Automatic auxiliary system for maintaining and carrying differential mechanism of loading machine

Abstract

The invention relates to the technical field of loader maintenance, and discloses an automatic auxiliary system for maintenance and carrying of a loader differential mechanism. A maintenance parameter acquisition module of the system acquires current pose deviation data and vibration spectrum characteristics of a differential assembly and clamping force distribution parameters of a maintenance tool clamp in real time to generate a maintenance parameter set; the pose compensation module generates a dynamic compensation vector sequence according to the pose deviation data and updates the compensation step length in combination with the vibration spectrum characteristics; the carrying path planning module generates a carrying track control point sequence based on the maintenance process time window constraint and the clamping force distribution parameters; and the execution control module activates the dynamic compensation vector sequence and generates a pose correction instruction when detecting that the pose deviation data exceeds a first activation threshold value, activates the carrying track control point sequence and generates a track correction instruction when the clamping force distribution parameter exceeds a second activation threshold value, and inputs the instruction to a maintenance mechanical arm execution unit.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of loader maintenance, in particular to a loader differential maintenance and carrying automation auxiliary system. BACKGROUND

[0002] As a key equipment in engineering construction, the differential assembly of the loader bears the important functions of transmitting power and adjusting the speed difference of the wheels, directly affecting the driving stability and work efficiency of the whole machine. The differential assembly structure is precise, and the internal gears, bearings and other components are prone to wear, deformation and other faults in long-term high-strength operation, and need to be disassembled, repaired and assembled regularly. During the differential maintenance process, carrying and positioning are the key links to ensure the maintenance quality, and the precise transfer of the differential assembly between different maintenance stations and the accurate docking with the maintenance tool fixture need to be realized. Traditional differential maintenance and carrying is mostly completed by manual or semi-automatic equipment. Manual carrying not only has high labor intensity, but also is affected by the experience, physical strength and other factors of the operator, and it is difficult to ensure the pose stability of the differential assembly during the carrying process, which may cause secondary damage to the components due to collision and tilting. Although the semi-automatic equipment can reduce manual intervention, it has obvious limitations: first, the parameter collection dimension is single, and only the position coordinate information of the differential is concerned, while the key parameters such as vibration state and fixture clamping force are ignored, which makes it difficult to find the potential pose deviation and unstable clamping in the maintenance process in time; second, the pose compensation mechanism is fixed, and the compensation step cannot be dynamically adjusted according to the real-time working condition after being preset, so when the differential produces additional displacement due to vibration, insufficient or excessive compensation may occur; third, the carrying path planning does not fully consider the time window requirements of the maintenance process, and the path often does not match the process rhythm, affecting the continuity of the maintenance process; fourth, the control logic is simple, and there is no hierarchical response mechanism for multiple parameter abnormalities, so when the clamping force is abnormal or the pose deviation exceeds the standard, the corresponding correction strategy cannot be quickly activated, increasing the maintenance risk. With the application of automation technology in the field of engineering machinery maintenance, higher requirements are put forward for the precision and adaptability of differential maintenance and carrying. In the existing technology, the automation auxiliary system has obvious shortcomings in the comprehensiveness of parameter perception, the dynamics of compensation strategy, the fitting of path planning and the pertinence of control response, which cannot meet the efficient operation demand in complex maintenance scenarios, and restricts the improvement of the automation level of loader differential maintenance. SUMMARY

[0003] The purpose of the present application is to provide a loader differential maintenance and carrying automation auxiliary system to solve the problems raised in the background.

[0004] To achieve the above purpose, the present application provides a loader differential maintenance and carrying automation auxiliary system, which comprises: The maintenance parameter acquisition module is used to collect the current posture deviation data, vibration spectrum characteristics and clamping force distribution parameters of the differential assembly in real time to generate a maintenance parameter set; a posture compensation module, configured to generate a dynamic compensation vector sequence according to the current posture deviation data in the maintenance parameter set, and update the compensation step size of the dynamic compensation vector sequence according to the vibration spectrum characteristics; A transport path planning module, configured to generate a transport trajectory control point sequence based on a maintenance process time window constraint and a clamping force distribution parameter in the maintenance parameter set; an execution control module, configured to activate the dynamic compensation vector sequence and generate a posture correction instruction when detecting that the current posture deviation data exceeds a first activation threshold; and activate the transport trajectory control point sequence and generate a trajectory correction instruction when detecting that the clamping force distribution parameter exceeds a second activation threshold; The execution control module is further used to input the posture correction instruction or trajectory correction instruction into the execution unit of the maintenance robot arm.

[0005] Preferably, the maintenance parameter acquisition module includes: A multi-dimensional sensor array is used to synchronously collect the axial offset and radial runout waveform of the differential assembly during the disassembly phase, as well as the pressure feedback values ​​of each contact point of the fixture; a material identification unit, configured to identify the material type of the differential housing and match the corresponding reference clamping force safety threshold based on a characteristic frequency interval in the radial runout waveform; a parameter fusion unit, configured to convert the axial offset into a three-dimensional spatial offset vector, and perform difference calculation between the pressure feedback value of each contact point and the reference clamping force safety threshold to generate a clamping force deviation distribution map; The set generating unit is used to integrate the three-dimensional space offset vector, radial runout waveform and clamping force deviation distribution diagram into the maintenance parameter set.

[0006] Preferably, the posture compensation module includes: a dynamic learning rate adjustment unit, configured to calculate a current compensation learning rate according to a rate of change of the main frequency amplitude of the vibration spectrum characteristic, wherein the current compensation learning rate is negatively correlated with the rate of change of the main frequency amplitude; a vector generating unit, configured to generate an initial compensation vector sequence based on the current posture deviation data in the maintenance parameter set; A step size optimization unit is used to iteratively update the initial compensation vector sequence using the current compensation learning rate to generate the dynamic compensation vector sequence including the compensation direction and compensation amount.

[0007] Preferably, the maintenance parameter acquisition module further outputs the radial run-out waveform to the pose compensation module; The dynamic learning rate adjustment unit is specifically configured to extract an energy attenuation slope of a preset frequency band in the radial run-out waveform, and adjust an attenuation coefficient of the current compensation learning rate according to the energy attenuation slope.

[0008] Preferably, the carrying path planning module comprises: A time window analysis unit is configured to analyze a disassembly process deadline and an installation process starting time in a maintenance process flow, and generate a carrying time constraint window; A trajectory generation unit is configured to identify a clamping force weak area and generate an initial carrying path by avoiding the clamping force weak area according to a clamping force distribution parameter in the maintenance parameter set; A control point optimization unit is configured to perform time-uniform segmentation on the initial carrying path within the carrying time constraint window, and generate a carrying trajectory control point sequence containing time stamp markers.

[0009] Preferably, the maintenance parameter acquisition module further outputs a mass distribution parameter of the differential assembly to the carrying path planning module; The control point optimization unit is specifically configured to calculate an inertia moment of each path segment according to the mass distribution parameter, and adjust a time stamp interval of the carrying trajectory control point sequence based on the inertia moment.

