CN116277040A

CN116277040A - Mechanical arm vibration suppression method, device, equipment and medium based on deep learning

Info

Publication number: CN116277040A
Application number: CN202310583831.5A
Authority: CN
Inventors: 赵伟峰; 何厥兴; 张建民; 张义; 万云辉
Original assignee: Wuhu Longshen Robot Co ltd; Foshan Longshen Robot Co Ltd
Current assignee: Wuhu Longshen Robot Co ltd; Foshan Longshen Robot Co Ltd
Priority date: 2023-05-23
Filing date: 2023-05-23
Publication date: 2023-06-23
Anticipated expiration: 2043-05-23
Also published as: CN116277040B

Abstract

The application discloses a mechanical arm vibration suppression method, device, equipment and medium based on deep learning, relates to the technical field of industrial robots, and the mechanical arm vibration suppression method based on deep learning comprises the following steps: collecting state data of each rotating shaft of the mechanical arm; inputting the state data into a target state detection model to judge whether the mechanical arm has abnormal vibration, wherein the target state detection model is obtained by training a plurality of groups of state data with state labels; if abnormal vibration exists in the mechanical arm, inputting the state data into a target vibration detection model to obtain a vibration label, wherein the target vibration detection model is obtained through training according to state data corresponding to multiple groups of mechanical arms in an abnormal vibration state; and inquiring a corresponding vibration compensation signal according to the vibration tag, and inhibiting abnormal vibration of the mechanical arm according to the vibration compensation signal. The technical problem of the low suppression efficiency of arm vibration has been solved to this application.

Description

Mechanical arm vibration suppression method, device, equipment and medium based on deep learning

Technical Field

The application relates to the field of industrial robots, in particular to a method, a device, equipment and a medium for inhibiting vibration of a mechanical arm based on deep learning.

Background

Industrial robots are an important component of the manufacturing industry in the modern industry and are widely applied to the production and manufacturing of machining, welding, assembly and the like. The mechanical arm of the industrial robot is also applied to some high-precision industries due to the advantages of multiple degrees of freedom, large working space, repeated programming, automatic control and the like, and under the condition, the requirement on the working stability of the mechanical arm is higher, for example, the product quality and the working efficiency are seriously affected once the mechanical arm has abnormal shake. In vibration detection and suppression of the mechanical arm, it is generally determined whether the mechanical arm generates abnormal vibration by using manual detection through experience or related indexes, and the mechanical arm is debugged by experience to suppress the abnormal vibration of the mechanical arm. The method cannot realize qualitative and quantitative analysis of the vibration of the mechanical arm, and needs repeated debugging, so that the inhibition efficiency of the vibration of the mechanical arm is lower.

Disclosure of Invention

The main aim of the application is to provide a method, a device, equipment and a medium for inhibiting the vibration of a mechanical arm based on deep learning, which aim to solve the technical problem of low inhibiting efficiency of the vibration of the mechanical arm.

In order to achieve the above object, the present application provides a method for suppressing vibration of a mechanical arm based on deep learning, the method for suppressing vibration of a mechanical arm based on deep learning comprising:

collecting state data of each rotating shaft of the mechanical arm;

inputting the state data into a target state detection model to judge whether the mechanical arm has abnormal vibration, wherein the target state detection model is obtained by training a plurality of groups of state data with state labels;

if abnormal vibration exists in the mechanical arm, inputting the state data into a target vibration detection model to obtain a vibration label, wherein the target vibration detection model is obtained by training according to state data corresponding to multiple groups of mechanical arms in abnormal vibration states, and each group of state data is provided with a corresponding vibration label;

and inquiring a corresponding vibration compensation signal according to the vibration tag, and inhibiting abnormal vibration of the mechanical arm according to the vibration compensation signal.

Optionally, the state data includes motion state data, amplitude data, and temperature data, and the step of collecting state data at each rotation axis of the mechanical arm includes:

collecting motion state data of each rotating shaft through a real-time controller, wherein the motion state data at least comprises speed, position, acceleration and moment;

Collecting amplitude data of each rotating shaft through a vibration sensor;

and acquiring temperature data of each rotating shaft through a temperature sensor.

Optionally, before the step of inputting the state data into a target state detection model to determine whether the mechanical arm has abnormal vibration, the method further includes:

acquiring first historical state data of the mechanical arm, wherein the first historical state data comprises state data corresponding to a plurality of groups of time points, each group of state data is provided with a corresponding state label, the state labels are determined according to the real state of the mechanical arm at each group of time points, and the state labels comprise normal labels and abnormal labels;

the proportion of the state data of the normal state label and the state data of the abnormal state label in the first historical state data is adjusted according to a preset state proportion, and a first target state data set is obtained;

training a preset state detection model according to the first target state data set to obtain a target state detection model.

Optionally, the training the preset state detection model according to the first target state data set, and the step of obtaining the target state detection model includes:

Dividing the first target state data set into a training set and a verification set;

inputting the training set into the preset state detection model, and iteratively optimizing model parameters of the preset state detection model;

inputting the verification set into an optimized preset state detection model, and determining the prediction precision of the preset state detection model;

if the prediction precision is smaller than the preset precision threshold, returning to the execution step: inputting the training set into the preset state detection model, and iteratively optimizing model parameters of the preset state detection model;

and if the prediction precision is greater than or equal to the preset precision threshold, setting the preset state detection model as a target state detection model.

