An intelligent control method and system for industrial robots

Abstract

The present invention relates to the technical field of industrial robots, and in particular to an intelligent control method and system for industrial robots, comprising the following steps: S1, data acquisition and preprocessing: real-time acquisition of operating data in the working environment of the industrial robot; S2, model construction: construction of the industrial robot control model; S3, real-time control and autonomous learning: regular update and optimization of the industrial robot control model; S4, adaptive friction compensation: real-time detection and compensation of friction changes of various joints and sliding parts of the robot; S5, fault detection and processing: real-time monitoring of the operating state of the robot, discovery and processing of potential faults through anomaly detection algorithms; S6, human-machine interaction: realization of interaction between operators and robots. The present invention dynamically adjusts drive instructions, compensates for friction changes, and improves the accuracy and consistency of robot operation.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial robots, and in particular to an intelligent control method and system for industrial robots. Background Art

[0002] With the development of industrial automation, industrial robots are being used more and more widely in the manufacturing industry. Industrial robots have played an important role in improving production efficiency, reducing production costs and improving product quality. However, with the complexity of application scenarios, the demand for intelligent control systems of industrial robots is also increasing. Existing industrial robot control systems still face many challenges in terms of real-time performance, accuracy and robustness. Especially when dealing with complex and changing working environments, traditional control methods are difficult to meet the requirements of efficient and stable operation.

[0003] Existing industrial robot control systems usually rely on preset control models and fixed control parameters, lacking adaptability and intelligent processing capabilities. These systems have many deficiencies in data acquisition, model building, real-time control, fault detection, and human-computer interaction. For example, improper handling of noise and outliers in the data acquisition process will affect the accuracy of the model, fixed control models cannot adapt to real-time changing working environments, fault detection lags lead to poor system stability, the human-computer interaction interface is not friendly enough, and the operation is complicated. These problems seriously restrict the application and performance improvement of industrial robots in complex environments.

[0004] The purpose of the present invention is to provide an intelligent control method for an industrial robot, which improves the convenience of operation and the level of intelligence, thereby significantly improving the working efficiency and product quality of the industrial robot. Summary of the invention

[0005] Based on the above objectives, the present invention provides an intelligent control method and system for an industrial robot.

[0006] An intelligent control method for an industrial robot comprises the following steps:

[0007] S1, data collection and preprocessing: collect the operation data of the industrial robot in the working environment in real time through a variety of sensors, and preprocess the collected operation data, including data cleaning, normalization and feature extraction;

[0008] S2, model construction: construct the industrial robot control model based on the preprocessed operation data;

[0009] S3, real-time control and autonomous learning: prediction and decision-making are carried out based on the real-time collected operation data to achieve real-time control of industrial robots. During the actual operation of the robot, new operation data is continuously collected to regularly update and optimize the industrial robot control model.

[0010] S4, Adaptive Friction Compensation: During the robot's motion, real-time detection and compensation of friction changes in the robot's joints and sliding parts, including:

[0011] S41, Friction force detection: Real-time detection of friction force data through sensors installed on robot joints and sliding parts;

[0012] S42, friction analysis: using an adaptive friction prediction model to analyze the friction data detected in real time and identify the change pattern and trend of friction;

[0013] S43, dynamic adjustment: dynamically adjust the robot's drive instructions according to the analysis results of the friction data, and compensate for the change of friction by optimizing the control parameters;

[0014] S5, fault detection and processing: During the control process, the robot's operating status is monitored in real time, and potential faults are discovered and processed through anomaly detection algorithms;

[0015] S6, Human-Robot Interaction: Enable interaction between operators and robots through user interface and natural language processing technology.

[0016] Optionally, the data collection and preprocessing in S1 includes:

[0017] S11, data collection: Use a variety of sensors to collect real-time operating data in the robot's working environment, including:

[0018] Position acquisition: Capture the robot's position information through visual sensors;

[0019] Force condition collection: Detect the force condition of the joint through force sensors;

[0020] Temperature collection: monitor the temperature of the robot through temperature sensors;

[0021] S12, data cleaning: using mean filtering to clean the collected operation data to remove noise and outliers;

[0022] S13, data normalization processing: using minimum-maximum normalization to normalize the data of different sensors to the same scale;

[0023] S14, feature extraction: Linear discriminant analysis (LDA) is used to extract key features from the normalized operating data, including position information features, force features and temperature features. The position information features include the spatial position coordinates, motion trajectory, velocity and acceleration of each joint of the robot; the force features include the magnitude, direction and torque change of the force on the joints; the temperature features include the temperature change, temperature gradient and temperature rise rate of each part of the robot.

