An ai-based industrial robot joint performance degradation diagnosis method

Abstract

The application is suitable for the technical field of industrial robot joint performance degradation diagnosis, and provides an AI-based industrial robot joint performance degradation diagnosis method, which comprises the following steps: acquiring a current equivalent eccentric load index of a target workpiece; when the current equivalent eccentric load index is greater than an equivalent eccentric load index corresponding to a critical turning point, generating an eccentric load correction factor according to an exceeding amplitude of the current equivalent eccentric load index relative to the equivalent eccentric load index corresponding to the critical turning point, so as to correct a lubrication health evaluation result of a target harmonic reducer joint. The application can solve the problem that the existing lubrication performance degradation diagnosis mainly depends on a conventional lubrication health evaluation result, and it is difficult to reflect the influence of eccentric workpieces on the lubrication bearing state of the harmonic reducer joint in low-speed small-angle reciprocating contact operation. Through the introduction of the equivalent eccentric load index, the abnormal fluctuation point frequency and the critical turning point, the drive response fluctuation which is originally within the conventional allowable range but has eccentric load sensitivity can be used for lubrication state correction.

Description

Technical Field

[0001] This invention belongs to the field of industrial robot joint performance degradation diagnosis technology, and particularly relates to an AI-based method for diagnosing industrial robot joint performance degradation. Background Technology

[0002] Collaborative robots typically employ harmonic reducer joints to achieve attitude adjustment and precise motion control of the end effector's gripping mechanism. Harmonic reducer joints are characterized by their compact structure, large transmission ratio, and high positioning accuracy, making them suitable for low-speed, small-angle reciprocating contact operations such as carving, grinding, polishing, and deburring. During long-term, high-frequency operation, the lubrication status of the harmonic reducer joint affects joint friction, transmission efficiency, and drive response stability. Therefore, existing technologies typically use data such as drive current, drive torque, temperature, operating time, or maintenance records to generate lubrication health assessment results, thereby determining whether the joint's lubrication performance has deteriorated.

[0003] However, when collaborative robots drive the gripping mechanism and workpiece to perform low-speed, small-angle reciprocating contact operations, the joint drive response data will exhibit periodic fluctuations due to the reciprocating motion and contact processing. Furthermore, due to factors such as contact force perturbations, servo compensation, and gripping disturbances, a small number of abnormal fluctuation points may appear in the variation curve. Existing diagnostic methods typically focus more on whether abnormal fluctuations exceed preset alarm thresholds or whether the lubrication health evaluation results themselves are below preset thresholds. However, for abnormal fluctuation point changes that are still within the allowable range, further analysis combined with the specific workpiece eccentricity state and the current working scenario is often lacking.

[0004] Furthermore, for workpieces with identical external machining characteristics but different internal mass distributions, different workpieces may have different centroid offset states relative to the clamping mechanism, causing the harmonic reducer joint to experience different off-center loads under the same machining background conditions. Existing lubrication performance degradation diagnostics often struggle to combine these off-center state differences with the local fluctuation characteristics in the joint drive response data to perform scenario-based corrections to the lubrication health evaluation results, resulting in insufficient matching between the diagnostic results and the actual working conditions of current low-speed, small-angle reciprocating contact operations. Summary of the Invention

[0005] The purpose of this invention is to provide an AI-based method for diagnosing joint performance degradation in industrial robots, aiming to solve the problems mentioned in the background art.

[0006] This invention is implemented as follows: an AI-based method for diagnosing joint performance degradation in industrial robots, the method comprising:

[0007] When it is detected that the target harmonic reducer joint drives the clamping mechanism and the target workpiece to perform low-speed, small-angle reciprocating contact operation, several historical processing samples are extracted from the preset processing database. Among them, the lubrication application conditions of the harmonic reducer joint corresponding to the sample are consistent with those of the target harmonic reducer joint and consistent with the current processing background conditions, but the eccentricity state of the workpiece relative to the clamping mechanism is different.

[0008] Analyze each sample, calculate the equivalent off-center load index caused by the workpiece centroid offset and acting on the corresponding harmonic reducer joint, and determine the frequency of abnormal fluctuation points of the joint drive response data change curve under the equivalent off-center load index corresponding to different samples.

[0009] Determine the relationship between the frequency of abnormal fluctuation points and the equivalent off-center load index, and identify the critical inflection point in the relationship where the frequency of abnormal fluctuation points changes from a stable state to an increasing state.

[0010] Obtain the current equivalent off-center load index of the target workpiece. When it exceeds the equivalent off-center load index corresponding to the critical inflection point, generate an off-center load correction factor based on the excess of the current equivalent off-center load index relative to the equivalent off-center load index corresponding to the critical inflection point, so as to correct the lubrication health evaluation result of the target harmonic reducer joint.

[0011] As a further limitation of the technical solution of the embodiment of the present invention, the low-speed small-angle reciprocating contact operation refers to the operation process in which the harmonic reducer joint reciprocates within a preset angle range at a speed lower than a preset speed threshold, and drives the workpiece held by the clamping mechanism to maintain contact or intermittent contact with the processing tool, and the workpiece has a center of mass offset state relative to the clamping mechanism.

[0012] As a further limitation of the technical solution of this invention, the specific meaning of "consistent with the lubrication application conditions of the target harmonic reducer joint" is:

[0013] The sample harmonic reducer joint and the target harmonic reducer joint are the same in one or more of the following: lubrication medium type, lubrication application amount, lubrication application position, lubrication application method, and running time from lubrication application to the execution of the low-speed small-angle reciprocating contact operation, or are within the corresponding preset allowable deviation range.

[0014] As a further limitation of the technical solution of this embodiment of the invention, the fact that the sample is consistent with the current processing background conditions specifically means:

[0015] The processing procedure corresponding to the sample is the same as or within one or more of the following: workpiece type, processing tool type, clamping method, working trajectory, reciprocating swing angle range, motion speed range, contact force range, robot posture range, and ambient temperature range; or it is within the corresponding preset allowable deviation range.

[0016] As a further limitation of the technical solution of this embodiment of the invention, the calculation process of the equivalent off-center load index includes:

[0017] Obtain the offset of the workpiece's center of gravity relative to the clamping mechanism and the workpiece's mass parameters;

[0018] Based on the centroid offset and the workpiece mass parameters, determine the eccentric load caused by the centroid offset of the workpiece.

[0019] The equivalent off-center load index of the corresponding harmonic reducer joint is determined based on the eccentric load amount.

[0020] As a further limitation of the technical solution of the embodiment of the present invention, the joint drive response data is one or more of drive current data, drive torque data, servo output power data, or joint load rate data;

[0021] The change curve refers to the curve formed by the change of the joint drive response data with the operation time or the number of reciprocating operations within a preset working cycle.

