阅读说明：本技术 一种基于改进最大似然投票算法（mlv）的速度传感器容错控制（ftc）策略 (Speed sensor fault-tolerant control (FTC) strategy based on improved maximum likelihood voting algorithm (MLV) ) 是由 袁宇浩 淡宁 沈谋全 于 2019-09-27 设计创作，主要内容包括：本发明公开了一种基于改进最大似然投票算法(MLV)的速度传感器容错控制(FTC)策略,包括扩展卡尔曼滤波器(EKF),利用数据融合技术处理转速信息以改进最大似然投票算法(MLV),在速度传感器故障时,能够自然完成故障隔离并重新配置系统结构。所述扩展卡尔曼滤波器(EKF)估计电机转速包括对状态矢量和误差协方差矩阵的预测、利用实测输出和预测输出对预测状态矢量和误差协方差矩阵进行校正；所述利用数据融合技术处理转速信息以改进最大似然投票算法(MLV)包括数据融合算法、最大似然投票算法(MLV)；所述在速度传感器故障时,能够自然完成故障隔离并重新配置系统结构包括可靠系数的确定；仿真结果证明,本发明在速度传感器故障时仍能维持PMSM高性能不间断运行。(The invention discloses a speed sensor fault-tolerant control (FTC) strategy based on an improved maximum likelihood voting algorithm (MLV), which comprises an Extended Kalman Filter (EKF), wherein the EKF utilizes a data fusion technology to process rotating speed information so as to improve the maximum likelihood voting algorithm (MLV), and when the speed sensor fails, the fault isolation can be naturally completed and a system structure can be reconfigured. The Extended Kalman Filter (EKF) estimates the rotating speed of the motor, and comprises the steps of predicting a state vector and an error covariance matrix, and correcting the predicted state vector and the error covariance matrix by using actual measurement output and prediction output; the method for processing the rotating speed information by using the data fusion technology to improve the maximum likelihood voting algorithm (MLV) comprises a data fusion algorithm and a maximum likelihood voting algorithm (MLV); when the speed sensor fails, the fault isolation can be naturally completed, and the system structure can be reconfigured, including the determination of the reliability coefficient; simulation results prove that the invention can still maintain the high-performance uninterrupted operation of the PMSM when the speed sensor fails.)

1. A speed sensor Fault Tolerant Control (FTC) strategy based on an improved maximum likelihood voting algorithm (MLV), characterized by: the method comprises an Extended Kalman Filter (EKF), utilizes a data fusion technology to process rotating speed information so as to improve a maximum likelihood voting algorithm (MLV), and can naturally complete fault isolation and reconfigure a system structure when a speed sensor fails.

The Extended Kalman Filter (EKF) includes a prediction of a state vector and an error covariance matrix, the predicted state vector and error covariance matrix being corrected using the measured output and the predicted output; the method for processing the rotating speed information by using the data fusion technology to improve the maximum likelihood voting algorithm (MLV) comprises a data fusion algorithm and a maximum likelihood voting algorithm (MLV); when the speed sensor fails, the fault isolation can be naturally completed, and the system structure can be reconfigured, including the determination of the reliability coefficient;

The data fusion algorithm is used for inputting MLV;

The maximum likelihood voting algorithm (MLV) is used to handle possible speed sensor failures;

The MLV is used for calculating the probability value of the correct result by utilizing the reliability of the input data of each algorithm, and selecting the input with the maximum probability value as the final output;

The determination of the reliability factor is used to ensure the correct output of the MLV.

2. A speed sensor fault-tolerant control (FTC) strategy based on an improved maximum likelihood voting algorithm (MLV) according to claim 1, characterized in that: the MLV can process the fault problem of the speed sensor only by inputting three speed modules, and under the existing condition, the EKF and the sensor can provide speed information and cannot realize fault isolation of a system, so a data fusion algorithm is additionally added.

