Real-time fault diagnosis and fault tolerance method of robot motion control system

Abstract

The invention discloses a real-time fault diagnosis and fault tolerance method for a robot motion control system, and relates to the technical field of robot control, and the method comprises the following steps: 1, carrying out the data fusion of multiple sensors; step 2, fault feature extraction; step 2, fault feature extraction; step 3, fault diagnosis; and step 4, fault-tolerant control. According to the real-time fault diagnosis and fault tolerance method of the robot motion control system, the running state of the robot can be reflected more comprehensively and accurately by fusing multi-sensor data and comprehensively utilizing information of different sensors, so that the accuracy of fault diagnosis is improved, and the fault tolerance efficiency is improved. Data preprocessing, feature selection and other links in the fault diagnosis step further improve the diagnosis precision, an advanced algorithm and an optimization strategy are adopted, fault feature extraction, diagnosis and fault-tolerant control decision can be completed in a short time, and it is ensured that the system responds to faults rapidly.

Description

Technical Field

[0001] The present invention relates to the field of robot control technology, in particular to a real-time fault diagnosis and fault tolerance method for a robot motion control system. Background Art

[0002] Fault diagnosis systems for mobile robot control systems are key technologies for ensuring reliable robot operation. Their design requires a combination of real-time monitoring, data analysis, and intelligent decision-making. Common fault diagnosis methods for mobile robots include identifying anomalies by comparing the differences (residuals) between sensor measurements and predicted values from dynamic models; training classification models (such as SVMs and random forests) using historical fault data to identify fault modes; using LSTM networks to predict time series data (such as vibration signals) and detect timing anomalies; and performing FFT transforms on vibration signals to identify characteristic frequencies (such as the frequency harmonics of bearing faults).

[0003] For example, patent publication CN113110406A discloses a fault diagnosis method for a mobile robot control system with front-wheel steering and rear-wheel drive. The method uses an onboard two-dimensional lidar to estimate the actual operating parameters of the mobile robot and constructs residual features and threshold logic for fault diagnosis.

[0004] However, this method has significant limitations. It focuses only on two types of faults: front-wheel steering system faults and rear-wheel drive system faults. It lacks the ability to handle common fault conditions such as sensor faults (such as lidar signal loss), controller faults (such as PID parameter drift), and multiple faults (such as steering faults and drive motor overheating). This makes it difficult for this method to cope with complex and changing fault scenarios in practical applications, and it cannot fully guarantee the stable operation of mobile robot control systems. In addition, the fixed threshold logic used in this method lacks adaptability, which poses a serious problem.

[0005] The above method doesn't account for changes in operating conditions, such as load variations that can cause speed threshold drift. Instead, it uses a fixed threshold and doesn't implement dynamic threshold updates. The threshold β is a preset constant. This approach is prone to false positives and missed negatives in the presence of environmental noise, reducing the accuracy and reliability of fault diagnosis. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention provides a real-time fault diagnosis and fault tolerance method for a robot motion control system, thereby solving the existing problems.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: A real-time fault diagnosis and fault tolerance method for a robot motion control system comprises the following steps:

[0008] Step 1: Multi-sensor data fusion: Use multiple sensors to collect the robot's kinematic and dynamic data in real time, and fuse the data through the Kalman filter algorithm to obtain accurate status information of the robot;

[0009] Step 2: Fault feature extraction: Extract time domain and frequency domain features that can reflect fault information from the fused multi-sensor data;

[0010] Step 3: Fault diagnosis: Use support vector machine as fault diagnosis classifier, take the extracted fault features as input, and classify and identify the fault type;

[0011] Step 4: Fault-tolerant control: Based on the fault diagnosis results, corresponding fault-tolerant control strategies are adopted for different types of faults, including adaptive compensation control, data reconstruction, and redundant sensor switching.

[0012] Preferably, the multi-sensor includes a laser radar, a joint encoder, a force sensor and an IMU.