[0022] As a further limitation of the technical solution of the present invention, the frequency of abnormal fluctuation points specifically refers to: within the preset working cycle, identifying multiple periodic fluctuation points or periodic fluctuation segments in the change curve, determining the periodic fluctuation points or periodic fluctuation segments whose fluctuation amplitude exceeds a preset amplitude threshold as abnormal fluctuation points, and determining the frequency based on the number of abnormal fluctuation points and the total number of all periodic fluctuation points or periodic fluctuation segments.

[0023] As a further limitation of the technical solution of this invention, the step of determining the relationship between the frequency of abnormal fluctuation points and the equivalent off-center load index, and identifying the critical inflection point where the frequency of abnormal fluctuation points changes from a stable state to an increasing state in the relationship, specifically includes:

[0024] Several samples are arranged in ascending order of equivalent off-load index to obtain a sample sequence. Based on the equivalent off-load index and the frequency of abnormal fluctuation points corresponding to each sample in the sample sequence, a curve showing the change between the equivalent off-load index and the frequency of abnormal fluctuation points is generated.

[0025] The above-mentioned change curves are segmented and identified to determine the stable region where the frequency of abnormal fluctuation points remains within a preset stable range as the equivalent off-center load index changes, and the rising region where the frequency of abnormal fluctuation points increases as the equivalent off-center load index increases.

[0026] The transition point between the stable region and the rising region is determined as the critical inflection point, and the equivalent off-center load index corresponding to the critical inflection point is taken as the critical equivalent off-center load index.

[0027] As a further limitation of the technical solution of this invention embodiment, when correcting the lubrication health evaluation results of the target harmonic reducer joint, a preset correction function is used, wherein the correction function is:

[0028] ;

[0029] in, This refers to the revised lubrication health assessment results. This refers to the initial lubrication health assessment results. This refers to the lower limit of the value for the lubrication health evaluation result. This refers to the current equivalent off-center load index of the target workpiece. This refers to the equivalent off-center load index corresponding to the critical inflection point. This refers to exceeding the allowable range. This refers to a preset correction strength coefficient, and satisfies... Greater than 0, This refers to the off-load correction factor.

[0030] As a further limitation of the technical solution of the present invention, the lubrication health evaluation result is a lubrication health score or lubrication health index that characterizes the lubrication health of the target harmonic reducer joint;

[0031] After correcting the lubrication health evaluation results of the target harmonic reducer joint, the diagnosis results of lubrication performance degradation of the target harmonic reducer joint are determined based on the corrected lubrication health evaluation results.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] This invention addresses the problem that existing diagnoses of lubrication performance degradation, primarily based on conventional lubrication health assessments, struggle to reflect the impact of eccentric workpieces on the lubrication load-bearing state of harmonic reducer joints during low-speed, small-angle reciprocating contact operations. By introducing an equivalent off-center load index, abnormal fluctuation frequency, and critical inflection point, drive response fluctuations that are within acceptable limits but sensitive to off-center loads can be used for lubrication condition correction. This avoids misjudging critical lubrication conditions as good based solely on conventional health scores, improves the ability to identify early lubrication performance degradation, and makes diagnostic results more consistent with specific operational scenarios. Consequently, it provides a more reliable basis for determining the timing of joint maintenance and providing early warnings of processing risks for collaborative robots. Attached Figure Description

[0034] Figure 1 A flowchart of the method provided in the embodiments of the present invention;

[0035] Figure 2 This is a flowchart illustrating the determination of critical inflection points in the method provided in this embodiment of the invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0037] Figure 1 A flowchart of the method provided by an embodiment of the present invention is shown.

[0038] Specifically, an AI-based method for diagnosing joint performance degradation in industrial robots includes the following steps:

[0039] Step S100: When it is detected that the target harmonic reducer joint drives the clamping mechanism and the target workpiece to perform low-speed, small-angle reciprocating contact operation, a number of historical processing samples are extracted from the preset processing database. Among them, the lubrication application conditions of the harmonic reducer joint corresponding to the sample are consistent with those of the target harmonic reducer joint and consistent with the current processing background conditions, but the eccentricity state of the workpiece relative to the clamping mechanism is different.

[0040] The low-speed, small-angle reciprocating contact operation refers to the process in which the harmonic reducer joint reciprocates within a preset angle range at a speed lower than a preset speed threshold, thereby driving the workpiece held by the clamping mechanism to maintain contact or intermittent contact with the processing tool, and the workpiece has a center of mass offset relative to the clamping mechanism.

[0041] The specific meaning of "consistent with the lubrication application conditions of the target harmonic reducer joint" is that the sample-corresponding harmonic reducer joint and the target harmonic reducer joint are the same in one or more of the following: lubrication medium type, lubrication application amount, lubrication application position, lubrication application method, and running time from the lubrication application to the execution of the low-speed small-angle reciprocating contact operation, or are within the corresponding preset allowable deviation range.

[0042] The fact that the sample is consistent with the current processing background conditions specifically means:

[0043] The processing procedure corresponding to the sample is the same as or within one or more of the following: workpiece type, processing tool type, clamping method, working trajectory, reciprocating swing angle range, motion speed range, contact force range, robot posture range, and ambient temperature range; or it is within the corresponding preset allowable deviation range.

[0044] In this embodiment of the invention, the target harmonic reducer joint can be a joint on the wrist or end effector of a collaborative robot used to drive the gripping mechanism for attitude adjustment. This type of joint typically uses harmonic reducers to achieve small size, high transmission ratio, and high positioning accuracy, making it suitable for low-speed precision operations such as grinding, polishing, engraving, deburring, and coating. Since the internal structure of the harmonic reducer joint is usually grease-lubricated, and it endures low-speed reciprocating, contact loads, and localized off-center loads for extended periods during high-frequency repetitive operations, its lubrication condition affects joint friction, transmission efficiency, drive response stability, and subsequent joint performance degradation. Therefore, analyzing the lubrication performance and diagnosing joint performance degradation in the target harmonic reducer joint is of practical significance.

[0045] In existing applications, for harmonic reducer joints operating at high frequencies, a lubrication performance analysis is typically performed after a specific service cycle or before the start of a particular joint operation. Current diagnostic methods generally involve first obtaining a lubrication health assessment result, and then determining whether the harmonic reducer joint exhibits lubrication performance degradation based on that result. The lubrication health assessment result can be a lubrication health score or lubrication health index characterizing the lubrication health of the harmonic reducer joint. Such scores or indices can be determined based on data such as temperature, drive current, drive torque, operating time, load rate, and historical maintenance records, representing a relatively mature health assessment foundation in this field.

