Multi-axis hole disassembling method for industrial robot in uncertain maintenance state of fusion mechanism

Abstract

The application discloses a kind of fusion mechanism's uncertain maintenance state industrial robot multi-class shaft hole disassembly method, comprising the following steps: 1) construct industrial robot shaft hole disassembly mechanism model;2) construct industrial robot multi-class shaft hole compliant disassembly simulation scene, realize industrial robot shaft hole disassembly process simulation, obtain the shaft hole disassembly data corresponding to disassembly action;3) construct SAC algorithm-based industrial robot multi-class shaft hole compliant disassembly strategy model;4) change the distance of compliant center to shaft bottom end center, retrain SAC algorithm-based industrial robot multi-class shaft hole compliant disassembly strategy model;Using mechanism model verifies disassembly strategy model, using the model that passes verification carries out compliant disassembly of industrial robot multi-class shaft hole under uncertain maintenance state.The method of the application constructs SAC algorithm-based industrial robot multi-class shaft hole compliant disassembly strategy model, can be better applicable to multi-class shaft hole compliant disassembly process, and have good generalization.

Description

Technical Field

[0001] The present invention relates to intelligent manufacturing technology, and in particular to a method for disassembling multiple types of shaft holes of an industrial robot in an uncertain maintenance state with a fusion mechanism. Background Art

[0002] In recent years, the rapid pace of product upgrades has led to a yearly increase in waste products. Recycling and reusing waste products can recycle manufacturing resources and minimize the environmental impact of waste products. Due to the uncertainty and complexity of disassembling waste products, the process is often performed manually. In recent years, the use of industrial robots to replace manual disassembly has effectively improved the intelligence and efficiency of the disassembly process.

[0003] Shaft and hole disassembly is a common step in the disassembly of scrap products. For industrial robot shaft and hole disassembly, improper process control parameters often result in excessive disassembly forces and torques, which can easily damage the robot's end effector. To address this, properly controlling the process control parameters and achieving smooth shaft and hole disassembly can effectively reduce the forces and torques involved, preventing damage to the robot.

[0004] To achieve compliant disassembly of an industrial robot's shaft hole, numerous physical and geometric parameters must be considered, including the surface friction coefficient of the shaft hole, the relative initial deflection angle of the shaft hole, the shaft hole radius, and the position of the robot's end-effector. Under quasi-static conditions, these physical and geometric parameters directly influence the disassembly force and torque during the disassembly process. Therefore, constructing a model for the disassembly mechanism of an industrial robot's shaft hole can provide a mathematical model foundation and theoretical basis for achieving compliant disassembly of an industrial robot's shaft hole.

[0005] For scrapped products, the uncertainty of their maintenance states during service life leads to uncertainty in physical and geometric parameters such as the surface friction coefficient of the shaft hole and the relative initial deflection angle of the shaft hole. In particular, the disassembly of batch products often involves multiple shaft holes of varying radii and lengths, placing higher demands on the generalization of control parameter decisions for the compliant disassembly of shaft holes in industrial robots. Therefore, a compliant disassembly strategy for multiple shaft holes in industrial robots under uncertain maintenance states is urgently needed to address the problems of a lack of mechanistic models for industrial robot shaft hole disassembly, low disassembly compliance, and poor generalization. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method for disassembling multiple types of shaft holes of an industrial robot in an uncertain maintenance state with a fusion mechanism in response to the defects in the prior art.

[0007] The technical solution adopted by the present invention to solve the technical problem is: a method for disassembling multiple types of shaft holes of an industrial robot in an uncertain maintenance state with a fusion mechanism, comprising the following steps:

[0008] 1) Based on the coupling relationship between the physical and geometric parameters of the shaft hole surface friction coefficient, the relative initial deflection angle of the shaft hole, the shaft hole radius, and the size of the two-point contact area during the shaft hole disassembly process under quasi-static conditions, a shaft hole disassembly mechanism model for industrial robots is constructed;

[0009] 2) Construct a simulation scenario for the compliant disassembly of multiple shafts and holes of an industrial robot, integrate the disassembly parameters of multiple shafts and holes of the industrial robot, control the movement of the target point during the disassembly process of the industrial robot by specifying the target position and target posture, simulate the disassembly process of the industrial robot shaft and hole, and obtain the shaft and hole disassembly data corresponding to the disassembly action, including the disassembly force / torque, the relative position of the shaft and hole, and the relative posture of the shaft and hole;

[0010] The shaft hole disassembly parameters include disassembly action, disassembly force / torque, shaft hole relative position, and shaft hole relative posture;

[0011] 3) Construct a flexible disassembly strategy model for multiple types of shaft holes of industrial robots based on the SAC algorithm;

[0012] 4) Change the distance L from the compliance center to the center of the shaft bottom C , retrain the industrial robot multi-type shaft hole flexible disassembly strategy model based on the SAC algorithm; use the trained strategy model to disassemble the shaft hole at different distances L from the flexible center to the shaft bottom center. C , shaft hole radius R / r, shaft hole friction coefficient μ, and shaft hole relative initial deflection angle θ0 parameters, the constructed mechanism model is used to verify the retrained industrial robot multi-type shaft hole flexible disassembly strategy model, retain the model that meets the consistency verification, and use the verified model to perform multi-type shaft hole flexible disassembly of industrial robots under uncertain maintenance conditions.