[0013] Preferably, the formulas for the prediction step and the update step of the Kalman filter algorithm are as follows:

[0014] Prediction steps:

[0015]

[0016] Update steps:

[0017]

[0018] P k|k =(IK k H k )P k|k-1 ;

[0019] in, is the current moment predicted state based on the state estimate of the previous moment, F k is the state transition matrix, is the state estimate at the previous moment, B k is the control input matrix, u k is the control input, P k|k-1 is the covariance matrix of the predicted state, Q k is the process noise covariance matrix, K k is the Kalman gain, H k is the observation matrix, z k is the observation value at the current moment, Rk is the observation noise covariance matrix, is the estimated value of the state at the current moment, P k|k is the covariance matrix of the current state estimate, and I is the identity matrix.

[0020] Preferably, in the fault feature extraction step, for the position signal x(t), its mean μ x and variance The calculation formula is as follows:

[0021]

[0022] Where N is the number of sampling points, t i is the i-th sampling moment.

[0023] Preferably, the fault diagnosis includes the following steps:

[0024] 1) Data preprocessing: normalize the extracted fault feature data, process missing values, and detect and process outliers;

[0025] 2) Feature selection: Select the most representative and discriminative features from the preprocessed features;

[0026] 3): Fault classifier training: Use support vector machine as the fault classifier and train it using the training dataset;

[0027] 4): Fault classification and diagnosis result output: Input the test data after preprocessing and feature selection into the trained fault classifier, obtain the fault classification result and output it.

[0028] Preferably, the normalization process in the data preprocessing step adopts the minimum-maximum normalization method, and the formula is:

[0029]

[0030] Among them, x is the original data, x min and x max are the minimum and maximum values of the feature data, respectively, x norm is the normalized data.

[0031] Preferably, in the feature selection step, correlation analysis or a method based on machine learning is used to select features. The correlation analysis uses the Pearson correlation coefficient r to measure the linear correlation between the feature and the fault label, and the formula is:

[0032]

[0033] Among them, x i and y iare the characteristic value and fault label value of the i-th sample, and are the means of feature x and fault label y respectively, and n is the number of samples.

[0034] Preferably, the formulas for the optimization problem and the dual problem of the support vector machine are as follows:

[0035] Optimization problem:

[0036]

[0037] Dual problem:

[0038]

[0039] Among them, ω is the normal vector of the hyperplane, b is the bias term, ξ i is the slack variable, C is the penalty factor, It is a function that maps the input feature vector x to a high-dimensional feature space, K(x i ,x j )=φ(x i ) T φ(x j ) is the kernel function, y i ∈{-1,1} is the category label of the i-th sample.

[0040] Preferably, in the fault-tolerant control step, for actuator failure, an adaptive compensation control strategy is adopted, and the compensated control input u comp The calculation formula is:

[0041]

[0042] Among them, u is the control input under normal conditions, u f is the actual output of the faulty actuator, is an estimate of the impact of a failure.

[0043] Preferably, in the fault-tolerant control step, for sensor failure, a data reconstruction and redundant sensor switching strategy is adopted; for a case where there is no redundant sensor, an extended Kalman filter algorithm is used to reconstruct the data of the faulty sensor.

[0044] Beneficial effects

[0045] The present invention provides a real-time fault diagnosis and fault tolerance method for a robot motion control system. Compared with the prior art, it has the following advantages:

[0046] By fusing multi-sensor data and comprehensively utilizing information from different sensors, the robot's operating status can be more comprehensively and accurately reflected, thereby improving the accuracy of fault diagnosis. At the same time, data preprocessing and feature selection in the fault diagnosis step further enhance diagnostic accuracy. By using advanced algorithms and optimization strategies, fault feature extraction, diagnosis, and fault-tolerant control decision-making can be completed in a short period of time, ensuring the system's rapid response to faults.

