A mechanical arm execution control method for industrial production

Abstract

The present application relates to the technical field of industrial robot control, and more particularly to a mechanical arm execution control method for industrial production, which comprises the following steps: collecting parameters in real time; generating a geometric risk index; generating a mechanical risk index; image feature extraction; abnormal mechanical arm determination; adjusting mechanical arm determination; and generating an adjustment instruction. Through real-time monitoring of the tilt angle, plane offset distance, contact image features, and torque and force values of the mechanical arm end, a dual risk judgment model based on geometric offset and mechanical loading is constructed, and further combined with image texture fluctuation and posture change, the continuously abnormal mechanical arm is accurately positioned, and after adjustment, the geometric and mechanical risk indexes are re-evaluated, thereby ensuring the aluminum shell positioning and bolt assembly precision, effectively solving the problems of assembly quality fluctuation due to mechanical arm positioning error and inaccurate force control, and thus reducing production efficiency and increasing product defect rate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial robot control, and in particular to a robot execution control method for industrial production. BACKGROUND

[0002] With the rapid development of industrial automation and intelligent manufacturing, robots as core execution equipment undertake more and more complex and high-precision operation tasks on the production line. How to realize the precise trajectory control of the robot execution end and ensure the stability, accuracy and real-time response of its motion has become a key technical problem to improve production efficiency and product quality. The existing trajectory control method still has deficiencies in computational complexity, adaptability and robustness, and a more efficient and reliable control strategy is needed to meet the needs of modern industrial production.

[0003] Chinese patent application publication No. CN107398903A discloses a trajectory control method for the execution end of an industrial robot arm, which includes: providing a trajectory control method for the execution end of an industrial robot arm, which includes obtaining the mechanical structure parameters of the execution end of the industrial robot arm, determining the Jacobian matrix according to the mechanical structure parameters and the initial griding precision ε; determining the motion trajectory of the execution end of the industrial robot arm according to the processing requirements; calculating the reference position variable of the operation space k+1 time according to the Jacobian matrix, and using the reference position variable of the operation space k+1 time as a parameter to constrain the quadratic optimization approximation of the target function, to obtain the actual operation space position variable Xk+1 at k+1 time and the input control amount uk of the joint space at k time; the execution end of the industrial robot arm performs trajectory motion according to the actual operation space position variable Xk+1 at k+1 time and the input control amount uk of the joint space at k time.

[0004] It can be seen that the trajectory control method for the execution end of the industrial robot arm has the following problems: the method relies on the accurate calculation of the Jacobian matrix, the computational complexity is high, and the real-time performance is limited; the setting of the initial griding precision ε lacks adaptive adjustment mechanism, which may lead to insufficient trajectory planning precision; the constrained quadratic optimization process requires a large amount of computing resources, which is difficult to meet the high-frequency control update; the trajectory planning does not fully consider the changes of external disturbance and dynamic characteristics of the robot, and the robustness is weak; the calculation of the joint space input control amount depends on the model precision, and the error accumulation may affect the motion precision. SUMMARY

[0005] Therefore, the present application provides a robot execution control method for industrial production, which overcomes the problem of fluctuation of assembly quality due to positioning error and inaccurate force control of the robot in the prior art, thereby reducing production efficiency and increasing product defect rate.

[0006] To achieve the above object, the application provides a mechanical arm execution control method for industrial production, comprising:

[0007] Real-time collection of normal direction and horizontal plane inclination angle of each work mechanical arm end in the process of infusion pump assembly on the industrial assembly line, end center and target bolt hole center plane offset, image of the contact area between the end and the infusion pump shell when the work mechanical arm clamps the infusion pump shell, and contact force and torque of the end when the target bolt is tightened, each work mechanical arm alternately clamps the infusion pump shell and the target bolt moving along the industrial assembly line based on a preset torque threshold;

[0008] Generation of a geometric risk index according to the inclination angle and the plane offset;

[0009] Generation of a mechanical risk index according to the geometric risk index, the contact force, the torque and the preset torque threshold;

[0010] Feature extraction of the image according to the comparison result of the mechanical risk index and a preset mechanical standard index, to obtain first extraction features and second extraction features;

[0011] Determination of a plurality of first abnormal mechanical arms based on the first extraction features and the fluctuation of the plane offset, or determination of a plurality of second abnormal mechanical arms based on the second extraction features and the fluctuation of the inclination angle;

[0012] Determination of a plurality of adjustment mechanical arms according to the time distribution characteristics and the overlap degree of the first abnormal mechanical arms and the second abnormal mechanical arms within a preset determination period, and adjustment of the preset torque threshold according to the position of the adjustment mechanical arms;

[0013] Generation of an instruction for adjusting the adjustment mechanical arms based on the geometric risk index and the mechanical risk index regenerated after the preset torque threshold is adjusted within a preset adjustment period.

[0014] Further, the process of generating a geometric risk index according to the inclination angle and the plane offset comprises:

[0015] Generation of the geometric risk index according to the ratio of each inclination angle to the maximum value in all inclination angles, a preset angle weight, the ratio of each plane offset to the maximum value in all plane offsets, and a preset offset weight.

[0016] Further, the process of generating a mechanical risk index according to the geometric risk index, the contact force, the torque and the preset torque threshold comprises:

[0017] determining that the mechanical risk index needs to be generated when the geometric risk index is greater than a preset geometric standard index, and generating the mechanical risk index according to a ratio of each contact force to a maximum value in all contact forces, a preset contact force weight, a ratio of each torque to the preset torque threshold, and a preset torque weight.

[0018] Further, a process of feature extraction of the image according to a comparison result of the mechanical risk index and a preset mechanical standard index includes:

[0019] when the mechanical risk index is greater than or equal to the preset mechanical standard index, extracting features of the image by using a preset first extraction method to obtain the first extracted features;

[0020] when the mechanical risk index is less than the preset mechanical standard index, extracting features of the image by using a preset second extraction method to obtain the second extracted features.

[0021] Further, a process of extracting features of the image by using the preset first extraction method to obtain the first extracted features includes:

[0022] converting the image into a gray-scale image and performing adaptive histogram equalization on the gray-scale image to obtain an equalized gray-scale image;

[0023] performing edge detection on the equalized gray-scale image to extract a contour edge of the contact area of the infusion pump shell to obtain an edge map;

[0024] applying a preset filtering algorithm to the same equalized gray-scale image to collect texture responses of a plurality of spatial frequencies and texture responses of a plurality of directions to obtain a plurality of texture response maps;

[0025] performing morphological opening and closing operations on the edge map and each texture response map to obtain a processed edge feature map and a texture feature map, respectively;

[0026] concatenating the processed edge feature map and the texture feature map by channel to obtain a high-dimensional feature matrix;

[0027] performing principal component analysis dimension reduction on the high-dimensional feature matrix to obtain the first extracted features.

