Adaptive control method for preload of heavy-duty joints based on wear state of transmission components

Abstract

This invention relates to an adaptive control method for preload of heavy-duty joints based on the wear state of transmission components. The method includes: real-time acquisition of joint load data and vibration signals; identification of initial wear characteristic indicators of the transmission components using a support vector machine algorithm; accumulation of running time; dynamic estimation of the wear state of the transmission components using a Kalman filter algorithm to obtain an estimated value including preload attenuation; separation of the wear acceleration factor from the estimated value and numerical integration to obtain the cumulative wear increment; determination of a compensation coefficient according to a preset wear-preload ratio; generation of a control command sequence based on the compensation coefficient, output by an embedded controller to the joint actuator, and application of real-time preload compensation values. This invention can dynamically sense the actual wear state of the transmission components and establish a mapping relationship between wear and preload compensation, achieving real-time adaptive adjustment of joint preload under complex heavy-duty conditions, and continuously maintaining high precision and high stability of joint operation.

Description

Technical Field

[0001] This invention relates to the field of heavy-duty industrial robot research and development technology, and in particular to an adaptive control method for heavy-duty joint preload based on the wear state of transmission components. Background Technology

[0002] In the field of industrial automation, high-precision, heavy-duty industrial robots are widely used in aerospace, automotive manufacturing, and heavy machinery processing industries. The stability and precision of their joints directly determine production efficiency and product quality. As core transmission components, robot joints contain precision transmission parts such as harmonic reducers, bearings, and gears. Under long-term heavy-duty conditions, they must withstand enormous dynamic loads and frequent start-stop and reversal operations. However, with the accumulation of operating time, the internal transmission parts of the joint inevitably experience progressive wear. This wear leads to a gradual decrease in preload, thereby reducing the rigidity and transmission precision of the joint. This can result in accumulated machining errors or, in severe cases, equipment failure or even production accidents. Therefore, maintaining the stability of preload during long-term joint operation has become a critical problem that urgently needs to be solved in this field.

[0003] To address the aforementioned issues, existing technologies primarily seek solutions in two directions: Firstly, some technologies employ a periodic maintenance strategy, where the joints are manually disassembled, inspected, and the preload readjusted after a certain period of operation. While this method can restore joint performance to some extent, its maintenance cycle is often based on experience and fails to reflect the actual wear and tear of the equipment, easily leading to over-maintenance or under-maintenance. Secondly, some more advanced solutions attempt to integrate force or displacement sensors within the joints to monitor changes in preload in real time, triggering alarms or automatic compensation when the preload falls below a set threshold.

[0004] However, these solutions generally suffer from a fundamental flaw: they often simply attribute preload decay to the accumulation of operating time, neglecting the complex coupling relationship between load variations and transmission component wear. In actual heavy-duty operations, the loads borne by robot joints are highly dynamic, potentially experiencing alternating periods of extremely high load impact and low-load idling within a short timeframe. This load fluctuation significantly accelerates the wear of specific transmission components, resulting in a highly nonlinear wear rate. Because existing technologies cannot dynamically sense the actual wear state of transmission components, preload adjustment always lags behind the actual wear process, especially in high-intensity, long-term continuous operation scenarios, where performance degradation is particularly pronounced. Summary of the Invention

[0005] Therefore, the technical problem to be solved by the present invention is to overcome the defect in the prior art that the preload adjustment lags behind the actual wear process due to the neglect of the complex relationship between load change and transmission component wear. The present invention provides a heavy-duty joint preload adaptive control method based on the wear state of transmission components. This method can dynamically sense the actual wear state of transmission components and establish a mapping relationship between wear amount and preload compensation, thereby realizing real-time adaptive adjustment of joint preload under complex heavy-duty working conditions, and thus continuously maintaining the high precision and high stability of joint operation.

[0006] To address the aforementioned technical problems, this invention provides an adaptive control method for preload force of heavy-duty joints based on the wear state of transmission components, comprising the following steps: The load data and vibration signals of the joints of industrial robots under heavy load conditions are collected in real time. Based on the load data and vibration signals, the initial wear characteristics of the transmission components inside the joints are identified using the support vector machine algorithm. Based on the initial wear characteristic index and combined with the cumulative running time, the Kalman filter algorithm is used to dynamically estimate the real-time wear state of the transmission component at the current moment, and the estimated value of the wear state of the transmission component is obtained. The estimated value of the wear state of the transmission component includes the amount of joint preload reduction caused by the wear of the transmission component. The wear acceleration factor caused by load change is separated from the wear state estimate of the transmission component, and the wear acceleration factor is numerically integrated with respect to the running time to obtain the cumulative wear increment of the transmission component. Based on the cumulative wear increment, and according to the preset wear-preload ratio, determine the compensation coefficient required for preload adjustment; A sequence of joint preload control commands is generated based on the compensation coefficient and output by the embedded controller to the joint actuator to apply real-time joint preload compensation values.

[0007] In one embodiment of the present invention, a support vector machine algorithm is used to identify the initial wear characteristic indicators of the transmission components inside the joint, including: The collected load data and vibration signal are aligned with time to form a two-dimensional input sample. The sample at each time point contains a load value and the corresponding vibration amplitude. A support vector machine classifier is trained in advance using sample data with pre-labeled wear levels. During training, the load value and vibration amplitude are used as input features, and the corresponding wear level is used as the classification label to obtain the classification hyperplane parameters. The real-time collected input samples are substituted into the trained support vector machine classifier, and the signed distance from the sample point to the classification hyperplane is calculated. The distance value is used as the initial wear feature index. The larger the distance value, the greater the degree to which the transmission component deviates from its healthy state.

[0008] In one embodiment of the present invention, the Kalman filter algorithm is used to dynamically estimate the real-time wear state of the transmission component at the current moment, including: Establish the state transition equation and observation equation for the wear state of the transmission components: The state transition equation is: x k =x k-1 +v0·△t+w k ; Where: x k Let v0 be the estimated wear state of the transmission component at time k, v0 be the preset nominal wear rate, Δt be the sampling time interval, and w be the mean wear rate. k This is process noise; The observation equation is z k =x k +n k , where z k Let n be the wear observation value obtained from the initial wear characteristic index at time k. k To observe noise; Perform the state prediction step and the state update step sequentially: The state prediction step predicts the prior estimate of the wear state and the prior estimate covariance at the current moment based on the wear state estimate and nominal wear rate at the previous moment. The state update step uses the wear observations at the current moment to calculate the Kalman gain, takes the weighted result of the prior estimate and the observations as the posterior estimate of the wear state at the current moment, and updates the posterior estimate covariance. The posterior estimate is output as the wear state estimate of the transmission component.

[0009] In one embodiment of the present invention, the specific method for converting the initial wear characteristic index into wear observation values ​​is as follows: Beforehand, the transmission components without wear are calibrated, and multiple sets of load-vibration samples under healthy conditions are collected and input into the support vector machine classifier. The average distance from the healthy sample points to the classification hyperplane is calculated and denoted as d0. The distance from the real-time sample point to the classification hyperplane is denoted as d; the wear observation value z is calculated according to the formula z=(d-d0)·k, where k is the wear depth coefficient corresponding to the pre-calibrated unit distance.

