CN115062672B - Method and system for predicting life cycle of SCARA robot - Google Patents
Method and system for predicting life cycle of SCARA robot Download PDFInfo
- Publication number
- CN115062672B CN115062672B CN202210879935.6A CN202210879935A CN115062672B CN 115062672 B CN115062672 B CN 115062672B CN 202210879935 A CN202210879935 A CN 202210879935A CN 115062672 B CN115062672 B CN 115062672B
- Authority
- CN
- China
- Prior art keywords
- life cycle
- wavelet packet
- joint
- acceleration
- scara robot
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J19/00—Accessories fitted to manipulators, e.g. for monitoring, for viewing; Safety devices combined with or specially adapted for use in connection with manipulators
- B25J19/0066—Means or methods for maintaining or repairing manipulators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Mechanical Engineering (AREA)
- Robotics (AREA)
- Manipulator (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a system for predicting a life cycle of a SCARA robot, which relate to the technical field of robots, and the method comprises the following steps: acquiring acceleration vibration signals in the running process through four acceleration sensors arranged on a SCARA robot arm; preprocessing acceleration vibration signals acquired by the four acceleration sensors to extract vibration signal feature vectors; taking each path of wavelet packet energy spectrum entropy vector as an input vector of a corresponding neural network, and acquiring a life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector; and processing the life cycle prediction result under the energy spectrum entropy vector of the four-way wavelet packet to obtain the life cycle of the SCARA robot. The life cycle of the SCARA robot is predicted by collecting acceleration vibration signals of four joints, so that the reliability and the precision of the life cycle of the robot can be effectively evaluated.
Description
Technical Field
The invention relates to the technical field of robots, in particular to a method and a system for predicting a life cycle of a SCARA robot.
Background
SCARA (Selective Compliance Assembly Robot Arm) is a selective compliance assembly robot arm, which is a cylindrical coordinate type industrial robot with four degrees of freedom and with a built-in four-axis servo motor, a harmonic reducer and a brake system. The SCARA robot has four degrees of freedom, including two horizontal joints and a link that can move vertically as well as rotate. The first degree of freedom and the second degree of freedom are rotary joints formed by a large mechanical arm and a small mechanical arm, so that the horizontal connecting rod can perform rotary motion in a horizontal plane to complete quick and accurate positioning in the plane; the third degree of freedom is a movable joint with a vertical lifting function, and can complete the movement vertical to the plane; the fourth degree of freedom is a rotational joint, which may allow the end effector to conveniently grasp a target. The SCARA robot can ensure stronger rigidity and higher precision in the vertical direction, and can freely and rapidly rotate in the horizontal plane, so that the SCARA robot is very suitable for plane positioning and completing the task of sorting and grabbing workpieces. The actuator parts of the four joints of the SCARA robot are all AC servo motors, so that larger torque can be provided, and the robot has good executing effect on the movement speed and precision of the mechanical arm. The first joint and the second joint of the SCARA robot are decelerated based on harmonics, while the third joint and the fourth joint are decelerated based on synchronous belts.
In the operation process of the SCARA robot, the corresponding operation process is finished by means of four joints, the four joints are worn or failed, and equipment operation data under normal operation, wear and failure behaviors can fully train a machine learning model, so that the trained model can predict the life cycle of the SCARA robot according to the input equipment operation data, but how to realize corresponding data acquisition for the four joints is particularly important to carry out life cycle expectation of the SCARA robot.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, and provides a method and a system for predicting the life cycle of a SCARA robot, which are used for predicting the life cycle of the SCARA robot by collecting acceleration vibration signals of four joints, so that the reliability and the accuracy of the life cycle of the robot can be effectively evaluated.
In order to solve the above problems, the present invention proposes a method for predicting life cycle of a SCARA robot, the method comprising the steps of:
acquiring acceleration vibration signals in the running process through four acceleration sensors arranged on a robot arm of the SCARA robot, wherein each of four joints of the SCARA robot is provided with an acceleration sensor;
preprocessing acceleration vibration signals acquired by four acceleration sensors to extract vibration signal feature vectors, wherein the vibration signal feature vectors comprise four-path wavelet packet energy spectrum entropy vectors;
taking each path of wavelet packet energy spectrum entropy vector as an input vector of a corresponding neural network, and acquiring a life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector;
and processing the life cycle prediction result under the energy spectrum entropy vector of the four-way wavelet packet to obtain the life cycle of the SCARA robot.
The acquisition of acceleration vibration signals in the running process through four acceleration sensors arranged on the SCARA robot arm comprises the following steps:
acquiring a simulated acceleration vibration signal on the joint based on an acceleration sensor on each joint;
filtering the analog acceleration vibration signal to filter a high-frequency vibration signal;
analog-to-digital conversion processing is carried out on the analog acceleration vibration signals with the high-frequency vibration signals filtered, and digital acceleration vibration signals are generated.
The preprocessing the acceleration vibration signals acquired by the four acceleration sensors to extract the vibration signal feature vectors comprises the following steps:
the wavelet packet decomposition principle is adopted to decompose the wavelet packet of the acceleration vibration signal;
reconstructing wavelet packet coefficients, and solving the proportion of energy contained in each frequency band component to the total energy of each acceleration signal;
and obtaining the energy distribution condition of each wavelet packet coefficient of the acceleration vibration signal by calculating the wavelet packet energy spectral entropy of each frequency band, and taking the wavelet packet energy spectral entropy as the vibration signal characteristic vector of the acceleration vibration signal.