[0046] This invention further focuses on the fact that when the target harmonic reducer joint drives the clamping mechanism and the target workpiece to perform low-speed, small-angle reciprocating contact operations, the joint drive response data typically exhibits periodic fluctuations that change with the reciprocating operation cycle. The joint drive response data may include drive current data, drive torque data, servo output power data, or joint load rate data. Since the low-speed, small-angle reciprocating contact operation itself has reciprocating oscillation and periodic contact characteristics, it is normal for the joint drive response data to exhibit regular rises and falls within a preset working cycle, and this is a common data manifestation in industrial robot contact processing scenarios.

[0047] Meanwhile, due to factors such as minute changes in machining contact force, local contact differences between the machining tool and the workpiece, servo control compensation actions, minor clamping disturbances, and low-speed friction fluctuations of the harmonic reducer joint itself, the change curve of the joint drive response data may also show a small number of periodic fluctuation points or periodic fluctuation segments that deviate from the normal fluctuation amplitude. For general working conditions, as long as the frequency, amplitude, or duration of such abnormal fluctuation points does not exceed the preset standard, they are usually not directly judged as abnormal joint lubrication, but are considered as normal fluctuations in low-speed, small-angle reciprocating contact operations.

[0048] However, after conducting AI analysis on a large number of historical processing samples, those skilled in the art discovered that for eccentric workpieces with identical external processing characteristics but different internal quality distributions, even if their corresponding lubrication application conditions, processing background conditions, and external surface processing requirements are the same, the eccentricity of different workpieces relative to the clamping mechanism will still lead to differences in the frequency of abnormal fluctuation points in the joint drive response data. In other words, even when the frequency of abnormal fluctuation points does not exceed the conventional alarm standard, the frequency of abnormal fluctuation points under different eccentric states will still exhibit distinguishable high and low variations.

[0049] Furthermore, those skilled in the art have discovered that the harmonic reducer joint is not linearly sensitive to the workpiece eccentricity, but rather has certain tolerance limits. When the workpiece eccentricity relative to the clamping mechanism is low, the target harmonic reducer joint can still maintain a relatively stable drive response through its own lubrication state, transmission structure, and servo control compensation, and the frequency of abnormal fluctuation points remains basically stable. When the eccentricity exceeds a certain range, the equivalent off-center load caused by the workpiece's center of mass shift will exacerbate the uneven local contact load inside the harmonic reducer joint, causing the stability of the local lubrication film to decrease, and low-speed friction instability or micro-stick-slip phenomenon to be more easily excited, thereby causing the frequency of abnormal fluctuation points in the joint drive response data to increase with the severity of eccentricity.

[0050] Therefore, it can be understood that although the frequency of the aforementioned abnormal fluctuation points may not yet reach the alarm or fault determination conditions under conventional standards, its changes can still reflect the lubrication load margin of the target harmonic reducer joint under the current off-center load condition. In other words, when the frequency of the abnormal fluctuation points changes from a stable state to an increasing state after the equivalent off-center load index increases, this change can be used as a basis for correcting the lubrication health evaluation results, thereby improving the accuracy of diagnosing the lubrication performance degradation of the target harmonic reducer joint.

[0051] The workpieces applicable to the embodiments of the present invention are preferably eccentric workpieces with consistent external processing features but different internal mass distributions. Specifically, each workpiece maintains or is substantially consistent in its external shape, external dimensions, surface to be processed, processing area, processing trajectory, and processing requirements, thus having the same contact conditions with the processing tool; however, the positions of internal components, the distribution of fillers, local density, or internal assembly states of each workpiece differ, resulting in varying degrees of offset of the overall center of mass of the workpiece relative to the clamping mechanism. For example, hollow shell workpieces whose external surfaces require the same engraving, grinding, or polishing processes may have different positions of internal counterweights, components, or fillers, thus creating different equivalent off-center loading states under the same external processing conditions.

[0052] The low-speed, small-angle reciprocating contact operation described can be understood as a common application scenario in collaborative robot precision machining. For example, in grinding, polishing, engraving, deburring, trimming, wiping, or coating operations, the harmonic reducer joint typically does not rotate at large angles and high speeds, but rather oscillates reciprocally within a preset angle range at a speed lower than a preset speed threshold, causing the workpiece held by the clamping mechanism to maintain contact or intermittent contact with the machining tool. Therefore, this operation scenario itself has a mature industrial application basis and is not a special scenario constructed solely for this invention.

[0053] The preset processing database can be derived from the collaborative robot's historical production records, servo drive records, robot controller records, process management system records, maintenance records, and quality inspection records. This database may include sample numbers, harmonic reducer joint identifiers, lubrication medium type, lubrication application amount, lubrication application location, lubrication application method, runtime after lubrication application, workpiece type, clamping method, work trajectory, reciprocating swing angle range, motion speed range, contact force range, robot posture range, ambient temperature range, joint drive response data, and workpiece eccentricity data relative to the clamping mechanism. This type of data can typically be obtained through robot control systems, servo drive systems, clamping mechanism calibration systems, weighing or center of mass detection equipment, process databases, and maintenance record systems, and is a type of data that can be accumulated and accessed in the industrial robot production site.

[0054] In step S100, the purpose of extracting several historical processing samples from the preset processing database is to obtain a sample set that differs only in the eccentricity of the workpiece relative to the clamping mechanism, under the premise that the lubrication application conditions and processing background conditions are consistent. This sample set can minimize the influence of factors such as lubrication medium, lubrication application amount, processing tools, contact force, work trajectory, and ambient temperature on the frequency of abnormal fluctuation points, making the subsequent relationship between the frequency of abnormal fluctuation points and the equivalent off-center load index more reflective of the influence of the eccentricity state on the lubrication load state of the target harmonic reducer joint.

[0055] Strict sample selection criteria are employed to ensure comparability between historical processing samples and the current processing. If samples differ significantly in tool type, clamping method, work trajectory, speed range, contact force range, or robot posture range, the frequency of abnormal fluctuations may not be caused by workpiece eccentricity but by changes in processing background conditions. Using such samples to determine critical inflection points in this case would lead to inaccurate generation of the off-center load correction factor. Therefore, this embodiment of the invention improves the reliability of subsequent critical inflection point identification results by using a selection method that ensures consistent lubrication application conditions and processing background conditions.

[0056] In practical implementation, in addition to the screening criteria listed above, further screening can be conducted by considering factors such as the collaborative robot model, the axis number of the target harmonic reducer joint, the specifications of the harmonic reducer, control parameters, sampling frequency, wear condition of the machining tool, clamping force of the clamping mechanism, workpiece batch, initial joint temperature, ambient humidity, and no-load running status before machining. The above criteria can be selected based on the actual application accuracy requirements and are not required to be used in all implementation methods.