[0024] Optionally, the model building in S2 includes:

[0025] S21, data segmentation: divide the preprocessed running data into a training set and a validation set;

[0026] S22, model selection: Select the convolutional neural network (CNN) model for model construction, including:

[0027] Convolutional layer calculation:

[0028] Among them, h i,j,k is the element of the convolution output feature map, x is the input data, w is the convolution kernel, b k is the bias, σ is the activation function;

[0029] Pooling layer calculation:

[0030] in, Elements of the pooled output feature map;

[0031] Fully connected layer calculation:

[0032] Among them, z i is the output of the fully connected layer, h j is the output of the pooling layer, w i,j is the weight of the fully connected layer, b i is bias;

[0033] S23, model training: using the training set data, the convolutional neural network (CNN) model is trained through the Adam optimizer, and the model parameters are continuously adjusted;

[0034] S24, Model Validation and Optimization: Use validation set data to evaluate the performance of the industrial robot control model and optimize the model by adjusting model hyperparameters (such as learning rate, convolution kernel size, and number of layers).

[0035] Optionally, the real-time control and autonomous learning in S3 include:

[0036] S31, real-time control: using the convolutional neural network (CNN) model, prediction and decision-making are carried out based on the real-time collected operation data, and control instructions are output to achieve real-time control of the industrial robot, which is expressed as:

[0037] y t =CNN(x t );

[0038] Among them, y t is the control instruction at time t, x t It is the running data collected in real time, and CNN is the convolutional neural network model;

[0039] S32, autonomous learning: During the actual operation of the robot, new operation data is collected in real time, and the stochastic gradient descent method (SGD) is used to regularly update and optimize the industrial robot control model using the new operation data, which is expressed as:

[0040]

[0041] Among them, θ t is the model parameter of the tth step, α is the learning rate, is the gradient of the loss function L with respect to the parameter θ.

[0042] Optionally, the friction force detection in S41 includes:

[0043] S411, sensor installation: force sensors are installed on each joint and sliding part of the robot to monitor and record the forces acting on the joints and sliding parts in real time;

[0044] S412, friction data collection: real-time collection of friction data on joints and sliding parts during movement;

[0045] S413, data processing: preliminary processing of the friction data collected in real time, including denoising and filtering.

[0046] Optionally, the adaptive friction prediction model in S42 adopts a mixed density network (MDN) model, and the mixed density network (MDN) model includes:

[0047] S421, input layer and adaptive sliding window: An adaptive sliding window is introduced to process the temporal information of the friction data, and its output is used as the input of the mixture density network (MDN) model, which is expressed as:

[0048]

[0049] Where W is the size of the sliding window, x i is the input data at the i-th moment, μ tis the mean of the sliding window, σ t is the standard deviation of the sliding window;

[0050] S422, output layer: Based on the sliding window output, the parameters of the mixed Gaussian distribution are output, expressed as:

[0051] Mixing weights:

[0052] Among them, z α,i is the weight of the i-th mixture component, M is the number of Gaussian components;

[0053] Mean of Gaussian distribution: μ i =z μ,i ;

[0054] Among them, z μ,i is the mean of the i-th mixture component;

[0055] The standard deviation of the Gaussian distribution: σ i =exp(z σ,i );

[0056] Among them, z σ,i is the standard deviation of the ith mixture component;

[0057] S423, loss function: The loss function is defined as negative log-likelihood, expressed as:

[0058]

[0059] Among them, y t is the actual observed value, represents the probability density function of the i-th Gaussian component;

[0060]

[0061] S424, model training: Use an optimization algorithm (Adam) to train the parameters of the mixture density network (MDN) model to minimize the loss function, expressed as:

[0062]

[0063] Among them, θ t is the parameter of the tth step, α is the learning rate, and are the bias-corrected values ​​of the first-order and second-order moment estimates, respectively;

[0064] S425, model prediction: input new friction data into the trained model and predict the change pattern and trend of friction, which is expressed as:

[0065]

[0066] in, is the predicted friction value, α i is the mixing weight, μ i is the mean of the mixture components.

[0067] Optionally, the dynamic adjustment in S43 includes:

[0068] S431, friction data analysis: using the adaptive friction prediction model, the friction data detected in real time is analyzed to obtain the change pattern and trend of the friction;

[0069] S432, based on predicted friction values Calculate the control parameter Δu that needs to be compensated t , expressed as:

[0070]

[0071] Among them, K p is the control gain, is the predicted friction value, y t is the actual friction value detected;

[0072] S433, adjust the drive command: the calculated compensation value Δu t Applied to the robot's drive instructions, the control parameters are optimized to compensate for the change in friction, expressed as:

[0073]

[0074] Among them, u t is the adjusted drive instruction, is the original drive instruction, Δu t is the compensation value.