The reconstructing the wavelet packet coefficients and calculating the specific gravity of the energy contained in each frequency band component to the total energy of the signal comprises the following steps:
different frequency band components of the signals are obtained through wavelet packet decomposition, characteristic information of each frequency band is obtained according to the distribution of the frequency bands of the acceleration vibration signals, and different acceleration vibration signals are analyzed through wavelet packet energy spectrum.
The step of taking each path of wavelet packet energy spectrum entropy vector as the input vector of the corresponding neural network and obtaining the life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector comprises the following steps:
acquiring sample entropy data on joints corresponding to energy spectrum entropy vectors of each path of wavelet packet used for model training templates;
performing an order analysis on the sample entropy data to extract each joint feature;
inputting a part of sample entropy data in each joint characteristic into a constructed neural network for predicting the life cycle of each joint, and executing verification processing through the other part of sample entropy data so as to train a life cycle prediction model of each joint;
and inputting the energy spectrum entropy vector of each path of wavelet packet to the corresponding life cycle prediction module to obtain a life cycle prediction result of each joint.
Processing the life cycle prediction result under the energy spectrum entropy vector of the four-way wavelet packet to obtain the life cycle of the SCARA robot comprises the following steps:
forming a current life cycle state curve graph of each joint from the life cycle prediction result of each joint;
and comparing and fusing the life cycle state graphs of the four joints to obtain a life cycle state result graph of the SCARA robot.
Correspondingly, the invention also provides a system for predicting the life cycle of the SCARA robot, which comprises:
the data acquisition module is used for acquiring acceleration vibration signals in the running process through four acceleration sensors arranged on the robot arm of the SCARA robot, and each of four joints of the SCARA robot is provided with one acceleration sensor;
the preprocessing module is used for preprocessing acceleration vibration signals acquired by the four acceleration sensors to extract vibration signal feature vectors, wherein the vibration signal feature vectors comprise four paths of wavelet packet energy spectrum entropy vectors;
the data processing module is used for taking each path of wavelet packet energy spectrum entropy vector as an input vector of the corresponding neural network and acquiring a life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector;
and the prediction processing module is used for processing the life cycle prediction result under the energy spectrum entropy vector of the four-way wavelet packet to obtain the life cycle of the SCARA robot.
The preprocessing module comprises:
the decomposition unit is used for decomposing the wavelet packet of the acceleration vibration signal by adopting the wavelet packet decomposition principle;
the reconstruction unit is used for reconstructing the wavelet packet coefficient and solving the proportion of the energy contained in each frequency band component to the total energy of each acceleration vibration signal;
the characteristic unit is used for obtaining the energy distribution condition of each wavelet packet coefficient of the acceleration vibration signal by calculating the wavelet packet energy spectrum entropy of each frequency band, and taking the wavelet packet energy spectrum entropy as the vibration signal characteristic vector of the acceleration vibration signal.
The data processing module comprises:
the acquisition unit is used for acquiring sample entropy data on the joints corresponding to the energy spectrum entropy vectors of each path of wavelet packet used for the model training template;
an order analysis unit for performing an order analysis on the sample entropy data to extract each joint feature;
the construction unit is used for inputting part of sample entropy data in the characteristics of each joint into the constructed neural network for predicting the life cycle of each joint, and executing verification processing through the other part of sample entropy data so as to train a life cycle prediction model of each joint;
the prediction unit is used for inputting the energy spectrum entropy vector of each path of wavelet packet to the corresponding life cycle prediction module to obtain a life cycle prediction result of each joint.
The prediction processing module comprises:
the generating unit is used for forming a life cycle state curve chart of each joint according to the life cycle prediction result of each joint;
and the synthesis unit is used for comparing and fusing the life cycle state graphs of the four joints to obtain a life cycle state result graph of the SCARA robot.
The method and the system realize life cycle prediction of the SCARA robot by extracting the characteristic quantity on each joint and combining the characteristic quantity with the neural network, have high efficiency in the whole realization process, and can intuitively embody the running health state of the SCARA robot through a life cycle model. The characteristic information of the acceleration vibration signals on each joint can be directly expressed by adopting a mode of combining wavelet packet decomposition and energy spectrum entropy, and extraction of the characteristic information is facilitated, so that learning and training of the acceleration vibration signals on each joint can be realized by combining a neural network, and further prediction of life cycle of the SCARA robot is realized. The method has higher accuracy in identifying the acceleration vibration signals, and can realize classification prediction of the acceleration vibration signals through the neural network model, so that the accuracy of the life cycle prediction of the SCARA robot is improved, and the reliability and the accuracy of the life cycle of the robot can be effectively evaluated.
Drawings
FIG. 1 is a schematic diagram of a system for predicting life cycle of a SCARA robot in accordance with an embodiment of the present invention;
FIG. 2 is a flow chart of a method of predicting a life cycle of a SCARA robot in an embodiment of the present invention;
FIG. 3 is a flow chart of a method for an acceleration sensor to collect acceleration vibration signals during operation in an embodiment of the present invention;
FIG. 4 is a flow chart of a method for extracting feature vectors of vibration signals by preprocessing in an embodiment of the invention;
FIG. 5 is a flowchart of a method for obtaining a life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector in an embodiment of the present invention;
fig. 6 is a graph showing the vibration signal of the joint and the degree of degradation of the joint in the embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Specifically, fig. 1 shows a schematic diagram of a system for predicting life cycle of a SCARA robot in an embodiment of the present invention, where the system includes:
the data acquisition module is used for acquiring acceleration vibration signals in the running process through four acceleration sensors arranged on the robot arm of the SCARA robot, and each of four joints of the SCARA robot is provided with one acceleration sensor;
the preprocessing module is used for preprocessing acceleration vibration signals acquired by the four acceleration sensors to extract vibration signal feature vectors, wherein the vibration signal feature vectors comprise four paths of wavelet packet energy spectrum entropy vectors;
the data processing module is used for taking each path of wavelet packet energy spectrum entropy vector as an input vector of the corresponding neural network and acquiring a life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector;
and the prediction processing module is used for processing the life cycle prediction result under the energy spectrum entropy vector of the four-way wavelet packet to obtain the life cycle of the SCARA robot.
