阅读说明：本技术 一种三相永磁同步电机控制系统的故障诊断方法 (Fault diagnosis method for three-phase permanent magnet synchronous motor control system ) 是由 周新秀 胡青瑶 王博阳 张旨 于 2020-12-25 设计创作，主要内容包括：本发明涉及一种应用于三相永磁同步电机控制系统的故障诊断方法。该方法首先对系统在各类故障下的故障特征进行分析总结；而后,利用差动电流状态观测器提取差动电流畸变(残差),并根据故障特征得出各类故障下的残差特征,进而设计故障检测和故障隔离算法；为提高故障检测算法的鲁棒性,设计了一种新的自适应故障检测阈值,可使算法在存在变速、变载和参数摄动的复杂工况下仍可实现快速而准确的故障检测,即赋予了算法良好的鲁棒性。该方法具有抗干扰能力强、鲁棒性强、检测速度快、检测正确率高等优点,且无需增加额外的传感电路,是适用于三相永磁同步电机控制系统的一种低成本、高性能的在线故障诊断方法。(The invention relates to a fault diagnosis method applied to a three-phase permanent magnet synchronous motor control system. Firstly, analyzing and summarizing fault characteristics of a system under various faults; then, extracting differential current distortion (residual) by using a differential current state observer, obtaining residual characteristics under various faults according to the fault characteristics, and further designing a fault detection and fault isolation algorithm; in order to improve the robustness of the fault detection algorithm, a new self-adaptive fault detection threshold is designed, so that the algorithm can still realize quick and accurate fault detection under the complex working conditions of variable speed, variable load and parameter perturbation, namely, the algorithm is endowed with good robustness. The method has the advantages of strong anti-interference capability, strong robustness, high detection speed, high detection accuracy and the like, does not need to add an additional sensing circuit, and is a low-cost and high-performance online fault diagnosis method suitable for a three-phase permanent magnet synchronous motor control system.)

1. A fault diagnosis method applied to a three-phase permanent magnet synchronous motor control system is characterized by comprising the following steps: on the premise of not increasing a sensor, the fast and accurate fault detection and isolation of inverter switching tube open circuit faults, motor winding open circuit faults and motor winding short circuit faults which are easy to occur in the system can be realized only by utilizing three-phase current, motor rotating speed, rotor angular position and voltage instruction information which can be obtained in a main control algorithm of a permanent magnet synchronous motor control system;

the method comprises the following steps:

step one, preprocessing data; firstly, the three-phase current of the motor, the rotating speed of the motor, the angular position of a rotor and voltage instruction information are processed to obtain a reference value u of the line voltage of the motor_ab ^*、u_bc ^*、u_ca ^*Differential back electromotive force e_ab、e_bc、e_ca(i.e. the result of the alternating differences of the three-phase back emf of the motor) and the differential current i₁、i₂、i₃(i.e., the result of the difference in the three phase currents of the motor by the rotation); wherein the expected value of the line voltage is a three-phase PWM wave duty ratio d generated by utilizing a motor main control algorithm_a、d_b、d_cAnd DC bus voltage U_dcObtaining a motor terminal voltage expected value through multiplication, correcting according to the nonlinear characteristic of the inverter, and then obtaining a difference through rotation; calculating two steps of differential counter electromotive force, namely, firstly obtaining the three-phase counter electromotive force of the motor according to the rotating speed of the motor, the angular position of the rotor and a pre-measured waveform function of the counter electromotive force of the motor, and then rotating the counter electromotive force to make a difference; differential current i₁₂₃(i₁₂₃＝[i₁,i₂,i₃]^T) Then the three-phase current i can be directly controlled_abc(i_abc＝[i_a,i_b,i_c]^T) The difference is calculated by rotation;

step two, calculating an estimated value of the differential current; the step of substituting the expected line voltage, the differential counter electromotive force and the differential current obtained in the step one into a state equation of a differential current observer of the motor to update the estimated value of the differential currentWherein, the state equation of the differential current observer is as follows:

wherein i₁、i₂And i₃Is a measure of the differential current of the motor, which can be expressed as [ i ]₁,i₂,i₃]^T＝[i_a-i_b,i_b-i_c,i_c-i_a]^TK is the sampling time;andare respectively i₁、i₂And i₃An estimated value of (d); G. h is a state transition matrix and a system control matrix of the system respectively, G ═ gE, H ═ hE, G ═ exp { -R_sT_s/(L-M)},h＝(1-g)/R_s，R_sIs the phase resistance of the motor, L is the mean value of three-phase inductance, M is the mean value of mutual inductance between phases, T_sIs the current sampling period, E is the identity matrix; l is_rIs a feedback matrix of an observer, L_r＝l_rE，l_rIs a feedback coefficient;andrespectively, are ideal inputs of the motor, from the desired line voltage (u)_ab ^*、u_bc ^*、u_ca ^*) And differential back electromotive force (e)_ab、e_bc、e_ca) To obtain the result of the above-mentioned method,

step three, generating a residual error; the differential current obtained in the first step and the second step is subtracted from the estimated value thereof, and the obtained difference value is filtered by a simple low-pass filter to generate a residual vector r;

step four, extracting residual error characteristics; the step is to carry out modular and normalized operation on the residual vector r to obtain the module value M of the residual_rAnd a unit direction vector r_n；

Step five, fault detection; the step is to calculate the module value M of the residual vector obtained in the step four_rAnd a set fault detection threshold value T_hComparing the sizes of the components, and judging the health condition of the system according to the comparison result; the rule of the fault judgment is as follows: if M is_r>T_hIf yes, judging the system to be in fault, otherwise, judging the system to be healthy;

step six, fault phase separation; the step is used for distinguishing the phase where the fault is located after the fault occurs; analysis shows that in the fault period, if errors are ignored, the residual error vector r is normalized_nThere are only six possible values (i.e., a)₁～a₆Called template vector after) and the direction is determined by the fault phase; therefore, can be according to r_nAnd positioning a fault phase. The specific method comprises the following steps: first, calculate r_nWith each template vector a_iA distance d between_i(i is 1, 2, … …, 6), finding out the value of i corresponding to the minimum distance (i.e. the value of i becomes the residual direction indicating variable later), and judging the fault phase according to the value of i, i.e. when i is 1 or 2, judging that the fault occurs in the A phase circuit; when i is 3 or 4, judging that the fault occurs in the A-phase circuit; when i is 5 or 6, judging that the fault occurs in the C-phase circuit; wherein, a₁～a₆The expression of (a) is:

step seven, distinguishing fault types; the method comprises the steps of determining the current fault type after the system fails; firstly, whether the current fault is a winding turn-to-turn short circuit fault can be distinguished according to whether the phase current of the fault phase is zero: if the current of the fault phase is not zero at the moment, the fault type can be judged to be winding turn-to-turn short circuit fault, otherwise, the fault type is winding open circuit fault or switching tube open circuit fault; the difference between the winding open-circuit fault and the switch tube open-circuit fault lies in whether the residual error direction in the fault period is unique or not, and whether the residual error direction is unique or not can be tested by changing a phase voltage reference value corresponding to the fault; if the phase voltage reference value changes for a period of time, representing that the residual direction indicating variable i changes, and i is obtained in the step six, the fault of the open circuit of the winding can be judged; otherwise, judging that the switching tube with the index i in the inverter has an open circuit fault;

the seven steps jointly form a fault diagnosis algorithm, and can provide real-time and accurate fault state, fault source and fault type information for the motor control system.

