CN111469128B - A method for separating and extracting current-coupling signals of an articulated robot - Google Patents

Abstract

The invention discloses a method for separating and extracting current coupling signals of an articulated robot, which belongs to the field of industrial robot state monitoring and fault diagnosis. The method firstly obtains the motor encoder values of the robot to be studied under the single-joint motion state and the double-joint linkage state of the robot. Feedback the pulse signal and calculate the real-time rotation angle; then obtain the joint driving torque based on the dynamic simulation model; obtain the winding current through the relationship between the driving torque and the current, and judge whether there is a coupling effect based on this; finally obtain the robot current signal, and extract it after filtering. Its envelope can realize the coupling and separation of the robot joint current signal; this method has no requirements on the noise level of the robot joint current, and can filter the noise of the joint current signal in any operating environment to obtain a smooth current signal, and then package the current signal. The network extraction can realize the separation of the current signal coupling action part and the coupling irrelevant part of the joint robot under any operating environment.

Description

A method for separating and extracting current-coupling signals of an articulated robot

technical field

The invention relates to a method for separating and extracting a current coupling signal of an articulated robot, and belongs to the technical field of industrial robot state monitoring and fault diagnosis.

Background technique

Six degrees of freedom serial industrial robots are widely used in industrial automation production. The wear of parts caused by long-term operation and the occurrence of some emergencies may cause the robot to suddenly stop running, thereby destroying the operation of the production line and causing losses to the enterprise. Therefore, it is particularly important to timely respond to some characteristics of the robot before it stops and make a response plan in advance, that is, to monitor the real-time running state of the robot. The main devices of robot joints are servo motors and reducers, so the monitoring of robots can learn from each other with the traditional state monitoring methods of motors and reducers. The traditional motor state monitoring uses its vibration signal or current signal, but for a six-degree-of-freedom series robot, compared with the current signal, the vibration signal has the problems of difficult acquisition and high cost, and the current signal can be obtained from the robot cabinet. It can be obtained directly inside, which is convenient for collection and low cost.

The use of current signals for fault diagnosis of gears, motors and other mechanical devices has achieved good results, and for robots, the use of current signals for robot state monitoring has many advantages, but the joints of series robots are connected together. , there is a coupling phenomenon between its joints, that is, except for the end motor, the current signals of other motors do not only express the state of the articulated arm corresponding to the motor, but include the running state of all articulated arms connected behind the motor, so it needs to be released. Only by the coupling effect in the joint current signal can the robot current signal be used to accurately monitor the running state of each joint arm of the robot; the coupling-related part of the original motor current signal is separated and extracted, and the extracted part is used for subsequent follow-up The current signal decoupling analysis of the robot realizes the decoupling of the current signal of the joint robot, that is, the current signal of each joint only contains the state information of the current joint, which is of great significance to the state detection of the robot joint using the current signal.

SUMMARY OF THE INVENTION

In view of the problems existing in the prior art, the present invention provides a method for separating and extracting the current coupling signal of an articulated robot. The method has no requirements on the noise of the robot joint current, and can filter the current signal of the articulated robot in any operating environment. Process and filter out noise to obtain smooth current signal, and realize the separation of the coupling action part and coupling irrelevant part of the current signal of the joint robot under any operating environment.

The specific steps of the method for separating and extracting the current coupling signal of the articulated robot of the present invention are as follows:

(1) When moving the angle θ in the single-joint motion state, collect the feedback pulse signals of the motor encoder of the i-th joint of the robot at different time points, and use the motor encoder feedback pulse signals to calculate the robot's i-th joint at different time points. Real-time rotation angle data; use ADAMS software to build a dynamic simulation model, and substitute the real-time rotation angle data into the dynamic simulation model to obtain the driving torque F ₁ of the i-th joint at different time points;

When the joint i and the i+1 joint move at an angle of θ in the double-joint linkage state, collect the motor encoder feedback pulse signals of the robot joint i and i+1 joint at different time points, and use the motor encoder feedback pulse The signal calculates the real-time rotation angle data of the i-th joint and the i+1-th joint of the robot at different time points; uses the ADAMS software to build a dynamic simulation model, and substitutes the real-time rotation angle data of the i-th joint and the i+1-th joint into the dynamic simulation model to obtain the driving torque F ₂ of the i-th joint at different time points;

The θ angle variation range of the joint motion does not exceed the maximum angle that the robot joint can rotate;

