Visual guidance positioning system and method for machine tool flexible shaft assembling process

Abstract

The invention discloses a visual guidance positioning system and method in the machine tool flexible shaft assembling process, and relates to the technical field of flexible shaft assembling, a visual system obtains three-dimensional pose data of the tail end of a flexible shaft and a target connector, a six-dimensional force sensor is combined to monitor contact force / torque in real time, the limitation of a single sensor is broken through, and the blind area of the single sensor is solved; visual and force sense coordinate systems are aligned in real time, and coordinate drifting caused by movement of the mechanical arm is eliminated; generating an approximation path based on the pose difference, introducing a compensation amount formula, inhibiting oscillation in combination with an attenuation coefficient, and improving the path smoothness; when the contact force exceeds the dynamic threshold value or the pose difference converges, path correction is automatically triggered, overload damage is avoided, and the assembly damage rate is effectively reduced. The method breaks through the limitation of traditional static path planning, has high precision, high reliability and complex working condition adaptability, and is especially suitable for precise instrument and heavy load machinery assembly scenes.

Description

Technical Field

[0001] The invention relates to the technical field of flexible shaft assembly, in particular to a visual guidance positioning system and method for a flexible shaft assembly process of a machine tool. Background Art

[0002] In the field of mechanical manufacturing, flexible shaft assembly is a key process in the assembly of precision machinery (such as automotive transmissions and aircraft engines). As a flexible transmission component, the flexible shaft must be precisely inserted into the target joint hole. Its assembly accuracy directly affects the transmission efficiency and service life of the equipment. However, the traditional flexible shaft assembly process faces the following technical bottlenecks:

[0003] While existing visual positioning methods based on monocular or binocular cameras can determine the position and orientation of the flexible shaft end, they are prone to cumulative positioning errors during the assembly contact phase due to issues such as flexible shaft deformation and occlusion of the target joint. Furthermore, relying solely on visual data makes it difficult to perceive assembly contact forces in real time, and overload can easily damage the flexible shaft or joint. Furthermore, flexible deformation during assembly can lead to unpredictable positional deviations at the flexible shaft end, which cannot be corrected in real time by existing static path planning methods, resulting in increased assembly failure rates. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the present invention provides a visually guided positioning system and method for the assembly process of a flexible shaft of a tool.

[0005] In order to achieve the above object, the technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention discloses a visually guided positioning method for a flexible shaft assembly process of a tool, comprising the following steps:

[0007] The three-dimensional coordinates and attitude angle data of the flexible shaft end are acquired through the visual system, and the center coordinates and axis direction vector data of the target joint hole are extracted;

[0008] The six-dimensional force sensor is used to collect six-dimensional mechanical data in the flexible shaft assembly process in real time;

[0009] Calibrate and align the coordinate system of the vision system with the coordinate system of the six-dimensional force sensor to generate unified spatial mapping relationship data;

[0010] The pose difference data between the flexible shaft end and the target joint hole is calculated based on the three-dimensional coordinates and attitude angle data of the flexible shaft end, the center coordinates of the target joint hole and the axis direction vector data, and the approximation path parameters are generated based on the pose difference data;

[0011] When the contact force data and torque data exceed a preset safety threshold or the posture difference dynamic data is less than a predetermined number of digits, a path correction parameter is constructed according to the six-dimensional mechanical data;

[0012] The path correction parameters are superimposed on the posture difference data to update the approximation path parameters, and the robot arm control instruction data is generated based on the updated approximation path parameters;

[0013] The constructing of path correction parameters according to the six-dimensional mechanical data includes:

[0014] The six-dimensional mechanical data is obtained, the difference between the six-dimensional mechanical data and the preset safety threshold is calculated, and the compensation amount is calculated. The calculation formula of the compensation amount is as follows:

[0015] Compensation amount = configurable compensation coefficient × (six-dimensional mechanical data - preset safety threshold) + compensation amount of the previous cycle × configurable attenuation coefficient.

[0016] In a second aspect, the present invention discloses a vision-guided positioning system for a flexible shaft assembly process of a machine tool, which implements the vision-guided positioning method for a flexible shaft assembly process of a machine tool, including:

[0017] The visual perception and posture calculation module is used to obtain the three-dimensional coordinates and posture angle data of the flexible shaft end through the visual system, and extract the center coordinates and axis direction vector data of the target joint hole;

[0018] The six-dimensional force sensing monitoring module is used to collect six-dimensional mechanical data in the flexible shaft assembly process in real time through a six-dimensional force sensor;

[0019] Multi-source data fusion and coordinate system calibration module, used to calibrate and align the coordinate system of the vision system with the coordinate system of the six-dimensional force sensor to generate unified spatial mapping relationship data;

[0020] An adaptive path planning module is used to calculate the posture difference data between the flexible shaft end and the target joint hole based on the three-dimensional coordinates and attitude angle data of the flexible shaft end, the center coordinates of the target joint hole, and the axis direction vector data, and generate the approximate path parameters based on the posture difference data;

[0021] A dynamic correction module, configured to construct a path correction parameter based on the six-dimensional mechanical data when the contact force data and the torque data exceed a preset safety threshold or the posture difference dynamic data is less than a predetermined number of digits;

[0022] The robot arm motion control and execution module is used to superimpose the path correction parameters with the posture difference data, update the approximation path parameters, and generate the robot arm control instruction data based on the updated approximation path parameters;

[0023] The constructing of path correction parameters according to the six-dimensional mechanical data includes:

[0024] The six-dimensional mechanical data is obtained, the difference between the six-dimensional mechanical data and the preset safety threshold is calculated, and the compensation amount is calculated. The calculation formula of the compensation amount is as follows:

[0025] Compensation amount = configurable compensation coefficient × (six-dimensional mechanical data - preset safety threshold) + compensation amount of the previous cycle × configurable attenuation coefficient.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] 1. The three-dimensional position of the flexible shaft end and the axis direction of the target hole are obtained through the vision system. In combination with the real-time feedback of the contact force / torque from the six-dimensional force sensor, the limitations of a single sensor are overcome. The vision system and the six-dimensional force sensor coordinate system are aligned in real time, solving the coordinate drift problem caused by the movement of the robot arm in traditional static calibration and effectively reducing the spatial mapping error.

[0028] 2. Generate the initial path based on the posture difference data, superimpose the compensation formula, and introduce the historical data attenuation mechanism into the compensation formula to suppress path oscillation and significantly reduce the axial deviation caused by the flexible deformation of the soft shaft. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The disclosure of the present invention is described with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components. Among them:

[0030] Figure 1 is a flow chart of the method of the present invention;

[0031] Figure 2 It is a workflow diagram of the present invention;

[0032] Figure 3 A flow chart of compensation for vibration and thermal deformation effects of the present invention;

[0033] Figure 4 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0034] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can propose a variety of interchangeable structural modes and implementation modes. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present invention and should not be regarded as the entire invention or as a limitation or restriction of the technical solution of the present invention.

[0035] Application Overview

[0036] During the assembly of flexible shafts in traditional precision machinery, the visual system is prone to deviations in the flexible shaft end-point pose measurement due to the complex three-dimensional topography of the target joint hole or interference from ambient lighting. The end-point offset caused by the flexible shaft's flexible deformation during the assembly contact phase is difficult to capture using static visual data, resulting in cumulative errors between the path planning parameters and the actual assembly trajectory. The lack of six-dimensional mechanical perception prevents the protection mechanism from being triggered when the assembly contact force exceeds the material's yield limit, causing scratches on the flexible shaft surface or plastic deformation of the joint's inner wall.