The preprocessing module comprises:
the decomposition unit is used for decomposing the wavelet packet of the acceleration vibration signal by adopting the wavelet packet decomposition principle;
the reconstruction unit is used for reconstructing the wavelet packet coefficient and solving the proportion of the energy contained in each frequency band component to the total energy of each acceleration vibration signal;
the characteristic unit is used for obtaining the energy distribution condition of each wavelet packet coefficient of the acceleration vibration signal by calculating the wavelet packet energy spectrum entropy of each frequency band, and taking the wavelet packet energy spectrum entropy as the vibration signal characteristic vector of the acceleration vibration signal.
The data processing module comprises:
the acquisition unit is used for acquiring sample entropy data on the joints corresponding to the energy spectrum entropy vectors of each path of wavelet packet used for the model training template;
an order analysis unit for performing an order analysis on the sample entropy data to extract each joint feature;
the construction unit is used for inputting part of sample entropy data in the characteristics of each joint into the constructed neural network for predicting the life cycle of each joint, and executing verification processing through the other part of sample entropy data so as to train a life cycle prediction model of each joint;
the prediction unit is used for inputting the energy spectrum entropy vector of each path of wavelet packet to the corresponding life cycle prediction module to obtain a life cycle prediction result of each joint.
The prediction processing module comprises:
the generating unit is used for forming a life cycle state curve chart of each joint according to the life cycle prediction result of each joint;
and the synthesis unit is used for comparing and fusing the life cycle state graphs of the four joints to obtain a life cycle state result graph of the SCARA robot.
The system in the embodiment of the invention realizes the life cycle prediction of the SCARA robot by extracting the characteristic quantity on each joint and combining with the neural network, has high efficiency in the whole realization process, and can intuitively embody the running health state of the SCARA robot through a life cycle model. The characteristic information of the acceleration vibration signals on each joint can be directly expressed by adopting a mode of combining wavelet packet decomposition and energy spectrum entropy, and extraction of the characteristic information is facilitated, so that learning and training of the acceleration vibration signals on each joint can be realized by combining a neural network, and further prediction of life cycle of the SCARA robot is realized. The method has higher accuracy in identifying the acceleration vibration signals, and can realize classification prediction of the acceleration vibration signals through the neural network model, so that the accuracy of the life cycle prediction of the SCARA robot is improved, and the reliability and the accuracy of the life cycle of the robot can be effectively evaluated.
The embodiment of the invention relates to a method for predicting the life cycle of a SCARA robot, which comprises the following steps: acquiring acceleration vibration signals in the running process through four acceleration sensors arranged on a robot arm of the SCARA robot, wherein each of four joints of the SCARA robot is provided with an acceleration sensor; preprocessing acceleration vibration signals acquired by four acceleration sensors to extract vibration signal feature vectors, wherein the vibration signal feature vectors comprise four-path wavelet packet energy spectrum entropy vectors; taking each path of wavelet packet energy spectrum entropy vector as an input vector of a corresponding neural network, and acquiring a life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector; and processing the life cycle prediction result under the energy spectrum entropy vector of the four-way wavelet packet to obtain the life cycle of the SCARA robot.
Specifically, fig. 2 shows a flowchart of a method for predicting a life cycle of a SCARA robot in an embodiment of the present invention, which specifically includes the following steps:
s201, acquiring acceleration vibration signals in the running process through four acceleration sensors arranged on a SCARA robot arm;
specifically, four acceleration sensors are arranged on the robot arm of the SCARA robot, namely, one acceleration sensor is arranged on each of four joints of the SCARA robot, and the acceleration sensor on each joint can collect acceleration vibration signals of the position where the SCARA robot is located in the running process of the SCARA robot, so that the final life cycle prediction process is completed by means of the acceleration vibration signals.
Fig. 3 shows a flowchart of a method for acquiring an acceleration vibration signal during operation by an acceleration sensor in an embodiment of the present invention, including:
s301, acquiring a simulated acceleration vibration signal on a joint based on an acceleration sensor on each joint;
the MEMS acceleration sensor can be selected to collect signals, and has the advantages of small volume, low price, strong output signal, simple post-circuit and the like. The vibration amplitude of each joint in the SCARA robot is different, the vibration frequency of equipment is generally 1-10 kHz, and the acceleration sensor has a larger frequency bandwidth and can meet the requirement of signal acquisition on different joints.
S302, filtering the analog acceleration vibration signal to filter a high-frequency vibration signal;
in the implementation process, the anti-aliasing filter circuit is used for filtering, and can effectively filter out high-frequency vibration signals in the analog acceleration vibration signals.
S303, performing analog-to-digital conversion processing on the analog acceleration vibration signal with the high-frequency vibration signal filtered, and generating a digital acceleration vibration signal.
The analog-to-digital converter ADC is used for converting the data into digital quantity, the sampling frequency of the ADC is set to be more than 5KHz, and the sampled data is output to a corresponding processor through a serial port according to a set format to carry out the extraction process of the acceleration vibration signal.
S202, preprocessing acceleration vibration signals acquired by four acceleration sensors to extract vibration signal feature vectors;
the acceleration vibration signals acquired by the four acceleration sensors are preprocessed to extract vibration signal characteristic vectors, wherein the vibration signal characteristic vectors comprise four-path wavelet packet energy spectrum entropy vectors.