[0047] An effective fault-tolerant control strategy can adjust the robot's operating status in a timely manner when a fault occurs, reduce the impact of the fault on the system, and improve the reliability and stability of the robot's motion control system. This method does not rely on precise mathematical models, can adapt to changes in the robot's working environment and operating conditions, and has strong versatility and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 Schematic diagram of the overall structure of the method of the present invention;

[0049] Figure 2 This is a schematic diagram of the fault diagnosis process of the present invention;

[0050] Figure 3 This is a schematic diagram of the dynamic selection of the control strategy of the present invention. DETAILED DESCRIPTION

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0052] See Figure 1-3 , the present invention provides two technical solutions:

[0053] The first embodiment: a real-time fault diagnosis and fault tolerance method for a robot motion control system, comprising the following steps:

[0054] Multi-sensor data fusion

[0055] The robot's kinematic and dynamic data are collected in real time using a variety of sensors (such as lidar, joint encoders, force sensors, etc.). The data from different sensors is fused and processed using the Kalman filter algorithm to obtain more accurate and reliable robot state information. The basic formula of the Kalman filter algorithm is as follows:

[0056] Prediction steps:

[0057]

[0058] Update steps:

[0059]

[0060] P k|k =(IK k H k )P k|k-1 ;

[0061] in, is the current moment predicted state based on the state estimate of the previous moment, F k is the state transition matrix, is the state estimate at the previous moment, B k is the control input matrix, u k is the control input, P k|k-1 is the covariance matrix of the predicted state, Q k is the process noise covariance matrix, K k is the Kalman gain, H k is the observation matrix, z k is the observation value at the current moment, R k is the observation noise covariance matrix, is the estimated value of the state at the current moment, P k|k is the covariance matrix of the current state estimate, and I is the identity matrix.

[0062] Fault feature extraction

[0063] Extract features that reflect fault information from the fused multi-sensor data. For time-domain features such as position, velocity, and acceleration, these can be described by calculating statistics such as mean, variance, and standard deviation. For frequency-domain features such as force and torque, a fast Fourier transform (FFT) can be used to convert the time-domain signal to the frequency domain and extract the characteristic frequency components.

[0064] Taking the position signal x(t) as an example, its mean μ x and variance The calculation formula is as follows:

[0065]

[0066] Where N is the number of sampling points, t i is the i-th sampling moment.

[0067] Troubleshooting

[0068] The troubleshooting steps are further broken down into the following steps:

[0069] Data preprocessing

[0070] Preprocess the extracted fault feature data to improve data quality and the accuracy of subsequent diagnosis. Specifically include:

[0071] Normalization: Unify data of different features to the same scale range to avoid excessive influence of certain features on the diagnosis results due to data scale differences. The commonly used normalization method is minimum-maximum normalization, the formula is:

[0072]

[0073] Among them, x is the original data, x min and x max are the minimum and maximum values of the feature data, respectively, x norm is the normalized data.

[0074] Missing value handling: Check if there are missing values in the data. If so, use appropriate methods to handle them, such as mean filling, median filling, or model-based filling. For example, for mean filling, if a missing value of a feature is missing, it is filled with the mean of all non-missing values of that feature.

[0075] Outlier detection and processing: Detect outliers in the data using statistical methods (such as those based on standard deviation) or machine learning methods (such as isolation forests). Detected outliers can be deleted or corrected.

[0076] Feature Selection

[0077] Select the most representative and discriminative features from the preprocessed features to reduce data dimensions and improve diagnostic efficiency and accuracy. Common feature selection methods include:

[0078] Correlation analysis: Calculate the correlation coefficient between each feature and the fault label, and select the features with higher correlation. For example, use the Pearson correlation coefficient r to measure the linear correlation between feature x and fault label y. The formula is:

[0079]

[0080] Among them, x i and y i are the characteristic value and fault label value of the i-th sample, and are the means of feature x and fault label y respectively, and n is the number of samples.