[0037] For example, in the assembly of a flexible shaft for an automotive transmission, when there is an angular deviation of more than 0.5° between the axial direction of the target joint hole and the posture of the flexible shaft end, the visual system's center coordinate extraction error exceeds ±0.3mm due to the reflective characteristics of the flexible shaft end. During assembly, the flexible shaft is bent and deformed by axial pressure, and the actual posture of the end deviates by 1.2mm from the visual measurement value, while traditional path planning still generates the robot arm's motion trajectory based on the initial measurement value. When the lateral contact force of the flexible shaft reaches 25N, the robot arm, lacking real-time force feedback, continues to apply thrust, causing permanent micron-level deformation of the joint locating pin.

[0038] If these issues are not addressed, posture deviations during the assembly contact phase will cause mechanical interference between the flexible shaft and the connector hole, resulting in an assembly success rate that drops below 78%. Microscopic damage caused by excessive contact forces will reduce the fatigue life of the transmission system and shorten the mean time between failures of key components by 40%. Irreversible plastic deformation will also increase the rework rate of precision machinery and tools, increasing the assembly cost of each piece of equipment by approximately 15%.

[0039] When faced with the above problems, this application first explores how to integrate vision and force perception to achieve dynamic control of the assembly process. The traditional solution relies on visual positioning but ignores contact mechanical feedback, resulting in the inability to correct the trajectory offset caused by flexible deformation. If a force sensor is introduced alone to monitor the contact force, although overload damage can be avoided, the contact force anomaly caused by the initial posture deviation cannot be solved. To this end, this application attempts to spatially map the visual coordinate system and the force sensor coordinate system so that the posture difference calculation and mechanical data have a unified reference benchmark. At the same time, the design generates the initial path direction based on the posture difference, and achieves rapid approach in the coarse adjustment stage through the inverse relationship between the moving speed and the displacement. When the contact force approaches the safety threshold, an incremental compensation algorithm is used to superimpose the path correction amount, which not only ensures timely avoidance when the mechanical limit is exceeded, but also avoids mechanical oscillations caused by sudden adjustments. By designing the attenuation coefficient in the compensation formula, the influence of high-frequency interference signals on path planning can be effectively suppressed.

[0040] like Figure 1 As shown, in this regard, the present application proposes a visually guided positioning method for the tool flexible shaft assembly process, comprising the following steps:

[0041] The three-dimensional coordinates and attitude angle data of the flexible shaft end are obtained through the visual system, and the center coordinates and axis direction vector data of the target joint hole are extracted; the visual system obtains the three-dimensional coordinates and attitude angle data of the flexible shaft end, which means using a three-dimensional vision sensor to collect the position and angle information of the flexible shaft end in space. Specifically, binocular stereo vision or structured light three-dimensional scanning technology can be used to achieve this, which is used to track the spatial posture changes of the flexible shaft end in real time.

[0042] The six-dimensional mechanical data of the flexible shaft assembly process is collected in real time through a six-dimensional force sensor; the six-dimensional mechanical data of the flexible shaft assembly process is collected in real time by a six-dimensional force sensor, which means that the forces and moments in three directions exerted on the flexible shaft during the assembly process are measured synchronously through a six-degree-of-freedom force sensor. Specifically, it can be implemented using a strain gauge or capacitive sensor to monitor the mechanical state of the assembly contact stage.

[0043] The coordinate system of the visual system is calibrated and aligned with the coordinate system of the six-dimensional force sensor to generate unified spatial mapping relationship data; calibrating and aligning the coordinate system of the visual system and the coordinate system of the six-dimensional force sensor means converting the measurement data of the two sensors into the same reference coordinate system through a coordinate transformation matrix. Specifically, this can be achieved by using a hand-eye calibration algorithm or an offline calibration method based on a calibration board to ensure the spatial consistency of the visual data and the force data.

[0044] Based on the three-dimensional coordinates and attitude angle data of the flexible shaft end, the center coordinates of the target joint hole and the axial direction vector data, the posture difference data between the flexible shaft end and the target joint hole is calculated, and the approximation path parameters are generated based on the posture difference data; generating approximation path parameters based on the posture difference data between the flexible shaft end and the target joint hole refers to generating the robot arm motion trajectory according to the displacement and angular deviation of the flexible shaft end relative to the target hole, which can be specifically implemented by a posture error feedback control algorithm, which is used to dynamically adjust the movement direction of the robot arm to reduce the posture deviation.

[0045] When the contact force data and torque data exceed the preset safety threshold or the posture difference dynamic data is less than a predetermined number of digits, path correction parameters are constructed based on the six-dimensional mechanical data; constructing path correction parameters based on six-dimensional mechanical data means generating a compensation amount through the difference between the mechanical data and the preset safety threshold, which can be specifically implemented using a proportional-integral control algorithm or an incremental compensation strategy, and is used to automatically adjust the path when the contact is overloaded to avoid damaging the flexible shaft or joint.

[0046] The path correction parameters are superimposed on the posture difference data to update the approximation path parameters, and the robot arm control instruction data is generated based on the updated approximation path parameters;

[0047] The path correction parameters constructed based on six-dimensional mechanical data include:

[0048] Obtain six-dimensional mechanical data, calculate the difference between the six-dimensional mechanical data and the preset safety threshold, and calculate the compensation amount. The calculation formula of the compensation amount is as follows:

[0049] Compensation Amount = Configurable Compensation Coefficient × (6-Dimensional Mechanical Data - Preset Safety Threshold) + Previous Cycle Compensation × Configurable Attenuation Coefficient. Superimposing path correction parameters with pose difference data involves fusing the mechanical compensation amount with the visual pose error. This can be achieved using weighted superposition or state-space fusion methods to generate more robust path parameters by integrating visual and force information.

[0050] The core innovation of this application lies in the fusion calibration and dynamic compensation mechanism of visual and force multimodal data, which simultaneously solves the problems of visual positioning error accumulation, contact force overload monitoring and real-time correction of flexible deformation during the flexible shaft assembly process, thereby achieving high-precision and high-safety adaptive assembly control.

[0051] like Figure 2 The following is a workflow diagram of the present invention; the working process and principle of this application are as follows:

[0052] First, the vision system acquires the 3D coordinates and attitude angle data of the flexible shaft end, and extracts the center coordinates and axis direction vector data of the target joint hole. This data is used to calculate the pose difference between the flexible shaft end and the target joint hole. Simultaneously, a 6D force sensor collects 6D mechanical data, including contact force and torque, during the assembly process in real time.

[0053] To achieve the fusion of visual and force data, the coordinate system of the visual system is calibrated and aligned with the coordinate system of the six-dimensional force sensor to generate unified spatial mapping data. This ensures that pose difference calculation and mechanical data analysis are performed in the same reference frame.

[0054] Based on the position and posture data of the flexible shaft end and the target joint hole, the system calculates the posture difference and generates initial approximate path parameters. If the contact force and torque data exceed the preset safety threshold or the dynamic posture difference data is less than a predetermined number of digits, the system constructs path correction parameters based on the six-dimensional mechanical data.

[0055] The path correction parameters are constructed using an incremental compensation algorithm. Specifically, six-dimensional mechanical data is acquired, the difference from the preset safety threshold is calculated, and then the compensation amount is calculated. The compensation amount is calculated as follows: Compensation = configurable compensation coefficient × (six-dimensional mechanical data - preset safety threshold) + previous cycle compensation × configurable attenuation coefficient. This algorithm design not only enables timely avoidance when mechanical limits are exceeded, but also suppresses the impact of high-frequency interference signals on path planning through the attenuation coefficient.