The acceleration vibration signal is a representation of the uncertainty degree in the state of the signal or the system, the randomness of the vibration energy under different joints can be described by converting the acceleration vibration signal into wavelet packet energy spectrum entropy, when the running state of each joint is changed, the wavelet packet energy spectrum entropy is changed along with the complexity of the running state, and therefore the characteristic parameters of the acceleration vibration signal can be represented by the wavelet packet energy spectrum entropy.
Specifically, fig. 4 shows a flowchart of a method for extracting feature vectors of vibration signals by preprocessing in an embodiment of the present invention, including the following steps:
s401, carrying out wavelet packet decomposition on the acceleration vibration signal by adopting a wavelet packet decomposition principle;
the acceleration vibration signals are subjected to wavelet packet decomposition through a wavelet packet decomposition principle, the wavelet packet decomposition is commonly used in judging and identifying fault vibration under a mechanical principle, the collected acceleration vibration signals form a random variable, a plurality of probability distribution values exist in the random variable, an integral characteristic entropy value exists in each random variable, the larger the probability distribution value of the random variable is, the larger the integral characteristic entropy value corresponds to the random variable is, and the smaller the probability distribution value of the random variable is, the smaller the integral characteristic entropy value corresponds to the random variable is.
For the random time sequence under the random variable, the energy of each frequency band under the random time sequence has a magnitude relation, and the energy spectrum entropy under the random time sequence reflects the energy distribution condition of the time sequence signal in the frequency domain.
The wavelet packet decomposition can decompose the decomposition space into the sum of different sub-decomposition spaces according to different scales, and if higher resolution is needed, the decomposition can be continued for each sub-decomposition space, and the multi-resolution sub-space and each sub-decomposition space are unified through a new space.
S402, reconstructing wavelet packet coefficients, and solving the proportion of energy contained in each frequency band component to the total energy of each acceleration vibration signal;
different frequency band components of the signal are obtained through wavelet packet decomposition, characteristic information of each frequency band is obtained according to the distribution of the frequency bands of the acceleration vibration signal, and the wavelet packet energy spectrum is utilized to analyze different acceleration vibration signals. The wavelet packet transformation has better processing capacity for non-stationary signals, and the analysis of the acceleration vibration signals can obtain better effects. Different frequency band components of the signal can be obtained through wavelet packet decomposition, the information distribution of the different frequency band components is different, and the characteristic information of each frequency band can be obtained according to the distribution of the acceleration vibration signal.
The energy of each frequency band component of the wavelet packet of each acceleration vibration signal can be calculated by the Pasteur theorem, the wavelet packet coefficient is utilized to analyze the energy of different frequency bands of the acceleration vibration signals, when the acceleration vibration signals are different, the energy corresponding to the wavelet packet coefficient is also different, and the wavelet packet energy spectrum can be utilized to analyze the different acceleration vibration signals.
Here, the energy of each band component of the wavelet packet of the acceleration vibration signal f (x) can be calculated by the pasteval theorem, and the expression thereof is as follows:
as can be seen from the formula (1), the wavelet packet coefficients can be used for analyzing the energy of different frequency bands of the acceleration vibration signals, when the acceleration vibration signals are different, the energy magnitudes corresponding to the wavelet packet coefficients are different, so that the wavelet packet energy spectrum can be used for analyzing the different acceleration vibration signals, and W ij The energy of the jth frequency band component under the ith acceleration vibration signal is represented, i represents the value of the acceleration vibration signal, and j represents the value of the frequency band component in the acceleration vibration signal.
Wherein the energy of each wavelet packet coefficient is:
the total energy of the acceleration vibration signal f (x) is as follows:
let the energy ratio of wavelet packet coefficient of each frequency band of signal f (x) be p j The expression is as follows:
s403, energy distribution conditions of all wavelet packet coefficients of the acceleration vibration signal are obtained through calculating the wavelet packet energy spectrum entropy of each frequency band, and the wavelet packet energy spectrum entropy is used as a vibration signal characteristic vector of the acceleration vibration signal.
The energy spectrum entropy is a quantitative description of the complexity of the energy distribution of the acceleration vibration signal in the frequency domain, is a reaction to the information quantity of the acceleration vibration signal in the whole frequency, and is characterized by the information entropy in all frequencies, and the detailed components of the frequency spectrum are not considered. The acceleration vibration signal is processed by wavelet packet transformation, so that the acceleration vibration signal can have the characteristic of broadband response to the non-stationary signal, has higher frequency resolution at a low frequency and higher time resolution at a high frequency, and is suitable for analyzing the non-stationary signal.
The wavelet packet energy spectrum entropy is taken as the vibration signal characteristic vector of the acceleration vibration signal, and the wavelet packet energy spectrum entropy H can be obtained through calculation in S402 w The expression is:
here H w Reflecting the energy distribution of each wavelet packet coefficient of the original acceleration vibration signal, when H w The larger the acceleration vibration signal is, the more complex the acceleration vibration signal is, and the poorer the stability is; when H is w The smaller the time, the smoother the trend of the acceleration vibration signal is indicated, and the lower the complexity is.