[0081] Machine learning-based feature selection: This approach uses machine learning algorithms such as decision trees and random forests to calculate the importance score of each feature and select features with higher scores. For example, random forests evaluate the importance of each feature by calculating the number of splits and information gain in the decision tree.

[0082] Fault classifier training

[0083] Select a suitable fault classifier and train it using the training data set. The present invention uses support vector machine (SVM) as the fault classifier, and the formulas of its optimization problem and dual problem are as follows:

[0084] Given a training sample set {(x1,y1),(x2,y2),…,(x n ,y n )}, where x i ∈R d is the feature vector of the i-th sample, y i ∈{-1,1} is the category label of the i-th sample, and the SVM optimization problem can be expressed as:

[0085]

[0086]

[0087] Among them, ω is the normal vector of the hyperplane, b is the bias term, ξ i is the slack variable, C is the penalty factor, It is a function that maps the input feature vector x to a high-dimensional feature space.

[0088] By introducing the Lagrange multiplier α i ≥0,β i ≥0, the above optimization problem can be transformed into a dual problem:

[0089]

[0090] Among them, K(x i ,x i )=φ(x i ) T φ(x j ) is the kernel function. Commonly used kernel functions include linear kernel, polynomial kernel, radial basis kernel, etc.

[0091] During the training process, methods such as cross-validation are used to select the optimal model parameters (such as penalty factor C and kernel function parameters) to improve the generalization ability of the classifier.

[0092] Fault classification and diagnosis result output

[0093] The preprocessed and feature-selected test data is fed into the trained fault classifier to generate a fault classification result. Based on the classification result, the robot motion control system is determined to have a fault and the type of fault. The diagnostic results are then output and displayed on a monitoring interface or sent to the relevant control module.

[0094] Fault-tolerant control

[0095] When a fault is diagnosed, the corresponding fault-tolerant control strategy is adopted according to the fault type and severity.

[0096] If the fault is an actuator fault, an adaptive compensation control strategy is used. Assuming that the output u of the actuator is affected by the fault and the actual output is uf, compensation can be performed using the following formula:

[0097]

[0098] Among them, u is the control input under normal conditions, u f is the actual output of the faulty actuator, It is an estimate of the impact of the fault. It can be estimated by the fault diagnosis results and the system dynamics model.

[0099] If a sensor failure occurs, data reconstruction and redundant sensor switching strategies are used. If redundant sensors are available, the system switches directly to the backup sensor. If redundant sensors are not available, state estimation methods (such as the extended Kalman filter) are used to reconstruct the data from the faulty sensor based on the system's historical data and dynamic model.

[0100] The second implementation method:

[0101] Example 1: Industrial Robotic Arm Fault Diagnosis and Fault Tolerance

[0102] System construction

[0103] The research used a six-degree-of-freedom industrial robotic arm. Joint encoders were installed at the arm's joints to measure joint angles, force sensors were installed on the end effector to measure force and torque, and a laser radar was used to obtain the arm's position and posture information. The data from these sensors was transmitted to a controller via a data acquisition card for processing.

[0104] Data collection and fusion

[0105] The data from each sensor is collected in real time at a set sampling frequency (e.g. 100 Hz). The collected data is fused and processed using the Kalman filter algorithm to obtain accurate status information of the robotic arm.

[0106] Fault feature extraction

[0107] Features such as position, velocity, acceleration, force, and torque are extracted from the fused data, and corresponding statistics and frequency domain features are calculated. For example, the mean, variance, and standard deviation of joint angles are calculated, and the force signal is fast Fourier transformed to extract characteristic frequency components.

[0108] Troubleshooting

[0109] Data preprocessing: Normalize the extracted fault feature data to bring the data of different features into the range of [0, 1]. Check for missing values in the data. If missing, use mean filling to address them. Use a standard deviation-based method to detect and address outliers.