[0010] Preferably, the execution control module comprises: A priority judgment unit is configured to preferentially execute the pose correction instruction when the current pose deviation data and the clamping force distribution parameter simultaneously exceed an activation threshold; An instruction fusion unit is configured to superimpose a residual pose deviation amount to the trajectory correction instruction after the execution of the pose correction instruction is completed; An execution unit driving unit is configured to drive the maintenance robot arm according to the fused trajectory correction instruction.

[0011] Preferably, the system further comprises: A feedback calibration module is configured to collect actual motion trajectory data of the maintenance robot arm, perform deviation comparison with an expected trajectory of the trajectory correction instruction to generate a trajectory error mapping table; The feedback calibration module is further configured to feed back the trajectory error mapping table to the carrying path planning module; The carrying path planning module corrects a generation rule of the carrying trajectory control point sequence according to the trajectory error mapping table.

[0012] Preferably, the system further comprises: A cooperative verification module is configured to acquire the load deformation data and the environmental vibration spectrum of the maintenance platform in real time during the carrying process.

[0013] Preferably, the cooperative verification module comprises the following steps of: When the load deformation data exceeds the deformation threshold, the pose compensation module is triggered to regenerate the dynamic compensation vector sequence. When the environmental vibration spectrum resonates with the vibration spectrum features in the maintenance parameter set, the carrying path planning module is triggered to re-plan the carrying trajectory control point sequence.

[0014] Compared with the prior art, the present application has the following advantages: The loader differential maintenance carrying automation auxiliary system provides comprehensive and dynamic technical support for the differential maintenance carrying process through the cooperation of multiple modules. The maintenance parameter acquisition module can capture the current pose deviation data, vibration spectrum features and clamping force distribution parameters of the differential assembly in real time, forming a maintenance parameter set that covers key state information in the maintenance carrying process, making the system's perception of the maintenance scene more comprehensive and able to timely detect potential pose deviation, vibration interference and clamping instability. The pose compensation module generates a dynamic compensation vector sequence based on the pose deviation data and updates the compensation step length in combination with the vibration spectrum features, so that the pose correction process no longer relies on fixed parameters, but can flexibly adjust the compensation rhythm according to the real-time vibration state of the differential. When the differential produces vibrations of different frequencies and amplitudes due to external interference or its own structural characteristics, the compensation step length can be adapted accordingly, avoiding the phenomenon of compensation lag or over-adjustment that may occur with fixed step length, and improving the accuracy and timeliness of pose correction.

[0015] The carrying path planning module takes into account the maintenance process time window constraints and clamping force distribution parameters to generate a carrying trajectory control point sequence that better meets the needs of the actual maintenance process. The time window constraints ensure that the differential is transferred between stations at a pace that matches the time nodes of the maintenance process, reducing the likelihood of process delays or overruns. The incorporation of clamping force distribution parameters allows the path planning to avoid trajectory segments that may cause abnormal fluctuations in clamping force, reducing the risk of instability caused by unreasonable paths, and making the carrying process more suitable for the structural characteristics of the differential and the stress characteristics of the clamp. The execution control module realizes the hierarchical response to different abnormal states by setting the first activation threshold and the second activation threshold. When the pose deviation data exceeds the first activation threshold, the system rapidly activates the sequence of dynamic compensation vectors and generates the pose correction instruction to timely correct the deviation. When the clamping force distribution parameter exceeds the second activation threshold, the system activates the sequence of carrying track control points and generates the track correction instruction to adjust the carrying path to improve the clamping state. The targeted response mechanism avoids the inadaptability of a single control logic to complex working conditions, so that the system can take precise intervention measures according to the specific abnormal type, and reduces invalid operations. The synergistic effect of each module enables the entire system to autonomously perceive, dynamically adjust and accurately control in a complex maintenance environment, reduces the dependence on artificial experience, improves the automation degree and stability of the loader differential maintenance carrying process, and better adapts to the diversified working condition requirements in differential maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The timing diagram of the loader differential maintenance carrying automation auxiliary system described in the present application; Figure 2 The work flow chart of the pose compensation module; Figure 3 The work flow chart of the dynamic learning rate adjustment in the pose compensation module; Figure 4 The work flow chart of the control point optimization in the carrying path planning module. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0018] Please refer to Figure 1 The present application provides a loader differential maintenance carrying automation auxiliary system, which comprises: The differential repair process is accurately positioned and safely transported through multi-module cooperation. The system includes a repair parameter acquisition module, a pose compensation module, a carrying path planning module, and an execution control module. The repair parameter acquisition module acquires the pose deviation, vibration characteristics and clamp force data of the differential assembly in real time through a sensor array, forming a structured repair parameter set. The pose compensation module generates a dynamic compensation vector based on the pose deviation data, and adjusts the compensation step size adaptively according to the vibration characteristics. The carrying path planning module generates a timestamped carrying trajectory control point sequence in combination with process time constraints and clamping force distribution. The execution control module selectively activates the pose correction or trajectory correction instructions according to the parameter threshold value judgment, and drives the repair robot to complete the accurate operation.

[0019] Embodiment 1: refer to Figure 2 The multi-dimensional sensor array of the repair parameter acquisition module adopts a modular design. Laser displacement sensors are installed on the orthogonal coordinate system of the end effector of the robot arm. Three measurement heads are arranged along the X, Y and Z axes, respectively, forming a spatial measurement network. Each measurement head uses a semiconductor laser with a wavelength of 650nm. The beam diameter is controlled within 0.5mm, and the measurement resolution reaches 0.01mm. The sensor array is connected to the main control system through an industrial Ethernet. The data transmission uses a timestamp synchronization mechanism to ensure the timing consistency of the measurement data of each axis. The three-axis accelerometer uses a micro sensor made of MEMS technology. The range covers ±50g, and the frequency response characteristic remains flat within 5-500Hz. The accelerometer is installed on the predetermined measurement reference surface of the differential housing and can be quickly installed and removed through a magnetic base. The signal conditioning circuit has an anti-aliasing filter built-in. The sampling rate is set to 1kHz, and the ADC conversion accuracy is 16 bits.

[0020] The material recognition unit runs on an embedded processor. The algorithm library contains feature frequency templates of common differential housing materials such as cast iron, cast steel, and aluminum alloy. After windowed Fourier transform of the vibration signal, the spectral features in the 50-250Hz frequency band are extracted and matched with the template library. The matching process uses the dynamic time warping algorithm, allowing a frequency offset tolerance of ±5%. The recognition result is output as a material type code and a confidence score. When the confidence is less than 85%, an artificial review request is triggered. The reference clamping force safety threshold corresponding to the material is stored in an encrypted database, including the static clamping force and the correction coefficient under dynamic working conditions. The data access uses a permission grading management mechanism.