[0075] Optionally, the fault detection and processing in S5 includes:

[0076] S51, real-time monitoring: using a variety of sensors to collect the robot's operating status data in real time;

[0077] S52, anomaly detection: using an anomaly detection algorithm based on K-means clustering to identify anomalies in the operating state;

[0078] S53, fault handling: If an abnormality is detected, the potential fault is handled through a preset processing mechanism (such as alarm, shutdown, automatic adjustment of control parameters).

[0079] Optionally, the human-computer interaction in S6 includes:

[0080] S61, User Interface Design: Design a graphical user interface through which operators can monitor the robot's operating status in real time, input control commands, and view feedback;

[0081] S62, Natural Language Processing: Integrate natural language processing (NLP) technology to enable operators to interact with industrial robots, expressed as:

[0082] P(W|X)=argmax W P(X|W)P(W);

[0083] Where P(W|X) is the probability of a word sequence W given a speech signal X, P(X|W) is the acoustic model, and P(W) is the language model;

[0084] I=argmax I P(I|T);

[0085] Where I is the intent, T is the text input, and P(I|T) is the probability of intent given the text input;

[0086] S63, Real-time interaction and feedback: Provide real-time interaction and feedback through user interface and natural language processing (NLP) technology.

[0087] An intelligent control system for an industrial robot, used to implement the above-mentioned intelligent control method for an industrial robot, comprises the following modules:

[0088] Data acquisition and preprocessing module: collects the operating data of the industrial robot in the working environment in real time through a variety of sensors, and preprocesses the collected operating data, including data cleaning, normalization and feature extraction;

[0089] Model building module: builds the industrial robot control model based on the preprocessed operation data;

[0090] Real-time control and autonomous learning module: Make predictions and decisions based on real-time collected operation data to achieve real-time control of industrial robots. During the actual operation of the robots, new operation data is continuously collected to regularly update and optimize the control model of industrial robots.

[0091] Adaptive friction compensation module: During the robot's motion, it detects and compensates for the friction changes of the robot's joints and sliding parts in real time;

[0092] Fault detection and processing module: During the control process, the robot's operating status is monitored in real time, and potential faults are discovered and processed through anomaly detection algorithms;

[0093] Human-computer interaction module: Through the user interface and natural language processing technology, the interaction between the operator and the robot is realized.

[0094] Beneficial effects of the present invention:

[0095] The present invention constructs a high-precision control model through data collection and preprocessing of multiple sensors, ensures the accuracy and effectiveness of the data, uses convolutional neural networks for model construction, and combines real-time control with autonomous learning mechanisms, so that the system can achieve efficient and precise control in complex and changeable environments. At the same time, through adaptive friction compensation technology, the drive instructions are dynamically adjusted to compensate for changes in friction, thereby improving the accuracy and consistency of robot operations.

[0096] The present invention monitors the operating status of the robot in real time through an anomaly detection algorithm based on K-means clustering, promptly discovers and handles potential faults, and significantly improves the stability and reliability of the system. In addition, the system sets thresholds through statistical characteristic analysis, which improves the accuracy and response speed of fault detection and ensures the safe operation of the robot under various working conditions.

[0097] The present invention realizes efficient human-computer interaction through the user interface and natural language processing technology, so that the operator can conveniently monitor and control the robot. The graphical user interface provides real-time monitoring and control functions, and the natural language processing technology enables the operator to input commands and interact with the system through voice or text, which significantly improves the convenience and intelligence level of operation, thereby greatly improving production efficiency and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0099] Figure 1 A schematic diagram of a control method flow chart of an embodiment of the present invention;

[0100] Figure 2 Schematic diagram of system function modules according to an embodiment of the present invention. DETAILED DESCRIPTION

[0101] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0102] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0103] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0104] like Figure 1 As shown, an intelligent control method for an industrial robot comprises the following steps:

[0105] S1, data collection and preprocessing: collect the operation data of the industrial robot in the working environment in real time through a variety of sensors, and preprocess the collected operation data, including data cleaning, normalization and feature extraction, to ensure the accuracy and effectiveness of the data;

[0106] S2, model construction: construct the industrial robot control model based on the preprocessed operation data;

[0107] S3, real-time control and autonomous learning: prediction and decision-making are carried out based on the real-time collected operation data to achieve real-time control of industrial robots. During the actual operation of the robot, new operation data is continuously collected to regularly update and optimize the industrial robot control model.