[0057] It should be noted that the consistency of lubrication application conditions and processing background conditions does not require all parameters to be absolutely identical, but rather that they are allowed to be within the corresponding preset allowable deviation range. This is because, even when performing the same type of operation according to the same process in an industrial setting, clamping errors, sensor acquisition errors, slight changes in ambient temperature, and slight fluctuations in contact force are unavoidable. As long as these differences are insufficient to substantially affect the relationship between the frequency of abnormal fluctuation points and the equivalent off-center load index, the sample can be considered to meet the screening requirements.

[0058] The preset allowable deviation range or preset similarity threshold can be determined based on process specifications, equipment calibration results, historical processing data statistics, or experimental verification results. For example, based on historical stable processing data under the same workpiece type, the same processing tool type, and the same work trajectory, the normal fluctuation range of each processing background parameter can be statistically analyzed, and this normal fluctuation range can be used as the corresponding preset allowable deviation range. Alternatively, the workpiece type, clamping method, work trajectory, motion speed range, contact force range, robot posture range, and ambient temperature range between the sample and the current processing process can be converted into similarity indices, and then the weighted result of each similarity index can be used to determine whether the sample meets the preset similarity threshold. In this way, it is possible to ensure that the sample has sufficient consistency with the current processing process, while avoiding an insufficient number of available historical processing samples due to the requirement for complete consistency.

[0059] Since historical processing samples are typically numerous and involve multi-dimensional parameters such as lubrication application conditions, processing background conditions, joint drive response data, and workpiece eccentricity, relying on manual comparison and screening would result in low processing efficiency and a high risk of screening bias. Therefore, in this embodiment of the invention, a trained AI model can be used to perform similarity identification, condition matching, and sample screening on historical processing samples in a pre-set processing database. This improves the efficiency and consistency of historical processing sample extraction and provides a reliable data foundation for subsequently determining the relationship between the frequency of abnormal fluctuation points and the equivalent off-center load index.

[0060] Furthermore, the AI-based industrial robot joint performance degradation diagnosis method also includes the following steps:

[0061] Step S200: Analyze each sample, calculate the equivalent off-center load index caused by the workpiece centroid offset and acting on the corresponding harmonic reducer joint, and determine the frequency of abnormal fluctuation points of the joint drive response data change curve under the equivalent off-center load index corresponding to different samples.

[0062] The calculation process of the equivalent off-center load index includes: obtaining the center-of-gravity offset of the workpiece relative to the clamping mechanism and the workpiece mass parameters; determining the eccentric load caused by the center-of-gravity offset of the workpiece based on the center-of-gravity offset and the workpiece mass parameters; and determining the equivalent off-center load index of the corresponding harmonic reducer joint based on the eccentric load.

[0063] The joint drive response data is one or more of the following: drive current data, drive torque data, servo output power data, or joint load rate data; the change curve refers to the curve formed by the change of the joint drive response data with the operation time or the number of reciprocating operations within a preset working cycle.

[0064] The frequency of abnormal fluctuation points specifically refers to: within the preset working cycle, identifying multiple periodic fluctuation points or periodic fluctuation segments in the change curve, determining periodic fluctuation points or periodic fluctuation segments whose fluctuation amplitude exceeds a preset amplitude threshold as abnormal fluctuation points, and determining the frequency based on the number of abnormal fluctuation points and the total number of all periodic fluctuation points or periodic fluctuation segments.

[0065] In this embodiment of the invention, step S200 is used to extract key parameters required for subsequent diagnosis from the historical processing samples screened in step S100, mainly including the equivalent off-center load index and the frequency of abnormal fluctuation points. The equivalent off-center load index can be understood as a quantitative parameter characterizing the degree of influence of the workpiece's eccentricity on the off-center load formation of the harmonic reducer joint. It can be obtained using existing load equivalence, torque conversion, or normalization calculation methods in the art. The focus of this invention is not on proposing the concept of "off-center load quantification" for the first time, but on establishing a correspondence between the equivalent off-center load index and the frequency of abnormal fluctuation points in the joint drive response data, and using this correspondence to identify critical inflection points, thereby correcting the lubrication health evaluation results of the target harmonic reducer joint.

[0066] Specifically, for each historical machining sample, the center-of-gravity offset of the workpiece relative to the clamping mechanism and the workpiece mass parameters can be obtained first. The center-of-gravity offset can represent the distance the overall center of gravity of the workpiece is offset from the clamping center, clamping reference point, or origin of the clamping coordinate system, or it can represent the component of the center-of-gravity offset in a preset direction. The workpiece mass parameters can be the total mass of the workpiece or the calibrated effective mass. The aforementioned center-of-gravity offset and workpiece mass parameters can be obtained from the workpiece design model, workpiece weighing data, center-of-gravity detection equipment, clamping mechanism calibration data, visual measurement results, or records in the historical machining database.

[0067] After obtaining the centroid offset and workpiece mass parameters, the eccentric load caused by the centroid offset can be determined based on these two parameters. For example, in one implementation, the product of the workpiece mass parameters and the centroid offset can be used as the base value of the eccentric load; in another implementation, the eccentric load in a rectangular form can be obtained by combining gravitational acceleration; in yet another implementation, the above-mentioned eccentric load can be equivalently converted based on the installation direction of the clamping mechanism, the force direction of the harmonic reducer joint, or the low-speed reciprocating characteristics during operation. Since step S100 has already screened the processing background conditions, ensuring that the working trajectory, reciprocating swing angle range, motion speed range, contact force range, and robot posture range among the samples are the same or within the preset allowable deviation range, in step S200, the difference in eccentric load among different samples can be mainly attributed to the different centroid offset states of the workpiece relative to the clamping mechanism.

[0068] Subsequently, the equivalent off-center load index of the corresponding harmonic reducer joint can be determined based on the eccentric load. The equivalent off-center load index can be a direct numerical value of the eccentric load, or an index obtained after normalizing, hierarchically mapping, or dimensionlessly processing the eccentric load. For example, a reference eccentric load can be used as a reference value, and the ratio of the current sample's eccentric load to this reference value can be used as the equivalent off-center load index; alternatively, the eccentric load can be mapped to a range of 0 to 1 or 0 to 100 to facilitate sorting and comparison between different historical processing samples. Through this equivalent off-center load index, the influence of the workpiece's eccentricity on the off-center load formed by the harmonic reducer joint can be transformed into a comparable and sortable quantitative parameter.