[0021] The pressure sensor matrix adopts flexible thin film pressure sensing technology, and 64 sensing units are evenly distributed on the contact surface of the clamp in an 8x8 array. The effective sensing area of each sensing unit is 10mmx10mm, the pressure measurement range is 0-5MPa, and the non-linear error is controlled within ±1% FS. The sensing matrix realizes parallel data reading through distributed acquisition nodes, and the sampling period is 10ms. After temperature compensation and zero point calibration, the pressure data is transmitted to the parameter fusion unit for subsequent processing. The contact pressure distribution adopts polar coordinate representation method, and the circumference of the clamp is divided into 32 sectors, and the average pressure value of each sector is calculated. The data processing flow of the parameter fusion unit includes three main stages. The spatial coordinate conversion stage converts the axial offset measured by the laser displacement sensor into a three-dimensional vector in the mechanical arm coordinate system through the homogeneous transformation matrix, and considers the installation bias of the clamp and the size change caused by temperature during the conversion process. The pressure data analysis stage adopts a sliding window statistical method to calculate the mean and standard deviation of each pressure sensing unit within the last 1 second time window, and the data points exceeding the 3σ range are marked as abnormal values. The clamping force evaluation stage compares the real-time pressure distribution with the safety threshold matched with the material point by point to generate a two-dimensional deviation map encoded with pseudo-color, and the color scale range is set to -30% to +30% of the safety threshold percentage.

[0022] The data integration process of the set generation unit adopts a hierarchical data structure, the top layer is a time synchronization index, recording the absolute timestamp of data acquisition and the progress percentage of the relative maintenance process. The middle layer includes three types of core parameters: spatial position data is represented in a right-handed Cartesian coordinate system, with coordinate system conversion parameters; vibration feature data includes time domain waveform and frequency domain spectrum two forms; clamping force data retains the original pressure matrix and derived statistical features. The bottom layer is quality control information, recording sensor self-checking state and data check code. The final generated maintenance parameter set is packaged as a standardized data package, which is read by other modules through shared memory.

[0023] The dynamic learning rate adjustment unit of the pose compensation module realizes the closed-loop coupling of vibration features and compensation parameters. The main frequency band identification of the vibration spectrum adopts the peak detection algorithm, and the local amplitude maximum value is found in the range of 50-150Hz, and the minimum frequency interval between adjacent peaks is not less than 10Hz. The main frequency amplitude change rate calculation adopts least squares linear fitting, and the time window length is adaptively adjusted according to the vibration intensity, ranging from 100ms to 1s. The learning rate adjustment curve adopts a piecewise linear function, which remains constant when the amplitude change rate is less than 2% / s, and decreases inversely when it is greater than 5% / s, and the intermediate transition region uses smooth interpolation.

[0024] The initial compensation vector calculation of the vector generation unit is based on the robot kinematics model. The input three-dimensional space offset is first processed by coordinate normalization to convert it into local coordinate system parameters with the robot end as the origin. Then the required compensation displacement of each joint is calculated by the inverse kinematics solver, and the solving process considers the current posture of the robot and the joint position constraint. For the redundant degree of freedom system, the gradient projection method is used to optimize the solution space selection. The generated initial vector sequence contains 6 degree of freedom components, each component is accompanied by confidence evaluation and error boundary estimation.

[0025] The iterative update process of the step optimization unit adopts the model predictive control strategy. Before each iteration, the system predicts the possible vibration response after executing the current compensation vector, and the prediction model is based on the ARMA time series model established based on historical data. The residual error between the actual vibration spectrum and the prediction result is used to correct the learning rate, and the larger the residual error is, the greater the learning rate adjustment amplitude is. The compensation direction optimization adopts the conjugate gradient method to ensure the convergence speed while avoiding oscillation. The compensation amount adjustment introduces a momentum term to retain part of the adjustment amount of the previous iteration to improve stability. The final output dynamic compensation vector sequence is accompanied by execution priority marking and time validity identification.

[0026] The calibration and maintenance mechanism of the multi-dimensional sensor array includes online and offline modes: online calibration uses a reference block installed on the fixture, and performs zero position calibration and sensitivity verification automatically every day; offline calibration is performed in a professional metrology room, using a standard vibration table and a laser interferometer as a reference, and the calibration period is not more than 6 months. The sensor health status monitoring is continuously performed, and real-time alarm is given for abnormal conditions such as drift overrun and noise increase. The data acquisition software realizes a double buffering mechanism to ensure that no effective data is lost during sensor calibration.

[0027] The template library of the material identification unit supports online updating and maintenance: when a new type of differential material is introduced, the typical vibration spectrum characteristics are collected through the expert mode, and after wavelet denoising and feature extraction, they are added to the template library. The template matching algorithm adopts dynamic weight distribution, and different feature weights are assigned to the low frequency band (50-100Hz) and the high frequency band (100-250Hz). The identification result output is accompanied by detailed spectrum comparison chart and similarity analysis report, supporting visual verification during manual review.

[0028] The temperature compensation algorithm of the pressure sensing matrix is based on the temperature sensor data installed beside each sensing unit, and the compensation model uses a cubic polynomial fitting, and the coefficients are determined through factory calibration experiments. The pressure data acquisition channel is configured with a programmable gain amplifier, and the range is automatically adjusted according to the current pressure range. The signal conditioning circuit has built-in hardware filter and overvoltage protection module to ensure the measurement stability in complex electrical environment of the maintenance site. The pressure data of the edge area of the matrix is compensated by using the spatial interpolation algorithm to improve the continuity of the overall distribution map.

[0029] Embodiment 2: Refer to Figure 3 The original radial runout waveform signals transmitted by the maintenance parameter acquisition module to the pose compensation module undergo multi-stage preprocessing. In the signal acquisition stage, an anti-aliasing filter is used with a cutoff frequency set to 500 Hz and a fixed sampling rate of 2 kHz to ensure the Nyquist sampling theorem. The original signal is first subjected to direct current component elimination, and then a fourth-order Butterworth bandpass filter is used to retain the effective frequency band of 50-200 Hz. The signal segmentation processing uses the overlapping window method, with a window length of 200 ms and an overlap rate of 50%. The Hanning window function is applied to each window to reduce spectral leakage. The preprocessed signal is converted to the frequency domain by fast Fourier transform, with a spectral resolution of 5 Hz.

[0030] The energy feature extraction of the dynamic learning rate adjustment unit uses complex Morlet wavelet transform with a center frequency set to 100 Hz and a bandwidth parameter of 3. After modulus calculation of the wavelet coefficient matrix, energy integration is performed in the 50-200 Hz frequency band. The calculation of energy decay slope uses the linear regression method, and the time window length is adaptively adjusted according to the current vibration intensity, ranging from 100 ms to 500 ms. When the absolute value of the energy slope exceeds the set threshold, the learning rate adjustment mechanism is started, and the adjustment process follows the following formula: ; where: represents the current compensation learning rate, is the initial learning rate benchmark value, is the decay sensitivity coefficient (default value 0.25), represents the energy decay slope. This formula ensures that when the high-frequency vibration energy decays rapidly, the system automatically reduces the learning rate to avoid overcompensation. The decay coefficient is dynamically adjusted based on historical data, and the average amplitude of the last 10 adjustments is used as a correction reference.