[0010] In one embodiment of the present invention, separating the wear acceleration factor caused by load change from the wear state estimate of the transmission component includes: Let x be the estimated wear state of the transmission component at the current moment. k The estimated wear condition at the previous moment is denoted as x. k-1 The sampling time interval is denoted as Δt, and calculated according to the formula r k = (x k -xk-1 )·△t, calculate the actual wear rate r at the current moment. k ; Let L be the load data at the current moment. k The preset reference load value is denoted as L0, and calculated according to formula β. k =L k / L0 calculates the load ratio β k The reference load value L0 is the standard load corresponding to the transmission component when it is running at the nominal wear rate; According to formula A k =r k / v0, calculate the total wear acceleration ratio A k v0 is the preset nominal wear rate; The acceleration factor caused by the load is separated from the total wear acceleration ratio. The separation method is as follows: When β k When >1, according to formula α k =(β k -1)·λ calculates the load acceleration factor α k When β k When ≤1, take α. k =0, where λ is the preset load sensitivity coefficient, which represents the additional increase in wear rate caused by the change in unit load ratio when the load exceeds the reference value; The calculated α k Output as a wear acceleration factor caused by load changes.

[0011] In one embodiment of the present invention, the cumulative wear increment of the transmission component is obtained by numerically integrating the wear acceleration factor with respect to the operating time, including: The wear acceleration factor is discrete-time integraled according to the sampling time interval, as shown in the formula: ; Calculate the cumulative wear increment D from the initial time to the current time. acc (t), where N is the total number of sampling steps from the initial time to the current time, and α i Let v0 be the wear acceleration factor at the i-th sampling time, v0 be the preset nominal wear rate, and Δt be the sampling time interval; Then follow formula D total = D acc (t) + v0·t, calculate the cumulative wear increment D of the transmission component. total , where v0·t is the cumulative wear caused by nominal wear, and t is the total running time from the initial moment to the current moment.

[0012] In one embodiment of the present invention, the compensation coefficient required for adjusting the preload is determined according to a preset wear-preload ratio, including: Obtain the wear-preload proportionality coefficient k, which has been experimentally calibrated beforehand. wp The proportionality coefficient represents the joint preload loss value corresponding to the unit wear depth of the transmission component; According to the formula ΔF=D total ·k wp Calculate the preload loss ΔF caused by the current cumulative wear; Obtain the initial preload value F0 of the joint, and calculate the compensation coefficient C required for preload adjustment according to the formula C=-ΔF / F0. The negative sign indicates that when the preload loss ΔF is positive, the preload needs to be increased to make up for the loss. The compensation coefficient C is a dimensionless ratio. When the compensation coefficient C is negative, it means that the current preload needs to be increased to (1-C) times the initial preload; when the compensation coefficient C is zero, it means that no adjustment is needed.

[0013] In one embodiment of the present invention, a joint preload control command sequence is generated based on a compensation coefficient, including: The calculated compensation coefficient is denoted as C, and the initial preload value of the joint is denoted as F0. The formula ΔF is then used to calculate the compensation coefficient. cmd =∣C∣·F0 Calculate the absolute value of the preload force ΔF that needs to be compensated. cmd , where |C| represents the absolute value of the compensation coefficient C; The force-command conversion coefficient of the joint actuator is denoted as η. This coefficient is obtained beforehand through calibration and represents the preload output value corresponding to a unit control command value, calculated according to formula U. target =ΔF cmd / η Calculate the target control command value U required to achieve the absolute value of the preload compensation force. target ; The value of the control command at the current moment is denoted as U. current According to formula U step =(U target -U current ) / M calculates the instruction change step size U for each control cycle. step , where M is the preset total number of transition cycles, and M is a positive integer; According to formula U j =U current +j·Ustep generates the instruction value for the j-th control cycle, where j=1,2,…,M, when U j When the calculation result exceeds the driver's maximum allowed instruction value, U will... j Limit to the maximum allowed instruction value; The generated M command values ​​are arranged in chronological order to form a joint preload control command sequence.

[0014] In one embodiment of the present invention, the embedded controller outputs to the joint actuator to apply a real-time joint preload compensation value, including: The embedded controller reads each instruction value U in the control instruction sequence of claim 8 sequentially according to a preset control cycle. j ; The currently read instruction value U j After being converted into an analog voltage signal or a pulse width modulation signal, it is output to the joint driver; The joint actuator drives the preload adjustment mechanism to change the amount of pressure on the internal transmission components of the joint based on the received signal, thereby changing the preload applied to the joint in real time. After all the command values ​​in the control command sequence have been executed, the preload stabilizes at the target value after the compensation value has been applied.

[0015] In one embodiment of the present invention, after applying the real-time joint preload compensation value, the following verification and cyclic adjustment steps are further included: After applying the preload compensation value, wait for a preset stabilization period until the preload adjustment actuator completes its action and the joint operation becomes stable. Re-execute the real-time acquisition step and the Kalman filter dynamic estimation step to obtain a new estimated value of the wear state of the transmission component after applying the compensation value; Extract the amount of joint preload reduction after applying compensation from the new wear condition estimate; The amount of preload reduction after applying the compensation value is compared with the preset allowable reduction threshold: If the attenuation is less than or equal to the allowable attenuation threshold, the preload adjustment is deemed qualified, and the adjustment process ends. If the attenuation amount is still greater than the allowable attenuation threshold, repeat all steps from separating the wear acceleration factor to applying the preload compensation value until the preload attenuation amount meets the requirements.

[0016] The technical solution of the present invention has the following advantages compared with the prior art: The adaptive control method for heavy-duty joint preload based on the wear state of transmission components described in this invention collects load data and vibration signals from the industrial robot joints in real time, identifies initial wear characteristic indicators using a support vector machine algorithm, and dynamically estimates the real-time wear state of the transmission components using a Kalman filter algorithm. This allows for precise perception of the accelerated wear effect caused by load changes under heavy-duty conditions. Therefore, the preload compensation coefficient can be adaptively determined based on the cumulative wear increment, and a control command sequence can be generated, thereby achieving real-time and accurate compensation of the joint preload. This method effectively solves the problem of preload adjustment lagging behind the actual wear process in traditional solutions, significantly improves the rigidity and transmission accuracy of heavy-duty industrial robot joints during long-term operation, avoids the accumulation of processing errors and equipment failures caused by wear, and ensures the stability of production efficiency and product quality. Attached Figure Description

[0017] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart of the steps of the adaptive control method for preload force of heavy-duty joints based on the wear state of transmission components in this invention; Figure 2 This is a flowchart illustrating the steps of identifying initial wear characteristic indicators using a support vector machine in this invention. Figure 3 This is a flowchart of the steps involved in dynamic estimation using Kalman filtering in this invention; Figure 4 This is a flowchart of the steps for separating wear acceleration factors in this invention; Figure 5 This is a flowchart of the steps for generating a control instruction sequence according to the present invention; Figure 6 This is a flowchart of the verification and cyclical adjustment steps of the present invention. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0019] Reference Figure 1As shown, this invention proposes an adaptive control method for preload force of heavy-duty joints based on the wear state of transmission components. The core technical solution of this method consists of the following steps: First, load data and vibration signals of industrial robot joints under heavy-duty conditions are collected in real time, and the initial wear characteristic index of the transmission components inside the joint is identified using a support vector machine algorithm. The principle of this step is that when the transmission component wears, the microscopic geometry of its contact surface changes, thereby altering the vibration spectrum characteristics and load transmission characteristics of the joint during operation. As a supervised learning algorithm, the support vector machine can effectively separate the feature patterns related to the wear state from high-dimensional vibration and load data, thereby obtaining a quantified initial wear index, solving the problem that the wear state is difficult to measure directly.

