CN111754567B - Comprehensive compensation method for static and dynamic errors in grinding and polishing processing of aircraft composite member robot - Google Patents

Abstract

The invention belongs to the field of robotic grinding and polishing processing of composite material components, and specifically discloses a comprehensive compensation method for static and dynamic errors in the robotic grinding and polishing processing of aircraft composite material components. The method includes: building a coordinate system measurement system for robot grinding and polishing of aircraft composite components; establishing a global coordinate system for the robot processing system; constructing a robot dynamics recursion model; identifying static parameters of the robot processing system, and establishing a robot distance error calibration model to compensate Robot body geometric error; construct a dynamic error compensation model for robot processing; optimize and train the dynamic error compensation model for robot processing based on the theoretical data set and prior knowledge data set for robot processing; train the robot based on the optimized dynamic error compensation model for robot processing Error compensation is performed on the terminal pose. This invention accurately constructs a coordinate system global control network under the constraints of the dynamic model, which facilitates comprehensive analysis of the error chain and ensures the accuracy of the robot grinding and polishing surface.

Description

Comprehensive compensation method for static and dynamic errors in robotic grinding and polishing of aircraft composite components

Technical field

The invention belongs to the field of robotic grinding and polishing of composite components, and more specifically, relates to a comprehensive compensation method for static and dynamic errors in the robotic grinding and polishing of aircraft composite components.

Background technique

Carbon fiber composite materials have been widely used in aerospace structural materials due to their excellent properties such as high temperature resistance, friction resistance, high strength, low density, corrosion resistance, lightweight and high strength, and designable performance. They are an important tool to achieve lightweight and high-performance aerospace equipment. The key to globalization, and its usage has become one of the symbols of the advancement and international competitiveness of aerospace vehicles. According to statistics, Z10 helicopter composite materials (hereinafter referred to as "composites") account for 90%, J20 fighter composite materials account for approximately 40%, CR929 passenger aircraft composite materials account for approximately 51%, and C919 passenger aircraft composite materials account for approximately 30% .

Since carbon fiber composite materials are prone to aging and have poor barrier properties, after the aircraft composite components are formed, their surfaces need to be polished so that a protective layer can be sprayed on them to improve the component's anti-corrosion, weather resistance, wear resistance and other protective properties. At present, the existing surface grinding technologies for aircraft composite components are mainly manual grinding and manual-assisted grinding. Workers are required to process large-scale aircraft composite components with hand-held or auxiliary grinding instruments. This has many disadvantages such as low processing accuracy, poor product consistency, and low processing efficiency, which seriously restricts the development of automated and intelligent production of aviation composite components and has become a field of aviation manufacturing technology. difficulties.

Compared with the first two surface grinding methods of aircraft composite components, robot grinding and polishing can take advantage of its advantages such as good flexibility, large operating space, and the ability to expand a variety of external sensing units, and is expected to realize intelligent processing of aircraft composite components. . However, the current global error chain in the robot grinding and polishing process is not fully established, and it is difficult to effectively capture errors in attitude, calibration, measurement and processing, which seriously affects the processing surface accuracy of the workpiece and limits the use of robot grinding and polishing in the intelligent production of aircraft composite components. applications in.

Based on the above defects and deficiencies, there is an urgent need in this field to propose a comprehensive compensation method for static and dynamic errors in the robotic grinding and polishing of aircraft composite components to solve the problem of low precision, poor product consistency, and low processing efficiency in the actual robotic grinding and polishing of aircraft composite components. Advanced questions.

Contents of the invention

In view of the above defects or improvement needs of the existing technology, the present invention provides a comprehensive compensation method for static and dynamic errors in robot grinding and polishing of aircraft composite components, which combines the characteristics of static and dynamic errors in robot grinding and polishing with its dynamic errors in robot processing. According to the characteristics of compensation model optimization, the robot body geometric parameter error is compensated based on the distance error calibration model of the robot processing system. On this basis, a new support vector machine algorithm that integrates prior knowledge is used to establish an error compensation model to correct the robot pose and posture in real time. It compensates the end loading deformation error and reduces the dynamic error of robot processing, so that the processed aircraft composite components have high precision, good product consistency and high processing efficiency.