[0110] Feature selection: Calculate the Pearson correlation coefficient between each feature and the fault label, and select features with an absolute value of the correlation coefficient greater than 0.5 as valid features.

[0111] Fault classifier training: Use a support vector machine as the fault classifier and select the radial basis kernel function. Use cross-validation to select the optimal penalty factor C and kernel function parameters, and train the classifier using the training dataset.

[0112] Fault classification and diagnosis output: The preprocessed and feature-selected test data is fed into the trained support vector machine classifier to generate a fault classification result. Based on the classification result, the robot arm is judged to have a fault and the type of fault, and the diagnosis result is displayed on the monitoring interface.

[0113] Fault-tolerant control

[0114] If an actuator fault is diagnosed, an adaptive compensatory control strategy is used to adjust the control input based on the fault’s severity and location. For example, if a motor failure in a joint results in insufficient output torque, the torque output of other joints can be increased to compensate while the motor in the faulty joint is repaired or replaced.

[0115] If a sensor failure is diagnosed, if there is a redundant sensor, it will immediately switch to the backup sensor; if there is no redundant sensor, the extended Kalman filter algorithm will be used to reconstruct the data of the faulty sensor to ensure the normal operation of the system.

[0116] Example 2: Mobile Robot Fault Diagnosis and Fault Tolerance

[0117] System construction

[0118] A four-wheeled mobile robot is installed with encoders to measure wheel speed, a lidar for environmental perception and positioning, and an IMU (inertial measurement unit) to measure the robot's posture and acceleration. The data from these sensors is transmitted to the robot's main control board for processing.

[0119] Data collection and fusion

[0120] The data of each sensor is collected at a certain sampling frequency (such as 50Hz), and the Kalman filter algorithm is used to fuse the data to obtain the accurate position, speed and posture information of the robot.

[0121] Fault feature extraction

[0122] Features related to the motion of the mobile robot, such as wheel speed, linear velocity, angular velocity, attitude angle, etc., are extracted from the fused data, and the corresponding statistics and frequency domain features are calculated.

[0123] Troubleshooting

[0124] Data preprocessing: normalize the fault feature data and handle missing values and outliers.

[0125] Feature selection: The importance score of each feature is calculated using the random forest algorithm, and the top 80% of features with high scores are selected.

[0126] Fault classifier training: A support vector machine (SVM) is used as the fault classifier, and a polynomial kernel function is selected. The optimal model parameters are determined through cross-validation, and training is performed using the training dataset.

[0127] Fault classification and diagnosis result output: Input the test data into the trained classifier to obtain the fault classification result, determine whether the robot has a fault and the fault type, and send the result to the robot's control module.

[0128] Fault-tolerant control

[0129] If a wheel slip fault is diagnosed, the friction between the wheel and the ground is increased by adjusting the output torque of the motor, and the robot's motion trajectory is adjusted to avoid further slipping.

[0130] If a sensor failure is diagnosed, if there is a redundant sensor, switch to the backup sensor; if there is no redundant sensor, use the state estimation method to reconstruct the data of the faulty sensor to ensure the navigation and positioning accuracy of the robot.

[0131] At the same time, the contents not described in detail in this specification belong to the existing technology known to those skilled in the art, and the model parameters of each electrical appliance are not specifically limited, and conventional equipment can be used.

[0132] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0133] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A real-time fault diagnosis and fault tolerance method for a robot motion control system, characterized in that: The following steps are involved: Step 1: Multi-sensor data fusion: Use multiple sensors to collect the robot's kinematic and dynamic data in real time, and fuse the data through the Kalman filter algorithm to obtain accurate status information of the robot; Step 2: Fault feature extraction: Extract time domain and frequency domain features that can reflect fault information from the fused multi-sensor data; Step 3: Fault diagnosis: Use support vector machine as fault diagnosis classifier, take the extracted fault features as input, and classify and identify the fault type; Step 4: Fault-tolerant control: Based on the fault diagnosis results, corresponding fault-tolerant control strategies are adopted for different types of faults, including adaptive compensation control, data reconstruction, and redundant sensor switching.