[0020] Building upon this foundation, this method utilizes the Kalman filter algorithm to fuse initial wear characteristic indicators with accumulated running time, dynamically estimating the real-time wear state of the transmission components. The Kalman filter's characteristic lies in its ability to recursively process noisy observation data. Even with measurement noise in the vibration signals and load data, it can output the optimal estimate of the true wear state through an iterative mechanism of state prediction and observation updates, obtaining an estimate that includes the reduction in joint preload caused by transmission component wear. This step cleverly solves the technical bottleneck of wear states being time-varying and not directly observable.

[0021] Furthermore, this method separates the wear acceleration factor caused by load changes from the wear state estimate and numerically integrates this acceleration factor with respect to operating time to obtain the cumulative wear increment of the transmission component. The core insight here is that the wear acceleration factor essentially reflects the amplification effect of the current load size on the basic wear rate. By separating this factor and integrating it, the contribution of load fluctuations to the entire wear process can be mathematically accumulated, thereby obtaining a physically meaningful cumulative wear increment.

[0022] Subsequently, the method determines the compensation coefficient based on the cumulative wear increment and a preset wear-preload ratio. Since there is a pre-calibrated linear or quasi-linear relationship between transmission component wear and preload loss, this ratio transforms the microscopic wear amount into the macroscopic preload adjustment range, thus bridging the information gap between the transmission component and the joint assembly.

[0023] Finally, a control command sequence is generated based on the compensation coefficient and output by the embedded controller to the joint actuator to apply a real-time preload compensation value, thus completing the closed-loop control from wear sensing to preload adjustment.

[0024] Through the above technical solution, the present invention achieves the following beneficial effects, including: First, by employing a two-stage wear estimation strategy that combines support vector machine and Kalman filtering, this method can accurately and in real time reconstruct the true wear state of transmission components from noisy sensor data, thus solving the fundamental problem of difficulty in predicting wear rate under dynamic load changes.

[0025] Second, by introducing a wear acceleration factor and calculating the cumulative wear increment using numerical integration, this method can fully encode the load fluctuation process into the wear contribution, so that the preload adjustment no longer depends on a fixed time period, but truly responds to the actual wear process, thus maintaining accurate compensation even under complex working conditions of ultra-high load impact and low load alternation.

[0026] Third, by establishing a pre-defined wear-preload ratio, this method creates a clear transmission chain from wear of transmission components to preload compensation of joints, overcoming the control mismatch problem caused by physical separation and making preload adjustment based on evidence.

[0027] Overall, this invention enables heavy-duty industrial robot joints to autonomously sense wear and adaptively adjust preload during long-term service, thereby maintaining high precision and high stability, significantly extending equipment life and reducing maintenance costs.

[0028] Under heavy-load conditions, identifying the wear condition of joint transmission components in industrial robots is crucial for adaptive preload control. However, accurately and effectively extracting characteristic indicators that characterize the initial wear degree of transmission components from complex load data and vibration signals presents a challenge for achieving subsequent accurate wear condition estimation and preload compensation.

[0029] To address this, this application proposes a more specific and accurate method for identifying initial wear characteristic indicators. This method first aligns real-time acquired load data and vibration signals according to time to form a two-dimensional input sample, where each time point sample contains a load value and a corresponding vibration amplitude. This processing aims to effectively integrate multi-source data from different sensors but related to the same physical event, ensuring a clear correlation between the load value and the vibration amplitude, thereby forming a two-dimensional feature vector that comprehensively describes the instantaneous working state of the joint. For example, high-precision timestamps can be used to synchronize sensor data, or signal processing techniques (such as resampling and interpolation) can be used to unify data with different sampling rates onto a common time reference.

[0030] Building upon this, this application pre-trains a Support Vector Machine (SVM) classifier using sample data labeled with wear levels. During training, load values ​​and vibration amplitudes are used as input features, and the corresponding wear levels are used as classification labels to obtain the classification hyperplane parameters. As a supervised learning model, the SVM establishes a mapping relationship between load-vibration features and the wear levels of transmission components by acquiring load-vibration response data of transmission components with different wear levels in the laboratory or actual operation and labeling them with wear levels. This training data is used to optimize the parameters of the SVM model, enabling it to learn one or more optimal classification hyperplanes to maximize the margin between samples of different wear levels. These classification hyperplane parameters are the core of the SVM model; they define how new, unknown load-vibration samples are classified into different wear levels.

[0031] Subsequently, the real-time collected input samples are fed into the trained support vector machine classifier to calculate the signed distance from the sample point to the classification hyperplane, and this distance value is used as the initial wear feature index. The larger the distance value, the greater the deviation of the transmission component from its healthy state. In the support vector machine, the distance from the sample point to the classification hyperplane can intuitively reflect the confidence level or deviation of the sample point from a certain category. Using this distance value as the initial wear feature index has the advantage of being a continuous numerical value, which can characterize the degree of wear more finely than discrete wear levels. The larger the distance value, the greater the difference between the current operating state of the transmission component and its healthy state, thus more accurately reflecting the severity of transmission component wear.

[0032] In some embodiments described above, a support vector machine algorithm is proposed to identify the initial wear characteristic index of the transmission components inside the joint, and to estimate the subsequent wear state based on this index. However, under actual heavy-load conditions of industrial robots, the wear of transmission components is a complex process that is dynamically changing and affected by multiple factors. Furthermore, vibration signals and load data collected by sensors inevitably introduce measurement noise. Relying solely on the initial wear characteristic index at a single moment is insufficient to accurately and robustly reflect the real-time wear state of the transmission components, especially in the presence of measurement noise and system uncertainties. Directly using this index for subsequent preload adjustment may lead to insufficient accuracy or over-adjustment, affecting the stability and reliability of the control system.

[0033] For reference Figure 3 As shown, this application further proposes to use the Kalman filter algorithm to dynamically estimate the real-time wear state of the transmission component at the current moment. Specifically, it includes: establishing the state transition equation and observation equation of the wear state of the transmission component; executing the state prediction step and the state update step in sequence; and outputting the posterior estimate as the estimated value of the wear state of the transmission component.

