Robot polishing flutter recognition and suppression method and system based on joint torque signals

Abstract

The invention discloses a robot polishing flutter recognition and suppression method and system based on joint torque signals. The method comprises the steps that S1, a kinetic model is established for a 3T2R type five-degree-of-freedom robot, and a driving torque calculation formula of each joint is obtained; s2, a built-in torque sensor is used for collecting the driving torque of each joint in real time; s3, the actual contact time of the tool and the workpiece is determined according to the torque signal of the joint 3; s4, torque signals of the joints 4 and 5 in the actual contact time period are extracted and preprocessed; s5, calculating a characteristic value of the processed signal, and comparing the characteristic value with a dynamic threshold value to determine a flutter occurrence time period; and S6, carrying out flutter suppression operation based on torque compensation on the flutter occurrence time period. The method is suitable for flutter identification and flutter suppression in industrial robot grinding, and has important significance for improving the quality and efficiency of robot grinding operation.

Description

Technical Field

[0001] The present invention belongs to the field of robot grinding technology and automation equipment, and particularly relates to a method and system for identifying and suppressing robot grinding chatter based on joint torque signals. Background Art

[0002] In modern industrial manufacturing, robot grinding technology is widely used in the surface treatment of various parts due to its high efficiency, flexibility, and high repeatability. However, during the robot grinding operation, the chatter phenomenon is a common problem, which seriously affects the grinding quality and production efficiency. Chatter refers to the self-excited vibration generated by the system due to factors such as the grinding force characteristics and the robot structure without external excitation. This phenomenon will lead to an increase in the surface roughness of the workpiece, a decrease in dimensional accuracy, and even damage to the workpiece and grinding tools in severe cases, affecting production efficiency and product quality.

[0003] Compared with traditional large machine tools, industrial robots have the advantages of low cost, high flexibility, and large working space. They have low requirements for the quality of operators and are easier to set up and transport. However, the relatively weak rigidity of the robot's serial structure makes chatter more likely to occur, resulting in problems such as poor surface quality of the workpiece, reduced processing efficiency, and shortened tool life. Therefore, how to effectively monitor, identify, and suppress the chatter phenomenon during robot grinding has become an important topic in the current research of industrial robot technology.

[0004] Existing chatter detection methods usually rely on external sensors such as accelerometers or microphones. These methods have problems such as complex installation, high cost, and sensitivity to environmental noise. In addition, most existing chatter suppression methods are achieved by adjusting the motion parameters of the robot or adding external damping devices. These methods are often difficult to control precisely and may affect the normal motion performance of the robot. For example, reducing the occurrence of chatter by adjusting the motion speed or feed rate of the robot will reduce the processing efficiency. While adding an external damping device can effectively suppress chatter, it will increase the complexity and cost of the system and may have a negative impact on the flexibility and working space of the robot. Summary of the Invention

[0004] The present invention provides a method and system for identifying and suppressing robot grinding chatter based on joint torque signals to solve the problems existing in the prior art and improve the stability and efficiency of robot grinding operations.

[0005] The implementation process of the present invention is as follows. A method and system for identifying and suppressing robot grinding chatter based on joint torque signals, and the method specifically includes the following steps:

[0006] S1. Establish a dynamic model for a 3T2R type five-degree-of-freedom robot to obtain the calculation formula for the driving torque of each joint;

[0007] S2. Use the torque sensors built into the five-degree-of-freedom robot to collect the driving torque of each joint in real time;

[0008] S3. According to the torque signal of joint 3, introduce the sliding variance method to determine the actual contact time between the tool and the workpiece;

[0009] S4. Extract the torque signals of joint 4 and joint 5 during the actual contact period between the tool and the workpiece, and perform preprocessing operations such as data normalization and denoising filtering on the extracted joint 4 and joint 5 signals;

[0010] S5. Calculate the CEI eigenvalue of the processed signal and compare it with the set dynamic threshold to determine the flutter occurrence period;

[0011] S6. Perform flutter suppression operations based on the torque compensation strategy during the flutter occurrence period.