[0056] Finally, the path correction parameters are superimposed on the pose difference data to update the approximate path parameters, and the robot arm control command data is generated based on the updated parameters. This dynamic adjustment mechanism can respond to mechanical feedback during the assembly process in real time and effectively compensate for pose deviations caused by flexible deformation of the flexible shaft.

[0057] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0058] During the assembly of a flexible shaft in an automotive transmission, a high-resolution industrial camera is first used to capture an image of the shaft end. Image processing algorithms are then used to extract the center coordinates (x, y, z) and attitude angles (α, β, γ) of the shaft end. Simultaneously, a structured light scanner is used to acquire 3D point cloud data of the target joint hole. A point cloud registration algorithm is then used to extract the hole center coordinates (X, Y, Z) and axis direction vectors (i, j, k).

[0059] A six-dimensional force / torque sensor is mounted at the end of the robotic arm, sampling at a 1000Hz frequency. It collects six-dimensional mechanical data (Fx, Fy, Fz, Mx, My, and Mz) in real time during the assembly process. Hand-eye calibration is used to align the vision system coordinate system with the force sensor coordinate system, establishing a unified spatial mapping matrix, T.

[0060] Based on the acquired position and posture data, the pose difference ΔP between the flexible shaft end and the target joint hole is calculated as (ΔX, ΔY, ΔZ, Δα, Δβ, Δγ). The initial approach path direction V is set to the normalized composite vector of the target joint axis direction vector (i, j, k) and the translation difference (ΔX, ΔY, ΔZ). The movement speed v is inversely proportional to the pose difference modulus |ΔP|.

[0061] When the contact force Fz exceeds the preset safety threshold Fz_max or the posture difference |ΔP| is less than the pre-positioning threshold ε, the path correction mechanism is triggered. The compensation amount C is calculated based on the six-dimensional mechanical data F = (Fx, Fy, Fz, Mx, My, Mz):

[0062] C=k×(F-F_threshold)+C_prev×λ.

[0063] Where k is a configurable compensation coefficient, F_threshold is the preset safety threshold, C_prev is the compensation amount in the previous cycle, and λ is a configurable attenuation coefficient.

[0064] The calculated compensation value C is added to the pose difference ΔP to update the approximation path parameters. Based on the updated parameters, the robot arm joint angle command θ is generated to control the robot arm to perform the adjusted assembly action.

[0065] Through the above scheme, the present application can achieve precise positioning and dynamic adjustment during the assembly process of the flexible shaft. The visual guidance system provides initial posture information, while the real-time force feedback mechanism can compensate for the position deviation caused by the flexible deformation of the flexible shaft. The incremental compensation algorithm effectively avoids mechanical interference and overload damage during the assembly process. This method of integrating vision and force perception significantly improves the success rate of flexible shaft assembly and reduces the probability of mechanical interference caused by posture deviation. At the same time, real-time mechanical monitoring and path dynamic adjustment mechanism effectively prevent overload contact force from damaging the flexible shaft and joint, thereby extending the service life of the transmission system. In addition, through adaptive path planning and real-time correction, the method reduces the need for manual intervention, improves assembly efficiency, and reduces the assembly cost of precision machinery.

[0066] In some of the above-mentioned schemes in this application, the approximation path parameters generated based on the posture difference data have problems such as large path direction deviation, inflexible adjustment of movement speed, and unreasonable setting of convergence conditions, which results in the end of the flexible shaft being unable to accurately and efficiently approach the target joint hole position, affecting the assembly accuracy and efficiency.

[0067] The present application further proposes that the posture difference dynamic data includes translation difference data and attitude angle difference data, and the approximation path parameters include path direction, movement speed and convergence condition;

[0068] The path direction is generated by normalizing the target joint axis direction vector and the translation difference data to generate a synthetic direction vector, ensuring that the path direction takes into account both the target axis alignment and the position deviation correction. The moving speed is calculated by dividing the preset proportional coefficient by the sum of the translation difference modulus and the minimum constant to prevent zero division to achieve dynamic speed adjustment. The convergence condition is set based on the sum of the squares of the translation differences, and its threshold range is dynamically determined according to the target aperture to avoid improper assembly or inefficiency caused by being too tight or too loose.

[0069] Specifically, the target joint axis direction vector and the translation difference data are normalized to form a synthetic direction vector, which is used as the initial approach direction, so that the end of the flexible shaft can keep the axis aligned while correcting the position deviation. The moving speed is automatically adjusted according to the translation difference module length. When the module length is large, the speed is reduced to prevent overshoot, and when the module length is small, the speed is increased to accelerate convergence. The preset proportional coefficient controls the speed reference, and the zero minimum constant is used to avoid calculation errors when the module length is zero. The convergence condition is achieved by comparing the sum of the squares of the translation difference with a preset threshold. The threshold is set to 80% to 120% of the aperture. When the actual deviation is less than the threshold, it is judged to be in place, ensuring that the end of the flexible shaft accurately enters the target hole position without being affected by the aperture processing error.

[0070] In specific implementation, a vision system can be used to obtain the three-dimensional coordinates (x, y, z) and attitude angles (α, β, γ) of the flexible shaft end, and simultaneously extract the center coordinates (x0, y0, z0) and axis direction vectors (a, b, c) of the target joint hole. The pose difference data between the flexible shaft end and the target joint hole is calculated, including the translation difference (Δx, Δy, Δz) and attitude angle difference (Δα, Δβ, Δγ).

[0071] Based on the above parameters, an initial approximation path can be generated, and the path parameters can be dynamically updated according to the real-time posture difference data during the assembly process to achieve precise positioning of the flexible shaft.

[0072] Through the above technical solution, the present application realizes precise positioning and path planning during the assembly process of the flexible shaft. By comprehensively considering the translation difference and attitude angle difference, a reasonable approach path direction is generated, which effectively avoids the positioning error that may be caused by relying solely on visual data. The association setting of the moving speed and the attitude difference enables adaptive deceleration when approaching the target position, thereby improving the assembly accuracy. At the same time, the convergence condition set based on the target joint aperture ensures the stability and reliability of the assembly process. This solution overcomes the limitations of traditional static path planning methods, and can dynamically adjust the approach path according to real-time attitude difference data, effectively dealing with the unpredictable attitude offset of the flexible shaft caused by flexible deformation during the assembly process, and significantly improving the success rate and efficiency of the flexible shaft assembly.

[0073] In some of the above-mentioned schemes of the present application, the moving speed is inversely proportional to the modulus of the translation difference data, but in actual applications, when the modulus of the translation difference approaches zero, it may cause abnormal speed calculation or division by zero error, thereby causing the risk of loss of control of the robotic arm movement.

[0074] The present application further proposes a calculation formula for the moving speed: speed parameter = preset proportional coefficient / (modulus length of translation difference data + minimum constant for preventing zero division).

[0075] The preset scaling factor is used to adjust the mapping between the velocity parameter and the translational differential modulus. The zero-elimination minimum constant is set to a very small positive number to prevent the denominator from being zero. The modulus of the translational differential data is calculated using the vector length, and its physical meaning is the straight-line distance between the end of the flexible shaft and the target joint hole. The zero-elimination minimum constant ranges from 0.001mm to 0.01mm, and the preset scaling factor is an adjustable parameter set within the range of 0.3-0.7 based on the maximum movement speed of the robot arm.