[0034] The Kalman filter algorithm is a highly efficient recursive filter used to estimate the state of dynamic systems. It provides an optimal estimate of the system state in the presence of noise by fusing system model predictions and sensor observations. In this application, the Kalman filter algorithm is used to dynamically track the wear state of transmission components to overcome the uncertainty and noise effects of single observation data.

[0035] Establishing the state transition equation and observation equation for the wear state of transmission components is the foundation of the Kalman filter algorithm. The state transition equation describes the change of the system state over time, i.e., the intrinsic evolution process of transmission component wear. The observation equation, on the other hand, correlates the actual system state with measurable observations, reflecting how wear information is obtained from sensor data.

[0036] The state transition equation is: x k =x k-1 +v0·△t+w k This equation describes the dynamic evolution of the wear state of the transmission components. Where x k This represents the estimated wear state of the transmission component at time k, which is derived from the wear state x at the previous time. k-1 In addition, the wear increment caused by the preset nominal wear rate v0 within the sampling time interval Δt, and the process noise w representing system uncertainty and unmodeled dynamics. k This is jointly determined. The equation provides a mechanism for predicting wear trends based on a physical model.

[0037] The observation equation is z k =x k +n k This equation establishes the relationship between observed wear values ​​and actual wear conditions. Where z k This represents the wear observation value obtained from the initial wear characteristic index at time k, which is considered to be the actual wear state x. k Add observation noise n k Observation noise n k This represents the sensor measurement error and the uncertainty introduced during the conversion from the initial wear characteristic index.

[0038] The core iterative process of the Kalman filter algorithm is to sequentially execute the state prediction step and the state update step. The state prediction step uses the system model (state transition equation) to predict the wear state at the current moment, while the state update step uses actual observation data (observation equation) to correct the prediction result, thereby obtaining a more accurate estimate.

[0039] The state prediction step predicts the prior estimate of the wear state and its prior covariance at the current moment based on the previous wear state estimate and nominal wear rate. In this step, the system uses the posterior estimate of the wear state from the previous moment as a starting point, combined with the nominal wear rate v0 of the transmission component and the sampling time interval Δt, to predict the wear state at the current moment through a state transition equation. Simultaneously, it also predicts the uncertainty of this prior estimate, i.e., the prior covariance, which reflects the confidence level of the prediction result.

[0040] The state update step calculates the Kalman gain using the wear observations at the current moment, uses the weighted sum of the prior estimate and the observations as the posterior estimate of the wear state at the current moment, and updates the posterior estimate covariance. In this step, the system first calculates the Kalman gain based on the covariance of the prior estimate and the covariance of the observation noise. The Kalman gain determines the weights of the predicted and observed values ​​in forming the final estimate. Then, the prior estimate is compared with the wear observation z. k A weighted fusion is performed to obtain the posterior estimate of the wear state at the current moment, which is the optimal combination of prediction and observation. Simultaneously, the posterior estimate covariance is updated to reflect the uncertainty of the new estimate.

[0041] The posterior estimate is output as the wear state estimate of the transmission component. After two iterative steps of state prediction and state update, the final posterior estimate output by the Kalman filter algorithm is the optimal estimate that integrates the system dynamic model and real-time observation data. It represents the most accurate wear state of the transmission component at the current moment.

[0042] Specifically, when the abstract initial wear characteristic index output by the support vector machine is directly applied to the observation equation of the Kalman filter, there may be a lack of direct quantitative relationship between the index and the actual physical wear, which limits the estimation accuracy of the Kalman filter and makes it difficult to accurately reflect the true wear degree of the transmission components.

[0043] In response, this application further proposes a specific method for converting initial wear characteristic indicators into wear observation values, including: pre-calibrating unworn transmission components, collecting multiple sets of load-vibration samples under healthy conditions and inputting them into a support vector machine classifier, calculating the average distance from healthy sample points to the classification hyperplane, denoted as d0; denoting the distance from real-time collected sample points to the classification hyperplane as d; and calculating the wear observation value z according to the formula z=(d-d0)·k, where k is the wear depth coefficient corresponding to the pre-calibrated unit distance.

[0044] Specifically, unworn transmission components are pre-calibrated, and multiple sets of load-vibration samples under healthy conditions are collected and input into a support vector machine (SVM) classifier. The average distance from the healthy sample points to the classification hyperplane is calculated, denoted as d0, to establish a healthy baseline. A series of tests are conducted on brand-new, unworn transmission components to obtain data under different load and vibration conditions. The distances from these healthy sample points to the classification hyperplane are then calculated using a pre-trained SVM classifier. The average distance d0 represents the "normal" distance when the transmission component is in a healthy state, serving as a reference point for subsequent wear observation calculations and quantifying the deviation of real-time wear from the healthy state. For example, before the transmission component is put into use, or in a laboratory environment, a series of load tests can be conducted on multiple brand-new transmission components of the same model, simultaneously collecting their load data and vibration signals. This data is input into a pre-trained SVM classifier, which outputs the signed distance from each sample point to the classification hyperplane. These distance values ​​are then statistically averaged to obtain d0.

[0045] The distance d from the real-time acquired sample point to the classification hyperplane is denoted as . This distance d is a quantitative representation of the real-time monitored state of the transmission component. It directly originates from the output of the support vector machine classifier mentioned above and reflects the relative position between the current load-vibration characteristics of the transmission component and the health / wear classification hyperplane. During the actual operation of the industrial robot, the system continuously acquires joint load data and vibration signals in real time. This real-time data is organized into two-dimensional input samples and input into a pre-trained support vector machine classifier. The classifier calculates the signed distance d from the current real-time sample point to the hyperplane based on its internal classification hyperplane parameters.

[0046] The wear observation value z is calculated using the formula z=(d-d0)·k, where k is a pre-calibrated wear depth coefficient corresponding to a unit distance. This step is the core of converting the abstract distance difference into a wear observation value with physical meaning. By subtracting the healthy baseline d0, the distance deviation of the current state relative to the healthy state can be obtained. Multiplying this by a calibration coefficient k converts this distance deviation into physical units related to the actual wear depth or wear amount, so that the wear observation value z can directly reflect the wear degree of the transmission component and can be used as the observation input for Kalman filtering. The calibration of the coefficient k usually needs to be completed through experiments or simulations. For example, accelerated wear experiments can be conducted on the transmission component, and load-vibration data can be collected at different wear stages (e.g., by measuring the actual wear depth or mass loss), and the corresponding d values ​​can be calculated. Then, the (d-d0) values ​​under different wear degrees are regressed with the actual wear depth to determine the wear depth coefficient k corresponding to the unit distance difference. In real-time operation, once d and d0 have been obtained and k has been pre-calibrated, the system can directly apply this formula to calculate the current wear observation value z.

[0047] Under actual heavy-load conditions, the wear of transmission components is not only related to the cumulative amount of operating time, but is also significantly affected by load changes. If the wear acceleration effect caused by load changes cannot be accurately distinguished and quantified from the wear state estimate, it may lead to inaccurate judgment of the actual wear degree of transmission components, thereby affecting the accuracy and effectiveness of preload compensation and making it impossible to achieve true adaptive control.