[0012] Furthermore, the specific content of S1 is as follows:

[0013] Establish a link coordinate system through the D-H parameter method and deduce the homogeneous transformation matrix of the forward kinematics of the robot;

[0014] Deduce the expression of the driving torque of each joint of the robot based on the Newton-Euler method;

[0015] The driving torque τ of joint i i :

[0016] τ i = i-1 z i-2 T · i-1 n i-1

[0017] Where, i-1 n i-1 The torque relative to the D-H reference coordinate system O associated with it i

[0018] Furthermore, the specific content of S2 is as follows:

[0019] Through the torque sensors built into the robot, collect the driving torque of each joint in real time and transmit it to the flutter recognition and flutter suppression module.

[0020] Furthermore, the specific content of S3 is as follows:

[0021] The torque signal sequence of the robot is x(t), where t = 1, 2,..., N represents each sampling moment, and the mean value within the window ​is as follows:

[0022]

[0023] The length of the moment signal sliding window is W. At each sampling point i, the variance of the data within the window is as follows:

[0024]

[0025] Repeat the above calculation to obtain the complete sliding variance sequence The sum of the mean μ and the double standard deviation 2σ of the sequence is used as the discrimination threshold T1:

[0026]

[0027] The discrimination function of the contact state is:

[0028]

[0029] where A(t) is the contact state indicator function, is the sliding variance value at time t.

[0030] Furthermore, the specific content of S4 is as follows:

[0031] For the joint 4 and joint 5 moment signals extracted, use a 4th-order Butterworth low-pass filter to filter them, retain their low-frequency signals, and ensure the accuracy of subsequent eigenvalue extraction.

[0032] The 4th-order Butterworth low-pass filter uses bilinear transformation to convert the analog filter into a digital filter, and its transfer function is:

[0033]

[0034] It is applicable to the following difference equation:

[0035]

[0036] where x[n] is the input original moment signal, y[n] is the output filtered moment signal, a i and b i are coefficients.

[0037] At the same time, perform two-way filtering on the moment signal, that is, first forward filtering and then reverse filtering, to ensure that the output filtered moment signal is aligned with the input original moment signal in the time axis. Its equivalent transfer function is:

[0038] H zero-phase (z) = H(z) · H(z -1 )

[0039] Further, S5 is specifically as follows:

[0040] Calculate the CEI eigenvalue of the processed signal:

[0041]

[0042] where P i is the normalized power spectrum of the i-th frequency band within the reference window, satisfying Q i is the normalized power spectrum of the i-th frequency band within the current analysis window, satisfying ε is a very small positive constant, usually taking 10 -8 ~10 -10 , S ref is the spectral amplitude of the reference window at the frequency f i , S curr (f i ) is the spectral amplitude of the current window at the frequency f i , and N is the number of frequency points for spectral analysis.

[0043] Perform CEI calculation on the torque signal during the contact time period to obtain the CEI value sequence {CEI1, CEI2,..., CEI n}}, for the k-th time point, use the first m values {CEI k-m , CEI k-m+1 , …, CEI k-1} to calculate the mean μ k and the standard deviation σ k , and the dynamic threshold T2 at the k-th time point can be obtained as:

[0044]

[0045] where κ is an adjustment coefficient.

[0046] Further, S6 is specifically as follows:

[0047] The torque τ raw (t) during the flutter time period can be expressed as:

[0048] τ raw (t) = τ stable (t) + τ chatter (t)

[0049] where τ stable (t) is the stable component, and τ chatter (t) is the flutter interference component.

[0050] Extract the frequency f chatter of τ chatter during the flutter time periodAnd the phase φ, design the compensation torque τ comp (t) is:

[0051] τ comp (t) = -A·sin(2πft chatter t + φ) = -τ chatter (t)

[0052] Where A is the amplitude.