[0013] According to the above scheme, the industrial robot shaft hole disassembly mechanism model constructed in step 1) is as follows:

[0014]

[0015] Among them, α, β, and γ are represented by the following formulas respectively;

[0016]

[0017] Among them, Lh represents the size of the contact area between the shaft and the hole, L C is the distance from the compliance center to the center of the shaft bottom, R is the hole radius, r is the shaft radius, d is the shaft diameter, D is the hole diameter, μ is the friction coefficient of the shaft hole surface, θ0 is the relative initial deflection angle of the shaft hole, and Lh is the size of the contact area between the two points of the shaft hole; Kx represents the lateral stiffness, K θ represents the angular stiffness, λ represents, δ0 represents the lateral error between the manipulator and the shaft hole, β0 represents the angular error between the manipulator and the shaft hole, and θ0 represents the initial tilt angle of the shaft; h1 and h2 represent the start and end positions of the two-point contact, respectively;

[0018] c and λ are expressed as follows;

[0019]

[0020] According to the above solution, in step 2), the industrial robot shaft hole disassembly process simulation is realized to obtain the shaft hole disassembly data corresponding to the disassembly action. The specific steps are as follows:

[0021] 2.1) Build a cube hole base model in the CoppeliaSim simulation environment. Construct a 3D model of the hole by combining a tubular body and a cube. Deploy the UR5 robot model integrated in CoppeliaSim into the environment. Build a cylindrical shaft model in the CoppeliaSim simulation environment and attach one end of it to the force / torque sensor at the end of the UR5. Generate a dummy point, UR5-ikTip, and construct an inverse kinematic relationship between the shaft and the UR5. Attach the dummy point to the shaft's central axis, making it a child of the shaft. Control the movement of the industrial robot during disassembly by specifying the target position and posture of the UR5-ikTip at the end of the robot.

[0022] 2.2) Enable data exchange between the CoppeliaSim server and the Pycharm client. Remotely control the movement of the UR5 robot in CoppeliaSim by transmitting the shaft-hole disassembly action containing the target position and target attitude data in Pycharm. After the industrial robot in the CoppeliaSim simulation environment completes the disassembly motion, the shaft-hole disassembly data (such as disassembly force / torque, shaft-hole relative position, shaft-hole relative attitude, etc.) is transmitted from the CoppeliaSim simulation environment to the Pycharm environment.

[0023] According to the above scheme, step 3) is aimed at the disassembly process of multiple types of shaft holes, combined with the SAC deep reinforcement learning algorithm, to build an industrial robot multi-type shaft hole flexible disassembly strategy model based on the SAC algorithm. The specific steps are as follows:

[0024] 3.1) Combine the disassembly force / torque, shaft-hole relative position, shaft-hole relative posture, industrial robot position offset and angle, and shaft-hole disassembly distance data to construct the state s of the SAC algorithm t 、Action a t and reward r t ; Use the simulation environment in step 2) to give the current state st and action a t The new state s t+1 ;

[0025] 3.2) Construct a neural network model of the SAC algorithm as a model for the flexible disassembly strategy of multiple types of shaft holes of industrial robots based on the SAC algorithm.

[0026] The neural network model consists of an input layer, a hidden layer, an output layer, and an activation function. The input of the actor network is the current state, and the output is the disassembly action performed on the current state. The input of the critic network is a tensor formed by the current state and the action at, and the output is the reward calculation value.

[0027] Among them, the current state performs disassembly action a t The subsequent update status and disassembly force / torque, shaft-hole relative position, and shaft-hole relative posture are given using the CoppeliaSim simulation environment;

[0028] 3.3) The disassembly action sequence is obtained through the model as a flexible disassembly strategy for multiple types of shaft holes of industrial robots.

[0029] According to the above scheme, in step 3.1), the state, action and reward of the SAC algorithm are constructed as follows:

[0030] The state is represented as a one-dimensional vector with a length of 12, rp x 、rp y and rp z Respectively represent the relative positions of the shaft hole in the x, y and z directions, ra α 、ra β and ra γ Respectively represent the relative deflection angles of the shaft hole in the x, y and z directions, f x 、f y 、f z 、m x 、m y and m z are the forces and moments in the x, y, and z directions read by the force / torque sensor, respectively, and are expressed as:

[0031]

[0032] Among them, F max is the force threshold 1, and the size is set to 60N; M max is the torque threshold 1, and its value is set to 2.5Nm;

[0033] Action is a one-dimensional vector of length 6, po x 、po y 、po z 、aox 、ao y Heao z They are the position offset and angle offset of the robot end (UR5-ikTip), respectively, expressed as:

[0034] action(a)=[po x ,po y ,po z ,ao x ,ao y ,ao z ];