[0031] The time window analysis unit of the transportation path planning module communicates with the maintenance management system using OPCUA protocol to obtain real-time process time data. The determination of the disassembly process cutoff time t1 involves double verification of the mechanical arm homing completion signal and the visual system confirmation signal. The synchronization of the installation process start time t2 considers the preheating time buffer, and the preparation program is started 30 seconds before the scheduled time. The boundaries of the time constraint window are set with a safety margin, with the window start time delayed by 200 ms and the end time advanced by 300 ms to prevent conflicts during process transitions.

[0032] The weak area identification algorithm of the trajectory generation unit adopts a morphological processing method. The clamping force deviation distribution map is first subjected to Gaussian smoothing filtering to eliminate local noise interference. Then, adaptive threshold segmentation is applied to mark the area with a pressure value lower than 70% of the safety threshold as a candidate area. The region growing algorithm connects adjacent candidate points to form a continuous area, and isolated areas with an area less than 5 cm² are filtered. The final weak area boundary is represented by the smallest convex hull, and a 3 mm safety buffer is extended outward. The path planning adopts an improved RRT* algorithm, which increases the weak area repulsion field on the basis of traditional random sampling, so that the planned path naturally moves away from the high-risk area.

[0033] The time uniform segmentation algorithm of the control point optimization unit considers the mechanical arm dynamics constraint, divides the total handling time into several time periods according to the equal division principle, and each time period has a length not less than 50 ms. In the time period, the feasible path point density is calculated according to the maximum acceleration limit of each joint of the mechanical arm. For straight path segments, the control point spacing is uniform; in the turning area, the control point density increases with the decrease of the curvature radius. The time stamp label adopts a double-track system of relative time and absolute time, which contains the millisecond offset relative to the process start and records the standard NTP time format.

[0034] The quality distribution parameters transmitted by the maintenance parameter acquisition module to the handling path planning module include the inertia tensor matrix and the mass center coordinates. The inertia parameters are pre-calculated by a three-dimensional modeling software and stored in a standardized data format. The quality distribution changes are monitored in real time during the handling process, and parameter updating is triggered when parts are loose or missing. The mass data is coupled with the mechanical arm dynamics model, and the load inertia ratio of each axis is pre-calculated in the trajectory planning stage.

[0035] The inertia torque compensation algorithm of the control point optimization unit is based on the Euler dynamics equation. For each path segment, the equivalent inertia of the mechanical arm end effector in the motion direction is calculated. The torque demand prediction considers the combined effects of the Coriolis force and the centripetal force, and adopts a feedforward control idea to compensate in advance. When the predicted torque exceeds the threshold, the system automatically extends the time allocation of the path segment, and the adjustment amplitude is proportional to the overrun degree. The adjustment of the time stamp interval follows the motion smoothness principle, and the time interval variation rate of adjacent control points does not exceed 20%.

[0036] The signal synchronization mechanism of the multi-dimensional sensor array adopts a hardware trigger method. The laser displacement sensor and the three-axis accelerometer share the same external clock source, and the clock jitter is controlled within 1 μs. The pressure sensor matrix adopts a time-division multiplexing acquisition scheme, and 64 channels are polled in 8 groups. The delay between channels in the group is less than 10 μs. The data packets of all sensors are attached with a unified time stamp label, and the time reference comes from the GPS disciplined clock module.

[0037] The spectral feature library of the material recognition unit adopts a hierarchical storage structure. The base layer contains the reference spectrum of standard materials, and the extended layer stores the variation features collected on site. The matching algorithm uses dynamic time warping distance as the similarity measure, allowing the spectrum to be stretched and contracted elastically on the frequency axis. The feature library weight is updated in real time during the recognition process, and the matching priority is increased for frequently occurring material types. When a new material appears, the system automatically starts the learning mode, records the typical vibration mode and requests manual annotation.

[0038] The spatial coordinate conversion of the parameter fusion unit uses quaternion representation. The offset measured by the laser displacement sensor is first converted to the fixture coordinate system, and then mapped to the mechanical arm base coordinate system through the homogeneous transformation matrix. The conversion parameters include installation bias, temperature compensation, and mechanical deformation correction. The pressure data fusion uses Bayesian estimation method to combine the current measurement value with the historical statistical characteristics to reduce the influence of random errors. The fusion result is accompanied by uncertainty evaluation, and the confidence of each parameter is represented in the form of covariance matrix.

[0039] The data packaging format of the set generation unit adopts a hybrid encoding scheme. The structured data part uses JSON format to record parameter names and values, and the binary part stores raw waveforms and image data. The data packet adds digital signature and time watermark to ensure integrity and non-tamperability during transmission. The memory sharing area implements partition locking mechanism to prevent data conflicts caused by simultaneous access by multiple modules. Data updates use the publish-subscribe mode, and change notifications are broadcast through the message queue.

[0040] The state monitoring of the dynamic learning rate adjustment unit includes multiple parallel threads. The vibration feature analysis thread continuously tracks the main frequency amplitude variation, the learning rate calculation thread maintains the current compensation parameters, and the system stability evaluation thread monitors the closed-loop performance indicators. The threads exchange data through a lock-free ring buffer to ensure real-time requirements. The learning rate adjustment process uses a gradual strategy, with each adjustment amplitude not exceeding 30% of the previous value, to avoid system oscillation caused by drastic changes.

[0041] The inverse kinematics solver of the vector generation unit supports multiple algorithm switching. The default is to use the analytical method to solve the closed solution, and automatically switch to the numerical iterative method near the singular configuration. The uniqueness of the solution is based on joint limit constraints and workspace accessibility analysis. For redundant degree of freedom systems, the zero space optimization goal is to minimize the sum of the squares of joint torques. The generated initial vector sequence is accompanied by a condition number evaluation of the Jacobian matrix, and when the numerical stability is below the threshold, it triggers a recalculation.

[0042] The model predictive controller of the step optimization unit adopts a rolling horizon strategy, and the prediction horizon length is dynamically adjusted according to the system response speed, ranging from 0.5s to 2s. The optimization objective function balances the pose error and energy consumption, and the weight coefficient changes adaptively with the vibration intensity. The compensation vector generated by each iteration is subjected to feasibility checks, including joint limit verification, speed and acceleration constraint verification, and collision detection. The execution result feedback is used to update the prediction model parameters, forming a closed-loop learning mechanism.

[0043] Embodiment 3: refer to Figure 4 The inertial measurement unit of the maintenance parameter acquisition module adopts a high-precision MEMS gyroscope and accelerometer combination device, which is installed on the reference measurement plane of the differential assembly. The gyroscope range covers The nonlinear error is controlled within 0.1% FS, and the accelerometer range is ±10g with a bandwidth set to 500Hz. The sensor data is transmitted through the SPI interface, and the sampling rate is set to 1kHz. The original signal is output after digital filtering and temperature compensation. The measurement process of the mass distribution parameters includes static and dynamic modes. In the static mode, the center of mass position is calculated by the multi-point weighing method, and in the dynamic mode, the rotational inertia is identified by the excitation response method. The measurement results are stored in the form of a standardized matrix, including mass value, center of mass coordinates and inertia tensor.