Optionally, before the step of inputting the state data into the target vibration detection model to obtain a vibration tag, the method further includes:

acquiring second historical state data of the mechanical arm when abnormal vibration exists, wherein the second historical state data comprises state data corresponding to a plurality of groups of time points, each group of state data is provided with a corresponding vibration label, and the vibration label is determined according to the vibration type corresponding to each group of state data;

The proportion of state data corresponding to various vibration tags in the second historical state data is adjusted according to a preset type proportion, and a second target state data set is obtained;

training a preset vibration detection model according to the second target state data set to obtain a target vibration detection model.

Optionally, the step of training a preset vibration detection model according to the second target state data set, and obtaining the target vibration detection model includes:

dividing the second target state data set into a training set and a verification set;

inputting the training set into the preset vibration detection model, and iteratively optimizing model parameters of the preset vibration detection model;

inputting the verification set into an optimized preset vibration detection model to obtain a corresponding predicted value;

constructing an confusion matrix based on the predicted value and the vibration label corresponding to the predicted value;

calculating a performance index of the preset vibration detection model based on the confusion matrix;

and if the performance index of the preset vibration detection model accords with a preset index threshold, setting the preset vibration detection model as a target vibration detection model.

Optionally, the step of querying a corresponding vibration compensation signal according to the vibration tag and suppressing abnormal vibration of the mechanical arm according to the vibration compensation signal includes:

Inquiring a vibration compensation signal corresponding to the vibration tag in a preset vibration compensation mapping table, wherein the preset vibration compensation mapping table is constructed according to effective vibration compensation signals corresponding to the vibration tags in a historical vibration inhibition process;

and sending the vibration compensation signal to a real-time controller corresponding to the mechanical arm, and outputting a control signal of the vibration compensation signal to the mechanical arm through the real-time controller so as to inhibit abnormal vibration of the mechanical arm.

The application also provides a mechanical arm vibration suppression device based on deep learning, mechanical arm vibration suppression device based on deep learning is applied to mechanical arm vibration suppression equipment based on deep learning, mechanical arm vibration suppression device based on deep learning includes:

the data acquisition module is used for acquiring state data of each rotating shaft of the mechanical arm;

the abnormal detection module is used for inputting the state data into a target state detection model to judge whether the mechanical arm has abnormal vibration or not, wherein the target state detection model is obtained by training a plurality of groups of state data with state labels;

The vibration analysis module is used for inputting the state data into a target vibration detection model to obtain a vibration label if the mechanical arm has abnormal vibration, wherein the target vibration detection model is obtained by training according to state data corresponding to the mechanical arm in a plurality of groups of abnormal vibration states, and each group of state data is provided with a corresponding vibration label;

and the vibration suppression module is used for inquiring the corresponding vibration compensation signal according to the vibration label and suppressing abnormal vibration of the mechanical arm according to the vibration compensation signal.

Optionally, the data acquisition module is further configured to:

collecting amplitude data of each rotating shaft through a vibration sensor;

Optionally, the anomaly detection module is further configured to:

Optionally, the vibration analysis module is further configured to:

Optionally, the vibration suppression module is further configured to:

The application also provides an electronic device, which is an entity device, and includes: the device comprises a memory, a processor and a program of the deep learning-based mechanical arm vibration suppression method, wherein the program of the deep learning-based mechanical arm vibration suppression method is stored in the memory and can run on the processor, and the program of the deep learning-based mechanical arm vibration suppression method can realize the steps of the deep learning-based mechanical arm vibration suppression method when being executed by the processor.

The present application also provides a computer-readable storage medium having stored thereon a program for implementing the deep learning-based mechanical arm vibration suppression method, which when executed by a processor, implements the steps of the deep learning-based mechanical arm vibration suppression method as described above.

The present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of a deep learning based mechanical arm vibration suppression method as described above.

The application provides a mechanical arm vibration suppression method, device, equipment and medium based on deep learning, firstly, state data of each rotating shaft of a mechanical arm are collected, then the state data are input into a target state detection model to judge whether abnormal vibration exists in the mechanical arm, wherein the target state detection model is obtained through training according to a plurality of groups of state data with state labels, if abnormal vibration exists in the mechanical arm, the state data are input into a target vibration detection model to obtain a vibration label, the target vibration detection model is obtained through training according to state data corresponding to the mechanical arm in a plurality of groups of abnormal vibration states, each group of state data is provided with a corresponding vibration label, the corresponding vibration compensation signal is inquired according to the vibration compensation signal, and abnormal vibration of the mechanical arm is suppressed according to the vibration compensation signal.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application.

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the description of the embodiments or the prior art will be briefly described below, and it will be obvious to those skilled in the art that other drawings can be obtained from these drawings without inventive effort.

Fig. 1 is a schematic flow chart of a first embodiment of a method for suppressing vibration of a mechanical arm based on deep learning;

fig. 2 is a schematic flow chart of a second embodiment of a method for suppressing vibration of a mechanical arm based on deep learning;

FIG. 3 is a schematic flow chart of a third embodiment of a method for damping vibration of a robotic arm based on deep learning;

fig. 4 is a schematic diagram of a composition structure of a mechanical arm vibration suppression device based on deep learning in an embodiment of the present application;

fig. 5 is a schematic device structure diagram of a hardware operating environment related to a deep learning-based mechanical arm vibration suppression method in an embodiment of the present application.

The implementation, functional features and advantages of the present application will be further described with reference to the accompanying drawings in conjunction with the embodiments.