[0108] S4, Adaptive Friction Compensation: During the robot's motion, the friction changes of the robot's joints and sliding parts are detected and compensated in real time. By dynamically adjusting the drive instructions, the motion error caused by the friction change is reduced, and the accuracy and consistency of the robot's operation are improved. Specifically, it includes:

[0109] S41, Friction force detection: Real-time detection of friction force data through sensors installed on robot joints and sliding parts;

[0110] S42, friction analysis: using an adaptive friction prediction model to analyze the friction data detected in real time and identify the change pattern and trend of friction;

[0111] S43, dynamic adjustment: According to the analysis results of friction data, dynamically adjust the robot's drive instructions, optimize control parameters, compensate for changes in friction, reduce motion errors, and improve operating accuracy;

[0112] S5, fault detection and processing: During the control process, the robot's operating status is monitored in real time, and potential faults are discovered and processed through anomaly detection algorithms to ensure the stability and reliability of the system;

[0113] S6, Human-machine interaction: Through the user interface and natural language processing technology, the interaction between the operator and the robot is realized, which is convenient for operation and management;

[0114] Through the above steps, it is ensured that the industrial robot can operate stably and efficiently in a complex and changeable working environment, the accuracy and consistency of the system are improved, the fault handling capability and operational convenience are enhanced, thereby greatly improving production efficiency and product quality.

[0115] Data collection and preprocessing in S1 include:

[0116] S11, data collection: Use a variety of sensors to collect real-time operating data in the robot's working environment, including:

[0117] Position acquisition: Capture the robot's position information through visual sensors;

[0118] Force condition collection: Detect the force condition of the joint through force sensors;

[0119] Temperature collection: monitor the temperature of the robot through temperature sensors;

[0120] S12, data cleaning: use mean filtering to clean the collected operation data, remove noise and outliers, and ensure the accuracy of the data, expressed as:

[0121]

[0122] in, represents the smoothed data, and N represents the size of the filter window;

[0123] S13, data normalization processing: use minimum-maximum normalization to normalize the data of different sensors to the same scale for subsequent processing, expressed as:

[0124]

[0125] Among them, x' represents the normalized data, x min and x max Respectively represent the minimum and maximum values ​​of the data;

[0126] S14, feature extraction: Linear discriminant analysis (LDA) is used to extract key features from the normalized operation data, including position information features, force features and temperature features. Position information features include the spatial position coordinates, motion trajectory, speed and acceleration of each joint of the robot. Force features include the magnitude, direction and torque change of the force on the joint. Temperature features include the temperature change, temperature gradient and temperature rise rate of each part of the robot.

[0127] Linear discriminant analysis (LDA) specifically includes:

[0128] Compute the mean vector for each class:

[0129] Among them, m i is the mean vector of the i-th class, n i is the number of samples in the i-th class, C i is the sample set of the i-th category;

[0130] Compute the overall mean vector:

[0131] Among them, m is the total mean vector, N is the total number of samples, and k is the total number of classes;

[0132] Calculate the intra-class scatter matrix S W and the inter-class scatter matrix S B :

[0133]

[0134] Calculating the Matrix The eigenvectors and eigenvalues ​​of , and the eigenvector corresponding to the largest eigenvalue is selected to form the projection matrix W:

[0135]

[0136] Where w is the projection vector and λ is the eigenvalue, which is used to measure the importance of the corresponding eigenvector. The larger the eigenvalue, the greater the inter-class scatter in that direction is relative to the intra-class scatter;

[0137] Project the original data into the new feature space: Z = XW;

[0138] Among them, X is the normalized data matrix, W is the projection matrix, and Z is the extracted feature matrix;

[0139] Through the above steps, not only the data quality is improved, but also a solid foundation is provided for the robustness and performance optimization of the system, ensuring that the industrial robot can operate stably and efficiently in a complex and changeable working environment.

[0140] Model building in S2 includes:

[0141] S21, data segmentation: the preprocessed running data is divided into a training set and a validation set. The training set is used to train the model, and the validation set is used to evaluate the model performance.

[0142] S22, model selection: Select the convolutional neural network (CNN) model for model construction, including:

[0143] Convolutional layer calculation:

[0144] Among them, h i,j,k is the element of the convolution output feature map, x is the input data, w is the convolution kernel, b k is the bias, σ is the activation function;

[0145] Pooling layer calculation:

[0146] in, Elements of the pooled output feature map;

[0147] Fully connected layer calculation:

[0148] Among them, z i is the output of the fully connected layer, h j is the output of the pooling layer, w i,j is the weight of the fully connected layer, b i is bias;

[0149] S23, model training: Use the training set data to train the convolutional neural network (CNN) model through the Adam optimizer, and continuously adjust the model parameters so that the model can accurately predict the robot's control instructions, including:

[0150] Parameter update:

[0151] Bias correction for first-order moment estimates:

[0152] Bias correction for second-order moment estimates:

[0153] First moment estimate: m t=β1m t-1 +(1-β1)g t ;

[0154] Second moment estimate:

[0155] Among them, θ t is the parameter of the tth step, α is the learning rate, and are the bias correction values ​​of the first-order and second-order moment estimates, respectively, and m t and v t are the first-order and second-order moment estimates, β1 and β2 are exponential decay rates (usually β1 = 0.9 and β2 = 0.999), ∈ is a small constant to prevent the denominator from being zero, and g t is the gradient of the tth step;

[0156] S24, model validation and optimization: Use validation set data to evaluate the performance of the industrial robot control model, and optimize the model by adjusting model hyperparameters (such as learning rate, convolution kernel size, number of layers, etc.) to make it have better generalization ability and robustness, expressed as:

[0157] Accuracy:

[0158] Among them, TP is true positive, TN is true negative, FP is false positive, and FN is false negative;

[0159] Cross Entropy Loss:

[0160] Where N is the number of samples, y i is the actual label, is the predicted probability;

[0161] Learning rate adjustment:

[0162] Among them, α t is the learning rate of the tth step, α0 is the initial learning rate, and k is the decay rate;

[0163] Hyperparameter Optimization: θ * =argmin θ∈Θ L(θ);

[0164] Among them, θ is a hyperparameter, Θ is a hyperparameter space, and L(θ) is a loss function;

[0165] The loss function L(θ) is expressed as:

[0166] Among them, L is the loss value, N is the number of samples, and y i is the actual value, is the predicted value;

[0167] Through the above content, data features can be efficiently extracted and model parameters can be optimized, thereby improving the prediction accuracy and generalization ability of the model, ensuring the stability and accuracy of the industrial robot control model in complex environments, and significantly improving the robot's work efficiency and operating accuracy.

[0168] Real-time control and autonomous learning in S3 include:

[0169] S31, real-time control: using the convolutional neural network (CNN) model, prediction and decision-making are carried out based on the real-time collected operation data, and control instructions are output to achieve real-time control of the industrial robot, which is expressed as:

[0170] y t =CNN(x t );

[0171] Among them, y t is the control instruction at time t, x t It is the running data collected in real time, and CNN is the convolutional neural network model;

[0172] S32, autonomous learning: During the actual operation of the robot, new operation data is collected in real time, and the stochastic gradient descent method (SGD) is used to regularly update and optimize the industrial robot control model using the new operation data to keep the model up to date and adaptable, which is expressed as:

[0173]

[0174] Among them, θ t is the model parameter of the tth step, α is the learning rate, is the gradient of the loss function L with respect to the parameter θ;

[0175] Through the above contents, it is ensured that the industrial robot control system can achieve efficient and accurate real-time control in complex and changing environments, and has the ability of self-learning and continuous optimization, which significantly improves the adaptability and intelligence level of the robot.

[0176] The friction detection in S41 includes:

[0177] S411, sensor installation: force sensors are installed on each joint and sliding part of the robot to monitor and record the forces acting on the joints and sliding parts in real time;

[0178] S412, friction data collection: real-time collection of friction data on joints and sliding parts during movement;

[0179] S413, data processing: preliminary processing of the friction data collected in real time, including denoising and filtering, to ensure the accuracy and reliability of the data;

[0180] Through the above steps, the friction force change data of each joint and sliding component of the robot during movement can be accurately obtained, providing accurate basic data for subsequent adaptive friction compensation, which significantly improves the accuracy and stability of the robot's movement.

[0181] The adaptive friction prediction model in S42 adopts the mixed density network (MDN) model, which includes:

[0182] S421, input layer and adaptive sliding window: An adaptive sliding window is introduced to process the temporal information of the friction data, and its output is used as the input of the mixture density network (MDN) model, which is expressed as:

[0183]

[0184] Where W is the size of the sliding window, x i is the input data at the i-th moment, μ t is the mean of the sliding window, σ t is the standard deviation of the sliding window;

[0185] S422, output layer: Based on the sliding window output, the parameters of the mixed Gaussian distribution are output, expressed as:

[0186] Mixing weights:

[0187] Among them, z α,i is the weight of the i-th mixture component, normalized by the softmax function, and M is the number of Gaussian components;

[0188] Mean of Gaussian distribution: μ i =z μ,i ;

[0189] Among them, z μ,i is the mean of the i-th mixture component;

[0190] The standard deviation of the Gaussian distribution: σ i =exp(z σ,i );

[0191] Among them, z σ,i is the standard deviation of the ith mixture component, using an exponential function to ensure that it is positive;

[0192] S423, loss function: The loss function is defined as negative log-likelihood, expressed as:

[0193]

[0194] Among them, y t is the actual observed value, represents the probability density function of the i-th Gaussian component;

[0195]

[0196] S424, model training: Use an optimization algorithm (Adam) to train the parameters of the mixture density network (MDN) model to minimize the loss function, expressed as:

[0197]

[0198] Among them, θ t is the parameter of the tth step, α is the learning rate, and are the bias-corrected values ​​of the first-order and second-order moment estimates, respectively;

[0199] S425, model prediction: input new friction data into the trained model and predict the change pattern and trend of friction, which is expressed as:

[0200]

[0201] in, is the predicted friction value, α i is the mixing weight, μ i is the mean of the mixture components;

[0202] Through the above content, we can better process the friction data detected in real time. The introduction of adaptive sliding window to process time series data enables the model to capture the time dependence and change trend of friction data. The output layer estimates the parameters of the mixed Gaussian distribution, which enhances the model's ability to recognize the friction change pattern and trend. Combined with the negative log-likelihood loss function and the Adam optimization algorithm, the model can be trained and predicted efficiently and accurately, thereby improving the accuracy and stability of the industrial robot control system.

[0203] Dynamic adjustments in S43 include:

[0204] S431, friction data analysis: using the adaptive friction prediction model, the friction data detected in real time is analyzed to obtain the change pattern and trend of the friction;

[0205] S432, based on predicted friction values Calculate the control parameter Δu that needs to be compensated t , so that the robot can effectively cope with the change of friction force, expressed as:

[0206]

[0207] Among them, K p is the control gain, is the predicted friction value, y t is the actual friction value detected;

[0208] S433, adjust the drive command: the calculated compensation value Δu t Applied to the robot's driving instructions, it optimizes the control parameters, compensates for the change in friction, and improves the operating accuracy, which can be expressed as:

[0209]

[0210] Among them, u t is the adjusted drive instruction, is the original drive instruction, Δu t is the compensation value;

[0211] Through the above contents, the efficient and stable operation of the robot under different friction conditions is ensured, the accuracy and consistency of the robot operation are significantly improved, and the system's adaptability and intelligence level are enhanced.

[0212] Fault detection and handling in S5 include:

[0213] S51, real-time monitoring: using a variety of sensors to collect the robot's operating status data in real time;

[0214] S52, anomaly detection: using an anomaly detection algorithm based on K-means clustering to identify anomalies in the running state, specifically including:

[0215] Calculate the distance between each data point and the cluster center: d ij =‖F t -C j ‖;

[0216] Among them, d ij is the i-th data point F t To the jth cluster center C j distance;

[0217] Assign data points to the nearest cluster center: Assign(F t ) = argmin j d ij ;

[0218] Calculate the anomaly score: s(F t )=min j d ij ;

[0219] Determine abnormality: When s(F t ) exceeds the set threshold, it is judged as abnormal;

[0220] S53, fault handling: If an abnormality is detected, the potential fault is handled through a preset processing mechanism (such as alarm, shutdown, automatic adjustment of control parameters) to ensure the stability and reliability of the system;

[0221] Through the above steps, the preset processing mechanism can be adopted in time, such as alarm or automatic adjustment of control parameters, to ensure the stable operation and safe operation of the robot, improve the accuracy and response speed of fault detection, and significantly improve the reliability and intelligence level of the industrial robot control system.

[0222] Setting the threshold By analyzing the distribution characteristics of historical data, a threshold based on statistical characteristics is set, which is expressed as:

[0223] Threshold = μ + k·σ;

[0224] Among them, μ is the mean of historical data, σ is the standard deviation, and k is a constant (2 or 3).

[0225] Human-computer interaction in S6 includes:

[0226] S61, User Interface Design: Design a graphical user interface through which operators can monitor the robot's operating status in real time, input control commands, and view feedback;

[0227] S62, Natural Language Processing: Integrate natural language processing (NLP) technology to enable operators to interact with industrial robots, expressed as:

[0228] P(W|X)=argmax W P(X|W)P(W);

[0229] Where P(W|X) is the probability of a word sequence W given a speech signal X, P(X|W) is the acoustic model, and P(W) is the language model;

[0230] I=argmax I P(I|T);

[0231] Where I is the intent, T is the text input, and P(I|T) is the probability of intent given the text input;

[0232] S63, Real-time interaction and feedback: Through the user interface and natural language processing (NLP) technology, real-time interaction and feedback are provided, enabling operators to quickly respond to changes in system status and adjust operating strategies;

[0233] Through the above steps, the convenience and intelligence level of operation are improved. Combined with voice recognition, intent recognition and real-time feedback, the system can quickly and accurately parse and execute operation instructions, thereby improving the operating efficiency and user experience of the industrial robot control system.

[0234] like Figure 2 As shown, an intelligent control system for an industrial robot, used to implement the above-mentioned intelligent control method for an industrial robot, includes the following modules:

[0235] Data acquisition and preprocessing module: collects the operating data of the industrial robot in the working environment in real time through a variety of sensors, and preprocesses the collected operating data, including data cleaning, normalization and feature extraction;

[0236] Model building module: builds the industrial robot control model based on the preprocessed operation data;

[0237] Real-time control and autonomous learning module: Make predictions and decisions based on real-time collected operation data to achieve real-time control of industrial robots. During the actual operation of the robots, new operation data is continuously collected to regularly update and optimize the control model of industrial robots.

[0238] Adaptive friction compensation module: During the robot's motion, it detects and compensates for the friction changes of the robot's joints and sliding parts in real time;

[0239] Fault detection and processing module: During the control process, the robot's operating status is monitored in real time, and potential faults are discovered and processed through anomaly detection algorithms;

[0240] Human-computer interaction module: Through the user interface and natural language processing technology, the interaction between the operator and the robot is realized.

[0241] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0242] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An intelligent control method for an industrial robot, characterized in that: The following steps are involved: S1, data collection and preprocessing: collect the operation data of the industrial robot in the working environment in real time through a variety of sensors, and preprocess the collected operation data, including data cleaning, normalization and feature extraction; S2, model construction: construct the industrial robot control model based on the preprocessed operation data; S3, real-time control and autonomous learning: prediction and decision-making are carried out based on the real-time collected operation data to achieve real-time control of industrial robots. During the actual operation of the robot, new operation data is continuously collected to regularly update and optimize the industrial robot control model. S4, Adaptive Friction Compensation: During the robot's motion, real-time detection and compensation of friction changes in the robot's joints and sliding parts, including: S41, Friction force detection: Real-time detection of friction force data through sensors installed on robot joints and sliding parts; S42, friction analysis: using an adaptive friction prediction model to analyze the friction data detected in real time and identify the change pattern and trend of friction; S43, dynamic adjustment: dynamically adjust the robot's drive instructions according to the analysis results of the friction data, and compensate for the change of friction by optimizing the control parameters; S5, fault detection and processing: During the control process, the robot's operating status is monitored in real time, and potential faults are discovered and processed through anomaly detection algorithms; S6, Human-Robot Interaction: Enable interaction between operators and robots through user interface and natural language processing technology.

2. The intelligent control method for an industrial robot according to claim 1, characterized in that: The data collection and preprocessing in S1 include: S11, data collection: Use a variety of sensors to collect real-time operating data in the robot's working environment, including: Position acquisition: Capture the robot's position information through visual sensors; Force condition collection: Detect the force condition of the joint through force sensors; Temperature collection: monitor the temperature of the robot through temperature sensors; S12, data cleaning: using mean filtering to clean the collected operation data to remove noise and outliers; S13, data normalization processing: using minimum-maximum normalization to normalize the data of different sensors to the same scale; S14, feature extraction: Linear discriminant analysis is used to extract key features from the normalized operating data, including position information features, force features and temperature features. The position information features include the spatial position coordinates, motion trajectory, speed and acceleration of each joint of the robot; the force features include the magnitude, direction and torque change of the force on the joints; the temperature features include the temperature change, temperature gradient and temperature rise rate of each part of the robot.

3. The intelligent control method for an industrial robot according to claim 1, characterized in that: The model construction in S2 includes: S21, data segmentation: divide the preprocessed running data into a training set and a validation set; S22, model selection: Select a convolutional neural network model for model construction, including: Convolutional layer calculation: Among them, h i,j,k is the element of the convolution output feature map, x is the input data, w is the convolution kernel, b k is the bias, σ is the activation function; Pooling layer calculation: in, Elements of the pooled output feature map; Fully connected layer calculation: Among them, z i is the output of the fully connected layer, h j is the output of the pooling layer, w i,j is the weight of the fully connected layer, b i is bias; S23, model training: Use the training set data to train the convolutional neural network model through the Adam optimizer and continuously adjust the model parameters; S24, Model Validation and Optimization: Use validation set data to evaluate the performance of the industrial robot control model and optimize the model by adjusting model hyperparameters.

4. The intelligent control method for an industrial robot according to claim 3, characterized in that: The real-time control and autonomous learning in S3 include: S31, real-time control: using the convolutional neural network model, prediction and decision-making are carried out based on the real-time collected operation data, and control instructions are output to achieve real-time control of the industrial robot, which is expressed as: y t =CNN(x t ); Among them, y t is the control instruction at time t, x t It is the running data collected in real time, and CNN is the convolutional neural network model; S32, autonomous learning: During the actual operation of the robot, new operation data is collected in real time, and the stochastic gradient descent method is used to regularly use the new operation data to update and optimize the industrial robot control model, which is expressed as: Among them, θ t is the model parameter of the tth step, α is the learning rate, is the gradient of the loss function L with respect to the parameter θ.

5. The intelligent control method for an industrial robot according to claim 1, characterized in that: The friction force detection in S41 includes: S411, sensor installation: force sensors are installed on each joint and sliding part of the robot to monitor and record the forces acting on the joints and sliding parts in real time; S412, friction data collection: real-time collection of friction data on joints and sliding parts during movement; S413, data processing: preliminary processing of the friction data collected in real time, including denoising and filtering.

6. The intelligent control method for an industrial robot according to claim 5, characterized in that: The adaptive friction prediction model in S42 adopts a mixed density network model, and the mixed density network model includes: S421, input layer and adaptive sliding window: An adaptive sliding window is introduced to process the temporal information of friction data, and its output is used as the input of the mixed density network model, which is expressed as: Where W is the size of the sliding window, x i is the input data at the i-th moment, μ t is the mean of the sliding window, σ t is the standard deviation of the sliding window; S422, output layer: Based on the sliding window output, the parameters of the mixed Gaussian distribution are output, expressed as: Mixing weights: Among them, z α,i is the weight of the i-th mixture component, M is the number of Gaussian components; Mean of Gaussian distribution: μ i =z μ,i ; Among them, z μ,i is the mean of the i-th mixture component; The standard deviation of the Gaussian distribution: σ i =exp(z σ,i ); Among them, z σ,i is the standard deviation of the ith mixture component; S423, loss function: The loss function is defined as negative log-likelihood, expressed as: Among them, y t is the actual observed value, represents the probability density function of the i-th Gaussian component; S424, model training: Use an optimization algorithm to train the parameters of the mixed density network model to minimize the loss function, expressed as: Among them, θ t is the parameter of the tth step, α is the learning rate, and are the bias-corrected values ​​of the first-order and second-order moment estimates, respectively; S425, model prediction: input new friction data into the trained model and predict the change pattern and trend of friction, which is expressed as: in, is the predicted friction value, α i is the mixing weight, μ i is the mean of the mixture components.

7. The intelligent control method for an industrial robot according to claim 6, characterized in that: The dynamic adjustment in S43 includes: S431, friction data analysis: using the adaptive friction prediction model, the friction data detected in real time is analyzed to obtain the change pattern and trend of the friction; S432, based on predicted friction values Calculate the control parameter Δu that needs to be compensated t , expressed as: Among them, K p is the control gain, is the predicted friction value, y t is the actual friction value detected; S433, adjust the drive command: the calculated compensation value Δu t Applied to the robot's drive instructions, the control parameters are optimized to compensate for the change in friction, expressed as: Among them, u t is the adjusted drive instruction, is the original drive instruction, Δu t is the compensation value.

8. The intelligent control method for an industrial robot according to claim 1, characterized in that: The fault detection and processing in S5 includes: S51, real-time monitoring: using a variety of sensors to collect the robot's operating status data in real time; S52, anomaly detection: using an anomaly detection algorithm based on K-means clustering to identify anomalies in the operating state; S53, fault handling: If an abnormality is detected, the potential fault is handled through a preset processing mechanism.

9. The intelligent control method for an industrial robot according to claim 1, characterized in that: The human-computer interaction in S6 includes: S61, User Interface Design: Design a graphical user interface through which operators can monitor the robot's operating status in real time, input control commands, and view feedback; S62, Natural Language Processing: Integrate natural language processing technology to enable operators to interact with industrial robots, expressed as: P(W|X)=argmax W P(X|W)P(W); Where P(W|X) is the probability of a word sequence W given a speech signal X, P(X|W) is the acoustic model, and P(W) is the language model; I=argmax I P(I|T); Where I is the intent, T is the text input, and P(I|T) is the probability of intent given the text input; S63, Real-time interaction and feedback: Provide real-time interaction and feedback through user interface and natural language processing technology.

10. An intelligent control system for an industrial robot, used to implement an intelligent control method for an industrial robot as claimed in any one of claims 1 to 9, characterized in that: Includes the following modules: Data acquisition and preprocessing module: collects the operating data of the industrial robot in the working environment in real time through a variety of sensors, and preprocesses the collected operating data, including data cleaning, normalization and feature extraction; Model building module: builds the industrial robot control model based on the preprocessed operation data; Real-time control and autonomous learning module: Make predictions and decisions based on real-time collected operation data to achieve real-time control of industrial robots. During the actual operation of the robots, new operation data is continuously collected to regularly update and optimize the control model of industrial robots. Adaptive friction compensation module: During the robot's motion, it detects and compensates for the friction changes of the robot's joints and sliding parts in real time; Fault detection and processing module: During the control process, the robot's operating status is monitored in real time, and potential faults are discovered and processed through anomaly detection algorithms; Human-computer interaction module: Through the user interface and natural language processing technology, the interaction between the operator and the robot is realized.

Images