S203, taking each path of wavelet packet energy spectrum entropy vector as an input vector of a corresponding neural network, and acquiring a life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector;
because each joint has different energy spectrums, the energy spectrums need to be analyzed and compared with the corresponding energy spectrums to obtain the life cycle prediction result corresponding to each joint, namely, the life cycle prediction result corresponding to each wavelet packet energy spectrum entropy vector is obtained by taking each wavelet packet energy spectrum entropy vector as the input vector of the corresponding neural network, and the life cycle prediction result corresponding to each wavelet packet energy spectrum entropy vector is obtained, fig. 5 is a flow chart of a method for obtaining the life cycle prediction result corresponding to each wavelet packet energy spectrum entropy vector in the embodiment of the invention, and the method comprises the following steps:
s501, acquiring sample entropy data on joints corresponding to energy spectrum entropy vectors of each path of wavelet packet used for model training templates;
the acceleration vibration data generated in the historical life cycle of the replacement of each joint in different SCARA robots can be used as training samples, and the wavelet packet energy spectrum entropy vector of the acceleration vibration data generated in the historical life cycle is the sample entropy data on the corresponding joint.
The acceleration vibration data generated by the joints which do not reach the replacement standard are taken as prediction data, namely the acceleration vibration data generated by each joint in the running process can be collected and converted into corresponding prediction data according to the embodiment of the invention, and the prediction data are input into a neural network model through wavelet packet energy spectrum entropy vectors.
S502, performing order analysis on the sample entropy data to extract each joint feature;
in the specific implementation process, extracting sample entropy data of each joint under the historical life cycle, expanding to frequency doubling and frequency tripling to serve as characteristic points of preliminary acquisition, sequencing the characteristic points of the preliminary acquisition, taking the maximum value of positive and negative preset sample entropy before and after the characteristic points as the value of a final characteristic point, wherein the frequency corresponding to the value of the final characteristic point is the characteristic point to be extracted and serves as the characteristic of a first batch of samples; selecting sample entropy data of a certain proportion of a machine for analysis, selecting characteristic points with sample entropy values showing extreme values, testing and comparing the sample entropy data of the residual proportion, retaining the common obvious characteristic points in all samples, collecting the obvious characteristic points of other SCARA robots, removing different memories of the characteristic points obtained by different SCARA robots, and analyzing and combining the common characteristic points to be used as the characteristics of a second batch of samples; comparing the first batch of sample features with the second batch of sample features, reserving frequency multiplication features obtained by order analysis and other common feature points, and merging feature points with close frequency to obtain life cycle features of joints of the SCARA robot needed to be used for creating the model.
S503, training a life cycle prediction model of each joint;
here, a part of the sample entropy data in the characteristics of each joint is input into the constructed neural network for predicting the life cycle of each joint, and verification processing is performed through another part of the sample entropy data so as to train a life cycle prediction model of each joint.
In the implementation process, a deep neural network for predicting the life cycle of each joint is constructed, and the number of corresponding input layer neuron nodes is selected according to the number of the acquired characteristic columns; randomly extracting a certain preset percentage of sample entropy data from all the sample entropy data to be used as a training sample set for training the deep neural network, and taking the rest sample entropy data as a verification set; and iteratively training the deep neural network to obtain a life cycle prediction model of each joint after training.
S504, inputting the energy spectrum entropy vector of each path of wavelet packet to the corresponding life cycle prediction module to obtain a life cycle prediction result of each joint.
All characteristic information can be used as input of a data driving model through the neural network, and operation data can be analyzed more comprehensively from high latitude, so that life cycle prediction accuracy is improved, a prediction process of the SCARA robot is more in line with an actual service scene, and prediction stability and applicability are improved. The wavelet packet energy spectrum entropy vector extracted from the acceleration vibration signal is used as an input characteristic column of the neural network by means of the calculation advantage of the neural network algorithm, the traditional mechanism model and personal experience can be replaced by the life cycle prediction module, complex formula calculation and logic among various characteristic quantities are avoided, the life cycle of each joint in the SCARA robot in the current state is directly predicted, the recognition efficiency is improved, the complexity required by the model is reduced, the external interference is eliminated, and the human factor is reduced.
S204, processing a life cycle prediction result under the energy spectrum entropy vector of the four-way wavelet packet to obtain the life cycle of the SCARA robot.
In the implementation process, the life cycle prediction result of each joint is formed into a current life cycle state curve chart of each joint; and comparing and fusing the life cycle state graphs of the four joints to obtain a life cycle state result graph of the SCARA robot.
In the implementation process, four joint model initial life cycle diagrams can be arranged in the life cycle state model diagram of the SCARA robot, the life cycle result of each joint is identified, then the life cycle result data of each joint is extracted, then data replacement is carried out on the initial life cycle diagram corresponding to each joint, and a life cycle state result diagram under the SCARA robot is formed on the four joint model initial life cycle diagrams.
Fig. 6 is a schematic diagram showing a joint vibration signal and a joint degradation degree curve in an embodiment of the present invention, a portion a in fig. 6 is a joint life cycle curve in an embodiment of the present invention, and a portion b in fig. 6 is a joint acceleration vibration signal curve in an embodiment of the present invention. The life cycle of the joint at 1 and 2 is obviously shown in fig. 6, so that the life cycle state of the joint can be roughly divided into three stages as shown in fig. 6, namely a stationary stage, a degradation stage and a rapid degradation stage, and the above can reflect the schematic diagram of the whole life cycle of the joint, wherein mm in vibration amplitude is the distance of the vibration amplitude, namely millimeter.
The life cycle state curve graph of the joint to be monitored from the beginning to the current moment is obtained through the acceleration vibration signals on each joint, the life cycle state curve graph of the joint at the current moment is compared with the life cycle curve graph of the joint corresponding to each joint, the life cycle state of the joint to be monitored at the monitoring moment can be obtained, if the joint is in a stable stage and a degradation stage, the joint does not need to be maintained, if the joint is in the rapid degradation stage, the corresponding joint is about to be damaged, a new joint or a SCARA robot needs to be replaced, and maintenance personnel can conveniently monitor and manage continuously.