The use of ADAMS software to build a dynamic simulation model is a conventional method, for example, refer to the literature "Liu Peisen, Jin Xingzi, etc. Modeling and Dynamics Simulation of Industrial Robots Based on ADAMS [J]. Journal of Chengdu Institute of Technology, 2018, 21(4) :10-13,59." method construction;

The calculation of the real-time rotation angle is based on the official data of the motor and driver company used by the robot to find the total number of pulses m per revolution of the motor encoder code disk, and calculate the angle α=360°/m corresponding to each pulse; Matlab The software calculates the real-time pulse number corresponding to the feedback pulse signal of the motor encoder at different time points, then the real-time rotation angle = real-time pulse number × α;

(2) Calculate the winding current I ₁ and the winding current I ₂ under the single-joint motion state and the double-joint linkage state when the i-th joint moves at an angle of θ by the following formulas, and obtain several winding currents I ₁ and winding currents I ₂ data;

其中K_t为转矩常数，驱动力矩F₁或驱动力矩F₂＝T_e×第i关节减速器的减速比；formula

Wherein K _t is the torque constant, the driving torque F ₁ or the driving torque F ₂ =T _e × reduction ratio of the ith joint reducer;

(3) Taking the time of step (1) as the abscissa and the winding current as the ordinate, draw the winding current I ₁ and winding current I ₂ curves of the i-th joint in the single-joint motion state and the double-joint linkage state, and compare the two Whether the two curves overlap, when the two curves overlap, the current of the i-th joint has no coupling effect in the double-joint linkage state;

(4) When the two curves do not overlap, the current of the i-th joint has a coupling effect in the double-joint linkage state; then when the single-joint motion state and the double-joint linkage state move the angle θ, the current sensor is used to obtain the first joint. Current signal data x ₁ (t) and x ₂ (t) of joint i in the single-joint motion state and the double-joint linkage state, where t is the acquisition time;

(5) The combined filtering method of zero-phase filtering combined with singular value denoising is used to filter the current signal to obtain a filtered signal;

(6) Substitute the filtered signal into the formula z(t)=x(t)+jH[x(t)] to obtain the analytical signal of the current signal, where x(t) is the real part of the complex signal z(t) , H[x(t)] is the imaginary part of the complex signal z(t), j is the imaginary unit;

得到滤波后信号在单关节运动状态下和双关节联动状态下的电流包络A₁(t)和电流包络A₂(t)，实现关节机器人电流信号耦合作用部分与和耦合无关部分的分离。Substitute H[x(t)] and x(t) into the formula

Obtain the current envelope A ₁ (t) and the current envelope A ₂ (t) of the filtered signal in the single-joint motion state and the double-joint linkage state, and realize the separation of the joint robot current signal coupling part and the coupling-independent part .

The above-mentioned combined filtering method using zero-phase filtering combined with singular value denoising to filter the current signal is a conventional method, and the details are as follows:

(1) Filter the current signals x ₁ (t) and x ₂ (t) through a zero-phase filter to obtain a filtered signal;

(2) Perform short-time Fourier transform on the filtered signal to obtain a time-frequency matrix A, where the time-frequency matrix is defined as each row represents a frequency point, each column represents a time point, and each value represents a certain frequency at the time of acquisition The magnitude of ;

(3) Substitute the time-frequency matrix A obtained in step (2) into the formula A=UΣV for singular value decomposition, and extract the value of the diagonal matrix Σ to obtain the singular value sequence d, where U and V are both orthogonal matrices, and Σ is the pair corner array;

(4) Using the singular value sequence d obtained in step (3), subtract the previous value from the latter value in turn to obtain a new sequence. When the quotient of the first value and a later value in the new sequence reaches 20, select Any value in the change interval corresponding to the difference value is taken as the threshold value, and the singular value smaller than the threshold value is taken as zero to form a new singular value sequence q;

(5) Use the new singular value sequence q to construct a new diagonal matrix Σ′, and substitute Σ′ into the formula A′=UΣ′V′ to obtain a new coefficient matrix A′;

(6) Perform inverse short-time Fourier transform on the coefficient matrix A' obtained in step (5) to obtain the final filtered signal.

The beneficial effects of the present invention are:

1. The present invention essentially seeks out the coupling mechanism of the joint current signal of the industrial robot, and combines the robot dynamics simulation model to obtain the essential embodiment of the coupling effect in the joint current signal of the robot;

2. The method of the present invention has no requirement on the noise of the robot joint current, and can filter the noise of the joint robot current signal in any operating environment to obtain a smooth current signal, and then perform envelope extraction to achieve any operating environment. The separation of the joint robot current signal coupling action part and the coupling irrelevant part.