[0053] The output torque τ out (t) after compensation is:

[0054] τ out (t) = τ raw (t) + τ comp (t) = τ stable (t)

[0055] A robot grinding operation chatter recognition and suppression system, comprising:

[0056] A chatter recognition and chatter suppression module, configured to run in MATLAB software to perform operations such as determining the actual contact section, extracting and preprocessing the torques of joint 4 and joint 5, calculating and comparing eigenvalues, and calculating the compensation torque;

[0057] A dSPACE controller, configured to transfer and execute control commands, and establish a data path with the torque sensor, the torque controller, and the chatter recognition and chatter suppression module;

[0058] A torque sensor, configured to complete high-speed acquisition and preliminary processing of the actual force data of each joint;

[0059] A torque controller, configured to achieve high-speed data interaction with the dSPACE controller to ensure real-time response and closed-loop execution of the torque compensation command.

[0060] Through the chatter recognition and chatter suppression method based on the joint torque signal proposed in this paper, the chatter phenomenon in the robot grinding process can be effectively detected, and the effective suppression of chatter can be achieved through the torque compensation strategy. This method can accurately identify the chatter area, and reduce the influence of chatter on the robot operation through the compensation torque, improving the stability and reliability of the robot. The optimized robot grinding process is more accurate in torque control, with a significant chatter suppression effect, which is beneficial to improving the efficiency and quality of the robot grinding operation. Brief Description of the Drawings

[0061] Figure 1 It is the flowchart of the present invention.

[0062] Figure 2 It is the 3T2R type five-degree-of-freedom robot which is the research object of the method proposed by the present invention.

[0063] Figure 3 is the D-H reference coordinate system established based on the robot.

[0064] Figure 4 Shows the actual contact time determination diagram of the tool and the workpiece.

[0065] Figure 5 Shows the chatter recognition result diagram based on the signal of joint 4.

[0066] Figure 6 Shows the chatter recognition result diagram based on the signal of joint 5.

[0067] Figure 7 Shows the chatter suppression result diagram based on the signal of joint 4.

[0068] Figure 8 Shows the chatter suppression result diagram based on the signal of joint 5. Detailed implementation manners

[0069] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0070] The present invention will be further described in detail below through specific embodiments and with reference to the accompanying drawings.

[0071] An embodiment of the present invention provides a method for robot grinding chatter recognition and suppression based on joint torque signals. Figure 1 Shows the robot chatter recognition and suppression process, including:

[0072] S1. Establish a dynamic model for a 3T2R type five-degree-of-freedom robot to obtain the calculation formula for the driving torque of each joint.

[0073] In a preferred embodiment, S1 includes:

[0074] Establish a link coordinate system by the D-H parameter method and deduce the homogeneous transformation matrix of the forward kinematics of the robot;

[0075] Deduce the expression of the driving torque of each joint of the robot based on the Newton-Euler method.

[0076] Figure 2The 3T2R high - dexterity robot under study is shown. The robot consists of translational joint axes J1 - J3 (all composed of ball screw pairs) and rotational joint axes J4 and J5. The translation of the robot end - effector in two directions within the horizontal plane is achieved by the combined method of parallel driving and differential driving of J1 and J2, forming a translational and rotational coupling motion. The change in the posture during the robot grinding process is determined by the change in the rotation angles of the rotational axes J4 and J5.

[0077] Figure 3 The D - H reference coordinate system of the studied 3T2R high - dexterity robot is shown. The pose matrix of the tool coordinate system of the robot end - effector relative to the base coordinate system obtained by forward kinematics solution is 0 T5 = 0 T1· 1 T2· 2 T3· 3 T4· 4 T5. Further, based on the attitude matrix and position vector of the homogeneous transformation matrix between adjacent coordinate systems, the moment of link i relative to its associated D - H reference coordinate system O i can be obtained, and then the driving moment of joint i corresponding to link i can be obtained:

[0078] τ i = i-1 z i-2 T · i-1 n i-1

[0079] S2. Use the torque sensors built into the 5 - degree - of - freedom robot to collect the driving torques of each joint in real - time;

[0080] In a preferred embodiment, S2 includes: collecting the driving torques of each joint in real - time through the torque sensors built into the robot and transmitting them to the chatter recognition and chatter suppression module.