In the implementation process, the current acceleration vibration signal of each joint can be combined with the corresponding joint life cycle graph of each joint to identify the stage of each acceleration vibration signal in the joint life cycle graph, and then the life cycle state of each joint, namely the life cycle state of the first joint, the life cycle state of the second joint, the life cycle state of the third joint and the life cycle state of the fourth joint, is generated by combining the operation time of the SCARA robot, wherein the life cycle state of each joint comprises: the life cycle state of each joint can be reflected by data superposition in a life cycle state model diagram of the SCARA robot, so that a life cycle state result diagram of the SCARA robot is formed, the life cycle state result diagram can reflect the life cycle of the SCARA robot, and then early warning reminding is carried out by the maximum value of the percentages of the used life cycles in the four joints, the life cycle of the current SCARA robot is fed back, and corresponding life cycle management early warning information is generated.
In the implementation process, the life cycle of the SCARA robot can be expressed by combining the life cycle states of the four joints, namely, the life cycle value of the SCARA robot is expressed as follows:
A1*b1+A2*b2+A3*b3+A4*b4；
wherein: a1, A2, A3 and A4 represent the percentage of the used life cycle of each joint, b1, b2, b3 and b4 represent the weight value corresponding to each joint, the corresponding weight value is set by combining the specific gravity of each joint in the SCARA robot, the influence weight value of the core part is larger, the influence weight value of the unnecessary part is smaller, the life cycle value obtained by the expression mode can be compared with the life cycle table to obtain the life cycle of the SCARA robot, and the life cycle of the SCARA robot comprises: the life cycle state of the SCARA robot can be reflected by data superposition in a life cycle state model diagram of the SCARA robot, so that a life cycle state result diagram of the SCARA robot is formed, the life cycle of the SCARA robot can be reflected by the life cycle state result diagram, the life cycle of the current SCARA robot is fed back, and corresponding life cycle management early warning information is generated.
The method provided by the embodiment of the invention realizes the life cycle prediction of the SCARA robot by extracting the characteristic quantity on each joint and combining with the neural network, has high efficiency in the whole implementation process, and can intuitively embody the running health state of the SCARA robot through a life cycle model. The characteristic information of the acceleration vibration signals on each joint can be directly expressed by adopting a mode of combining wavelet packet decomposition and energy spectrum entropy, and extraction of the characteristic information is facilitated, so that learning and training of the acceleration vibration signals on each joint can be realized by combining a neural network, and further prediction of life cycle of the SCARA robot is realized. The method has higher accuracy in identifying the acceleration vibration signals, and can realize classification prediction of the acceleration vibration signals through the neural network model, so that the accuracy of the life cycle prediction of the SCARA robot is improved, and the reliability and the accuracy of the life cycle of the robot can be effectively evaluated.
The foregoing has outlined rather broadly the more detailed description of embodiments of the invention, wherein the principles and embodiments of the invention are explained in detail using specific examples, the description of the embodiments being merely intended to facilitate an understanding of the method of the invention and its core concepts; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present invention, the present description should not be construed as limiting the present invention in view of the above.
Claims (7)
1. A method of predicting a life cycle of a SCARA robot, the method comprising the steps of:
acquiring acceleration vibration signals in the running process through four acceleration sensors arranged on a robot arm of the SCARA robot, wherein each of four joints of the SCARA robot is provided with an acceleration sensor;
preprocessing acceleration vibration signals acquired by four acceleration sensors to extract vibration signal feature vectors, wherein the vibration signal feature vectors comprise four-path wavelet packet energy spectrum entropy vectors;
taking each path of wavelet packet energy spectrum entropy vector as an input vector of a corresponding neural network, and acquiring a life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector;
processing life cycle prediction results under the energy spectrum entropy vectors of the four-way wavelet packet to obtain life cycle of the SCARA robot;
the step of taking each path of wavelet packet energy spectrum entropy vector as the input vector of the corresponding neural network and obtaining the life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector comprises the following steps:
acquiring sample entropy data on joints corresponding to each path of wavelet packet energy spectrum entropy vector used for model training templates, and aiming at acceleration vibration data generated under a historical life cycle of replacement of each joint in different SCARA robots as a training sample, wherein the wavelet packet energy spectrum entropy vector of the acceleration vibration data generated under the historical life cycle is the sample entropy data on the corresponding joint;
performing an order analysis on the sample entropy data to extract each joint feature;
inputting a part of sample entropy data in each joint characteristic into a constructed neural network for predicting the life cycle of each joint, and executing verification processing through the other part of sample entropy data so as to train a life cycle prediction model of each joint;
inputting each path of wavelet packet energy spectrum entropy vector to a corresponding life cycle prediction module to obtain a life cycle prediction result of each joint; the acquisition of acceleration vibration signals in the running process through four acceleration sensors arranged on the SCARA robot arm comprises the following steps:
acquiring a simulated acceleration vibration signal on the joint based on an acceleration sensor on each joint;
filtering the analog acceleration vibration signal to filter a high-frequency vibration signal;
analog-to-digital conversion processing is carried out on the analog acceleration vibration signals with the high-frequency vibration signals filtered, and digital acceleration vibration signals are generated;
the performing an order analysis on the sample entropy data to extract each joint feature comprises: extracting sample entropy data of each joint under the historical life cycle, expanding to frequency doubling and frequency tripling to serve as characteristic points of preliminary collection, sequencing the characteristic points of the preliminary collection, taking the maximum value of each positive and negative preset sample entropy before and after the characteristic points as the value of a final characteristic point, wherein the frequency corresponding to the value of the final characteristic point is the characteristic point to be extracted and serves as the characteristic of a first batch of samples; selecting sample entropy data with a certain proportion for analysis, selecting characteristic points with sample entropy values showing extreme values, testing and comparing the sample entropy data with the residual proportion, retaining the common obvious characteristic points in all samples, collecting the obvious characteristic points of other SCARA robots, removing different memory of the characteristic points obtained by different SCARA robots, and analyzing and combining the common characteristic points to be used as the characteristics of a second batch of samples; comparing the first batch of sample features with the second batch of sample features, reserving frequency multiplication features and common feature points obtained by order analysis, and combining feature points with close frequency to obtain life cycle features of joints of the SCARA robot needed to be used for creating the model.