In order to achieve the above objectives, the present invention proposes a comprehensive static and dynamic error compensation method for robot grinding and polishing of aircraft composite components, which includes the following steps:

S1 builds a coordinate system measurement system for robot grinding and polishing of aircraft composite components;

S2 establishes the global coordinate system of the robot processing system;

S3 builds a robot dynamics recursive model based on the global coordinate system of the robot processing system;

S4 identifies the static parameters of the robot processing system based on the robot dynamics recursion model, and establishes a robot distance error calibration model based on the static parameters of the robot processing system to compensate for the robot body geometric error;

S5 builds a dynamic error compensation model for robot processing in the global coordinate system of the robot processing system based on the geometric error of the robot body;

S6 optimizes and trains the robot processing dynamic error compensation model based on the robot processing dynamic error theory data set and the prior knowledge data set;

S7 performs error compensation on the robot end pose based on the optimized robot processing dynamic error compensation model.

As a further preference, in step S1, the robot grinding and polishing processing coordinate system measurement system includes a laser tracker and binocular measurement equipment.

As a further preference, in step S2, the global coordinate system of the robot processing system includes a workpiece coordinate system, a tool coordinate system, a measurement coordinate system and a robot global reference coordinate system.

As a further preference, in step S2, the global coordinate system of the robot processing system is established using a nonlinear optimization algorithm based on discrete differential evolution.

As a further preference, in step S5, a new support vector machine regression algorithm that integrates prior knowledge is used to construct a dynamic error compensation model for robot processing.

As a further preference, in step S6, a chaotic particle swarm optimization algorithm is used to optimize and train the robot processing dynamic error compensation model.

As further preferred, step S6 specifically includes the following steps:

S61 selects a set of target points A scattered in the robot grinding and polishing processing space, and measures the pose P _AM of the target point set A in the measurement coordinate system W ₁ through the robot grinding and polishing processing coordinate system measurement system. Obtain the theoretical pose P _A of the target point set A in the robot's global reference coordinate system W ₂ , and then obtain the coordinate system transformation matrix T between the measurement coordinate system W ₁ and the robot's global reference coordinate system W ₂ ;

S62 determines a set B of target points to be compensated at the end of the robot, and divides the set B of target points to be compensated into a robot processing dynamic error prior knowledge data set B _p and a remaining data set C _B B _p according to a preset ratio;

S63 Measure the target pose P _Bpm of each point in the prior knowledge data set B _p under the measurement coordinate system W ₁ , and then obtain the prior knowledge data set B _p through the coordinate system transformation matrix T. The actual pose P _Bpr of each point in the robot's global reference coordinate system W ₂ ;

S64 obtains the corrected pose P _Bpr ' corresponding to each point in the prior knowledge data set B _p according to the actual pose P _Bpr and the theoretical pose P _Bpt of each point in the prior knowledge data set B _p ;

S65 uses the actual pose P _Bpr of each point in the prior knowledge data set B _p as the input of the robot processing dynamic error compensation model, and the corrected pose P _Bpr ' as the output of the robot processing dynamic error compensation model. The machining dynamic error compensation model is optimized and trained;

S66 uses the target pose of the remaining data set C _B B _p as the input of the optimized and trained robot processing dynamic error compensation model, and the model outputs the corrected target point set B', and the corrected target point set B' serves as the basis for static and dynamic error compensation in robot grinding and polishing.

As a further preference, in step S61, the target point set A is obtained as follows:

S611 sets the number of sample points that constitute the set;

S612 divides the robot grinding and polishing processing space into regions;

S613 randomly selects sample points in each interval after the robot grinding and polishing processing space is divided according to the robot grinding and polishing processing path and the number of sample points that constitute the set, thereby obtaining the target point set A.

As a further preference, in step S7, error compensation for the robot end pose includes compensation for the robot end position and attitude.

As a further preference, the robot end position is compensated by a position compensator, and the robot end attitude is compensated by an attitude compensator.