[0044] After receiving the mass distribution parameters, the control point optimization unit of the carrying path planning module first performs coordinate system alignment. The inertia parameters of the differential body are converted to the end effector coordinate system of the robot arm, and the conversion process considers the additional mass and geometric offset of the clamp. The dynamics analysis of the path segment uses a piecewise linearization method to discretize the continuous trajectory into several characteristic motion states. The inertia moment calculation of each state point includes translation and rotation components. The translation component is based on Newton's second law, and the rotation component is based on the Euler dynamics equation. The moment demand prediction algorithm uses a forward recursion method to calculate the load inertia from the robot arm base to each joint.

[0045] The priority judgment unit of the execution control module sets a multi-level trigger mechanism. The pose deviation monitoring uses a sliding window statistical method to calculate the average value and standard deviation of the last 10 sampling periods. When the deviation exceeds the 3σ range for 3 consecutive periods, a first-level alarm is triggered, and the system enters a warning state without interrupting the current operation. The clamping force deviation monitoring uses regional aggregation analysis to divide the clamp contact surface into 8 functional regions. If the average pressure of any region deviates from the safety threshold by 20%, a second-level alarm is triggered. When both alarms occur simultaneously, the system starts the interruption processing program, saves the current state register, and performs the on-site protection action.

[0046] The processing flow of the instruction fusion unit contains two stages of spatial mapping and time alignment. The residual pose deviation is first projected to the manipulator workspace and decomposed into components along the tangent and normal directions of the trajectory. The tangent component is absorbed by adjusting the trajectory velocity profile, and the normal component is compensated by path offset. The time alignment stage inserts the modified instructions into the existing instruction queue, considering the motion synchronization of each axis of the manipulator, to avoid the time difference of different axes reaching the target point exceeding the allowed range. The weight coefficient in the fusion process is dynamically adjusted according to the clamp pressure distribution, and the higher the pressure imbalance, the greater the normal direction compensation weight.

[0047] The execution unit driving unit adopts a distributed control architecture, with an independent servo driver configured for each joint axis, and synchronization control is achieved through an EtherCAT bus. The instruction issuance adopts a periodic synchronous position mode, with a cycle time set to 2ms. The trajectory interpolation algorithm adopts a seven-segment S-type velocity curve to ensure continuous acceleration. The actual position feedback of each axis is compared with the instruction position, and a protective stop is triggered when the deviation exceeds the safety threshold. The driving parameters are automatically adjusted according to the load inertia, including the velocity feedforward gain and the acceleration feedforward gain.

[0048] The dynamics model of the maintenance manipulator is maintained in the system database, including the mass properties of each link, the joint friction characteristics, and the motor dynamics parameters. The model updating mechanism runs a self-identification program periodically to obtain the latest parameters through the excitation response method. During online identification, the manipulator moves along a predefined test trajectory, and the current, speed, and position signals of each joint are collected. The data processing uses the recursive least squares method, and the identification results are fused with the theoretical model to generate a hybrid parameter set. Model verification is achieved by comparing the predicted torque and the actual motor current, and re-identification is triggered when the error exceeds 15%.

[0049] The real-time updating mechanism of the mass distribution parameters monitors the state changes of the differential assembly, and the vision system periodically scans the appearance of the differential to detect missing or position offset of parts. When an anomaly is found, the mass re-estimation program is started, and the parameterized adjustment interface of the three-dimensional model is called. The dynamics parameter update caused by mass change adopts incremental processing, and only the affected part is recalculated. The system maintains a change history record for analyzing the correlation between mass distribution and vibration characteristics.

[0050] The temperature compensation algorithm of the inertial measurement unit is based on the temperature probe data distributed around the sensor. Each probe uses a PT100 platinum resistance with a measurement accuracy of ±0.1℃. The compensation model is established through polynomial fitting, and the coefficients are determined during factory calibration. The real-time compensation process considers the temperature gradient effect and performs weighted averaging on the temperature differences of different parts of the sensor. The calibration data is stored in non-volatile memory, supporting power saving and automatic loading on power-up.

[0051] The timestamp interval adjustment of the control point optimization unit follows the energy optimization principle. Under the premise of ensuring motion accuracy, the control point distribution that minimizes the total energy consumption of the robot arm is selected. The energy consumption model includes motor copper loss, iron loss, and mechanical friction loss, which are calculated according to the real-time working point of each joint. The adjustment algorithm uses a heuristic search method to perform local optimization based on the initial uniform distribution. The change rate of the interval between adjacent control points is limited to within 20% to avoid sudden acceleration.

[0052] The multi-task scheduling of the priority judgment unit adopts a preemptive strategy, with the pose correction task set as high priority and the trajectory modification task of low priority interrupted. The task switching process saves the context environment of the interrupted task, including the instruction pointer and register state. The resource conflict processing adopts a priority inheritance protocol to prevent high-priority tasks from being blocked by low-priority tasks. The task execution time monitoring is performed in real time, and the watchdog reset is triggered for tasks that are not completed within the specified time.

[0053] The space mapping algorithm of the instruction fusion unit is based on the robot Jacobian matrix, which converts the pose deviation in Cartesian space to the joint space, considering the singularity characteristics of the current configuration of the robot arm. The damping least squares method is used to handle the singular region, adding a regularization term when inverting the matrix. The fused joint instructions are subjected to amplitude limiting to ensure that the motion parameters of each axis are within the allowed range. Collision detection is performed before issuing the instructions to verify the safety of the path using the bounding box algorithm.

[0054] The fault detection mechanism of the execution unit driving unit includes multiple protections. The current loop monitors the balance of the motor phase current to detect open-phase or short-circuit faults. The speed loop monitors the tracking error, and a persistent deviation beyond the limit is judged as a mechanical jam. The position loop uses dual encoders for verification, with a high-resolution absolute encoder as the main encoder and an incremental encoder as a redundant backup. The cooling system monitors the temperature of the driver in real time, and automatically reduces the power when the temperature exceeds the safety threshold. All fault events are recorded in the black box memory for post-analysis.

[0055] The calibration and maintenance of the maintenance robot includes daily inspection and regular maintenance. Daily inspection is automatically performed at each system startup, including zero operation of each axis and load detection. Regular maintenance is triggered according to the running time count, including harmonic reducer lubrication, cable inspection, and encoder calibration. The remaining life prediction of key components is based on the running hours and load statistics to provide early warning of possible failures. The calibration data management adopts version control, supporting backtracking to any historical version.