[0035] The reward is expressed as a weighted sum of penalty terms, where P F and P M Denote the penalty terms of force and torque respectively, P p The penalty term for the relative position between the shaft and the hole in the x and y directions, P a P represents the penalty term for the relative deflection angle between the shaft and the hole in the x, y and z directions. Z Represents the penalty term for the position offset of the axis in the z direction, R H Indicates the bonus item for the axis disassembly distance, a_n indicates the total number of actions in the current round, and m_a_n indicates the maximum number of actions allowed in a single round. i is the weight coefficient of the corresponding penalty term in the reward;

[0036] reward(r)=σ1P F +σ2P M +σ3P p +σ4P a +σ5P Z +σ6R H

[0037]

[0038] P a =||ra α ,ra β ,ra γ ||

[0039]

[0040] Among them, F max ' is the force threshold 2, the size is set to 40N, M max ' is the torque threshold 2, which is set to 1.5Nm.

[0041] The present invention also provides an electronic device, comprising: one or more processors; and a storage device for storing one or more programs.

[0042] When the one or more programs are executed by the one or more processors, the one or more processors execute the method described in any one of the above solutions.

[0043] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and the method described in any one of the above solutions when the computer program is executed by a processor.

[0044] The beneficial effects produced by the present invention are:

[0045] 1. A model of the industrial robot shaft hole disassembly mechanism was constructed, providing mathematical model support and theoretical basis for the construction of a flexible disassembly strategy model for multiple types of shaft holes in industrial robots;

[0046] 2. A simulation environment for the flexible disassembly process of various shaft holes of industrial robots was built for data interaction. An online training environment for the flexible disassembly strategy of various shaft holes of industrial robots was built;

[0047] 3. Construct a multi-type shaft hole flexible disassembly strategy model for industrial robots based on the SAC algorithm, which can be well applied to the flexible disassembly process of multiple types of shaft holes and has good generalization. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0049] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0050] Figure 2 is a schematic diagram of a shaft hole compliant support model according to an embodiment of the present invention;

[0051] Figure 3 Schematic diagram of four contact states of the shaft hole in an embodiment of the present invention;

[0052] Figure 4 Schematic diagram of initial error of shaft hole disassembly according to an embodiment of the present invention;

[0053] Figure 5 This is a force diagram of the shaft hole disassembly process according to an embodiment of the present invention;

[0054] Figure 6 Schematic diagram of the functional relationship between the geometric and physical parameters of an embodiment of the present invention and the two-point contact area between the shaft and the hole;

[0055] Figure 7 This is a schematic diagram of the online training environment for constructing a flexible disassembly strategy for multiple types of shaft holes of an industrial robot according to an embodiment of the present invention;

[0056] Figure 8 This is a flow chart of the SAC algorithm for compliant disassembly of an industrial robot shaft hole according to an embodiment of the present invention;

[0057] Figure 9 Schematic diagram of the SAC training process according to an embodiment of the present invention;

[0058] Figure 10 This is a diagram for verifying the compliance and generalization of the flexible disassembly strategy model for multiple types of shaft holes of an industrial robot according to an embodiment of the present invention;

[0059] Figure 11 2 is a schematic diagram comparing the SAC and TD3 algorithms according to an embodiment of the present invention;

[0060] Figure 12 is L in the embodiment of the present invention C Box plots of parameter mechanism verification;

[0061] Figure 13 d and d parameter mechanism verification box plots of an embodiment of the present invention;

[0062] Figure 14 2 is a box plot of the μ and θ0 parameter mechanism verification of an embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0064] like Figure 1 As shown, a method for disassembling multiple types of shaft holes of an industrial robot in an uncertain maintenance state with a fusion mechanism includes the following steps:

[0065] 1) Based on the coupling relationship between the physical and geometric parameters (including the friction coefficient of the shaft hole surface, the relative initial deflection angle of the shaft hole, and the shaft hole radius) and the disassembly force during the shaft hole disassembly process under quasi-static conditions, a model of the industrial robot shaft hole disassembly mechanism is constructed;

[0066] In step 1), the force on the shaft and the geometric constraint relationship between the shaft and the hole during the disassembly process are analyzed to construct the distance L from the compliance center to the center of the shaft bottom. C The mathematical relationship between the shaft hole radius R / r, the shaft hole surface friction coefficient μ, the shaft hole relative initial deviation angle θ0 and the size of the two-point contact area of ​​the shaft hole Lh was established, and a shaft hole disassembly mechanism model of an industrial robot based on mathematical description was established.

[0067] The specific process is as follows:

[0068] 1.1) Shaft-hole compliant support model Figure 2 As shown. Among them, O c Indicates the center of compliance, Kx and K θ are the lateral stiffness and angular stiffness respectively, L C is the distance from the compliance center to the center of the shaft bottom, F Z 、F X and M represent the force and moment on the shaft respectively, and θ represents the inclination angle of the shaft. There are four contact states of the shaft hole during the disassembly process, such as Figure 3 As shown. Among them, two-point contact is most likely to cause wedging and blocking during the shaft hole disassembly process, resulting in excessive disassembly force. Therefore, in order to protect the end effector of the industrial robot from damage, the possibility of two-point contact during the shaft hole disassembly process should be minimized. During the industrial robot shaft hole disassembly process, there are two types of position errors: lateral error δ0 and angular error β0. As shown Figure 4 As shown in Figure 1, the initial tilt angle of the axis is θ0. When there is a lateral error δ0 between the robot arm and the shaft hole, an angular error β0 will be generated. Due to the existence of the initial error, when the axis is grasped by the robot arm, it will perform translational and rotational motion around the compliance center.

[0069] 1.2) When the industrial robot disassembles the shaft hole, due to the influence of the lateral error δ0 and the angular error β0, the shaft will translate and rotate, resulting in a two-point contact state. Figure 4 and Figure 5 As shown, the initial horizontal distances from the compliance center and the bottom center of the shaft to the axis of the hole are U0 and ε0 respectively. The shaft-hole disassembly process in this patent is assumed to be a quasi-static process. During the shaft-hole disassembly process, the values ​​of U and ε will change continuously. t where d = 2r and f = 2r are the shaft-hole length and the disassembly depth at time t, respectively. The diameters of the shaft and hole are d (d = 2r) and D (D = 2R), respectively. The support forces on the shaft from the hole are f1 and f2, and the friction forces between them are μf1 and μf2, where μ is the coefficient of friction on the shaft-hole surface and M represents the torque on the shaft.

[0070] 1.3) During the shaft-hole disassembly process, equation (1) describes the clearance ratio between the shaft and the hole; the geometric constraints between the shaft and the hole are described by equations (2)-(5). Based on the approximate calculation of equation (6), equation (5) can be simplified to equation (7). Combining equations (2) and (3) yields equation (8).

[0071]

[0072] U0=L C θ0-ε0 (2)

[0073] U=L C θ-cR (3)

[0074] ε0=R-(D-Hsinθ0-dcosθ0)-rcosθ0 (4)

[0075] 2R=h t sinθ+2rcosθ (5)

[0076] cosθ≈1,cosθ0≈1,sinθ≈0,sinθ0≈0 (6)

[0077] h t θ=2(Rr) (7)

[0078] U0-U=L C θ0-ε0-L C θ+cR (8)

[0079] In order to simplify the calculation, define intermediate variables:

[0080] λ=L C θ0-ε0+cR (9)

[0081] The force analysis of the shaft is carried out. According to the horizontal force balance and moment balance relationship of the shaft hole, formulas (10) and (11) can be obtained:

[0082] f2+K x (U0-U+δ0)=f1 (10)

[0083] K x (U0-U+δ0)L C +K θ (θ-θ0+β0)+μrf1=f1h t +μrf2 (11)

[0084] When the shaft-hole two-point contact state is about to disappear:

[0085] f2=0 (12)

[0086] Substituting formulas (8) and (12) into formulas (10) and (11), we can obtain:

[0087]

[0088] Combining formulas (7) and (13), we can obtain:

[0089]

[0090] Simplifying formula (14), we can get the disassembly depth h t The quadratic equation of :

[0091] αh t 2 +βht +γ=0 (15)

[0092] in,

[0093]

[0094] The solution of formula (15) can be described by formula (17):

[0095]

[0096] If formula (17) has no real number solution, it indicates that the two-point contact state between the shaft and the hole will not occur; otherwise, the two solutions h1 and h2 of the equation represent the starting and ending positions of the two-point contact, respectively.

[0097] 1.4) Assume H = 40 mm, K x =4N / mm,K θ =30Nmm / rad,δ0=2mm,β0=0rad, define Lh as the size of the contact area between the shaft and the hole, as shown in formula (18). At this time, the size of the contact area between the shaft and the hole, Lh, is only related to L C , D, d, μ and θ0 are related:

[0098]

[0099] In formula (18), if D, d, θ0 and μ are 25.2 mm, 25 mm, 0 rad and 0.15 respectively, Lh and L can be obtained. C The functional relationship is shown in formula (19); if d, L C , θ0 and μ are 25mm, 20mm, 0rad and 0.15 respectively, the functional relationship between Lh and D can be obtained as shown in formula (20); if D, L C , θ0 and μ are 25.2mm, 20mm, 0rad and 0.15 respectively, and the functional relationship between Lh and d can be obtained as shown in formula (21); if D, d, θ0 and L C are 25.2 mm, 25 mm, 0 rad and 0 mm respectively, the functional relationship between Lh and μ can be obtained as shown in formula (22); if D, d, μ and L C The functional relationship between Lh and θ0 is 25.2mm, 25mm, 0.5 and 0mm respectively, as shown in formula (23). At the same time, the functional curve of formula (19)-(23) is as follows Figure 6 shown.

[0100] Lh=f(L C )=0.8264L C 2 +3.4091L C +6.2429 (19)

[0101]

[0102] Lh = f(μ) = 156.25μ 2 + 2.73 (22)

[0103]

[0104] Formula (19) and Figure 6 (a) shows that when L C > 0, Lh decreases as L C decreases, that is, reducing the distance between the compliant center and the center of the bottom end of the shaft can reduce the size of the two - point contact area Lh between the shaft and the hole. Formulas (20) - (21), Figure 6 (b) and 6(c) show that when 25 mm < D < 29 mm, Lh decreases as D increases; when 21.2 mm < d < 25.2 mm, Lh increases as d increases, that is, when the diameters of the shaft and the hole are closer, the two - point contact area Lh between the shaft and the hole will increase. Formulas (22) - (23), Figure 6 (d) and 6(e) show that when 0 rad < θ0 < 0.005 rad, 0 < μ < 0.005, Lh increases as θ0 and μ increase, that is, a poor maintenance state (a large friction coefficient on the shaft - hole surface and a relative initial angular deviation between the shaft and the hole) will increase the size of the two - point contact area Lh between the shaft and the hole. When the two - point contact state appears between the shaft and the hole, a large disassembly force will appear between them, and the magnitude of the disassembly force can reflect the size of the two - point contact area Lh between the shaft and the hole.

[0105] 2) Construct a multi - type shaft - hole compliant disassembly simulation scenario for industrial robots, integrate multi - type shaft - hole disassembly parameters of industrial robots, and control the movement of the target point during the disassembly process of the industrial robot by specifying the target position and target posture, so as to realize the simulation of the shaft - hole disassembly process of the industrial robot and obtain the shaft - hole disassembly data corresponding to the disassembly action, including disassembly force / torque, relative position between the shaft and the hole, and relative posture between the shaft and the hole;

[0106] The shaft - hole disassembly parameters include disassembly action, disassembly force / torque, relative position between the shaft and the hole, and relative posture between the shaft and the hole;

[0107] The specific operations are as follows:

[0108] 2. In Solidworks, establish a tubular model and import it into the CoppeliaSim simulation environment in the format of an STL file. In the CoppeliaSim simulation environment, construct a cubic - hole base model, and construct a three - dimensional model of the hole by combining the tubular body and the cube, as Figure 7As shown in the figure. The UR5 robot model is integrated into CoppeliaSim, and the end of the robot arm is equipped with a force / torque sensor, so it can be directly deployed in the environment. In the CoppeliaSim simulation environment, a cylindrical axis model is constructed and one end is fixed to the force / torque sensor at the end of the UR5. A dummy point UR5-ikTip is generated, as shown in the figure. Figure 7 As shown in the figure, the inverse kinematic relationship between the axis and the UR5 is established and fixed on the axis's centerline, making it a child of the axis. By specifying the target position and posture of the UR5-ikTip at the end of the robot, the movement of the industrial robot during the disassembly process is controlled.

[0109] 2.2) Call the sim.RemoteAPI.start function in conjunction with the CoppeliaSim simulation environment, and call the sim.simxStart and sim.simxStartSimulation functions in Pycharm to start data interaction between the CoppeliaSim server and the Pycharm client. The sim.rmlMoveToPositions function is a spatial path planning function in the RML library. By specifying the target position and target posture, the movement of the target point in the disassembly process of the industrial robot can be controlled. In Pycharm, call the sim.simxSetIntegerSignal and sim.simxSetFloatSignal functions to implement the shaft hole disassembly action. t Data transmission from Pycharm environment to CoppeliaSim simulation environment; in CoppeliaSim simulation environment, the shaft hole disassembly action a is read through sim.getIntegerSignal and sim.getFloatSignal functions t Finally, the sim.rmlMoveToPositions function is used to realize the movement of the target point in the disassembly process of the industrial robot. The UR5-ikTip point is set as the target point, and the inverse kinematic relationship between it and the UR5 industrial robot is established. By transferring the shaft hole disassembly action containing the target position and target posture data in Pycharm t, you can remotely control the movement of the UR5 robot in CoppeliaSim. After the industrial robot completes the disassembly movement in the CoppeliaSim simulation environment, call the sim.simxReadForceSensor, sim.simxGetObjectionPosition, and sim.simxGetObjectionOriebtation functions in the PyCharm environment to transfer the shaft-hole disassembly data (such as disassembly force / torque, shaft-hole relative position, and shaft-hole relative orientation) from the CoppeliaSim simulation environment to the PyCharm environment.

[0110] 3) Construct a flexible disassembly strategy model for multiple types of shaft holes of industrial robots based on the SAC algorithm;

[0111] The specific steps are as follows:

[0112] 3.1) Combine the disassembly force / torque, shaft-hole relative position, shaft-hole relative posture, industrial robot position offset and angle, and shaft-hole disassembly distance data to construct the state s of the SAC algorithm t 、Action a t and reward r t ; Use the simulation environment in step 2) to give the current state s t and action a t The new state s t+1 ;

[0113] In step 3.1), construct the state, action, and reward of the SAC algorithm as follows:

[0114] The state is represented as a one-dimensional vector with a length of 12, rp x 、rp y and rp z Respectively represent the relative positions of the shaft hole in the x, y and z directions, ra α 、ra β and ra γ Respectively represent the relative deflection angles of the shaft hole in the x, y and z directions, f x 、f y 、f z 、m x 、m y and m z are the forces and moments in the x, y, and z directions read by the force / torque sensor, respectively, and are expressed as:

[0115]

[0116] Among them, F max is the force threshold 1, and the size is set to 60N; M maxis the torque threshold 1, and its value is set to 2.5Nm;

[0117] Action is a one-dimensional vector of length 6, po x 、po y 、po z 、ao x 、ao y Heao z They are the position offset and angle offset of the robot end (UR5-ikTip), respectively, expressed as:

[0118] action(a)=[po x ,po y ,po z ,ao x ,ao y ,ao z ];

[0119] The reward is expressed as a weighted sum of penalty terms, where P F and P M Denote the penalty terms of force and torque respectively, P p The penalty term for the relative position between the shaft and the hole in the x and y directions, P a P represents the penalty term for the relative deflection angle between the shaft and the hole in the x, y and z directions. Z Represents the penalty term for the position offset of the axis in the z direction, R H Indicates the bonus item for the axis disassembly distance, a_n indicates the total number of actions in the current round, and m_a_n indicates the maximum number of actions allowed in a single round. i is the weight coefficient of the corresponding penalty term in the reward;

[0120] reward(r)=σ1P F +σ2P M +σ3P p +σ4P a +σ5P Z +σ6R H

[0121]

[0122] P a =||ra α ,ra β ,ra γ ||

[0123]

[0124] Among them, F max ' is the force threshold 2, the size is set to 40N, M max' is the torque threshold 2, which is set to 1.5Nm.

[0125] In this embodiment, σ1=σ2=-0.5, σ3=-5000, σ4=-12.5, σ5=-0.3, σ6=1. If a_n>m_a_n, P F =2, P M =2, P p >0.0003 or P a If it is >0.04, it means that the disassembly task of the current round has failed, the environment of the SAC algorithm is reset, and a new disassembly process starts again.

[0126] 3.2) Construct a neural network model of the SAC algorithm as a strategy model for the flexible disassembly of multiple types of shaft holes of industrial robots based on the SAC algorithm. The process of the SAC algorithm for the flexible disassembly of shaft holes of industrial robots is as follows: Figure 8 ;

[0127] The neural network model consists of an input layer, a hidden layer, an output layer, and an activation function. The input of the actor network is the current state, and the output is the disassembly action performed on the current state. The input of the critic network is a tensor formed by the current state and the action at, and the output is the reward calculation value.

[0128] Among them, the current state performs disassembly action a t The subsequent update status and disassembly force / torque, shaft-hole relative position, and shaft-hole relative posture are given using the CoppeliaSim simulation environment;

[0129] The actor network consists of an input layer (12 input neurons, 512 output neurons), three hidden layers (512 neurons in both input and output), and an output layer (512 input neurons, 6 output neurons). The critic network consists of an input layer (18 input neurons, 512 output neurons), one hidden layer (512 input neurons, 256 output neurons), and an output layer (256 input neurons, 1 output neuron). The training hyperparameters for the actor and critic networks of the SAC algorithm are shown in Table 1. The axis and hole parameters of the training network are selected as number ① in Table 2.

[0130] Table 1 SAC hyperparameters

[0131]

[0132] Table 2 Shaft hole parameters

[0133]

[0134]

[0135] The training process is as follows: At the beginning of training, the agent action a t Generated by a random function. When the total number of actions of the agent during the entire training process exceeds the threshold 256, its action a t The actor neural network is based on the current state s t Generate. When the agent performs action a t Afterwards, the environment updates the state s t+1 , update the done value and give the reward calculation value r t In the process of training SAC, the interaction between the agent and the environment will continuously generate five-tuple experience t ,a t ,s t+1 ,r t ,done> is stored in the experience pool E. When the number of experiences stored in the experience pool E exceeds the threshold of 256, the parameters of the actor and critic neural networks in SAC are updated by randomly sampling a fixed number of 256 in the experience pool E. After the saving conditions are met, the industrial robot multi-class shaft hole flexible disassembly strategy model based on the SAC algorithm will be saved. The reward, temperature coefficient and loss function curves during the SAC training process are as follows: Figure 9 shown.

[0136] 3.3) The disassembly action sequence is obtained through the model as a flexible disassembly strategy for multiple types of shaft holes of industrial robots.

[0137] The model trained in step 3) is tested for its flexibility and generalization in completing the flexible disassembly of various types of shaft holes of industrial robots. Select shaft hole parameters ①, ②, ③, ④ and ⑤, and record the disassembly force / torque and disassembly depth h during the disassembly process. t relationship, such as Figure 10 As shown. In the same training environment as the SAC algorithm, the TD3 algorithm is used to train the policy model (TwinDelayed Deep Deterministic policy gradient algorithm) for comparison. The reward curves of the two algorithms during training are shown as follows: Figure 11 Using the trained strategy model, we conducted tests under different shaft-hole parameters (①, ⑥, ⑦, and ⑧). We recorded the accumulated absolute force and torque values ​​(F (N) and T (Nm)) during each disassembly process. The average values ​​of F (N) and T (MT (Nm)) over 200 repeated experiments were calculated. The results are shown in Table 3. The magnitudes of MF and MT, respectively, reflect the magnitude of the disassembly force and torque, and are used to evaluate the flexibility of different algorithms in completing shaft-hole disassembly.

[0138] Table 3 Algorithm comparison​

[0139]

[0140] 4) Change the distance L from the compliance center to the center of the shaft bottom C , retrain the industrial robot multi-type shaft hole flexible disassembly strategy model based on the SAC algorithm; use the trained strategy model to disassemble the shaft hole at different distances L from the flexible center to the shaft bottom center. C , shaft hole radius R / r, shaft hole friction coefficient μ, and shaft hole relative initial deflection angle θ0 parameters, the constructed mechanism model is used to verify the retrained industrial robot multi-type shaft hole flexible disassembly strategy model, retain the model that meets the consistency verification, and use the verified model to perform multi-type shaft hole flexible disassembly of industrial robots under uncertain maintenance conditions.

[0141] Select the shaft hole parameter ⑨, the distance L from the center of the flexible center to the center of the shaft bottom C The strategy model is trained under four different random seeds, which are set to 0mm, 20mm and 40mm respectively. C The relationship between the two: select the shaft hole parameter ⑨ to carry out the shaft hole disassembly experiment, and calculate the average value MF (N) and MT (Nm) of F and T under 200 repeated experiments. The MF and MT data under four random seeds are shown in Table 4. The F and T data under four random seeds are combined to obtain the relationship between L C Mechanism verification box line Figure 12 ; Verify the relationship between Lh, D and d: Select the shaft hole parameters ⑩ The shaft hole disassembly experiment was carried out, and the MF and MT data under four random seeds are shown in Tables 5 and 6. The F and T data under four random seeds were combined to obtain the mechanism verification box line about D and d Figure 13 ; Verify the relationship between Lh, μ and θ0: Select the shaft and hole parameters The shaft hole disassembly experiment was carried out, and the MF and MT data under four random seeds are shown in Tables 7 and 8. The F and T data under four random seeds are combined to obtain the mechanism verification box line about μ and θ0 Figure 14 .

[0142] Table 4L C Mechanism verification (shaft hole parameters ⑨)

[0143]

[0144] Table 5D Mechanism verification (shaft hole parameters ⑩)

[0145]

[0146] Table 6d Mechanism verification (shaft hole parameters )

[0147]

[0148] Table 7μ mechanism verification (axis hole parameters )

[0149]

[0150] Table 8 θ0 mechanism verification (axis hole parameters )

[0151]

[0152] The model of this method uses the magnitude of the disassembly force to reflect the size of the contact area Lh between the shaft and the hole. The mechanism model shows that the size of the contact area Lh between the shaft and the hole increases with L C The simulation results show that reducing L C The disassembly force can be reduced; the closer the diameters D and d of the shaft hole are, the greater the disassembly force; and large values ​​of θ0 and μ increase the disassembly force. Therefore, the simulation results of this method are consistent with the conclusions drawn from the mechanism model and can be used to generate a compliant disassembly strategy for multiple shaft holes in industrial robots under uncertain maintenance conditions.

[0153] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A disassembly method for multiple types of shaft holes of industrial robots in uncertain maintenance states with a fusion mechanism, characterized by: The following steps are involved: 1) A model of the industrial robot's shaft hole disassembly mechanism is constructed based on the coupling relationship between the physical and geometric parameters (including the shaft hole surface friction coefficient, the relative initial deflection angle of the shaft hole, the shaft hole radius) and the size of the two-point contact area during the shaft hole disassembly process under quasi-static conditions. 2) Construct a simulation scenario for the compliant disassembly of multiple shafts and holes using an industrial robot. This scenario integrates the disassembly parameters of multiple shafts and holes in the industrial robot. By specifying the target position and target attitude, the motion of the target point in the disassembly process of the industrial robot is controlled. This allows for the simulation of the shaft and hole disassembly process and the acquisition of the corresponding shaft and hole disassembly data, including the disassembly force / torque, the relative position of the shaft and hole, and the relative attitude of the shaft and hole. The shaft hole disassembly parameters include disassembly action, disassembly force / torque, shaft hole relative position, and shaft hole relative posture; 3) Construct a flexible disassembly strategy model for multiple types of shaft holes in industrial robots based on the SAC algorithm; 4) Change the distance from the compliance center to the center of the shaft bottom, and retrain the industrial robot multi-type shaft hole compliance disassembly strategy model based on the SAC algorithm; use the constructed mechanism model to verify the retrained industrial robot multi-type shaft hole compliance disassembly strategy model, retain the model that meets the consistency verification, and use the verified model to perform multi-type shaft hole compliance disassembly of the industrial robot under uncertain maintenance conditions.

2. The method for disassembling multiple types of shaft holes of an industrial robot in an uncertain maintenance state with a fusion mechanism according to claim 1 is characterized in that: The industrial robot shaft hole disassembly mechanism model constructed in step 1) is as follows: ； Among them, α, β, and γ are represented by the following formulas respectively; in, Lh Indicates the size of the contact area between the shaft and the hole. LC is the distance from the compliance center to the center of the shaft bottom, R is the hole radius, r is the shaft radius, d Indicates the shaft diameter, D Indicates the diameter of the hole, μ is the friction coefficient of the shaft hole surface, θ0 is the relative initial deflection angle of the shaft hole; Kx represents the lateral stiffness, Kθ represents the angular stiffness, δ0 Indicates the lateral error between the robot arm and the shaft hole, β0 Indicates the angular error between the robot arm and the shaft hole; h1 and h2 Respectively represent the starting and ending positions of the two-point contact; c and λ are expressed as follows; ； 。 3. The method for disassembling multiple types of shaft holes of an industrial robot in an uncertain maintenance state with a fusion mechanism according to claim 1 is characterized in that: In step 2), the industrial robot shaft hole disassembly process is simulated to obtain shaft hole disassembly data corresponding to the disassembly action. The specific steps are as follows: 2.1) Build a cube hole base model in the CoppeliaSim simulation environment. Create a 3D model of the hole by combining a tubular body and a cube. Deploy the UR5 robot model integrated in CoppeliaSim into the environment. Build a cylindrical shaft model in the CoppeliaSim simulation environment and attach one end of it to the force / torque sensor at the end of the UR5. Generate a dummy point, UR5-ikTip, and construct an inverse kinematic relationship between the shaft and the UR5. Attach the dummy point to the shaft's central axis, making it a child of the shaft. Control the movement of the industrial robot during disassembly by specifying the target position and posture of the UR5-ikTip at the end of the robot. 2.2) Enable data exchange between the CoppeliaSim server and the Pycharm client. Remotely control the movement of the UR5 robot in CoppeliaSim by transmitting the shaft-hole disassembly action containing the target position and target posture data in Pycharm. After the industrial robot in the CoppeliaSim simulation environment completes the disassembly action, the shaft-hole disassembly data is transferred from the CoppeliaSim simulation environment to the Pycharm environment.

4. The method for disassembling multiple types of shaft holes of an industrial robot in an uncertain maintenance state with a fusion mechanism according to claim 1 is characterized in that Step 3) is aimed at the disassembly process of multiple types of shaft holes. In combination with the SAC deep reinforcement learning algorithm, a SAC algorithm-based industrial robot multi-type shaft hole flexible disassembly strategy model is constructed. The specific steps are as follows: 3.1) Combine the disassembly force / torque, shaft-hole relative position, shaft-hole relative posture, industrial robot position offset and angle, and shaft-hole disassembly distance data to construct the state, action, and reward of the SAC algorithm. Use the simulation environment from step 2) to generate the new state of the current state and action. 3.2) Construct a neural network model of the SAC algorithm as a model for the flexible disassembly strategy of multiple types of shaft holes of industrial robots based on the SAC algorithm. 3.3) The disassembly action sequence is obtained through the model as a compliant disassembly strategy for multiple types of shaft holes of industrial robots.

5. The method for disassembling multiple types of shaft holes of an industrial robot in an uncertain maintenance state with a fusion mechanism according to claim 1 is characterized in that: In step 3.1), the state, action, and reward of the SAC algorithm are constructed as follows: The state is represented as a one-dimensional vector with a length of 12. rp x 、 rp y and rp z Respectively indicate the shaft hole x 、 y and z Relative position in direction, ra α 、 ra β and ra γ Respectively indicate the shaft hole x 、 y and z The relative deflection angle in the direction, f x 、 f y 、 f z 、 m x 、 m y and m z Read by force / torque sensor x 、 y and z The forces and moments in the directions are expressed as: in, Fmax is the first force threshold set; Mmax is the first torque threshold to be set; Action is a one-dimensional vector of length 6. po x 、 po y 、 po z 、 ao x 、 ao y and ao z are the position offset and angle offset of the robot's end, respectively, expressed as: ； The reward is expressed as a weighted sum of penalty terms, where P F and P M denote the penalty terms of force and torque respectively, P p Indicates the distance between the shaft and the hole x and y The penalty term for the direction relative position, P a Indicates the distance between the shaft and the hole x 、 y and z The penalty term for the direction relative to the deflection angle, P Z Indicates that the axis is z The penalty term for position offset in the direction, R H Reward item indicating the distance of shaft disassembly, a _ n Indicates the total number of actions in the current round. m _ a _ n represents the maximum number of actions allowed in a single round; σ i is the weight coefficient of the corresponding penalty term in the reward; in, F max ′ The second force threshold is set to M max ′ The second torque threshold is set 。 6. The method for disassembling multiple types of shaft holes of an industrial robot in an uncertain maintenance state with a fusion mechanism according to claim 1 is characterized in that: In step 3.2), the neural network model includes an input layer, a hidden layer, an output layer, and an activation function; the input of the actor network is the current state, and the output is the disassembly action performed by the current state; the input of the critic network is the current state and the action a t The combined tensor output is the reward calculation value; Among them, the disassembly action a is performed based on the current state t The subsequent update status and disassembly force / torque, shaft-hole relative position, and shaft-hole relative posture are given using the CoppeliaSim simulation environment.

7. An electronic device, characterized in that: include: one or more processors; as well as a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors are enabled to perform the method according to any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images