Detailed Description

In order to make the above objects, features and advantages of the present application more comprehensible, the following description will make the technical solutions of the embodiments of the present application clear and complete with reference to the accompanying drawings of the embodiments of the present application. It will be apparent that the described embodiments are only some, but not all, of the embodiments of the present application. All other embodiments, based on the embodiments herein, which are within the scope of the protection of the present application, will be within the purview of one of ordinary skill in the art without the exercise of inventive faculty.

Example 1

In a first embodiment of the present application, referring to fig. 1, the method for suppressing the vibration of the mechanical arm based on the deep learning includes:

step S10, collecting state data of each rotating shaft of the mechanical arm;

step S20, inputting the state data into a target state detection model to judge whether the mechanical arm has abnormal vibration, wherein the target state detection model is obtained by training a plurality of groups of state data with state labels;

Step S30, if abnormal vibration exists in the mechanical arm, inputting the state data into a target vibration detection model to obtain a vibration label, wherein the target vibration detection model is obtained by training according to state data corresponding to the mechanical arm in a plurality of groups of abnormal vibration states, and each group of state data is provided with a corresponding vibration label;

and S40, inquiring a corresponding vibration compensation signal according to the vibration label, and inhibiting abnormal vibration of the mechanical arm according to the vibration compensation signal.

In this embodiment of the present application, it should be noted that the state data includes motion state data, amplitude data and temperature data at all rotation axes of a mechanical arm of an industrial robot, the motion state of the mechanical arm is determined by the motion data, how high the vibration degree of the mechanical arm can be determined by the amplitude data, the temperature data can represent how much the vibration of the mechanical arm affects the rotation axes, and also represents how severe the vibration is, specifically, the motion state data includes a motion speed, a coordinate position, a motion acceleration and a moment. The target state detection model is obtained through training of a neural network model based on deep learning and used for detecting whether abnormal vibration exists in the mechanical arm, and the state labels comprise normal labels and abnormal labels, are obtained through manually analyzing, judging and labeling the state data in historical state data of a plurality of groups of mechanical arms and are used as real labels in the supervised model training process. In addition, the target vibration detection model is a vibration label of the mechanical arm, which is obtained by training based on a neural network model of deep learning and is used for detecting the existence of abnormal vibration, the vibration label can comprise reasons of vibration, such as loose structure, circuit faults, improper control parameters, external interference and the like, and can also comprise specific vibration attributes of abnormal vibration, such as transverse abnormal vibration, longitudinal abnormal vibration, circular arc abnormal vibration, irregular abnormal vibration, slight vibration, severe abnormal vibration and the like, for example, when the mechanical arm has abnormal vibration at a certain time, the detected vibration label is (improper control parameters, transverse abnormal vibration and slight abnormal vibration), and before the state of the mechanical arm is detected, vibration compensation signals are pre-debugged by manpower for the mechanical arm corresponding to each vibration label and are stored in a database.

According to the technical scheme, the abnormal condition of the mechanical arm is detected and analyzed step by step through the abnormal detection and the label detection respectively, and finally vibration suppression is carried out according to the queried vibration compensation signals, wherein the target state detection model only needs to detect whether the state of the mechanical arm is normal or not, the detection speed is high, the target vibration detection model needs to analyze the corresponding label in detail, the detection time is long, the follow-up detection step is not needed for the label in the normal state, and compared with the analysis of all vibration state data of the mechanical arm by using a single target vibration detection model, the time is saved, and the vibration suppression efficiency of the mechanical arm is improved.

As an example, steps S10 to S40 include: acquiring motion state data, amplitude data and temperature data of each rotating shaft of the mechanical arm through a real-time controller and a sensor; inputting the motion state data, the amplitude data and the temperature data into a pre-trained target state detection model, and analyzing the input motion state data, amplitude data and temperature data through the target state detection model to obtain a mechanical arm state detection result, wherein the target state detection model is obtained through training according to a plurality of groups of state data with state labels, and the state labels comprise normal and abnormal states; if the state detection result of the mechanical arm is abnormal, the abnormal vibration of the mechanical arm is represented, the motion state data, the amplitude data and the temperature data are input into a pre-trained target vibration detection model, the input motion state data, the amplitude data and the temperature data are analyzed through the target vibration detection model, and a vibration label corresponding to the mechanical arm is obtained, wherein the target vibration detection model is obtained through training according to state data corresponding to the mechanical arm in a plurality of groups of abnormal vibration states, each group of state data is provided with a corresponding vibration label, and the vibration label is used for representing vibration reasons and vibration attributes of the state data; inquiring a vibration compensation signal corresponding to the vibration tag in a preset database according to the vibration tag, and inputting the vibration compensation signal to a real-time controller of the mechanical arm so as to inhibit abnormal vibration of the mechanical arm.

Specifically, the state data includes motion state data, amplitude data and temperature data, and the step of collecting state data at each rotation axis of the mechanical arm includes:

step S11, acquiring motion state data of each rotating shaft through a real-time controller, wherein the motion state data at least comprises speed, position, acceleration and moment;

step S12, collecting amplitude data of each rotating shaft through a vibration sensor;

and S13, acquiring temperature data of each rotating shaft through a temperature sensor.

In this embodiment of the present application, it should be noted that the real-time controller is configured to control a workflow of the mechanical arm by controlling rotation of each rotation axis of the mechanical arm, where the motion state data may be extracted by extracting state feedback data received by the real-time controller. Wherein the position may be represented by coordinates of the rotational axis in a spatial coordinate system established with a certain stationary point on the robotic arm.