[0056] Example 4: The optical tracking system of the feedback calibration module adopts a six-degree-of-freedom infrared camera networking scheme. Four capture units are arranged around the maintenance work area to form a three-dimensional measurement space covering 3m x 3m x 2m. Each capture unit integrates a 120Hz sampling rate CMOS sensor and an infrared LED array, and tracks the reflective ball on the end effector of the mechanical arm through active marker point recognition technology. The marker point arrangement adopts an asymmetric geometric pattern, with four 15mm diameter reflective balls installed around the flange of the effector, and the center point as the main reference point, with the remaining three points used for attitude calculation. The system calibration process uses a standard length rod, and the spatial accuracy is verified at multiple positions in the measurement space to ensure that the positioning error throughout the space does not exceed 0.1mm.

[0057] The construction of the trajectory error mapping table adopts a space-time alignment method: the actual trajectory data collected by the optical system is synchronized with the expected trajectory issued by the handling path planning module on the time axis, and the synchronization signal is derived from the PPS pulse of the system master clock. Error analysis is performed in three dimensions: position error is calculated as Euclidean distance, direction error is represented as the angle between quaternions, and velocity error is obtained by differentiating position data. Error data is counted in 100ms time windows to generate structured records containing multiple indicators. A typical trajectory error mapping example is shown in Table 1.

[0058] Table 1: Trajectory error mapping example (time window 100-400ms) After receiving the error mapping table, the handling path planning module starts the control point sequence correction program. The correction process is divided into offline and online stages. The offline stage analyzes historical error data to identify systematic deviation patterns, such as persistent positive deviation in specific path intervals or speed-related errors. The online stage monitors current error trends in real time and dynamically adjusts subsequent control points during trajectory execution. The correction algorithm uses error feedforward compensation principles to pre-add the average error amount obtained by statistics to the target position. For periodic error patterns, the system establishes an error prediction model to calculate the compensation amount in advance based on the current motion state.

[0059] The deformation monitoring system of the collaborative verification module is arranged with twelve fiber Bragg grating sensors at key positions of the maintenance platform. The sensors are distributed in a network along the main beam of the platform, with a measurement point spacing of 300mm, covering all major load-bearing structures. The wavelength drift of the grating is read by a demodulator with a resolution of 1pm, corresponding to a strain measurement accuracy of 1με. The deformation reconstruction algorithm is based on a finite element model, which converts discrete point strain data into a three-dimensional deformation field of the platform as a whole. Temperature compensation uses the reference grating method, with compensation sensors arranged in non-stressed areas of the platform to eliminate the effects of environmental temperature. The deformation threshold is set considering the material properties, with a warning threshold of 0.1mm in the elastic deformation stage and an emergency threshold of 0.3mm in the plastic deformation stage.

[0060] The environmental vibration monitoring unit adopts a three-axis seismic accelerometer array, with four monitoring points located at the base of the platform. The sensor range is ±2g, and the low-frequency response extends to 0.5Hz, meeting the requirements of quasi-static vibration measurement. The frequency spectrum analysis uses 1 / 3 octave segmentation, with 31 frequency bands from 1Hz to 250Hz. The coherence function of the differential vibration spectrum and the environmental vibration spectrum is calculated by the resonance detection algorithm, and a warning is triggered when the coherence coefficient exceeds 0.7. The vibration data and deformation monitoring results are correlated and analyzed to establish an excitation-response relationship model, which is used to distinguish between structural vibration and external interference.

[0061] The neural network training of the feedback calibration module uses an incremental learning strategy, with the input layer receiving a twenty-dimensional feature vector containing current position, velocity, acceleration, and historical error motion state parameters. The hidden layer is designed as a three-layer structure, with 64, 32, and 16 neurons in each layer, and the activation function uses the ReLU variant. The output layer predicts the error compensation of six degrees of freedom. The training data comes from the long-term accumulation of the error mapping table, and the dataset is automatically updated after each system run. The network parameter optimization uses the stochastic gradient descent method with momentum, and the learning rate is adaptively adjusted according to the proportion of new data.

[0062] The calibration and maintenance of the optical tracking system includes daily quick verification and regular comprehensive calibration. Daily verification uses a reference target fixed in the working area to check system zero drift and measurement stability. Comprehensive calibration is performed every three months by professional metrology personnel using a laser tracker as a reference device. Calibration data management uses version control, and historical calibration records are retained for error trend analysis. The system health status is monitored by the built-in self-test program, including LED light source intensity detection and image sensor noise evaluation.

[0063] The time alignment algorithm for trajectory error analysis processes multi-source heterogeneous sampling data. The optical system data is collected at a fixed frequency of 120Hz, and the robot controller data is recorded at a frequency of 500Hz. The system uses cubic spline interpolation to achieve data rate matching. Clock synchronization is based on the IEEE1588 precision time protocol, with a master-slave clock deviation controlled within 100μs. Data delay compensation considers signal transmission time and processing time, and a complete time chain model from collection to display is established.

[0064] The fiber sensor network of the deformation monitoring system uses wavelength division multiplexing technology, with twelve sensing gratings with center wavelengths spaced 0.8nm apart, distributed in the 1525nm to 1535nm wavelength range. The demodulator scan rate is set to 1kHz, and the dynamic strain measurement range is ±5000 The signal conditioning circuit includes a programmable gain amplifier and digital filter to adapt to the acquisition requirements of signals of varying strengths. The sensor is mounted with a special adhesive to withstand the oil, dirt, and vibration environments found at maintenance sites.

[0065] The coherence analysis of the ambient vibration spectrum utilizes a sliding window method with a calculation window length of 1 second and a sliding step size of 200ms. Vibration characteristics are analyzed jointly in the time and frequency domains. Resonant frequency band identification combines power spectrum density and phase information to eliminate misidentification of antiresonance points. For identified resonant frequency bands, the system automatically generates bandstop filter parameters and injects them into the control system's feedforward channel. Vibration monitoring data is stored in association with maintenance process parameters to establish a correspondence between vibration characteristics and operation types.

[0066] The iterative optimization process for error compensation utilizes a closed-loop verification mechanism. The results of each corrected trajectory are fed back into the training set for neural network parameter updates. Compensation effectiveness is evaluated not only by considering the reduction in the absolute value of the error but also by monitoring the rate of change of higher-order derivatives to ensure smoothness. Compensation is validated through simulation before application, and the effects are previewed on the digital twin model. Compensation rules that have not been used for a long time automatically decay their weight to prevent outdated parameters from affecting current performance.

[0067] The collaborative verification module's data fusion center integrates multi-source information, correlating and analyzing deformation, vibration, and error data against a unified timeframe to establish a multidimensional state vector. Abnormal event detection utilizes a multi-metric, combined judgment system, requiring abnormal readings from at least two independent sensors to trigger an alarm. The data visualization interface simultaneously displays spatial deformation cloud maps and frequency spectrum waterfall plots, assisting engineers in understanding system status. All monitoring data is recorded in complete time series, supporting backtracking queries by time point or event type.