[0082] S3. According to the torque signal of joint 3, introduce the sliding variance method to determine the actual contact time between the tool and the workpiece.

[0083] In a preferred embodiment, S3 includes:

[0084] Figure 4 It shows that in a preferred embodiment, based on the torque signal of joint 3, the sliding variance method is used to determine the actual contact time between the tool and the workpiece.

[0085] The torque signal sequence of the robot is x(t), where t = 1, 2, …, N represents each sampling moment, and the mean value within the window is:

[0086]

[0087] The length of the torque signal sliding window is W. At each sampling point i, the variance of the data within the window is:

[0088]

[0089] Repeat the above calculation to obtain the complete sliding variance sequence The sum of the mean μ and twice the standard deviation 2σ of the sequence is used as the discrimination threshold T1:

[0090]

[0091] The discrimination function for the contact state is:

[0092]

[0093] where A(t) is the contact state indicator function, is the sliding variance value at time i.

[0094] S4. Extract the joint 4 and joint 5 torque signals during the actual contact period between the tool and the workpiece, and perform preprocessing operations such as data normalization and denoising filtering on the extracted joint 4 and joint 5 signals;

[0095] In a preferred embodiment, S4 includes:

[0096] For the extracted joint 4 and joint 5 torque signals, use a 4th-order Butterworth low-pass filter to filter them, retain their low-frequency signals, and ensure the accuracy of subsequent eigenvalue extraction.

[0097] The 4th-order Butterworth low-pass filter converts the analog filter into a digital filter using the bilinear transformation, and its transfer function is:

[0098]

[0099] It is applicable to the following difference equation:

[0100]

[0101] where x[n] is the input original torque signal, y[n] is the output filtered torque signal, a i and b i are coefficients.

[0102] At the same time, perform two-way filtering on the torque signal, that is, filter forward first and then backward, to ensure that the output filtered torque signal is aligned with the input original torque signal in the time axis, and its equivalent transfer function is:

[0103] Hzero-phase (z) = H(z) · H(z -1 )

[0104] S5. Calculate the CEI eigenvalue of the processed signal and compare it with the set dynamic threshold to determine the flutter occurrence period

[0105] In the preferred embodiment, the above S5 includes:

[0106] Figure 5 Shows the flutter recognition result based on the signal of joint 4 in the preferred embodiment; Figure 6 Shows the flutter recognition result based on the signal of joint 5 in the preferred embodiment.

[0107] Calculate the CEI eigenvalue of the processed signal:

[0108] In frequency-domain analysis, assume the original time-series signal is a(t), and its corresponding spectrum is A(f). Obtained through discrete Fourier transform:

[0109]

[0110] where N is the number of frequency points for spectrum analysis, A k represents the complex spectrum value of the k-th frequency component. After taking the square of its magnitude and normalizing, the power spectral density distribution within this window can be obtained.

[0111]

[0112] where P i is the normalized power spectrum of the i-th frequency band within the reference window, satisfying Q i is the normalized power spectrum of the i-th frequency band within the current analysis window, satisfying ε is a very small positive constant, usually taking 10 -8 ~10 -10 , S ref is the spectrum amplitude of the reference window at frequency f i , S curr (f i ) is the spectrum amplitude of the current window at frequency f i , and N is the number of frequency points for spectrum analysis.

[0113] Perform CEI calculation on the torque signal during the contact period to obtain the CEI value sequence {CEI1, CEI2,..., CEI n} For the k-th time point, use the previous m values {CEI k-m , CEI k-m+1 ,..., CEI k-1} to calculate the mean μk and standard deviation σ k , the dynamic threshold T2 at the k-th time point can be obtained as follows:

[0114]

[0115] where κ is an adjustment coefficient.

[0116] The discrimination function for the grinding state is:

[0117]

[0118] where B(t) is the contact state indicator function, and CEI t is the CEI value at time t.

