Industrial robot temperature prediction model training method, prediction method and related device

Abstract

The embodiment of the invention provides a training method and a prediction method of an industrial robot temperature prediction model and a related device. The method comprises the following steps: constructing a training sample; wherein each training sample comprises a plurality of operation process data sets arranged according to a time sequence; wherein the operation process data set comprises a sample temperature value and state parameters corresponding to working states in the operation process of a plurality of industrial robots; training a specified model base based on the training sample to obtain an industrial robot temperature prediction model; wherein the specified model base comprises a long-term dependency relationship extraction layer and a nonlinear feature extraction layer; the long-term dependency relationship extraction layer is used for extracting an interdependency relationship among parameters of the operation process data set; the non-linear feature extraction layer is used for extracting non-linear features of the operation process data set. The accuracy of temperature prediction is improved by capturing the long-term dependency relationship and nonlinear characteristics in the operation process data of the industrial robot.

Description

Technical Field

[0001] The present application relates to the field of temperature prediction, and specifically to a training method, a prediction method, and related devices for an industrial robot temperature prediction model. Background Art

[0002] In modern manufacturing, industrial robots are widely used in a variety of fields, including automotive manufacturing, instrumentation, warehousing and logistics, and semiconductors, playing an indispensable role in modern industrial production. During extended operation, mechanical friction, motor heating, and other factors can cause internal temperature fluctuations in industrial robots, directly impacting their performance and service life. Therefore, accurately predicting temperature fluctuations during industrial robot operation is crucial for maintaining equipment stability, extending service life, and optimizing maintenance strategies.

[0003] Currently, temperature prediction for industrial robots primarily relies on traditional statistical methods and some basic machine learning models. For example, the Autoregressive Integrated Moving Average (ARIMA) model and simple Multilayer Perceptron (MLP) neural networks are used. While these methods can provide basic temperature prediction, they often exhibit insufficient accuracy when dealing with complex nonlinear relationships and long-term data dependencies. Summary of the Invention

[0004] In view of this, multiple embodiments of the present application are dedicated to providing a training method for an industrial robot temperature prediction model, which can improve the temperature prediction accuracy of the industrial robot to a certain extent.

[0005] In a first aspect, an embodiment of the present application provides a method for training an industrial robot temperature prediction model, comprising:

[0006] Construct training samples; wherein each training sample includes multiple operation process data groups arranged in chronological order; wherein the operation process data groups include sample temperature values ​​and state parameters corresponding to the working states of multiple industrial robots during operation; based on the training samples, train a specified model base to obtain an industrial robot temperature prediction model; wherein the specified model base includes a long-term dependency extraction layer and a nonlinear feature extraction layer; the long-term dependency extraction layer is used to extract the mutual dependence between the parameters of the operation process data group; the nonlinear feature extraction layer is used to extract the nonlinear features of the operation process data group.

[0007] Optionally, the step of constructing a training sample includes: obtaining multiple categories of operation process data generated by the industrial robot during operation; arranging the operation process data in chronological order, and forming an operation process data group with different categories of operation process data corresponding to the same time step; dividing the multiple operation process data groups into multiple process data sets based on a specified time length, each process data set includes multiple operation process data groups, and each process data set serves as a training sample.

[0008] Optionally, the categories of the operating process data include at least one of the following in addition to temperature: current, speed, torque or angle.

[0009] Optionally, in the specified model base, the output of the long-term dependency extraction layer serves as the input of the nonlinear feature extraction layer.

[0010] Optionally, the designated model base further includes a fully connected layer, and the output of the long-term dependency extraction layer and the output of the nonlinear feature extraction layer are both used as inputs of the fully connected layer, and the fully connected layer outputs the predicted temperature.

[0011] In a second aspect, an embodiment of the present application also provides an industrial robot temperature prediction method, comprising: obtaining state parameters of the industrial robot during operation to form multiple operation process data groups arranged in chronological order; inputting the multiple operation process data groups into an industrial robot temperature prediction model to obtain a temperature prediction value calculated by the industrial robot temperature prediction model; wherein the industrial robot temperature prediction model is obtained based on the aforementioned industrial robot temperature prediction model training method.

[0012] In a third aspect, an embodiment of the present application also provides a training device for an industrial robot temperature prediction model, comprising: a sample construction module for constructing training samples; wherein each training sample includes multiple operation process data groups arranged in chronological order; wherein the operation process data groups include sample temperature values ​​and state parameters corresponding to the working states of multiple industrial robots during operation; a model training module for training a specified model base based on the training samples to obtain an industrial robot temperature prediction model; wherein the specified model base includes a long-term dependency extraction layer and a nonlinear feature extraction layer; the long-term dependency extraction layer is used to extract the mutual dependence between the sample temperature value and the operation process data group; the nonlinear feature extraction layer is used to extract the nonlinear features of the sample temperature value and the operation process data group.

[0013] In a fourth aspect, an embodiment of the present application also provides an industrial robot temperature prediction device, comprising: a parameter acquisition module, used to obtain state parameters of the industrial robot during operation, forming multiple operation process data groups arranged in chronological order; a temperature prediction module, used to input the multiple operation process data groups into the industrial robot temperature prediction model, and obtain the temperature prediction value obtained by the industrial robot temperature prediction model; wherein, the industrial robot temperature prediction model is obtained based on the aforementioned industrial robot temperature prediction model training method.

[0014] In a fifth aspect, an embodiment of the present application further provides an electronic device, comprising a memory and a processor, wherein the memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor to implement the method as described above.

[0015] In a sixth aspect, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores at least one computer program, and when the at least one computer program is executed by a processor, it can implement the aforementioned method.

[0016] In various embodiments provided herein, a designated model base is constructed using a long-term dependency extraction layer and a nonlinear feature extraction layer. Training samples are then constructed using operational data generated during the operation of an industrial robot. The training samples are then used to train the designated model base to obtain an industrial robot temperature prediction model. This improves the prediction accuracy of the industrial robot temperature prediction model when processing data containing complex nonlinear relationships and long-term dependencies. This industrial robot temperature prediction model training method combines the advantages of the long-term dependency extraction layer and the nonlinear feature extraction layer to capture complex features and patterns in industrial robot operational data, improving the accuracy of temperature prediction. The method also exhibits strong generalization capabilities and is applicable to diverse industrial environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart of a method for training an industrial robot temperature prediction model provided in one embodiment of the present application.

[0018] Figure 2 A flowchart of a method for predicting temperature of an industrial robot provided in one embodiment of the present application.

[0019] Figure 3 A diagram illustrating data set division is provided for one embodiment of the present application.

[0020] Figure 4 A structural diagram of the LSTM-DNN model provided for one embodiment of the present application.

[0021] Figure 5 A comparison chart of LSTM-DNN predictions and actual results provided for one embodiment of this application.

[0022] Figure 6 A schematic diagram of a module of a training device for an industrial robot temperature prediction model provided in one embodiment of the present application.

[0023] Figure 7 A schematic diagram of a module of an industrial robot temperature prediction device provided in one embodiment of the present application.

[0024] Figure 8 A schematic diagram of an electronic device provided in accordance with one embodiment of the present application. DETAILED DESCRIPTION

[0025] 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. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0026] See also Figure 1 . An embodiment of the present application provides a training method for an industrial robot temperature prediction model. The training method for an industrial robot temperature prediction model can be applied to a training device for an industrial robot temperature prediction model. The training device for the industrial robot temperature prediction model can be an electronic device with certain computing capabilities. The electronic device can have a controller and a memory, etc. Of course, in some embodiments, the training device for the industrial robot temperature prediction model can also refer to a program module running in an electronic device. The training method for the industrial robot temperature prediction model can include the following steps.

[0027] Step S110: constructing training samples; wherein each training sample includes a plurality of operation process data groups arranged in chronological order; wherein the operation process data groups include sample temperature values ​​and state parameters corresponding to the working states of the plurality of industrial robots during operation.

[0028] Step S120: Training a designated model base based on the training samples to obtain an industrial robot temperature prediction model; wherein the designated model base includes a long-term dependency extraction layer and a nonlinear feature extraction layer; the long-term dependency extraction layer is used to extract the mutual dependency between the parameters of the operation process data group; the nonlinear feature extraction layer is used to extract the nonlinear features of the operation process data group.

[0029] In this embodiment, the training sample can be a data set used to train a specified model base. A data set can include multiple groups of operational process data arranged in chronological order, and a single operational process data group can include multiple different categories of operational process data at the same time step. The categories of operational process data can include sample temperature values ​​and multiple state parameters corresponding to the working states of the industrial robot during operation. For example, in some embodiments, the aforementioned state parameters can include current, speed, torque, and angle. The training device for the industrial robot temperature prediction model can learn the temporal trends of the sample temperature values ​​and the relationships between the sample temperature values ​​and other state parameters, while simultaneously using some sample temperature values ​​to verify the temperature prediction results, thereby optimizing the specified model base to obtain the industrial robot temperature prediction model. The specified model base can be an untrained neural network model, which can include a long-term dependency extraction layer and a nonlinear feature extraction layer. For example, in some embodiments, the long-term dependency extraction layer can be an LSTM, the nonlinear feature extraction layer can be a DNN, and the specified model base can be an untrained LSTM-DNN model.

[0030] The training device for the industrial robot temperature prediction model can collect multiple operation process data sets during the operation of the industrial robot over a period of time, and preprocess the collected multiple operation data sets to obtain training samples. For example, preprocessing can include processing methods such as data cleaning, data aggregation, and normalization. The training device for the industrial robot temperature prediction model can train a specified model base based on the training samples, learn the long-term dependencies of the sample data with a long-term dependency extraction layer, and extract the nonlinear features of the sample data with a nonlinear feature extraction layer. The long-term dependency extraction layer and the nonlinear feature extraction layer can improve the temperature prediction accuracy of the industrial robot temperature prediction model when processing data containing long-term dependencies and complex nonlinear features. Furthermore, the training device for the industrial robot temperature prediction model can optimize the specified model base based on the training results to obtain the industrial robot temperature prediction model.

[0031] In various embodiments provided herein, multiple sets of collected operational data are preprocessed to generate training samples. A designated model base is trained based on the training samples. A long-term dependency extraction layer within the designated model base learns long-term dependencies within the sample data, and a nonlinear feature extraction layer within the designated model base extracts nonlinear features within the sample data. This training method for an industrial robot temperature prediction model can improve the accuracy of temperature prediction when processing data containing long-term dependencies and complex nonlinear features.

[0032] In some embodiments, a training device for an industrial robot temperature prediction model can obtain multiple categories of operation process data generated by the industrial robot during operation; arrange the operation process data in chronological order, and form an operation process data group with different categories of operation process data corresponding to the same time step; divide the multiple operation process data groups into multiple process data sets based on a specified time length, each process data set includes multiple operation process data groups, and each process data set serves as a training sample.

[0033] In this embodiment, the training device for an industrial robot temperature prediction model can use a digital collector to collect multiple categories of operating process data generated during the operation of the industrial robot over a period of time. The period of time can include multiple time steps, each of which can be an evenly spaced point within the period of time. The training device for the industrial robot temperature prediction model can aggregate the operating process data of different categories at the same time step into a column to generate an operating process data group. Specifically, for example, for a fixed time length, the smaller the time interval between the time steps, the greater the number of time steps, and the greater the number of operating process data groups. At least some of the multiple categories of operating process data included in an operating process data group may affect the temperature of the industrial robot during operation. The training device for the industrial robot temperature prediction model can divide the multiple operating process data groups within the period of time into multiple process data sets, each of which includes multiple operating process data groups, and each process data set serves as a training sample. Specifically, for example, the greater the number of process data groups in a training sample, the smaller the time interval between corresponding time steps, which results in higher accuracy in industrial robot temperature prediction, but also increases the computational complexity. The training device of the industrial robot temperature prediction model can determine the number of operation process data groups contained in a training sample according to actual needs, so as to take into account both the accuracy and training efficiency of the industrial robot temperature prediction model.

[0034] The training sample construction method in this embodiment arranges the collected operational data of multiple categories generated by the industrial robot in chronological order to obtain multiple operational data groups, which are then divided into multiple training samples. This training sample construction method can balance the accuracy and training efficiency of the industrial robot temperature prediction model.

[0035] In some embodiments, the training device for the temperature prediction model of the industrial robot can also perform data cleaning on the operation process data to remove noise and outliers, and then normalize the cleaned operation process data to obtain a high-quality training data set. Furthermore, the training device for the temperature prediction model of the industrial robot can set the time series length and divide the training data set into a training set and a test set according to a certain time length ratio. For example, in some embodiments, refer to Figure 3 , assuming that the total time series length is 10, the ratio of the time length of the training set to the time length of the test set can be 8:2.

[0036] This implementation cleans and normalizes multiple categories of collected operational data generated by industrial robots during operation to generate a high-quality training dataset. This method can improve the accuracy of the industrial robot temperature prediction model and enhance its generalization capabilities across different scenarios.

[0037] In some embodiments, the categories of the operating process data include at least one of the following in addition to temperature: current, speed, torque, or angle.

[0038] In this embodiment, different types of data, such as current, speed, torque, and angle, generated during the operation of an industrial robot, all affect the temperature of the industrial robot during operation. Specifically, for example, an increase in current may lead to an increase in temperature, while changes in speed and angle may cause changes in temperature. Furthermore, different types of data, such as current, speed, torque, and angle, may influence each other. For example, changes in torque may cause changes in current and speed. Furthermore, these different types of operational process data may contain multiple deep-level features, such as time domain features, frequency domain features, and energy features. These deep-level features may affect the temperature prediction value of the industrial robot. For example, the time domain features of current can reflect the variation patterns of current over a time series, thereby affecting the variation of temperature over a time series. Through the aforementioned training sample construction method, the training device for an industrial robot temperature prediction model can maintain the mutual influence of different types of operational process data while retaining the deep-level features of these different types of operational data, thereby obtaining a high-quality training dataset, thereby improving the accuracy and training efficiency of the industrial robot temperature prediction model and enhancing the performance of the industrial robot temperature prediction model.

[0039] The training sample construction method in this embodiment arranges the collected operational data generated by the industrial robot in chronological order, generating multiple operational data groups. These groups are then divided into multiple training samples. This method can minimize the impact of different operational data categories while preserving the underlying characteristics of these data. This training sample construction method can improve the accuracy and training efficiency of the industrial robot temperature prediction model, enhancing its performance.

[0040] In some embodiments, in the specified model base, the output of the long-term dependency extraction layer serves as the input of the nonlinear feature extraction layer.

[0041] During industrial robot operation, different types of process data may be interdependent. For example, an increase in current may lead to a rise in temperature, and a change in torque may cause changes in current and speed. The long-term dependency extraction layer captures the complex interactions between these process data. Similarly, during industrial robot operation, the robot's operating state may exhibit certain periodicities. For example, temperature and torque may peak at specific times of the day or within specific operating cycles. The long-term dependency extraction layer captures the periodic variations in these process data. Furthermore, in some special cases, past events may affect the current and future operating state of the industrial robot. For example, maintenance on an industrial robot may affect its operating state for weeks or months to come. The long-term dependency extraction layer preserves this long-term memory. By learning these long-term dependencies within the industrial robot's operation, the long-term dependency extraction layer can more accurately predict the temporal trends of various process data of the industrial robot.

[0042] The nonlinear feature extraction layer can capture the interactions between different process data during the operation of an industrial robot. For example, the combination of current and speed may have a nonlinear effect on temperature. The nonlinear feature extraction layer can learn this interaction through nonlinear changes in the hidden layer. The nonlinear feature extraction layer can be a multi-layered structure, which allows it to combine lower-level features to form higher-level features. For example, the first layer of the nonlinear feature extraction layer can learn the basic characteristics of the process data, such as current, while deeper layers can combine these features with process data such as temperature and torque to form more complex feature representations. Furthermore, each layer in the multi-layered nonlinear feature extraction layer can perform a nonlinear transformation on the input data. These transformations can be highly complex, allowing the nonlinear feature extraction layer to learn nonlinear trends and relationships in the input data.

[0043] See also Figure 4 In this embodiment, for example, the long-term dependency extraction layer may be an LSTM layer, the nonlinear feature extraction layer may be a DNN layer, and the designated model base may be an LSTM-DNN model, wherein the output of the LSTM layer serves as the input to the DNN layer. Specifically, in some embodiments, the LSTM layer may include two layers, each with the same number of neurons. The output of the first LSTM layer may serve as the input to the second LSTM layer. For example, both LSTM layers may contain 50 neurons. The first LSTM layer returns the entire time series data, i.e., 50 chronologically ordered sets of operational process data processed by the first LSTM layer; the second LSTM layer returns the last output. The two-layer LSTM structure can process more complex input data, improving the designated model base's ability to abstract temporal changes in input data, and enhancing the training efficiency and accuracy of the designated model base. In some embodiments, the DNN layer may be composed of multiple fully connected layers, which further extract complex features from the input data. The DNN layer can learn nonlinear features in the input data, improving the designated model base's ability to fit temperature changes. For example, the first fully connected layer in the DNN layer can contain 100 neurons. The output of the LSTM layer will be input into the first fully connected layer in the DNN layer. The activation function can be ReLU. The number of neurons in each subsequent fully connected layer decreases layer by layer to reduce the complexity of the model base and improve computational efficiency.

[0044] In this embodiment, the input data of the industrial robot temperature prediction model involves multiple categories of operating process data, such as current, speed, torque, angle, and temperature. Therefore, the long-term dependencies and nonlinear features in the input data are relatively complex. In the designated model base, the output of the LSTM layer serves as the input of the DNN layer. On the basis of capturing the long-term dependencies and time-varying trends in the input data, the extracted time features can be further nonlinearly mapped and fused, thereby enhancing the designated model base's ability to understand complex features. In addition, the architecture combining the LSTM layer and the DNN layer can make full use of the long-term dependencies and high-dimensional nonlinear features in the input data to improve the prediction accuracy of the designated model base; connecting the DNN layer on the basis of the LSTM layer can further enhance the generalization ability of the designated model base to adapt to diverse industrial environments.

[0045] This implementation uses the output of the long-term dependency extraction layer as the input to the nonlinear feature extraction layer, building a designated model foundation based on this architecture. This method combines the advantages of the long-term dependency extraction layer and the nonlinear feature extraction layer to enhance the industrial robot temperature prediction model's ability to understand complex features, improving its training efficiency, accuracy, and generalization capabilities.

[0046] In some embodiments, the designated model base further includes a fully connected layer, and the output of the long-term dependency extraction layer and the output of the nonlinear feature extraction layer are both used as inputs of the fully connected layer, and the fully connected layer outputs the predicted temperature.

[0047] In this embodiment, a fully connected layer fuses the features extracted by the long-term dependency extraction layer and the nonlinear feature extraction layer to output a predicted temperature. In some embodiments, a training device for an industrial robot temperature prediction model can use an activation function to increase the nonlinear capability of a specified model base when outputting the temperature prediction result from the fully connected layer, thereby improving the accuracy of temperature prediction.

[0048] This implementation uses a fully connected layer to fuse and output the features extracted by the long-term dependency extraction layer and the nonlinear feature extraction layer. This method can enhance the model's nonlinear capabilities, enabling a given model base to capture more complex data features and improving the accuracy of the industrial robot temperature prediction model.

[0049] In some embodiments, the training device for the industrial robot temperature prediction model can train a specified model base using training samples from the aforementioned training set, validate the model using data from the aforementioned test set, employ mean squared error as the loss function, and perform iterative optimization using an Adam optimizer to obtain the industrial robot temperature prediction model. The performance of the industrial robot temperature prediction model can be evaluated using the following evaluation metrics.

[0050] (1) Mean Squared Error (MSE)

[0051]

[0052] MSE is the average of the squares of the errors between the predicted values ​​and the true values, where y i is the actual value, is the predicted value and n is the number of samples.

[0053] (2) Mean Absolute Error (MAE)

[0054]

[0055] MAE is the mean absolute error between the predicted value and the actual value, where y i is the actual value, is the predicted value and n is the number of samples.

[0056] (3) Coefficient of determination (R 2 , Coefficient of Determination)

[0057]

[0058] R 2 Used to measure the goodness of fit of the model, where y i is the actual value, is the predicted value, is the mean of the actual values, and n is the number of samples.

[0059] See also Figure 5 In some embodiments, the industrial robot temperature prediction model can be a trained LSTM-DNN model. The temperature prediction results of the LSTM-DNN are compared with the actual results. Figure 5 As shown, the corresponding evaluation index calculation values ​​are as follows:

[0060] MSE=0.002118424707690151

[0061] MAE=0.013357529887547093

[0062] R 2 =0.9967894380287233

[0063] The values ​​of MSE and MAE are extremely small, and R 2 The value of is very close to 1, which shows that the temperature prediction model achieves a relatively ideal prediction effect.

[0064] See also Figure 2 . An embodiment of the present application provides an industrial robot temperature prediction method. The industrial robot temperature prediction method can be applied to an industrial robot temperature prediction device. The industrial robot temperature prediction device can be an electronic device with certain computing capabilities. The electronic device can have a controller and a memory, etc. In some embodiments, the industrial robot temperature prediction device can also refer to a program module running in the electronic device. The industrial robot temperature prediction method can include the following steps.

[0065] Step S210: Acquire state parameters of the industrial robot during operation to form a plurality of operation process data groups arranged in chronological order.

[0066] Step S220: inputting the multiple operation process data groups into the industrial robot temperature prediction model to obtain the temperature prediction value obtained by the industrial robot temperature prediction model; wherein, the industrial robot temperature prediction model is obtained based on the aforementioned industrial robot temperature prediction model training method.

[0067] In this embodiment, the industrial robot temperature prediction device can collect data corresponding to multiple state parameters during the operation of the industrial robot in real time, and preprocess the data corresponding to the multiple state parameters to obtain multiple operation data groups arranged in chronological order. The multiple state parameters may include current, speed, torque, angle, and temperature, and the preprocessing may include the aforementioned data cleaning, data aggregation, and normalization. The industrial robot temperature prediction device can input the multiple operation process data groups into the industrial robot temperature prediction model to obtain the temperature prediction value obtained by the industrial robot temperature prediction model. For example, in some embodiments, the industrial robot temperature prediction model can be an LSTM-DNN neural network trained using the training method of the industrial robot temperature prediction model.

[0068] See also Figure 6 One embodiment of the present application further provides a training device for an industrial robot temperature prediction model, comprising: a sample construction module for constructing training samples; wherein each training sample includes a plurality of operation process data groups arranged in chronological order; wherein the operation process data groups include sample temperature values ​​and state parameters corresponding to the working states of a plurality of industrial robots during operation; a model training module for training a specified model base based on the training samples to obtain an industrial robot temperature prediction model; wherein the specified model base includes a long-term dependency extraction layer and a nonlinear feature extraction layer; the long-term dependency extraction layer is used to extract the mutual dependency between the sample temperature value and the operation process data group; and the nonlinear feature extraction layer is used to extract the nonlinear features of the sample temperature value and the operation process data group.

[0069] In this embodiment, the specific functions and effects achieved by the CNC machine tool control device can be explained by referring to other embodiments of the present application and will not be repeated here.

[0070] See also Figure 7One embodiment of the present application further provides an industrial robot temperature prediction device, comprising: a state parameter acquisition module, configured to acquire state parameters of the industrial robot during operation, forming a plurality of operation process data groups arranged in chronological order; a temperature prediction module, configured to input the plurality of operation process data groups into an industrial robot temperature prediction model, and obtain a temperature prediction value obtained by the industrial robot temperature prediction model; wherein the industrial robot temperature prediction model is obtained based on the aforementioned industrial robot temperature prediction model training method.

[0071] In this embodiment, the specific functions and effects achieved by the CNC machine tool control device can be explained by referring to other embodiments of the present application and will not be repeated here.

[0072] The embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the processor implements the method as described above.

[0073] The embodiments of the present application further provide a computer program product comprising instructions, which implements the aforementioned method when executed by a processor.

[0074] See also Figure 8 . An embodiment of the present application may provide an electronic device, comprising: a memory, and one or more processors communicatively connected to the memory; the memory stores instructions executable by the one or more processors, and the instructions are executed by the one or more processors to enable the one or more processors to implement the aforementioned method.

[0075] In some embodiments, the electronic device may include a processor, a storage medium, and a communication interface connected by a system bus. The storage medium may store a related computer program.

[0076] The user information or user account information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, etc.) involved in multiple implementation methods of this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0077] It should be understood that the specific examples herein are only intended to help those skilled in the art better understand the embodiments of the present application, rather than to limit the scope of the present invention.

[0078] It can be understood that in the various implementation methods of this application, the size of the serial number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the implementation method of this application.

[0079] It can be understood that the various embodiments described in this application can be implemented individually or in combination, and the embodiments of this application are not limited to this.

[0080] Unless otherwise indicated, all technical and scientific terms used in the embodiments of the present application have the same meaning as those commonly understood by those skilled in the art in the technical field of the present application. The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit the scope of this application. The term "and / or" used in this application includes any and all combinations of one or more related listed items. The singular forms "a", "above", and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates otherwise.

[0081] It is understood that the processor in the embodiments of the present application can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above-mentioned method embodiment can be completed by hardware integrated logic circuits in the processor or software instructions. The above-mentioned processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The various methods, steps, and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above-mentioned method.

[0082] It will be understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (programmable ROM, PROM), an erasable programmable read-only memory (erasable PROM, EPROM), an electrically erasable programmable read-only memory (EEPROM) or flash memory. The volatile memory may be a random access memory (RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0083] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0084] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, devices and units can refer to the corresponding processes in the aforementioned method implementation methods and will not be repeated here.

[0085] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0086] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of this embodiment.

[0087] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0088] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling an electronic device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0089] The above description is merely a specific embodiment of the present application, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A training method for an industrial robot temperature prediction model, characterized in that: The method comprises: Constructing training samples; wherein each training sample includes a plurality of operation process data groups arranged in chronological order; wherein the operation process data groups include sample temperature values ​​and state parameters corresponding to the working states of the plurality of industrial robots during operation; A designated model base is trained based on the training samples to obtain an industrial robot temperature prediction model; wherein the designated model base includes a long-term dependency extraction layer and a nonlinear feature extraction layer; the long-term dependency extraction layer is used to extract the mutual dependence between the parameters of the operation process data group; the nonlinear feature extraction layer is used to extract the nonlinear features of the operation process data group.

2. The method according to claim 1, characterized in that The steps to construct training samples include: Acquiring multiple categories of operation process data generated by the industrial robot during operation; Arrange the operation process data in chronological order, and form an operation process data group with different categories of operation process data corresponding to the same time step; The plurality of operating process data groups are divided into a plurality of process data sets based on a specified time length, each process data set includes a plurality of operating process data groups, and each process data set serves as a training sample.

3. The method according to claim 2, characterized in that The categories of the operating process data include at least one of the following in addition to temperature: current, speed, torque or angle.

4. The method according to claim 1, wherein In the specified model base, the output of the long-term dependency extraction layer serves as the input of the nonlinear feature extraction layer.

5. The method according to claim 4, characterized in that The designated model base also includes a fully connected layer. The output of the long-term dependency extraction layer and the output of the nonlinear feature extraction layer are both used as inputs of the fully connected layer, and the fully connected layer outputs the predicted temperature.

6. A temperature prediction method for an industrial robot, characterized in that: include: Acquire state parameters of the industrial robot during operation to form multiple operation process data groups arranged in chronological order; The multiple operation process data groups are input into the industrial robot temperature prediction model to obtain the temperature prediction value obtained by the industrial robot temperature prediction model; wherein, the industrial robot temperature prediction model is obtained based on the training method of the industrial robot temperature prediction model as described in any one of claims 1 to 5.

7. A training device for an industrial robot temperature prediction model, characterized in that: include: A sample construction module is used to construct training samples; wherein each training sample includes multiple operation process data groups arranged in chronological order; wherein the operation process data groups include sample temperature values ​​and state parameters corresponding to the working states of multiple industrial robots during operation; A model training module is used to train a specified model base based on the training samples to obtain an industrial robot temperature prediction model; wherein the specified model base includes a long-term dependency extraction layer and a nonlinear feature extraction layer; the long-term dependency extraction layer is used to extract the mutual dependency between the sample temperature values ​​and the operation process data group; the nonlinear feature extraction layer is used to extract the nonlinear features of the sample temperature values ​​and the operation process data group.

8. A temperature prediction device for an industrial robot, characterized in that: include: A parameter acquisition module is used to obtain the state parameters of the industrial robot during operation and form multiple operation process data groups arranged in chronological order; A temperature prediction module is used to input the multiple operation process data groups into the industrial robot temperature prediction model to obtain the temperature prediction value obtained by the industrial robot temperature prediction model; wherein, the industrial robot temperature prediction model is obtained based on the training method of the industrial robot temperature prediction model as described in any one of claims 1 to 5.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, wherein at least one computer program is stored in the memory, and the at least one computer program is loaded and executed by the processor to implement the method according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores at least one computer program, and when the at least one computer program is executed by a processor, it can implement the method according to any one of claims 1 to 6.

Images