[0068] Example 5: The load-bearing deformation monitoring system of the collaborative verification module adopts a distributed fiber optic sensing network, and sixteen strain measurement points are arranged on the key support structure of the maintenance platform. A fiber Bragg grating sensor is installed at each measuring point, and the central wavelength of the grating is evenly distributed in the range of 1525nm to 1545nm. The demodulator adopts tunable laser scanning technology, with a wavelength resolution of 1pm, corresponding to a strain measurement accuracy of ±2με. The sensor layout scheme has been optimized through finite element analysis and covers the stress concentration areas of the platform's main beams, cross beams and connection nodes. Temperature compensation is achieved through a reference grating set in the non-stressed area to eliminate measurement deviations caused by ambient temperature fluctuations. The deformation data acquisition period is set to 50ms. When the deformation exceeds 0.05mm for three consecutive sampling periods, an early warning is triggered, and when it exceeds 0.1mm, it is determined to be a deformation threshold exceeding event.

[0069] The environmental vibration monitoring unit is configured with four sets of three-axis acceleration sensor arrays arranged at the maintenance platform base and the differential mechanism clamp installation position. The sensor range covers ±5g, and the frequency response range is 0.5Hz to 500Hz, meeting the full-band monitoring requirements from low-frequency shaking to high-frequency vibration. The signal conditioning circuit is built-in with an anti-aliasing filter and a 24-bit analog-to-digital converter, with a sampling rate set to 2kHz. The spectrum analysis adopts a real-time parallel processing architecture, with 1 / 3 octave analysis completed on a dedicated digital signal processor, and the power spectral density calculation of each frequency band using the Welch average periodogram method. The resonance detection algorithm compares the differential mechanism vibration characteristic frequency with the environmental vibration spectrum coherence function, and determines the resonance risk in the frequency band where the coherence coefficient exceeds 0.7 and the phase difference is stable.

[0070] The dynamic compensation vector regeneration process is started immediately after detecting the deformation overrun event, first freezing the compensation instructions currently being executed, and saving the state parameters of each axis of the robot arm. The deformation field reconstruction algorithm converts the discrete point strain measurement data into a three-dimensional displacement field of the platform as a whole, and calculates the deformation gradient directly related to the differential mechanism installation position. The new compensation vector generation considers the deformation trend prediction, and uses time series analysis method to estimate the deformation increment in the next 500ms. The compensation amount calculation introduces a safety factor, adding a 20% margin to the predicted value. The vector verification process simulates the execution effect through the digital twin model, and confirms that there is no collision risk before issuing to the actuator.

[0071] The carrying trajectory re-planning mechanism is activated after the resonance frequency band is detected and confirmed, the system analyzes the coherence characteristics of the vibration spectrum, and identifies the key frequency band that has the greatest impact on the differential mechanism maintenance quality. The trajectory optimization objective function is adjusted to minimize the vibration sensitivity, and the frequency domain vibration suppression requirement is added to the traditional path planning constraint conditions. The timestamp distribution of the control point sequence is adjusted according to the vibration characteristics to avoid integer multiple periods of the resonance frequency. After the new trajectory is generated, it enters the trial operation mode, and the robot arm executes the previous path at 50% of the rated speed, while monitoring the vibration response changes. The vibration amplitude is reduced by more than 30% before confirming that the trajectory update is effective, otherwise it returns to re-planning.

[0072] The installation process of the strain sensor uses special epoxy resin glue to fix, and after curing, a uniform thickness adhesive layer is formed. The metal surface is sandblasted before installation to improve the adhesive strength of the glue layer. The optical fiber wiring uses stainless steel hose protection, and enough excess length is reserved at the moving parts. The sensor network adopts a double-ring topology structure, the main and standby channels run independently, and single-point failure does not affect the overall function. The optical cable joint is treated with waterproof and oil-proof, suitable for the harsh environment of the maintenance workshop. The initial reference value of each measurement point is collected under the empty load state of the platform and stored in the calibration database for subsequent comparison.

[0073] The frequency band energy calculation of the vibration monitoring unit uses a digital filter set method, which divides the 0.5-500Hz frequency domain into 31 1 / 3 octave bandpass filter sets, each filter having strict amplitude and phase characteristics. The real-time energy integration window is set to 100ms, with a sliding step of 10ms, ensuring that rapidly changing vibration characteristics are not missed. The frequency band coherence analysis uses complex fast Fourier transform to calculate the cross-power spectral density of the differential vibration signal and the environmental vibration signal. The coherence threshold setting has a hysteresis characteristic, with a trigger value of 0.7 and a reset value of 0.5, to avoid frequent state switching.

[0074] The time series analysis of the deformation prediction model uses the ARIMA method. The model parameters are automatically trained based on historical deformation data, including autoregressive terms, difference orders and moving average terms. The prediction step is aligned with the control system sampling period, and the development of the deformation of the future ten sampling points is predicted each time. Model confidence evaluation is achieved through residual analysis, and model retraining is triggered when the prediction error continues to increase. Deformation data is stored in association with platform load distribution information, and a typical deformation mode library is established under different maintenance procedures.

[0075] The trajectory optimization of resonance suppression uses a frequency domain sensitivity weighting method, which gives higher optimization weight to vibration sensitive frequency bands based on frequency response function analysis. The path smoothness constraint is changed to a frequency domain representation, limiting the energy distribution of the trajectory spectrum in the sensitive frequency band. Time stamp adjustment avoids integer multiple of resonance period time, so that the mechanical arm motion state change point and the vibration peak phase are staggered. The optimization algorithm uses a hybrid strategy of genetic algorithm and gradient descent method to balance global search and local fine tuning.

[0076] The verification process of dynamic compensation includes three-dimensional space interference check, and the digital twin model loads the latest platform deformation data to simulate the whole process of the mechanical arm executing compensation motion. The collision detection algorithm uses the hierarchical bounding box method to quickly screen potential interference areas. Dynamic gap analysis calculates the minimum distance change during motion to ensure that the safety margin is always greater than 10mm. The compensation scheme that fails the verification returns to the correction link, adjusts the compensation direction or reduces the compensation amplitude, and then re-verifies.

[0077] The signal demodulation of the optical fiber sensing network uses a parallel scheme of wavelength scanning and intensity detection, and the wavelength scanning range of the tunable laser covers all the reflection bands of the grating sensors, with a scanning rate matching the deformation monitoring requirements. The light intensity detection channel monitors the reflectivity change of each sensor to identify abnormal loss events. The signal processing unit realizes real-time wavelength peak detection, and uses cubic spline interpolation to improve the positioning accuracy. Network fault diagnosis is realized through optical time domain reflection analysis technology, which accurately locates the breakpoint or loss mutation position.

[0078] The time-frequency analysis of vibration monitoring data adopts wavelet transform method. The continuous wavelet transform provides the time-frequency joint representation of the vibration signal and identifies the instantaneous frequency variation characteristics. The Morlet wavelet is selected as the wavelet basis function, and the center frequency is matched with the characteristic vibration frequency band of the differential. The energy aggregation area detection on the time-frequency plane adopts image processing technology, and the significant vibration events are located through edge detection and region growing algorithm. The analysis results are verified with each other by the spectrum analysis method to improve the reliability of the resonance detection.