[0028] Further, a process of extracting features of the image by using the preset second extraction method to obtain the second extracted features includes:

[0029] down-sampling the image to a preset resolution by equal proportion and performing Gaussian smoothing on the down-sampled image to obtain a smoothed image;

[0030] performing a fast corner detection algorithm on the smoothed image to obtain a plurality of key points;

[0031] calculating a binary descriptor for each of the key points to generate a plurality of local binary feature vectors;

[0032] performing a Boolean aggregation on all the local binary feature vectors to obtain a global binary feature string;

[0033] applying Hamming distance normalization to the global binary feature string to obtain the second extracted feature.

[0034] Further, the process of determining a plurality of first abnormal mechanical arms based on the first extracted feature and fluctuation of the plane offset includes:

[0035] determining a first edge fluctuation intensity according to the Euclidean distance of the first extracted feature of the spatially adjacent work mechanical arms in each two pipeline directions within a preset determination period, and determining a maximum offset according to all the plane offsets within the preset determination period;

[0036] determining the work mechanical arm as the first abnormal mechanical arm according to a comparison result of the first edge fluctuation intensity and a preset first value edge fluctuation threshold, and a comparison result of the maximum offset and a preset offset threshold, to determine a plurality of first abnormal mechanical arms.

[0037] Further, the process of determining a plurality of second abnormal mechanical arms based on the second extracted feature and fluctuation of the tilt angle includes:

[0038] determining a second edge fluctuation intensity according to the Hamming distance of the second extracted feature of the spatially adjacent work mechanical arms in each two pipeline directions within a preset determination period, and determining a maximum angle according to all the tilt angles within the preset determination period;

[0039] determining the work mechanical arm as the second abnormal mechanical arm according to a comparison result of the second edge fluctuation intensity and a preset second value edge fluctuation threshold, and a comparison result of the maximum angle and a preset angle threshold, to determine a plurality of second abnormal mechanical arms.

[0040] Further, the process of determining a plurality of adjustment mechanical arms according to the time distribution characteristics and overlap degree of the first abnormal mechanical arms and the second abnormal mechanical arms within a preset determination period includes:

[0041] determining a first abnormal distribution degree according to the time stamp of the first abnormal mechanical arms within the preset determination period, and determining a second abnormal distribution degree according to the time stamp of the second abnormal mechanical arms within the preset determination period;

[0042] When the first abnormal distribution degree and the second abnormal distribution degree are both greater than a preset distribution degree threshold, the overlap degree is determined according to a number of times that each of the work robot arms is determined as the first abnormal robot arm or the second abnormal robot arm in a preset determination period;

[0043] When the overlap degree is greater than a preset overlap degree threshold, the work robot arm is determined as the adjustment robot arm, so as to determine a number of adjustment robot arms.

[0044] Further, the process of adjusting the preset torque threshold according to the position of the adjustment robot arm comprises:

[0045] The adjustment concentration degree is determined according to the position of the adjustment robot arm.

[0046] The preset torque threshold is adjusted based on a comparison result of the adjustment concentration degree and a preset concentration degree standard range.

[0047] Compared with the prior art, the beneficial effects of the present application are that, by monitoring the tilt angle of the end of the robot arm, the plane offset distance, the contact image features, and the multi-dimensional real-time monitoring of the tightening torque and force value, a dual-risk judgment model based on geometric offset and mechanical loading is constructed, and further combined with image texture fluctuation and posture change, the continuously abnormal robot arm is accurately positioned. By introducing time distribution and multi-arm overlap degree analysis, only the most needed adjustment of the robot arm group is driven by the spatial concentration degree of the torque threshold for adaptive correction, so that the force control sensitivity and the geometric safety range can dynamically match the current work layout. After adjustment, the geometric and mechanical risk indexes are re-evaluated and closed-loop fine-tuning instructions are generated, effectively avoiding single misjudgment and overcompensation, which not only guarantees the aluminum shell positioning and bolt assembly precision, but also maximally reduces the collision risk of arm group cross operation, achieves the stability and safety double requirements of high rhythm and multi-arm cooperation, and effectively solves the problems of reduced production efficiency and increased product defect rate caused by the fluctuation of assembly quality due to the positioning error of the robot arm and the inaccurate force control.

[0048] Further, by weighting and normalizing the tilt angle of the end of the work robot arm and the plane offset of the end center, the spatial posture and position deviation can be quantitatively reflected. The tilt angle reflects the stability of the end in the vertical direction, and too large an angle can cause the deflection of the contact surface, thereby affecting the clamping or alignment accuracy. The plane offset directly reflects the deviation of the end positioning in the two-dimensional working plane, and is a direct indicator for judging the assembly deviation. By normalizing the dimension difference and introducing a preset weight to adjust the relative importance of the two in different assembly tasks, the generated geometric risk index has physical meaning, adjustability and engineering adaptability, which is beneficial to the rapid perception and quantitative evaluation of spatial errors in a complex assembly environment with multiple stations and multiple arms.

[0049] Further, by introducing the contact force and torque parameter to generate the mechanical risk index after the geometric risk index exceeds the preset threshold, the identification of mechanical abnormalities in the assembly process is more targeted and timely. Since assembly errors are often accompanied by abnormal distribution of force or excessive torque, the greater the contact force or the closer or exceeding the threshold of the torque, the more obvious the impact on the assembly structure. Therefore, the ratio of the contact force to the maximum contact force and the ratio of the torque to the preset torque threshold are introduced as standardized indicators, and a preset weight is used for weighted fusion, which can further evaluate whether the actual physical action can cause stress concentration, connection failure or material damage to the structure when the geometric deviation has appeared, so as to timely warn and guide subsequent adjustment, effectively improving the stability and product consistency of the assembly process.

[0050] Further, by dynamically comparing the mechanical risk index with the preset standard index, different image feature extraction methods are intelligently selected to realize fine analysis of the assembly state. When the risk index is high, a more detailed first extraction method is used to capture key details and minor abnormalities, improving the accuracy of fault identification. When the risk index is low, a simplified second extraction method is used to ensure computational efficiency while still effectively reflecting overall assembly quality. This parameter-driven adaptive feature extraction strategy, combined with the mutual verification of mechanical and image information, promotes scientific judgment and optimal control of the system for complex assembly processes.