Generally speaking, compared with the existing technology, the above technical solution conceived by the present invention mainly has the following technical advantages:

1. The present invention establishes a dynamic recursion model based on the global coordinate system, identifies the static parameters of the robot processing system, obtains the distance error calibration model of the robot processing system based on the static parameters, and compensates the robot body geometric parameter error; finally adopts The new support vector machine algorithm that integrates prior knowledge establishes an error compensation model, corrects the robot's posture and compensates for end loading deformation errors in real time, and reduces dynamic errors in robot processing. This invention accurately constructs a coordinate system global control network under the constraints of the dynamic model, which facilitates comprehensive analysis of the error chain and ensures the accuracy of the robot grinding and polishing surface.

2. The distance error calibration model established by the present invention optimizes the initial value of dynamic compensation of the robot processing system, significantly improving iteration efficiency and iteration accuracy.

3. The present invention adopts the chaotic particle swarm optimization algorithm, which has the characteristics of simple structure, fast learning speed and outstanding generalization performance. It realizes real-time dynamic compensation of robot posture and processing trajectory, and improves the accuracy and robustness of the robot grinding and polishing processing system. Great sex.

4. The present invention divides the robot grinding and polishing processing space into regions, and improves the traversability of the sample space according to the robot grinding and polishing processing path and the number of sample points constituting the set, so that the optimized robot processing dynamic error compensation model can adapt to The accuracy is stronger, and the obtained robot end pose compensation is more accurate.

Description of drawings

Figure 1 is a schematic flow chart of a comprehensive compensation method for static and dynamic errors in robotic grinding and polishing of aircraft composite components according to an embodiment of the present invention;

Figure 2 is a flow chart of robot posture and terminal loading deformation error compensation involved in the embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

As shown in Figures 1 and 2, an embodiment of the present invention provides a comprehensive compensation method for static and dynamic errors in robot grinding and polishing of aircraft composite components, using a new support vector machine algorithm that integrates prior knowledge to achieve the robot end target pose and posture. Correct the mapping of pose so that the actual pose after compensation is consistent with the target pose to improve the accuracy of robot grinding and polishing. In the present invention, prior knowledge refers to the theoretical pose of the robot's end target point. The specific implementation method is as follows:

Step 1: Build a coordinate system measurement system for robot grinding and polishing of aircraft composite components.

In the present invention, the coordinate system measurement system includes a laser tracker and binocular measurement equipment. Or, in order to further improve the accuracy, speed and resolution of measurement, the coordinate system measurement system is an Optitrack system composed of multiple cameras, multiple calibration balls and calibration plates.

Step 2: Establish the global coordinate system of the robot processing system based on the nonlinear optimization algorithm of discrete differential evolution.

According to the coordinate system measurement system built in step 1, the nonlinear optimization algorithm of discrete differential evolution is used to establish the global coordinate system of the robot processing system. Among them, the global coordinate system of the robot processing system at least includes the workpiece coordinate system, the tool coordinate system, the measurement coordinate system and the robot global reference coordinate system. The workpiece coordinate system is the coordinate system where the complex aircraft components to be processed are located, the tool coordinate system is the coordinate system where the tool held by the end of the robot is located, the measurement coordinate system is the coordinate system where the measurement system is located, and the robot global reference coordinate system is the base of the robot. Coordinate System. Through the relative position relationship and the kinematic relationship of the robot, the transformation relationship between each coordinate system is obtained, and then the robot end pose information can be measured and obtained, which is the actual pose of the robot end. In the present invention, a nonlinear optimization algorithm model of discrete differential evolution is first constructed, and then the pose relationships between workpieces, tools, robots and reference objects measured by the coordinate system measurement system are input into the discrete differential algorithm as variable parameters. The evolved nonlinear optimization algorithm model obtains the workpiece coordinate system, tool coordinate system, measurement coordinate system, robot global reference coordinate system and the conversion relationship between each coordinate system. The difference between the actual pose of the robot end measured by the coordinate measurement system and the theoretical pose of the robot end input by the control system is the robot end pose error. Among them, the pose error includes position error and attitude error. The position error includes the error of the robot end along the X-axis, Y-axis and Z-axis. The attitude error includes the angle error of the α angle, β angle and γ angle of the robot end.