[0048] Reference Figure 4 As shown, this application further proposes to separate the wear acceleration factor caused by load changes from the estimated wear state of transmission components, specifically including: denoting the estimated wear state of the transmission components at the current moment as x. k The estimated wear condition at the previous moment is denoted as x. k-1 The sampling time interval is denoted as Δt, and calculated according to the formula r k = (x k -x k-1 )·△t, calculate the actual wear rate r at the current moment. k The load data at the current moment is denoted as L. k The preset reference load value is denoted as L0, and calculated according to formula β. k =L k / L0 calculates the load ratio β k Where the reference load value L0 is the standard load corresponding to the transmission component operating at the nominal wear rate; according to formula A k =r k / v0, calculate the total wear acceleration ratio Ak, where v0 is the preset nominal wear rate; separate the acceleration factor caused by the load from the total wear acceleration ratio, the separation method is: when βk When >1, according to formula α k =(β k -1)·λ calculates the load acceleration factor αk; when β k When ≤1, take α. k =0, where λ is a preset load sensitivity coefficient, representing the additional increase in wear rate caused by a change in unit load ratio when the load exceeds the reference value; the calculated α k Output as a wear acceleration factor caused by load changes.

[0049] To achieve the above separation, firstly, the estimated wear state of the transmission component at the current moment is used... k Compared with the previous moment's wear state estimate x k-1 Dividing the difference by the sampling time interval Δt, the actual wear rate r at the current moment is calculated. k The actual wear rate r k This reflects the wear trend of the transmission components over a short period of time, providing a quantitative basis for subsequent analysis. Secondly, to assess the impact of load on wear, the load data L at the current moment is obtained. k The load ratio β is then calculated by comparing it with the preset baseline load value L0. k The reference load value L0 is the standard load corresponding to the transmission component operating at the nominal wear rate. It is typically determined through experimental calibration or design specifications and represents the load level under normal operating conditions. The load ratio β... k It can intuitively reflect the degree of deviation of the current load from the standard load. Based on this, the actual wear rate r at the current moment is calculated. k Divide by the preset nominal wear rate v0 to calculate the total wear acceleration ratio A. k The nominal wear rate v0 is the theoretical wear rate of the transmission component under standard operating conditions without considering additional acceleration effects, and the total wear acceleration ratio A is... k This reflects the total effect of wear acceleration caused by all factors. Finally, the acceleration factor α caused by the load is separated from the total wear acceleration ratio. k Specifically, when the load ratio is β k When it is greater than 1, that is, the current load L k When the load exceeds the reference load L0, it is considered that the load accelerates wear; at this point, the load acceleration factor α is... k via (β) k -1)·λ is calculated, where λ is a preset load sensitivity coefficient, characterizing the additional increase in wear rate caused by a unit change in load ratio. This coefficient is usually obtained through experimental calibration. When the load ratio β k When less than or equal to 1, i.e., the current load L kIf the load does not exceed the baseline load L0, it is considered that the load does not cause additional wear acceleration, and the load acceleration factor α is used in this case. k Set it to 0. Finally, calculate α. k Output as a wear acceleration factor caused by load changes.

[0050] To more comprehensively and accurately assess the cumulative wear of transmission components, this application further proposes to numerically integrate the wear acceleration factor relative to operating time to obtain the cumulative wear increment of the transmission components, specifically including: First, the wear acceleration factor is discrete-time integraled according to the sampling time interval, as shown in the formula: ; Calculate the cumulative wear increment D from the initial time to the current time. acc (t), where N is the total number of sampling steps from the initial time to the current time, and α i Let v0 be the wear acceleration factor at the i-th sampling time, v0 be the preset nominal wear rate, and Δt be the sampling time interval.

[0051] This step aims to quantify the additional wear accumulation caused by non-nominal operating conditions (such as heavy loads). Wear acceleration factor α i It is derived from the wear condition estimate based on load changes, reflecting the ratio of the wear rate to the nominal rate at a specific moment. By multiplying this acceleration factor by the nominal wear rate v0 in each sampling time interval Δt and summing the results of all sampling steps N, the cumulative wear depth D accelerated by the change in operating conditions can be accurately calculated. acc (t). This discrete-time cumulative integration method can effectively transform the instantaneous wear acceleration effect into a quantifiable cumulative wear amount, ensuring accurate capture of the nonlinear wear process.

[0052] Secondly, follow formula D. total = D acc (t) + v0·t, calculate the cumulative wear increment D of the transmission component. total , where v0·t is the cumulative wear caused by nominal wear, and t is the total running time from the initial moment to the current moment.

[0053] This step will determine the accelerated wear amount D caused by changes in operating conditions. acc The final cumulative wear increment D of the transmission component is obtained by superimposing (t) with the inherent wear amount v0·t accumulated over time under nominal operating conditions. totalThe term v0·t represents the baseline wear that the transmission component will experience over the total operating time t, even without additional acceleration factors. The nominal wear rate v0 is typically predetermined based on the material properties of the transmission component, design parameters, and experimental data under standard operating conditions. By adding these two wear parameters, a comprehensive assessment of the actual wear state of the transmission component is ensured, taking into account both the baseline wear from continuous operation and the accelerated wear caused by special operating conditions such as heavy loads.

[0054] In some embodiments described above, the real-time wear state of the transmission components is dynamically estimated and the cumulative wear increment is calculated. However, in practical applications, how to effectively convert this cumulative wear increment into a specific adjustment amount for the joint preload to achieve adaptive compensation of the preload remains a problem to be solved. Simply obtaining the wear amount cannot directly guide the precise adjustment of the preload; a quantitative relationship between the wear amount and the preload adjustment needs to be established.

[0055] In response, this application further proposes determining the compensation coefficient required for preload adjustment based on a preset wear-preload ratio, specifically including: obtaining the wear-preload ratio coefficient k, which has been experimentally calibrated beforehand. wp The proportionality coefficient represents the joint preload loss value corresponding to the unit wear depth of the transmission component; according to the formula ΔF=D total ·k wp Calculate the preload loss ΔF caused by the current cumulative wear; obtain the initial preload value F0 of the joint, and calculate the compensation coefficient C required for preload adjustment according to the formula C=-ΔF / F0, where the negative sign indicates that when the preload loss ΔF is positive, the preload needs to be increased to make up for the loss, and the compensation coefficient C is a dimensionless ratio; when the compensation coefficient C is negative, it means that the current preload needs to be increased to (1-C) times the initial preload; when the compensation coefficient C is zero, it means that no adjustment is needed.