As an example, steps S11 to S13 include: determining the corresponding speed, position, acceleration and moment of each rotating shaft through the state feedback data received by the real-time controller; acquiring amplitude data of each rotating shaft through vibration sensors arranged on the rotating shafts; and acquiring the temperature of each rotating shaft motor through a temperature sensor to obtain temperature data.

In addition, the step of inquiring the corresponding vibration compensation signal according to the vibration tag and suppressing abnormal vibration of the mechanical arm according to the vibration compensation signal includes:

step S41, inquiring a vibration compensation signal corresponding to the vibration tag in a preset vibration compensation mapping table, wherein the preset vibration compensation mapping table is constructed according to effective vibration compensation signals corresponding to the vibration tags in a historical vibration inhibition process;

and step S42, the vibration compensation signal is sent to a real-time controller corresponding to the mechanical arm, and a control signal of the vibration compensation signal is output to the mechanical arm through the real-time controller so as to inhibit abnormal vibration of the mechanical arm.

In this embodiment of the present application, it should be noted that the preset vibration compensation mapping table is stored in the database, and may be a two-dimensional mapping table, where the two-dimensional mapping table includes each vibration tag and vibration compensation signals corresponding to each vibration tag, where the vibration compensation signals are vibration compensation signals that can be summarized and used to effectively inhibit abnormal vibration of the mechanical arm when abnormal vibration conditions corresponding to each vibration tag are encountered manually in long-term working practice. In the method of the embodiment of the application, after the current vibration label of the mechanical arm is analyzed through the target vibration detection model, the corresponding vibration compensation signal can be inquired, the mechanical arm with abnormal vibration can be quickly subjected to vibration suppression, the timeliness of vibration suppression is improved, and adverse effects such as mechanical damage and production quality reduction caused by the long-term abnormal vibration of the mechanical arm are avoided.

As an example, step S41 to step S42 include: inquiring corresponding vibration compensation signals according to a preset vibration compensation mapping table of a vibration tag in a database, wherein the preset vibration compensation mapping table is constructed according to effective vibration compensation signals corresponding to the vibration tags in the vibration inhibition process of the mechanical arm, and the effective vibration compensation signals are obtained through manual repeated debugging; and inputting the vibration compensation signal into a real-time controller corresponding to the mechanical arm, and controlling the motion state of the mechanical arm through the real-time controller according to the vibration compensation signal so as to inhibit abnormal vibration.

The embodiment of the application provides a mechanical arm vibration suppression method based on deep learning, firstly, state data of each rotating shaft of a mechanical arm are collected, then the state data are input into a target state detection model to judge whether abnormal vibration exists in the mechanical arm, wherein the target state detection model is obtained through training according to a plurality of groups of state data with state labels, if abnormal vibration exists in the mechanical arm, the state data are input into a target vibration detection model to obtain vibration labels, wherein the target vibration detection model is obtained through training according to the state data corresponding to the mechanical arm in a plurality of groups of abnormal vibration states, each group of state data is provided with a corresponding vibration label, corresponding vibration compensation signals are inquired according to the vibration labels, and abnormal vibration of the mechanical arm is suppressed according to the vibration compensation signals.

Example two

Further, in another embodiment of the present application, the same or similar content as the first embodiment may be referred to the above description, and will not be repeated. On this basis, referring to fig. 2, before the step of inputting the state data into a target state detection model to determine whether there is abnormal vibration of the robot arm, the method further includes:

step A10, acquiring first historical state data of the mechanical arm, wherein the first historical state data comprises state data corresponding to a plurality of groups of time points, each group of state data is provided with a corresponding state label, the state labels are determined according to the real state of the mechanical arm at each group of time points, and the state labels comprise normal labels and abnormal labels;

step A20, adjusting the proportion of the state data of the normal state label and the state data of the abnormal state label in the first historical state data according to a preset state proportion to obtain a first target state data set;

and step A30, training a preset state detection model according to the first target state data set to obtain a target state detection model.

In this embodiment of the present application, it should be noted that, the first historical state data is state data of each rotating shaft of a plurality of groups of mechanical arms based on a time point collected during a working process of the mechanical arms, where the state data includes state data of a normal state and state data of abnormal vibration, the real state includes a normal state and an abnormal state, the real state corresponds to a normal label and an abnormal label in a state label, and the abnormal state is state data of abnormal vibration. The preset proportion is used for adjusting the state data proportion of the normal tag and the state data proportion of the abnormal tag in the first historical state data to be in accordance with the proportion condition under the natural condition, for example, the probability of abnormal vibration of a certain mechanical arm during working is 10%, and the preset proportion can be set to be 9:1.

In addition, the preset state detection model is a neural network identification model, and the structure of the preset state detection model can comprise a full connection layer, a hidden layer, a full connection layer, an activation layer and the like; in addition, considering that the rotation shafts of the mechanical arm are connected through a mechanical structure, the rotation shafts can be mutually influenced, and abnormal vibration is possibly caused by the combined action of a plurality of rotation shafts, so that the state data of all the rotation shafts is taken as input; the self-attention mechanism is added, so that the preset state detection model automatically learns the input features needing attention when detecting whether the rotating shaft generates abnormal vibration or not. Examples: intuitively, abnormal vibration of a single rotating shaft is most relevant to states of other rotating shafts closest to the rotating shaft, and difficulty in modeling is great in quantifying the abnormal vibration, and through a deep neural network recognition model introducing an attention mechanism, the preset state detection model can learn local focusing characteristics by itself, so that whether abnormal vibration exists in the mechanical arm can be detected more accurately.

As an example, steps a10 to a30 include: acquiring state data corresponding to the mechanical arm at different time points respectively and manually marking the state data with state labels to acquire first historical state data, wherein the state labels comprise normal labels and abnormal labels; the method comprises the steps of adjusting the proportion of state data of a normal state label and state data of an abnormal state label in first historical state data according to a preset state proportion to obtain a first target state data set, wherein the preset state proportion is determined according to the probability of abnormal vibration of a mechanical arm under natural conditions; dividing the first target state data set into a training set and a verification set; constructing a preset state detection model based on a deep learning algorithm; training the preset state detection model according to the training set to optimize model parameters of the preset state detection model; evaluating whether the prediction precision of the preset state detection model accords with a preset threshold value according to the verification set; if the prediction precision of the preset state detection model accords with the preset threshold, stopping training and setting the preset state detection model as a target state detection model; and if the prediction precision of the preset state detection model does not accord with the preset threshold value, returning to the execution step, wherein the preset state detection model is trained according to the training set so as to optimize the model parameters of the preset state detection model.

Specifically, the training the preset state detection model according to the first target state data set, and the step of obtaining the target state detection model includes:

step A31, dividing the first target state data set into a training set and a verification set;

step A32, inputting the training set into the preset state detection model, and iteratively optimizing model parameters of the preset state detection model;

step A33, inputting the verification set into an optimized preset state detection model, and determining the prediction precision of the preset state detection model;

step a34, if the prediction precision is less than the preset precision threshold, returning to the execution step: inputting the training set into the preset state detection model, and iteratively optimizing model parameters of the preset state detection model;

and step A35, setting the preset state detection model as a target state detection model if the prediction precision is greater than or equal to the preset precision threshold.

In this embodiment of the present application, it should be noted that, in the process of dividing the first target state data set into the training set and the verification set, the sample data amounts corresponding to the training set and the verification set respectively may be set according to requirements, for example, when a supervised model is trained, a ratio of the sample data amounts between the training set and the verification set is generally 9: the training set is used for inputting a preset state detection model to be trained so as to continuously perform iterative optimization on model parameters of the preset state detection model, and the model parameters specifically comprise a learning rate, a regularization parameter, the number of layers of a neural network, the number of neurons in each hidden layer, the number of iterations and the like as an example.

As an example, steps a31 to a35 include: dividing each group of state data in the first target state data set into a training set and a verification set, wherein the group number ratio of the group data of the state data in the training set to the group data of the state data in the verification set is 9:1; training the preset state detection model by using the state data in the training set as input by adopting a gradient descent method and iteratively optimizing model parameters of the preset state detection model until the preset state detection model converges; inputting each group of state data in the verification set into a converged preset state detection model to obtain a predicted value corresponding to each state data; judging a first number of accurate prediction in each predicted value according to the state label corresponding to each state data; calculating the ratio between the first number and the total group number of each state data to obtain prediction precision; judging whether the prediction precision is larger than or equal to a preset precision threshold value; if the prediction precision is smaller than the preset precision threshold, returning to the execution step A32: inputting the training set into the preset state detection model, and iteratively optimizing model parameters of the preset state detection model; and if the prediction precision is greater than or equal to the preset precision threshold, stopping training and setting the preset state detection model as a target state detection model.

The embodiment of the application provides a training method of a target state detection model before the step of inputting state data into the target state detection model to judge whether abnormal vibration exists in the mechanical arm, first historical state data of the mechanical arm is obtained, the first historical state data comprise state data corresponding to a plurality of groups of time points, each group of state data are provided with corresponding state labels, the state labels are determined according to the real state of the mechanical arm at each group of time points, each state label comprises a normal label and an abnormal label, the proportion of the state data of the normal state label and the state data of the abnormal state label in the first historical state data is adjusted according to a preset state proportion, a first target state data set is obtained, and finally the preset state detection model is trained according to the first target state data set, so that the target state detection model is obtained. According to the technical scheme, the state data of each rotating shaft of the mechanical arm and the manually marked state label are adopted to train the supervised neural network identification model, so that the target state detection model which has the prediction accuracy meeting the requirement and can detect whether the mechanical arm generates abnormal vibration according to the current state data is obtained, the automation of the abnormal detection of the mechanical arm is realized, the real-time requirement can be met, the method is more stable and higher in real-time than the method for manually judging the abnormal vibration of the mechanical arm, the detection result can be timely obtained after the abnormal vibration occurs, and the negative influence caused by the long-term abnormal vibration of the mechanical arm is avoided.

Example III

Further, in another embodiment of the present application, the same or similar content as the first embodiment may be referred to the description above, and the description is omitted herein. On this basis, referring to fig. 3, before the step of inputting the state data into the target vibration detection model to obtain a vibration label, the method further includes:

step B10, acquiring second historical state data of the mechanical arm when abnormal vibration exists, wherein the second historical state data comprises a plurality of groups of state data corresponding to time points, each group of state data is provided with a corresponding vibration label, and the vibration label is determined according to the vibration type corresponding to each group of state data;

step B20, the proportion of the state data corresponding to various vibration tags in the second historical state data is adjusted according to the preset type proportion, and a second target state data set is obtained;

and step B30, training a preset vibration detection model according to the second target state data set to obtain a target vibration detection model.

In this embodiment of the present application, it should be noted that, the second historical state data is state data of each rotation axis of the mechanical arm, where the state data is collected during operation of the mechanical arm and is based on a time point as a group, and includes state data corresponding to various vibration labels, where the vibration labels are used to characterize vibration reasons and vibration attributes corresponding to the state data, the vibration reasons include loosening of a structure, circuit faults, improper control parameters, external interference, and the like, and the vibration attributes include spatial attributes and degree attributes, for example, the spatial attributes include transverse abnormal vibration, longitudinal abnormal vibration, arc abnormal vibration, irregular abnormal vibration, and the like, and the degree attributes include mild, severe, and the like. The preset proportion is used for adjusting the corresponding quantity value of each vibration label in the second historical state data to be in accordance with the proportion condition under the natural condition, for example, the probability of transverse abnormal vibration of a certain mechanical arm during working is 20%, the probability of longitudinal abnormal vibration is 30%, the probability of circular arc abnormal vibration is 10%, and the probability of irregular abnormal vibration is 40%, and the proportion corresponding to the transverse abnormal vibration, the longitudinal abnormal vibration, the circular arc abnormal vibration and the irregular abnormal vibration can be set to be 2:3:1:4. Moreover, the labels of each set of state data may include multiple dimensions, for example, the vibration labels of a certain set of state data are (external disturbance, transverse abnormal vibration, mild), so the total number of types of all vibration labels is the product of the label types of each dimension, namely, 4×4×3=48, and each dimension includes vibration cause, spatial attribute and degree attribute.

In addition, the preset vibration detection model is a neural network identification model, and the structure of the preset vibration detection model can comprise a full connection layer, a hidden layer, a full connection layer, an activation layer and the like; in addition, considering that the rotation shafts of the mechanical arm are connected through a mechanical structure, the rotation shafts can be mutually influenced, and abnormal vibration is possibly caused by the combined action of a plurality of rotation shafts, so that the state data of all the rotation shafts is taken as input; and adding a self-attention mechanism to enable the preset vibration detection model to learn the input features needing attention by self when detecting abnormal vibration of the mechanical arm.

As an example, steps B10 to B30 include: acquiring state data corresponding to the mechanical arm at different time points respectively and manually marking vibration labels on the state data to acquire second historical state data, wherein the vibration labels are used for representing the vibration types of the mechanical arm; dividing the proportion of the state data corresponding to each vibration tag in the second historical state data according to a preset type proportion to obtain a second target state data set, wherein the preset vibration proportion is determined according to the probability of abnormal vibration corresponding to various vibration tags of the mechanical arm under natural conditions; dividing the second target state data set into a training set and a verification set; constructing a preset vibration detection model based on a deep learning algorithm; training the preset vibration detection model according to the training set to optimize model parameters of the preset vibration detection model; inputting each group of state data in the verification set into the preset vibration detection model to obtain a predicted value corresponding to each state data; constructing a confusion matrix according to each predicted value and the corresponding vibration label; and evaluating whether the performance of the preset vibration detection model meets preset conditions or not according to the confusion matrix, and if so, setting the preset vibration detection model as a target vibration detection model.

Specifically, the step of training a preset vibration detection model according to the second target state data set to obtain a target vibration detection model includes:

step B31, dividing the second target state data set into a training set and a verification set;

step B32, inputting the training set into the preset vibration detection model, and iteratively optimizing model parameters of the preset vibration detection model;

step B33, inputting the verification set into an optimized preset vibration detection model to obtain a corresponding predicted value;

step B34, constructing an confusion matrix based on the predicted value and the vibration label corresponding to the predicted value;

step B35, calculating performance indexes of the preset vibration detection model based on the confusion matrix;

and step B36, setting the preset vibration detection model as a target vibration detection model if the performance index of the preset vibration detection model meets a preset index threshold.

In this embodiment of the present application, it should be noted that, in addition to this, the confusion matrix is a standard format for representing precision evaluation, and is represented by a matrix form of n rows and n columns, where each column of the confusion matrix represents a prediction class (i.e., a predicted value), and the total number of each column represents the number of data predicted as the class; each row represents the true home class of data (i.e., vibration tags), the total number of data for each row represents the number of instances of data for that class, and the values in each column represent the number of classes for which the true data is predicted. Through the confusion matrix, the detection of the preset vibration detection model on which vibration labels is accurate can be clearly analyzed, the detection errors of which vibration labels are larger, the performance index can be the prediction precision corresponding to each vibration label respectively, the preset index threshold can also be the prediction precision threshold corresponding to each vibration label according to specific conditions, for example, the vibration label with higher occurrence probability can be provided with a higher prediction precision threshold, for example, 99%, because the application is more, the detection result with higher precision can be stably output by the target vibration detection model, the vibration label with lower occurrence probability cannot be expected to have too high precision due to insufficient sample data for training, for example, 90% of the lower prediction precision threshold is set, and the model training time is avoided from being overlong, and the training efficiency is prevented from being influenced.

As an example, steps B31 to B36 include: dividing each group of state data in the second target state data set into a training set and a verification set, wherein the group number ratio of the group data of the state data in the training set to the group data of the state data in the verification set is 9:1; training the preset vibration detection model by using the state data in the training set as input and adopting a gradient descent method and iteratively optimizing model parameters of the preset vibration detection model until the preset vibration detection model converges; inputting each group of state data in the verification set into a converged preset vibration detection model to obtain a predicted value corresponding to each state data; constructing a confusion matrix according to each predicted value and vibration labels corresponding to each predicted value, wherein the confusion matrix is 48 rows and 48 columns, and 48 is the total number of types of the vibration labels; respectively calculating the prediction precision corresponding to the state data of each vibration tag according to the confusion matrix; comparing the prediction precision corresponding to each vibration tag with a preset prediction precision threshold corresponding to each vibration tag to judge that the performance index of the preset vibration detection model accords with a preset index threshold, wherein the preset index threshold comprises the prediction precision threshold corresponding to each vibration tag; if the prediction precision corresponding to each vibration label is higher than the preset prediction precision threshold corresponding to each vibration label, judging that the performance index of the preset vibration detection model accords with the preset index threshold, and setting the preset vibration detection model as a target vibration detection model; if there is a prediction accuracy not higher than each of the prediction accuracy thresholds, the process returns to step B32: and inputting the training set into the preset vibration detection model, and iteratively optimizing model parameters of the preset vibration detection model until the performance index of the preset vibration detection model accords with a preset index threshold.

The embodiment of the application provides a training method of a target vibration detection model before the step of inputting the state data into the target vibration detection model to obtain the vibration label, wherein the second historical state data of the mechanical arm when abnormal vibration exists is firstly obtained, the second historical state data comprises a plurality of groups of state data corresponding to time points, each group of state data is provided with a corresponding vibration label, the vibration label is determined according to the vibration type corresponding to each group of state data, then the proportion of the state data corresponding to various vibration labels in the second historical state data is adjusted according to the preset type proportion to obtain a second target state data set, and further the preset vibration detection model is trained according to the second target state data set to obtain the target vibration detection model. According to the technical scheme, the state data of each rotating shaft of the mechanical arm and the manually marked vibration label are adopted to train the supervised neural network identification model, so that the target vibration detection model, with the prediction accuracy meeting the requirement, of the vibration label of the mechanical arm can be detected according to the current state data, so that the corresponding vibration compensation signal is inquired according to the vibration label, the detection automation of the abnormal type of the mechanical arm is realized, the real-time requirement can be met, compared with a method for manually analyzing the vibration type of the mechanical arm and debugging the compensation signal, the method is more stable and higher in real-time, the detection result of the vibration label can be timely obtained after abnormal vibration occurs, and the damage to the mechanical arm caused by long-term abnormal vibration of the mechanical arm and the influence on the production quality are avoided.

Example IV

The embodiment of the application also provides a mechanical arm vibration suppression device based on deep learning, the mechanical arm vibration suppression device based on deep learning is applied to mechanical arm vibration suppression equipment based on deep learning, referring to fig. 4, the mechanical arm vibration suppression device based on deep learning includes:

the data acquisition module 101 is used for acquiring state data of each rotating shaft of the mechanical arm;

the anomaly detection module 102 is configured to input the state data into a target state detection model to determine whether the mechanical arm has abnormal vibration, where the target state detection model is obtained by training multiple sets of state data with state labels;

the vibration analysis module 103 is configured to input the state data into a target vibration detection model to obtain a vibration tag if abnormal vibration exists in the mechanical arm, where the target vibration detection model is obtained by training state data corresponding to multiple groups of mechanical arms in abnormal vibration states, and each group of state data has a corresponding vibration tag;

and the vibration suppression module 104 is used for inquiring the corresponding vibration compensation signal according to the vibration label and suppressing abnormal vibration of the mechanical arm according to the vibration compensation signal.

Optionally, the data acquisition module is further configured to:

collecting amplitude data of each rotating shaft through a vibration sensor;

Optionally, the anomaly detection module is further configured to:

Optionally, the vibration analysis module is further configured to:

Optionally, the vibration suppression module is further configured to:

According to the mechanical arm vibration suppression device based on the deep learning, the mechanical arm vibration suppression method based on the deep learning in the embodiment is adopted, and the technical problem of low mechanical arm vibration suppression efficiency is solved. Compared with the prior art, the beneficial effects of the mechanical arm vibration suppression device based on deep learning provided by the embodiment of the application are the same as those of the mechanical arm vibration suppression method based on deep learning provided by the embodiment, and other technical features in the mechanical arm vibration suppression device based on deep learning are the same as those disclosed in the method of the previous embodiment, so that the description is omitted.

Example five

The embodiment of the application provides electronic equipment, the electronic equipment includes: at least one processor; and a memory communicatively linked to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the deep learning-based mechanical arm vibration suppression method in the first embodiment.

Referring now to fig. 5, a schematic diagram of an electronic device suitable for use in implementing embodiments of the present disclosure is shown. The electronic devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistant, personal digital assistants), PADs (tablet computers), PMPs (Portable Media Player, portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and the like, and stationary terminals such as digital TVs, desktop computers, and the like. The electronic device shown in fig. 5 is merely an example and should not be construed to limit the functionality and scope of use of the disclosed embodiments.

As shown in fig. 5, the electronic device may include a processing means (e.g., a central processing unit, a graphic processor, etc.) that may perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) or a program loaded from a storage means into a random access memory (RAM, random access memory). In the RAM, various programs and data required for the operation of the electronic device are also stored. The processing device, ROM and RAM are connected to each other via a bus. Input/output (I/O) interfaces are also linked to the bus.

In general, the following systems may be linked to I/O interfaces: input devices including, for example, touch screens, touch pads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices including, for example, liquid crystal displays (LCDs, liquid crystal display), speakers, vibrators, etc.; storage devices including, for example, magnetic tape, hard disk, etc.; a communication device. The communication means may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. While electronic devices having various systems are shown in the figures, it should be understood that not all of the illustrated systems are required to be implemented or provided. More or fewer systems may alternatively be implemented or provided.

In particular, according to embodiments of the present disclosure, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from ROM. The above-described functions defined in the methods of the embodiments of the present disclosure are performed when the computer program is executed by a processing device.