[0119] S6. Perform flutter suppression operation based on the torque compensation strategy during the flutter occurrence period.

[0120] In a preferred embodiment, the S6 includes:

[0121] Figure 7 shows the flutter suppression result of calculating the eigenvalue based on the signal of joint 4 in a preferred embodiment; Figure 8 shows the flutter suppression result of calculating the eigenvalue based on the signal of joint 5 in a preferred embodiment.

[0122] The torque τ raw (t) during the flutter time period can be expressed as:

[0123] τ raw (t) = τ stable (t) + τ chatter (t)

[0124] where τ stable (t) is the stable component, and τ chatter (t) is the flutter interference component.

[0125] Extract the frequency f chatter and phase φ of τ chatter (t) during the flutter time period, and design the compensation torque τ comp (t) as:

[0126] τ comp (t) = -A·sin(2πf chatter t + φ) = -τ chatter (t)

[0127] where A is the amplitude.

[0128] The output torque τ out (t) after compensation is:

[0129] τout τ(t) = τ raw τ(t) + τ comp τ(t) = τ stable τ(t)

[0130] For joint 4, from Equation (1), when chatter is detected, a compensation torque term Δτ is introduced. comp The corrected torque equation is:

[0131]

[0132] where Δn comp is the external torque corresponding to the compensation torque.

[0133] For joint 5, from Equation (1), when chatter is detected, a compensation torque term Δτ is introduced. comp The corrected torque equation is:

[0134]

[0135] where Δn comp is the external torque corresponding to the compensation torque.

[0136] The present invention has been described above in an exemplary manner with reference to the accompanying drawings. Obviously, the specific implementation of the present invention is not limited by the above methods. As long as various non-substantive improvements are made by adopting the method concept and technical solution of the present invention, or the concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A method and system for robot grinding chatter recognition and suppression based on joint torque signals, including the steps: S1. Establish a dynamic model for a 3T2R type five-degree-of-freedom robot to obtain the calculation formula for the driving torque of each joint; S2. Use the torque sensors built in the 5-degree-of-freedom robot to collect the driving torque of each joint in real time; S3. According to the torque signal of joint 3, introduce the sliding variance method to determine the actual contact time between the tool and the workpiece; S4. Extract the torque signals of joints 4 and 5 during the actual contact period between the tool and the workpiece, and perform preprocessing operations such as data normalization and denoising filtering on the extracted signals of joints 4 and 5; S5. Calculate the CEI (Cross-Entropy Information) eigenvalue of the processed signal, and compare it with the set dynamic threshold to determine the chatter occurrence period; S6. Perform chatter suppression operations based on the torque compensation strategy during the chatter occurrence period.

2. The robot grinding chatter recognition and suppression method and system based on joint torque signals according to claim 1, wherein The S1 for establishing a dynamic model for a 3T2R type five-degree-of-freedom robot includes: Establish a link coordinate system through the D-H parameter method, and deduce the homogeneous transformation matrix of the forward kinematics of the robot; Deduce the expression of the driving torque of each joint of the robot based on the Newton-Euler method.

3. A method and system for robot grinding chatter recognition and suppression based on joint torque signals according to claim 1, characterized in that, The S2 collects the driving torque of each joint in real time by using the torque sensors built in the 5-degree-of-freedom robot, collects the driving torque of each joint in real time, and transmits it to the chatter recognition module.

4. A method and system for identifying and suppressing robot grinding chatter based on joint torque signals according to claim 1, characterized in that, The S3 determines the actual contact time between the tool and the workpiece according to the torque signal of joint 3 by introducing the sliding variance method, specifically: The torque signal sequence of the robot is x(t), where t = 1, 2, …, N represents each sampling moment, and the mean value within the window is: The length of the torque signal sliding window is W. At each sampling point i, the variance of the data within the window is as follows: Repeat the above calculation to obtain the complete sliding variance sequence The sum of the mean μ and twice the standard deviation 2σ of the sequence is used as the discrimination threshold T1: The discriminant function of the contact state is: where A(t) is the contact state indication function, is the sliding variance at time t.