2. The method for predicting a life cycle of a SCARA robot of claim 1, wherein preprocessing the acceleration vibration signals acquired by the four acceleration sensors to extract vibration signal feature vectors comprises:
the wavelet packet decomposition principle is adopted to decompose the wavelet packet of the acceleration vibration signal;
reconstructing wavelet packet coefficients, and solving the proportion of energy contained in each frequency band component to the total energy of each acceleration vibration signal;
and obtaining the energy distribution condition of each wavelet packet coefficient of the acceleration vibration signal by calculating the wavelet packet energy spectral entropy of each frequency band, and taking the wavelet packet energy spectral entropy as the vibration signal characteristic vector of the acceleration vibration signal.
3. The method for predicting a life cycle of a SCARA robot of claim 2, wherein reconstructing the wavelet packet coefficients and finding a specific gravity of energy contained in each frequency band component to total energy of the signal comprises:
different frequency band components of the signals are obtained through wavelet packet decomposition, characteristic information of each frequency band is obtained according to the distribution of the frequency bands of the acceleration vibration signals, and different acceleration vibration signals are analyzed through wavelet packet energy spectrum.
4. A method of predicting a life cycle of a SCARA robot according to any one of claims 1 to 3, wherein processing the life cycle prediction result under the four-way wavelet packet spectral entropy vector to obtain the life cycle of the SCARA robot comprises:
forming a current life cycle state curve graph of each joint from the life cycle prediction result of each joint;
and comparing and fusing the life cycle state graphs of the four joints to obtain a life cycle state result graph of the SCARA robot.
5. A system for predicting a life cycle of a SCARA robot, the system comprising:
the data acquisition module is used for acquiring acceleration vibration signals in the running process through four acceleration sensors arranged on the SCARA robot arm, each of the four joints of the SCARA robot is provided with one acceleration sensor, and the acquisition of the acceleration vibration signals in the running process through the four acceleration sensors arranged on the SCARA robot arm comprises the following steps: acquiring a simulated acceleration vibration signal on the joint based on an acceleration sensor on each joint; filtering the analog acceleration vibration signal to filter a high-frequency vibration signal; analog-to-digital conversion processing is carried out on the analog acceleration vibration signals with the high-frequency vibration signals filtered, and digital acceleration vibration signals are generated;
the preprocessing module is used for preprocessing acceleration vibration signals acquired by the four acceleration sensors to extract vibration signal feature vectors, wherein the vibration signal feature vectors comprise four paths of wavelet packet energy spectrum entropy vectors;
the data processing module is used for taking each path of wavelet packet energy spectrum entropy vector as an input vector of the corresponding neural network and acquiring a life cycle prediction result corresponding to each path of wavelet packet energy spectrum entropy vector;
the prediction processing module is used for processing the life cycle prediction result under the energy spectrum entropy vector of the four-way wavelet packet to obtain the life cycle of the SCARA robot;
the data processing module comprises:
the acquisition unit is used for acquiring sample entropy data on joints corresponding to each path of wavelet packet energy spectrum entropy vector used for the model training template, and aiming at acceleration vibration data generated in a historical life cycle of replacement of each joint in different SCARA robots as a training sample, wherein the wavelet packet energy spectrum entropy vector of the acceleration vibration data generated in the historical life cycle is the sample entropy data on the corresponding joint;
an order analysis unit for performing an order analysis on the sample entropy data to extract each joint feature;
the construction unit is used for inputting part of sample entropy data in the characteristics of each joint into the constructed neural network for predicting the life cycle of each joint, and executing verification processing through the other part of sample entropy data so as to train a life cycle prediction model of each joint;
the prediction unit is used for inputting the energy spectrum entropy vector of each path of wavelet packet to the corresponding life cycle prediction module to obtain a life cycle prediction result of each joint;
the performing an order analysis on the sample entropy data to extract each joint feature comprises: extracting sample entropy data of each joint under the historical life cycle, expanding to frequency doubling and frequency tripling to serve as characteristic points of preliminary collection, sequencing the characteristic points of the preliminary collection, taking the maximum value of each positive and negative preset sample entropy before and after the characteristic points as the value of a final characteristic point, wherein the frequency corresponding to the value of the final characteristic point is the characteristic point to be extracted and serves as the characteristic of a first batch of samples; selecting sample entropy data with a certain proportion for analysis, selecting characteristic points with sample entropy values showing extreme values, testing and comparing the sample entropy data with the residual proportion, retaining the common obvious characteristic points in all samples, collecting the obvious characteristic points of other SCARA robots, removing different memory of the characteristic points obtained by different SCARA robots, and analyzing and combining the common characteristic points to be used as the characteristics of a second batch of samples; comparing the first batch of sample features with the second batch of sample features, reserving frequency multiplication features and common feature points obtained by order analysis, and combining feature points with close frequency to obtain life cycle features of joints of the SCARA robot needed to be used for creating the model.