According to the electronic equipment, the technical problem of low suppression efficiency of mechanical arm vibration is solved by adopting the mechanical arm vibration suppression method based on deep learning in the embodiment. Compared with the prior art, the beneficial effects of the electronic device provided by the embodiment of the present application are the same as those of the mechanical arm vibration suppression method based on deep learning provided by the first embodiment, and other technical features of the electronic device are the same as those disclosed by the method of the previous embodiment, which are not repeated herein.

It should be understood that portions of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof. In the description of the above embodiments, particular features, structures, materials, or characteristics may be combined in any suitable manner in any one or more embodiments or examples.

The foregoing is merely specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily think about changes or substitutions within the technical scope of the present application, and the changes and substitutions are intended to be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Example six

The present embodiment provides a computer-readable storage medium having computer-readable program instructions stored thereon for performing the method for deep learning-based arm vibration suppression in the above embodiment one.

The computer readable storage medium provided by the embodiments of the present application may be, for example, a usb disk, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, or device, or a combination of any of the foregoing. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical link having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-Only Memory (ROM), an erasable programmable read-Only Memory (EPROM, erasable Programmable Read-Only Memory, or flash Memory), an optical fiber, a portable compact disc read-Only Memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this embodiment, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, or device. Program code embodied on a computer readable storage medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, fiber optic cables, RF (radio frequency), and the like, or any suitable combination of the foregoing.

The above-described computer-readable storage medium may be contained in an electronic device; or may exist alone without being assembled into an electronic device.

The computer-readable storage medium carries one or more programs that, when executed by the electronic device, cause the electronic device to collect status data at each rotational axis of the robotic arm; inputting the state data into a target state detection model to judge whether the mechanical arm has abnormal vibration, wherein the target state detection model is obtained by training a plurality of groups of state data with state labels; if abnormal vibration exists in the mechanical arm, inputting the state data into a target vibration detection model to obtain a vibration label, wherein the target vibration detection model is obtained by training according to state data corresponding to multiple groups of mechanical arms in abnormal vibration states, and each group of state data is provided with a corresponding vibration label; and inquiring a corresponding vibration compensation signal according to the vibration tag, and inhibiting abnormal vibration of the mechanical arm according to the vibration compensation signal.

Computer program code for carrying out operations of the present disclosure may be written in one or more programming languages, including an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be linked to the user's computer through any kind of network, including a local area network (LAN, local area network) or a wide area network (WAN, wide Area Network), or it may be linked to an external computer (e.g., through the internet using an internet service provider).

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The modules described in the embodiments of the present disclosure may be implemented in software or hardware. Wherein the name of the module does not constitute a limitation of the unit itself in some cases.

The computer readable storage medium stores computer readable program instructions for executing the mechanical arm vibration suppression method based on deep learning, and solves the technical problem of low mechanical arm vibration suppression efficiency. Compared with the prior art, the beneficial effects of the computer readable storage medium provided by the embodiment of the present application are the same as the beneficial effects of the mechanical arm vibration suppression method based on deep learning provided by the above embodiment, and are not described herein.

Example seven

The application provides a computer program product which solves the technical problem of low damping efficiency of mechanical arm vibration. Compared with the prior art, the beneficial effects of the computer program product provided by the embodiment of the present application are the same as the beneficial effects of the mechanical arm vibration suppression method based on deep learning provided by the above embodiment, and are not described herein.

The foregoing description is only of the preferred embodiments of the present application and is not intended to limit the scope of the claims, and all equivalent structures or equivalent processes using the descriptions and drawings of the present application, or direct or indirect application in other related technical fields are included in the scope of the claims.

Claims

1. The mechanical arm vibration suppression method based on the deep learning is characterized by comprising the following steps of:

collecting state data of each rotating shaft of the mechanical arm;

2. The deep learning-based mechanical arm vibration suppression method according to claim 1, wherein the state data includes motion state data, amplitude data, and temperature data, and the step of collecting the state data at each rotation axis of the mechanical arm includes:

collecting amplitude data of each rotating shaft through a vibration sensor;

3. The deep learning based robot vibration suppression method according to claim 2, wherein before the step of inputting the state data into a target state detection model to determine whether there is abnormal vibration of the robot, the method further comprises:

4. The method for suppressing vibration of a robot arm based on deep learning as claimed in claim 3, wherein the step of training a preset state detection model according to the first target state data set to obtain a target state detection model comprises:

5. The deep learning-based mechanical arm vibration suppression method according to claim 2, further comprising, before the step of inputting the state data into a target vibration detection model to obtain a vibration label:

6. The method for suppressing vibration of a robot arm based on deep learning of claim 5, wherein the step of training a preset vibration detection model based on the second target state data set to obtain a target vibration detection model includes:

7. The deep learning-based mechanical arm vibration suppression method according to claim 1, wherein the step of inquiring a corresponding vibration compensation signal according to the vibration tag and suppressing abnormal vibration of the mechanical arm according to the vibration compensation signal comprises:

8. The utility model provides a robotic arm vibration suppression device based on degree of depth study which characterized in that, robotic arm vibration suppression device based on degree of depth study includes:

9. An electronic device, the electronic device comprising:

at least one processor; the method comprises the steps of,

a memory communicatively linked to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the steps of the deep learning based robotic arm vibration suppression method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a program for realizing the deep learning-based robot vibration suppression method, the program for realizing the deep learning-based robot vibration suppression method being executed by a processor to realize the steps of the deep learning-based robot vibration suppression method according to any one of claims 1 to 7.