent with the width of the rear commutator segment), the ripple angle θ r =4π. Ripple angle θ r According to the mechanical angle θ m To convert.

[0139] According to the ripple angle θ r , you can Fig. 22 The contact resistance Rc between the specified brush and the specified commutator segment is calculated as shown in the table. Fig. 22 R B Equivalent to the brush resistance R_B. In addition, Fig. 22 The ripple angle θ shown r Outside the range of , the contact resistance value is ∞. However, as described above, when the calculated contact resistance Rc exceeds the saturation value Sat, the contact resistance Rc=Sat.

[0140] Fig.23 1 is a diagram showing the structure of a motor physical model 211 when modeling is performed using Simscape (registered trademark) / Simulink (registered trademark).

[0141] The motion equation unit 2111A receives the motor terminal current i output from the winding circuit unit 2111B. m , output mechanical angle θ m and the mechanical angular velocity ω m And fed back to the winding circuit section 2111B.

[0142] The winding circuit section 2111B includes an induced electromotive force generating section 2A, a winding circuit model section 2B, a ripple angle converting section 2C, a contact resistance value generating section 2D, and a switching signal generating section 2E.

[0143] The induced electromotive force generating unit 2A generates an induced electromotive force according to the mechanical angle θ. m , generating the induced electromotive force of each winding WR. Here, for an example of the winding circuit model part 2B, Fig.24 A part of the structure is shown. Here, the number of pole pairs p=2 is considered, and actually half of the total 16 windings WR, that is, 8 windings WR are modeled. Alternatively, the total number of windings WR may be used for modeling.

[0144] Therefore, if Fig.24 As shown, the winding WR is modeled by WR1 to WR8, and the brush BR is modeled by a pair of a positive brush and a negative brush. The induced electromotive force e generated by the induced electromotive force generating unit 2A is output from the voltage source E inserted in the winding WR of the corresponding winding number. coil，No. . Fig.24 It shows the induced electromotive force e output from the voltage source E8 in the winding WR8 coil，8 .

[0145] In addition, if Fig.24 As shown, the winding circuit model part 2B is modeled in such a way that switches SW1, SW2 and variable resistors VR1, VR2 are set according to the winding WR. Specifically, in the winding WR, the inductor L_1 and the resistor R_1 are connected in series, but one end of the resistor R_1 (the opposite side to the side connected to the inductor L_1) is connected to one end of the variable resistor VR1, and the switch SW1 is connected between the other end of the variable resistor VR1 and the positive line LP. The positive line LP is connected to the positive brush. In addition, one end of the resistor R_1 is also connected to one end of the variable resistor VR2, and the switch SW2 is connected between the other end of the variable resistor VR2 and the negative line LN. The negative line LN is connected to the negative brush.

[0146] When the commutator segment CM connected to the lead wire of the winding WR contacts the positive brush, the switch SW1 is in the on state, and the resistance value of the variable resistor VR1 is set to the contact resistance value. When the commutator segment CM connected to the lead wire of the winding WR contacts the negative brush, the switch SW2 is in the on state, and the resistance value of the variable resistor VR2 is set to the contact resistance value. In addition, when the commutator segment CM is not in contact with the positive brush or the negative brush, the switch SW1 or SW2 is in the off state. In addition, sometimes both switches SW1 and SW2 are in the off state.

[0147] Fig.23 The switching signal generating unit 2E shown in FIG. m The on / off of the switches SW1 and SW2 is determined for the winding WR, and a switching signal is generated. According to the generated switching signal, the switches SW1 and SW2 are turned on or off. In addition, the ripple angle conversion unit 2C converts the mechanical angle θ m Converted to ripple angle θ rThe contact resistance value generating unit 2D is based on the ripple angle θ r The contact resistance value between the commutator segment CM and the brush BR is generated. The generated contact resistance value is set as the resistance value of the variable resistors VR1 and VR2.

[0148] The winding circuit model unit 2B calculates the input voltage V under the condition that the induced electromotive force of the voltage source E, the on / off states of the switches SW1 and SW2, and the resistance values ​​of the variable resistors VR1 and VR2 are determined. in The motor terminal current i m Furthermore, when the pole pair number is considered to be 2 and the model is constructed by 8 windings WR as described above, the motor terminal current i m Therefore, the calculated motor terminal current i m The signal is input to an amplifier with a gain of 2 and output to the equation of motion section 2111A.

[0149] <Simulation results>

[0150] Fig.25 : is a diagram showing an example of the result of a simulation performed in the present disclosure. in The motor terminal current i m Generates ripple waveforms in the same way as the actual machine. Fig.26 It means that the input voltage V in The figure shows an example of comparing the relationship between the actual machine and the simulation. In this way, the relationship between the mechanical angular speed and the input voltage V is reproduced by simulation. in As long as it is the simulation disclosed in the present invention, the amount of calculation can be greatly reduced and the simulation time can be shortened.

[0151] <Abnormal status model>

[0152] like Figure 4 As shown, the motor model 211 has an abnormal state model 211A. The abnormal state model 211A has a bearing poor lubrication model 2112 and a bearing damage model 2113. The reason for modeling the abnormality of the bearing 20E as an abnormal state is as follows. The bearing 20E is a mechanical element that supports the shaft 20C and is independent of other mechanical elements and loads inside the motor. That is, the abnormal state of the bearing 20E can be handled independently of other abnormal states. In addition, regardless of the type of motor and the type of load, the bearing E is a set element, and the abnormal state model of the bearing can be flexibly used without relying on the type of motor and the type of load.

[0153] Bearing abnormalities can be roughly divided into two types: poor lubrication and damage. Therefore, a bearing poor lubrication model 2112 and a bearing damage model 2113 are modeled. The bearing 20E is mechanically connected to the shaft 20C via friction and vertical resistance. Therefore, the abnormal state model 211A is modeled as a model that outputs the friction torque as the difference (deviation) between the normal state and the abnormal state and the vertical resistance caused by the abnormality.

[0154] <Bearing lubrication failure model>

[0155] Fig.29 2 shows the front view and side sectional view of the bearing 20E and the shaft 20C. The bearing 20E has an outer ring OR and an inner ring IR. The shaft 20C is fixed inside the inner ring IR. A lubricant LUB is provided between the outer ring OR and the inner ring IR. As a result, the inner ring IR and the shaft 20C rotate relative to the stationary outer ring OR. In addition, Fig.29 This is an example in which the bearing is a rolling bearing, but the bearing may be a sliding bearing.

[0156] The poor lubrication of the bearing 20E depends on whether the poor lubrication is overall or local, or the viscosity of the lubricant or the contribution of the fluid. In view of the above, if the poor lubrication is expressed by a mathematical formula, the mode that depends on the mechanical angle and rotation speed of the rotating shaft and the mode that depends only on the rotation speed are considered to be the superposition of these modes. The mode that depends on the mechanical angle and rotation speed of the rotating shaft is a mode related to the mechanical deviation or change in one rotation of the shaft. For example, it is related to the surface condition, dimensional deviation, offset relative to the perfect circle, fitting condition, foreign matter attachment site, damaged site, radial load, etc. The mode that depends only on the rotation speed is related to the dynamic viscosity of the fluid (air, lubricating oil, grease, etc.) when sliding on the surface. Therefore, the friction torque (deviation from the normal state) caused by poor lubrication is expressed by the following formula (3).

[0157]

[0158] B lub_cof1 and B lub_cof2 is a constant coefficient, B lub_index1 and B lub_index2 is the exponent of the speed. m ) is the mechanical angle θ m According to the switch SW_B lub_cof2 The value of (an integer from 1 to 7), f(θ m ) is switched as shown in the following formula. In the following formula, theta_m = θ m .

[0159] 1: f(theta_m)=1-|sin(0.5*theta_m)|

[0160] 2: f(theta_m)={1-|sin(0.5*theta_m)|}^2

[0161] 3: f(theta_m)={1-|sin(0.5*theta_m)|}^4

[0162] 4: f(theta_m)=|sin(0.5*theta_m)|

[0163] 5: f(theta_m)={|sin(0.5*theta_m)|}^2

[0164] 6: f(theta_m) = sin(theta_m)

[0165] 7: f(theta_m) = 0

[0166] like Fig. 27 As shown, the mechanical angle θ output from the motion equation unit 2111A is m and the mechanical angular velocity ω m It is input into the above formula (3) to calculate the friction torque T lubrication The calculated friction torque T lubrication The loss torque component ΔT of the abnormal state in the above equation (1) is input to the motion equation unit 2111A. The progress of the abnormal state (ie, degradation) is expressed by increasing and decreasing the parameters in the above equation (3) as a time function.

[0167] In addition, in the bearing lubrication failure model 2112, the vertical resistance caused by the abnormality is calculated by the following formula (4).

[0168]

[0169] μ lub is the friction coefficient, and R is the radius of the shaft 20C.

[0170] That is, the vertical resistance is calculated using the friction torque calculated by the above formula (3). The vertical resistance calculated in this way is input as the excitation force to the vibration model of the support system described later.

[0171] <Bearing damage model>

[0172] like Fig.29 As shown in FIG. 2 , in the bearing 20E, a rolling element RE is provided between the outer ring OR and the inner ring IR. A plurality of rolling elements RE are provided in the circumferential direction. Fig.29 In the figure, the lubricant LUB and the rolling element RE are shown separately for convenience.

[0173] Bearing damage is classified into three types according to which mechanical element of the bearing is damaged. Specifically, there are three types of damage: outer ring damage, inner ring damage, and rolling element damage.

[0174] The rolling element RE rotates and moves in the circumferential direction by revolution. When the outer ring is damaged, an impact force (hereinafter referred to as impact pulse) is generated each time the rolling element RE slides relative to the damaged part. The impact pulse acts on the shaft 20C as a vertical resistance and friction torque. Therefore, when the outer ring is damaged, each time the rolling element RE rotates at an angle θ revolution When the following formula (A) is satisfied, a vertical resistance represented by the following formula (5) is generated.

[0175]

[0176] N bearing_daeage =P hight … (5)

[0177] Z: number of rolling elements, P hight : Height of the shock pulse, P width : Width of the impulse pulse, n: integer starting from 0 (n = 0, 1, 2...)

[0178] In addition, when the inner ring is damaged, an impact pulse is generated each time the rolling element RE slides relative to the damaged part. revolution When the following formula (B) is satisfied, the vertical resistance represented by the above formula (5) is generated.

[0179]

[0180] In addition, when the rolling element is damaged, an impact pulse is generated every time the rolling element RE rotates and causes the outer ring or inner ring to slide relative to the damaged part. rotation When the following formula (C) is satisfied, the vertical resistance represented by the above formula (5) is generated.

[0181] πn≤θ rotation ≤πn+P width …(C)

[0182] According to the vertical resistance N caused by the damage mentioned above bearing_damage The friction torque caused by damage is expressed as follows (6).

[0183] T bearing_damage =μ bearing ·N bearing_damage ·R…(6)

[0184] μ bearing: Friction coefficient, R: Radius of shaft 20C

[0185] Revolution angle θ revolution and the rotation angle θ rotation According to the mechanical angle θ m It is calculated as follows.

[0186] θ inner =θ m

[0187]

[0188] Among them, θ inner Indicates the rotation angle of the inner ring, D pitch Indicates the diameter of the pitch circle of the bearing, D rolling represents the diameter of the rolling element and α represents the contact angle. Fig.30 In addition, Fig.30 It also represents the radius r of the inner orbit i and the radius r of the outer orbit o .

[0189] Here, refer to Fig.31 The derivation of the above formula will be described. Fig.31 Indicates that the inner circle IR rotates by θ inner (=θ m ) when the rolling element RE orbital angle θ revolution and the rotation angle θ rotation . Points Ar and Br represent the contact points with the outer ring OR and inner ring IR on the rolling element RE side. Points Ar' and Br' represent the points after points Ar and Br move due to the revolution of the rolling element RE. At this time, the contact points on the outer ring OR side and the inner ring IR side change to Ao→Ao”, Bi→Bi”, and the contact point Bi moves to Bi'.

[0190] The length of Ar'Ao" = the length of AoAo", therefore, the following formula (D) holds.

[0191] r r θ rotation =r o θ revolution (D)

[0192] On the other hand, the length of Br'Bi"=the length of Bi'Bi", and therefore, the following formula (E) holds.

[0193] r r θ rotation =r i (θ m -θ revolution )(E)

[0194] According to the above formula (D) (E),

[0195] r o θ revolution =r i θ m -r i θ revolution ,

[0196] θ revolution =(r i / (r o +r i ))θ m .

[0197] in,

[0198] r i =(D pitch -D rolling ·cosα) / 2

[0199] r o =(D pitch +D rolling ·cosα) / 2,

[0200] Therefore, θ revolution =1 / 2·(1-D rolling / D pitch ·cosα)·θ m .

[0201] On the other hand, according to the above formula (D),

[0202] θ rotation =r o / r r ·θ revolution .

[0203] Among them, the above θ revolution =1 / 2·(1-D rolling / D pitch ·cosα)·θ m 、r r =D rolling / 2,

[0204] Therefore, θ rotation =1 / 2·D pitch / D rolling ·(1-(D rolling / D pitch ·cosα) 2 )·θ m .

[0205] like Fig.28 As shown, the mechanical angle θ output from the motion equation unit 2111A is mis input to the bearing damage model 2113. The friction torque T calculated by the bearing damage model 2113 bearing_damage The loss torque component ΔT of the abnormal state in the above equation (1) is input to the motion equation unit 2111A. The progress of the abnormal state (ie, degradation) is measured by the height P of the impulse pulse. hight and width P width It is expressed as a function of time. In addition, the vertical resistance N bearing_damage This is input to the vibration model of the support system described later.

[0206] <Relationship between abnormal state model and motor physical model>

[0207] like Figure 4 As shown, there is signal exchange between the motor physical model 2111 (motion equation part) and the bearing poor lubrication model 2112, and between the motor physical model 2111 and the bearing damage model 2113. However, there is no signal exchange between the bearing poor lubrication model 2112 and the bearing damage model 2113. The background is that the transition from a normal state to an abnormal state first changes to one abnormal state according to various insights such as statistics, and then changes to a complex abnormal state along with the development of the abnormality. Since it is a model for studying the early detection of abnormal signs, it is sufficient to be able to express the initial stage of the transition to an abnormal state.

[0208] <Vibration model of support system>

[0209] like Figure 4 As shown, the motor model 211 has a vibration model 2114 of the support system. First, the support system is defined as all mechanical elements constituting the motor 20 including the housing 20A. Figure 3 ) is not a support system, but is stationary. The coordinate system for dealing with the vibration of the support system is Figure 3 The stationary coordinate system is as shown. The coordinate axes in the stationary coordinate system are determined as described above. The reason for determining the coordinate axes in this way is that the rigidity of the motor 20 is generally anisotropic in the direction perpendicular to the installation surface and in the horizontal direction.

[0210] The vibration of the support system is generated by applying the force (vibration force) caused by the abnormality to the support system. In the case of poor lubrication and damage of the bearing 20E mentioned above, the vertical resistance caused by the abnormality becomes the vibration force. In addition, when the bearing is abnormal, the vibration force is a vector in the XY plane.

[0211] The vibration of the support system is determined by the combination of the stiffness (spring constant) and damping of each mechanical element that constitutes the support system. The vibration model of the support system is as follows: Fig.32As shown, it is assumed that the system is composed of springs Kx, Ky, dampers Cx, Cy and a mass point of mass M. The entire motor 20 including the housing 20A and the load connected to the shaft 20C are all regarded as a mass point of mass M. On both sides of the above mass point in the X direction, the spring Kx and the damper Cx are connected in parallel. On both sides of the above mass point in the Y direction, the spring Ky and the damper Cy are connected in parallel.

[0212] The translational motion equation in such a support system is expressed by the following equation (7): The translational motion equation needs to be established for each vibration measurement location of the vibration sensor (position of the vibration sensor).

[0213] This is because the parameters M, k, and c may vary depending on the vibration measurement location.

[0214]

[0215] k x : Spring constant of spring Kx, k y : Spring constant of spring Ky, c x : Attenuation coefficient of damper Cx, c y : attenuation coefficient of damper Cy, F: external force applied to the mass point, θ f : Indicates the angle of the direction of the external force F relative to the X-axis

[0216] It is unrealistic to synthesize the rigidity and damping of each mechanical element constituting the support system by accumulating them. Therefore, the rigidity and damping of the support system as a whole can be estimated, and the estimated values ​​can be used as representative values ​​for parameter setting. In addition, the motion equation for each of the X-axis, Y-axis, and Z-axis can be established. In addition, for example, when measuring the vibration on an axis inclined at 45 degrees in the XY plane, the X-axis vibration and the Y-axis vibration can be synthesized.

[0217] As described above, the vertical resistance N output from the bearing poor lubrication model 2112 is lubrication (Formula (4) above), or the vertical resistance N output from the bearing damage model 2113 bearing_damage (Formula (5) above) is input as external force F to the vibration model 2114 of the support system. The motion equation (Formula (7) above) in the vibration model 2114 of the support system outputs the time series data of the displacement in the X direction and the displacement in the Y direction. In addition, by performing a first-order differentiation on the displacement data, the velocity data can be obtained, and by further performing a first-order differentiation, the acceleration data can be obtained.

[0218] In addition, there is no signal input from the vibration model 2114 of the support system to the abnormal state model 211A ( Figure 4 ). The reason is that since the applied force is the cause, the result is treated as vibration.

[0219] <Calculation of Motor Model>

[0220] When the simulation is executed, the motor model 211 is operated by the model operation unit 3. At this time, the motion equation unit 2111A and the abnormal state model 211A exchange signals, and the motor physical model 2111 (motion equation unit 2111A and winding circuit unit 2111B) is numerically calculated, and the motor terminal current i is output from the winding circuit unit 2111B. m On the other hand, the vibration model 2114 of the support system receives input from the abnormal state model 211A and performs numerical calculations to output time series data of X-direction displacement, Y-direction displacement, X-direction acceleration, and Y-direction acceleration.

[0221] Thus, the motor terminal current i is output from the motor model 211. m , and the time series data of the X-direction displacement, the Y-direction displacement, the X-direction acceleration, and the Y-direction acceleration are used as the physical signal waveform data. By setting the abnormal state model 211A to the abnormal state, the physical signal waveform in the abnormal state can be virtually generated.

[0222] Even if the signal waveform data of the abnormal state of the motor is generated, the simulation speed is very slow in a model based on the finite element method. In addition, in a model that uses experimental data as a reference and adds frequency component noise, the relationship between the physical signals is contradictory. Therefore, the motor model 211 of this embodiment can generate the physical signal waveform data of the abnormal state with the necessary accuracy and appropriate simulation speed. At this time, no contradiction occurs between the physical signals.

[0223] In addition, by setting the abnormal state model 211A to the normal state, it is also possible to virtually generate a physical signal waveform in the normal state. At this time, in the bearing lubrication failure model 2112, the coefficient B in the above formula (3) is lub_cof1 and B lub_cof2 are set to 0 respectively. In the bearing damage model 2113, the height P of the impact pulse in the above formula (5) is hight Set to 0.

[0224] <Sensor model>

[0225] Next, the sensor model 22 will be described. Fig.332 is a diagram showing a configuration example of the sensor model 22. The sensor model 22 includes a transfer function model 221 and an AD conversion model 222. In addition, the sensor model 22 includes a transfer function model 221 and an AD conversion model 222 according to the type of sensor. The types of sensors are, for example, current sensors and vibration sensors. Thus, the type of sensor can be selected as described later.

[0226] The transfer function model 221 receives an input signal Sin1 and outputs a sensing signal SS as an analog signal. The sensing signal SS is AD-converted by the AD conversion model 222 and output as an output signal (sensing signal) Sout1 as a digital signal. The transfer function model 221 includes a filter. As described later, the type of the filter and the characteristics of the filter (cutoff frequency, etc.) can be set. In addition, the characteristics of the AD conversion model 222 (sampling speed, etc.) can also be set.

[0227] The physical signal waveform data output from the motor model 211 (or data based on the physical signal waveform data) is input to the sensor model 22 as the input signal Sin1. m , it is input to the sensor model 22 as a current sensor. If the physical signal waveform data (or data based on the physical signal waveform data) is X-direction displacement, Y-direction displacement, X-direction acceleration and Y-direction acceleration, it is input to the sensor model 22 as a vibration sensor.

[0228] In addition, if Fig.33 As shown, noises Ns1, Ns2, and Ns3 can be inputted respectively in the front stage of the transfer function model 221, between the transfer function model 221 and the AD conversion model 222, and in the back stage of the AD conversion model. With respect to noise, a noise type such as normal distribution noise or uniform distribution noise is specified, and parameters such as the mean value or variance (standard deviation) can be set according to the type.

[0229] <Machine Learning Model>

[0230] Next, the machine learning model 23 is explained. Fig.34 2 is a diagram showing a configuration example of the machine learning model 23. The machine learning model 23 includes a preprocessing unit 231 and a machine learning unit 232.

[0231] The preprocessing unit 231 preprocesses the input signal Sin2. The preprocessing includes envelope processing, window function processing and FFT (fast Fourier transform) processing. As described later, envelope processing, window function processing and FFT processing can be selected whether to be performed. As a mode, it is possible to select to perform only window function processing, only envelope processing, only FFT processing, window function processing and FFT processing, envelope processing and FFT processing, or envelope processing, window function processing and FFT processing. In addition, it is not limited to FFT processing, for example, frequency analysis processing such as wavelet transform can also be used.

[0232] In addition, in the preprocessing unit 231, a normalization process is also performed for proper learning in the machine learning unit 232. Normalization processing is a process of multiplying data by a normalization coefficient so that the data converges to a range of approximately 0 to 1 (or -1 to +1). When the input data is a value outside the specified range, learning is not performed or learning is performed as a saturated value. Therefore, in order to learn all the data, a normalization process is required as a preprocessing.

[0233] When preprocessing is selected, after performing at least one of window function processing and FFT processing on the input signal Sin2, normalization processing is performed and the result is set as input data Din to the machine learning unit 232. In addition, when no preprocessing is selected, the input signal Sin2 is normalized and the result is set as input data Din.

[0234] The machine learning unit 232 performs learning and inference on the input data Din.

[0235] As an AI model used in the machine learning unit 232, for example, Fig.35 A three-layer neural network 30 as shown.

[0236] like Fig.35 As shown, the three-layer neural network 30 is an AI model having an input layer 30A, a hidden layer 30B, and an output layer 30C. Generally speaking, in the three-layer neural network 30, for n-dimensional input data x∈R k×n , n'-dimensional inference result y∈R k×n’ It is obtained as y=G(x·α+b)β. Here, α∈R n×m is the weight that combines the input layer 30A with the hidden layer 30B, β∈R m×n’ is the weight that connects the hidden layer 30B and the output layer 30C. In addition, b∈R m is the bias of hidden layer 30B, and G is the activation function of hidden layer 30B.

[0237] In this embodiment, an algorithm is used that can sequentially learn the three-layer neural network 30 with an arbitrary batch size. i The i-th learning data {x i ∈R ki×n , t i ∈R ki×n’}, it is necessary to find β that minimizes the error expressed by the following formula (8) i .

[0238]

[0239] In addition, the i-th hidden layer matrix H i =G(x i ·α+b). In addition, t is the training data corresponding to the inference result y.

[0240] The optimized weight β i It is calculated by the following formula (9).

[0241] P i =P i-1 -P i-1 H i T (I+H i P i-1 H i T ) -1 H i P i-1

[0242] β i =β i-1 +P i H i T (t i -H i β i-1 ) (9)

[0243] Here, P0 and β0 are obtained by the following formula (10).

[0244] P0=(H0 T H0) -1

[0245] β0=P0H0 T t0 (10)

[0246] The learning algorithm is as follows.

[0247] (1) Initialize the values ​​of weight α and bias b using random numbers.

[0248] (2) Calculate H0 relative to x0, and calculate P0 and β0.

[0249] (3) Whenever we get the batch size k i When learning data of the ith time, P is calculated successively i and β i In addition, instead of using the calculation formula for β0 in formula (10), a value initialized by a random number may be set as β0.

[0250] In addition, in this embodiment, learning is performed using an autoencoder. The autoencoder uses the input data directly as training data and performs learning in a manner that can reconstruct the input data into an inference result. That is, as described above, learning is performed by setting t=x. The autoencoder does not need to produce training data separately, and therefore becomes a type of unsupervised learning algorithm.

[0251] According to the AI ​​model in the machine learning unit 232, learning can be performed by a microcomputer-level computing device in the edge device. In particular, the bottleneck of the calculation amount in the above formula (9) is (I+H i P i-1 H i T ) -1 , but (I+H i P i-1 H i T ) is k×k, so when k=1, the inverse matrix operation can be replaced by the reciprocal operation. Therefore, by fixing the batch size to k=1, the operation becomes easy even for a microcomputer-level operation device. That is, when such on-device learning is introduced into the abnormality detection of the motor, the effect of the abnormality detection can be confirmed by simulation. In addition, the input data x is time series data when there is no FFT processing in the preprocessing unit 231, and is frequency domain data when there is FFT processing.

[0252] In the machine learning unit 232, the abnormality is calculated by the loss function L(y, t) representing the error between the inference result y and the training data t. The loss function is, for example, MAE (Mean Absolute Error) or MSE (Mean Squared Error). When the loss function is MAE, the loss function L is represented as shown in the following formula (11).

[0253]

[0254] In addition, when the loss function is MSE, the loss function L is expressed as shown in the following formula (12).

[0255]

[0256] Since learning is performed using an autoencoder, the error is calculated by setting the loss function L(y, t)=L(y, x), and the calculated error is set as the abnormality. The calculated abnormality is output from the machine learning unit 232 as abnormality data Dab.

[0257] In addition, in this embodiment, a forgetting rate can be set. The forgetting rate is a parameter indicating the degree to which the learning result is forgotten. As a method of not reflecting the learning result, for example, an initialization learning result using past learning results is listed.

[0258] <Abnormality judgment model>

[0259] Next, the abnormality determination model 24 is described. An abnormality determination is performed based on the abnormality degree data Dab output from the machine learning model 23. The abnormality determination model 24, for example, compares the abnormality degree with a threshold value to determine whether it is abnormal or normal. In addition, the abnormality determination model 24, for example, can compare the abnormality degree with multiple thresholds to determine the abnormality level in stages. As described later, such an abnormality determination method can be selected. In addition, in the abnormality determination model 24, it is also possible to compare with the threshold value after performing processing such as integration or averaging of the abnormality degree.

[0260] <Simulation Setting GUI>

[0261] Next, a GUI (Graphical User Interface) capable of setting simulation conditions in the simulation device 1 of this embodiment will be described. Examples of various setting screens described below are provided by the display control unit 5 ( Figure 2 ) is displayed on the display unit 7. Selection and setting in the setting screen, or switching of the screen, etc. are performed based on inputs of the operation input unit 6. The contents set in the various setting screens are set by the model setting unit 4.

[0262] Fig.36 1 is a diagram showing a setting screen for motor type and abnormal state (hereinafter referred to as a first setting screen). In the first setting screen, labels TB are displayed on the upper left and right sides. By switching labels TB, the setting screen can be switched. In addition, in the first setting screen, a motor type selection unit SG1 and an abnormal state selection unit SG2 are displayed.

[0263] In the motor type selection unit SG1, selection candidate motors are displayed and the motors can be selected. Fig.36 In the figure, as an example, a BDC motor and a BLDC motor are displayed as motor types, and the BDC motor is selected.

[0264] In the abnormal state selection unit SG2, the abnormal state of the motor can be selected. Specifically, poor lubrication of the bearing, damage to the outer ring of the bearing, damage to the inner ring of the bearing, and damage to the rolling element of the bearing are displayed as selection candidates, and any abnormal state can be selected. Fig.36 In this example, inner race damage is selected.

[0265] In addition, the first setting screen displays a lamp LP0 indicating a mechanical element corresponding to a selectable abnormal state. The lamp LP0 is displayed in the display showing the overall structure of the motor and can be turned off and on. The lamp LP0 of the mechanical element corresponding to the selected abnormal state is lit. Fig.36 In the example, inner ring damage is selected, and therefore the lamp LP0 of the bearing as the corresponding mechanical element is lit. With such lamp LP0, it is easy to confirm the mechanical element corresponding to the abnormal state of the simulation object.

[0266] In addition, the first setting screen displays lamps LP1 to LP4 that indicate selectable locations where abnormal conditions have occurred. Fig.36 As an example, the light LP1 indicating the lubricant of the bearing, the light LP2 indicating the outer ring, the light LP3 indicating the inner ring, and the light LP4 indicating the rolling element are displayed. LP1 corresponds to poor bearing lubrication, LP2 corresponds to outer ring damage, LP3 corresponds to inner ring damage, and LP4 corresponds to rolling element damage. LP1 to LP4 can be turned off and on. The lights LP1 to LP4 corresponding to the selected abnormal state are lit. Fig.36 In the example of FIG. 1 , inner race damage is selected, so lamp LP3 is lit. By using such lamps LP1 to LP4, it is possible to easily confirm the location where the abnormal state of the simulated object occurs.

[0267] In addition, at the bottom of the first setting screen, a simulation time setting section ST, a simulation start button SB, and an abnormal state lamp FLP are displayed. In addition, these displays are displayed in common even when the setting screen is switched. In the simulation time setting section ST, the simulation time can be set. When the time of the value input to the simulation time setting section ST has passed since the start of the simulation, the simulation is stopped. In addition, when the time to stop the simulation before the inference start time described later is set, the simulation is stopped before starting the inference.

[0268] The simulation can be started by pressing the simulation start button SB. The abnormal state lamp FLP is a lamp corresponding to the selectable abnormal state, and can be turned off and on. Fig.36 In the example, as abnormal state lights FLP, lights indicating poor bearing lubrication, outer ring damage, inner ring damage, and rolling element damage are displayed side by side. Fig.36 In the example, since the inner race damage is selected, the inner race damage abnormal status lamp FLP is lit.

[0269] When switching from the first setting screen to the adjacent tab TB, the display Fig.37 The motor basic setting screen is shown as shown. In the motor basic setting screen, a basic parameter setting unit SG3 and a drive setting unit SG4 are displayed.

[0270] In the basic parameter setting unit SG3, the basic parameters of the motor can be set for the motor type selected in the first setting screen. Fig.37 As an example, the winding resistance (R_Wire), winding inductance (L_Wire), and inertia (J m ).

[0271] In the drive setting unit SG4, the drive method and drive voltage can be set for the motor type selected in the first setting screen. The drive method can be set to constant voltage drive, PWM drive, three-phase drive (rectangular wave drive, sine wave drive), etc. The drive voltage can be set to the voltage applied to the motor terminal during constant voltage drive, or the high level voltage when the voltage is applied to the motor terminal by switching between high level and low level.

[0272] Furthermore, the basic parameter setting unit SG3 and the drive setting unit SG4 may change the items that can be set according to the selected motor type.

[0273] When switching from the Motor Basic Setting screen to the adjacent tab TB, the display Fig.38 The support system setting screen is shown as shown. The support system setting screen displays the rigidity / damping property setting section SG5. In the rigidity / damping property setting section SG5, spring constants (rigidity) kx, ky and damping coefficients cx, cy in the vibration model 2114 of the support system can be set.

[0274] When switching from the support system setting screen to the adjacent tab TB, the display Fig.39 The first abnormal state setting screen is shown in FIG. 1. The abnormal state setting screen displays an abnormal parameter setting unit SG6. In the abnormal parameter setting unit SG6, B lub_cof1 , B lub_index1 , B lub_cof2 SW_B lub_cof2 , B lub_index2 , μ lub .

[0275] When switching from the first abnormality setting screen to the adjacent tab TB, the display Fig.40 The second abnormal state setting screen is as shown. In the second abnormal state setting screen, the poor lubrication position setting part SG7 and the bearing damage position setting part SG8 are displayed.

[0276] In the poor lubrication position setting unit SG7, the position where the poor lubrication occurs in the bearing can be set by angle. In the bearing damage position setting unit SG8, ​​the position where the outer ring damage occurs can be set by angle. The angular positions set in the poor lubrication position setting unit SG7 and the bearing damage position setting unit SG8 correspond to the angular position θ f , where the angular position θ f The vibration model 2114 ( Fig.32 ) in the direction of the external force F applied as the vertical resistance caused by the anomaly.

[0277] When switching from the second abnormal state setting screen to the adjacent tab TB, the display Fig.41 The time / degradation setting screen is as shown. The time / degradation setting screen displays a time setting section SG9, a lubrication failure degradation setting section SG10, and a bearing damage degradation setting section SG11. In addition, the time / degradation setting screen also displays an explanation display ED for explanation.

[0278] In the time setting unit SG9, the learning start time, the inference start time and the degradation step time can be set. The learning time and the inference start time are represented by the elapsed time from the start of the simulation. The learning start time and the inference start time are displayed on the explanation display ED. According to the setting of the learning start time, the data at the time of motor startup can also be not used for learning.

[0279] As shown in the explanation display ED, the progress of degradation is represented by gain. The state of gain = 0 is the normal state, and degradation progresses from the normal state to the first degradation state Gain0, the second degradation state Gain1, and the third degradation state Gain2 in stages. The degradation step time is the maximum time for the normal state and each degradation state to be maintained. When the simulation ends when the degradation step time is not completed, the normal state or a degradation state is maintained for a time shorter than the degradation step time.

[0280] In the lubrication failure setting unit SG10, the constant coefficient B in the above formula (3) for calculating the friction torque caused by the bearing lubrication failure can be set. lub_cof1 and B lub_cof2 The gain values ​​in the respective first to third degradation states. For example, when the gain value = 0, it indicates the coefficient value in the normal state, when the gain value = 1, it indicates the coefficient value in the predetermined abnormal state (set by the abnormal parameter setting unit SG6), when the gain value = 0.5, it indicates 0.5 times the coefficient value in the predetermined abnormal state, when the gain value = 2, it indicates 2 times the coefficient value in the predetermined abnormal state, and when the gain value = 4, it indicates 4 times the coefficient value in the predetermined abnormal state. The larger the gain value, the larger the coefficient, indicating the progression of degradation of poor lubrication.

[0281] In the bearing damage degradation setting unit SG11, the height P of the impact pulse caused by bearing damage (damage to the outer ring, inner ring or rolling element) can be set. hight and width P width The gain values ​​in the first to third degradation states. The specific example of the gain value is the same as that of the coefficient. The larger the gain value, the larger the height and width of the impulse pulse, indicating that the degradation is progressing.

[0282] When switching from the time / degradation setting screen to the adjacent tab TB, the Fig.42 The sensor setting screen is as shown in FIG. 1. The sensor setting screen displays a sensor type setting unit SG12, a filter setting unit SG13, a sensor characteristic setting unit SG14, a sampling rate setting unit SG15, and a noise setting unit SG16.

[0283] In the sensor type setting unit SG12, the sensor type in the sensor model 22 can be selected. Fig.42 As an example, at least one of a current sensor, a first vibration sensor, and a second vibration sensor can be selected. Setting by sensor type Fig.33 The transfer function model 221 and the AD conversion model 222 are shown. The current sensor inputs the motor terminal current from the motor model 211 as a physical signal waveform and outputs a current sensing signal. The first vibration sensor and the second vibration sensor respectively input displacement and acceleration from the motor model 211 as physical signal waveforms and output displacement sensing signals and acceleration sensing signals respectively.

[0284] In the filter setting unit SG13, settings related to the filter (included in the transfer function model 221) in the sensor model 22 can be made. This filter is added in consideration of noise removal or the frequency range and Nyquist frequency of the sensor. The filter can be set according to the sensor type. For example, the filter type (low-pass filter (LPF), band-pass filter (BPF), etc.) and parameters related to the filter (cutoff frequency, etc.) can be set.

[0285] The sensor characteristic setting unit SG14 can set the characteristics of the sensor itself for each sensor type. For example, the frequency characteristics of the sensor itself can be expressed by a transfer function (LPF, BPF, etc.). Fig.42 G(s) shown represents a transfer function, and s is a Laplace operator.

[0286] The sampling rate setting unit SG15 can set the sampling rate in the AD conversion model 222. The sampling rate can be set for each sensor type.

[0287] In the noise setting unit SG16, the type and parameter of each of the noises Ns1, Ns2, and Ns3 can be set.

[0288] In addition, the sensor setting screen also displays a sensor position setting unit SG17. In the sensor position setting unit SG17, the position of the sensor can be set. The position of the sensor can be set by selecting a sensor using the operation input unit 6 and moving it on the screen. Fig.42 In FIG. 1 , as an example, the first vibration sensor VS1 and the second vibration sensor VS2 are displayed, and the positions of the first vibration sensor VS1 and the second vibration sensor VS2 can be set. Fig.42 In the diagram, the first vibration sensor VS1 is arranged on the X-axis, and the second vibration sensor VS2 is arranged on the Y-axis.

[0289] When switching from the sensor setting screen to the adjacent tab TB, the Fig.43 The AI ​​setting screen is shown as shown. In the AI ​​setting screen, a normalization coefficient setting unit SG18 and a machine learning setting unit SG19 are displayed.

[0290] The preprocessing unit 231 and the machine learning unit 232 in the machine learning model 23 are set according to the sensing signals output from the sensor model 22. As an example, the preprocessing unit 231 and the machine learning unit 232 are respectively set according to the current sensing signal output from the current sensor, the displacement sensing signal and the acceleration sensing signal output from the first vibration sensor, and the displacement sensing signal and the acceleration sensing signal output from the second vibration sensor (five sensing signals).

[0291] In the normalization coefficient setting unit SG18, it is possible to set a normalization coefficient for each sensing signal used for normalization in the preprocessing unit 231. Fig.43 As an example, the normalization coefficients of the five sensing signals can be set.

[0292] In the machine learning setting unit SG19, various items related to the machine learning unit 232 can be set. Fig.43 As an example, for a three-layer neural network 30 ( Fig.35 ), the number of input data n, the number of hidden layer nodes m, the activation function, the loss function and the forgetting rate can be set. The activation function can be set to, for example, Sigmoid or ReLU. In addition, the loss function can be set to, for example, MAE or MSE.

[0293] In addition, the AI ​​setting screen also displays a preprocessing setting unit SG20. In the preprocessing setting unit SG20, the presence or absence of each of envelope processing, FFT processing (frequency analysis processing), and window function processing can be set.

[0294] When switching from the AI ​​setting screen to the adjacent tab TB, the Fig.44 The abnormality determination setting screen is as shown. The abnormality determination setting screen displays an abnormality determination method setting unit SG21 and an abnormality determination threshold value setting unit SG22.

[0295] In the abnormality determination method setting unit SG21, an abnormality determination method according to the degree of abnormality can be selected in the abnormality determination model 24. Fig.44 As an example, a first determination method (Method 1) and a second determination method (Method 2) can be selected. The first determination method is a method of comparing the abnormality with a threshold value to determine abnormality or normality. The second determination method is a method of comparing the abnormality with multiple threshold values ​​to determine the abnormality level in stages.

[0296] In the abnormality determination threshold value setting unit SG22, a threshold value used in the selected abnormality determination method can be set.

[0297] If all setting items are set through the above setting screen, and the simulation start button SB is pressed through the operation input unit 6, the model calculation unit 3 executes the simulation according to the content set by the model setting unit 4. Fig.45 By clicking the tab TB of the simulation result screen as shown, the simulation result is displayed on the display unit 7 through the display control unit 5.

[0298] In the simulation result screen, the time series data of the sensing signal of the selected sensor and the time series data of the abnormality corresponding to the sensing signal can be displayed. Fig.45 As an example, the time series data of the motor terminal current, X-direction displacement, Y-direction displacement, X-direction acceleration, and Y-direction acceleration, as well as the time series data of the abnormality of the motor terminal current, X-direction displacement, Y-direction displacement, X-direction acceleration, and Y-direction acceleration are displayed. In addition, as Fig.42 As shown, since the first vibration sensor VS1 is configured on the X-axis and the second vibration sensor VS2 is configured on the Y-axis, the X-direction displacement and the X-direction acceleration are detected by the first vibration sensor VS1, and the Y-direction displacement and the Y-direction acceleration are detected by the second vibration sensor.

[0299] In addition, the time series data of the abnormality determination results in the abnormality determination model 24 or the preprocessing unit 231 ( Fig.34 ) in the FFT processing result (power spectrum).

[0300] In this way, simulation conditions can be intuitively set through the GUI, and simulation results can be confirmed by executing the simulation, thereby confirming the effectiveness of motor abnormality sign detection based on machine learning.

[0301] <Other>

[0302] In addition, in addition to the above-mentioned embodiments, various technical features disclosed in this specification can also be subjected to various changes within the scope of the technical creation spirit. That is, the above-mentioned embodiments should be considered to be illustrative and non-restrictive in all aspects, and the technical scope of the present disclosure is not limited to the above-mentioned embodiments, and should be understood to include all changes within the meaning and scope equivalent to the claims.

[0303] <Note>

[0304] As described above, the simulation device (1) according to one aspect of the present disclosure is configured to include:

[0305] The model storage unit (2) stores the following models:

[0306] A motor physical model (2111) modeled by a winding circuit portion (2111B) and a rotational motion equation portion (2111A); and

[0307] an abnormal state model (211A) obtained by modeling the abnormal state of the motor; and

[0308] A model calculation unit (3) is configured to perform calculation processing using the motor physical model.

[0309] An abnormality parameter indicating a deviation from a normal state calculated by the abnormal state model is used as an input to the motor physical model (first configuration).

[0310] In addition, in the above-mentioned first structure, it can also be configured as follows:

[0311] The abnormal state model includes two or more motor abnormality models (2112, 2113),

[0312] The two or more motor abnormality models do not exchange signals between the models (second structure).

[0313] In addition, in the first or second structure, it can also be configured as follows:

[0314] The abnormal state model includes: a bearing poor lubrication model (2112) obtained by modeling poor lubrication in the bearing of the motor,

[0315] The bearing poor lubrication model is configured to input a mechanical angle velocity from at least the motion equation section, and calculate a friction torque caused by poor lubrication based on the mechanical angle velocity.

[0316] The calculated friction torque is input as the abnormal parameter to the motion equation unit (third configuration).

[0317] In addition, in the third structure, it can also be configured as follows:

[0318] The bearing poor lubrication model calculates the friction torque T based on the mechanical angle velocity ωm and the mechanical angle θm input from the motion equation unit and the following equation: lubrication ,

[0319]

[0320] Among them, B lub_cof1 and B lub_cof2 is the coefficient, B lub_index1 and B lub_index2 is the exponent of the mechanical angular velocity, f(θ m ) is the mechanical angle θ m Function (fourth structure).

[0321] In addition, in the fourth structure, it can also be configured as follows:

[0322] The bearing poor lubrication model calculates the vertical resistance N caused by poor lubrication according to the following formula: lubrication ,

[0323]

[0324] Among them, μ lub is the friction coefficient, and R is the radius of the rotating shaft of the motor (fifth structure).

[0325] In addition, in any one of the first to fifth structures described above, it may also be configured as follows:

[0326] The abnormal state model includes: a bearing damage model (2113) obtained by modeling damage in a bearing of the motor,

[0327] The bearing damage model is configured to calculate the friction torque caused by bearing damage.

[0328] The calculated friction torque is input as the abnormal parameter to the motion equation unit (sixth configuration).

[0329] In addition, in the sixth structure, it can also be configured as follows:

[0330] The bearing damage model is constructed based on the vertical resistance N caused by the bearing damage. bearing_damage The friction torque T is calculated by the following formula: bearing_damage ,

[0331] T bearing_damage =μbearing ·N bearing_damage ·R

[0332] Among them, μ bearing is the friction coefficient, and R is the radius of the rotating shaft of the motor (seventh structure).

[0333] In addition, in the seventh configuration, it may also be configured as follows:

[0334] The bearing damage model is modeled as follows: when the outer ring is damaged, each time the rolling element's revolution angle θ revolution When the following formula is satisfied, the height P of the shock pulse is generated. hight The vertical resistance N bearing_damage ,

[0335]

[0336] Where Z is the number of rolling elements, P width is the width of the pulse impact, and n is an integer starting from 0 (eighth structure).

[0337] In addition, in the seventh configuration, it may also be configured as follows:

[0338] The bearing damage model is modeled as follows: when the inner ring is damaged, each time the rolling element's revolution angle θ revolution When the following formula is satisfied, the height P of the shock pulse is generated. hight The vertical resistance N bearing_damage ,

[0339]

[0340] Among them, θ m is the mechanical angle input from the equation of motion, Z is the number of rolling elements, and P width is the width of the pulse impact, and n is an integer starting from 0 (ninth structure).

[0341] In addition, in the seventh configuration, it may also be configured as follows:

[0342] The bearing damage model is modeled as follows: when the rolling element is damaged, whenever the rolling element's rotation angle θ rotation When the following formula is satisfied, the height P of the shock pulse is generated. hight The vertical resistance N bearing_damage ,

[0343] πn≤θ rotation ≤πn+P width

[0344] Among them, P width is the width of the pulse impact, and n is an integer starting from 0 (the tenth structure).

[0345] In addition, in any of the fifth, seventh to tenth structures described above, it may also be configured as follows:

[0346] The model storage unit stores: a vibration model (2114) of the support system when all mechanical elements constituting the motor are defined as a support system,

[0347] The vertical resistance is input to the vibration model of the support system (eleventh structure).

[0348] In addition, in the above eleventh structure, it can also be configured as follows:

[0349] The vibration model of the support system is modeled by the equation of motion, which is a structure (twelfth structure) obtained by connecting the parallel connection structure of the spring (Kx, Ky) and the damper (Cx, Cy) and the mass point containing the support system.

[0350] In addition, in the above-mentioned eleventh or twelfth structure, it can also be configured as follows:

[0351] There is no signal input from the vibration model of the support system to the abnormal state model (thirteenth configuration).

[0352] In addition, in any one of the first to thirteenth structures described above, it may be configured as follows:

[0353] The model storage unit stores:

[0354] A motor model (211), comprising the motor physical model and the abnormal state model;

[0355] A sensor model (22) configured to receive the physical signal waveform data output from the motor model and output a sensing signal;

[0356] a machine learning model (23) configured to be input with the sensing signal to perform learning and inference; and

[0357] The abnormality determination model (24) is configured to receive the abnormality degree output from the machine learning model as error data and perform abnormality determination (fourteenth configuration).

[0358] In addition, a program (P) according to one aspect of the present disclosure is for causing a computer (100) to function as a simulation device according to any one of the first to fourteenth structures (a fifteenth structure).

Claims

1. A simulation device comprising: The model storage unit stores the following models: A physical model of the motor, which is modeled by a winding circuit portion and a rotational motion equation portion; and an abnormal state model obtained by modeling an abnormal state of the motor; and a model calculation unit configured to perform calculation processing using the motor physical model; An abnormality parameter representing a deviation from a normal state calculated by the abnormal state model is used as an input to the motor physical model.

2. The simulation device according to claim 1, wherein: The abnormal state model includes two or more motor abnormality models. The two or more motor abnormality models do not have signal exchange between the models.

3. The simulation device according to claim 1 or 2, wherein: The abnormal state model includes: a bearing lubrication failure model obtained by modeling the lubrication failure in the bearing of the motor; The bearing poor lubrication model is configured to input a mechanical angle velocity from at least the motion equation section, and calculate a friction torque caused by poor lubrication based on the mechanical angle velocity. The calculated friction torque is input as the abnormal parameter to the motion equation unit.

4. The simulation device according to claim 3, wherein: The bearing poor lubrication model is based on the mechanical angle angular velocity ω input from the motion equation unit. m and mechanical angle θ m The friction torque T is calculated by the following formula: lubrication , Among them, B lub_cof1 and B lub_cof2 is the coefficient, B lub_index1 and B lub_index2 is the exponent of the mechanical angular velocity, f(θ m ) is the mechanical angle θ m function.

5. The simulation device according to claim 4, wherein: The bearing poor lubrication model calculates the vertical resistance N caused by poor lubrication according to the following formula lubrication , Among them, μ lub is the friction coefficient, and R is the radius of the motor's rotating shaft.

6. The simulation device according to any one of claims 1 to 5, wherein: The abnormal state model includes: a bearing damage model obtained by modeling damage in a bearing of the motor; The bearing damage model is configured to calculate the friction torque caused by bearing damage. The calculated friction torque is input as the abnormal parameter to the motion equation unit.

7. The simulation device according to claim 6, wherein: The bearing damage model is constructed based on the vertical resistance N caused by the bearing damage. bearing_damage The friction torque T is calculated by the following formula: bearing_damage , T bearing_damage =μ bearing ·N bearing_damage ·R Among them, μ bearing is the friction coefficient, and R is the radius of the motor's rotating shaft.

8. The simulation device according to claim 7, wherein: The bearing damage model is modeled as follows: when the outer ring is damaged, each time the rolling element's revolution angle θ revolution When the following formula is satisfied, the height P of the shock pulse is generated. hight The vertical resistance N bearing_damage , Where Z is the number of rolling elements, P width is the width of the pulse impact, and n is an integer starting from 0.

9. The simulation device according to claim 7, wherein: The bearing damage model is modeled as follows: when the inner ring is damaged, each time the rolling element's revolution angle θ revolution When the following formula is satisfied, the height P of the shock pulse is generated. hight The vertical resistance N bearing_damage , Among them, θ m is the mechanical angle input from the equation of motion, Z is the number of rolling elements, and P width is the width of the pulse impact, and n is an integer starting from 0.

10. The simulation device according to claim 7, wherein: The bearing damage model is modeled as follows: when the rolling element is damaged, whenever the rolling element's rotation angle θ rotation When the following formula is satisfied, the height P of the shock pulse is generated. hight The vertical resistance N bearing_damage , πn≤θ rotation ≤πn+P width Among them, P width is the width of the pulse impact, and n is an integer starting from 0.

11. The simulation device according to any one of claims 5, 7 to 10, wherein: The model storage unit stores: a vibration model of the support system when all mechanical elements constituting the motor are defined as a support system; The vertical resistance is input to a vibration model of the support system.

12. The simulation device according to claim 11, wherein: The vibration model of the support system is modeled by an equation of motion, which is a structure obtained by connecting a parallel connection structure of springs and dampers to a mass point including the support system.

13. The simulation device according to claim 11 or 12, wherein: There is no input signal from the vibration model of the support system to the abnormal state model.

14. The simulation device according to any one of claims 1 to 13, wherein: The model storage unit stores: a motor model, comprising the motor physical model and the abnormal state model; a sensor model configured to receive the physical signal waveform data output from the motor model and output a sensing signal; a machine learning model configured to be input with the sensing signal to perform learning and inference; as well as The abnormality determination model is configured to perform abnormality determination by receiving the abnormality degree output as error data from the machine learning model. 15 . A program for causing a computer to function as the simulation device according to claim 1 .

Images