[0051] Further, the image is processed by adopting a preset first extraction method, the contrast and detail performance of the image are improved by converting to a gray image and adaptive histogram equalization, and the key contour information of the contact area of the infusion pump shell is effectively enhanced; clear contour edges are extracted by edge detection, combined with multi-scale and multi-direction texture response, noise is removed and feature performance is strengthened by morphological opening and closing operation, so that the edge and texture information are complementary, forming a rich and stable high-dimensional feature matrix; then principal component analysis is used for dimension reduction, redundant information is removed, and main change features are retained, realizing data compression and feature highlighting. This process reasonably utilizes the spatial frequency and structural features of the image, so that the finally extracted features can accurately reflect the slight changes of the mechanical arm end and the contact area of the infusion pump shell, thereby improving the sensitivity and reliability of the anomaly detection.

[0052] Further, the image is processed by adopting a preset second extraction method, noise and data volume are effectively reduced by downsampling and Gaussian smoothing, and the stability and continuity of key features are ensured; the significant local information in the image is accurately captured by a fast corner detection algorithm, and the generated binary descriptor is converted into a compact and representative global feature string by Boolean aggregation, and Hamming distance normalization ensures the comparability and robustness between features. These processing steps cooperate with each other, so that the extracted features can accurately reflect the image details of the mechanical arm end and the contact area of the infusion pump shell, thereby effectively assisting the judgment and adjustment of the mechanical arm state, and improving the precision and stability of the assembly process.

[0053] Further, by combining the Euclidean distance fluctuation of adjacent mechanical arms in the image feature space and the maximum plane offset in the physical space, the stability and position consistency of the mechanical arm end action are cross-verified in double dimensions, the calculated first edge fluctuation strength can effectively reflect the continuous change trend of the local assembly quality in the time window, and the maximum offset further limits the physical amplitude of the overall assembly deviation. When both exceed the corresponding threshold value, it is determined as the first abnormal mechanical arm, which can significantly enhance the sensitivity and discrimination accuracy of abnormal identification, avoid false positives or false negatives caused by single parameter fluctuation, and is beneficial to timely discover and intervene potential assembly deviation sources, and improve the overall operation stability and consistency of the assembly system.

[0054] Further, by jointly considering the volatility of image features and the deviation amplitude of the pose angle, more accurate identification of abnormal mechanical arms can be achieved on the basis of coupling of multi-dimensional information. Among them, the second edge fluctuation intensity reflects the consistency degree of adjacent mechanical arms in the expression of texture features, and the maximum tilt angle as a quantitative indicator of the stability of the end pose can reflect the significant degree of pose deviation in the assembly process. When the image feature fluctuation is severe and the pose deviation exceeds the normal operation threshold, it is often closely related to potential problems such as execution errors, structural looseness or work station interference, and the deviation degree of the mechanical arm running state can be comprehensively judged by using the cooperative change trend of image perception and pose perception, so as to improve the sensitivity and accuracy of abnormal detection, thereby providing a high-quality data basis for subsequent dynamic adjustment and precise operation and maintenance.

[0055] Further, by introducing the distribution statistics and judgment mechanism of the abnormal period, the time concentration trend of the abnormal behavior of the mechanical arm and the coincidence degree of the abnormal type can be comprehensively measured. On the one hand, by calculating the standard deviation of the time period corresponding to the first abnormality and the second abnormality, it is evaluated whether the abnormal distribution presents a wide dispersion, and then it is judged whether it has representativeness and persistence; on the other hand, the overlap index is constructed by combining the judgment frequencies of the first and second abnormalities, and the abnormal accumulation degree of a single mechanical arm in different dimensions is measured. If the time distribution fluctuation is significant and the abnormal overlap degree is high, it means that the mechanical arm shows persistent deviation in multiple feature domains, has a high adjustment priority, and can effectively avoid misjudgment caused by accidental abnormality or single indicator fluctuation, thereby improving the stability and pertinence of adjustment decision.

[0056] Further, by calculating the concentration of the adjusted mechanical arm position deviation, the preset torque threshold is dynamically adjusted, so that the adjustment of the threshold is closely related to the spatial distribution characteristics of the abnormal mechanical arm. When the adjustment concentration exceeds the preset upper limit, it means that the abnormal mechanical arms are concentrated on the industrial assembly line, and increasing the torque threshold can improve the clamping and tightening force of the mechanical arm, which helps to relieve local overload and improve operation stability; on the contrary, when the adjustment concentration is lower than the preset lower limit, it means that the abnormal distribution is relatively dispersed, and by appropriately reducing the torque threshold, component damage and assembly errors caused by excessive force can be avoided, and the flexibility and adaptability of the overall assembly are enhanced. Relying on the Euclidean distance of the position of the abnormal mechanical arm and its statistical characteristics, the spatial fluctuation of the operation state between the mechanical arms can be effectively reflected. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 A flowchart of the mechanical arm execution control method for industrial production of the present embodiment;

[0058] Figure 2 A determination logic diagram for determining whether to generate the mechanical risk index of the present embodiment;

[0059] Figure 3A determination logic diagram for feature extraction of an image in the embodiment;

[0060] Figure 4 A determination logic diagram for determining a first abnormal robot arm in the embodiment. DETAILED DESCRIPTION

[0061] In order to make the objects and advantages of the present application clearer, the present application will be further described below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0062] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. It should be understood by those skilled in the art that the embodiments are only used to explain the technical principles of the present application and are not used to limit the protection scope of the present application.

[0063] Please refer to Figure 1 As shown in the figure, it is a flowchart of a robot arm execution control method for industrial production in the embodiment. The embodiment provides a robot arm execution control method for industrial production, which comprises the following steps:

[0064] Real-time collection of a normal direction of an end of each work robot arm and a tilt angle with a horizontal plane, a plane offset between an end center and a target bolt hole center, an image of a contact area between the end and a transfusion pump shell when the work robot arm clamps the transfusion pump shell, a contact force and a torque of the end when a target bolt is tightened during the process of assembling the transfusion pump on an industrial assembly line, and each work robot arm alternately clamping the transfusion pump shell and the target bolt moving along the industrial assembly line based on a preset torque threshold value;

[0065] Generating a geometric risk index according to the tilt angle and the plane offset;

[0066] Generating a mechanical risk index according to the geometric risk index, the contact force, the torque and the preset torque threshold value;

[0067] Feature extraction of the image according to a comparison result of the mechanical risk index and a preset mechanical standard index, to obtain a first extracted feature and a second extracted feature;

[0068] Determining a plurality of first abnormal robot arms based on the first extracted feature and fluctuation changes of the plane offset, or determining a plurality of second abnormal robot arms based on the second extracted feature and fluctuation changes of the tilt angle;

[0069] Determining a plurality of adjustment robot arms according to time distribution characteristics and an overlap degree of the first abnormal robot arms and the second abnormal robot arms within a preset determination period, and adjusting the preset torque threshold value according to the positions of the adjustment robot arms;

[0070] Generate an instruction for adjusting the adjustment robot based on the geometric risk index and the mechanical risk index regenerated after adjusting the preset torque threshold within a preset adjustment period.