[0076] Specifically, during the flexible shaft assembly process, the translation differential modulus between the flexible shaft end and the target hole position is calculated in real time. When the modulus value decreases, the overall value of the denominator decreases, causing the speed parameter to increase, but the anti-zero minimum constant ensures that the denominator is always greater than zero. For example, when the translation differential modulus is 0.5mm, the preset proportional coefficient of 0.5 and the anti-zero constant of 0.001mm are used to calculate the speed parameter to be 0.5 / (0.5+0.001)=0.998mm / s. When the translation differential modulus approaches zero, the speed parameter approaches 0.5 / 0.001=500mm / s. At this time, the actual output speed is constrained by the preset proportional coefficient upper limit not to exceed the safety threshold of the robot arm. This formula uses a mathematical constraint mechanism to ensure the inverse relationship between speed and posture difference while eliminating calculation singularities and ensuring the continuous stability of motion control instructions.

[0077] Therefore, as the end of the flexible shaft gradually approaches the target position, the modulus of the translation difference data gradually decreases, and the speed parameter will increase accordingly, realizing adaptive speed control of the end of the flexible shaft.

[0078] Through the above-mentioned technical solution, this application achieves adaptive speed control during the flexible shaft assembly process. By establishing an inverse relationship between movement speed and the modulus length of the translation difference data, a slower movement speed is used when the flexible shaft end is farther from the target position, and the speed is gradually increased as the distance decreases. This adaptive speed control method effectively improves the accuracy and efficiency of flexible shaft assembly. Furthermore, the introduction of a zero-prevention minimum constant avoids possible division errors during the calculation process, further enhancing the stability and reliability of the algorithm.

[0079] In some of the above-mentioned solutions of this application, a method of constructing path correction parameters based on six-dimensional mechanical data is proposed. However, in the actual assembly process, it is difficult to accurately correspond to the changes in contact force and torque in different directions using only a single compensation parameter, resulting in the inability to perform differentiated adjustments for axial force, lateral force and axial torque during the path correction process, which may cause local overload or correction direction deviation.

[0080] This application further proposes that six-dimensional mechanical data includes contact force data and torque data, wherein the contact force data includes axial force Fz, lateral force Fx, and lateral force Fy, and the torque data includes axial torque Mx, My, and Mz; there are six types of compensation amounts, which correspond one-to-one to the data types in the six-dimensional mechanical data.

[0081] The six-dimensional mechanical data is decomposed into contact force components in three orthogonal directions and torque components in three rotational directions, and each component independently corresponds to a compensation amount. The axial force Fz in the contact force data is defined as the longitudinal force along the insertion direction of the flexible shaft, and the lateral forces Fx and Fy are defined as the lateral force components perpendicular to the axis of the flexible shaft. The axial moments Mx, My, and Mz in the torque data correspond to the rotational moments around the three orthogonal axes, respectively. The generation logic of the six compensation amounts is bound one by one to the components of the six-dimensional mechanical data. For example, the compensation amount of the axial force Fz only affects the path adjustment in the insertion direction of the flexible shaft, and the compensation amount of the axial moment Mx only affects the posture correction around the X-axis.

[0082] Specifically, when the axial force Fz exceeds the safety threshold, the compensation amount acts on the moving speed of the flexible shaft end along the axial direction, avoiding overload by reducing the propulsion speed; when the lateral force Fx or Fy exceeds the threshold, the compensation amount generates a lateral offset correction instruction to adjust the position of the flexible shaft end in the XY plane to reduce lateral friction. For torque data, when the torque Mx around the X-axis exceeds the limit, the compensation amount drives the robot arm to fine-tune the posture around the X-axis to release the torque. The six compensation amounts are independently calculated and superimposed to achieve six-degree-of-freedom dynamic correction. For example, the axial force compensation coefficient can be set to 0.5-0.8, the lateral force compensation coefficient to 0.3-0.6, and the torque compensation coefficient to 0.2-0.4 according to the material stiffness, thereby ensuring the differentiation and controllability of the correction amount when the mechanics in each direction exceeds the limit.

[0083] In practical applications, a six-dimensional force sensor can be used to collect six-dimensional mechanical data during the flexible shaft assembly process in real time. For example, an ATI Gamma six-dimensional force / torque sensor with a sampling frequency of 1000 Hz can be used. The measurement range of contact force data is ±32 N, and the measurement range of torque data is ±2.5 Nm.

[0084] The compensation amount is calculated for six types of mechanical data. Taking the axial force Fz as an example, when it is detected that Fz exceeds the preset safety threshold, the compensation amount of Fz is calculated:

[0085] Fz compensation amount = 0.5 × (Fz - 10N) + Fz compensation amount of the previous cycle × 0.8;

[0086] Among them, 0.5 is the configurable compensation coefficient, 10N is the preset safety threshold of Fz, and 0.8 is the configurable attenuation coefficient.

[0087] Similarly, a similar compensation calculation method is used for the other five types of mechanical data, except that the preset safety thresholds and compensation coefficients may be different. In this way, fine-grained compensation adjustments can be made for each component of the six-dimensional mechanical data.

[0088] Through the above technical solution, the present application can achieve comprehensive monitoring and precise compensation of six-dimensional mechanical data during the flexible shaft assembly process. By subdividing the six-dimensional mechanical data into two categories, contact force and torque, and further dividing it into six specific components, independent compensation can be performed for forces and torques in different directions. This refined compensation strategy helps to more accurately adjust the assembly path of the flexible shaft and improve assembly accuracy and success rate. At the same time, by setting configurable compensation coefficients and attenuation coefficients, the compensation strategy has strong flexibility and adaptability, and can be adjusted according to different assembly scenarios and requirements. In addition, considering the influence of the compensation amount of the previous cycle, a smooth transition of compensation can be achieved, avoiding sudden changes in the compensation process, thereby improving the stability and reliability of the flexible shaft assembly process.

[0089] In some of the above-mentioned schemes in this application, the static setting of the preset safety threshold and the predetermined number of bits is difficult to adapt to the real-time changes in the flexible deformation and dynamic contact force of the soft shaft during the assembly process, resulting in the safety threshold being unable to accurately reflect the actual assembly mechanical state, and the predetermined number of bits being unable to be dynamically adjusted to match the posture offset caused by the deformation of the soft shaft end, which poses a risk of overload damage or excessive positioning error.

[0090] This application further proposes that the preset safety threshold is the dynamic fluctuation range of the six-dimensional mechanical data generated based on historical assembly data, and the predetermined digit is the three-dimensional coordinate difference between the end of the flexible shaft and the target joint hole ≤ the dynamic convergence value.

[0091] The dynamic fluctuation range is generated by statistically analyzing the maximum, minimum, and standard deviation of six-dimensional mechanical data recorded during the historical assembly process. The dynamic convergence value is adjusted in real time based on the target joint aperture size and the deformation of the flexible shaft end. The safety threshold is updated using a sliding time window based on the mean and variance of the historical data. The dynamic fluctuation range is calculated as follows: Safety threshold = historical data mean ± 3 × historical data standard deviation. The dynamic convergence value is set at 90% of the target joint aperture size and is adjusted to 110% of the aperture size when the flexible shaft deformation exceeds the preset deformation threshold.