[0069] For example, for hollow shell workpieces with consistent external shape and processing requirements, the positions of internal counterweights or internal components differ in different samples, resulting in varying center-of-gravity offsets relative to the clamping mechanism. If sample A has a workpiece mass of 1.5 kg and a center-of-gravity offset of 5 mm, while sample B also has a workpiece mass of 1.5 kg but a center-of-gravity offset of 15 mm, then sample B corresponds to a greater eccentric load than sample A. Under otherwise identical processing conditions, sample B also corresponds to a larger equivalent off-center load index for the harmonic reducer joint. Therefore, the eccentricity states of different samples can be transformed into lateral comparison parameters for establishing subsequent relationships.

[0070] Joint drive response data reflects the drive response state of the harmonic reducer joint during low-speed, small-angle reciprocating contact operations. This data can be one or more of the following: drive current data, drive torque data, servo output power data, or joint load rate data. Because the low-speed, small-angle reciprocating contact operation has repetitive oscillation and contact processing characteristics, the joint drive response data typically forms a periodic variation curve within a preset working cycle. This variation curve can be formed by the joint drive response data changing with the operation time, or it can be formed by the joint drive response data changing with the number of reciprocating operations.

[0071] The preset working cycle can be determined according to the specific processing technology. In one embodiment, the preset working cycle can be the time interval corresponding to the target workpiece completing one complete processing action; in another embodiment, the preset working cycle can be the time interval corresponding to the harmonic reducer joint completing a preset number of reciprocating swings; in yet another embodiment, the preset working cycle can also be the time interval from the start of contact between the workpiece and the processing tool to the completion of a stable processing stage. The purpose of setting the preset working cycle is to enable the joint drive response data of different historical processing samples to be analyzed at the same or comparable time scale and period scale, avoiding the incomparability of statistical results of abnormal fluctuation point frequencies due to different sample collection lengths.

[0072] After determining the change curve of the joint drive response data, multiple periodic fluctuation points or periodic fluctuation segments can be identified within the preset working cycle. Specifically, the change curve can be divided into several periodic fluctuation segments based on the number of reciprocating operations, joint movement commands, joint position changes, peak and trough positions, or time window division results; alternatively, peaks, troughs, local abrupt change points, or periodic fluctuation points formed by adjacent peaks and troughs in the change curve can be directly identified. For each periodic fluctuation point or periodic fluctuation segment, its fluctuation amplitude can be calculated. The fluctuation amplitude can be the difference between adjacent peaks and troughs, the difference between the maximum and minimum values ​​within a certain periodic fluctuation segment, or the deviation of the periodic fluctuation segment from the average fluctuation amplitude within the same preset working cycle.

[0073] After determining the amplitude of each periodic fluctuation point or periodic fluctuation segment, periodic fluctuation points or periodic fluctuation segments with amplitudes exceeding a preset amplitude threshold are identified as abnormal fluctuation points. The preset amplitude threshold can be determined based on the average fluctuation amplitude, median fluctuation amplitude, standard deviation, historical stable sample statistical results, or the process-allowed fluctuation range of each periodic fluctuation segment within the same preset working cycle. For example, the preset amplitude threshold can be the average fluctuation amplitude of multiple periodic fluctuation segments within the same preset working cycle plus a preset multiple of the standard deviation; alternatively, it can be determined based on the upper limit of normal fluctuation amplitudes in historical stable samples. The purpose of this approach is to avoid misjudging the normally occurring periodic rise and fall fluctuations in low-speed, small-angle reciprocating contact operations as abnormal, and to only count fluctuation points or fluctuation segments that significantly exceed the normal fluctuation amplitude.

[0074] After identifying abnormal fluctuation points, the frequency of these points can be determined by comparing their number with the total number of all periodic fluctuation points or periodic fluctuation segments. For example, if 100 periodic fluctuation segments are identified within a preset work cycle, and the amplitude of 8 of these segments exceeds a preset amplitude threshold, then 8 / 100 can be used as the frequency of abnormal fluctuation points for that sample. Alternatively, the frequency of abnormal fluctuation points can be obtained by dividing the number of abnormal fluctuation points by the duration of the preset work cycle. The former method is more suitable for cases where different samples have the same number of reciprocating operations, while the latter method is more suitable for cases where different samples have the same work duration.

[0075] In practical implementation, the determination of abnormal fluctuation frequency can be achieved using techniques such as time series analysis, peak-valley identification, sliding window analysis, period segmentation, amplitude statistics, or identification by trained AI models. For example, the joint drive response data can be filtered to remove acquisition noise and abnormal sampling points, and then the change curve can be segmented according to the reciprocating operation cycle; then the fluctuation amplitude of each periodic fluctuation segment can be calculated, and abnormal fluctuation points can be determined according to a preset amplitude threshold. Alternatively, the change curve can be input into a trained AI model, which can then identify the periodic fluctuation segments, abnormal fluctuation points, and their corresponding abnormal fluctuation frequencies. When using an AI model, historical labeled data can be used to train the model, enabling it to identify fluctuation points or fluctuation segments with abnormal amplitude characteristics under different joint drive response data types.

[0076] Step S200 allows each historical machining sample to be assigned a set of parameters: the equivalent off-center load index and the abnormal fluctuation point frequency. The equivalent off-center load index characterizes the degree of influence of the workpiece's eccentricity on the off-center load formed by the corresponding harmonic reducer joint, while the abnormal fluctuation point frequency characterizes the periodic fluctuation stability of the joint drive response data under the influence of this off-center load. Since step S100 has ensured the consistency of lubrication application conditions and machining background conditions among the historical machining samples as much as possible, the correspondence between the above two types of parameters can provide a data basis for subsequently identifying the critical inflection point where the abnormal fluctuation point frequency changes from a stable state to an increasing state.

[0077] Furthermore, the AI-based industrial robot joint performance degradation diagnosis method also includes the following steps:

[0078] Step S300: Determine the relationship between the frequency of abnormal fluctuation points and the equivalent off-center load index, and identify the critical inflection point in the relationship where the frequency of abnormal fluctuation points changes from a stable state to an increasing state.

[0079] Specifically, Figure 2 A flowchart for determining critical inflection points is shown.

[0080] The specific steps involved in determining the relationship between the frequency of abnormal fluctuations and the equivalent off-center load index, and identifying the critical inflection point where the frequency of abnormal fluctuations transitions from a stable state to an increasing state, are as follows:

[0081] Step S301: Arrange several samples in ascending order according to the equivalent off-load index to obtain a sample sequence, and generate a curve showing the change between the equivalent off-load index and the frequency of abnormal fluctuation points based on the equivalent off-load index and the frequency of abnormal fluctuation points corresponding to each sample in the sample sequence.