3. The extracted envelope can be used for the decoupling analysis of the robot current signal, to release the joint coupling effect in the robot current signal, so as to realize the robot joint state detection by using the current signal.

Description of drawings

Figure 1 is a schematic diagram of the structure of the robot;

FIG. 2 is a schematic diagram of the winding current I ₁ and the winding current I ₂ of the second joint in a single-joint motion state and a double-joint linkage state;

3 is a schematic diagram of the original waveform of the current signal x ₁ (t) of the second joint in a single-joint motion state;

Figure 4 is a schematic diagram of the original waveform of the current signal x ₂ (t) of the second and third joints in the double-joint linkage state;

5 is a schematic diagram of the waveform of the current signal filtered by the second joint in a single-joint motion state;

Fig. 6 is the waveform schematic diagram after filtering the current signal of the second joint under the state of double joint linkage;

7 is a schematic diagram of the envelope of the current signal of the second joint in a single-joint motion state;

8 is a schematic diagram of the envelope of the current signal of the second joint in the state of double joint linkage;

In Figure 1: 1-joint 1, 2-joint 2, 3-joint arm I, 4-joint 3, 5-joint 4, 6-joint arm II, 7-joint 5, 8-joint 6 Joint, 9-electric cabinet, 10-current sensor.

Detailed ways

The present invention will be further described in detail below through the examples, but the protection scope of the present invention is not limited to the content.

Embodiment 1: The current coupling signal separation and extraction method of the articulated robot is as follows:

The Qianjiang QJR6-1 welding robot was used, and the SolidWorks three-dimensional model was established according to the official data of Qianjiang Robot Company. ,2018,21(4):10-13,59." The three-dimensional model is imported into the ADAMS software to build a dynamic simulation model. Figure 1 is a schematic diagram of the robot structure, which includes the first joint 1, the second joint Joint 2, connecting arm I3, joint 3 4, joint 4 5, connecting arm Ⅱ6, joint 5 7, joint 6 8, electric cabinet 9, current sensor 10; the implementation objects are joint 2 and joint 3, The movement angle θ is rotated by 90°. The specific operation process is as follows:

1. In the state of single-joint motion of the second joint, collect the feedback pulse signals of the motor encoder at continuous time points when the second joint of the robot rotates 90°, and check the motor encoder code disk according to the official data of the motor and driver company used by the robot. The total number of pulses per revolution is m=1500, and the angle corresponding to each pulse is calculated as α=360°/m=0.24°; Matlab software is used to calculate the feedback pulse signal of the motor encoder at continuous time points when the second joint rotates 90°. The number of real-time pulses, then the real-time rotation angle = the number of real-time pulses × α; the real-time rotation angle data is substituted into the dynamic simulation model to obtain the driving torque F ₁ of the second joint at continuous time points when it rotates 90°, as shown in the following table:

Table 1

2. In the double-joint linkage state of the second and third joints, collect the feedback pulse signals of the motor encoder at continuous time points when the second and third joints of the robot rotate 90°. According to the data, the total number of pulses m=1500 per revolution of the motor encoder code disc, and the angle corresponding to each pulse is calculated α=360°/m=0.24°; Matlab software is used to calculate the rotation of the second joint and the third joint 90 When °, the real-time pulse number corresponding to the feedback pulse signal of the motor encoder at continuous time points, then the real-time rotation angle = real-time pulse number × α; substitute the real-time rotation angle data of the i-th joint and the i+1-th joint into the dynamic simulation model to obtain the second joint The driving torque F ₂ at successive time points at a rotation of 90° is shown in the table below:

Table 2

3. Calculate the winding current I _{1 and the winding current I 2 under the single-joint motion state and the double-joint linkage state when the second joint rotates 90° by the following formulas, and obtain several winding currents I 1 and winding} currents I ₂ For the data, see Table 3. In the table, current_m2 is I ₁ , current_m2m3 is I ₂ , and the time point corresponding to the current;

其中K_t为转矩常数＝0.91，驱动力矩F₁或驱动力矩F₂＝T_e×第2关节减速器的减速比，第2关节减速器的减速比＝81；formula

Wherein K _t is the torque constant = 0.91, the driving torque F ₁ or the driving torque F ₂ =T _e × the reduction ratio of the joint 2 reducer, the reduction ratio of the joint 2 reducer = 81;