Step 3: Construct a robot dynamics recursion model based on the global coordinate system of the robot processing system to obtain the robot joint torque relationship. The robot joint torque relationship includes the robot joint torque, angle, angular velocity and angular acceleration.

Step 4: Identify the static parameters of the robot processing system based on the robot dynamics recursive model, and establish a robot distance error calibration model based on the static parameters of the robot processing system to compensate for the geometric error of the robot body, and calibrate the robot based on the distance error calibration model. Body geometry errors are compensated. Generally speaking, the theoretical identification method can be used to identify the static parameters of the robot processing system. Among them, the geometric error of the robot body is mainly caused by the robot installation error, tool installation error and other errors caused by geometric construction and installation. By constructing a robot dynamics recursion model based on the global coordinate system of the robot processing system, and using this to identify the static parameters of the robot processing system, the control and compensation of the static errors of the robot can be achieved.

Step 5: Construct a dynamic error compensation model for robot processing in the global coordinate system of the robot processing system based on the geometric error of the robot body. Among them, the robot machining dynamic error compensation model is constructed using a new support vector machine regression algorithm that integrates prior knowledge. That is, prior knowledge is integrated with a new support vector machine regression algorithm to construct a dynamic error compensation model for robot processing in the global coordinate system of the robot processing system. In the present invention, prior knowledge refers to the theoretical pose of the robot's end target point, and a group of theoretical poses of the robot's end target point constitutes a priori knowledge data set. In the present invention, a new support vector machine regression algorithm based on prior knowledge first constructs a dynamic error compensation model for robot processing, and combines the robot processing dynamic parameters, robot end processing deformation, transmission error, control error, threshold parameters, and weight coefficients. etc. are used as parameters of a new support vector machine regression algorithm based on prior knowledge to construct a dynamic error compensation model for robot machining.

Step 6: Optimize and train the robot processing dynamic error compensation model based on the robot processing dynamic error theory data set and the prior knowledge data set. The entire iterative optimization training process is carried out using the chaotic particle swarm optimization algorithm. It specifically includes the following steps:

(1) Select a set of target points A scattered in the robot grinding and polishing processing space, and measure the pose P of the target point set A in the measurement coordinate system W ₁ through the robot grinding and polishing processing coordinate system measurement system. _AM , obtain the theoretical pose P _A of the target point set A in the robot's global reference coordinate system W ₂ , and then obtain the coordinate system transformation between the measurement coordinate system W ₁ and the robot's global reference coordinate system W ₂ matrixT;

(2) Determine a set B of target points to be compensated at the end of the robot, and divide the set B of target points to be compensated into the robot machining dynamic error prior knowledge data set B _p and the remaining data set C _B B _p according to a preset ratio;

(3) Measure the target pose P _Bpm of each point in the prior knowledge data set B _p under the measurement coordinate system W ₁ , and then obtain the prior knowledge data set through the coordinate system transformation matrix T The actual pose P _Bpr of each point in B _p under the robot's global reference coordinate system W ₂ ;

(4) According to the actual pose P _Bpr and theoretical pose P _Bpt of each point in the prior knowledge data set B _p , obtain the corrected pose P _Bpr ' corresponding to each point in the prior knowledge data set B _p ;

(5) Use the actual pose P _Bpr of each point in the prior knowledge data set B _p as the input of the robot processing dynamic error compensation model, and the corrected pose P _Bpr ' as the output of the robot processing dynamic error compensation model, Optimize and train the dynamic error compensation model for robot processing;

(6) Use the target pose of the remaining data set C _B B _p as the input of the optimized and trained robot processing dynamic error compensation model, and output the corrected target point set B' from the model, and use the corrected target Point set B' serves as the basis for static and dynamic error compensation in robot grinding and polishing processing.