3. A speed sensor fault-tolerant control (FTC) strategy based on an improved maximum likelihood voting algorithm (MLV) according to claim 1, characterized in that: the data fusion algorithm is a weighted average algorithm, the accuracy of the sensor is sacrificed when the sensor is normal, and the data fusion algorithm is combined with the MLV, so that the accuracy loss can be effectively avoided.

4. A speed sensor fault-tolerant control (FTC) strategy based on an improved maximum likelihood voting algorithm (MLV) according to claim 1, characterized in that: the MLV selects a sensor with the highest reliability as an output, but when the speed sensor is completely failed or has intermittent faults (the measured value is 0), the output value of the data fusion algorithm is the same as the estimated value of the EKF, the MLV can automatically isolate the sensor faults at the moment, and the data fusion algorithm is selected as the output of the MLV due to higher reliability, so that the natural reconstruction of the rotating speed is realized.

5. A speed sensor fault-tolerant control (FTC) strategy based on an improved maximum likelihood voting algorithm (MLV) according to claim 1, characterized in that: the reliability of the data fusion algorithm is slightly higher than that of the EKF algorithm due to the fact that the data fusion algorithm contains partial sensor information, only the fault that the output of a speed sensor is 0 is considered, under the fault, the speed output by the data fusion algorithm is the estimated rotating speed of the EKF, and therefore the reliability coefficient determining method guarantees the correct output of the MLV.

Technical Field

The invention relates to the technical field of a Permanent Magnet Synchronous Motor (PMSM) vector control system, in particular to a speed sensor fault-tolerant control (FTC) strategy technology based on an improved maximum likelihood voting algorithm (MLV).

Background

in the field of industrial control requiring variable speed driving, a Permanent Magnet Synchronous Motor (PMSM) is favored due to its advantages of high efficiency and high power density. In many practical applications, such as electric locomotives and the like, it is still necessary to maintain satisfactory operating performance even in the event of a failure of the drive system, in order to avoid a threat to the personal and property safety of the operator. Therefore, it is important that the drive control system itself be capable of fault isolation and reconfiguration. For a vector control system of a permanent magnet synchronous motor, a voltage sensor, a two-phase current sensor and a speed sensor are generally required to be equipped. The speed sensor is influenced by the installation position, the electromagnetic environment and the application working condition of the motor, and the problems of complete failure, intermittence or precision reduction and the like easily occur. As an important link of PMSM feedback control, a failure of the speed sensor inevitably reduces system performance, even causes instability of the system. Therefore, many scholars pay more attention to the fault tolerance research on the speed sensor fault problem and achieve a series of results.

An extended Kalman filter is designed herein to perform real-time estimation of PMSM rotation speed and is used as an input to MLV. The data fusion technology is used for improving the MLV, the reliability determination method is simplified, and meanwhile, after a speed sensor breaks down, only one rotating speed observer is needed to complete fault isolation and system reconfiguration, so that the system is smoothly switched from a control mode with a sensor to a vector control mode without the sensor. The simulation result proves the effectiveness of the method for the fault-tolerant control of the speed sensor.

Disclosure of Invention

in view of the above technical problems, the present invention aims to provide a speed sensor fault-tolerant control (FTC) strategy technique based on an improved maximum likelihood voting algorithm (MLV).

In order to achieve the purpose, the invention adopts the following technical scheme:

A speed sensor Fault Tolerant Control (FTC) strategy based on an improved maximum likelihood voting algorithm (MLV), characterized by: the method comprises an Extended Kalman Filter (EKF), utilizes a data fusion technology to process rotating speed information so as to improve a maximum likelihood voting algorithm (MLV), and can naturally complete fault isolation and reconfigure a system structure when a speed sensor fails.

the Extended Kalman Filter (EKF) includes a prediction of a state vector and an error covariance matrix, the predicted state vector and error covariance matrix being corrected using the measured output and the predicted output; the method for processing the rotating speed information by using the data fusion technology to improve the maximum likelihood voting algorithm (MLV) comprises a data fusion algorithm and a maximum likelihood voting algorithm (MLV); when the speed sensor fails, the fault isolation can be naturally completed, and the system structure can be reconfigured, including the determination of the reliability coefficient;

The data fusion algorithm is used for inputting MLV;

The maximum likelihood voting algorithm (MLV) is used to handle possible speed sensor failures;

The MLV is used for calculating the probability value of the correct result by utilizing the reliability of the input data of each algorithm, and selecting the input with the maximum probability value as the final output;

The determination of the reliability factor is used to ensure the correct output of the MLV.