[0082] Step S302: Segment the above-mentioned change curve to identify the stable region where the frequency of abnormal fluctuation points remains within a preset stable range as the equivalent off-center load index changes, and the rising region where the frequency of abnormal fluctuation points increases as the equivalent off-center load index increases.

[0083] Step S303: The transition position between the stable region and the rising region is determined as the critical inflection point, and the equivalent off-center load index corresponding to the critical inflection point is taken as the critical equivalent off-center load index.

[0084] In this embodiment of the invention, step S300 is used to verify, based on the equivalent off-center load index and abnormal fluctuation point frequency obtained in step S200, whether the current processing process and the corresponding workpiece type exhibit a change pattern where "after the equivalent off-center load index increases to a certain extent, the abnormal fluctuation point frequency changes from a stable state to an increasing state." If this change pattern can be identified, it indicates that the workpiece eccentricity state has a quantifiable impact on the drive response stability of the target harmonic reducer joint, and this pattern can be used to correct the lubrication health evaluation results subsequently.

[0085] In step S301, several samples are arranged in ascending order of equivalent off-center loading index to obtain a sample sequence. This sorting method allows the sample sequence to reflect the change process of workpiece eccentricity from light to heavy. Subsequently, the equivalent off-center loading index and abnormal fluctuation point frequency corresponding to each sample in the sample sequence are extracted, and a change curve between the equivalent off-center loading index and the abnormal fluctuation point frequency is generated. This change curve can be generated using methods such as scatter plot fitting, curve smoothing, regression analysis, or fitting with a trained AI model.

[0086] In step S302, the change curve is segmented to distinguish between a stable region and an increasing region. The stable region refers to the interval where the frequency of abnormal fluctuation points remains within a preset stable range as the equivalent off-center load index changes, indicating that the workpiece eccentricity has not significantly disrupted the periodic stability of the joint drive response data within this interval. The increasing region refers to the interval where the frequency of abnormal fluctuation points shows an upward trend as the equivalent off-center load index increases, indicating that the eccentricity has begun to cause more frequent abnormal fluctuations within this interval. Segmentation identification can be achieved using slope change detection, piecewise linear fitting, inflection point detection, trend classification models, or trained AI models.

[0087] In step S303, the transition point between the stable region and the rising region is determined as the critical inflection point, and the equivalent off-center load index corresponding to this critical inflection point is taken as the critical equivalent off-center load index. The critical inflection point can be understood as the stable bearing boundary of the harmonic reducer joint to the workpiece eccentricity under the current lubrication application conditions and current processing background conditions. When the equivalent off-center load index does not exceed the critical equivalent off-center load index, the frequency of abnormal fluctuation points remains basically stable; when the equivalent off-center load index exceeds the critical equivalent off-center load index, the frequency of abnormal fluctuation points begins to increase with the increase of the equivalent off-center load index.

[0088] The significance of determining the critical inflection point lies in providing a benchmark for the subsequent generation of the off-center load correction factor. Since conventional lubrication health assessment results may only reflect the general lubrication state of the target harmonic reducer joint, without fully considering the impact of the current workpiece eccentricity on low-speed, small-angle reciprocating contact operations, the critical inflection point can be used to determine whether the current equivalent off-center load index has entered the abnormal fluctuation frequency rise zone. If the current equivalent off-center load index is greater than the critical equivalent off-center load index, it indicates that the eccentric load under the current operating condition has exceeded the stable bearing range of the corresponding harmonic reducer joint under this processing background, and the lubrication health assessment results need to be corrected in subsequent steps based on the excess magnitude.

[0089] Furthermore, the AI-based industrial robot joint performance degradation diagnosis method also includes the following steps:

[0090] Step S400: Obtain the current equivalent off-center load index of the target workpiece. When the current equivalent off-center load index is greater than the equivalent off-center load index corresponding to the critical inflection point, generate an off-center load correction factor based on the excess of the current equivalent off-center load index relative to the equivalent off-center load index corresponding to the critical inflection point to correct the lubrication health evaluation result of the target harmonic reducer joint. When the current equivalent off-center load index is not greater than the equivalent off-center load index corresponding to the critical inflection point, use the lubrication health evaluation result of the target harmonic reducer joint as the basis for lubrication performance degradation diagnosis.

[0091] When correcting the lubrication health evaluation results of the target harmonic reducer joint, a preset correction function is used, which is:

[0092] ;

[0093] in, This refers to the revised lubrication health assessment results. This refers to the initial lubrication health assessment results. This refers to the lower limit of the value for the lubrication health evaluation result. This refers to the current equivalent off-center load index of the target workpiece. This refers to the equivalent off-center load index corresponding to the critical inflection point. This refers to exceeding the allowable range. This refers to a preset correction strength coefficient, and satisfies... Greater than 0, This refers to the off-load correction factor.

[0094] In this embodiment of the invention, step S400 is used to apply the critical inflection point identified in step S300 to the current operation process, thereby performing a scenario-based correction to the lubrication health evaluation result of the target harmonic reducer joint. It should be noted that this correction does not represent actual intervention actions such as adding oil, reducing speed, stopping the machine, or adjusting control parameters on the target harmonic reducer joint. Rather, it is a diagnostic-level auxiliary correction to the existing lubrication performance evaluation result, enabling it to more accurately reflect the lubrication load state of the target harmonic reducer joint under the current low-speed, small-angle reciprocating contact operation.

[0095] Specifically, existing evaluation systems can output initial lubrication health assessment results for the target harmonic reducer joint based on data such as temperature, runtime, drive current, drive torque, or maintenance records. However, this initial lubrication health assessment result typically reflects the lubrication health of the target harmonic reducer joint under normal operating conditions and may not fully consider the impact on the stability of the joint's drive response when the center of gravity of the target workpiece is offset relative to the clamping mechanism. Therefore, when the current equivalent off-center load index is greater than the equivalent off-center load index corresponding to the critical inflection point, it indicates that the current operation has entered the range where the frequency of abnormal fluctuation points increases with the increase of the equivalent off-center load index. At this time, it is necessary to deduct and correct the initial lubrication health assessment result according to the excess magnitude.

[0096] The rationale for using the excess magnitude as the off-center load correction factor is as follows: when the current equivalent off-center load index is only slightly higher than the critical inflection point, it indicates that the target harmonic reducer joint has just entered the off-center load sensitive zone, and its impact on the lubrication health evaluation results is relatively small. However, when the current equivalent off-center load index is significantly higher than the critical inflection point, it indicates that the off-center load degree of the target harmonic reducer joint in the current operation has significantly exceeded the stable bearing boundary, and the frequency of abnormal fluctuation points is more easily amplified, and its impact on the lubrication health evaluation results should be greater. Therefore, generating the off-center load correction factor based on the excess magnitude of the current equivalent off-center load index relative to the equivalent off-center load index corresponding to the critical inflection point can ensure that the correction result corresponds to the actual severity of the off-center load.