[0056] Among them, the wear-preload proportionality coefficient k wp This value, ΔF, is obtained through prior experimental calibration and represents the preload loss per unit wear depth of the transmission component. In practice, transmission components with varying degrees of wear can be tested in a laboratory environment to measure preload decay under specific loads, thus establishing a mapping relationship between wear depth and preload loss. For example, a preload can be applied to a new transmission component and measured. Then, an accelerated wear experiment can be conducted to wear the component to a certain depth, and the preload can be measured again. This proportionality coefficient can be obtained by fitting multiple experimental data. This coefficient is the key bridge connecting the physical wear of the transmission component with the system performance (preload) decay. The calculation of the preload loss value ΔF is based on the aforementioned cumulative wear increment. Dtotal Wear-preload proportionality coefficient k wpThis was carried out. Specifically, it was achieved by increasing the cumulative wear increment D. tota l multiplied by the proportionality coefficient k wp This calculation directly quantifies the actual attenuation of joint preload due to wear of transmission components. This step transforms the abstract wear into concrete mechanical loss, providing a direct basis for subsequent preload compensation. The initial preload value F0 of the joint refers to the standard preload set when the transmission components are in a healthy state (no wear or wear within an acceptable range). This value is usually calibrated and set when the robot leaves the factory or after a major overhaul, and is the benchmark preload for normal joint operation. Obtaining this value allows for comparison and quantification with a relatively stable benchmark when calculating the compensation coefficient. The formula for calculating the compensation coefficient C is C = -ΔF / F0. The negative sign is introduced to clearly indicate the direction of preload adjustment: when the preload loss ΔF is positive (i.e., the preload has indeed attenuated), it is necessary to increase the preload to compensate for this loss, and the calculated compensation coefficient C will be negative. The compensation coefficient C is a dimensionless ratio, representing the proportion of preload that needs to be adjusted relative to the initial preload F0. This dimensionless processing makes the compensation coefficient more universal, facilitating its application in joints of different specifications or models. When the calculated compensation coefficient C is negative, it indicates that the current preload of the joint has decayed and compensation is needed. Specifically, the current preload needs to be increased to (1-C) times the initial preload F0. For example, if C is -0.05, it means the preload needs to be increased to 1.05 times F0. When the compensation coefficient C is zero, it indicates that the current cumulative wear has not yet led to a significant loss of preload, or the loss is within an acceptable range, therefore no adjustment of the preload is necessary. This explicit judgment mechanism ensures that preload adjustment is only performed when necessary, avoiding unnecessary intervention.

[0057] In some of the embodiments described above in this application, the compensation coefficient required for adjusting the joint preload can be determined by dynamically estimating the wear state of the transmission components and calculating the cumulative wear increment. However, if the joint preload is adjusted directly based on this compensation coefficient in one go, it may cause a sudden change in the preload, which may lead to problems such as mechanical shock, system oscillation, or control instability. Especially under heavy-load conditions, such a sudden change will have an adverse effect on the smoothness and accuracy of the industrial robot's operation.

[0058] Reference Figure 5 As shown, this application further proposes a method for generating a joint preload control command sequence based on a compensation coefficient, specifically including: First, multiply the absolute value of the calculated compensation coefficient C by the initial preload value F0 of the joint to calculate the absolute value of the preload that needs to be compensated, ΔF. cmdThis step aims to convert the dimensionless compensation coefficient C into the actual amount of preload that needs to be compensated. For example, if the compensation coefficient C is -0.05, it means that a 5% increase in preload is required, then the absolute value of the preload to be compensated is ΔF. cmd It is 0.05 multiplied by the initial preload value F0.

[0059] Next, the force-command conversion coefficient η of the joint actuator is obtained. This force-command conversion coefficient η is obtained in advance through calibration and represents the preload output value corresponding to a unit control command value. Then, according to formula U... target =ΔF cmd / η Calculate the target control command value U required to achieve the absolute value of the preload compensation force. target Joint actuators typically control preload adjustment mechanisms by receiving electrical signals. The purpose of this step is to calculate the absolute value ΔF of the preload to be compensated. cmd Converted into an electrical signal value U that the driver can recognize and execute. target The force-command conversion factor η is usually obtained through experimental calibration, for example, by measuring the preload output of the drive under different control command inputs and establishing the relationship between the two.

[0060] Subsequently, the control command value U at the current moment is obtained. current And according to formula U step =(U target -U current ) / M calculates the instruction change step size U for each control cycle. step Where M is the preset total number of transition cycles, and M is a positive integer. To avoid abrupt changes during the preload adjustment process, the total command change (from the current command U) needs to be adjusted. current To target instruction U target The change is distributed across multiple control cycles and completed gradually. The instruction change step size U step The increment or decrement of the command value within each control cycle is defined to achieve a smooth transition. The setting of M should take into account the system's response speed, mechanical inertia, and stability requirements.

[0061] Based on this, according to formula U j =U current +j·U step Generate the instruction value for the j-th control cycle, where j = 1, 2, ..., M. During the generation process, when U... j When the calculation result exceeds the driver's maximum allowed instruction value, U will... j The limit is set to the maximum allowed instruction value. This step is based on the calculated instruction change step size U. stepThis process generates a series of discrete command values, which form a gradual control sequence. By applying these commands one by one, the preload will smoothly transition from the current value to the target value. Simultaneously, to protect the drive and mechanical structure, the generated command values ​​are limited to ensure they do not exceed the drive's maximum permissible command value, preventing overload or damage.

[0062] Finally, the generated M instruction values ​​are arranged in chronological order to form a joint preload control instruction sequence. This sequence is the core of achieving smooth and gradual adjustment of the preload, and will then be read by the embedded controller and output one by one to the joint actuator.

[0063] Furthermore, this application proposes an embedded controller outputting a real-time joint preload compensation value to a joint actuator, comprising: the embedded controller sequentially reading each instruction value Uj in the aforementioned control instruction sequence according to a preset control cycle; converting the currently read instruction value Uj into an analog voltage signal or a pulse width modulation signal and outputting it to the joint actuator; the joint actuator driving the preload adjustment actuator to change the amount of pressure on the internal transmission components of the joint according to the received signal, thereby changing the preload applied to the joint in real time; after all instruction values ​​in the control instruction sequence have been executed, the preload stabilizes at the target value after the compensation value is applied.

[0064] Specifically, the embedded controller is the core hardware unit for implementing this method, and its function is to receive and process the control command sequence generated by the upper-level algorithm. This controller typically employs a high-performance microprocessor or digital signal processor, possessing a real-time operating system or bare-metal programming capability to ensure timely response to control tasks. The preset control cycle refers to the time interval between the controller executing one control loop, such as 1 millisecond, 10 milliseconds, or 100 milliseconds. Its selection must comprehensively consider the system's dynamic response characteristics, control accuracy requirements, and computational resource limitations. The controller accesses and sequentially reads each instruction value U from the control command sequence through internal memory or external storage media. j This ensures the sequentiality and completeness of instructions, providing accurate input for subsequent signal conversion and driving.

[0065] In order for the joint actuator to understand and execute control commands, the embedded controller needs to convert the command values ​​U in digital form. jThis is converted into a physical signal that the driver can recognize. This conversion typically involves a digital-to-analog converter (DAC) that converts digital commands into analog voltage signals, or a pulse-width modulation (PWM) module that generates pulse-width modulated signals. Analog voltage signals can be directly used to control certain types of servo amplifiers to adjust their output current or voltage; while PWM signals control the motor driver by changing the duty cycle of the pulses, thereby adjusting the motor's speed, torque, or position. This conversion acts as a bridge between digital control logic and physical actuators, ensuring that control commands are effectively transmitted.