2. The real-time fault diagnosis and fault tolerance method for a robot motion control system according to claim 1, characterized in that: The multi-sensor includes a lidar, a joint encoder, a force sensor and an IMU.

3. The real-time fault diagnosis and fault tolerance method for a robot motion control system according to claim 1, characterized in that: The formulas for the prediction step and update step of the Kalman filter algorithm are as follows: Prediction steps: Update steps: P k|k =(I-K k H k )P k|k-1 ; in, is the current moment predicted state based on the state estimate of the previous moment, F k is the state transition matrix, is the state estimate at the previous moment, B k is the control input matrix, u k is the control input, P k|k-1 is the covariance matrix of the predicted state, Q k is the process noise covariance matrix, K k is the Kalman gain, H k is the observation matrix, z k is the observation value at the current moment, R k is the observation noise covariance matrix, is the estimated value of the state at the current moment, P k|k is the covariance matrix of the current state estimate, and I is the identity matrix.

4. The real-time fault diagnosis and fault tolerance method for a robot motion control system according to claim 1, characterized in that: In the fault feature extraction step, for the position signal x(y), its mean μ x and variance The calculation formula is as follows: Where N is the number of sampling points, t i is the i-th sampling moment.

5. The real-time fault diagnosis and fault tolerance method for a robot motion control system according to claim 1, characterized in that: The fault diagnosis comprises the following steps: 1) Data preprocessing: normalize the extracted fault feature data, process missing values, and detect and process outliers; 2) Feature selection: Select the most representative and discriminative features from the preprocessed features; 3): Fault classifier training: Use support vector machine as the fault classifier and train it using the training dataset; 4): Fault classification and diagnosis result output: Input the test data after preprocessing and feature selection into the trained fault classifier, obtain the fault classification result and output it.

6. The real-time fault diagnosis and fault tolerance method for a robot motion control system according to claim 5, characterized in that: The normalization process in the data preprocessing step adopts the minimum-maximum normalization method, and the formula is: Among them, x is the original data, x min and x max are the minimum and maximum values of the feature data, respectively, x norm is the normalized data.

7. The real-time fault diagnosis and fault tolerance method for a robot motion control system according to claim 5, characterized in that: In the feature selection step, correlation analysis or a machine learning-based method is used to select features. Correlation analysis uses the Pearson correlation coefficient r to measure the linear correlation between features and fault labels. The formula is: Among them, x i and y i are the characteristic value and fault label value of the i-th sample, and are the means of feature x and fault label y respectively, and n is the number of samples.

8. The real-time fault diagnosis and fault tolerance method for a robot motion control system according to claim 5, characterized in that: The formulas for the optimization problem and dual problem of the support vector machine are as follows: Optimization problem: Dual problem: Among them, ω is the normal vector of the hyperplane, b is the bias term, ξ i is the slack variable, C is the penalty factor, It is a function that maps the input feature vector x to a high-dimensional feature space, K(x i ,x j )=φ(x i ) T φ(x j ) is the kernel function, y i ∈{-1,1} is the category label of the i-th sample.

9. The real-time fault diagnosis and fault tolerance method for a robot motion control system according to claim 1, characterized in that: In the fault-tolerant control step, for actuator failure, an adaptive compensation control strategy is adopted, and the compensated control input u comp The calculation formula is: Among them, u is the control input under normal conditions, u f is the actual output of the faulty actuator, is an estimate of the impact of a failure.

10. The real-time fault diagnosis and fault tolerance method for a robot motion control system according to claim 1, characterized in that: In the fault-tolerant control step, for sensor failure, a data reconstruction and redundant sensor switching strategy is adopted. For the case where there is no redundant sensor, an extended Kalman filter algorithm is used to reconstruct the data of the faulty sensor.

Images