[0097] The correction function in this embodiment of the invention can be understood as an intuitive and effective deduction-type correction method. Its basic idea is: when the current equivalent off-center load index does not exceed the equivalent off-center load index corresponding to the critical inflection point, the excess is 0, and no deduction is made from the initial lubrication health evaluation result; when the current equivalent off-center load index exceeds the equivalent off-center load index corresponding to the critical inflection point, the deduction amount is determined based on the excess ratio and a preset correction strength coefficient, and after deduction, it is protected by the lower limit of the lubrication health evaluation result to avoid the correction result being too low and losing its evaluation significance. In other words, this correction function simultaneously possesses three characteristics: "no correction if not exceeding," "the more it exceeds, the more it deducts," and "the deduction result is not lower than the lower limit," making it suitable for correcting the lubrication health evaluation result in this embodiment of the invention.

[0098] The preset correction strength coefficient can be determined based on statistical results of historical machining samples, test calibration results, maintenance experience, or expert experience. For example, harmonic reducer joints with known good lubrication, slight degradation, and moderate degradation can be selected as calibration objects. Under the same low-speed, small-angle reciprocating contact operation, their current equivalent off-center load index, abnormal fluctuation frequency, and post-maintenance verification results can be statistically analyzed to determine the correction strength coefficient that makes the corrected lubrication health evaluation result more consistent with the actual lubrication condition. This preset correction strength coefficient can also be set separately according to different workpiece types, different machining tool types, or different target harmonic reducer joint models.

[0099] Besides the aforementioned deduction-type correction function, other correction methods can also be used. For example, a segmented deduction method can be used, setting different deduction ratios according to different excess ranges after the current equivalent off-center load index exceeds the equivalent off-center load index corresponding to the critical inflection point; a non-linear correction method can also be used, making the deduction slower when the excess is small and faster when the excess is large; a lookup table correction method can also be used, pre-storing the excess range and the correction amount, and calling the corresponding correction amount according to the current excess. All of the above methods can achieve the purpose of correcting the lubrication health evaluation results based on the degree of off-center load.

[0100] For example, a collaborative robot uses a target harmonic reducer joint to drive a clamping mechanism and a target workpiece to perform low-speed, small-angle reciprocating contact operations. Twenty historical machining samples were extracted from a pre-set machining database. These samples had the same lubrication application conditions and machining background conditions, but differed in the workpiece's eccentricity relative to the clamping mechanism. After processing in step S200, each historical machining sample yielded an equivalent off-center load index and an abnormal fluctuation point frequency.

[0101] For example, when the equivalent off-center load index is 20, 30, 40, and 50, the frequencies of abnormal fluctuation points are 0.06, 0.05, 0.06, and 0.07, respectively, indicating an overall stable state. When the equivalent off-center load index is 60, 70, and 80, the frequencies of abnormal fluctuation points rise to 0.12, 0.18, and 0.25, respectively, indicating that the frequency of abnormal fluctuation points begins to increase significantly with the increase of the equivalent off-center load index. After segmentation identification in step S300, the transition position between the equivalent off-center load index of 50 and 60 is determined as the critical inflection point, and the equivalent off-center load index corresponding to the critical inflection point is determined to be 55.

[0102] In the current operation, the current equivalent off-center load index of the target workpiece is 70, the initial lubrication health evaluation result of the target harmonic reducer joint is 82 points, the lower limit of the lubrication health evaluation result is set to 40 points, and the preset correction strength coefficient is set to 0.5. Since the current equivalent off-center load index of 70 is greater than the equivalent off-center load index of 55 corresponding to the critical inflection point, correction is required. The excess ratio of the current equivalent off-center load index relative to the equivalent off-center load index corresponding to the critical inflection point is 70 minus 55 divided by 55, which is approximately 0.273. Multiplying this excess ratio by the preset correction strength coefficient 0.5 yields a deduction ratio of approximately 0.1365. The initial lubrication health evaluation result of 82 points is deducted according to this deduction ratio, resulting in a corrected lubrication health evaluation result of approximately 70.8 points. If the existing evaluation system corresponds to excellent lubrication with a score above 80 points and average lubrication with a score between 70 and 80 points, the target harmonic reducer joint might originally be judged as having excellent lubrication, but after the off-center load correction of this embodiment, it is judged as having average lubrication. This result suggests that under current low-speed, small-angle reciprocating contact operation, although the target harmonic reducer joint has not reached a significant fault state, its lubrication health is no longer suitable for evaluation as it is in good condition under normal operating conditions.

[0103] As can be seen from the above examples, the embodiments of the present invention do not simply judge lubrication performance degradation based on whether the joint drive response data exceeds the limit. Instead, in low-speed, small-angle reciprocating contact operations, the initial lubrication health evaluation results are corrected by combining the eccentricity of the workpiece relative to the clamping mechanism, the equivalent off-center load index, the frequency of abnormal fluctuation points, and the critical inflection point. This can identify some cases that do not meet the conventional alarm standards but have already shown a decrease in lubrication load margin under specific off-center load conditions, thereby improving the accuracy and sensitivity of the diagnosis of lubrication performance degradation of the target harmonic reducer joint.

[0104] This invention addresses the core research point raised in step S100: in the machining scenario of eccentric workpieces with consistent external machining characteristics but different internal mass distributions, even if the lubrication application conditions and machining background conditions are consistent, the eccentricity of the workpiece relative to the clamping mechanism may still lead to regular differences in the frequency of abnormal fluctuation points. When this difference shows an upward trend after the critical inflection point, it can reflect the change in the lubrication load margin of the target harmonic reducer joint under the current operation. By identifying this pattern and using it to correct the lubrication health evaluation results, this invention makes existing lubrication performance evaluations more adaptable to the specific application scenario of low-speed, small-angle reciprocating contact operations.

[0105] The beneficial effects of this invention are as follows: On the one hand, it can introduce an auxiliary correction mechanism under off-center loading scenarios without changing the original lubrication evaluation system of the collaborative robot, thereby improving the matching degree between the lubrication performance degradation diagnosis results and the current working state; on the other hand, it can make full use of historical processing samples and joint drive response data, and can further judge the lubrication health of the target harmonic reducer joint without adding additional complex sensors; and on the other hand, it can transform the frequency changes of abnormal fluctuation points that were originally considered to be within the normal range into features with diagnostic value, thereby improving the ability to identify early lubrication performance degradation.