As a preferred solution of the present invention, in step (6), the target pose of each point in the remaining data set C _B B _p is used as the input of the robot processing dynamic error compensation model optimized and trained in step (5), and the preliminary training Optimize the ability of the trained robot processing dynamic error compensation model to transform the compensation value output into the corrected pose P _CBpr ', so that the difference between the compensation value and the corrected pose P _CBpr ' is within the threshold range, completing the robot processing dynamic error compensation model. Optimization. Among them, the corrected pose P _CBpr ' is the actual pose P _CBpm and the theoretical pose P _CBpt of each point in the remaining data set C _B B _p to obtain the corrected pose _P corresponding to each point in the remaining data set C _B B p _CBpr '. The actual pose P _CBpr is the pose P _CBpm of each point in the remaining data set C _B B _p measured in the measurement coordinate system W ₁ , and then the remaining data set is obtained through the coordinate system transformation matrix T The actual pose P _CBpr of each point in C _B B _p under the robot's global reference coordinate system W ₂ .

More specifically, in the present invention, in order to improve the applicability of training samples, it is necessary to set the number of samples that constitute the target point set A. In principle, the more the number of samples, the more accurate the result. Then the robot grinding and polishing processing space is divided into regions. According to the robot grinding and polishing processing path and the number of sample points that constitute the set, sample points are randomly selected from each interval after the robot grinding and polishing processing space is divided, thereby obtaining the target point set A. Preferably, the number of sample points is selected according to the coverage amount of the robot grinding and polishing processing path. For example, the greater the coverage amount, the more sample points are selected.

Step 7: Perform error compensation on the end position of the robot based on the optimized robot processing dynamic error compensation model. That is, the robot machining dynamic error compensation model is used to correct the robot posture and compensate for the terminal loading deformation error.

Figure 2 shows a flow chart of robot posture and end loading deformation error compensation in the embodiment of the present invention. Through the measurement of the robot grinding and polishing processing coordinate system measurement system, a set of robots with target point set A under the measurement coordinate system _W1 is obtained. At the same time, the end pose P _AM is obtained through the robot software end, and the corresponding theoretical pose P _A in a set of robot global reference coordinate systems W ₂ is obtained. The LM algorithm is used to obtain the measurement coordinate system W ₁ and the robot global The coordinate system transformation matrix T between the reference coordinate system W ₂ . Secondly, for the target point set B to be compensated, a small number of target points are randomly selected to form a priori knowledge data set B _p , and its pose P _Bpm in the measurement coordinate system W ₁ is measured, and the coordinate system is used to transform The matrix T obtains the actual pose P _Bpr of the prior knowledge data set B _p in the robot global reference coordinate system W ₂ , and compares the actual pose P _Bpr with its corresponding theoretical pose P _Bpt . Obtain the corrected pose P _Bpr ' of the corresponding target point in the prior knowledge data set B _p , use the actual coordinates P _Bpr of the target point in the prior knowledge data set B _pre as input, and use its corresponding corrected position Pose P _Bpr ' is used as the output to train a robot machining dynamic error compensation model established by a new support vector machine regression algorithm that integrates prior knowledge. Finally, the set composed of the remaining target points of the target point set B to be compensated, that is, the remaining data set C _B B _p , is used as input, and the model output established by the new support vector machine regression algorithm integrating prior knowledge is corrected. The target point set B' is input into the robot controller to compensate for the end error of the robot.

It is easy for those skilled in the art to understand that the above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements, etc., made within the spirit and principles of the present invention, All should be included in the protection scope of the present invention.

Claims (10)