[0066] The joint actuator is the intermediate link that receives the output signal from the embedded controller and converts it into mechanical action. Based on the received analog voltage signal or pulse width modulation signal, it controls the preload adjustment actuator. The preload adjustment actuator can be an electric lead screw mechanism, hydraulic cylinder, pneumatic cylinder, piezoelectric actuator, or electromagnetic actuator, etc., and its function is to directly or indirectly change the degree of clamping on the internal transmission components of the joint. For example, an electric lead screw mechanism drives the lead screw to rotate via a motor, which in turn moves the clamping component, thereby increasing or decreasing the clamping amount of the transmission component. In this way, the joint actuator and the actuator work together to adjust the preload applied to the joint in real time and with precision to compensate for preload attenuation caused by wear.

[0067] This step describes the final state of the preload compensation process. Once the embedded controller outputs and executes the entire sequence of control commands according to the preset control cycle, it means the system has completed the gradual adjustment of the preload. At this point, the clamping amount of the internal transmission components of the joint has reached the preset compensation level, allowing the joint's preload to recover to or approach its target value. Preload stability means that, without new external disturbances or wear changes, the joint preload will remain near the compensated target value, thereby ensuring the joint can operate continuously, stably, and efficiently under heavy load conditions, and effectively suppressing further vibration and wear.

[0068] In practice, due to the uncertainty of the system model, the response lag of the actuator, and the interference of environmental factors, a one-time preload compensation may not be able to completely eliminate the preload decay caused by wear, resulting in residual deviations in the joint preload after compensation, which affects the long-term stable operation and performance accuracy of the industrial robot.

[0069] Reference Figure 6As shown, this application further proposes a verification and cyclic adjustment step after applying the real-time joint preload compensation value. Specifically, after applying the preload compensation value, the system waits for a preset stabilization period to ensure that the preload adjustment actuator can fully complete its action and that the joint's operating state tends to stabilize. This preset stabilization period is crucial; it allows the mechanical system (especially systems involving force adjustment) to overcome inertia and response delay, ensuring that the system has reached a new steady state when evaluating the compensation effect, thereby avoiding misjudgments caused by transient responses. The specific duration of this period can be experimentally calibrated or set according to the dynamic characteristics of the joint actuator.

[0070] Once the system stabilizes, the real-time data acquisition and Kalman filter dynamic estimation steps will be re-executed. This means the system will again acquire load data and vibration signals of the industrial robot joints under heavy-duty conditions, and use a support vector machine algorithm to identify the initial wear characteristics of the internal transmission components. Subsequently, based on these new initial wear characteristics and combined with the accumulated runtime, the Kalman filter algorithm is used to dynamically estimate the real-time wear state of the transmission components at the current moment, thereby obtaining a new estimated value of the transmission component wear state after applying compensation. This re-evaluation step is crucial, as it provides the latest and most accurate feedback information on the actual wear state of the joint after compensation.

[0071] Next, the amount of preload reduction after applying compensation is extracted from the new wear condition estimate. Since the wear condition estimate of the transmission component output by the Kalman filter algorithm itself includes the amount of preload reduction caused by wear, this step directly separates this reduction component from the updated estimate as a direct indicator of the current preload loss.

[0072] Subsequently, the amount of preload reduction after applying the compensation value is compared with a preset allowable reduction threshold. This "preset allowable reduction threshold" is pre-set based on the industrial robot's design requirements, performance indicators, and safety margins, and represents the maximum acceptable amount of preload reduction for the system. Through this comparison, the system can objectively determine whether the current joint preload has recovered to an acceptable level.

[0073] If the attenuation is less than or equal to the allowable attenuation threshold, the preload adjustment is deemed qualified, and the adjustment process ends. This indicates that the current preload compensation has achieved the expected effect, the joint preload has returned to the normal working range, and no further adjustment is required, thus avoiding unnecessary resource consumption and system disturbance.

[0074] However, if the attenuation amount still exceeds the allowable attenuation threshold, the system will repeat all steps from separating the wear acceleration factor to applying the preload compensation value until the preload attenuation amount meets the requirements. This constitutes a closed-loop adaptive adjustment mechanism. When a single compensation is insufficient, the system will not simply stop, but will recalculate the required compensation based on the latest wear state and apply it again, repeating this cycle until the joint preload attenuation amount reaches the preset qualified standard.

[0075] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. An adaptive control method for preload of heavy-duty joints based on the wear state of transmission components, characterized in that, Includes the following steps: The load data and vibration signals of the joints of industrial robots under heavy load conditions are collected in real time. Based on the load data and vibration signals, the initial wear characteristics of the transmission components inside the joints are identified using the support vector machine algorithm. Based on the initial wear characteristic index and combined with the cumulative running time, the Kalman filter algorithm is used to dynamically estimate the real-time wear state of the transmission component at the current moment, and the estimated value of the wear state of the transmission component is obtained. The estimated value of the wear state of the transmission component includes the amount of joint preload reduction caused by the wear of the transmission component. The wear acceleration factor caused by load change is separated from the wear state estimate of the transmission component, and the wear acceleration factor is numerically integrated with respect to the running time to obtain the cumulative wear increment of the transmission component. Based on the cumulative wear increment, and according to the preset wear-preload ratio, determine the compensation coefficient required for preload adjustment; A sequence of joint preload control commands is generated based on the compensation coefficient and output by the embedded controller to the joint actuator to apply real-time joint preload compensation values.

2. The adaptive control method for preload of heavy-duty joints based on wear state of transmission components according to claim 1, characterized in that: The initial wear characteristics of the internal transmission components of the joint were identified using a support vector machine algorithm, including: The collected load data and vibration signal are aligned with time to form a two-dimensional input sample. The sample at each time point contains a load value and the corresponding vibration amplitude. A support vector machine classifier is trained in advance using sample data with pre-labeled wear levels. During training, the load value and vibration amplitude are used as input features, and the corresponding wear level is used as the classification label to obtain the classification hyperplane parameters. The real-time collected input samples are substituted into the trained support vector machine classifier, and the signed distance from the sample point to the classification hyperplane is calculated. The distance value is used as the initial wear feature index. The larger the distance value, the greater the degree to which the transmission component deviates from its healthy state.

3. The adaptive control method for preload of heavy-duty joints based on wear state of transmission components according to claim 1, characterized in that: The Kalman filter algorithm is used to dynamically estimate the real-time wear state of the transmission components at the current moment, including: Establish the state transition equation and observation equation for the wear state of the transmission components: The state transition equation is: x k =x k-1 +v0·△t+w k ; Where: x k Let v0 be the estimated wear state of the transmission component at time k, v0 be the preset nominal wear rate, Δt be the sampling time interval, and w be the mean wear rate. k This is process noise; The observation equation is z k =x k +n k , where z k Let n be the wear observation value obtained from the initial wear characteristic index at time k. k To observe noise; Perform the state prediction step and the state update step sequentially: The state prediction step predicts the prior estimate of the wear state and the prior estimate covariance at the current moment based on the wear state estimate and nominal wear rate at the previous moment. The state update step uses the wear observations at the current moment to calculate the Kalman gain, takes the weighted result of the prior estimate and the observations as the posterior estimate of the wear state at the current moment, and updates the posterior estimate covariance. The posterior estimate is output as the wear state estimate of the transmission component.