[0092] Specifically, during the assembly process, the six-dimensional mechanical data in the historical assembly data is called in real time, and the mean and standard deviation in the current time window are calculated by the moving average algorithm, and the upper and lower limits of the safety threshold are dynamically updated. When the contact force data collected in real time exceeds the dynamic fluctuation range, the path correction parameters are triggered to avoid overload damage. At the same time, the deformation of the end of the flexible shaft is measured in real time by the visual system. When the deformation exceeds the preset deformation threshold, the dynamic convergence value is adjusted to 110% of the aperture size, allowing a greater tolerance for positioning errors. This solution realizes dynamic adaptive adjustment of the safety threshold and the positioning error threshold by integrating the statistical laws of historical data with real-time deformation monitoring, solving the problem that the static threshold cannot adapt to the changes in flexible deformation and dynamic contact force. For example, when the flexible shaft is bent, the dynamic convergence value automatically increases to 1.1 times the aperture size to avoid repeated oscillations of the assembly path due to rigid positioning error limitations.

[0093] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0094] Preset safety thresholds generate the dynamic fluctuation range of six-dimensional mechanical data based on historical assembly data. For example, by collecting six-dimensional mechanical data from multiple successful assembly processes, the mean and standard deviation of the mechanical data in each dimension are calculated, and the dynamic fluctuation range of each dimension is calculated by adding or subtracting twice the standard deviation from the mean. Specifically, the safety threshold for axial force Fz can be set to 10±2N, the safety threshold for lateral forces Fx and Fy to 5±1N, and the safety threshold for axial moments Mx, My, and Mz to 0.5±0.1N·m.

[0095] Pre-positioning is achieved when the 3D coordinate difference between the flexible shaft end and the target connector hole is less than or equal to the dynamic convergence value. This dynamic convergence value can be adjusted dynamically based on the target connector hole diameter. For example, if the target connector hole diameter is 10mm, the dynamic convergence value can be set to 1mm. This means that pre-positioning is considered achieved when the 3D coordinate difference between the flexible shaft end and the target connector hole is less than or equal to 1mm in all directions.

[0096] Furthermore, the dynamic convergence value can be gradually reduced as the assembly process progresses to improve the final assembly accuracy. For example, the initial dynamic convergence value can be set to 10% of the target joint aperture, and then the convergence value can be reduced by 1% after each approximation action until it reaches 1% of the aperture.

[0097] Through the above technical solution, this application achieves dynamic adjustment of safety thresholds based on historical data, improving the safety and stability of the assembly process. At the same time, the use of dynamic convergence values ​​as the basis for pre-positioning judgment enables the assembly process to be flexibly adjusted according to target joints of different sizes, improving the adaptability and accuracy of the assembly system. Furthermore, by gradually reducing the dynamic convergence value, a smooth transition from coarse positioning to fine positioning is achieved, effectively reducing vibration and impact during the assembly process and extending the service life of the flexible shaft and joint.

[0098] In some of the aforementioned solutions, the generation of approximate path parameters based on pose difference data fails to account for mechanical vibration and temperature deformation in the assembly environment, resulting in dynamic interference errors in the path planning results. Mechanical vibrations cause high-frequency oscillations at the flexible shaft end, reducing the stability of the assembly trajectory. Changes in ambient temperature cause the flexible shaft material to expand and contract, resulting in deviations between the actual end position and the theoretically calculated value.

[0099] like Figure 3 As shown, it is a flow chart of compensation for vibration and thermal deformation effects of the present invention; the present application further proposes that generating approximation path parameters based on posture difference data also includes:

[0100] Real-time monitoring of vibration frequency data and ambient temperature data during the flexible shaft assembly process;

[0101] Build a vibration compensation model based on vibration frequency data and generate vibration suppression path correction;

[0102] Generate thermal deformation compensation based on the mapping relationship between ambient temperature data and thermal expansion coefficient of flexible shaft material;

[0103] The vibration suppression path correction amount and thermal deformation compensation amount are added to the posture difference data to update the approximation path parameters.

[0104] Vibration frequency data is collected by an accelerometer. The vibration compensation model generates a periodic correction by multiplying the preset critical frequency band by the amplitude attenuation coefficient. Ambient temperature data is acquired in real time by a temperature sensor. Thermal deformation compensation is calculated by multiplying the material's thermal expansion coefficient, the effective assembly length of the flexible shaft, and the temperature difference. The vibration suppression path correction is activated only when the vibration frequency reaches the critical frequency band to avoid overcompensation. The thermal deformation compensation is linearly proportional to temperature changes, and the compensation strength is controlled by the temperature difference.

[0105] Specifically, the accelerometer acquires vibration frequency data at a sampling rate of 1000Hz. When vibration in the critical frequency range of 200-400Hz is detected, a vibration suppression path correction is calculated as the product of the amplitude attenuation coefficient of 0.2 and the vibration sine function. A temperature sensor collects ambient temperature once per second. When the assembly workshop temperature rises from a base temperature of 20°C to 25°C, thermal deformation compensation is calculated as the material thermal expansion coefficient multiplied by the effective length of the flexible shaft of 300mm, then multiplied by a 5°C temperature difference, resulting in an axial compensation of 0.018mm. The vibration suppression path correction and thermal deformation compensation are added to the translational component of the pose difference data, allowing the robot arm control instructions to simultaneously eliminate vibration offset and thermal deformation errors. The updated approximation path parameters enable the flexible shaft end to maintain submillimeter assembly accuracy even in the presence of environmental interference.

[0106] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0107] During the flexible shaft assembly process, the vibration frequency data and ambient temperature data are monitored in real time. The vibration frequency data is collected by the acceleration sensor, and the ambient temperature data is collected by the temperature sensor.

[0108] A vibration compensation model is constructed based on vibration frequency data to generate a vibration suppression path correction. Specifically, when the vibration frequency is detected to be within a preset danger band, such as 50-100 Hz, the vibration suppression path correction is calculated. With the amplitude attenuation coefficient set to 0.5 and a vibration frequency of 75 Hz, the vibration suppression path correction is calculated as 0.5 × sin(2π × 75 × t), where t is the time variable. If the vibration frequency exceeds the danger band, the vibration suppression path correction is reset to 0.

[0109] The thermal deformation compensation is generated based on the mapping relationship between the ambient temperature data and the thermal expansion coefficient of the flexible shaft material. For example, the flexible shaft material is stainless steel, and the thermal expansion coefficient is 16.5×10 -6 / ℃, the effective assembly length of the flexible shaft is 500mm, the real-time temperature is 30℃, and the reference temperature is 20℃, then the thermal deformation compensation is calculated to be 16.5×10 -6 ×500×(30-20)=0.0825mm.

[0110] Furthermore, the calculated vibration suppression path correction and thermal deformation compensation are added to the pose difference data to update the approximation path parameters. The robotic arm control system can then generate optimized motion control instructions based on the updated approximation path parameters.

[0111] Through the above technical solution, this application realizes real-time compensation for the effects of vibration and thermal deformation during the flexible shaft assembly process. Specifically, by constructing a vibration compensation model, the impact of vibration in the dangerous frequency band on assembly accuracy can be effectively suppressed. At the same time, considering the thermal deformation caused by changes in ambient temperature, real-time compensation is performed to improve the positioning accuracy of the flexible shaft assembly. As a result, this solution significantly improves the stability and reliability of the flexible shaft assembly process, reduces assembly failures caused by vibration and thermal deformation, and improves assembly efficiency and product quality.

[0112] In some of the above-mentioned solutions of this application, the vibration compensation model does not effectively distinguish the vibration frequency bands, resulting in the generation of compensation amounts under vibration interference in non-dangerous frequency bands, which in turn introduces additional path deviations and affects the alignment accuracy between the flexible shaft end and the target joint.

[0113] This application further proposes the construction logic of the vibration compensation model as follows:

[0114] When the vibration frequency is within the preset dangerous frequency band, the vibration suppression path correction amount = amplitude attenuation coefficient × sin(2π × vibration frequency × time);

[0115] When the vibration frequency exceeds the dangerous frequency band, the vibration suppression path correction amount = 0.

[0116] The preset dangerous frequency band is determined through spectrum analysis and covers the inherent resonant frequency range of the flexible shaft material. The amplitude attenuation coefficient is dynamically adjusted based on the mean amplitude value collected by the vibration sensor. For example, when the amplitude exceeds 0.1mm, the attenuation coefficient is set to 0.8; when the amplitude is less than 0.05mm, the attenuation coefficient is set to 0.2. The time parameter is derived from the real-time clock signal of the assembly system to ensure that the phase of the sine function is synchronized with the actual vibration waveform. The boundary value of the dangerous frequency band is calibrated based on experimental data. For example, 80-120Hz is defined as the dangerous frequency band, within which the flexible shaft is prone to lateral swing.

[0117] Specifically, the vibration frequency is sampled in real time by an accelerometer at a sampling rate of 1000Hz, and the spectral distribution is obtained through a fast Fourier transform. When the main frequency component is detected to fall within the preset dangerous frequency band, a vibration suppression path correction is generated based on a sine function, whose amplitude is controlled by the amplitude attenuation coefficient and the phase is synchronized with the real-time clock. This correction is superimposed on the posture difference data to form a compensation trajectory that is in phase with the vibration, offsetting the path deviation caused by the vibration. When the vibration frequency is outside the dangerous frequency band, the suppression correction is reset to zero to avoid ineffective compensation for low-frequency or high-frequency noise. For example, when 105Hz vibration is detected during assembly, the system generates a correction with an amplitude of 0.5×sin(2π×105t) with an attenuation coefficient of 0.5, which directly acts on the XY plane motion axis of the robot arm.

[0118] During the flexible shaft assembly process, the vibration frequency data of the flexible shaft end is first collected in real time through an acceleration sensor. Furthermore, the dangerous frequency band range is pre-set to 20-50Hz. Therefore, when the vibration frequency is detected to fall into the 20-50Hz range, the vibration compensation model is activated. The amplitude attenuation coefficient can be set to 0.5mm, and the vibration frequency and time are obtained by real-time acquisition. For example, if the vibration frequency is detected to be 30Hz at a certain moment, the calculated vibration suppression path correction amount is 0.5×sin(2π×30×t)mm. As a preferred embodiment, this correction amount is superimposed on the original approximation path parameters to suppress the vibration of the flexible shaft during the assembly process. When the vibration frequency exceeds the range of 20-50Hz, the vibration suppression path correction amount is set to 0, that is, no vibration compensation is performed.

[0119] Through the above-mentioned technical solution, this application can effectively suppress vibration of the flexible shaft during assembly, improving assembly accuracy and stability. By real-time monitoring of vibration frequency and dynamic compensation, the adverse effects of vibration on assembly accuracy are avoided. Furthermore, by setting critical frequency bands, selective vibration suppression is achieved, avoiding unnecessary compensation and improving system efficiency. Furthermore, this solution utilizes a simple sinusoidal function model, which requires minimal computation and is easy to process in real time, making it suitable for high-speed assembly scenarios.

[0120] In some of the above-mentioned schemes of the present application, when the ambient temperature changes, the effective assembly length of the flexible shaft material will be deformed due to thermal expansion and contraction. If the thermal deformation is not accurately calculated, the path planning parameters will not match the actual physical deformation, thereby causing assembly position deviation.

[0121] The present application further proposes a calculation formula for the thermal deformation compensation amount: thermal deformation compensation amount = material thermal expansion coefficient × flexible shaft effective assembly length × (real-time temperature - reference temperature).

[0122] Among them, the material thermal expansion coefficient is the inherent physical property of the flexible shaft material, reflecting the rate of length change under unit temperature change; the effective assembly length of the flexible shaft is defined as the actual effective length from the end of the flexible shaft to the contact point of the target joint; the real-time temperature collects the thermodynamic data of the assembly environment through the temperature sensor; the reference temperature is set as the benchmark temperature when the flexible shaft is in a stress-free state.

[0123] Specifically, the calculation process of the compensation amount is as follows: First, the thermal expansion coefficient of the material corresponding to the flexible shaft is called from the material database. This parameter is obtained through laboratory calibration. Secondly, the effective assembly length of the flexible shaft is determined based on the geometric relationship between the end position of the flexible shaft and the target joint hole position fed back by the visual system. The real-time temperature is collected at a frequency of 10 times per second by the infrared sensor deployed in the assembly area, and the difference is calculated with the pre-stored reference temperature. Finally, the linear expansion caused by temperature change is quantified as a path compensation parameter through the product relationship. For example, when the flexible shaft material is stainless steel, its thermal expansion coefficient is 16×10 -6 / ℃, if the effective assembly length is 500 mm and the ambient temperature rises by 20℃, the thermal deformation compensation amount is 16×10 -6 ×500×20=0.16mm. This compensation is added to the posture difference data, allowing the robot arm control command to dynamically correct the motion trajectory of the flexible shaft end, offsetting the axial displacement deviation caused by temperature changes, thereby improving assembly accuracy.

[0124] As a preferred embodiment, the solution of the present application is specifically implemented as follows: During the assembly process of the flexible shaft, the generation of the thermal deformation compensation amount is achieved in the following manner: the built-in temperature sensor of the assembly system collects the ambient temperature data of the assembly area in real time, and the temperature data is calculated as a difference with the preset reference temperature value; the system calls the pre-stored thermal expansion coefficient parameters of the flexible shaft material, and combines the effective assembly length of the flexible shaft in the assembly direction to generate the axial thermal deformation compensation amount by multiplying the thermal expansion coefficient, the effective assembly length and the temperature difference; the compensation amount is converted into a displacement correction amount in the coordinate system, and directly superimposed on the translation difference component in the posture difference data, so that the robot arm control instruction automatically corrects the axial dimensional deviation caused by temperature changes.

[0125] Through the above technical solution, this application effectively eliminates the thermal deformation error of the flexible shaft caused by ambient temperature fluctuations. Through real-time temperature monitoring and linked calculation of material parameters, the path planning system can dynamically compensate for the axial displacement deviation caused by thermal expansion, avoiding the assembly jamming caused by axial misalignment between the flexible shaft and the joint hole due to temperature changes, and significantly improving the assembly reliability under complex working conditions.

[0126] In some of the above-mentioned schemes of the present application, the coordinate system of the visual system and the coordinate system of the six-dimensional force sensor may be relatively offset during the assembly process due to the movement of the robotic arm or environmental interference, resulting in inaccurate mapping between the force data and the visual posture data, affecting the accuracy of path correction.

[0127] This application further proposes to calibrate and align the coordinate system of the visual system with the coordinate system of the six-dimensional force sensor, including: updating the coordinate system transformation matrix data of the visual system and the six-dimensional force sensor through real-time calibration plate data, and mapping the force dynamic data to the dynamically updated visual coordinate system.

[0128] Among them, the real-time calibration plate contains multiple identifiable geometric marks, whose spatial positions can be solved by image processing algorithms in both the visual system and the six-dimensional force sensor coordinate systems; the coordinate system transformation matrix is ​​solved by constructing a least squares optimization model based on the coordinate difference of the calibration plate in the dual coordinate systems; the dynamically updated visual coordinate system corrects the parameters of the transformation matrix in real time by periodically collecting calibration plate data.

[0129] Specifically, the calibration plate is fixed to the end effector of the robotic arm and moves synchronously with the flexible shaft. The visual system captures the image of the calibration plate at a fixed frequency and solves its posture. The six-dimensional force sensor synchronously collects the force data in the current posture. The calculation process of the transformation matrix is ​​as follows: the coordinate data of the calibration plate in the visual coordinate system and the coordinate data in the six-dimensional force sensor coordinate system are input into the matrix solving algorithm to generate the translation vector and rotation component of the transformation matrix. When the movement of the robotic arm causes the posture of the calibration plate to change, the transformation matrix is ​​updated in real time through iterative optimization. The force dynamic data is mapped to the visual coordinate system through the current transformation matrix to ensure the spatial consistency of the visual posture data and the force data, thereby eliminating the path correction error caused by the coordinate system offset.

[0130] As a preferred embodiment, the solution of the present application is specifically implemented as follows: During the initialization stage of the assembly equipment, a calibration plate with a checkerboard pattern is fixed to the mounting base of the six-dimensional force sensor, and the image data of the calibration plate in different postures is collected through the binocular vision system. The internal and external parameters of the visual system are calculated using the Zhang calibration method, and the origin offset relative to the base coordinate system is collected through the force sensor. The initial transformation matrix of the visual coordinate system and the force sensor coordinate system is established based on the least squares method, and the matrix is ​​stored in the calibration database of the control system. During the assembly process, when it is detected that the posture change of the end effector of the robotic arm exceeds the preset threshold, the image data of the current calibration plate is reacquired, and the rotation and translation parameters of the transformation matrix are updated through the feature point matching algorithm to achieve dynamic alignment of the coordinate system. The contact force vector data collected by the force sensor is mapped to the visual coordinate system through the real-time updated transformation matrix to generate force feedback data synchronized with the three-dimensional coordinates of the end of the flexible shaft.

[0131] Through the above technical solution, this application effectively solves the problem of dynamic assembly error accumulation caused by static calibration of the visual system and force sensor coordinate system. By updating the transformation matrix in real time, the spatial consistency of force data and visual data is ensured during the movement of the robot arm, avoiding control errors caused by coordinate system offset due to mechanical vibration or temperature changes. The dynamic mapping mechanism enables the contact force feedback to accurately correspond to the real-time position and posture of the flexible shaft end, providing an accurate force-position synchronization data foundation for path correction during the assembly process.

[0132] In some of the above-mentioned solutions in this application, there is a problem that the real-time performance of multi-source data fusion and the closed-loop feedback of the robotic arm motion control are not considered during the dynamic correction process, resulting in the inability to compensate in time for sudden offsets caused by deformation or environmental interference during the flexible shaft assembly process, thereby causing assembly overload or decreased accuracy.

[0133] like Figure 4 As shown, the present application further proposes a visually guided positioning system for the tool flexible shaft assembly process, including a visual perception and posture solution module, a six-dimensional force perception monitoring module, a multi-source data fusion and coordinate system calibration module, an adaptive path planning module, a dynamic correction module and a robotic arm motion control and execution module.

[0134] The visual perception and posture calculation module collects the three-dimensional coordinates and posture angle data of the flexible shaft end through the visual system, and extracts the center coordinates and axis direction vector of the target joint hole;

[0135] The six-dimensional force perception monitoring module collects six-dimensional mechanical data in the assembly process in real time, including axial force, lateral force and moment around the axis;

[0136] The multi-source data fusion and coordinate system calibration module updates the coordinate transformation matrix of the visual system and the six-dimensional force sensor through real-time calibration plate data, realizing spatial synchronous mapping of force data and visual data;

[0137] The adaptive path planning module generates the path direction, movement speed and convergence conditions based on the dynamic data of the posture difference, where the movement speed is inversely proportional to the translation difference modulus length;

[0138] When the contact force or torque exceeds the safety threshold, the dynamic correction module uses the compensation coefficient and attenuation coefficient to calculate the compensation amount and superimposes it on the posture difference data;

[0139] The robot arm motion control and execution module converts the updated path parameters into robot arm joint motion instructions.

[0140] Specifically, the visual perception and pose calculation module uses a binocular camera to acquire 3D point cloud data of the flexible shaft end, applies the ICP algorithm to match the target joint hole geometry model, and outputs the center coordinates and axis vector. The six-dimensional force perception and monitoring module collects contact force and torque data at a 1000Hz sampling rate, eliminating noise interference through Kalman filtering. The multi-source data fusion and coordinate system calibration module performs dynamic calibration every 50ms, using singular value decomposition to calculate the coordinate transformation matrix to ensure spatial consistency between visual and force data within a unified coordinate system. The adaptive path planning module normalizes and combines the target joint axis direction vector with the translation difference data to generate the initial approach path direction. It also adjusts the movement speed based on a preset scaling factor and a zero-prevention constant. The dynamic correction module uses an incremental compensation algorithm. For example, when the axial force Fz exceeds a threshold, the compensation amount is iteratively calculated according to the formula ΔFz = 0.8 × (Fz - 10N) + 0.2 × ΔFz_prev, where 10N is the preset safety threshold, 0.8 is the compensation factor, and 0.2 is the attenuation factor. The robotic arm's motion control and execution module converts the corrected path parameters into a Cartesian trajectory at the end of the robotic arm. Using inverse kinematics, it solves for the angular displacement of each joint, ultimately executing the assembly action with 0.1mm positional accuracy. This system utilizes a closed-loop feedback mechanism based on visual and force data to suppress vibration interference and thermal deformation errors in real time during the flexible shaft contact phase, preventing path deviations caused by flexible deformation.

[0141] As a preferred embodiment, the solution of the present application is specifically implemented as follows: The visual guidance positioning system for the assembly process of the flexible shaft of the machine tool is composed of six functional modules. The visual perception and posture solution module uses a binocular stereo vision camera, which is installed 300 mm above the end effector of the robotic arm. The three-dimensional coordinates and Euler angle posture parameters of the end of the flexible shaft are obtained in real time through the binocular parallax solution algorithm. At the same time, the Hough circle detection algorithm is used to extract the hole center coordinates and axis direction vector from the target joint image. The six-dimensional force perception monitoring module is integrated in the wrist of the robotic arm and uses a strain-type six-dimensional force sensor to collect XYZ three-axis contact force and torque data around the axis during the assembly process at a sampling frequency of 1000Hz. The multi-source data fusion and coordinate system calibration module aligns the origin of the visual coordinate system with the origin of the force sensor coordinate system through the hand-eye calibration algorithm, and establishes a spatial transformation model containing a rotation matrix and a translation vector. The adaptive path planning module uses a proportional-differential control algorithm based on pose difference data to generate the movement trajectory of the flexible shaft end. The path direction is determined by the vector synthesis of the target joint axis and the pose translation difference, and the movement speed is inversely proportional to the pose difference modulus length. The dynamic correction module uses an incremental PID controller. When the contact force exceeds the preset threshold, the compensation amount is calculated based on the deviation of the six-dimensional mechanical data, with the compensation coefficient set to 0.5 and the attenuation coefficient set to 0.3. The robot arm motion control and execution module converts the updated path parameters into six-axis joint angle commands, and drives the servo motor to perform the assembly action through a real-time operating system with a 1ms cycle.

[0142] Through the above technical solutions, this application effectively solves the assembly failure problem caused by the accumulation of visual positioning errors and the lack of force feedback during the traditional flexible shaft assembly process. Through the spatial fusion of visual and force data, the precise position and posture of the flexible shaft end under occlusion conditions is achieved; the path correction mechanism based on dynamic mechanical compensation avoids component damage caused by assembly overload; adaptive path planning combined with vibration suppression and thermal deformation compensation significantly improves the assembly success rate of the flexible shaft under flexible deformation conditions, especially in precision assembly scenarios such as aircraft engine gearboxes, where the assembly accuracy can be stably maintained within the aperture matching tolerance range.

[0143] The technical scope of the present invention is not limited to the contents of the above description. Those skilled in the art can make various deformations and modifications to the above embodiments without departing from the technical idea of ​​the present invention, and these deformations and modifications should all fall within the protection scope of the present invention.

Claims

1. A visually guided positioning method for tool flexible shaft assembly, characterized by: The steps include: The three-dimensional coordinates and attitude angle data of the flexible shaft end are acquired through the visual system, and the center coordinates and axis direction vector data of the target joint hole are extracted; The six-dimensional force sensor is used to collect six-dimensional mechanical data in the flexible shaft assembly process in real time; Calibrate and align the coordinate system of the vision system with the coordinate system of the six-dimensional force sensor to generate unified spatial mapping relationship data; The pose difference data between the flexible shaft end and the target joint hole is calculated based on the three-dimensional coordinates and attitude angle data of the flexible shaft end, the center coordinates of the target joint hole and the axis direction vector data, and the approximation path parameters are generated based on the pose difference data; When the contact force data and torque data exceed a preset safety threshold or the posture difference dynamic data is less than a predetermined number of digits, a path correction parameter is constructed according to the six-dimensional mechanical data; The path correction parameters are superimposed on the posture difference data to update the approximation path parameters, and the robot arm control instruction data is generated based on the updated approximation path parameters; The constructing of path correction parameters according to the six-dimensional mechanical data includes: The six-dimensional mechanical data is obtained, the difference between the six-dimensional mechanical data and the preset safety threshold is calculated, and the compensation amount is calculated. The calculation formula of the compensation amount is as follows: Compensation amount = configurable compensation coefficient × (six-dimensional mechanical data - preset safety threshold) + compensation amount of the previous cycle × configurable attenuation coefficient.

2. The visually guided positioning method for tool flexible shaft assembly according to claim 1, characterized in that: The posture difference dynamic data includes translation difference data and attitude angle difference data, and the approximation path parameters include path direction, moving speed and convergence condition; The process of generating the path direction includes: normalizing the target joint axis direction vector and the translation difference data to generate a synthetic direction vector, which is used as the direction parameter of the initial approximation path; The moving speed is inversely proportional to the modulus of the translation difference data; The convergence condition is that the sum of squares of the translation difference data is less than a preset convergence threshold, and the preset convergence threshold is 80% to 120% based on the target joint aperture.

3. The visually guided positioning method for tool flexible shaft assembly according to claim 2, characterized in that: The process of generating the path direction includes: The calculation formula of the moving speed is: Speed ​​parameter = preset proportional coefficient / (modulus length of translation difference data + minimum constant to prevent zero elimination).

4. The visually guided positioning method for tool flexible shaft assembly according to claim 1, characterized in that: The six-dimensional mechanical data includes contact force data and torque data, wherein the contact force data includes axial force Fz, lateral force Fx, and lateral force Fy, and the torque data includes axial torques Mx, My, and Mz; the compensation amounts are set in six types, which correspond one-to-one to the data types in the six-dimensional mechanical data.

5. The visually guided positioning method for tool flexible shaft assembly according to claim 1, characterized in that: The preset safety threshold is the dynamic fluctuation range of the six-dimensional mechanical data generated based on the historical assembly data, and the predetermined digit is the three-dimensional coordinate difference between the end of the flexible shaft and the target joint hole position ≤ the dynamic convergence value.

6. The visually guided positioning method for tool flexible shaft assembly according to claim 1, characterized in that: Generating the approximation path parameters based on the posture difference data also includes: Real-time monitoring of vibration frequency data and ambient temperature data during the flexible shaft assembly process; Build a vibration compensation model based on vibration frequency data and generate vibration suppression path correction; Generate thermal deformation compensation based on the mapping relationship between ambient temperature data and thermal expansion coefficient of flexible shaft material; The vibration suppression path correction amount and thermal deformation compensation amount are added to the posture difference data to update the approximation path parameters.

7. The visually guided positioning method for tool flexible shaft assembly according to claim 5, characterized in that: The construction logic of the vibration compensation model is: When the vibration frequency is within the preset dangerous frequency band: Vibration suppression path correction amount = amplitude attenuation coefficient × sin(2π × vibration frequency × time); When the vibration frequency exceeds the dangerous frequency band: Vibration suppression path correction amount = 0.

8. The visually guided positioning method for tool flexible shaft assembly according to claim 5, characterized in that: The calculation formula of the thermal deformation compensation amount is: Thermal deformation compensation amount = material thermal expansion coefficient × flexible shaft effective assembly length × (real-time temperature - reference temperature).

9. The visually guided positioning method for tool flexible shaft assembly according to claim 1, characterized in that: The calibration and alignment of the coordinate system of the visual system and the coordinate system of the six-dimensional force sensor includes: updating the coordinate system conversion matrix data of the visual system and the six-dimensional force sensor through real-time calibration board data, and mapping the force dynamic data to the dynamically updated visual coordinate system.

10. The visual guidance positioning system for the tool flexible shaft assembly process is characterized by: The method for visually guiding positioning during the assembly of a flexible shaft of a tool as claimed in any one of claims 1 to 9 is implemented, comprising: The visual perception and posture calculation module is used to obtain the three-dimensional coordinates and posture angle data of the flexible shaft end through the visual system, and extract the center coordinates and axis direction vector data of the target joint hole; The six-dimensional force sensing monitoring module is used to collect six-dimensional mechanical data in the flexible shaft assembly process in real time through a six-dimensional force sensor; Multi-source data fusion and coordinate system calibration module, used to calibrate and align the coordinate system of the vision system with the coordinate system of the six-dimensional force sensor to generate unified spatial mapping relationship data; An adaptive path planning module is used to calculate the posture difference data between the flexible shaft end and the target joint hole based on the three-dimensional coordinates and attitude angle data of the flexible shaft end, the center coordinates of the target joint hole, and the axis direction vector data, and generate the approximate path parameters based on the posture difference data; A dynamic correction module, configured to construct a path correction parameter based on the six-dimensional mechanical data when the contact force data and the torque data exceed a preset safety threshold or the posture difference dynamic data is less than a predetermined number of digits; The robot arm motion control and execution module is used to superimpose the path correction parameters with the posture difference data, update the approximation path parameters, and generate the robot arm control instruction data based on the updated approximation path parameters; The constructing of path correction parameters according to the six-dimensional mechanical data includes: The six-dimensional mechanical data is obtained, the difference between the six-dimensional mechanical data and the preset safety threshold is calculated, and the compensation amount is calculated. The calculation formula of the compensation amount is as follows: Compensation amount = configurable compensation coefficient × (six-dimensional mechanical data - preset safety threshold) + compensation amount of the previous cycle × configurable attenuation coefficient.

Images