1. A comprehensive static and dynamic error compensation method for robot grinding and polishing of aircraft composite components, which is characterized by including the following steps: S1 builds a coordinate system measurement system for robot grinding and polishing of aircraft composite components; S2 establishes the global coordinate system of the robot processing system; S3 builds a robot dynamics recursive model based on the global coordinate system of the robot processing system; S4 identifies the static parameters of the robot processing system based on the robot dynamics recursion model, and establishes a robot distance error calibration model based on the static parameters of the robot processing system to compensate for the robot body geometric error; S5 builds a dynamic error compensation model for robot processing in the global coordinate system of the robot processing system based on the geometric error of the robot body; S6 optimizes and trains the robot processing dynamic error compensation model based on the robot processing dynamic error target data set and the prior knowledge data set; S7 performs error compensation on the robot end pose based on the optimized robot processing dynamic error compensation model. 2. A comprehensive static and dynamic error compensation method for robot grinding and polishing of aircraft composite components according to claim 1, characterized in that, in step S1, the robot grinding and polishing coordinate system measurement system includes a laser tracker and a dual Visual measuring equipment. 3. A comprehensive static and dynamic error compensation method for robot grinding and polishing of aircraft composite components according to claim 1, characterized in that, in step S2, the global coordinate system of the robot processing system includes a workpiece coordinate system and a tool coordinate system. , measurement coordinate system and robot global reference coordinate system. 4. A comprehensive static and dynamic error compensation method for robot grinding and polishing of aircraft composite components according to claim 1, characterized in that, in step S2, the global coordinate system of the robot processing system adopts a non-linear method based on discrete differential evolution. Linear optimization algorithm is established. 5. A comprehensive static and dynamic error compensation method for robot grinding and polishing of aircraft composite components according to claim 1, characterized in that, in step S5, a support vector machine regression algorithm integrating prior knowledge is used to construct the dynamic error of robot processing Compensation model. 6. A comprehensive compensation method for static and dynamic errors in robot grinding and polishing of aircraft composite components according to claim 1, characterized in that, in step S6, a chaotic particle swarm optimization algorithm is used to optimize and train the dynamic error compensation model of robot processing. . 7. A comprehensive static and dynamic error compensation method for robot grinding and polishing of aircraft composite components according to claim 3, characterized in that step S6 specifically includes the following steps: S61 selects a set of target points A scattered in the robot grinding and polishing processing space, and measures the pose P _AM of the target point set A in the measurement coordinate system W ₁ through the robot grinding and polishing processing coordinate system measurement system. Obtain the theoretical pose P _A of the target point set A in the robot's global reference coordinate system W ₂ , and then obtain the coordinate system transformation matrix T between the measurement coordinate system W ₁ and the robot's global reference coordinate system W ₂ ; S62 determines a set B of target points to be compensated at the end of the robot, and divides the set B of target points to be compensated into a robot processing dynamic error prior knowledge data set B _p and a remaining data set C _B B _p according to a preset ratio; S63 Measure the target pose P _Bpm of each point in the prior knowledge data set B _p under the measurement coordinate system W ₁ , and then obtain the prior knowledge data set B _p through the coordinate system transformation matrix T. The actual pose P _Bpr of each point in the robot's global reference coordinate system W ₂ ; S64 obtains the corrected pose P _Bpr ' corresponding to each point in the prior knowledge data set B _p according to the actual pose P _Bpr and the theoretical pose P _Bpt of each point in the prior knowledge data set B _p ; S65 uses the actual pose P _Bpr of each point in the prior knowledge data set B _p as the input of the robot processing dynamic error compensation model, and the corrected pose P _Bpr ' as the output of the robot processing dynamic error compensation model. The machining dynamic error compensation model is optimized and trained; S66 uses the target pose of the remaining data set C _B B _p as the input of the optimized and trained robot processing dynamic error compensation model, and the model outputs the corrected target point set B', and the corrected target point set B' serves as the basis for static and dynamic error compensation in robot grinding and polishing. 8. A comprehensive static and dynamic error compensation method for robot grinding and polishing of aircraft composite components according to claim 7, characterized in that, in step S61, the acquisition method of target point set A is as follows: S611 sets the number of sample points that constitute the set; S612 divides the robot grinding and polishing processing space into regions; S613 randomly selects sample points in each interval after the robot grinding and polishing processing space is divided according to the robot grinding and polishing processing path and the number of sample points that constitute the set, thereby obtaining the target point set A. 9. A comprehensive static and dynamic error compensation method for robot grinding and polishing of aircraft composite components according to claim 1, characterized in that in step S7, error compensation for the end position and posture of the robot includes adjusting the end position and attitude of the robot. compensate. 10. A comprehensive static and dynamic error compensation method for robot grinding and polishing of aircraft composite components according to claim 9, characterized in that the robot end position is compensated through a position compensator, and the robot end attitude is compensated through an attitude compensator. accomplish.