4. The adaptive control method for preload of heavy-duty joints based on wear state of transmission components according to claim 1, characterized in that: The specific method for converting initial wear characteristic indicators into wear observation values ​​is as follows: Beforehand, the transmission components without wear are calibrated, and multiple sets of load-vibration samples under healthy conditions are collected and input into the support vector machine classifier. The average distance from the healthy sample points to the classification hyperplane is calculated and denoted as d0. The distance from the real-time sample point to the classification hyperplane is denoted as d; the wear observation value z is calculated according to the formula z=(d-d0)·k, where k is the wear depth coefficient corresponding to the pre-calibrated unit distance.

5. The adaptive control method for preload of heavy-duty joints based on wear state of transmission components according to claim 1, characterized in that: The wear acceleration factor caused by load variation is separated from the wear state estimate of transmission components, including: Let x be the estimated wear state of the transmission component at the current moment. k The estimated wear condition at the previous moment is denoted as x. k-1 The sampling time interval is denoted as Δt, and calculated according to the formula r k = (x k -x k-1 )·△t, calculate the actual wear rate r at the current moment. k ; Let L be the load data at the current moment. k The preset reference load value is denoted as L0, and calculated according to formula β. k =L k / L0 calculates the load ratio β k The reference load value L0 is the standard load corresponding to the transmission component when it is running at the nominal wear rate; According to formula A k =r k / v0, calculate the total wear acceleration ratio A k v0 is the preset nominal wear rate; The acceleration factor caused by the load is separated from the total wear acceleration ratio. The separation method is as follows: When β k When >1, according to formula α k =(β k -1)·λ calculates the load acceleration factor α k When β k When ≤1, take α. k =0, where λ is the preset load sensitivity coefficient, which represents the additional increase in wear rate caused by the change in unit load ratio when the load exceeds the reference value; The calculated α k Output as a wear acceleration factor caused by load changes.

6. The adaptive control method for preload of heavy-duty joints based on wear state of transmission components according to claim 1, characterized in that: The cumulative wear increment of the transmission component is obtained by numerically integrating the wear acceleration factor relative to the operating time, including: The wear acceleration factor is discrete-time integraled according to the sampling time interval, as shown in the formula: ； Calculate the cumulative wear increment D from the initial time to the current time. acc (t), where N is the total number of sampling steps from the initial time to the current time, and α i Let v0 be the wear acceleration factor at the i-th sampling time, v0 be the preset nominal wear rate, and Δt be the sampling time interval; Then follow formula D total = D acc (t) + v0·t, calculate the cumulative wear increment D of the transmission component. total , where v0·t is the cumulative wear caused by nominal wear, and t is the total running time from the initial moment to the current moment.

7. The adaptive control method for preload of heavy-duty joints based on wear state of transmission components according to claim 6, characterized in that: Based on the preset wear-preload ratio, determine the compensation coefficient required for preload adjustment, including: Obtain the wear-preload proportionality coefficient k, which has been experimentally calibrated beforehand. wp The proportionality coefficient represents the joint preload loss value corresponding to the unit wear depth of the transmission component; According to the formula ΔF=D total ·k wp Calculate the preload loss ΔF caused by the current cumulative wear; Obtain the initial preload value F0 of the joint, and calculate the compensation coefficient C required for preload adjustment according to the formula C=-ΔF / F0. The negative sign indicates that when the preload loss ΔF is positive, the preload needs to be increased to make up for the loss. The compensation coefficient C is a dimensionless ratio. When the compensation coefficient C is negative, it means that the current preload needs to be increased to (1-C) times the initial preload; when the compensation coefficient C is zero, it means that no adjustment is needed.

8. The adaptive control method for preload of heavy-duty joints based on wear state of transmission components according to claim 7, characterized in that: A sequence of joint preload control commands is generated based on the compensation coefficient, including: The calculated compensation coefficient is denoted as C, and the initial preload value of the joint is denoted as F0. The formula ΔF is then used to calculate the compensation coefficient. cmd =∣C∣·F0 Calculate the absolute value of the preload force ΔF that needs to be compensated. cmd , where |C| represents the absolute value of the compensation coefficient C; The force-command conversion coefficient of the joint actuator is denoted as η. This coefficient is obtained beforehand through calibration and represents the preload output value corresponding to a unit control command value, calculated according to formula U. target =ΔF cmd / η Calculate the target control command value U required to achieve the absolute value of the preload compensation force. target ; The value of the control command at the current moment is denoted as U. current According to formula U step =(U target -U current ) / M calculates the instruction change step size U for each control cycle. step , where M is the preset total number of transition cycles, and M is a positive integer; According to formula U j =U current +j·Ustep generates the instruction value for the j-th control cycle, where j=1,2,…,M, when U j When the calculation result exceeds the driver's maximum allowed instruction value, U will... j Limit to the maximum allowed instruction value; The generated M command values ​​are arranged in chronological order to form a joint preload control command sequence.

9. The adaptive control method for preload of heavy-duty joints based on wear state of transmission components according to claim 8, characterized in that: The embedded controller outputs to the joint actuator to apply real-time joint preload compensation values, including: The embedded controller reads each instruction value U in the control instruction sequence of claim 8 sequentially according to a preset control cycle. j ; The currently read instruction value U j After being converted into an analog voltage signal or a pulse width modulation signal, it is output to the joint driver; The joint actuator drives the preload adjustment mechanism to change the amount of pressure on the internal transmission components of the joint based on the received signal, thereby changing the preload applied to the joint in real time. After all the command values ​​in the control command sequence have been executed, the preload stabilizes at the target value after the compensation value has been applied.

10. The adaptive control method for preload of heavy-duty joints based on wear state of transmission components according to claim 1, characterized in that: After applying the real-time joint preload compensation value, the following verification and cyclic adjustment steps are also included: After applying the preload compensation value, wait for a preset stabilization period until the preload adjustment actuator completes its action and the joint operation becomes stable. Re-execute the real-time acquisition step and the Kalman filter dynamic estimation step to obtain a new estimated value of the wear state of the transmission component after applying the compensation value; Extract the amount of joint preload reduction after applying compensation from the new wear condition estimate; The amount of preload reduction after applying the compensation value is compared with the preset allowable reduction threshold: If the attenuation is less than or equal to the allowable attenuation threshold, the preload adjustment is deemed qualified, and the adjustment process ends. If the attenuation amount is still greater than the allowable attenuation threshold, repeat all steps from separating the wear acceleration factor to applying the preload compensation value until the preload attenuation amount meets the requirements.

Images