6. The system for predicting a life cycle of a SCARA robot of claim 5, wherein the preprocessing module comprises:
the decomposition unit is used for decomposing the wavelet packet of the acceleration vibration signal by adopting the wavelet packet decomposition principle;
the reconstruction unit is used for reconstructing the wavelet packet coefficient and solving the proportion of the energy contained in each frequency band component to the total energy of each acceleration vibration signal;
the characteristic unit is used for obtaining the energy distribution condition of each wavelet packet coefficient of the acceleration vibration signal by calculating the wavelet packet energy spectrum entropy of each frequency band, and taking the wavelet packet energy spectrum entropy as the vibration signal characteristic vector of the acceleration vibration signal.
7. The system for predicting a life cycle of a SCARA robot of claim 5 or 6, wherein the prediction processing module comprises:
the generating unit is used for forming a life cycle state curve chart of each joint according to the life cycle prediction result of each joint;
and the synthesis unit is used for comparing and fusing the life cycle state graphs of the four joints to obtain a life cycle state result graph of the SCARA robot.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210879935.6A CN115062672B (en) | 2022-07-25 | 2022-07-25 | Method and system for predicting life cycle of SCARA robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210879935.6A CN115062672B (en) | 2022-07-25 | 2022-07-25 | Method and system for predicting life cycle of SCARA robot |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115062672A CN115062672A (en) | 2022-09-16 |
| CN115062672B true CN115062672B (en) | 2023-04-28 |
Family
ID=83206400
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210879935.6A Active CN115062672B (en) | 2022-07-25 | 2022-07-25 | Method and system for predicting life cycle of SCARA robot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115062672B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116079696B (en) * | 2023-04-07 | 2023-06-13 | 广东毕要科技有限公司 | SCARA robot positioning updating method and device based on vibration signals |
| CN117370878B (en) * | 2023-12-07 | 2024-03-05 | 山东第一医科大学第一附属医院(山东省千佛山医院) | Epidermis extraction and positioning method and system based on spine joint vibration information |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106017879A (en) * | 2016-05-18 | 2016-10-12 | 河北工业大学 | Universal circuit breaker mechanical fault diagnosis method based on feature fusion of vibration and sound signals |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102072829B (en) * | 2010-11-04 | 2013-09-04 | 同济大学 | Iron and steel continuous casting equipment oriented method and device for forecasting faults |
| CN107464023B (en) * | 2017-08-16 | 2020-04-17 | 新疆金风科技股份有限公司 | Online learning method and equipment of life cycle prediction model |
| CN110757510B (en) * | 2019-10-31 | 2021-05-11 | 广东工业大学 | Method and system for predicting remaining life of robot |
| CN114611551A (en) * | 2022-03-04 | 2022-06-10 | 浙江浙能兰溪发电有限责任公司 | Electromechanical fault classification method based on wavelet packet energy spectrum entropy |
-
2022
- 2022-07-25 CN CN202210879935.6A patent/CN115062672B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106017879A (en) * | 2016-05-18 | 2016-10-12 | 河北工业大学 | Universal circuit breaker mechanical fault diagnosis method based on feature fusion of vibration and sound signals |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115062672A (en) | 2022-09-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115062672B (en) | Method and system for predicting life cycle of SCARA robot | |
| Jaber et al. | Fault diagnosis of industrial robot gears based on discrete wavelet transform and artificial neural network | |
| Lou et al. | Bearing fault diagnosis based on wavelet transform and fuzzy inference | |
| Abed et al. | A robust bearing fault detection and diagnosis technique for brushless DC motors under non-stationary operating conditions | |
| Jaber et al. | Industrial robot backlash fault diagnosis based on discrete wavelet transform and artificial neural network | |
| Jaber et al. | Development of a Condition Monitoring Algorithm for Industrial Robots based on Artificial Intelligence and Signal Processing Techniques. | |
| Zhang et al. | A novel intelligent fault diagnosis method based on variational mode decomposition and ensemble deep belief network | |
| Hwang et al. | SVM-RBM based predictive maintenance scheme for IoT-enabled smart factory | |
| Zhou et al. | A multistage deep transfer learning method for machinery fault diagnostics across diverse working conditions and devices | |
| Medina et al. | Deep learning-based gear pitting severity assessment using acoustic emission, vibration and currents signals | |
| Jaber et al. | The state of the art in research into the condition monitoring of industrial machinery | |
| Zhou et al. | Harmonic reducer in-situ fault diagnosis for industrial robots based on deep learning | |
| Pham et al. | Efficient fault diagnosis of rolling bearings using neural network architecture search and sharing weights | |
| CN116956215A (en) | Fault diagnosis method and system for transmission system | |
| CN113733164B (en) | Industrial robot fault diagnosis method based on current detection | |
| Nacer et al. | A novel method for bearing fault diagnosis based on BiLSTM neural networks | |
| Almounajjed et al. | Stator fault diagnosis of induction motor based on discrete wavelet analysis and neural network technique | |
| Miao et al. | Fault diagnosis of wheeled robot based on prior knowledge and spatial-temporal difference graph convolutional network | |
| Massol et al. | An exTS based neuro-fuzzy algorithm for prognostics and tool condition monitoring | |
| Kuo et al. | Study on bandwidth analyzed adaptive boosting machine tool chatter diagnosis system | |
| Nguyen et al. | An integrated generalized discriminant analysis method and chemical reaction support vector machine model (GDA-CRSVM) for bearing fault diagnosis | |
| Sun et al. | Fault diagnosis of planetary gearbox based on signal denoising and convolutional neural network | |
| CN113639985B (en) | Mechanical fault diagnosis and state monitoring method based on optimized fault characteristic frequency spectrum | |
| Li et al. | Deep learning based end-to-end rolling bearing fault diagnosis | |
| Puchalski et al. | Real-time UAV fault detection and classification using measurement data from the PADRE database |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |