CN112247898B - Robot non-rigid body assembling method based on deflection analysis - Google Patents

Abstract

A robot non-rigid body assembly method based on deflection analysis is characterized in that a contact force sense model provided based on deflection analysis accords with contact deformation characteristics of non-rigid body parts in the assembly process, and the deflection angle theta and the deflection amount d of a shaft can be well predicted_x，d_y(ii) a Based on the contact state recognition model and the contact force sense model, the current contact state and the current posture of the part can be better reflected, and the part assembly can be realized more quickly.

Description

Robot non-rigid body assembling method based on deflection analysis

Technical Field

The invention belongs to the field of robot assembly and control, and is suitable for the field of robot non-rigid body part assembly. In particular to a robot non-rigid body assembling method based on deflection analysis.

Background

In recent years, with rapid development of industrial robot industry and related technologies, industrial robots are widely used in various fields such as automated assembly, manufacturing, and the like. In the field of automatic assembly, the materials of the assembly parts are gradually diversified, and the assembly parts are not limited to metal materials with high rigidity and hard texture, such as non-rigid materials of plastics, rubber and the like, and are widely applied. Unlike rigid bodies, non-rigid body parts are prone to elastic deformation during assembly contact, and such deformation can cause inaccurate force feedback information to affect overall assembly, so that it is necessary to study the assembly method of the non-rigid body parts of the robot.

The existing robot part assembling method is to combine a contact force sense model and robot inverse kinematics analysis to realize the assembling task of parts. However, the existing contact force sense model is only suitable for rigid body parts, that is, deformation of the parts is negligible in the assembling and contacting process, and is not suitable for non-rigid body parts which generate non-negligible deformation in the assembling and contacting process.

Disclosure of Invention

In order to solve the problem that a contact force sense model is not suitable for a non-rigid part in the existing robot assembly method, the invention provides a non-rigid part assembly method of a robot based on deflection analysis.

The technical scheme adopted by the invention for solving the technical problems is as follows:

a robot non-rigid body assembling method based on deflection analysis comprises the following steps:

step 1: the shaft hole assembly is divided into three states: a single point contact state, a two point contact state, and a planar contact state. Acquiring force sense data aiming at different contact states, analyzing the force sense data characteristics of the different states, and establishing a non-rigid body assembling contact body state recognition model by combining a machine learning algorithm and statics characteristics;

step 2: designing a non-rigid contact force sense model based on deflection analysis according to different contact states,

single-point contact state and two-point contact state:

M_x＝-G(0.5sinθl-l_g)-F_f(cosθl+sinθr-l_v+δ₁)

+F_n(sinθl-cosθr+l_h+δ₂)

(1)

l_g＝cosθW_g (2)

l_v＝-sinθ(cosθ_er+W_e-r)+cosθsinθ_er (3)

l_h＝cosθ(cosθ_er+W_e-r)+sinθsinθ_er (4)

object＝|M_x+G(0.5sinθl-l_g)+F_f(cosθl+sinθr-l_v+δ₁)

-F_n(sinθl-cosθr+l_h+δ₂)|

(6)

G＝max(|tanθ_e-θ_e|-σ₁，0)+max(θ_e-θ，0)+max(|l_h|-σ₂，0) (7)

fitness(i)＝object+(C×i)^γG (8)

in the formula: m_xMoment on the shaft, G gravity on the shaft, F_fThe friction force acting on the shaft, F_nThe supporting force applied to the shaft, l is the length of the shaft, r is the radius of the shaft, theta is the deflection angle of the shaft relative to the vertical line, l_gIs the lateral offset of the center of gravity of the shaft relative to the clamping end,/_vLongitudinal offset of the shaft contact relative to the clamping end,/_hIs the transverse offset, δ, of the shaft contact with respect to the clamping end₁，δ₂For local deformation of the contact point, W_gDeflection of the midpoint of the shaft, W_eIs the deflection of the free end of the shaft, theta_eIs a corner of the free end face of the shaft, G₁，G₂，G₃Are all single point contact state constraints, σ₁，σ₂The constraint constant object is an objective function of the model, G is a constraint function of the model, fit is an adaptive function, C and gamma are penalty constants of a dynamic penalty function, i is the iteration number of an optimization algorithm, the constraint function can limit variables in a feasible domain, and the dynamic penalty function can fully search the whole feasible domain, so that an optimal value is found in the feasible domain;

predicting a deviation angle theta of the shaft relative to the vertical line by combining the formulas (1) - (8) and a parameter optimization algorithm, and providing a deviation correction parameter for subsequent assembly;

plane contact state:

M_z＝F₁(y₁+ε_y)-F₂(x₁+ε_x) (11)

fitness(i)＝|M_z-F₁(y₁+ε_y)+F₂(x₁+ε_x)| (12)

in the formula: x is the number of₁，y₁The centroid position of the portion of the shaft end face in contact with the bore plane, F₁，F₂，F_nForces acting on the shaft in all directions at the centroid location, M_x，M_y，M_zRespectively the moments, e, of the axes about the x, y, z axes_x，ε_yFor compensation of the centroid due to elastic deformation of the shaft, d_x，d_yAlpha is the relative distance and the relative deflection angle of the shaft end face center and the hole center respectively, and the centroid compensation quantity, namely epsilon, is predicted by combining the formulas (9) - (12) and a parameter optimization algorithm_x，ε_yCombining formulae (13) - (15) and ∈_x，ε_yThe relative distance between the center of the end face of the shaft and the center of the hole, namely the offset distance d, is obtained by calculating the predicted value_x，d_yProviding deviation correction parameters for subsequent assembly;

and step 3: the robot starts to execute the assembly task to move downwards after clamping the non-rigid body part, and the shaft holes are contacted and reach the longitudinal force F_zThe threshold value of (1) is quickly lifted to separate the shaft holes, and meanwhile, the next step of instruction is waited;

and 4, step 4: transmitting force sense data acquired in the shaft hole contact process to a PC (personal computer) end, calling corresponding software and a contact state identification model by the PC end, and judging the contact state according to the data characteristics;

and 5: calling corresponding contact force sense models according to different contact states, if the part is in the stage of single-point contact and two-point contact, predicting to obtain an included angle theta between the axis and the vertical line through contact force sense model formulas (1) - (5) and an optimization algorithm, and if the part is in the stage of plane contact, predicting to obtain a relative distance between the center of the end face of the axis and the center of the hole, namely an offset d through contact force sense model formulas (13) - (15) and the optimization algorithm_x，d_y；

Step 6: and (4) performing inverse kinematics analysis of the robot according to the prediction result, and calling different robot deviation rectifying programs to achieve the purpose of adjusting the posture and the position of the shaft. After the posture or the position of the shaft is adjusted, the assembly action is carried out again;

and 7: and (4) judging whether the assembly condition is met, continuing to operate until the assembly is finished after the assembly condition is met, and otherwise, executing the step 4.

The invention has the following beneficial effects: 1. the contact force sense model based on deflection analysis accords with the contact deformation characteristic of the non-rigid body part in the assembling process, and can better predict the deflection angle theta and the offset d of the shaft_x， d_y. 2. Based on the contact state recognition model and the contact force sense model, the current contact state and the current posture of the part can be better reflected, and the part assembly can be realized more quickly.

Drawings

FIG. 1 is a single point contact statics analysis.

Fig. 2 is a non-rigid body deflection analysis.

Fig. 3 is a plane contact statics analysis.

FIG. 4 is a flat contact interface analysis.

Fig. 5 is a flow chart of the overall assembly of non-rigid body parts.

Detailed Description

The invention is further described below with reference to the accompanying drawings.

Referring to fig. 1 to 5, a robot non-rigid body assembling method based on deflection analysis includes the following steps:

step 1: the shaft hole assembly is divided into three states: a single point contact state, a two point contact state, and a planar contact state. Acquiring force sense data aiming at different contact states, analyzing the force sense data characteristics of the different states, and establishing a non-rigid body assembling contact body state recognition model by combining a machine learning algorithm and statics characteristics;

step 2: non-rigid contact force sense model established based on deflection analysis and designed for different contact states

The single point contact state is shown in FIGS. 1 and 2, where F_y，F_zThe force borne by the fixed end of the shaft, M and F are respectively the moment and the force equivalent to the center of the end face;

M_x＝-G(0.5sinθl-l_g)-F_f(cosθl+sinθr-l_v+δ₁)

+F_n(sinθl-cosθr+l_h+δ₂)

(1)

l_g＝cosθW_g (2)

l_v＝-sinθ(cosθ_er+W_e-r)+cosθsinθ_er (3)

l_h＝cosθ(cosθ_er+W_e-r)+sinθsinθ_er (4)

object＝|M_x+G(0.5sinθl-l_g)+F_f(cosθl+sinθr-l_v+δ₁)

-F_n(sinθl-cosθr+l_h+δ₂)|

(6)

G＝max(|tanθ_e-θ_e|-σ₁，0)+max(θ_e-θ，0)+max(|l_h|-σ₂，0) (7)

fitness(i)＝object+(C×i)^γG (8)

in the formula: m_xMoment on the shaft, G gravity on the shaft, F_fThe friction force acting on the shaft, F_nThe supporting force applied to the shaft, l is the length of the shaft, r is the radius of the shaft, theta is the deflection angle of the shaft relative to the vertical line, l_gIs the lateral offset of the center of gravity of the shaft relative to the clamping end,/_vLongitudinal offset of the shaft contact relative to the clamping end,/_hIs the transverse offset, δ, of the shaft contact with respect to the clamping end₁，δ₂The local deformation of the contact point has a value range of 0, 0.5r]，[0，r]，W_gDeflection of the midpoint of the shaft, W_eIs the deflection of the free end of the shaft, theta_eIs a corner of the free end face of the shaft, G₁，G₂，G₃Are all single point contact state constraints, σ₁，σ₂The constraint constants are respectively 0.001 and 0.002, the penalty constants are respectively 0.5 and 2 for C and gamma;

the method comprises the steps of optimizing parameters by using a gray wolf algorithm (GWO), setting initial parameters including a wolf group number SearchAgents which is 20, an individual dimension dim which is 3 and an iteration number Max _ iteration which is 100, establishing a contact force sense model by combining formulas (1) to (5), and calculating an adaptive function value of each individual by combining formulas (6) to (8) so as to obtain the optimal individual of the current iteration. Obtaining optimal parameters theta, delta through multiple iterations₁，δ₂Thereby predicting the deviation angle theta of the shaft relative to the vertical line and providing deviation correction parameters for subsequent assembly;

the contact force sensation model in the two-point contact state is similar to that in the single-point contact state, and is not repeated;

the planar contact state is shown in FIGS. 3 and 4, h, p₁The center of the hole, the center of the end face of the shaft and the centroid of the contact part of the shaft and the platform, respectively, the shaft should be deformed in fig. 3, which deformation is not shown in the figure.

M_z＝F₁(y₁+ε_y)-F₂(x₁+ε_x) (11)

fitness(i)＝|M_z-F₁(y₁+ε_y)+F₂(x₁+ε_x)| (12)

In the formula: x is the number of₁，y₁The centroid position of the portion of the shaft end face in contact with the bore plane, F₁，F₂，F_nForces acting on the shaft in all directions at the centroid location, M_x，M_y，M_zRespectively the moments, e, of the axes about the x, y, z axes_x，ε_yThe value range of the centroid compensation quantity caused by the elastic deformation of the shaft is [ -0.01, 0.01 [)]，d_x，d_yAnd alpha is the relative distance and the relative deflection angle of the center of the shaft end surface and the center of the hole respectively

The gray wolf algorithm (GWO) is adopted to carry out parameter optimization, initial parameters are set to include the wolf group quantity Searchgents which is 20, the individual dimension dim which is 3 and the iteration times Max _ iteration which is 100, and by combining the formulas (9) - (12), the centroid compensation quantity, namely epsilon_x，ε_y. Combining formulae (13) - (15) and ∈_x，ε_yThe relative distance between the center of the end face of the shaft and the center of the hole, namely the offset distance d, is obtained by calculating the predicted value_x，d_yProviding deviation correction parameters for subsequent assembly;

and step 3: robot clampStarting to execute the assembly task to move downwards after holding the non-rigid body part until the shaft hole is contacted and reaches the longitudinal force F_zThe threshold value of (1) is quickly lifted to separate the shaft holes, and meanwhile, the next step of instruction is waited;

and 4, step 4: transmitting force sense data acquired in the shaft hole contact process to a PC (personal computer) end, calling corresponding software and a contact state identification model by the PC end, and judging the contact state according to the data characteristics;

and 5: and calling the corresponding contact force sense model according to different contact states. If the part is in the single-point contact and two-point contact stages, the included angle theta between the axis and the vertical line is predicted by the contact force sense model formulas (1) - (5) and the optimization algorithm GWO, and if the part is in the plane contact stage, the relative distance between the center of the end face of the axis and the center of the hole, namely the offset d, is predicted by the contact force sense model formulas (13) - (15) and the optimization algorithm GWO_x， d_yProviding deviation correction parameters for subsequent assembly;

step 6: performing inverse kinematics analysis of the robot according to the prediction result, calling different robot deviation rectifying programs to achieve the purpose of adjusting the posture and the position of the shaft, and performing assembly again after the posture or the position of the shaft is adjusted;

and 7: and (4) judging whether the assembly condition is met, continuing to operate until the assembly is finished after the assembly condition is met, and otherwise, executing the step 4.

The foregoing lists merely illustrate specific embodiments of the invention. It is obvious that the invention is not limited to the above embodiments, but that many variations are possible. All modifications which can be derived or suggested directly from the disclosure of the invention by a person skilled in the art are to be considered within the scope of the invention.

Claims (1)

1. A robot non-rigid body assembling method based on deflection analysis is characterized by comprising the following steps:

step 1: the shaft hole assembly is divided into three states: the method comprises the steps of acquiring force sense data aiming at different contact states, analyzing force sense data characteristics of different states and establishing a non-rigid body assembly contact state recognition model by combining a machine learning algorithm and statics characteristics under the conditions of a single-point contact state, a two-point contact state and a plane contact state;

step 2: designing a non-rigid contact force sense model based on deflection analysis according to different contact states,

single-point contact state and two-point contact state:

l_g＝cosθW_g (2)

l_v＝-sinθ(cosθ_er+W_e-r)+cosθsinθ_er (3)

l_h＝cosθ(cosθ_er+W_e-r)+sinθsinθ_er (4)

object＝|M_x+G(0.5sinθl-l_g)+F_f(cosθl+sinθr-l_v+δ₁)-F_n(sinθl-cosθr+l_h+δ₂)| (6)

G＝max(|tanθ_e-θ_e|-σ₁,0)+max(θ_e-θ,0)+max(|l_h|-σ₂,0) (7)

fitness(i)＝object+(C×i)^γG (8)

in the formula: m_xMoment on the shaft, G gravity on the shaft, F_fThe friction force acting on the shaft, F_nThe supporting force applied to the shaft, l is the length of the shaft, r is the radius of the shaft, theta is the deflection angle of the shaft relative to the vertical line, l_gIs the lateral offset of the center of gravity of the shaft relative to the clamping end,/_vLongitudinal offset of the shaft contact relative to the clamping end,/_hThe lateral offset of the shaft contact with respect to the clamping end,δ₁，δ₂for local deformation of the contact point, W_gDeflection of the midpoint of the shaft, W_eIs the deflection of the free end of the shaft, theta_eIs a corner of the free end face of the shaft, G₁，G₂，G₃Are all single point contact state constraints, σ₁，σ₂Is a constraint constant

object is an objective function of the model, G is a constraint function of the model, fit is an adaptive function, C and gamma are penalty constants of dynamic penalty functions, i is the iteration times of an optimization algorithm, the constraint function can limit variables in a feasible domain, and the dynamic penalty function can fully search the whole feasible domain, so that an optimal value is found in the feasible domain;

predicting a deviation angle theta of the shaft relative to the vertical line by combining the formulas (1) - (8) and a parameter optimization algorithm, and providing a deviation correction parameter for subsequent assembly;

plane contact state:

M_z＝F₁(y₁+ε_y)-F₂(x₁+ε_x) (11)

fitness(i)＝|M_z-F₁(y₁+ε_y)+F₂(x₁+ε_x)| (12)

in the formula: x is the number of₁，y₁The centroid position of the portion of the shaft end face in contact with the bore plane, F₁，F₂，F_nForces acting on the shaft in all directions at the centroid location, M_x，M_y，M_zRespectively the moments, e, of the axes about the x, y, z axes_x，ε_yFor compensation of the centroid due to elastic deformation of the shaft, d_x，d_yAlpha is the relative distance and the relative deflection angle of the shaft end face center and the hole center respectively, and the centroid compensation quantity, namely epsilon, is predicted by combining the formulas (9) - (12) and a parameter optimization algorithm_x，ε_yCombining formulae (13) - (15) and ∈_x，ε_yThe relative distance between the center of the end face of the shaft and the center of the hole, namely the offset distance d, is obtained by calculating the predicted value_x，d_yProviding deviation correction parameters for subsequent assembly;

and step 3: the robot starts to execute the assembly task to move downwards after clamping the non-rigid body part, and the shaft holes are contacted and reach the longitudinal force F_zThe threshold value of (1) is quickly lifted to separate the shaft holes, and meanwhile, the next step of instruction is waited;

and 4, step 4: transmitting force sense data acquired in the shaft hole contact process to a PC (personal computer) end, calling corresponding software and a contact state identification model by the PC end, and judging the contact state according to the data characteristics;

and 5: calling corresponding contact force sense models according to different contact states, if the part is in the stage of single-point contact and two-point contact, predicting to obtain an included angle theta between the axis and the vertical line through contact force sense model formulas (1) - (5) and an optimization algorithm, and if the part is in the stage of plane contact, predicting to obtain a relative distance between the center of the end face of the axis and the center of the hole, namely an offset d through contact force sense model formulas (13) - (15) and the optimization algorithm_x，d_y；

Step 6: performing inverse kinematics analysis of the robot according to the prediction result, calling different robot deviation rectifying programs to achieve the purpose of adjusting the posture and the position of the shaft, and performing assembly again after the posture and the position of the shaft are adjusted;

and 7: and (4) judging whether the assembly condition is met, continuing to operate until the assembly is finished after the assembly condition is met, and otherwise, executing the step 4.

Images