5. The robot grinding chatter recognition and suppression method and system based on joint torque signals according to claim 1, characterized in that, The S4 extracts the torque signals of joints 4 and 5 during the actual contact period between the tool and the workpiece, and performs preprocessing operations such as data normalization and denoising filtering on the extracted signals of joints 4 and 5, specifically: for the extracted torque signals of joints 4 and 5, use a 4th-order Butterworth low-pass filter to filter them, retain their low-frequency signals, and ensure the accuracy of subsequent eigenvalue extraction. The 4th-order Butterworth low-pass filter converts the analog filter into a digital filter by using the bilinear transformation, and its transfer function is: Applicable to the following difference equation: where x[n] is the original torque signal as input, y[n] is the filtered torque signal as output, a i and b i are coefficients. At the same time, perform two-way filtering on the torque signal, that is, filter forward first and then backward, to ensure that the output filtered torque signal is aligned with the input original torque signal in the time axis, and its equivalent transfer function is: H zero-phase (z) = H(z) · H(z -1 )。 6. A method and system for identifying and suppressing robot grinding chatter based on joint torque signals according to claim 1, characterized in that The S5 calculates the CEI (Cross-Entropy Information) eigenvalue of the processed signal, and compares it with the set dynamic threshold to determine the chatter occurrence period, specifically: The calculation formula for the eigenvalue CEI is: Among them, P i is the normalized power spectrum of the ith frequency band in the reference window, satisfying Q i is the normalized power spectrum of the ith frequency band in the current analysis window, satisfying ε is a very small positive constant, usually 10 -8 ~10 -10 , S ref The reference window is at frequency f i The spectrum amplitude at S curr (f i ) is the current window at frequency f i The spectrum amplitude at , N is the number of frequency points for spectrum analysis. Calculate the CEI for the torque signal during the contact time period to obtain the CEI value sequence {CEI1, CEI2,..., CEI n}, for the k-th time point, use the first m values {CEI k-m , CEI k-m+1 , …, CEI k-1} to calculate the mean μ k and the standard deviation σ k , and the dynamic threshold T2 at the k-th time point can be obtained: Among them, κ is the adjustment coefficient.

7. A method and system for robot grinding chatter recognition and suppression based on joint torque signals according to claim 1, characterized in that, The S6 performs chatter suppression operations based on the torque compensation strategy during the chatter occurrence period, specifically: The torque τ during the flutter time period raw (t) can be expressed as: τ raw ψ(t) = τ stable φ(t) + τ chatter ω(t) Among them, τ stable (t) is the stable component, and τ chatter (t) is the flutter interference component. Extract τ within the flutter time period chatter (t) of the frequency f chatter and the phase φ, and design the compensation torque τ comp (t) is as follows: τ comp (t) = -A·sin(2πf chatter t + φ) = -τ chatter (t) Among them, A is the amplitude. The output torque τ out (t) after compensation is as follows: τ out ψ(t) = τ raw φ(t) + τ comp χ(t) = τ stable ω(t).

8. A robot grinding operation chatter recognition and suppression system, characterized in that, Including: A chatter recognition and chatter suppression module, configured to run in the MATLAB software to perform operations such as determining the actual contact section, extracting and preprocessing the torque of joints 4 and 5, calculating and comparing eigenvalues, and calculating the compensation torque; The dSPACE controller is configured to control the relay and execution of control commands and establish data paths with the torque sensor, the torque controller, and the flutter identification and flutter suppression module; The torque sensor is configured to complete the high-speed acquisition and preliminary processing of the actual force data of each joint; The torque controller is configured to achieve high-speed data interaction with the dSPACE controller to ensure the real-time response and closed-loop execution of torque compensation commands.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program is capable of implementing any of the method steps described in claims 1 to 7 and invoking the functional modules of any of the systems described in claim 8.

Images