[0079] The spatial interpolation algorithm for generating compensation vectors adopts the radial basis function method to fit the discrete deformation measurement data into a continuous spatial displacement field, and the interpolation function form is a multi-quadratic surface. The deformation gradient calculation adopts the central difference method to estimate the displacement rate on the three-dimensional grid nodes. The compensation direction is determined considering the normal vector of the differential installation surface to ensure that the compensation motion is along the effective degree of freedom. The vector amplitude is determined according to the deformation gradient and the historical compensation effect, and the fuzzy logic reasoning is used to process uncertain factors.

[0080] The environment modeling of trajectory re-planning includes dynamic obstacle prediction. The vibration sensitive area is regarded as a virtual obstacle and is represented as a repulsive potential field in the configuration space. The potential field strength is proportional to the vibration sensitivity, which affects the calculation of the cost function of trajectory optimization. The mechanical arm dynamics constraint is converted into a speed-acceleration feasible region, which is directly embedded into the optimization model when parameterizing the trajectory. The planning algorithm adopts a variant of the fast exploration random tree based on sampling, which improves the convergence speed under the premise of ensuring the probability completeness.

[0081] The abnormal event log record of the collaborative verification module adopts a structured format. Each event contains fields such as timestamp, event type, triggered sensor list, severity level, and disposal measures. The event correlation analysis is implemented through a graph database to establish a causal relationship model between different sensor alarms. The log data visualization displays the spatio-temporal evolution relationship between deformation, vibration, and control instructions, which helps engineers understand the system behavior. Long-term running data is used for reliability analysis to identify potential improvement points and maintenance needs.

[0082] It should be noted that, in this text, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0083] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A loader differential service handling automation assistance system characterized by, The method comprises the following steps: a maintenance parameter acquisition module is configured to acquire current pose deviation data, vibration frequency spectrum characteristics and clamping force distribution parameters of a maintenance tooling fixture of the differential assembly in real time, and generate a maintenance parameter set; a pose compensation module is configured to generate a dynamic compensation vector sequence according to the current pose deviation data in the maintenance parameter set, and update a compensation step of the dynamic compensation vector sequence according to the vibration frequency spectrum characteristics; a carrying path planning module is configured to generate a carrying track control point sequence based on a maintenance process time window constraint and the clamping force distribution parameters in the maintenance parameter set; an execution control module is configured to activate the dynamic compensation vector sequence and generate a pose correction instruction when it is detected that the current pose deviation data exceeds a first activation threshold, and activate the carrying track control point sequence and generate a track correction instruction when it is detected that the clamping force distribution parameters exceed a second activation threshold; the execution control module is further configured to input the pose correction instruction or the track correction instruction to an execution unit of a maintenance robot arm.

2. The loader differential service handling automation assist system of claim 1, wherein, The maintenance parameter acquisition module comprises: a multi-dimensional sensor array configured to synchronously acquire an axial offset amount, a radial runout waveform and pressure feedback values of each contact point of the tooling fixture of the differential assembly in a disassembly stage; a material identification unit configured to identify a material type of the differential housing according to a characteristic frequency interval in the radial runout waveform, and match a corresponding reference clamping force safety threshold; a parameter fusion unit configured to convert the axial offset amount into a three-dimensional space offset vector, and perform difference calculation on the pressure feedback values of each contact point and the reference clamping force safety threshold to generate a clamping force deviation distribution map; a set generation unit configured to integrate the three-dimensional space offset vector, the radial runout waveform and the clamping force deviation distribution map into the maintenance parameter set.

3. The loader differential service handling automation assist system of claim 1, wherein, The pose compensation module comprises: a dynamic learning rate adjustment unit configured to calculate a current compensation learning rate according to a main frequency amplitude change rate of the vibration frequency spectrum characteristics, the current compensation learning rate being negatively correlated with the main frequency amplitude change rate; a vector generation unit configured to generate an initial compensation vector sequence based on the current pose deviation data in the maintenance parameter set; a step optimization unit configured to perform iterative update on the initial compensation vector sequence by using the current compensation learning rate, and generate the dynamic compensation vector sequence containing a compensation direction and a compensation amount.

4. The loader differential maintenance carrying automation auxiliary system according to claim 3, wherein: the maintenance parameter acquisition module further outputs the radial runout waveform to the pose compensation module; the dynamic learning rate adjustment unit is specifically configured to extract an energy attenuation slope of a preset frequency band in the radial runout waveform, and adjust an attenuation coefficient of the current compensation learning rate according to the energy attenuation slope.

5. The loader differential service handling automation assist system of claim 1, wherein, The carrying path planning module comprises: a time window analysis unit configured to analyze a disassembly process cutoff time and an installation process starting time in a maintenance process flow, and generate a carrying time constraint window; a track generation unit configured to identify a clamping force weak area according to the clamping force distribution parameters in the maintenance parameter set, and generate an initial carrying path by avoiding the clamping force weak area. A control point optimization unit is configured to divide the initial carrying path evenly in time within the carrying time constraint window to generate the carrying trajectory control point sequence containing time stamp markers.

6. The loader differential repair handling automation assistance system of claim 5, wherein: The repair parameter acquisition module further outputs a mass distribution parameter of the differential assembly to the carrying path planning module; The control point optimization unit is specifically configured to calculate an inertia moment of each path segment according to the mass distribution parameter, and adjust the time stamp interval of the carrying trajectory control point sequence based on the inertia moment.

7. The loader differential service handling automation assist system of claim 1, wherein, The execution control module includes: A priority judgment unit is configured to preferentially execute the pose correction instruction when the current pose deviation data and the clamping force distribution parameter simultaneously exceed the activation threshold; An instruction fusion unit is configured to add a residual pose deviation to the trajectory correction instruction after the execution of the pose correction instruction is completed; An execution unit driving unit is configured to drive the repair robot arm according to the fused trajectory correction instruction.

8. The loader differential service handling automation assist system of claim 7, wherein, Further comprising: A feedback calibration module is configured to acquire actual motion trajectory data of the repair robot arm, and perform deviation comparison with an expected trajectory of the trajectory correction instruction to generate a trajectory error mapping table; The feedback calibration module is further configured to feed back the trajectory error mapping table to the carrying path planning module; The carrying path planning module corrects the generation rule of the carrying trajectory control point sequence according to the trajectory error mapping table.

9. The loader differential service handling automation assist system of claim 1, wherein, Further comprising: A cooperative verification module is configured to acquire load deformation data and environmental vibration spectrum of the repair platform in real time during the carrying process.

10. The loader differential maintenance and handling automated auxiliary system according to claim 9, characterized in that: The cooperative verification module includes execution: When the load deformation data exceeds the deformation threshold, the pose compensation module is triggered to regenerate a dynamic compensation vector sequence; When the environmental vibration spectrum and the vibration spectrum characteristics in the repair parameter set produce a resonance frequency band, the carrying path planning module is triggered to re-plan the carrying trajectory control point sequence.

Images