Further, the speed sensor fault-tolerant control (FTC) strategy based on the improved maximum likelihood voting algorithm (MLV) includes that the data fusion algorithm is a weighted average algorithm, the accuracy of the sensor is sacrificed when the sensor is normal, and the accuracy loss can be effectively avoided by combining the data fusion algorithm with the MLV.

Further, the speed sensor fault-tolerant control (FTC) strategy based on the improved maximum likelihood voting algorithm (MLV) comprises that the MLV selects the sensor with the highest reliability as an output. However, when the speed sensor is completely failed or intermittently failed (the measured value is 0), the output value of the data fusion algorithm is the same as the estimated value of the EKF, the MLV can automatically isolate the sensor failure, and the data fusion algorithm is selected as the output of the MLV due to higher reliability, so that the natural reconstruction of the rotating speed is realized.

Further, the speed sensor fault-tolerant control (FTC) strategy based on the improved maximum likelihood voting algorithm (MLV) comprises that the data fusion algorithm has a slightly higher reliability than the EKF algorithm due to the fact that the data fusion algorithm contains partial sensor information. Only the fault that the output of the speed sensor is 0 is considered, under the fault, the speed output by the data fusion algorithm is the estimated rotating speed of the EKF, and therefore the reliability coefficient determination method ensures the correct output of the MLV.

the invention has the beneficial effects that:

The MLV is improved by the data fusion algorithm, so that the MLV can successfully realize automatic isolation of sensor faults only by one speed estimation module, the system design is simplified, and meanwhile, the MLV can output expected rotating speed no matter in the fault state or the normal state of the speed sensor, so that the natural reconstruction of the rotating speed is completed. On one hand, the real-time estimation of the motor rotating speed is completed by utilizing the EKF and the advantage of robustness of the EKF to system random noise. On the other hand, the measured rotating speed and the estimated rotating speed are weighted based on the data fusion technology and are used for improving the MLV, the problem that the MLV algorithm can process sensor faults only by relying on a multi-observer module is solved, occupation of software resources is reduced, and determination of the MLV reliability coefficient is simplified. The faults of the speed sensor can be isolated in time, meanwhile, the uninterrupted operation of the system is ensured through information reconstruction, and the reliability of the permanent magnet synchronous motor driving system is further improved.

Drawings

FIG. 1 is a PMSM fault tolerant vector control diagram of the present invention; FIG. 2 is a diagram of EKF-based state estimation; FIG. 3 is a graph of the output of the improved MLV; FIG. 4 is a rotational speed response after fault reconstruction;

Wherein, in fig. 1, a modified maximum likelihood voting algorithm (MLV); 2. an Extended Kalman Filter (EKF); 3. space Vector Pulse Width Modulation (SVPWM); 4. an inverter; 5. permanent Magnet Synchronous Machines (PMSM).

Detailed Description

The invention is further described with reference to the accompanying drawings and the detailed description below:

referring to fig. 1, a speed sensor Fault Tolerant Control (FTC) strategy based on an improved Maximum Likelihood Voting (MLV) of the present embodiment is applied to improve reliability of a Permanent Magnet Synchronous Motor (PMSM) vector control system, and mainly includes: improved maximum likelihood voting algorithm (MLV) 1; an Extended Kalman Filter (EKF) 2; space Vector Pulse Width Modulation (SVPWM) 3; an inverter 4; a Permanent Magnet Synchronous Motor (PMSM) 5. An Extended Kalman Filter (EKF)2 performs a real-time estimation of PMSM5 rotational speed and is an input to MLV 1.