[0071] In this embodiment, the industrial pipeline is composed of 12 consecutive stations (S1-S12), each station is 1m wide and 0.8m long, and two six-axis robots (numbered M1-M24) are fixed on the left and right sides at equal intervals, the base is aligned with the global coordinate system through the ground calibration board, the horizontal distance between the robots is 900mm, and the end motion covers a 300mm x 300mm assembly sub-area. Each robot end is equipped with a 200Hz six-axis IMU to collect inclination angles, and a 2MP industrial camera is arranged 50mm away from the suction cup on the arm end side to synchronously capture images of the shell contact area; a pair of RGB-D depth cameras is erected every 1.6m at 1.2m above the bridge, and through a calibration matrix, the depth map and camera intrinsic parameters are mapped to the station plane to realize millimeter-level plane distance measurement between the end center and the bolt hole center; a 1kHz six-degree-of-freedom force / torque sensor is integrated between the arm end flange and the tool, and real-time data is uploaded to the PLC through the EtherCAT network.

[0072] In this embodiment, the plane offset between the end center and the target bolt hole center refers to the difference in horizontal plane distance between the center position of the end tool (such as the action point of the clamp or electric screwdriver) of the industrial robot and the preset target bolt hole center on the infusion pump during the process of clamping the infusion pump shell and performing tightening operation. This offset reflects the deviation of the robot positioning accuracy and the clamping posture, and is affected by the relative relationship between the robot structure, joint layout, end effector geometry, and infusion pump shell and bolt hole position. The size of the offset is directly related to the accuracy of the tightening operation and the assembly quality, and excessive offset may cause uneven tightening torque distribution and even damage to the parts, so it is an important parameter for risk judgment and adjustment control.

[0073] The preset torque threshold is the maximum torque limit allowed by the robot when tightening the target bolt, which depends on the specifications, material strength, and assembly process requirements of the infusion pump bolt, and is usually set between 0.5 and 5 Newton-meters. In this embodiment, it is set to 2.5 Newton-meters, which can effectively prevent damage to parts caused by over-tightening while ensuring that the connection strength meets the process standards.

[0074] The preset mechanical standard index is used to determine whether the mechanical risk exceeds the safe range, and its value depends on the mechanical load distribution of the robot working environment and the mechanical bearing capacity of the assembly workpiece, and is usually set between 0.3 and 0.8. In this embodiment, it is set to 0.6, which can accurately distinguish between normal assembly state and potential abnormality and ensure the mechanical stability of the assembly process.

[0075] The preset determination period is a time window for statistics and determination of abnormal robots, which depends on the assembly line beat and robot action frequency, and is usually set between 5 minutes and 30 minutes. In the embodiment, it is set to 15 minutes, which can balance the timeliness of data acquisition and the stability of statistics, and ensure the accuracy of abnormal determination.

[0076] The preset adjustment period is a time interval for preset torque threshold adjustment, which depends on the production rhythm and the response speed of the adjustment strategy, and is usually set between 30 minutes and 2 hours. In the embodiment, it is set to 1 hour, which can respond to changes in the abnormal state of the robot in a timely manner, dynamically optimize the torque control strategy, and improve the overall assembly efficiency and quality stability.

[0077] By monitoring the tilt angle, plane offset distance, contact image features, and tightening torque and force value of the robot end in multiple dimensions in real time, a dual-risk judgment model based on geometric offset and mechanical loading is constructed, and further combined with image texture fluctuation and posture change, the continuously abnormal robot is accurately positioned. By introducing time distribution and multi-arm overlap analysis, only the most need to adjust the robot group is driven for adaptive correction of the spatial aggregation degree of the torque threshold, so that the force control sensitivity and the geometric safety range can dynamically match the current work layout. After adjustment, re-evaluate the geometric and mechanical risk indexes and generate closed-loop fine-tuning instructions, effectively avoiding single misjudgment and overcompensation, which not only guarantees the aluminum shell positioning and bolt assembly precision, but also maximally reduces the collision risk of arm group cross operation, achieves the stability and safety double requirements of high beat and multi-arm cooperation, and effectively solves the problems of assembly quality fluctuation caused by robot positioning error and inaccurate force control, and the problems of production efficiency reduction and product defect rate rising.

[0078] Specifically, the process of generating a geometric risk index according to the tilt angle and the plane offset amount includes:

[0079] The geometric risk index is generated according to the ratio of each tilt angle to the maximum value in all tilt angles, a preset angle weight, each plane offset amount, and the ratio of the maximum value in all plane offset amounts to a preset offset amount weight, Q = w1 x R / Rmax + w2 x D / Dmax, wherein Q is the geometric risk index, w1 is the preset angle weight, R is the plane offset amount, Rmax is the maximum value in all plane offset amounts, w2 is the preset offset amount weight, D is the plane offset amount, and Dmax is the maximum value in all plane offset amounts.

[0080] The preset angle weight and the preset offset weight are respectively used for adjusting the relative influence degree of the tilt angle and the plane offset in the geometric risk index, the values of the two depend on the requirements for the posture stability and the position accuracy in the specific assembly task, are usually set between [0.3, 0.7], and satisfy w1+w2=1. In the embodiment, in order to balance the requirements for the posture and the position in the process of assembling the infusion pump shell, w1=0.4 and w2=0.6 are set, so that the assembly risk caused by the position offset of the end can be effectively identified while ensuring the clamping stability.

[0081] By weighting and normalizing the tilt angle of the end of the working mechanical arm and the plane offset of the end center, the comprehensive influence of the spatial posture and the position deviation on the assembly accuracy can be quantitatively reflected. The tilt angle reflects the posture stability of the end in the vertical direction, and an excessively large angle is easy to cause the deflection of the contact surface, thereby affecting the clamping or alignment accuracy; and the plane offset directly reflects the deviation degree of the end positioning on the two-dimensional working plane, and is a direct index for judging the assembly deviation. By normalizing the dimension difference and introducing the preset weight to adjust the relative importance of the two in different assembly tasks, the generated geometric risk index has physical meaning, adjustability and engineering adaptability, and is beneficial to the rapid perception and quantitative evaluation of the spatial error in a complex assembly environment with multiple stations and multiple arms cooperating. The index is not only convenient for realizing the horizontal comparison of the performance between the mechanical arms, but also provides a unified quantitative basis for subsequent risk assessment and abnormality judgment.

[0082] Referring to FIG. 8, Figure 2 FIG. 8 is a determination logic diagram for determining whether the mechanical risk index needs to be generated, according to the embodiment. In the embodiment, the process of generating the mechanical risk index according to the geometric risk index, the contact forces, the torques and the preset torque threshold value includes:

[0083] When the geometric risk index is greater than the preset geometric standard index, it is determined that the mechanical risk index needs to be generated, and the mechanical risk index is generated according to the ratio of each contact force to the maximum value of all contact forces, a preset contact force weight, the ratio of each torque to the preset torque threshold value and a preset torque weight, L=w3xF / Fmax+w4xU / U', wherein L is the mechanical risk index, w3 is the preset contact force weight, F is the contact force, Fmax is the maximum value of all contact forces, w4 is the preset torque weight, U is the torque, and U' is the preset torque threshold value.

[0084] The preset contact force weight is a parameter for measuring the influence degree of the contact force in the mechanical risk index, depends on the specific requirements of the assembly process for the sensitivity of the contact force, is usually set between 0.3 and 0.7, and is set to 0.5 in the embodiment, so that the contribution of the contact force to the overall risk assessment can be balanced, and the abnormal force action can be effectively identified.

[0085] The preset torque weight is used to adjust the weight of the torque parameter in the mechanical risk index, and is set between 0.3 and 0.7 according to the mechanical connection strength and safety margin of the assembled product, and is set to 0.5 in the embodiment, which can reasonably reflect the influence of torque abnormality on assembly quality and avoid excessive or insufficient response to torque fluctuation.

[0086] The preset geometric standard index is a threshold for determining whether the geometric risk needs further mechanical analysis, and is set between 0.4 and 0.7 according to the assembly accuracy requirement and error tolerance, and is set to 0.6 in the embodiment, which can effectively screen out the working conditions with significant geometric deviation and provide basis for subsequent mechanical risk assessment.

[0087] By introducing the process of generating the mechanical risk index by combining the contact force and the torque parameter when the geometric risk index exceeds the preset threshold, the identification of mechanical abnormalities in the assembly process is more targeted and timely. Since assembly errors are often accompanied by abnormal distribution of force or over-torque phenomenon, the greater the contact force or the closer or exceeding the threshold of the torque, the more obvious the influence on the assembly structure. Therefore, the ratio of the contact force to the maximum contact force and the ratio of the torque to the preset torque threshold are introduced as standardized indexes, and a preset weight is used for weighted fusion, which can further evaluate whether the actual physical action may cause stress concentration, connection failure or material damage when the geometric deviation has appeared, so as to timely warn and guide subsequent adjustment, and effectively improve the stability and product consistency of the assembly process.

[0088] Referring to FIG. 8, Figure 3 The process of extracting features from the image according to the comparison result of the mechanical risk index and the preset mechanical standard index to obtain the first extracted feature and the second extracted feature includes:

[0089] When the mechanical risk index is greater than or equal to the preset mechanical standard index, a preset first extraction method is used to extract features from the image to obtain the first extracted feature;

[0090] When the mechanical risk index is less than the preset mechanical standard index, a preset second extraction method is used to extract features from the image to obtain the second extracted feature.

[0091] By dynamically comparing the mechanical risk index with the preset standard index, different image feature extraction methods are intelligently selected to realize fine analysis of the assembly state. When the risk index is high, a more detailed first extraction method is used to capture key details and minor abnormalities, improving the accuracy of fault identification. When the risk index is low, a simplified second extraction method is used to ensure computational efficiency while still effectively reflecting overall assembly quality. This parameter-driven adaptive feature extraction strategy, combined with the mutual verification of mechanical and image information, promotes scientific judgment and optimal control of complex assembly processes.

[0092] Specifically, the process of extracting features from the image using the preset first extraction method includes:

[0093] Converting the image into a grayscale image and performing adaptive histogram equalization on the grayscale image to obtain an equalized grayscale image;

[0094] Performing edge detection on the equalized grayscale image to extract the contour edges of the infusion pump shell contact area, obtaining an edge map;

[0095] Applying a preset filtering algorithm to the same equalized grayscale image to collect texture responses of several spatial frequencies and texture responses of several directions, obtaining several texture response maps;

[0096] Performing morphological opening and closing operations on the edge map and each texture response map, respectively, to obtain processed edge feature maps and texture feature maps;

[0097] Concatenating the processed edge feature maps and the texture feature maps by channel to obtain a high-dimensional feature matrix;

[0098] Performing principal component analysis dimension reduction on the high-dimensional feature matrix to obtain the first extraction features.

[0099] The image is processed using the preset first extraction method. By converting to a grayscale image and performing adaptive histogram equalization, the contrast and detail performance of the image are improved, effectively enhancing the key contour information of the infusion pump shell contact area. Edge detection extracts clear contour edges, combined with multi-scale and multi-directional texture responses, through morphological opening and closing operations to remove noise and enhance feature performance, making the edge and texture information complementary, forming a rich and stable high-dimensional feature matrix. Subsequently, principal component analysis dimension reduction is used to remove redundant information and retain main variation features, achieving data compression and feature highlighting. This process reasonably utilizes the spatial frequency and structural features of the image, enabling the final extracted features to accurately reflect the minor changes of the mechanical arm end and the infusion pump shell contact area, thereby improving the sensitivity and reliability of abnormal detection.

[0100] Specifically, the process of performing feature extraction on the image by using a preset second extraction method to obtain the second extraction feature includes:

[0101] down-sampling the image to a preset resolution by equal proportion, and performing Gaussian smoothing on the down-sampled image to obtain a smoothed image;

[0102] performing a fast corner detection algorithm on the smoothed image to obtain a plurality of key points;

[0103] calculating a binary descriptor for each key point to generate a plurality of local binary feature vectors;

[0104] performing Boolean aggregation processing on all the local binary feature vectors to obtain a global binary feature string;

[0105] applying Hamming distance normalization to the global binary feature string to obtain the second extraction feature.

[0106] The preset resolution refers to the target pixel size obtained by down-sampling the original image by equal proportion, which depends on the richness of image details and the computational complexity of subsequent feature extraction algorithms, and is usually set between 64x64 and 256x256 pixels. In this embodiment, it is set to 128x128 pixels, which can effectively reduce the computational burden while ensuring the integrity of the key image information, improving the speed and accuracy of feature extraction.

[0107] By using a preset second extraction method to process the image, noise and data volume are effectively reduced through down-sampling and Gaussian smoothing, ensuring the stability and continuity of key features; the fast corner detection algorithm accurately captures the significant local information in the image, and the generated binary descriptor is converted into a compact and representative global feature string through Boolean aggregation, and Hamming distance normalization ensures the comparability and robustness between features. These processing steps cooperate with each other, so that the extracted features can accurately reflect the image details of the contact area between the end of the mechanical arm and the shell of the infusion pump, thereby effectively assisting the determination and adjustment of the state of the mechanical arm, improving the precision and stability of the assembly process.

[0108] Referring to Figure 4 the determination logic diagram of the first abnormal mechanical arm determined by the embodiment, in this embodiment, the process of determining a plurality of first abnormal mechanical arms based on the fluctuation of the first extraction feature and the plane offset includes:

[0109] In a preset determination period, the standard deviation of the Euclidean distance of the first extraction feature corresponding to the work mechanical arm in each two pipeline directions is calculated to determine the first edge fluctuation strength, and the maximum value of all the plane offsets in the preset determination period is selected to determine the maximum offset.

[0110] When the first edge fluctuation intensity is greater than a preset value edge fluctuation threshold and the maximum offset is greater than a preset offset threshold, the work robot arm is determined as the first abnormal robot arm to determine a plurality of first abnormal robot arms.

[0111] The preset value edge fluctuation threshold is a judgment criterion for measuring whether the fluctuation degree of the image feature of the adjacent robot arm is abnormal, and the value thereof usually depends on the historical statistical range of the image feature change in the collaborative assembly process between robot arms and the tolerance requirement of the image consistency for the current task, and is usually set to be between 0.05 and 0.25 (unit: standardized Euclidean distance standard deviation), and is set to 0.12 in this embodiment, which can effectively screen out the robot arm with large image feature fluctuation, so that the system can more sensitively identify the potential assembly quality abnormality caused by the image sensing difference, and improve the efficiency of assembly consistency management.

[0112] The preset offset threshold is used to define the maximum plane offset range allowed between the work robot arm end and the target bolt hole, and the value thereof depends on the product structure size tolerance, positioning accuracy requirement and tool alignment capability, and is usually set to be between 1.0 mm and 3.5 mm, and is set to 2.0 mm in this embodiment, which can ensure that when the robot arm end deviates from the target position by a range exceeding the normal process tolerance range, the system can be marked as an abnormal state in time, thereby avoiding the risk of mispositioning assembly or poor connection, and ensuring the product structure precision and assembly reliability.

[0113] By combining the Euclidean distance fluctuation of the adjacent robot arm in the image feature space and the maximum plane offset in the physical space, the stability and position consistency of the robot arm end action are verified in double dimensions, the calculated first edge fluctuation intensity can effectively reflect the continuous change trend of the local assembly quality in the time window, and the maximum offset further limits the physical amplitude of the overall assembly deviation. When both of them exceed the corresponding threshold value, it is determined that the first abnormal robot arm can significantly enhance the sensitivity and discrimination accuracy of abnormal identification, avoid false positives or false negatives caused by single parameter fluctuation, and is beneficial to timely discovering and intervening potential assembly deviation sources, and improving the overall operation stability and consistency of the assembly system.

[0114] Specifically, the process of determining a plurality of second abnormal robot arms based on the fluctuation change of the second extracted features and the inclination angles includes:

[0115] In a preset determination period, the standard deviation of the Hamming distance of the second extracted features of the work robot arms spatially adjacent in each two pipeline directions is calculated to determine a second edge fluctuation intensity, and the maximum value of all the inclination angles in the preset determination period is selected to determine a maximum angle.

[0116] determining the second abnormal mechanical arm when the second edge fluctuation intensity is greater than a preset binary edge fluctuation threshold value and the maximum angle is greater than a preset angle threshold value, to determine a plurality of second abnormal mechanical arms.

[0117] The preset determination period is a time window length for abnormality detection during system operation, and is determined by the standard assembly tempo of a single product on the assembly line, the acquisition frequency, and the linkage rhythm between multiple mechanical arms. The preset determination period is usually set to be between 10 seconds and 120 seconds, and is set to 30 seconds in the embodiment. The preset determination period can ensure the response speed of the system while accumulating sufficient sampling data, so that the fluctuation characteristics have statistical significance, thereby improving the stability and robustness of abnormality identification.

[0118] The preset binary edge fluctuation threshold value is a fluctuation intensity threshold value for representing whether the change degree of the image feature is abnormal, and is determined by the sensitivity of the image feature extraction method, the natural difference degree of the image structure of adjacent workstations, and the image acquisition definition. The preset binary edge fluctuation threshold value is usually set to be between 0.05 and 0.5, and is set to 0.18 in the embodiment. The preset binary edge fluctuation threshold value can determine potential abnormalities in time when the image difference between adjacent mechanical arms is too large, and effectively reflect the problem of reduced consistency of workstation operation.

[0119] The preset angle threshold value is a reference standard value for determining whether the deviation of the end posture of the mechanical arm is too large, and is determined by the assembly tolerance requirement of the assembly object, the influence degree of the end posture on the contact precision, and the measurement accuracy of the IMU sensor. The preset angle threshold value is usually set to be between 1° and 10°, and is set to 3.5° in the embodiment. The preset angle threshold value can effectively identify the risk of contact misalignment or assembly failure caused by posture drift, and improve the precise control ability of the mechanical arm positioning.

[0120] By jointly considering the fluctuation of the image feature and the deviation amplitude of the posture angle, more accurate identification of the abnormal mechanical arm can be achieved on the basis of the coupling of multi-dimensional information. The second edge fluctuation intensity reflects the consistency degree of the adjacent mechanical arms in the expression of the texture feature, and the maximum inclination angle as a quantitative index of the stability of the end posture can reflect the significant degree of the posture deviation in the assembly process. When the image feature fluctuates sharply and the posture deviation exceeds the normal operation threshold value, it is often closely related to potential problems such as execution error, structure loosening, or workstation interference. The fluctuation trend of the image perception and the posture perception can be used to comprehensively judge the deviation degree of the mechanical arm running state, thereby improving the sensitivity of the abnormality detection and the accuracy of the determination, and providing a high-quality data basis for subsequent dynamic adjustment and precise operation and maintenance.

[0121] Specifically, the process of determining a plurality of adjustment mechanical arms according to the time distribution characteristics and the overlap degree of the first abnormal mechanical arm and the second abnormal mechanical arm in the preset determination period includes:

[0122] respectively obtain the time length of each timestamp and initial time of the first abnormal mechanical arm and the second abnormal mechanical arm in the preset determination period, and obtain a plurality of first time periods and a plurality of second time periods respectively;

[0123] calculate the standard deviation of all first time periods to obtain the first abnormal distribution degree;

[0124] calculate the standard deviation of all second time periods to obtain the second abnormal distribution degree;

[0125] When the first abnormal distribution degree and the second abnormal distribution degree are both greater than a preset distribution degree threshold, determine the overlap degree according to the number of times each work mechanical arm is determined to be the first abnormal mechanical arm or the second abnormal mechanical arm in the preset determination period, G=(N1+N2) / N', wherein G is the overlap degree of the work mechanical arm, N1 is the number of times the work mechanical arm is determined to be the first abnormal mechanical arm, N2 is the number of times the work mechanical arm is determined to be the second abnormal mechanical arm, and N' is the total number of determination rounds in the preset determination period;

[0126] When the overlap degree is greater than a preset overlap degree threshold, determine the work mechanical arm to be the adjustment mechanical arm to determine a plurality of adjustment mechanical arms.

[0127] The preset distribution degree threshold is a reference standard for measuring the distribution dispersion degree of abnormal time periods in the preset determination period, and is in seconds, and depends on the work pace, the action period of each mechanical arm, and the environmental interference frequency. It is usually set between 5 seconds and 30 seconds, and is set to 15 seconds in this embodiment, which can effectively eliminate isolated abnormalities caused by accidental factors and screen out mechanical arms with relatively widespread abnormal distribution and systematic problem tendency in the entire period, providing a stable and reliable basis for subsequent adjustment.

[0128] The preset overlap degree threshold is used to determine whether the same mechanical arm shows high consistency in abnormal behavior in different risk identification dimensions (first extraction feature and second extraction feature), and is a dimensionless ratio, which depends on the abnormality determination frequency, the detection period length, and the task complexity. It is usually set between 0.2 and 0.6, and is set to 0.45 in this embodiment, which can comprehensively reflect the superposition strength of the abnormal behavior of a single mechanical arm, thereby avoiding false adjustment caused by local index fluctuations and improving the accuracy of overall diagnosis and the precision of adjustment decision.

[0129] By introducing the distribution statistics and judgment mechanism of the abnormal period, the time concentration trend and the coincidence degree of the abnormal type of the abnormal behavior of the robot arm can be comprehensively measured. On the one hand, by calculating the standard deviation of the corresponding time period of the first abnormality and the second abnormality, it is evaluated whether the abnormal distribution presents a wide dispersion, and then it is judged whether it has representativeness and persistence; on the other hand, an overlap index is constructed combining the judgment frequencies of the first and second abnormalities to measure the abnormal accumulation degree of a single robot arm in different dimensions. If the time distribution fluctuation is significant and the abnormal overlap degree is high, it means that the robot arm shows persistent deviation in multiple feature domains, has a high adjustment priority, can effectively avoid misjudgment caused by accidental abnormality or single indicator fluctuation, and thus improves the stability and pertinence of the adjustment decision.

[0130] Specifically, the process of adjusting the preset torque threshold according to the position of the adjustment robot arm includes:

[0131] Obtaining the Euclidean distance between the position of each adjustment robot arm in the industrial pipeline direction and the initial position, obtaining a plurality of adjustment distances;

[0132] Calculating the reciprocal of the standard deviation of all adjustment distances to determine the adjustment concentration;

[0133] When the adjustment concentration is greater than the maximum value of the preset concentration standard range, the preset torque threshold U'' is increased according to the relative deviation of the adjustment concentration and the maximum value of the preset concentration standard range and the preset adjustment coefficient, U''=U'×[1+s×(Y-Ymax) / Ymax], wherein U'' is the adjusted preset torque threshold, s is the preset adjustment coefficient, Y is the adjustment concentration, and Ymax is the maximum value of the preset concentration standard range;

[0134] When the adjustment concentration is less than the minimum value of the preset concentration standard range, the preset torque threshold is decreased according to the relative deviation of the minimum value of the preset concentration standard range and the adjustment concentration and the preset adjustment coefficient, U''=U'×[1-s×(Ymin-Y) / Y], wherein Ymin is the minimum value of the preset concentration standard range.

[0135] The preset concentration standard range refers to the numerical interval used to judge whether the spatial distribution of the adjustment robot arm is too concentrated or dispersed, which depends on the layout density and operation distance of the robot arm on the industrial pipeline, and is usually set between [0.1, 0.5], and is set to [0.2, 0.4] in this embodiment. It can accurately reflect the spatial fluctuation degree of the robot arm position, thereby providing a scientific basis for reasonable adjustment of the torque threshold, and avoiding affecting the assembly quality due to too concentrated or dispersed abnormal robot arms.

[0136] The preset adjustment coefficient is a scale factor for adjusting the torque threshold adjustment range, and is usually set between 0.01 and 0.1 according to the force sensitivity of the robot arm in the assembly process and the process tolerance requirement, and is set to 0.05 in the embodiment, which can ensure that the adjustment of the torque threshold is neither too violent to damage the equipment nor too conservative to affect the work efficiency, and can realize smooth and flexible adjustment of the torque control, and improve the adaptability and safety of the robot arm as a whole.

[0137] By calculating the concentration of the position offset of the adjustment robot arm, the preset torque threshold is dynamically adjusted, so that the adjustment of the threshold is closely related to the spatial distribution characteristics of the abnormal robot arm. When the adjustment concentration exceeds the preset upper limit, it indicates that the abnormal robot arms are concentrated on the industrial assembly line, and increasing the torque threshold can improve the clamping and tightening force of the robot arm, which helps to relieve local overload and improve work stability; on the contrary, when the adjustment concentration is lower than the preset lower limit, it indicates that the abnormal distribution is relatively dispersed, and by appropriately reducing the torque threshold, component damage and assembly errors caused by excessive force can be avoided, and the flexibility and adaptability of the overall assembly are enhanced, which effectively reflects the spatial fluctuation of the work state between the robot arms depending on the Euclidean distance of the position of the abnormal robot arm and its statistical characteristics.

[0138] Specifically, the process of generating an instruction for adjusting the adjustment robot arm based on the recalculated geometric risk index and the recalculated mechanical risk index after adjusting the preset torque threshold within the preset adjustment period includes:

[0139] In the embodiment, the process of generating an instruction for adjusting the adjustment robot arm based on the recalculated geometric risk index and the recalculated mechanical risk index after adjusting the preset torque threshold within the preset adjustment period includes:

[0140] The position adjustment amount of the robot arm end is calculated as ΔP = w1 x (Qnew-Q / Q) x vg + w2 x (Lnew-L) / L x vl, where ΔP is the position adjustment amount of the robot arm end, Qnew is the recalculated geometric risk index, Lnew is the recalculated mechanical risk index, vg is the preset geometric velocity coefficient corresponding to the adjustment speed of the geometric risk, vl is the preset mechanical velocity coefficient corresponding to the adjustment speed of the mechanical risk, and the unit is meter (m / s);

[0141] The adjustment instruction is generated according to the current position and the adjustment amount of the end of each adjustment robot arm.

[0142] The adjustment amount of the end position of the mechanical arm is calculated based on the change of the geometric risk index and the mechanical risk index before and after adjustment. First, the change proportions of the geometric risk index and the mechanical risk index relative to the adjustment before are calculated, and these proportions reflect the error change trend of the mechanical arm in the spatial position and the plane attitude. Then, according to the change proportions of the two risk indexes, the direction and amplitude of the end of the mechanical arm that needs to be adjusted are determined in combination with the preset weight coefficient. The adjustment direction is jointly determined by the error directions indicated by the geometric risk and the mechanical risk, and it is ensured that the adjustment is aimed at the specific deviation. The overall adjustment amount is controlled by the proportion coefficient to avoid that the adjustment is too large to cause the mechanical arm to be unstable or to exceed the working range. Finally, the calculation result of the adjustment amount is applied to the position and attitude correction of the end of the mechanical arm, so as to realize the accurate control of the clamping and tightening action of the mechanical arm and improve the stability of the assembly process and the product quality.

[0143] The above only describes the preferred embodiments of the present application and is not used to limit the present application; for those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A robotic arm execution control method for industrial production, characterized in that, include: The system collects real-time data on the tilt angle of the normal direction of the end of each robotic arm relative to the horizontal plane, the plane offset between the end center and the center of the target bolt hole, the contact area between the end of the robotic arm and the end of the robotic arm when clamping the infusion pump housing, and the contact force and torque of the end when the target bolt is tightened. Each robotic arm takes turns clamping the infusion pump housing and the target bolt moving along the industrial assembly line based on a preset torque threshold. A geometric risk index is generated based on the tilt angle and the plane offset; A mechanical risk index is generated based on the geometric risk index, the contact force, the torque, and the preset torque threshold. When the mechanical risk index is greater than or equal to the preset mechanical standard index, the image is extracted using a preset first extraction method to obtain the first extracted features. When the mechanical risk index is less than the preset mechanical standard index, the image is extracted using a preset second extraction method to obtain the second extracted features. Based on the first extracted features and the fluctuation of the plane offset, a number of first abnormal robotic arms are determined, or based on the second extracted features and the fluctuation of the tilt angle, a number of second abnormal robotic arms are determined. Based on the time distribution characteristics and overlap of the first abnormal robotic arm and the second abnormal robotic arm within a preset period, a number of adjustment robotic arms are determined, and the preset torque threshold is adjusted according to the position of the adjustment robotic arms. The instructions for adjusting the adjustment robot arm are generated based on the geometric risk index and the mechanical risk index regenerated after adjusting the preset torque threshold within a preset adjustment period; The process of extracting features from the image using a preset first extraction method to obtain the first extracted features includes: The image is converted into a grayscale image, and adaptive histogram equalization is performed on the grayscale image to obtain a balanced grayscale image. Edge detection is performed on the uniform grayscale image to extract the contour edges of the contact area of ​​the infusion pump housing, thus obtaining an edge map; A preset filtering algorithm is applied to the same equalized grayscale image to collect texture responses at several spatial frequencies and in several directions, resulting in several texture response images. Morphological opening and closing operations are performed on the edge map and each of the texture response maps to obtain the processed edge feature map and texture feature map, respectively. The processed edge feature map and the texture feature map are concatenated by channel to obtain a high-dimensional feature matrix; Principal component analysis is performed on the high-dimensional feature matrix to reduce its dimensionality, thereby obtaining the first extracted feature; The process of extracting features from the image using a preset second extraction method to obtain the second extracted features includes: The image is downsampled proportionally to a preset resolution, and then Gaussian smoothed to obtain a smooth image. A fast corner detection algorithm is performed on the smoothed image to obtain several key points; For each of the key points, a binary descriptor is calculated to generate several local binary feature vectors; Boolean aggregation is performed on all the local binary feature vectors to obtain a global binary feature string; Hamming distance normalization is applied to the global binary feature string to obtain the second extracted feature.

2. The robotic arm execution control method for industrial production according to claim 1, characterized in that, The process of generating a geometric risk index based on the tilt angle and the plane offset includes: The geometric risk index is generated based on the ratio of each tilt angle to the maximum value among all tilt angles, a preset angle weight, the ratio of each plane offset to the maximum value among all plane offsets, and a preset offset weight.

3. The robotic arm execution control method for industrial production according to claim 2, characterized in that, The process of generating a mechanical risk index based on the geometric risk index, the contact force, the torque, and the preset torque threshold includes: When the geometric risk index is greater than the preset geometric standard index, it is determined that the mechanical risk index needs to be generated. The mechanical risk index is generated based on the ratio of each contact force to the maximum value among all contact forces, the preset contact force weight, the ratio of each torque to the preset torque threshold, and the preset torque weight.

4. The robotic arm execution control method for industrial production according to claim 3, characterized in that, The process of determining several first abnormal robotic arms based on the first extracted features and the fluctuation changes of the planar offset includes: Within a preset period, the intensity of the first edge fluctuation is determined based on the Euclidean distance between the first extracted features corresponding to the two spatially adjacent robotic arms in the production line direction, and the maximum offset is determined based on all the planar offsets within the preset period. Based on the comparison results of the first edge fluctuation intensity and the preset edge fluctuation threshold, and the comparison results of the maximum offset and the preset offset threshold, the operating robot arm is determined to be the first abnormal robot arm, thereby identifying a number of first abnormal robot arms.

5. The robotic arm execution control method for industrial production according to claim 4, characterized in that, The process of determining several second abnormal robotic arms based on the second extracted features and the fluctuation changes of the tilt angle includes: Within a preset period, the intensity of the second edge fluctuation is determined based on the Hamming distance of the second extracted features of the two spatially adjacent robotic arms in each production line direction, and the maximum angle is determined based on all the tilt angles within the preset period. Based on the comparison results of the second edge fluctuation intensity and the preset binary edge fluctuation threshold, and the comparison results of the maximum angle and the preset angle threshold, the operating robot arm is determined to be the second abnormal robot arm, so as to identify a number of second abnormal robot arms.

6. The robotic arm execution control method for industrial production according to claim 5, characterized in that, The process of determining several adjustment robots based on the time distribution characteristics and overlap of the first and second abnormal robots within a preset period includes: The first abnormality distribution degree is determined based on the timestamp of the first abnormal robot arm within the preset period, and the second abnormality distribution degree is determined based on the timestamp of the second abnormal robot arm within the preset period. When both the first abnormality distribution degree and the second abnormality distribution degree are greater than the preset distribution degree threshold, the overlap degree is determined according to the number of times each of the operating robotic arms is identified as the first abnormal robotic arm or the second abnormal robotic arm within a preset determination period. When the overlap is greater than a preset overlap threshold, the working robotic arm is determined to be the adjusting robotic arm, thereby identifying a number of adjusting robotic arms.

7. The robotic arm execution control method for industrial production according to claim 6, characterized in that, The process of adjusting the preset torque threshold based on the position of the robotic arm includes: The adjustment concentration is determined based on the position of the robotic arm. The preset torque threshold is adjusted based on the comparison between the adjusted concentration and the preset concentration standard range.

Images