Table 3 Winding current I ₁ and winding current I ₂ data

4. Taking the continuous time points of step (1) as the abscissa, that is, time_m2 and time_m2m3 in Table 3 as the abscissa, and the winding current as the ordinate, draw the second joint in the single-joint motion state and the double-joint linkage state. The curves of winding current I ₁ and winding current I ₂ are shown in Figure 2. It can be seen from the figure that the two curves do not overlap, indicating that the current of the second joint has a coupling effect in the double-joint linkage state;

5. Use the current sensor to obtain the current signals x ₁ (t) and x ₂ (t) of the second joint in the single-joint motion state and the double-joint linkage state, respectively, as shown in Figure 3 and Figure 4, t is the acquisition time , the acquisition card used is NI9215, the acquisition software is SignalExpress, and the sampling frequency is 8192Hz;

6. Use the combined filtering method of zero-phase filtering and singular value denoising to filter the current signal to obtain the filtered signal, as follows:

(1) Filter the current signals x ₁ (t) and x ₂ (t) through a zero-phase filter to obtain a filtered signal;

(2) Perform short-time Fourier transform on the filtered signal to obtain time-frequency matrices A1 and A2, where the time-frequency matrix is defined as each row represents a frequency point, each column represents a time point, and each value represents a certain point at the acquisition time Amplitude at frequency;

(3) Substitute the time-frequency matrices A1 and A2 obtained in step (2) into the formula A=U∑V to perform singular value decomposition, and extract the values of the diagonal matrices ∑ ₁ and ∑ ₂ to obtain the singular value sequences d1 and d2, as shown in Table 4 where U and V are orthogonal matrices, and ∑ is a diagonal matrix;

Table 4 Singular value sequences of single-joint motion and double-joint linkage

d_m2 d_m2m3 2432.421 2270.41 906.9608 773.4315 561.6425 462.6762 302.3551 242.9642 177.0268 142.4626 103.7739 77.7333 51.65359 41.4323 19.79018 19.01606 17.71604 17.49756 16.84157 16.9581 15.78628 16.32425 14.80652 15.48301 12.72295 13.58477 12.12206 12.01732 11.04193 11.14144 10.51087 10.57703 10.20449 9.98004 ... ...

(4) Using the singular value sequences d1 and d2 obtained in step (3), subtract the previous value from the latter value in turn to obtain a new sequence. When the quotient of the five values reaches 20, a value of 150 in the change interval [100, 174] corresponding to the difference is selected as the threshold value; The value of is 100, which is used as the threshold, and the singular values smaller than the threshold are set to zero to form new singular value sequences q1 and q2;

(5) Construct new diagonal matrices ∑' ₁ and ∑' ₂ using the new singular value sequences q1 and q2, and substitute ∑' ₁ and ∑' ₂ into the formula A'=U∑'V' to obtain a new coefficient matrix A ' ₁ and A'₂;

(6) Inverse short-time Fourier transform is performed on the coefficient matrices A' ₁ and A' ₂ obtained in step (5) to obtain the final filtered signal, as shown in Figure 5 and Figure 6;

7. Substitute the filtered signal into the formula z(t)=x(t)+jH[x(t)] to obtain the analytical signal of the current signal, where x(t) is the real part of the complex signal z(t), H[x(t)] is the imaginary part of the complex signal z(t), and j is the imaginary unit;

得到滤波后信号在单关节运动状态下和双关节联动状态下的电流包络A₁(t)和电流包络A₂(t)，如图7和图8所示，实现关节机器人电流信号耦合作用部分与和耦合无关部分的分离；利用提取的部分进行后续的机器人电流信号解耦分析，进一步实现对机器人每个关节臂的运行状态的精确监测，及时发现机器人故障的早期特征，并采取相应措施，防止机器人因为零件磨损以及一些突发事件的发生而导致机器人突然停止运行，破坏生产线。Substitute H[x(t)] and x(t) into the formula

The current envelope A ₁ (t) and current envelope A ₂ (t) of the filtered signal in the single-joint motion state and the double-joint linkage state are obtained, as shown in Figures 7 and 8, to realize the current signal coupling of the joint robot Separation of the active part from the uncoupling part; using the extracted part to carry out the subsequent decoupling analysis of the robot current signal, to further realize the accurate monitoring of the running state of each joint arm of the robot, to find the early characteristics of the robot fault in time, and to take corresponding measures. Measures to prevent the robot from suddenly stopping operation due to the wear of parts and the occurrence of some emergencies, destroying the production line.