The data fusion technology is used for improving the MLV1, the reliability determination method is simplified, and meanwhile, after a speed sensor fails, only one rotating speed observer is needed to complete fault isolation and system reconfiguration, so that the system is smoothly switched from a control mode with a sensor to a vector control mode without the sensor. In order to reduce the calculation burden and improve the state estimation precision, the surface-mounted permanent magnet synchronous motor 5 is taken as a research object, and the equation of the surface-mounted permanent magnet synchronous motor in an alpha-beta static coordinate system is established as

In the formula: i.e. i_α，i_β，u_α，u_βThe components of the stator current and the stator voltage on an alpha beta axis are respectively; psi_fIs a permanent magnet flux linkage, R, L_sStator resistance and inductance; theta_e，ω_ethe electrical angular position and the angular velocity of the rotor.

In view of

the nonlinear state of the motor 5 can be modeled as

y(t)＝Cx(t)+v(t) (4)

in the formula: x (t) ═ i_α i_β ω_e θ_e]^T，u＝[u_α u_β]^T，y(t)＝[i_α i_β]^T；

w (t), v (t) are system noise and measurement noise, which are uncorrelated zero-mean white noise, respectively, and the EKF2 algorithm performs recursive computation by using their covariance matrices Q and R.

The gradient matrices F and H of the system are as follows:

The recursive algorithm design of EKF2 mainly comprises two steps in a discrete state. The first step is the prediction of the state vector and the error covariance matrix, i.e.:

the second step is to correct the predicted state vector and the error covariance matrix using the measured output and the predicted output, i.e.:

K_k＝P_k|k-1H^T(HP_k|k-1H^T+R)^-1 (9)

Wherein: k_kIs a kalman gain matrix.

an inaccurate majority voting algorithm and a weighted average algorithm are commonly used in fault-tolerant control systems based on a static redundancy analysis method, but both algorithms have inherent disadvantages. An inaccurate majority voting algorithm has difficulty determining a suitable threshold, while a weighted average algorithm sacrifices the accuracy of the measurement information in the case of a properly functioning system. Therefore, to avoid the above problems, the present invention employs the MLV1 algorithm to address possible speed sensor failures. MLV1 operates on the principle of computing the probability values for correct results using the reliability of the input data for each algorithm and selecting the input with the highest probability value as the final output. The probability value of each input algorithm as the correct result is

Considering the actual PMSM5 servo system, the measured and estimated rotational speeds may not be exactly the same at the same time, so Δ in equation (14) must be considered_k(i) Make a correction

In the formula: n is the input algorithm number; r is_ithe reliability coefficient of each input algorithm; x is the number of_iSample space for each algorithm input; d_maxIs a set threshold. The purpose of the thresholds set here is to consider the algorithm inputs equal within the threshold range, thus ensuring that MLV1 can output the sensor's measurement information in the absence of a fault.

Since the EKF2 does not require the measurement value of the speed sensor when estimating the state, the system reconstruction can be realized by using the rotation speed and position information estimated by the EKF2 when the speed sensor fails. The MLV1 must have three speed module inputs to handle the speed sensor fault problem, but under the existing conditions, only the EKF2 and the sensors can provide speed information, and fault isolation of the system cannot be achieved. The MLV1 design is improved herein in view of the fact that the additional design of a speed observer occupies more software resources, which greatly increases the system load. In addition to the velocity measurements and estimates, a data fusion algorithm is added as a third input to MLV1, the mathematical expression of which is

In the formula: omega_r、θ_rSpeed and position information, ω, output for the data fusion algorithm,the rotational speed was measured for the sensor and estimated for EKF2, respectively. The essence of data fusion is a weighted average algorithm, which can effectively avoid the problem by combining with MLV1, which sacrifices the accuracy of the sensor when the sensor is normal. This is because when the sensors are normal, the algorithm inputs are approximately equal within the threshold range, so MLV1 will select the sensor with the highest reliability as the output. However, when the speed sensor is completely failed or intermittently failed (the measured value is 0), the output value of the data fusion algorithm is the same as the estimated value of the EKF2, the MLV1 automatically isolates the sensor failure, and the data fusion algorithm is selected as the output of the MLV1 due to higher reliability, so that the natural reconstruction of the rotating speed is realized.