[0106] This method is applicable to scenarios where collaborative robots perform low-speed, small-angle reciprocating contact operations such as carving, grinding, polishing, deburring, coating, and finishing. It is particularly suitable for batch processing of eccentric workpieces with consistent external processing features but different internal quality distributions. With the increasing application of collaborative robots in flexible manufacturing, small-batch customized processing, and precision contact machining, the lubrication status of the target harmonic reducer joints will have a more significant impact on operational stability and maintenance decisions. Therefore, the embodiments of this invention have good engineering application prospects and can be used for joint status monitoring, lubrication maintenance timing determination, pre-processing risk assessment, and predictive maintenance systems for collaborative robots.

[0107] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0108] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0109] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0110] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0111] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AI-based method for diagnosing joint performance degradation in industrial robots, characterized in that, The method includes: When it is detected that the target harmonic reducer joint drives the clamping mechanism and the target workpiece to perform low-speed, small-angle reciprocating contact operation, several historical processing samples are extracted from the preset processing database. Among them, the lubrication application conditions of the harmonic reducer joint corresponding to the sample are consistent with those of the target harmonic reducer joint and consistent with the current processing background conditions, but the eccentricity state of the workpiece relative to the clamping mechanism is different. Analyze each sample, calculate the equivalent off-center load index caused by the workpiece centroid offset and acting on the corresponding harmonic reducer joint, and determine the frequency of abnormal fluctuation points of the joint drive response data change curve under the equivalent off-center load index corresponding to different samples. Determine the relationship between the frequency of abnormal fluctuation points and the equivalent off-center load index, and identify the critical inflection point in the relationship where the frequency of abnormal fluctuation points changes from a stable state to an increasing state. Obtain the current equivalent off-center load index of the target workpiece. When it exceeds the equivalent off-center load index corresponding to the critical inflection point, generate an off-center load correction factor based on the excess of the current equivalent off-center load index relative to the equivalent off-center load index corresponding to the critical inflection point, so as to correct the lubrication health evaluation result of the target harmonic reducer joint.

2. The AI-based method for diagnosing joint performance degradation in industrial robots according to claim 1, characterized in that, The low-speed, small-angle reciprocating contact operation refers to the process in which the harmonic reducer joint reciprocates within a preset angle range at a speed lower than a preset speed threshold, thereby driving the workpiece held by the clamping mechanism to maintain contact or intermittent contact with the processing tool, and the workpiece has a center of mass offset relative to the clamping mechanism.

3. The AI-based method for diagnosing joint performance degradation in industrial robots according to claim 1, characterized in that, The specific meaning of "consistent with the lubrication application conditions of the target harmonic reducer joint" is: The sample harmonic reducer joint and the target harmonic reducer joint are the same in one or more of the following: lubrication medium type, lubrication application amount, lubrication application position, lubrication application method, and running time from lubrication application to the execution of the low-speed small-angle reciprocating contact operation, or are within the corresponding preset allowable deviation range.

4. The AI-based method for diagnosing joint performance degradation in industrial robots according to claim 1, characterized in that, The fact that the sample is consistent with the current processing background conditions specifically means: The processing procedure corresponding to the sample is the same as or within one or more of the following: workpiece type, processing tool type, clamping method, working trajectory, reciprocating swing angle range, motion speed range, contact force range, robot posture range, and ambient temperature range; or it is within the corresponding preset allowable deviation range.

5. The AI-based method for diagnosing joint performance degradation in industrial robots according to claim 1, characterized in that, The calculation process of the equivalent off-center load index includes: Obtain the offset of the workpiece's center of gravity relative to the clamping mechanism and the workpiece's mass parameters; Based on the centroid offset and the workpiece mass parameters, determine the eccentric load caused by the centroid offset of the workpiece. The equivalent off-center load index of the corresponding harmonic reducer joint is determined based on the eccentric load amount.

6. The AI-based method for diagnosing joint performance degradation in industrial robots according to claim 1, characterized in that, The joint drive response data is one or more of the following: drive current data, drive torque data, servo output power data, or joint load rate data. The change curve refers to the curve formed by the change of the joint drive response data with the operation time or the number of reciprocating operations within a preset working cycle.

7. The AI-based method for diagnosing joint performance degradation in industrial robots according to claim 6, characterized in that, The frequency of abnormal fluctuation points specifically refers to: within the preset working cycle, identifying multiple periodic fluctuation points or periodic fluctuation segments in the change curve, determining periodic fluctuation points or periodic fluctuation segments whose fluctuation amplitude exceeds a preset amplitude threshold as abnormal fluctuation points, and determining the frequency based on the number of abnormal fluctuation points and the total number of all periodic fluctuation points or periodic fluctuation segments.

8. The AI-based method for diagnosing joint performance degradation in industrial robots according to claim 1, characterized in that, The steps for determining the relationship between the frequency of abnormal fluctuation points and the equivalent off-center load exponent, and for identifying the critical inflection point where the frequency of abnormal fluctuation points transitions from a stable state to an increasing state, specifically include: Several samples are arranged in ascending order of equivalent off-load index to obtain a sample sequence. Based on the equivalent off-load index and the frequency of abnormal fluctuation points corresponding to each sample in the sample sequence, a curve showing the change between the equivalent off-load index and the frequency of abnormal fluctuation points is generated. The above-mentioned change curves are segmented and identified to determine the stable region where the frequency of abnormal fluctuation points remains within a preset stable range as the equivalent off-center load index changes, and the rising region where the frequency of abnormal fluctuation points increases as the equivalent off-center load index increases. The transition point between the stable region and the rising region is determined as the critical inflection point, and the equivalent off-center load index corresponding to the critical inflection point is taken as the critical equivalent off-center load index.

9. The AI-based method for diagnosing joint performance degradation in industrial robots according to claim 1, characterized in that, When correcting the lubrication health evaluation results of the target harmonic reducer joint, a preset correction function is used, which is: ； in, This refers to the revised lubrication health assessment results. This refers to the initial lubrication health assessment results. This refers to the lower limit of the value for the lubrication health evaluation result. This refers to the current equivalent off-center load index of the target workpiece. This refers to the equivalent off-center load index corresponding to the critical inflection point. This refers to exceeding the allowable range. This refers to a preset correction strength coefficient, and satisfies... Greater than 0, This refers to the off-load correction factor.

10. The AI-based method for diagnosing joint performance degradation in industrial robots according to claim 1, characterized in that, The lubrication health evaluation result is a lubrication health score or lubrication health index that characterizes the lubrication health of the target harmonic reducer joint. After correcting the lubrication health evaluation results of the target harmonic reducer joint, the diagnosis results of lubrication performance degradation of the target harmonic reducer joint are determined based on the corrected lubrication health evaluation results.

Images