2. The fault diagnosis method according to claim 1, characterized in that: the designed fault detection threshold is a self-adaptive fault detection threshold, and the fault diagnosis algorithm can still realize quick and accurate fault detection under the complex working conditions of speed change, load change and parameter perturbation by adopting the threshold, and the calculation formula of the threshold is as follows:

T_h＝T₀+m₁·||i||₂+m₃·||Δi||₂

wherein, T₀The part of the threshold is a small normal number so as to avoid false detection caused by noise interference; i is the differential current of the motor, delta i is the variation of the differential current between two adjacent sampling points, | · includes₂Is a modulo operation of a vector, m₁、m₃The constant coefficient is obtained by calculating the magnitude of the model error and setting the constant coefficient by a trial and error method.

3. The fault diagnosis method according to claim 1, characterized in that: after fault phase separation is completed, distinguishing winding turn-to-turn short circuit fault and winding/switching tube open circuit fault according to the magnitude of residual error component in the fault period, said method includes the following steps:

firstly, calculating the equivalent voltage distortion component delta u corresponding to the open-circuit fault of the winding/switching tube_jAnd the equivalent voltage distortion component calculation formula is as follows:

wherein, x is a fault phase, and j is 1, 2 and 3 respectively when x is a, b and c respectively;

then, according to the state equation of the residual error, the residual error component r is obtained_jEstimated value of (a):

and finally, residual error size matching: estimating residual componentsWith its actual value r_jComparing, and if the difference between the estimated value of the residual error and the actual value is always small (namely the relative difference is within 10%) within a certain time (namely the electrical period of 1-2 motors), classifying the current fault as a winding open circuit fault or a switch tube open circuit fault; otherwise, judging the turn-to-turn short circuit fault.

4. The fault diagnosis method according to claim 1, characterized in that: the designed open circuit fault distinguishing method comprises the following steps: forcibly changing a phase voltage instruction of a fault phase, and if a new nonzero value of a fault direction indication variable i appears within a certain time (namely an electrical time constant of 1-2 motors) after the voltage instruction is changed, determining that the fault direction indication variable i is a winding open-circuit fault, otherwise, determining that the fault direction indication variable i is a switching tube open-circuit fault;

the above-described manner of changing the phase voltage command is related to the PWM modulation method of the motor, and in the motor control system using the carrier-based PWM modulation method, the phase voltage command changing formula is:

(u_xn ^*)_new＝-λu_xn ^*

in the formula, x is a fault phase, x belongs to { a, b, c }, subscript new represents an updated value, and lambda is a constant between 0.25 and 0.35;

for a system adopting an SVPWM modulation mode, the voltage reference value u of the motor under an alpha-beta coordinate system can be changed_α ^*、u_β ^*To indirectly change the phase voltage reference:

when the A phase fails:

when the B phase fails:

when the C phase fails:

Technical Field

The invention relates to a fault diagnosis method of a permanent magnet synchronous motor control system, which is mainly used for detecting three faults of open circuit of an inverter switching tube, turn-to-turn short circuit of a motor winding and open circuit of the motor winding with higher frequency in the operation process of the motor control system in real time and positioning a fault source.

Background

The permanent magnet synchronous motor has the advantages of wide speed regulation range, good dynamic response, strong controllability, high power factor and the like, and is widely applied to the fields of industry, military, aerospace and the like. The faults of the motor driving system of the frame of the inertia actuating mechanism are mainly divided into motor body and driver faults. An inverter switching tube open circuit fault is one of the most common faults in a permanent magnet synchronous motor control system, which can greatly reduce the control performance of the motor, increase the power loss of the motor, and in extreme cases can cause catastrophic failures. For the motor body, the motor winding turn-to-turn short circuit and open circuit faults can cause the three-phase load asymmetry of the motor, and the voltage and current of the non-fault phase winding and the power inverter are increased. If corresponding fault-tolerant control measures are not taken timely, more serious secondary faults are easily caused, and finally the whole motor driving system loses functions. In order to avoid the above disadvantages, it is necessary to research a fault diagnosis method for a permanent magnet synchronous motor control system to match with a fault tolerance method, so as to ensure that the motor driving system can still safely, reliably and high-performance operate after a fault. Most of the existing fault diagnosis methods for the open circuit of the inverter switching tube have the defects of high false detection rate, narrow application range, low detection speed, poor anti-interference capability and the like, and the correctness of the detection result is difficult to ensure. Therefore, the research of an accurate, universal and rapid diagnosis method is of great significance.

Disclosure of Invention

The technical problem to be solved by the invention is as follows: on the premise of not adding an additional sensor, the traditional online fault diagnosis method for the motor control system has long diagnosis time and poor robustness, and can not accurately distinguish and position the motor winding fault and the inverter switch tube fault.

The technical scheme adopted by the invention for solving the technical problems is as follows: aiming at the problems, firstly, the distortion rule of current when three faults occur independently is analyzed, and an online fault diagnosis method based on a differential current state observer is designed; obtaining the expected value of the differential current of the permanent magnet synchronous motor and the distortion of the measured value; according to the change rule of the current residual error when the inverter and the winding are in fault, a corresponding fault detection and positioning algorithm is designed. In order to reduce the false detection rate and the missed detection rate of fault diagnosis, a self-adaptive detection threshold value is designed. The method based on the differential current state observer does not need to add an additional sensor, has the advantages of high diagnosis speed, high algorithm efficiency, good anti-interference performance, strong robustness and the like, can be applied to the speed change stage of the motor, and is a high-performance universal fault diagnosis method.

The method comprises the following steps:

step one, data preprocessing. Firstly, processing information such as three-phase current of a motor, rotating speed of the motor, angular position of a rotor, voltage instruction and the like to obtain a reference value u of line voltage of the motor_ab ^*、u_bc ^*、u_ca ^*Differential back electromotive force e_ab、e_bc、e_ca(the result of the difference in the three-phase back emf of the motor) and the differential current i₁、i₂、i₃(the three phase currents of the motor are a result of the difference in rotation). Wherein the expected value of the line voltage is a three-phase PWM wave duty ratio d generated by utilizing a motor main control algorithm_a、d_b、d_cAnd DC bus voltage U_dcObtaining a motor terminal voltage expected value through multiplication, correcting according to the nonlinear characteristic of the inverter, and then obtaining a difference through rotation; calculating two steps of differential counter electromotive force, namely, firstly obtaining the three-phase counter electromotive force of the motor according to the rotating speed of the motor, the angular position of the rotor and a pre-measured waveform function of the counter electromotive force of the motor, and then rotating the counter electromotive force to make a difference; differential current i₁₂₃(i₁₂₃＝[i₁,i₂,i₃]^T) Then can be directlyFor three-phase current i_abc(i_abc＝[i_a,i_b,i_c]^T) The difference is calculated by rotation.

And step two, calculating an estimated value of the differential current. The step of substituting the expected line voltage, the differential counter electromotive force and the differential current obtained in the step one into a state equation of a differential current observer of the motor to update the estimated value of the differential currentWherein, the state equation of the differential current observer is as follows:

wherein i₁、i₂And i₃Is a measure of the differential current of the motor, which can be expressed as [ i ]₁,i₂,i₃]^T＝[i_a-i_b,i_b-i_c,i_c-i_a]^TK is the sampling time;andare respectively i₁、i₂And i₃An estimated value of (d); G. h is a state transition matrix and a system control matrix of the system respectively, G ═ gE, H ═ hE, G ═ exp { -R_sT_s/(L-M)},h＝(1-g)/R_s，R_sIs the phase resistance of the motor, L is the mean value of three-phase inductance, M is the mean value of mutual inductance between phases, T_sIs the current sampling period, E is the identity matrix; l is_rIs a feedback matrix of an observer, L_r＝l_rE，l_rIs a feedback coefficient;andrespectively, are ideal inputs of the motor, from the desired line voltage (u)_ab ^*、u_bc ^*、u_ca ^*) And differential back electromotive force (e)_ab、e_bc、e_ca) To obtain the result of the above-mentioned method,

and step three, generating residual errors. And the step of subtracting the differential current obtained in the step one and the step two from the estimated value thereof, and filtering the obtained difference value by a simple low-pass filter to generate a residual vector r.

And step four, extracting residual error characteristics. The step is to carry out modular and normalized operation on the residual vector r to obtain the module value M of the residual_rAnd a unit direction vector r_n。

And step five, fault detection. The step is to calculate the module value M of the residual vector obtained in the step four_rAnd a set fault detection threshold value T_hAnd comparing the sizes, and judging the health condition of the system according to the comparison result. The rule of the fault judgment is as follows: if M is_r>T_hIf yes, judging the system to be in fault, otherwise, judging the system to be healthy.

And step six, fault phase separation. This step is used to identify the phase in which the fault is located after the fault has occurred. Analysis shows that in the fault period, if errors are ignored, the residual error vector r is normalized_nThere are only six possible values (a)₁～a₆Later called template vector) and the direction is determined by the faulted phase. Therefore, can be according to r_nAnd positioning a fault phase. The specific method comprises the following steps: first, calculate r_nWith each template vector a_iA distance d between_i(i is 1, 2, … …, 6), finding out the value of i corresponding to the minimum distance (which becomes the residual direction indication variable later), and judging the fault phase according to the value of i, namely when i is 1 or 2, judging that the fault occurs in the A phase circuit; when i is 3 or 4, judging that the fault occurs in the A-phase circuit; when i is 5 or 6, it is determined that a fault occurs in the C-phase circuit.

Wherein, a₁～a₆The expression of (a) is:

and step seven, distinguishing fault types. This step is used to clarify the current type of fault after a system fault has occurred. Firstly, whether the current fault is a winding turn-to-turn short circuit fault can be distinguished according to whether the phase current of the fault phase is zero: if the current of the fault phase is not zero at the moment, the fault type can be judged to be winding turn-to-turn short circuit fault, and if not, the fault type is winding open circuit fault or switching tube open circuit fault. The difference between the winding open-circuit fault and the switch tube open-circuit fault lies in whether the residual error direction in the fault period is unique or not, and whether the residual error direction is unique or not can be tested by changing a phase voltage reference value corresponding to the fault; if the phase voltage reference value changes for a period of time, representing that the residual direction indicating variable i changes, and i is obtained in the step six, the fault of the open circuit of the winding can be judged; and conversely, judging that the switching tube with the index i in the inverter has an open-circuit fault.

The seven steps jointly form a fault diagnosis algorithm, and can provide real-time and accurate fault state, fault source and fault type information for the motor control system.

The principle of the invention is as follows: in a motor control system, three faults of open circuit of an inverter switch tube, turn-to-turn short circuit of a motor winding and open circuit of the motor winding can cause distortion of differential current of the motor, so that the health condition of the system can be judged as long as the current distortion can be extracted quickly and accurately, and fault detection is realized. The differential current state observer can be used to generate an ideal output of the differential current, and the difference between the ideal output and the differential current measurement value in the actual system is the distortion (also called residual) of the differential current. The residual error reflects the degree of the actual system deviating from the ideal system, so that the system can be judged to be in fault as long as the residual error is overlarge. After the fault is detected, in order to realize fault location, a fault location algorithm needs to be designed according to differences presented by the system under various fault conditions. Analysis shows that the following differences exist in the system under various faults:

1. during the fault period, the direction of the residual vector is fixed (residual orientation), and the direction is related to the fault source;

2. in a fault period, the residual error direction under the fault of a single switch tube is unique, and two opposite residual error directions exist under the fault of a winding;

3. in the open-circuit fault period of the switching tube/winding, the phase current of the phase where the fault exists is continuously zero; in the turn-to-turn short circuit fault period of the winding, the phase current of the fault phase still approximates to sinusoidal change;

4. the open circuit fault of the switch tube/winding and the turn-to-turn short circuit fault of the winding can be equivalent to cause voltage distortion to the motor, but the equivalent voltage distortion is different, and the residual errors are different.

According to the difference, a corresponding positioning algorithm can be designed to accurately distinguish and position the three types of faults. The invention designs a differential current observer to estimate the differential current of the motor by using a current state equation of the motor, and makes a difference between the estimated differential current and a feedback value of the differential current to obtain a distortion quantity of the differential current as a residual error; and extracting the characteristics of the residual error by utilizing the modular operation and the normalization operation for fault diagnosis. The following are details of the design principle of the invention:

(1) firstly, the distortion rule of the voltage under various faults is analyzed.

I. Inverter switching tube failure

When a certain switching tube in the inverter has an open-circuit fault, the fault switching tube is in a constant cut-off state, and the voltage waveform of the output end of the fault bridge arm of the inverter is seriously distorted. Upper switch tube T of A phase bridge arm₁For example, when an open circuit fault occurs, T₁At constant turn-off, affecting phase current i_aThe current path causes the voltage at the winding end of the phase A to be distorted.

When S is present as shown in FIG. 1(a)_a1 and i_a>At 0, due to T₁In a constant cut-off state, the phase current passes through a diode D₂Freewheeling is performed which causes the terminal voltage u_aoIs clamped at approximately 0V. In a similar manner, at S_a1 and i_a<0、S_a0 and i_a>0、S_a0 and i_a<In the case of 0, the phase current paths of the a-phase arm are as shown in fig. 1(b), 1(c), and 1(d), and the corresponding output terminal voltages are as shown in table 1.

TABLE 1T₁Voltage of phase-a terminal after open circuit fault

The analysis only considers the output condition of the bridge arm end voltage when the phase current is not zero, and when the phase current is close to zero, the switch tube and the diode on the bridge arm are both in a cut-off state, and the output end voltage of the bridge arm is determined by the electric state of one side of the motor. According to phase current i_aWhen the terminal voltage of the phase a winding is 0:

the terminal voltage of the A phase winding under different conditions is subjected to state space averaging to obtain T₁Average value of the voltage of the A-phase winding end after the fault in each PWM period:

wherein u is_xo ^*(x is a, b, c) is an expected value of a three-phase terminal voltage, and u may be represented_xo ^*＝d_x·U_dc，d_xIs a switching signal S_aThe duty cycle of (c). From the above formula, it can be seen that at T₁Average terminal voltage output by fault bridge arm of inverter after circuit breakIs equal to its expected value u_ao ^*If and only if i_a<When 0 is true; and in i_aAt a voltage of 0 or more, the actual terminal voltage deviates from the expected value, resulting in serious distortion of the terminal voltage at the output. Due to the system T₁Open circuit thereforeBarrier at i_a<At 0, the control of the system on the motor terminal voltage is not affected, and therefore the system operation is not affected, so the system can be considered healthy in this case, but at i_aIf 0 or more, it must be considered as a failure.

While in the actual operation, T₁I will not appear after circuit break_a>0, etc. This is because, for a normal motor control system, the phase current i is under closed-loop control_aWill track its reference value i_a ^*. Thus i_a>0 may only occur at i_a ^*>0, in case of a high frequency signal. And when the phase current reference value i_a ^*>At 0, assume the actual phase current i_aCan track its reference value and become positive, due to the failure of the switching tube, the terminal voltage u at that time_anWill become negative, which will also cause i to become negative_aRapidly decrease until i_aIs negative. Once i is completed_a<0, the system can be considered normal, control i_aIncreasing to track its positive reference. Once i is_aIf the value is positive again, the above process is repeated again. Thus, macroscopically, when i_a ^*>0, actual phase current i_aWill continue to be zero as shown in fig. 2.

Easily obtain T according to the expression of the winding end voltage of the phase A₁Expression of the amount of terminal voltage distortion caused by a tube break fault:

although the above formula quantitatively describes the amount of terminal voltage distortion, it is difficult to intuitively obtain the variation trend, so a certain conversion is required. In a three-phase permanent magnet synchronous motor control system, the following coupling relation exists between a motor terminal voltage expected value and a phase voltage expected value:

wherein u is_an ^*、u_bn ^*、u_cn ^*The expected motor phase voltage value is obtained. With this equation, the terminal voltage distortion can be expressed as:

as can be seen from the above formula, the terminal voltage distortion of the fault phase is only related to the expected value of the phase voltage and the back electromotive force. In a normal permanent magnet synchronous motor control system, the relation formula satisfied between various voltages and currents is generalized to be approximately satisfied between corresponding reference values. Therefore, the phase voltage desired values that are readily available approximately satisfy the following relationship between the phase current desired values:

therefore, the terminal voltage distortion amount is simplified as:

in the formula, Z_LIs the load impedance of the motor windings.

It can be seen that at switch T₁Terminal voltage distortion Deltau u after open circuit fault_aoPhase-dependent current reference value i_a ^*May vary. If neglecting the phase lag caused by the inductance, Δ u_aoPhase-dependent current reference value i_a ^*The trend of the change will be shown in fig. 2. As can be seen from FIG. 2, T₁After failure, voltage distortion Δ u_aoAt i_a ^*<0 is zero, and i is_a ^*>0 is negative and since the phase current expected value is approximately a periodic function of electrical angle, the terminal voltage also followsThe electrical angle appears periodic. Similarly, the polarities of the voltage distortion at the lower ends of the different open-circuit faults of the switching tube can be obtained, as shown in table 2.

TABLE 2 Voltage distortion law under open-circuit fault of switching tube

Through analysis of the fault of the switching tube, the fault of the switching tube can be found out, the voltage of the output end of the fault bridge arm is distorted due to the open-circuit fault of the switching tube, and further, the voltage of the motor phase is all generated, so that the three-phase current is seriously distorted, and the fluctuation of the output torque of the motor is caused. The polarity of the terminal voltage distortion is determined by the position of a fault source and the polarity of a fault phase current expected value.

In addition, for a single switch tube open circuit fault, the terminal voltage distortion time occupies a half cycle in each electrical cycle, and the phase current corresponding to the fault is zero in the half cycle; in the other half period, the fault does not affect the operation rule of the system, so the system can be considered as normal. For convenience of description, a period in which the voltage distortion is not zero after the failure is hereinafter referred to as a failure period, and a period in which the voltage distortion disappears is hereinafter referred to as a recovery region. For a single tube open circuit fault, the fault period and recovery zone alternate and each time last half an electrical cycle.

II. Open circuit fault of winding

The open circuit fault of the motor winding mainly comes from faults of fusing of a winding wire, failure of a conducting ring and the like of the motor caused by factors such as long time, large current, multiple oscillations and the like, and typically shows that the phase current of a fault phase is continuously zero. Note that the open-circuit fault of the switching tube also shows that the phase current is continuously zero in the fault period, so that the motor winding fault can be equivalent to no fault of the motor winding, and the upper and lower switching tubes of the connected bridge arms are alternately opened, and each opening lasts for half an electrical cycle. As can be seen from the above conclusion about the open-circuit fault of the switching tube, the open-circuit fault of the x-phase winding can be further equivalent to the terminal voltage of the phase winding, and the expression of the distortion is:

as can be seen from the equation, the terminal voltage distortion is approximately sinusoidal, and the polarity of the terminal voltage distortion is opposite to the polarity of the phase current reference value, so the corresponding equivalent voltage distortion law is shown in table 3.

TABLE 3 distortion law of equivalent voltage under open circuit fault of winding

III, turn-to-turn short circuit fault of winding

When a winding turn-to-turn short circuit fault occurs, a short circuit loop is added between turns of the fault phase winding. Taking the turn-to-turn short circuit fault of the phase a winding as an example, the equivalent physical model is shown in fig. 3. At this time, the voltage balance equation in the ABC coordinate system is:

where q is the failure rate, which is numerically equal to the ratio of the number of short circuit turns to the total number of turns. R_fFor short-circuit resistance, assuming that the air-gap field is still uniform after a fault and neglecting inverter nonlinearities, the above equation can be simplified as:

in the formula i_faIs the short-circuit current in the case of a fault and has the value:

in the formula u_an ^*Is the phase voltage command for phase a of the motor. The motor can be known from the simplified voltage balance equationAfter the turn-to-turn short circuit fault occurs, the three-phase load of the system is asymmetric. For the convenience of analysis, the effect caused by the fault can be equivalent to the fact that interference is injected into the three-phase voltage of the motor from the outside, and the last term of the simplified voltage balance equation is a quantitative expression of the simplified voltage balance equation, namely:

in the expression of the equivalent voltage disturbance, i is_faAnd u_an ^*Linear relationship, under the action of closed-loop control_an ^*Is approximately sinusoidal, so i_faIt can also be approximated to a sine wave of the same frequency, the time derivative of which is also a sine wave of the same frequency, so that the equivalent voltage disturbance is also approximated to a sine signal of the same frequency.

In addition, as can be seen from the expression of the equivalent phase voltage interference, after a turn-to-turn short circuit fault occurs in a certain phase winding of the motor, the equivalent voltage interference has components on three axes ABC, which causes the integral distortion of three-phase current and the torque ripple of the motor.

(2) Fault detection algorithm design

The continuous time state equation of the system can be obtained by the voltage balance equation of the normal three-phase permanent magnet synchronous motor as follows:

let the current sampling frequency of the system be T_sThen the discrete state equation of the system is:

wherein G, H is a state transition matrix and a system control matrix of the system, G ═ gE, H ═ hE, G { -exp { -R {, respectively_sT_s/(L-M)}，h＝(1-g)/R_sAnd k is the sampling instant. According to the discrete state equation of the systemDesigning a current state observer:

wherein the content of the first and second substances,is an estimate of the phase current. In order to improve the anti-interference capability of the observer, the result of alternate differential of the phase current can be used as a system state quantity, and feedback is introduced to form the differential current closed-loop observer:

wherein i₁、i₂And i₃Is a measure of the differential current of the motor, which can be expressed as [ i ]₁,i₂,i₃]^T＝[i_a-i_b,i_b-i_c,i_c-i_a]^T， Andare respectively i₁、i₂And i₃Estimated value of, L_rIs a feedback matrix of an observer, L_r＝l_rE。Andthe ideal input of the motor is respectively, and the calculation formula is as follows:

according to the relation between the phase voltage and the terminal voltage of the motor, another approximate expression of the equivalent input voltage can be obtained:

in the formula, the three-phase counter electromotive force can be determined according to the electrical angular velocity omega of the motor_eElectrical angle theta_eThe expected value of the terminal voltage can be obtained by calculating the duty ratio d of the three-phase PWM waveform according to a master control algorithm_a、d_b、d_cAnd DC bus voltage U_dcThe product is obtained. The nonlinear property of the inverter can distort the terminal voltage of the motor, and the estimation precision of an observer is influenced. Therefore, the terminal voltage needs to be corrected according to the nonlinearity of the inverter, so that the false detection of the nonlinear trigger system of the inverter is avoided. The amount of terminal voltage distortion caused by inverter nonlinearity can be expressed as:

wherein, t_dFor added dead time, t_onFor the conduction time of the power switch tube, t_offFor turn-off time of power switching tube, U_SAnd U_DThe tube voltage drops of the power switch tube and the freewheeling diode are respectively, and sign (·) is a sign function. Thus, the structure of the entire observer is as shown in fig. 4.

The performance of the closed-loop observer depends heavily on the choice of the feedback coefficient, which can be determined according to the desired pole in practical applications. Generally, the closer the pole of the discrete observer is to the origin, the faster its tracking speed is, and the higher the estimation accuracy is, but this will also result in a decrease in the disturbance rejection capability of the observer. In order to realize better comprehensive performance, the speed of the pole of the closed loop of the observer is usually set to be 3-5 times of the speed of the pole of the actual system.

By analyzing the faults of the switching tube and the motor winding, if only the response to the phase current is considered, the two faults can be equivalent to that the input voltage of the motor is injected with regular interference, so that the phase voltage is distorted (equivalent voltage distortion for short), and finally the phase current is distorted. Therefore, a unified voltage equation of the system under various faults can be obtained:

wherein, Δ u_an、Δu_bnAnd Δ u_cnThe equivalent phase voltage distortion caused by the fault. The differential current state equation of the motor can be obtained according to the voltage equation as follows:

wherein the voltage Deltau₁、Δu₂、Δu₃Is the equivalent differential voltage distortion caused by the fault.

The observer equation and the actual differential current state equation are subtracted, and the residual error state equation is obtained as follows:

the Z transformation of the above equation yields the Z transfer function from each voltage distortion to the residual as:

wherein R is_j(z) and. DELTA.U_j(z) are residual components r_jAnd equivalent differential voltageDistortion component Δ u_jZ transformation of (2)_pClosed loop pole of observer, z_pG-l. The Z transfer function indicates that the voltage distortion to the residual is a first order inertial element, the time constant of which is determined by the pole of the observer. Therefore, the change rule of the residual error is consistent with the voltage distortion rule.

For convenient analysis of residual features, a residual vector is defined as r ═ r₁,r₂,r₃]^TThe modulus and normalized vector are M_r＝||r||₂(||·||₂For modulo arithmetic), r_n＝r/||r||₂. When the system is normal, the distortion of the motor is very small, so that the residual error is close to zero; once the system fails, the motor distortion becomes significantly larger, which also results in a rapid increase in residual error. Thus, M can be substituted_rAs a fault detection index, when M_rIf the value is larger than a certain threshold value, the system can be judged to be in fault.

After a system fault is detected, the fault needs to be located according to other information. Analysis finds that the residual vector is determined by the fault source. By T₁For example, from the above analysis, when T is₁After the circuit is broken, the polarities of the three-phase terminal voltage distortion are respectively as follows: Δ u_ao<0，Δu_bo＝Δu_co0. Therefore, the differential voltage distortion has a polarity Δ u₁＝-Δu₃<0,Δu₂0. Because the first-order inertia link is formed between the distortion of the differential voltage and the residual error, if the inertia delay is ignored, r exists₁＝-r₃<0，r₂This orients the residual direction as 0: r is_n ^T＝[-1,0,1]V 2. In the same way, the equivalent differential voltage distortion and the characteristics of the normalized residual vector under different fault sources can be obtained, as shown in table 4.

As can be seen from this table, when a system fault is detected, the phase of the fault can be determined according to the normalized vector of the residual error, but it may be difficult to distinguish the switching tube from the winding fault in a short time. For example, when r is detected_n≈a₁ ^TWhen the fault source is the switch tube of the A-phase winding or the upper arm of the A-bridge arm. If the phase current i at this time_aNot equal to 0, the fault type can be judged to be A-phase winding turn-to-turn short circuit fault, otherwise, the A-phase winding is open circuit or T₁Open circuit failure. In this case, additional information is required to distinguish between a winding break or a switch tube open fault. One possible solution is to wait for or destroy T₁Residual error orientation condition corresponding to pipe open circuit fault even if system occurs i_a ^*<Please case 0. If the system detects a fault at the moment, the system can be judged to have a winding open-circuit fault, otherwise, the fault source is considered to be T₁. This will greatly reduce the speed of fault diagnosis due to the longer time to wait for the residual error oriented condition to naturally disappear. Thus, the condition may be quickly violated by changing the phase voltage reference.

TABLE 4 differential Voltage distortion law under different failure sources

Based on the above analysis, after a fault occurs, the magnitude of the residual vector will increase rapidly, and a phenomenon (residual orientation) in which the direction of the residual is fixed will occur. The fault diagnosis method shown in fig. 5 can be designed based on these two features. The method mainly comprises a differential current state observer, a low-pass filter, feature extraction, fault detection and positioning. The differential current state observer can calculate the feedback value and the estimated value of the differential current as shown in fig. 4. The residual vector r can be obtained by subtracting the feedback value from the measured value of the dynamic current. In order to further improve the anti-interference capability of the method, all components of the residual vector r enter the same low-pass filter in parallel, and the obtained residual vector r_LObtaining residual error module value M after module value operation and normalization operation_rAnd normalized residual vector r_nRespectively as the basis for fault detection and fault location. Wherein, the fault detection logic is:

F＝(sign(M_r-T_h)+1)/2

wherein F is a fault markVariable (F ═ 1 marks system fault, otherwise system health), T_hIs a fault detection threshold.

(3) Fault location algorithm

When F is 1, the fault location algorithm starts. If the switching tube on the A-phase bridge arm of the inverter fails or the A-phase winding of the motor fails, the residual vector r is normalized_nWill be oriented at a₁、a₂I.e. the residual direction at phase-A fault appears as r_n≈a₁Or a₂(ii) a The same way can obtain the residual error direction r when the B phase fault occurs_n≈a₃Or a₄And r in case of C-phase fault_n≈a₅Or a₆. Based on this characteristic, the actually obtained normalized vector can be directly compared with the template vector a₁～a₆And carrying out template matching according to a minimum distance principle, and carrying out fault phase positioning according to a matching result.

After the faulty phase is determined, the possible fault types are three types, switching tube fault, winding turn-to-turn short fault and winding open fault. In order to distinguish the fault types, unique fault characteristics under each fault type need to be further mined and utilized.

From the foregoing analysis, it can be seen that the directional condition of the residual vector under a single switching tube fault is unique, and the directional condition of the residual vector after a single winding fault is two and includes the directional condition under the switching tube fault in the same phase. Therefore, after the fault phase is determined, the current residual error orientation condition can be directly destroyed, and whether the system still has the residual error orientation phenomenon is detected, so that the fault of the switching tube and the fault of the winding are distinguished.

By switching tube T₁For example, when a fault occurs, when the system satisfies i_a ^*>And 0, the fault detection algorithm and the phase positioning algorithm can judge that the A-phase circuit of the system has faults in a short time. To further distinguish the fault types, i may be made in some way_a ^*<0 to break the previous residual orientation condition. Due to the fact that in i_a ^*<At 0, T₁The fault hardly affects the normal operation of the system, so that the fault phase indicating variable at this time is changed into 0 to be displayedThe system is fault-free. If the fault source occurs in the A phase winding of the motor, i is even at the moment_a ^*<0, the system still has residual error orientation phenomenon, and the fault phase indicator variable FP is 1. It follows that in order to distinguish between a switching tube fault and a winding fault, the key difficulty is to break the original residual error orientation condition.

In order to destroy the residual directional condition, the phase current expectation value i of the fault phase needs to be made_a ^*And (4) inverting.

According to the formula, the following can be obtained: if changes in the motor parameters are ignored, i_a ^*Phase voltage reference u only with faulted phase_an ^*And back electromotive force e_aIt is related. Considering that the motor is a large inertia system, the change of the motor speed and the angular position of the rotor is small in a short time when a fault is detected, namely, the back electromotive force is almost unchanged. Thus, the motor reference phase voltage u can be varied by varying the motor reference phase voltage u_an ^*To change i_a ^*Of (c) is used. Generally, to achieve high efficiency of the motor, the voltage drop across the resistance and inductance of the motor winding will be much less than the back emf across the winding, i.e., when u should be present_an ^*≈e_aAnd u is_an ^*>e_a>0. If at this time u is directly_an ^*Become negative, will result in i_a ^*<0, thereby destroying the original residual orientation condition.

In order to get the system out of the original orientation as soon as possible, i should be taken_a ^*A large negative value. However, an excessively large negative value will in turn greatly reduce the system torque, thereby causing a large speed fluctuation of the electric machine. Therefore, u_an ^*Values should be compromised. Simulation and experiment results show that when u is_an ^*When the residual error is modified according to the following formula, the time for the residual error to be separated from the original orientation is short, and the rotation speed fluctuation caused by the residual error is equivalent to the fluctuation amplitude caused by the fault:

(u_an ^*)_new＝-λu_an ^*

wherein, λ is an adjustment coefficient, and the effect is better when λ ∈ (0.25, 0.35).

In the carrier-based PWM mode, the phase voltage reference value is used as the input of the PWM module, so that the value can be assigned according to the above formula directly before the input, and the phase voltage reference value is changed. However, in the widely used control system of the SVPWM modulation scheme, the phase voltage reference value does not need to be calculated in the control algorithm, and therefore, the calculation of the phase point voltage reference value cannot be realized by the above formula, but the expected voltage value u in the α - β coordinate system can be changed_α ^*And u_β ^*To be implemented. This can be achieved indirectly by means of Clark transformation and its inverter:

in the second expression above, only the three-phase voltage reference values need to be updated (u)_xn ^*)_new＝-λu_xn ^*. Therefore, u_α ^*And u_β ^*The updating algorithm is as follows:

when the A phase fails:

when the B phase fails:

when the C phase fails:

because the first-order inertia link is formed between the voltage distortion and the residual vector, namely the original orientation is deduced from the change of the voltage reference value to the residual, and a certain response time exists. Therefore, in order to distinguish the winding fault from the switching tube fault, a certain waiting time needs to be reserved, and before the waiting time is finished, if the reverse residual error orientation phenomenon of the system is detected, the system is judged to be in a winding fault; otherwise, the fault of the inverter switch tube is judged. Considering the step response characteristic of the inertia link, the waiting time can be set to be 1-2 times of the inertia time constant. Once the fault distinguishing between the winding and the switching tube is completed, the change of the reference value of the phase voltage is immediately released, and further interference to the system operation is avoided.

After fault location is completed, the open-circuit fault and the short-circuit fault of the winding can be further distinguished. From the foregoing analysis, the difference characteristics after the two types of failures are mainly represented by the following two points:

1. after the winding is disconnected, the phase current of the fault phase is continuously zero; after the winding turns are short-circuited, the phase current of the fault phase is distorted, but the zero-crossing time is still short.

2. The equivalent voltage distortion after the winding is broken and the turn-to-turn short circuit is different.

The difference between the two points can be used as the basis for distinguishing turn-to-turn short circuit and open circuit faults of the winding. The fault distinguishing algorithm designed by the first difference is simpler, but the method can judge the small current as the zero current state, so that the turn-to-turn short circuit fault is easily judged as the open circuit fault by mistake under the condition of light load and stable speed (the phase current is smaller).

The second method estimates the equivalent voltage distortion mainly according to a quantitative expression of the equivalent voltage distortion, and estimates the residual error according to a discrete transfer function from the voltage distortion to the residual error. If the difference between the estimated value and the actual value of the residual error is always small (the relative difference is within 10%) within a certain time (within 1-2 electrical time constants), determining that the winding is in open circuit fault; otherwise, it is turn-to-turn short circuit fault.

(4) And (4) designing an adaptive threshold.

It is considered that model errors reduce the accuracy of the observer, which may lead to false detections and missed detections. In order to reduce the false detection rate and the missing detection rate, the invention designs a self-adaptive fault detection threshold.

When there is a model error, the actual differential current state equation is:

g 'and H' are the actual state transition matrix and the system control matrix, respectively. And (3) subtracting the actual differential current state equation with the state observer to obtain a residual error state equation with model errors:

wherein Δ G and Δ H are errors of the state transition matrix and the system control matrix, respectively, Δ G is G '-G, and Δ H is H' -H, and Z transformation is performed to obtain a residual error:

in the formula, R_j(z)、I_j(z)、And Δ U (z) is r_j、i_j、And Δ u_jCorresponding Z transformation, M_i、M_u、M_fFor corresponding Z transfer functions, M_i＝ΔG/(z-z_p)，M_u＝ΔH/(z-z_p)，M_f＝H’/(z-z_p). As can be seen from the equation, when there is a model error, the residual error is not only caused by a fault, etcThe effect difference is related to voltage distortion, and is also related to factors such as motor current, voltage command and the like. Because the motor current and voltage commands change along with the operation conditions of the motor, the amplitude of the residual vector under different working conditions has obvious difference. If the failure detection threshold is set to a fixed value, once the working conditions such as load or motor rotation speed are changed, errors or missed detection may occur. The conditions for realizing correct detection are as follows: threshold value M under normal conditions_r<T_hBut after a fault threshold M_r>T_h. Therefore, to avoid false detection, normal M should be analyzed first_rThe upper bound of (c).

Due to M_i、M_uAll the steps are first-order inertia links, if inertia lag is ignored, the residual vector amplitude M can be obtained by a residual Z transformation equation_rAn upper bound:

wherein m is₁＝(1-z_p)^-1·sum(|△G|)，m₂＝(1-z_p)^-1·sum(|△H|)，i,A differential current vector and an equivalent differential voltage vector, i ═ i₁,i₂,i₃]^T，| is the absolute operator, sum (-) is the summation of the elements in the matrix. With this upper bound, the threshold may be set to:

wherein, T₀The part of the threshold is a small normal number so as to avoid false detection caused by noise interference; however M_i、M_uAnd for residual errorThe filtering process performed by the filtering will be such that the threshold value T is_h1And a fault detection index M_rThere is a certain phase lag between them, which may also lead to false detections. To avoid false detection due to phase lag, T is required_h1And M_rPhase lag compensation is performed.

By introducing the self-adaptive fault detection threshold and the phase compensation method, the false detection rate of the system under various working conditions can be greatly reduced. However, after the fault occurs, the motor output torque will drop significantly. Under closed-loop control, the equivalent differential voltage amplitude will rise. Thus, after a fault occurs, the threshold T is adapted_h1It will appear that an upward trend will be present. T is_h1The fault diagnosis time and the missed detection rate will be increased at the same time.

In order to reduce the false detection rate and detection time, an adaptive threshold is required which is lowered by the occurrence of a fault. Therefore, the adaptive threshold needs to be redesigned, and the equivalent differential voltage can be obtained through the actual differential current state equation when the system is normalCan be calculated from the following equation:

wherein, the approximate conditions include H '≈ H ═ hE, G' ≈ G { -R {, G { -exp { -E {, and_sT_sv (L-M) } ≈ 1, bringing it into T_h1A new adaptive threshold is obtained from the equation:

T_h2＝T₀+m₁·||i||₂+m₃·||△i||₂

wherein Δ i (k) ═ Δ i (k +1) -i (k), m₃＝m₂H, parameter m₁And m₃The selection of (a) requires the acquisition of model errors (Δ G and Δ H), which are often difficult to acquire in real time. In practical applications, Δ G and Δ H are mainly caused by the variation of inductance and resistance parameters of the motor, and the variation range is usually not more than 30% and 50%. Therefore, the parameter m can be set in accordance with such an error upper limit₁And m₃The value of (c). For T₀The rated current of the motor can be set to 10% firstly, and then fine adjustment is carried out under the no-load condition.

When the system is normal, T_h1And T_h2Approximately equal and therefore also a low false detection rate can be achieved. In addition, the fault can cause the three-phase load of the motor to be asymmetric, so that the amplitude of the differential current of the motor is reduced, and therefore | | | i | | and | | | Δ i | | non-woven fabrics₂Will exhibit a downward trend after the failure has occurred.

Compared with the prior art, the invention has the advantages that:

1. the online fault diagnosis algorithm designed by the invention can run by depending on the information (rotating speed, angular position, three-phase current and voltage instruction) acquired by the permanent magnet synchronous motor control system, and the online fault diagnosis of the motor control system can be realized without an additional sensor.

2. The invention designs the self-adaptive fault threshold value for fault detection, thereby greatly improving the robustness and the detection accuracy of the algorithm.

3. The method can distinguish three fault types at one time in the running process of the motor, and can accurately judge that a certain path of the switch tube of the inverter is broken or a certain phase winding is broken or short-circuited.

Drawings

FIG. 1 shows T in the inverter₁The A phase current in the open circuit flows to the schematic diagram;

FIG. 2 is a diagram of the switching tube T₁A graph of terminal voltage and phase current distortion after disconnection;

FIG. 3 is a fault model for a winding turn-to-turn short circuit;

FIG. 4 is a signal flow diagram of signal pre-processing and an observer;

FIG. 5 is a block diagram of a fault diagnosis algorithm contemplated by the present invention;

FIG. 6 is a flow chart of a fault diagnosis algorithm.

Detailed Description

The invention is further described with reference to the following figures and detailed description.

Fig. 5 is a structural diagram of the present invention, which mainly includes a motor master control algorithm, a differential current state observer, a low-pass filter, a fault feature extraction module, a threshold calculation module, a fault diagnosis logic module, and six templates, and these six modules jointly operate to complete online fault diagnosis of a motor.

The algorithm flow diagram of the present invention is shown in fig. 6, where each step corresponds to one or more of the modules in fig. 5:

after the system is started, the whole algorithm starts:

first, an initial value is assigned. This step performs an initial assignment of all variables involved in the algorithm (all assigned to zero). The fault flag variable F is assigned to be zero to represent that the system is healthy, the fault location variable Loc is assigned to be zero to represent that a fault is not located temporarily, and the voltage reference value update flag variable New is zero to represent that the voltage reference value is not required to be modified currently.

Secondly, starting a motor master control algorithm to obtain three-phase current i of the motor_abcElectrical angle theta of motor_eAnd electrical angular velocity omega_eAnd obtaining the duty ratio d of the three-phase PWM wave according to a motor control algorithm_abc. Wherein i_abc(three-phase Current of the Motor, i_abc＝[i_a,i_b,i_c]^T)、θ_eAnd omega_eThe feedback information of the motor control system can realize data acquisition through the original sensor and the AD converter of the control system; and d_abcDuty ratio command for the output of the master control algorithm, d_abc＝[d_a,d_b,d_c]^T。

And thirdly, preprocessing the data. This step is performed in the differential current state observer module of fig. 5, the specific algorithm of which corresponds to fig. 4, the task of which is to calculate the variables required by the current observer, the steps of which are to calculate the equivalent machine terminal voltage reference u_ao ^*、u_bo ^*、u_co ^*And the back electromotive force e of the motor_a、e_b、e_cThen, the differential voltage and the differential current are calculated.

Wherein, the calculation formula of the terminal voltage reference value is as follows:

u_xo ^*＝d_xU_dc，x＝a，b，c

in order to reduce the influence of the inverter nonlinearity on the current estimation accuracy, the obtained terminal voltage reference value can be corrected by the following correction quantity:

the back electromotive force calculation can firstly measure the waveform of the back electromotive force of the motor under the unit rotating speed under the line and store the waveform into the memory in the form of a table. When the system runs, the back electromotive force waveform table can be searched according to the current electrical angle and the memory, the back electromotive force of the motor at the unit rotating speed is obtained, and then the back electromotive force is multiplied by the current rotating speed.

The differential current and differential voltage calculation equations are then as follows:

[i₁,i₂,i₃]^T＝[i_a-i_b,i_b-i_c,i_c-i_a]^T

and fourthly, estimating the differential current. This step corresponds to the differential current observer of fig. 5, whose core algorithm is:

wherein the parameter L_rThe selection principle of the method is already explained in the invention principle and is not repeated.

And fifthly, residual calculation and filtering. The step mainly corresponds to the low-pass filter module in fig. 5, and is responsible for subtracting the differential current and the estimated value thereof obtained in the steps III and IV to obtain a residual r, and then filtering each component of the obtained residual by using a digital low-pass filter, thereby increasing the anti-interference capability of the subsequent fault diagnosis algorithm.

Sixthly, calculating the modulus M of the residual error_rAnd direction r_n. This step corresponds to the threshold calculation module in fig. 5, and its algorithm is:

seventhly, calculating a threshold value T_h. This step is accomplished in the fault diagnosis logic module of fig. 5. The threshold calculation formula is:

T_h＝T₀+m₁·||i||₂+m₃·||△i||₂

in the formula, T₀、m₁、m₃The values of the coefficients are shown in the description of the invention.

And eighthly, fault detection. The step is completed in the fault diagnosis logic module in fig. 5, and is responsible for calculating the residual modulus M obtained in the step (c)_rAnd step (c) the obtained adaptive threshold value T_hAnd (3) comparing the sizes, and judging whether the system has a fault according to a comparison result: if M is_r>T_hIf so, judging the system fault (making the fault flag variable F equal to 1), and performing subsequent fault positioning; otherwise, the system is considered to be fault-free (the fault flag variable F is made to be 0).

And ninthly, matching the template and positioning the fault phase. This step is performed in the fault diagnosis logic module in fig. 5, and can be performed in two steps: firstly, according to the principle of minimum distance, the residual error direction vector r is divided into_nAnd six template vectors a₁～a₆And matching to obtain a template serial number i with the highest matching degree, and determining the phase of the fault according to the value of i.

The matching algorithm is as follows:

d_j＝||r_n-a_k||，d_i＝min(d_k)，k＝1,2,……,6

wherein the template vector is:

and after the value i is obtained, judging a fault phase: if i is 1 or 2, judging the fault phase as A phase; if i is 3 or 4, the fault phase is judged to be the B phase; and if i is 5 or 6, determining that the fault phase is the C phase.

And judging the fault type in the (R). This step is used in the fault diagnosis logic block of fig. 5 to distinguish switching tube open circuit faults, winding open circuit faults and winding turn-to-turn short circuit faults.

Firstly, whether the fault is a winding short-circuit fault is judged. And after fault phase separation is finished, residual error matching is carried out according to the size of a residual error component in a fault period so as to distinguish the winding turn-to-turn short circuit fault and the winding/switching tube open circuit fault. The method comprises the following specific steps:

firstly, selecting a residual error component to be calculated according to a phase x where a fault is located: when the failure phase x is a, b, c, j is 1, 2, 3, respectively.

Then calculating the equivalent voltage distortion component delta u corresponding to the open circuit fault of the winding/switching tube_jAnd the equivalent voltage distortion component calculation formula is as follows:

in the formula, when x is a fault phase and x is a, b and c, j is 1, 2 and 3 respectively.

Then, according to the state equation of the residual error, the residual error component r is obtained_jEstimated value of (a):

and finally, residual error size matching: estimating residual componentsWith its actual value r_jComparing, and if the difference between the estimated value of the residual error and the actual value is small (the relative difference is within 10%) within a certain time (the electrical period of 1-2 motors), classifying the current fault as a winding open circuit fault or a switch tube open circuit fault; otherwise, it is determined as inter-turn short circuitAnd (4) a barrier.

If an open circuit fault is determined, it is necessary to further determine whether the fault is located on the switching tube or the winding. At this time, the voltage reference value corresponding to the fault is forcibly modified. If a new i value is output in the step ninthly within a certain time (the electrical time constant of 1-2 motors), judging that the system has a winding open circuit fault; otherwise, the inverter switch tube T is judged_iAn open circuit fault occurs. And after the fault type is determined, removing the voltage instruction to modify the related instruction.

The above-described manner of changing the phase voltage command is related to the PWM modulation method of the motor, and in the motor control system using the carrier-based PWM modulation method, the phase voltage command changing formula is:

(u_xn ^*)_new＝-λu_xn ^*

in the formula, x is a fault phase, x belongs to { a, b, c }, subscript new represents an updated value, and λ is a constant between 0.25 and 0.35.

For a system adopting an SVPWM modulation mode, the voltage reference value u of the motor under an alpha-beta coordinate system can be changed_α ^*、u_β ^*To indirectly change the phase voltage reference:

when the A phase fails:

when the B phase fails:

when the C phase fails:

those skilled in the art will appreciate that the details of the present invention not described in detail herein are well within the skill of those in the art.