The data fusion algorithm is used for improvement of MLV1, and a determination method of the reliability coefficient of each input is simplified. If the estimated speeds of the two observers are used as the input of MLV1, the speed estimation effect of the observers must be considered in combination when determining the reliability coefficient, which now only needs to be determined in the normal state of the system. Under normal conditions, the sensors are generally considered to have the highest reliability, and the data fusion algorithm has slightly higher reliability than the EKF2 algorithm because the data fusion algorithm contains partial sensor information. The method only considers the fault that the output of the speed sensor is 0, and under the fault, the speed output by the data fusion algorithm is the estimated rotating speed of the EKF2, so the reliability coefficient determination method ensures the correct output of the MLV 1.

The Matlab/simulink simulation is utilized to verify the effectiveness of the proposed speed sensor fault-tolerant control method based on the improved MLV 1. Fig. 1 is a block diagram of a vector control of a permanent magnet drive system.The PMSM5 parameters used were: stator resistance R2.875 Ω, stator inductance L_s＝8.5e^-3H, permanent magnetic linkage psi_f0.3Wb, and a moment of inertia J of 0.001kg m²the number of pole pairs P is 4, and the coefficient of viscous friction is 0. To demonstrate the effectiveness of EKF2, FIG. 2 shows the estimates and errors for each state quantity under different operating conditions. As can be seen from the figure, the estimated quantity can quickly approach the actual quantity no matter the motor is in stable operation or in an acceleration and deceleration state. For ease of observation, only a partial map of the position and current estimates is given in (a) and for this reason, a full range of state estimation errors is given in (b) to demonstrate the superior performance of EKF2, which lays the foundation for fault isolation and system reconstruction afterwards. The initial value of the EKF2 recursive calculation is randomly given, which makes it have a certain estimation error in the low speed section, so the threshold D of MLV1 is set in the low speed section_maxIt must be made small enough otherwise MLV1 will fail. The invention sets D at the low-speed stage_maxat 5rpm, in the medium-high speed stage, D is set_maxThe reliability coefficients of the sensor, data fusion, and EKF2 algorithms were 0.99, 0.96, and 0.92, respectively, at 30 rpm.

In order to verify the fault-tolerant feasibility of the improved MLV in the PMSM system, the working conditions are set as follows: the expected rotation speed of the system is 600rpm, and the speed sensor fails within 0.2-0.3s and 0.4-0.6 s. Fig. 3 depicts the real-time probability values of each input algorithm of the improved MLV1, and the system outputs corresponding speed information to complete the feedback control of the PMSM 5. It can be seen that the MLV1 is improved by the data fusion algorithm, so that the MLV1 can successfully realize automatic isolation of sensor faults only by one speed estimation module, the system design is simplified, and meanwhile, the MLV1 can output expected rotating speed no matter in a fault state or a normal state of the speed sensor, thereby completing natural reconstruction of the rotating speed. Fig. 4 shows a rotation speed response diagram after system reconfiguration, and it can be seen that the improved MLV1 still maintains the excellent performance of the static redundancy fault-tolerant method, the transition process of the system from the control with the sensor to the control without the speed sensor is very stable, no large oscillation phenomenon occurs, and the rotation speed of the motor can quickly reach the expected value.

the above-mentioned embodiments are not intended to limit the present invention, and it will be apparent to those skilled in the art that other modifications may be made based on the above-mentioned embodiments and concepts, and any modifications within the spirit and principle of the present invention shall be covered by the scope of the present invention.