Claims (2)

1. an articulated robot current coupling signal separation and extraction method, is characterized in that, concrete steps are as follows: (1) When moving the angle θ in the single-joint motion state, collect the feedback pulse signals of the motor encoder of the i-th joint of the robot at different time points, and use the motor encoder feedback pulse signals to calculate the robot's i-th joint at different time points. Real-time rotation angle; use ADAMS software to build a dynamic simulation model, and substitute the real-time rotation angle data into the dynamic simulation model to obtain the driving torque F ₁ of the i-th joint at different time points; When the joint i and the i+1 joint move at an angle of θ in the double-joint linkage state, collect the motor encoder feedback pulse signals of the robot joint i and i+1 joint at different time points, and use the motor encoder feedback pulse The signal calculates the real-time rotation angle data of the i-th joint and the i+1-th joint of the robot at different time points; uses the ADAMS software to build a dynamic simulation model, and substitutes the real-time rotation angle data of the i-th joint and the i+1-th joint into the dynamic simulation model. Obtain the driving torque F ₂ of the i-th joint at different time points; (2) Calculate the winding current I ₁ and the winding current I ₂ under the single-joint motion state and the double-joint linkage state when the i-th joint moves at an angle of θ by the following formulas, and obtain several winding currents I ₁ and winding currents I ₂ data; formula

where K _t is the torque constant, _Te is the torque, I is the winding current, the driving torque F ₁ or the driving torque F ₂ =T _e × the reduction ratio of the ith joint reducer, (3) Taking the time of step (1) as the abscissa and the winding current as the ordinate, draw the winding current I ₁ and winding current I ₂ curves of the i-th joint in the single-joint motion state and the double-joint linkage state, and compare the two Whether the two curves overlap, when the two curves overlap, the current of the i-th joint has no coupling effect in the double-joint linkage state; (4) When the two curves do not overlap, the current of the i-th joint has a coupling effect in the double-joint linkage state; then when the single-joint motion state and the double-joint linkage state move the angle θ, the current sensor is used to obtain the first joint. Current signal data x ₁ (t) and x ₂ (t) of joint i in the single-joint motion state and the double-joint linkage state, where t is the acquisition time; (5) The combined filtering method of zero-phase filtering combined with singular value denoising is used to filter the current signal to obtain a filtered signal; (6) Substitute the filtered signal into the formula z(t)=x(t)+jH[x(t)] to obtain the analytical signal of the current signal, where x(t) is the real part of the complex signal z(t) , H[x(t)] is the imaginary part of the complex signal z(t), j is the imaginary unit; Substitute H[x(t)] and x(t) into the formula

Obtain the current envelope A ₁ (t) and the current envelope A ₂ (t) of the filtered signal in the single-joint motion state and the double-joint linkage state, and realize the separation of the joint robot current signal coupling part and the coupling-independent part . 2. The articulated robot current coupling signal separation and extraction method according to claim 1, is characterized in that, the combined filtering method is as follows: A. Filter the current signals x ₁ (t) and x ₂ (t) through a zero-phase filter to obtain a filtered signal; B. Perform short-time Fourier transform on the filtered signal to obtain a time-frequency matrix A, where the time-frequency matrix is defined as each row represents a frequency point, each column represents a time point, and each value represents a certain frequency at the time of acquisition Amplitude; C. Substitute the time-frequency matrix A obtained in step B into the formula A=UΣV to perform singular value decomposition, and extract the value of the diagonal matrix Σ to obtain the singular value sequence d, wherein U and V are both orthogonal arrays, and Σ is a diagonal matrix; D. Using the singular value sequence d obtained in step C, subtract the previous value from the next value in turn to obtain a new sequence. When the quotient of the first value and a later value in the new sequence reaches 20, select the latter value Any value in the corresponding change interval is taken as the threshold value, and the singular value smaller than the threshold value is set to zero to form a new singular value sequence q; E. Use the new singular value sequence q to construct a new diagonal matrix Σ′, and substitute Σ′ into the formula A′=UΣ′V′ to obtain a new coefficient matrix A′; F. Perform inverse short-time Fourier transform on the coefficient matrix A' obtained in step E to obtain the final filtered signal.

Images